ARISTOTLE UNIVERSITY OF THESSALONIKI VOICE MORPHOLOGY AND TRANSITIVITY ALTERNATIONS IN GREEK: EVIDENCE FROM CORPORA AND PSYCHOLINGUISTIC EXPERIMENTS by Georgia Fotiadou A thesis submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Psycholinguistics School of English Department of Linguistics Aristotle University Thessaloniki 2010
424
Embed
VOICE MORPHOLOGY AND TRANSITIVITY ALTERNATIONS IN GREEK: EVIDENCE FROM CORPORA AND PSYCHOLINGUISTIC EXPERIMENTS
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
ARISTOTLE UNIVERSITY OF THESSALONIKI
VOICE MORPHOLOGY AND
TRANSITIVITY ALTERNATIONS IN GREEK:
EVIDENCE FROM CORPORA AND PSYCHOLINGUISTIC EXPERIMENTS
by Georgia Fotiadou
A thesis submitted in partial fulfillment of the requirements for the degree
of Doctor of Philosophy in Psycholinguistics
School of English Department of Linguistics
Aristotle University
Thessaloniki 2010
To raise a newborn child is a Miracle that takes all of one’s strength and courage.
To write a PhD Dissertation is a demanding achievement that needs dedication to form a Miracle.
The simultaneous conception of them turned me into a better person
Their growing up helped me believe that effort matters….
I dedicate this dissertation to my daughter Katerina who was born an Angel to light my way through life….
iii
ACKNOWLEDGMENTS
For the completion of this Thesis I need to thank many people who offered me their
support and help throughout this long and heavy task.
First and foremost, I am indebted to my primary supervisor, Ianthi-Maria Tsimpli. I
am grateful to her for her commitment, her generosity in sharing her knowledge, her
unfailing support and her faith in me. I almost nostalgically recall our first meetings
when she supervised my MA dissertation and the inspired Reading Groups she held at
that time with PhD students from the department of Theoretical and Applied
Linguistics, at the School of English of the Aristotle University. Countless meetings
followed of course; every occasion sitting in one of Ianthi’s talks and lectures were
fruitful, educating and inspiring. For all that and many other unsaid reasons I consider
Ianthi-Maria Tsimpli my tutor, my mentor in linguistics and life.
Anna Roussou, showed trust in me by offering me the chance to collaborate with her,
in the project she directed (Karatheodoris grant, FK 2780, Syntactic Features in Greek
L2: the Case of Article and Clitics, financed by the University of Patras). I especially
thank her for this opportunity which allowed me to gain from a very important
linguist the knowledge of responsibility and the strength of positive encouragement.
Although it was most definitely Ianthi’s supervision of my MA dissertation and her
way of working that got me interested in syntax, I would not have decided to begin a
PhD had it not been for Anna Anastassiadi-Simeonidi’s encouragement during the
years of my MA studies. I thank her for agreeing to be a member in my supervision
committee and for constantly supporting my work.
I am grateful to Despoina Papadopoulou for introducing me to the exciting world of
sentence processing and for never, since that time, failing to reply to any of my
questions on experimental methodology and statistics, among other things.
Maria Dimitrakopoulou, has always been the person I could count on, from my very
first day in the Department of English Language. I thank her, for explaining all
difficult parts of linguistic theories, for reading manuscripts and commenting on my
English during earlier stages of this project.
iv
Also, I thank all the members of the administration staff who always kept a smile for
me, and especially Lia Efstathiadi who supported the project with her caring work and
stood by me every time I needed help. The corpus project would not have
materialized without the precious help of Tasos Paschalis in technologically
supporting our needs. His guidance and committed personal work in creating
databases, web-pages and other useful tools for the study has been of invaluable help
in the course of developing the project. Kalliopi Katsika has also been a tireless
companion throughout the whole journey of doctorate research. I also need to thank
other members of staff at the Department of Theoretical and Applied Linguistics who
made me feel at home during all this time. In particular, I am grateful to Professor
Niovi Antonopoulou who has for so many years kindly agreed to share her office with
us, the Language Development Lab, people.
I am greatly indebted to all the participants who voluntarily offered their intuitions in
the studies I conducted. Almost two hundred adults and children participated in the
experimental tasks designed for the present study. I particularly thank the Lilliputian
participants for the enthusiasm they showed for the study, sacrificing some of their
playtime in order to cooperate and doing so with great professionalism. I would also
like to thank all the parents of the child participants for showing interest in my study
and especially nursery schools for being so cooperative and generous with their time
and helping me with recruiting the participants. I am grateful to the Directors of the
nursery schools: Militidis Margaritis (Άνοιξη ‘Spring’), in the area of Panorama in
Thessaloniki), Zighanitidou Eleni (Το Μικρό Σπίτι στο Λιβάδι ‘Little House ιn the
Prairie’, in Veria), Gkeliri Soultana (Στρουµφάκια ‘Smurfs’, in Veria) and Tzima
Rodanthi (3ος ∆ηµοτικός Σταθµός Βέροιας ‘3rd State Nursery School of Veria’), in
Veria).
Parts of this thesis have been presented in the following conferences:
-- “Η Μορφολογία της Φωνής και η από-φράση στην Νέα Ελληνική” [“Voice
Morphology and the apo-phrase in Modern Greek”], 3η Συνάντηση
Eugenia, … for assisting me in any way you could, but mostly for offering me
yourselves, invaluable friends.
I could not name all those who offered me help in any kind, but I specifically thank
them for adding something to my theorizing of life, academic and personal, and for
encouraging me to follow my heart, remaining skeptical, objective and decisive at the
same time!
This thesis is part of the 03ED375 research project, implemented
within the framework of the “Reinforcement Programme of Human
Research Manpower” (PENED) and co-financed by National and
Community Funds (25% from the Greek Ministry of Development-
General Secretariat of Research and Technology and 75% from
E.U.-European Social Fund).
Thessaloniki, January 2010
vii
TABLE OF CONTENTS
Acnowledgments ……………………………………………………………………………………….… iii Table of Contents………………………………………………………………..……………….….…… vii List of Tables……………………………………………………………………………..………………… xii List of Graphs……………………………………………………………………………………..………... xv List of Pictures…...………………………………………………………………..……….…………..… xiii List of Appendices …………………………………………………………...….……………….…..… xiii Abbreviations……………………………………………………………………………………...……… xix
Abstract ………………………………………………………………...………………………..………...… xx
2.2. Specific rationale………………………………………………............................……….. 211 3. The SPR and AJ tasks……………………………………………………………………….….… 216
3.1. Design and materials ……………………………………………………………………… 217 3.2. Procedure……………………………..………………………………………………….….…....220 3.3. Participants ………………………………………………...............................……..……. .222 3.4. Research questions ………………………………….………………………………..……. 222
4. Results………………………………..……………………………………………...…………………. 222 4.1. Analyses and measurements……………………………………...…………………… 224 4.2. Results of the Acceptability Judgment Task (rates and RTs) …………... 224
x
4.2.1. Total sentences not evaluated ……………………………………………224
4.2.2. Total Mean Rates with respect to voice morphology and subject
2. The Rationale …………………………………………………………….......................…....... 267
2.1 Notes on previous research on language acquisition ……………......…... 267 2.2 Specific rationale……………………………………….………………............................. 271
3. The SPM task…………………………………………………………..............…….................. 273
3.1. Design and Materials…………………………………………………….………………... 273
1996, a.o.) which are based on the idea that frequencies of the alternative readings of
ambiguous structures affect the way adults parse sentences. The second aim is to
evaluate ‘usage-based’ approaches to language acquisition (Tomasello, 2003 a.o.)
since the role of frequency of particular phenomena in the input is assumed to
determine the pattern of acquisition. With respect to this second aim, usage-based
approaches will be juxtaposed with nativist approaches which prioritize the role of
grammar acquisition over frequency information in the input. These issues are
explored by looking at whether and how frequency in the occurrence of particular
verbs in active and non-active voice morphology and the corresponding transitivity
changes is a more deterministic factor for L1 acquisition than the interaction of Voice
morphology and animacy in on-line adult and child L1 data. Frequencies of structures
and of verb types will be measured in corpora of formal and informal registers of
adult Greek.
1. Motivation and Aim of the Study
Transitivity alternations have been a main focus of research in the field of theoretical
linguistics. More specifically, one-place predicates have been claimed to actually
consist of two classes, the unergative (1) and the unaccusative (2) class although they
share the same surface syntactic realization, namely a single argument in subject
position. The Unaccusativity Hypothesis, formulated by Perlmutter (1978) and later
INTRODUCTION CHAPTER 1
3
adapted by Burzio (1986) associates the two verb classes with distinct syntactic
configurations, since (1) takes a deep-structure subject and no object, whereas (2)
takes a deep-structure object and no subject, as illustrated below:
(1) NP [VP V]
(2) [VP V NP]
Evidence of morphological and syntactic syncretism is cross-linguistically
found not only with respect to the above mentioned structures but also with respect to
passive, reflexive and unaccusative structures (Kalluli, 2004, 2006 for Albanian;
Borer, 1994 and subsequent work for Hebrew; Kayne, 1988, for Romance, to mention
a few): a pronoun, a clitic or verbal inflection is shared by reflexive, passive and anti-
causative predicates.
Greek, which is the language examined in the present thesis, is a language
with a morphologically rich inflectional system. Voice morphology on the Greek verb
has a binary value, active (ACT) and non-active (NACT), but does not always show a
one-to-one correlation with diathesis1. Diathesis is a semantic notion which concerns
the relationship between the verb and the subject; in previous literature the terms
active and (medio)-passive were used instead (Holton et al., 1997; Joseph &
Philippaki-Warburton, 1987; Tsimpli, 1989, 2005): verbs in the active diathesis (eg.
treho ‘run’) include examples of unergatives as in o Petros trehi sto jipedo (‘Peter
runs in the stadium’) and transitives as in o Petros trehi to programma ston ipolojisti
(‘Peter runs the program in the computer’). Verbs in the (medio)-passive diathesis are
morphologically marked and describe structures where the subject is an experiencer
or an agent and the action affects directly (eg. dinome ‘get dressed’ (dress + NACT))
or indirectly the subject (stenohorieme ‘get upset’ (upset + NACT)) (middle
diathesis); in this category structures where the subject does not act but is affected by
an action or cause are also included (eg. metaferome ‘be transferred’ (transfer +
NACT)) (passive diathesis).
1 For the description of the morphological inflection attached to verb entries referred to as voice and the semantic notion of diathesis in Greek see Joseph & Philippaki-Warburton, 1987; Τζάρτζανος, 1989 [Tzartzanos, 1989]; Holton, Mackridge & Philippaki-Warburton, 1997; Κλαίρης & Μπαµπινιώτης, 1999, 2005 [Klairis & Babiniotis, 1999, 2005]; Σετάτος, 1997 [Setatos 1997] a.o.
INTRODUCTION CHAPTER 1
4
The terms active (ACT) and non-active (NACT) are used instead (Rivero,
2005, 2006 a.o., cf. Haspelmath, 1993)2. In addition, based on previous empirical
evidence (Tsimpli, 2006), it has been shown that the animacy of the syntactic subject,
as well as lexical/encyclopaedic properties of the verb interact with voice morphology
in that strong preferences for one of the possible readings, e.g. the reflexive, the
passive or the anti-causative can be explained in relation to these factors.
Due to the morphological syncretism in Greek, ambiguity can be found in the
domain of thematic structures as in (3a) and (3b) below:
(3) a. To aghori plithike
The boy wash-NACT. 3Sg
“The boy washed himself/ The boy was washed.”
b. To aftokinito plithike (apo ton ipalilo tu garaz) / (apo ti vrohi).
The car wash-NACT.3Sg (by the employee of the car station) / (by the rain)
“The car got/was washed by the employee of the car station / with the rain.”
In (3) both (a) and (b) are considered at least temporarily ambiguous because the Non-
active (NACT) Voice morphology marked on the Verb signals a transitivity
alternation with more than one available readings. Thus, in (3a) the passive and the
2 The Greek middle structure is also morphologically non-active (Tsimpli, 1989; Papastathi, 1999; Sioupi, 1998; cf. Condoravdi, 1989). However, contextual factors as the generic interpretation, the arbitrary agent by-phrase and the dynamic modal reading associated with the Greek middle render it difficult to test in ways similar with passives, reflexives and anti-causatives. Thus, they are not included in the present study.
INTRODUCTION CHAPTER 1
5
reflexive readings are both possible, while extra-clausal information (such as the
boy’s age) and a continuation of the sentence (such as me to nero tis vrohis ‘with the
rain’ or apo tin megali tu adhelfi ‘by his older sister’) disambiguates the reading.
Moreover, in (3b), where the subject is inanimate, the interpretation could be either
passive or anti-causative, as shown by the two alternative continuations of the
sentence. Thus, the ambiguity in (3) results from the fact that Greek NACT
morphology is amenable to different interpretations even when it appears with the
same verb.
Active (ACT) Voice morphology marked on the Verb does not signal
transitivity alternations in a transparent way either, since it is associated with
unaccusative, unergative and transitive structures (Anagnostopoulou & Alexiadou,
(6) a. To trapezomandilo lerose/lerothike (apo mono tou / apo ti saltsa / apo to
moro).
the tablecloth dirty-ACT/NACT.3Sg (by itself / by/from the sauce / by the
baby)
“The tablecloth got dirty (by itself / from the sauce) / was dirtied by the
baby.”
b. I Maria *lerose/ lerothike ((apo) moni tis / apo ti saltsa / apo to moro).
The Maria dirty-ACT/ NACT.3Sg (by herself / by/from the sauce / by the
baby)
“Maria dirtied herself / got dirty from the sauce / was dirtied by the baby.”
While in (6a) both ACT and NACT verb forms are considered grammatical with an
anti-causative reading, in (6b) only the NACT verb form is. Nevertheless, according
to the discussion of examples (3) as well as (4) and (5), (6a) is temporarily ambiguous
between anti-causative & passive readings while (6b) with the NACT verb form is
ambiguous between reflexive, anti-causative and passive readings.
It is worth noting at this point that the unaccusative verbs which optionally
appear either with active or non-active morphology without this change in voice
morphology affecting the availability of the unaccusative reading are referred to in
Greek as ‘ρήµατα διτυπίας’ (lit. of two forms, Theophanopoulou-Kontou, 2000)3. We
3 The term ‘wrinkle/crumple verbs’ (Anagnostopoulou & Alexiadou, 2004) has been used instead to refer to the verbs which can appear either with active or non-active morphology with an anti-causative reading; furthermore, the term ‘ergative verbs’ has been used to denote the verbs which undergo the causative/anti-causative alternation but only appear in the active voice morphology in both structures
INTRODUCTION CHAPTER 1
7
use the term ‘Voice Alternating Anti-causatives’ to refer to verbs as in (6), while the
term ‘Voice Non-alternating Anti-causatives’ to refer to verbs which participate in the
Swinney, 2008), or passive and transitive ones (Rohde & Gibson, 2003; Ferreira,
2003; Marinis, 2007). However, there are discrepancies in the results and
disagreement on how to interpret them. Peristeri et al. (submitted) investigated
animacy of the syntactic subject in ‘non-canonical’ structures of Greek with Greek
(Theophanopoulou-Kontou, 2000). We opt for the term ‘voice (non)-alternating anti-causative’ verbs (Tsimpli, p.c.) as more accurate and broad in use, without close reference to semantically restricted verb classes.
INTRODUCTION CHAPTER 1
8
agrammatics and unimpaired controls. These studies will be presented in Chapter 4
where the on-line experimental study is presented.
Previous studies on other phenomena involving ambiguity resolution in Greek
sentence processing investigate Relative Clause (RC) attachment ambiguities in L1
and L2 (Papadopoulou, 2006; Papadopoulou & Clahsen, 2003, 2006; Papadopoulou
& Marinis, 2004), subject-object ambiguities with optionally transitive verbs
(Papadopoulou & Tsimpli, 2005) and prepositional phrase (PP) attachment
ambiguities (Katsika, 2009 a, b).
With respect to research on first language acquisition, the question of how
children acquire transitivity alternations has been a central one. Although, children
acquire language through developmental stages commonly identified, mastery of the
involved structures has been attested at different developmental stages cross-
not unambiguously, we investigate whether the parser relies on structures such as
DP(subject)-VACT and DP(subject)-VNACT verb forms to resolve the interpretational
ambiguity arising on the verb.
In line with coarse-grained models of parsing, the more frequent interpretation
of a given structure should be processed faster than the remaining, less frequent ones
of this same structure. In addition to Voice morphology, animacy distinctions on the
surface syntactic subject have been argued to play a role in the choice of interpretation
of transitivity alternations in Greek (Tsimpli, 2006). The interaction of [+/- animacy]
and Voice morphology with the preferred interpretations (Peristeri et al., submitted) is
thus also addressed in this study: if, as assumed, subject animacy is a factor
contributing to the availability of alternative readings, processing of voice
‘alternating’ and ‘non-alternating’ anti-causatives in Greek should mirror frequent
4 Note however that the authors admit that there must exist ‘at least some non-statistical influences’ during initial parsing (Mitchell & Brysbaert, 1998:324).
INTRODUCTION CHAPTER 1
12
uses of structures characterized by animacy distinctions on the subject as well as
morphological Voice distinctions.
Other models propose the existence of a more fine-grained storing system of
lexical, semantic, syntactic, pragmatic or other discourse frequency information
(MacDonald, 1997; MacDonald et al., 1994; Spivey-Knowlton, Trueswell &
Tanenhaus, 1993 and subsequent work). These ‘Constraint Satisfaction’ models
assume that parsing is a continuous (one-stage) process during which multiple sources
of information (lexical, syntactic and/or discourse related) are used in parallel and
affect sentence interpretation on the basis of their frequency. However, some models
distinguish between syntactic and semantic information, claiming that ‘only
information that is correlated with syntactic alternatives will have effects’ (Trueswell
et al., 1994:176), while others propose that lexical constraints are the dominant ones
(MacDonald 1993, 1997) suggesting that lexical representations contain not only
phonological and orthographic information, but also all relevant morpho-syntactic
information as well as all alternative structures in which specific lexical items
participate categorized with respect to their frequency of occurrence in the language
(MacDonald et al., 1994).
Subsequently, although no explcit reference to a morphological analysis is
made in these models, but rather a storing of item records, in our case, this would
imply that the lexical entry of a verb such as htipai (‘hit’) in (N)ACT which may
receive a transitive or an anti-causative reading in ACT and a reflexive, anti-causative
or passive reading in NACT, would activate all possible readings when processed on-
line. In line with these fine-grained lexicalist accounts, the parser would resolve the
ambiguity by means of lexical frequency counts of each interpretation, thus rendering
more accessible the most frequent reading per voice form.
Research in the field has given rise to contradictory results and seems to pose
rather than answer questions as to whether the parser is based on input frequencies
measurable in natural language corpora or not, and as to what kind of information
directs the parser’s initial choices.
The present thesis aims to evaluate coarse-grained and fine-grained accounts
of experience-based models on sentence processing with respect to transitivity
INTRODUCTION CHAPTER 1
13
alternations. The evaluation is based on a comparison between data from an on-line
self-paced reading task with adult NS of Greek and frequency results from two adult
language corpora of formal and informal written speech respectively with respect to
verbs in the anti-causative set (Voice (Non) - alternating Anti-causatives5).
3. Frequency-based vs UG- based Language Acquisition
Chomsky (1965, 1969, 1986 and subsequent work) has argued that language
acquisition cannot be based on imitation or habit-formation but there must be some
innate knowledge of linguistic structure which predisposes children to learn any
natural language they are exposed to. This innate endowment is referred to as
Universal Grammar, a cognitive system that contains Principles, applying to all
languages and Parameters, which cater for crosslinguistic variation. Accordingly,
language acquisition is the task of acquiring (setting) parameters (i.e. the particular
values of functional features/categories) on the basis of the input children are exposed
to. Principles of UG determine the basic architecture of any linguistic system, its
constraints and operations that apply to all languages.
Within the nativist approach, different analyses of transitivity alternations and
argument structure have been proposed. While all of these analyses argue that
knowledge of syntactic categories is innate, some of them argue for the priority of
lexical semantics over syntax: The Semantic Bootstrapping hypothesis (Grimshaw,
1981; Pinker, 1984; Randall et al., 2004), suggests that children learn the meaning of
(some) content words and use them to construct semantic representations of simple
sentences they hear. The Syntactic Bootstrapping Hypothesis (Gleitman, 1990; Borer,
2004) assumes that children analyse the structure of sentences they hear and interpret
the meaning of new material based on syntactic abstract knowledge they already have.
Deviation from adult-like performance can be accounted for in terms of maturation
(eg. Wexler & Borer, 1987) up to the end of the critical period for language
5 We use the term ‘anti-causatives’ for all the verbs used in the present study since they have been labelled as such in the Greek generative grammar (Alexiadou & Anagnostopoulou, 2004; Theophanopoulou-Kontou 2001 a.o.). Nevertheless, other readings may be assigned to the verbal stimuli used in the task, i.e. the passive, the middle or the reflexive readings.
INTRODUCTION CHAPTER 1
14
acquisition: language knowledge matures just like other cognitive systems.
Computational immaturity within the language system may lead the child to perform
in a non-adult fashion both in comprehension and production (Wexler, 1998).
From the non-generative perspective of usage-based acquisition, abstract
syntactic knowledge is viewed as a derivative of language not a prerequisite. In
particular, the ‘Usage-Based Theory’ (Tomasello, 2003) attributes language
development to the learning of items and constructions on the basis of input
frequency, while abstract and complex syntactic constructions are acquired later.
Within this framework, it has been proposed that children reproduce “verb-
island” (utterance schemas revolving around verbs) and other “item-based”
which however are not unambiguously specified as passive, anti-causative or others
even when NACT appears on the same verb. As shown by the examples under (7) an
ACT form is morphologically distinct from its NACT counterpart (7a)7; in that use of
either form indicates to the native speaker that the verb is being used transitively or
intransitively. However, what is left unspecified until the end of the sentence (if not
even later, at the level of contextual integration) is whether the preferred
interpretation is a passive, reflexive or anti-causative. As (7b) shows, the reflexive,
the passive and the anti-causative readings are all available for the same NACT verb
6 With the exception of deponent verbs which will not concern us in this study (for analyses on deponents see for example Vassilaki, 1988; Manney, 1993, 2000; Zombolou 1996, 1997, 2004). 7 For an overview of the different approaches for the verb morphology in Greek see Ralli (2003, 2005), Janda & Joseph (2002). See also Galani (2005)
THE PHENOMENON: CHAPTER 2 TRANSITIVITY ALTERNATIONS
32
form. In (7c), where the sentence subject is inanimate, the interpretation could be
either passive, i.e. with an agent by-phrase or anti-causative8:
(7) a. plen-o (wash-ACT.1Sg) / plen-ome (wash-NACT.1Sg)
b. To aghori plithike (mono tu) apo tis vromies / (apo ti mitera tu) / (apo ti
vrohi)
the boy washed-NACT.3Sg (alone his) from dirt / (by the mother his) / (by
the rain)
“The boy washed himself out of his dirt / The boy was washed (by someone)
/The boy got washed (with/from the rain).”
c. To aftokinito plithike (apo tin eteria katharismu) / (apo ti vrohi)
the car washed-NACT.3Sg (by the company cleaning) / (by the rain)
“The car was washed by the cleaning company / with/from the rain.”
In passive structures there is an optional prepositional phrase (apo-phrase ‘by-
phrase’), denoting the agent or the causer, as in apo ti mitera tu (by his mother).
However, due to the underspecification of apo in terms of semantic features, this PP
is found in different syntactic structures (Alexiadou et al., 2006; Zombolou, 2004) and
can have an (internal) argument status as in to aghori/to aftokinito plithike apo tis
vromies (the boy/ the car (was) washed out of his/its dirt) or an adjunct status as in the
passive or anti-causative structures in (7c), among other possibilities (Levin &
b; Lavidas, 2007a, b). Accordingly, it can express agentivity or cause, as well as
instrument, source, directionality, location (when part of complex locative
prepositions), comparison, time or other meanings. Moreover, the use of an overt
agent ‘by-phrase’ is considered marked in Greek (Laskaratou & Philippaki-
Warburton, 1984; Joseph & Philippaki-Warburton, 1987, a.o.). This leads to a very
rare use of the apo-phrase even in cases in which the passive reading is independently
available. Evidence in favour of this claim is also provided by empirical data
(Fotiadou, 2007) in a study investigating the interaction between voice morphology,
8 For theoretical approaches to mediopassive/non-active morphology see Warburton (1970, 1975), Theophanopoulou-Kontou (1982, 1983), Laskaratou (1984), Campos (1987), Vassilaki (1988), Tsimpli (1989), Rivero (1990), Smirniotopoulos (1992), Joseph & Smirniotopoulos (1993), Philippaki-Warburton (1998), Ralli (1999), Papastathi (1999), Papastathi & Tsimpli (2004), Tsimpli (2006), Roussou (2008).
THE PHENOMENON: CHAPTER 2 TRANSITIVITY ALTERNATIONS
33
subject animacy and the meaning of the apo-phrase in Greek; evidence provided from
the present study is also supportive of the earlier findings.
When the apo-phrase has an adjunct status there may be co-indexation with
the θ-role, absorbed by the NACT morpheme (Baker et al., 1989) or it may express
cause as part of the lexical conceptual rather than the argument structure of the verb
(Jackendoff, 1990). Thus, the presence of an apo-phrase cannot be used as a criterion
for passivization as is usually suggested for the corresponding by-phrase in English
(Tsimpli, 2006; but Alexiadou & Anagnostopoulou, 2004; Zombolou, 2004, for a
different opinion).
With respect to the passive reading in (7 b & c) some approaches consider that
the NACT marking appears post-syntactically in environments where there is no
external argument and hence does not affect the derivation (Embick, 1998). In this
way the possibility of more than one readings (i.e. passive, anti-causative, reflexive)
being morphologically expressed in the same way is accounted for. However, other
approaches, mentioned below, consider that the NACT morpheme actually affects the
alternation.
For Theophanopoulou-Kontou (2004) passives are vP constructions: the
NACT morpheme checks the missing external θ-role, implied by light-v. Lavidas
(2007), following Collins (2005), considers that the external argument is present in
passive, while Voice absorbs (or blocks) the ACC Case. Alexiadou &
Anagnostopoulou (2004) assume (following Kratzer, 1994, 2000) that in NACT, a
Voice head is located above vP. Voice introduces the external argument of the verb,
bears [+/-agentive] features and absorbs ACC case.9
9 Roussou (2008) provides a morphosyntactic analysis of the morphological marking involved in the above mentioned structures and demonstrates how the imperfective formation is similar to the participial inflection in Romance and the perfective formation is similar to the Romance se/si (see also Papangeli, 2004 for a comparison between Greek and Romance) and the Albanian u (see also Kalluli, 2006), given that all the remaining pieces of the inflection in Greek, in this latter case, involve actually active morphology. Moreover, she suggests that the imperfective inflection –me (specialized inflection) associates with the internal argument, which promotes to subject position to satisfy the EPP condition (while it is the auxiliary that directs this movement in Romance); the perfective –θ-ik-a (θ-formation) is associated with the se/si object clitic. Thus, she concludes that the addition of a Voice feature or head is not really necessary to explain intransitivity. In line with Tsimpli (2006), she proposes that the passive, reflexive and anti-causative readings are all available and a preferred interpretation is attributed with respect to lexical properties of the verb and other pragmatic factors.
THE PHENOMENON: CHAPTER 2 TRANSITIVITY ALTERNATIONS
34
As already implied, the same example (7b) can also receive a reflexive
reading. Some theoretical accounts of reflexives assume that they are lexically derived
and are further distinguished with respect to the status of the subject. According to
some researchers the subject of reflexives is considered an underlying object just like
the subject of unaccusative/anti-causative verbs and is possible only with certain
verbs because of their lexical semantics (Embick, 2004; Anagnostopoulou & Everaert,
1999; Alexiadou & Anagnostopoulou, 2004). According to alternative analyses
(Tsimpli, 1989; Papangeli, 2004) reflexives are unergative entries, i.e. their subject is
an external (non-derived) argument, which can control a purpose clause, formed in the
Lexicon by the attachment of the morphological suffix to the verb saturating the
internal θ-role from their transitive counterpart. Moreover, all the above consider that
in Greek there is a subcategory of reflexives which are syntactically derived (cf.
Theophanopoulou-Kontou, 1980). These differ from the ‘inherent’ lexical reflexives
in that they can appear with the prefix afto (’self’), as in aftokatastrefome (self-
destroy) but not *aftoplenome (self-wash), they can participate in active reflexive
constructions as in katastrefo ton eafto mu (I destroy myself) but not ??pleno ton
eafto mu (I wash myself) and they can also receive a passive interpretation as in
katastrefome (I destroy myself/I am destroyed).
According to Tsimpli (2006), there is a derivational difference between the
reflexive on one hand and the passive or the anti-causative, on the other. Specifically,
the animate subject is the external argument in the reflexive derivation while it is a
derived subject in the passive (and the anti-causative). Given that only animate
subjects can give rise to the structural ambiguity between the reflexive and the non-
reflexive derivation (i.e. passive or anti-causative), the interaction between the
animacy of the syntactic subject with the semantics of the predicate can lead to the
disambiguation of NACT verbs. The claim is that the grammar distinguishes between
reflexive and non-reflexive derivations of NACT morphology verbs, but does not
distinguish between the passive and the anti-causative. These readings are
distinguished only at an interface level where the cause or agent interpretation of the
external argument becomes relevant.
THE PHENOMENON: CHAPTER 2 TRANSITIVITY ALTERNATIONS
35
A syntactic analysis of the derivational difference between the reflexive and
the non-reflexive structure is provided. Based on the assumption that theta-roles are
features that can be attracted by DPs, clitics or affixes (cf. Manzini & Roussou, 2000),
Tsimpli (2006) assumes (following Alexiadou & Anagnostopoulou, 2004) that Voice
projects as a feature of light v in the non-active. Accordingly, the NACT morpheme
has the property of attracting a theta-feature which can be either internal or external,
thus giving rise to a reflexive or non-reflexive derivation respectively. In the
reflexive, the DP in the subject position is a true subject whereas in the non-reflexive
it is a derived subject (bearing the internal theta-feature). (8) and (9) below illustrate
the dress got-dirty-ACT.3Sg (*from the children) / (from the paint)
b. To forema lero-thi-ke (apo ta pedhia) / (apo tis bojies)
the dress got-dirty-NACT.3Sg (from the paint)
“The dress got dirty.”
c. Ta pedhia lerothikan / *lerosan apo tis bojies (katalathos / jia na
nevriasoun ti dhaskala tous).
the children got-dirty-NACT.3Pl /*got-dirty-ACT.3Pl by the paint (by
mistake / for SUBJ anger the teacher their)
“The children got dirty by the paint (by mistake) / The children dirtied
themselves (to make their teacher angry).”
Following Kratzer (1994) and Chomsky (1995), Alexiadou &
Anagnostopoulou (2004) suggest that a functional category Voice (or Event Phrase or
little v) located above the VP determines the exact nature of the predicate, for
example whether it is transitive or intransitive: this category comes in two types
depending on whether it introduces an external argument or not. Moreover, they
assume a predicate decomposition (Dowty, 1979; Hale & Keyser, 1993) according to
which predicates are distinguished on the basis of properties such as the existence of
operators like BECOME (BECOME/RESULT) and CAUSE which result in predicate
types such as states, achievements, activities and accomplishments. The addition of a
THE PHENOMENON: CHAPTER 2 TRANSITIVITY ALTERNATIONS
37
separate Voice head, embedded under this BECOME/RESULT operator, associated
with voice morphology, includes only manner features (see also Zombolou, 2004),
hence sentences such as *to pani skistike epitides (the cloth tore.NACT deliberately)
or to pani skistike prosektika *apo ti Maria (the cloth tore.NACT carefully *from the
Mary) (examples adapted from Alexiadou and Anagnostopoulou, 2004:131-2) are ill-
formed. What is allowed in anti-causatives (and unergatives) is the anaphoric element
apo mono tu (by himself) followed by a possessive, by analogy to the Italian da sé
(Chierchia, 2004).
In contrast to the examples (10 a & b) where the subject is inanimate, ‘voice
alternating anti-causative’ verbs cannot appear in the active form when the subject is
animate as in (10c), indicating that animacy clearly blocks the ‘alternating’ status of
these anti-causatives (Alexiadou & Anagnostopoulou, 2004). Moreover, with respect
to the apparent voice optionality of both ACT and NACT forms with inanimate
subjects Alexiadou & Anagnostopoulou (2004) observe that there is an
interpretational difference in that the ACT denotes a partial change, while the NACT
a complete one. Thus, ACT verbs have a ‘part-whole or possessive substructure’ in
their representation, while the NACT forms have a BECOME/RESULT operator
embedded under a Voice head, associated with NACT morphology, as illustrated
below:
(11) RESULT + VoiceP (crumple-non-active)
[as in to poukamiso tsalakothike (the shirt was crumpled)]
(12) BECOME + possessive construction (crumple-active)
[to poukamiso tsalakose (se ena simio) (the shirt crumpled (in one
spot))]’10
(adapted from Alexiadou & Anagnostopoulou, 2004: 135)
10 The examples in brackets added for demonstration, are adapted from Alexiadou & Anagnostopoulou (2004).
THE PHENOMENON: CHAPTER 2 TRANSITIVITY ALTERNATIONS
38
Theophanopoulou-Kontou (2000, 2004) and Lavidas (2007) assume that all
anti-causatives, irrespective of their (N)ACT morphological marking, are related to
the semantic properties of the predicate, represented as shown below:
(13) [CAUSE [Y BECOME STATE]]
(adapted from Lavidas, 2007: 71)
More specifically, anti-causatives are VP-constructions derived in the lexicon,
irrespective of their (optional) ACT/NACT morphology.
With respect to the two different (ACT/NACT) forms, Theophanopoulou-
Kontou (2000) observes some interpretational differences: NACT morphology raises
a passive interpretation, with the syntactic subject being ‘affected’ by the action
denoted by the verb, while ACT morphology presents the event as a natural or
spontaneous process that takes place without the intervention of an ‘agent’ and
implies that the speaker is unaware or wants to avoid reference to a cause or agent.
Lavidas (2007a, b) considers that the NACT morpheme reflects the initial
morphological marking of the intransitive verb. All anti-causatives are basic and
cauzativize11: evidence for this claim is provided from a historical perspective
(Lavidas 2007a, b), since innovative transitive/causative uses appear in some periods
11 Alexiadou, Anagnostopoulou & Shäfer (2006) argue against derivational approaches suggesting that the change of state verbs are syntactically decomposed into a Voice and a CAUS component, giving rise to the causative/anti-causative alternation built on the combination of different types of roots with Voice and CAUS heads. Other approaches (eg. Haspelmath, 1993; Klairis & Babiniotis, 1999), using morphological cues suggest that the morphological marking illustrates the source of derivation. Verbs are distinguished, according to the complexity of their morphology; the less complex is the base core. Three categories are identified among verbs participating in the inchoative/causative alternation (Haspelmath, 1993): (i) causative where the inchoative verb is basic and the causative derived, (ii) anti-causative where the causative verb is basic and the inchoative derived (as also suggested by
Nedjalkov & Sil’nickij, 1969) and (iii) non-directed, subdivided into (iii.a) labile’ where the same verb is used in both senses, (iii.b)
‘equipollent’ where both forms are derived from the same stem and (iii.c) ‘suppletive alternations’ where different roots are used like in the pair die/kill.
However, this type of approaches face serious problems, since in many languages the causative verb is more marked than the anti-causative, while in many cases, the pair of verbs participating in the causative/inchoative alternation use either the same exact form (as in iii.a), or two different derived forms from the same base core (as in iii.b) or completely different roots to describe the two meanings (as in iii.c).
THE PHENOMENON: CHAPTER 2 TRANSITIVITY ALTERNATIONS
39
of Greek, while the reverse type of innovation is not observed; further evidence is
provided by comparing data from several stages of Greek.
Finally, some ACT verbs can participate in the causative/inchoative
alternation, as exemplified in (14a) below, illustrating the case of the morphologically
active ergatives found in languages like English as well (Sioupi, 1998;
2006; Zombolou, 2004, a.o.). Throughout the thesis, these verbs will be referred to as
‘voice non-alternating anti-causatives’. Consider the relevant examples of the verb
lijizi ‘bend’ in (14a-c). The examples in (14a&b) instantiate the transitive/causative
version, the difference between them being the animacy of the subject:
(14) a.To aghori lijise to kutali (apo ta nevra tou).
the boy bent-ACT.3Sg the spoon (because of his nerves)
b. O aeras lijise to dentro.
the wind bent-ACT.3Sg the tree
c. To sidero lijise (?* apo to palesti12/ apo to varos / apo mono tou).
the iron bent-ACT.3Sg (by* the wrestler/ from the weight / from self it)
“The iron bent (* by the wrestler/ because of the weight / by itself).”
In (14c) only the anti-causative reading is accepted, although the structure
involves suppression of the external argument (not in the syntax though) the resulting
sentence bearing similarities with the passive without the non-active morphological
specification though. The absence of NACT morphology has been attributed to the
lexical or thematic feature-based operation reducing the predicate’s arguments by one
prior to the syntactic derivation (Chierchia, 1989; Levin & Rappaport-Hovav, 1995;
Reinhart, 2003). Alternatively, the lack of NACT has been attributed to the absence of
12 But see on the results of an sentence-completionn task with adult NS of Greek (Fotiadou, 2007), where apo- agent PPs were provided in sentences with ACT anti-causatives as in O lekes katharise topika apo tin Maria (‘the stain cleaned locally by Maria’) and To sidhero lijise poli efkola apo ton Koutaliano (‘the iron bent very easily by Koutalianos’), although the English translations remain ungrammatical.
THE PHENOMENON: CHAPTER 2 TRANSITIVITY ALTERNATIONS
40
a light v or Voice/v category responsible for introducing the external argument (Hale
2004). Moreover, Alexiadou & Anagnostopoulou (2004) argue that these verbs are all
‘deadjectival’, in that they consist of an adjectival stem. Following Alexiadou (2001)
they propose that the –iz, -iaz, -ev, -en and –on affixes attached to the verb stem in
this class are overt reflexes of the feature of BECOME/RESULT. For them, these
deadjectival verbs are built upon a BECOME v which embeds an adjective, as
illustrated below:
(15) BECOME + predicate (deadjectival)
[as in I sakoula adjiase (the bag emptied)]
(adapted from Alexiadou & Anagnostopoulou, 2004: 135)
For Tsimpli (2006) also, the structure involved in (14) is similar to the one in
(10a) but different from the one in (10b) since the NACT voice marking denotes an
implicit external cause or agent, whereas in the ACT the external argument is
suppressed, as illustrated below:
(16) [TP DP [vP lerose [VP [V<θ> ]]]]
(adapted from Tsimpli 2006: 23)
The phrase apo to varos (by the heavy weight) indicating the cause is thus
analysed as a true adjunct in (14c) and, similarly, the phrase apo tis bojies (lit. ‘from
the paint) in (10a). However, the apo-phrase is analysed as co-indexed with the
NACT morpheme carrying the external theta-feature in (10b) (Tsimpli, 2006; also
Lavidas, 2007a, b).
Despite the availability of ACT anti-causatives such as (14), there is a
tendency to productively use NACT forms of verbs that have been classified as ‘voice
non-alternating anti-causatives’, as is shown in (17) below, especially in instances
where the targeted meaning is the passive13 (Tsimpli, 2006; Fotiadou, in press).
13 Similarly, ‘unergative’ entries are often transitivized in idiomatic expressions to promote the agent/cause for discourse purposes (cf. Theophanopoulou-Kontou, 2002, 2003) (for relevant discussion see also Roussou & Tsimpli, 2007; Mavrogiorgos, 2007)
THE PHENOMENON: CHAPTER 2 TRANSITIVITY ALTERNATIONS
41
According to Tsimpli (2006) it is mostly the externally-caused change-of-state verbs
such as klino ‘close’, anigho ‘open’, vrazo ‘boil’ that can also appear in non-active
form. As will be shown in Chapter 3, frequency data drawn from corpora of adult
written formal and informal speech show that NACT morphology remains available
depending on the syntactic environment and the interpretation of specific verbs (in the
form of neologisms) in both formal and, specifically, informal register, as shown in
(17).
(17) και είναι αρκετά ανθεκτική να αντισταθεί στην επίδραση "cheesecutter"
όταν λυγίζεται η λουρίδα, παραδείγµατος χάριν µε τη σύνδεση σε έναν
In order to summarize the above discussion, we need to observe that there are
at least two types of ambiguity arising from NACT voice morphology on the Greek
verb. The first lies in the distinction between the reflexive vs. the non-reflexive
(passive or anti-causative) readings of sentences such as (7b) and the second between
the anti-causative vs. passive readings of sentences such as (7c) and (10b). For 14 A tendency of transitivization is attested also with unergative verbs in Greek as in horepsan (enan horo/ ton Kosta) ‘they danced (a dance/the Kostas)’, where (non)-cognate objects are also available (cf. Theophanopoulou-Kontou, 2003 a.o.; but see Roussou & Tsimpli, 2007 for an elaborate discussion).
THE PHENOMENON: CHAPTER 2 TRANSITIVITY ALTERNATIONS
43
Tsimpli (2006), which is the theory adopted in this thesis, the former ambiguity is
structural and it is associated with the derived or non-derived status of the sentence
subject; reflexives have a true subject while non-reflexives a derived one. The
passive/anti-causative distinction is an interpretative one. In both cases, the final
unambiguous interpretation is subject to the consideration of the semantics of the
predicate as well as other extra-clausal factors, while the animacy of the subject is
independently of great importance. With respect to the ACT voice morphology,
ambiguity is raised only temporarily in that the same form is used to denote both
transitive and intransitive readings.
Furthermore, questions are raised with respect to which verbs can ‘voice
alternate’: evidence from colloquial speech shows that even internally-caused change-
of-state verbs which are assumed to surface only in ACT, are productively used with
the NACT morpheme too. Also, transitivity alternations are suggested to be
productive in that unaccusatives/anti-causatives and unergatives are found in
instances of colloquial speech.
The present thesis is an empirical research which addresses the above
questions. We examine whether the more frequent interpretations in corpora of adult
written formal and informal speech mirror these ambiguities. Furthermore, empirical
data from on-line and off-line experiments are used to investigate whether they
provide evidence for the importance of the most frequent readings on the processing
as well as on the acceptability judgments and more preferred interpretations of adult
NS speakers of Greek, in line with the assumption that disambiguation of information
in sentences is driven by exposure facts. Finally, empirical research addresses the
question of whether children differentiate between their preferred interpretations
according to the most frequent interpretations drawn from the corpora examined, on
the assumption that the pattern of language acquisition of particular phenomena is
determined by input frequency.
PART II EMPIRICAL RESEARCH
CORPUS ANALYSIS CHAPTER 3
44
CHAPTER 3
CORPUS ANALYSIS
1. Introduction
The shift to corpus-based approaches has entailed a focus on naturally occurring
language (eg. Bybee & Hopper, 2001). While research is usually based on constructed
example sentences and self-inspection, the present thesis includes research based on
corpus analysis and a comparison of frequencies found in the research corpora with
empirical data of psycholinguistic experiments designed to examine sentence
processing (using self-paced-reading and sentence-picture matching tasks) including
the verbs whose frequencies were calculated in the corpora (see following chapters).
Corpora in general consist of collections of language samples produced in
natural contexts and without experimental interference. They are used to bring actual
usage evidence to theoretical and applied linguistic questions. In this chapter we
present the frequency of use of Voice (Non)-alternating Anti-causative and Activity
verbs of Greek drawn from the ILSP corpus of written language (Hellenic National
2005, 2006). They constitute the ‘Voice Alternating Anti-causative’ verbs (Class
II) of the following analyses.
iii. htenizi (comb), pleni (wash), dini (dress), kitai (look), troi (eat), metaferi
(transfer), krivi (hide). While these are all activity verbs, the first three are
classified as ‘inherently’ reflexives in previous literature when appearing in
NACT (cf. Alexiadou & Anagnostopoulou 2004). They constitute the ‘Activity’
verbs (Class III) of the following analyses.
The sample compiled consists of the total number of queries run for all the
above verbs in active (ACT) and non-active (NACT) voice morphology. All the
15 Notice that the externally-caused change-of-state verbs such as klini ‘close, vrazi ‘boil’ and ljoni ‘melt’ also appear in the non-active form. NACT morphology is available in these cases depending on the syntactic environment and the interpretation (examples adapted from Tsimpli, 2006): (i) Oles I eksodhi tu aerodhromiu klistikan ke asfalistikan.
all the gates the-Gen airport closed-NACT.3Pl and secured-NACT.3Pl “All airport gates were closed and secured.”
(ii) Afta ta ergalia prepi na vrazonde prin xrisimopiithun. these the tools must Sub. boil-NACT.3Pl before use-NACT.3Pl “These tools must be boiled before being used.”
(iii) To vutiro ljoni / ljonete (jia na bi sta makaronia). The butter melt-ACT.3Sg/melt-NACT.3Sg (to be put on the spaghetti) “The butter melts / is being melted (to be put on the spaghetti).”
CORPUS ANALYSIS CHAPTER 3
47
sentences of the sample were next annotated with respect to the test variables, as
shown in picture 1. Note furthermore that other criteria were also taken into
consideration in order to facilitate measurements and analyses that followed. They are
also available for future research on their potential interaction with verb readings.
Picture 1: example of annotation Sentence Με τη ζέστη του ήλιου έλιωσε το κερί που τα κρατούσε ενωµένα).
Τελικά µετά από καταδίωξη του βασιλιά Μίνωα ο ∆αίδαλος κατέφυγε στην ...
Intransitive Literal Anticausative 0 µε cause informal
As shown in Picture 1, for each verb examined, we cite the sentence and its source
and we next annotate the main variables of the study, as well as additional coding for
methodological reasons.
More specifically, we coded each verb for ACT when it was marked for active
voice morphology and NACT when marked for non-active morphology (cf. Ralli,
2005) (see also Chapter 1: 3-4). As discussed in Chapter 2, we remind the reader that
ACT morphology can give rise to transitive structures with a cause (inanimate) or
agent (animate) syntactic subject as well as intransitive structures with an inanimate
(ergative/anti-causative) or an animate (unergative) syntactic subject. ‘Intransitives’
can also bear the NACT morpheme, which gives rise to a passive reading with the
cause of change-of-state being external (either inanimate or animate). Furthermore the
NACT voice marking can raise ambiguity between passive/anti-causative and
reflexive readings, constrained by the interaction of subject animacy with the
semantic features of the predicate and the lexical preferences of adult speakers
depending on the prototypicality of the subject in relation to the predicate used (cf.
Tsimpli, 2006)16. We thus annotated also whether the syntactic subject of each verb
16 Ambiguity between anti-causative and reflexive (i) or passive and reflexive (ii) is raised among animate subjects, while between anti-causative and passive is raised among inanimate subjects (iii). (i) To pedhi kriftike (apo to dhendro/ ??apo ton adelfo tu/ mono tu /apo mono tu) the child hid-NACT.3Sg (by the tree /??by the brother his/*own his/ by himself) “The child was hidden (by the tree/ *by himself / *on his own).”
CORPUS ANALYSIS CHAPTER 3
48
occurrence was animate or inanimate. Frequency of the various interpretations in
combination with subject animacy is expected to provide evidence to whether
animacy affects verb interpretation, as in the example of reflexive readings preferred
in the presence of animate subjects (see Tsimpli, 2005, 2006 for a syntactic analysis).
With respect to the verb readings discussed in the thesis, we need to specify
that further analyses would provide us with safer results with respect to the
distribution of readings attested when considering only voice morphology and subject
animacy. Thus, ACT verbs perceived as transitives show different patterns with
respect to the verb class discussed each time; moreover, preliminary calculations of
the presence of an overt or null object (for transitive) in the Web Corpus also show
variation in the distribution across the various classes. However, this issue together
with the possibility of the subject’s intention to act upon the object (cf. Kallulli, 2006
a, b) or the affectedness of the object are not examined in the present thesis but are
considered for future research. ACT verbs that are annotated as transitives include
utterances with and without an explicit object; transitivity was established on the basis
of the context. ACT verbs that are annotated as unergatives include utterances where
the subject is an agent, while non-literal interpretations are also included (as in the
example of I Maria eklise san jineka ‘Mary closed (=feels fulfilled) as a woman’).
Finally, ACT verbs annotated as anti-causatives include the cases where the subject
(animate or inanimate) is acted upon, irrespective of whether there is information
about an agent (animate) or a cause (inanimate) initiating the event described by the
verb’s semantics, while volition was not taken into consideration. This is not the case
for NACT anti-causatives in which we included only the cases where a cause or an
instrument is included in the structure, while when an agent is implicitly or explicitly
present the NACT verbs are characterized as passives. Finally, under the
characterization of reflexives we included all the cases where the subject (animate or
inanimate, but with a non-literal sense) of a NACT verb is a true agent; reciprocals are
the NACT verbs where the subject (always in the plural) is the agent and the theme is
(ii) To pedhi plithike (apo ti mitera tu /?apo ti vroxi/ mono tu/ ?*apo mono tu)
The child wash- NACT.3Sg (by the mother his /own his /*by himself) “The child was washed by his mother / *by himself / *on his own)
(iii) I simea skistike (apo moni tis /apo ton aera / apo tus mathites) The flag tore- NACT.3Sg (by itself / by the wind/ by the students) “The flag was torn (by itself / by the wind /by the students)
CORPUS ANALYSIS CHAPTER 3
49
bound by the agent showing that the action is initiated by the agent and affects him at
the same time.
In the calculations discussed in the thesis we also included any potential PPs
(for each verb reading). More specifically we coded the existence of an apo
(‘by/from’)-phrase and its status. We specified whether it denotes the agent, cause or
instrument (for passive/anti-causative readings), whether it has a temporal,
directional, locative meaning, whether it is part of complex locative prepositions
(construed with locatives such as epano (over), dhipla (next to)), or it illustrates other
meanings (e.g. comparison), or even it is the true complement of the verb); other
prepositional phrases (e.g. me-instrument) modifying the verb or the predicate were
also coded.
Annotation included also the following verb characteristics, in order to
facilitate further analyses and remain available for future research:
i. 1st, 2nd or 3rd person singular / plural
ii. past / non-past tense
iii. perfective / imperfective aspect
iv. modality
v. literal / non-literal (metaphorical) use of the verb
vi. speech genre
CORPUS ANALYSIS CHAPTER 3
50
3. The research sample
3.1. The ILSP Corpus For the present study we used the morphological and syntactic corpus developed by
the Institute for Language and Speech Processing (ILSP): Research and Innovation
Center “Athena’, financed by PAROLE – Le II (1/4/1996 - 31/4/1998). The Institute
for Language and Speech Processing (ILSP) was founded in Athens (Greece) in 1991
under the auspices of the General Secretariat of Research and Technology of the
Ministry of Development.
The ILSP Corpus (Hellenic National Corpus™ (HNC) Web version 2.0
3.2. The Web-Based Corpus For the present study we compiled a sample of utterances found in documents that are
spontaneously produced by non-professional writers, providing us with some
characteristics of oral communication. More specifically, in order to include different
types of informal language, we compiled a sample of sentences (‘snippets’) found in
chat pages, fora, mails, blogs and on-line newspapers or other informal written
material (prospecti, reports and documentation are included). The source texts dealt
with highly diversified topics of every day life (sport, religion, politics and economy
among others).
The Web-based Corpus has been created with the use of a system developed to
search the Web and save the results in a database21. The database was created in an
MS Access environment whereas for the Internet Search Engine we used the Google
Web APIs (‘Application Programmable Interface’) technology in Active Server
Pages. The coding procedure was done with MS Access forms.
More specifically the procedure we followed was the following: We signed up
for a Google account and obtained a license key which provided us a daily limit of
1000 queries to Active Server Pages. Furthermore, given that the system does not give
access to information beyond the 1000th result for any given query, we developed a
‘randomize’ program which would provide us with random numbers from the total
number indicated in the Google home page: for example the verb ‘klini’ returns
560.000 snippets (accessed in October 6, 2009) but no access is allowed beyond
1000th result. In order to retrieve the larger possible number of results we asked the
system to access groups of ten results per time, choosing among the sets of ten
suggested by the ‘randomize’ program. An example of the utterances returned in a
given query and the random numbers returned by the program ‘randomize’ are
illustrated in Picture 4.
21 Many special thanks go to Tasos Paschalis for his most valuable technical support, more specifically for his contribution to the design of the Web Based Corpus.
CORPUS ANALYSIS CHAPTER 3
55
Picture 4: example of the environment with the output of specific queries in
Active Server Pages with Google Web APIs and the program ‘randomize’
Furthermore, in order to manipulate representativeness, we deliberately biased
query terms using combined criteria which enabled us to have access to different
results every time. More specifically, we created queries with specific verbs (eg. klini)
and we added key words frequently used in the URL sources (as forum, view topic
a.o.) which returned a different ranking every time: the results returned were not the
same sentences. All these results were automatically saved in the MSAccess database
created for this purpose.
In order to eliminate repetition, when this was not automatically done by the
program, we manually scanned all the sentences and deleted all the snippets that
occurred more than once, as in the example of jokes, repeated under different URLs
or the quotation of various extracts of texts under personal blogs created by non-
professional users of the Internet.
Furthermore, we diminished noise by manually eliminating tokens returned by
pornographic sites: some verbs were used in specific contexts, risking to affect the
CORPUS ANALYSIS CHAPTER 3
56
distribution of the most frequent interpretations attested. Utterances which did not fit
in the research questions were also manually eliminated: For example, Google APIs
returned automatically to our Access Base all findings of ‘τρυπα’ not distinguishing
the stress, providing us thus with many examples of the noun ‘hole’ instead of
occurrences of the verb ‘pierce’ in 3rd singular. Another characteristic example was
‘µετέφερε’ for which the system returned examples of both perfective and
imperfective aspect in 3rd singular past tense, as well as the imperative in 2nd singular,
which was not of interest in the study. Moreover, all the examples that came from
educational sites, consisting of examples demonstrating the language use were also
excluded from the research.
The coding and the analyses in terms of semantic and syntactic environments
of the saved sentences were completed in the MS Access environment created, as
illustrated in Picture 5:
Picture 5: Example of MS Access Environment
After eliminating all irrelevant tokens, the sample used for the present study consists
of 44873 tokens found in Active Server Pages with the Google Web APIs interface.
CORPUS ANALYSIS CHAPTER 3
57
4. Results
Results that are next reported involve only 3rd Singular and Plural occurrences drawn
from the compilation of samples treated in the MSAccess database and involve the
tokens drawn from the ILSP corpus (hereafter ILSP Corpus) and the Active Server
Pages in the Web (hereafter Web Corpus). The 1st and 2nd Singular and Plural are not
included given that they always involve animate subjects and furthermore they are
less frequent than the 3rd Singular and Plural.
The Corpora used in the present research (ILSP and Web) differ in size, as
shown in the Table 1.
Table 1: Corpus – Total number of utterances indexed
Corpus Total N of utterances
ILSP Corpus 18395 (29.07%)
Web Corpus 44873 (70.92%)
TOTAL 63268 (100%)
It is noteworthy that data examined represent mostly the ‘Web Corpus’, the difference
of utterances drawn from each of the corpora used being significant (ILSP (29.07%)
vs Web (70.92%): χ2= 11081.186, p<.01).
CORPUS ANALYSIS CHAPTER 3
58
4.1. The distribution of Voice morphology in the corpora used
If we compare the two corpora with respect to the Voice morphology of the verbs
indexed we notice that both ACT and NACT verb forms examined represent mostly
the ‘Web Corpus’ (ACT: ILSP (33.95%) vs Web (66.05%): χ2=4479.688, p<.01;
NACT: ILSP (18.38%) vs Web (81.62%): χ2=7920.946, p<.01), as shown in Table 2:
Table 2: Distribution of utterances indexed with respect to Voice morphology
(ILSP / Web Corpus)
Corpus Total N of ACT Total N of NACT
ILSP Corpus 14756 (33.95%) 3639 (18.38%)
Web Corpus 28710 (66.05%) 16163 (81.62%)
TOTAL 43466 (100%) 19802 (100%)
With regard to each corpus examined, notice that in the ILSP Corpus ACT verbs were
more frequent than NACT (ACT (80.21%) vs NACT (19.78%); χ2=6718.548, p<.01),
a pattern also attested in the Web Corpus (ACT (64.26%) vs NACT (36.01%);
χ2=3508.284, p<.01). However, NACT Verbs were more frequent in the Web than the
ILSP Corpus, if we consider the total size of each corpus (χ2=1599.615, p<.01,
η=.159). Overall, ACT verbs represent the 68.70% (43466/63268) of the total number
of verbs indexed in the two corpora examined, while NACT verbs represent only
31.30% (19802/63268) of the total, their difference being significant (χ2= 8850.997,
p<. 01).
The different sizes of the two corpora result from variation in the frequency of
the verbs examined. We next compare occurrences of these verbs with respect to
voice morphology (from ILSP and Web Corpus) in each verb class examined.
CORPUS ANALYSIS CHAPTER 3
59
Graph 1: The distribution of Class I verbs with respect to Voice morphology
(ILSP/Web Corpus)
3193
118
1789
1590
224
7
729
313216
0
1257
8183
0
1563
21106
0
642
57
276
0
1534
2145 2
1653
144
0
500
1000
1500
2000
2500
3000
3500
ACT NACT ACT NACT
ILSP Corpus Web Corpus
klini vrazi jerni sapizi ljoni lijizi stegnoni
More specifically, six of the total seven Class I verbs (Voice Non-alternating Anti-
causatives) were rare in the ILSP Corpus (mean: 175; frequency range: 45 – 276).
Only the verb klini (close) was highly frequent (act: 3193). Moreover, very few
occurrences of NACT forms were found, i.e. the verbs klini (close) (nact: 118), vrazi
(boil) (nact: 7) and stegnoni (dry) (nact: 2), as shown in the Graph.
On the other hand, the Web Corpus revealed a more frequent use of active
forms (mean: 1309; frequency range: 642 – 1789). Furthermore, some instances of
NACT verbs were also found: i.e. the verb klini (close) was frequently used, while the
frequency of the remaining verbs was low (mean: 93.833; frequency range: 8 – 313).
When we compare data from the two corpora we notice that ACT verb forms
are in general more frequent in the Web than the ILSP corpus, with the exception of
the verb klini (close) which is more frequent in the ILSP corpus (ILSP (3193) vs Web
CORPUS ANALYSIS CHAPTER 3
60
(1789); χ2=395.668, p<.01). This finding may be attributed to the different discourse
factors involved (register, mode of discourse) in the two corpora.
Turning to NACT verb forms, though they were not frequently attested, which
is to be expected since the verbs examined are prototypically ergatives, we note that
they were used more in the Web than in the ILSP corpus. Only the verb stegnoni (dry)
presented a similar pattern of distribution with respect to Voice morphology in the
two corpora (ILSP vs Web: χ2=.887, p=.346, η=.022). The NACT forms of the verbs
klini (close) and vrazi (boil), were significantly more frequent in the Web than in the
ILSP corpus (ILSP vs Web: χ2=1663.757, p<.01, η=.499; ILSP vs Web: χ2=73.028,
p<.01 η=.240, respectively), while NACT forms of the rest of the verbs were attested
only in the Web corpus.
Overall, Voice Non-alternating Anti-causatives were not largely used in active
morphology and were only marginally found in non-active morphology in both
corpora. However, the significant differences between the two corpora in specific
verbs with regard to NACT forms suggest productivity of the NACT morpheme even
in this verb Class.
CORPUS ANALYSIS CHAPTER 3
61
Graph 2: The distribution of Class II verbs with respect to Voice morphology
NACT verb forms in the presence of an animate subject received reflexive
readings more frequently than any other available reading (anti-causative, passive),
irrespective of corpus (ILSP: χ2 =11.560, p=.001; Web: χ2 =10.240, p=.001 and total:
χ2 =14.440, p<.01). Furthermore, in the presence of an inanimate subject, we found
CORPUS ANALYSIS CHAPTER 3
75
that passive readings are more frequent than any other available reading (reflexive,
anti-causative), irrespective of corpus (ILSP: χ2 =54.760, p<.01; Web: χ2 =70.560,
p<.01 and total: χ2 =67.240, p<.01).
Note finally that with respect to ACT morphology, unergative readings are
found in the presence of an inanimate subject and with respect to NACT morphology,
reflexive and reciprocal readings are found in the presence of an inanimate subject
when this is used non-literally (metaphorically).
CORPUS ANALYSIS CHAPTER 3
76
b. Specific Tokens
As noticed in the previous section, the frequency of Class I verbs differed between the
two corpora, their total number being higher in the Web than the ILSP corpus, in both
(ACT-NACT) voice morphology, irrespective of [+/- animacy] of the syntactic
subject (Graph 8).
Graph 8: Interpretations of Class I verbs in the ILSP and Web Corpus (voice
morphology and subject animacy)
0
1000
2000
3000
4000
5000
6000
trans
itive
antic
ausa
tive
uner
gativ
etra
nsiti
vean
ticau
sativ
eun
erga
tive
trans
itive
antic
ausa
tive
uner
gativ
etra
nsiti
vean
ticau
sativ
eun
erga
tive
refle
xive
antic
ausa
tive
pass
ive
refle
xive
antic
ausa
tive
pass
ive
refle
xive
antic
ausa
tive
pass
ive
refle
xive
antic
ausa
tive
pass
ive
anim inanim anim inanim anim inanim anim inanim
ILSP Web ILSP Web
ACT NACT
stegnonilijizi ljoni sapizi jerni vrazi klini
In the ILSP corpus, the distributions presented in Graph 8 are not always
representative, since they rather illustrate frequencies of individual verbs. For
example, some readings attested for active verb forms in the presence of an animate
CORPUS ANALYSIS CHAPTER 3
77
subject come mostly from frequencies drawn from specific verbs: the transitive use
comes from klini (close) (n=1150), the anti-causative reading comes from lijizi (bend)
(n=95) and the unergative reading comes from jerni (lean) (n= 57). Moreover, active
verbs in the presence of an inanimate subject come mostly from the verb klini (close):
transitive (n=267), anti-causative (n=1656), unergative (n=71). On the other hand,
non-active verbs are not frequently found in this corpus, as expected. Hence, the
reflexive and the passive uses in the presence of an animate subject are drawn
exclusively from the verb klini (close). In the presence of an inanimate subject, the (i)
anti-causative, (ii) passive and (iii) reflexive (metaphorically used) readings are drawn
from the verbs klini (close) (n=6) and stegnoni (dry) (n=2) (for (i)), the verbs klini
(close) (n=53) and vrazi (boil) (n=7) (for (ii)), and the verb klini (close) (n=1) (for
(iii)) respectively.
In the Web corpus, no such large individual differences occurred among the
various readings attested with respect to voice morphology or subject animacy. In
other words, we found occurrences of almost all the verbs examined in each research
condition.
If we consider the total number of data (from ILSP and Web corpus), Voice
Non-alternating Anti-causatives appear to be more often used with active voice
morphology, with the exception of klini (close), which is largely used in the non-
active voice morphology as well. Also, in their majority, active forms used
intransitively receive an anti-causative reading. On the other hand, the largest part of
NACT occurrences examined come from the Web Corpus, where the various verbs
are mostly used in friendly communication, or consist of jargon language; hence,
examples drawn commonly illustrate non-literal interpretations, as in the example of
unergative uses of active verbs with inanimate subjects.
CORPUS ANALYSIS CHAPTER 3
78
1. The verb klini (close)
The distribution of the various readings that the verb received in the two corpora
examined is shown in Table:
Table 8: the verb klini (close) (ACT-NACT)
verb
morphology
interpretation [+/-animacy] of syntactic subject
animate inanimate
ILSP Web total ILSP Web total
active transitive 1199 805 2004 267 147 414
% 99.5 100 99.7 13.39 14.94 13.9
anti-causative 6* 0 6 1727 837 2564
% 0.5 0.3 86.61 85.06 86.1
non-active reflexive** 38 706 744 1 8 9
% 65.52 69.83 69.59 1.67 1.12 1.39
anti-causative 0 21 21 6 70 76
% 2.08 1.97 10 11.92 11.74
passive 20 284 304 53 509 562
% 34.48 18.09 28.44 88.33 86.71 86.86
* As such are characterized sentences with non-literal meanings, attested : Εφόσον αληθεύει το σχετικό ρεπορτάζ της « Αυριανής », επιβεβαιώνεται ότι όντως η κοπέλα µε τον Ανδρέα έκλεισε σα γυναίκα. (ILSP: 563444) ** There is also an utterance with a non-literal reciprocal reading of nact-inanim in the Web corpus: Αιµορραγούν και κλείνονται. Τρακάρουν και φλέγονται και µυρίζει σάρκα καµένη και δεν ακούγεται…http://e-missos.gr /forum/phpBB2/viewtopic.php?t=1430&start=30&sid=4d72d3d8bac6020019613b394dfe8091
Illustrative examples of the various readings with respect to voice morphology and
subject animacy are found below:
CORPUS ANALYSIS CHAPTER 3
79
Animate subject:
Active Voice morphology
Transitive: ΚΑΙ ΩΣ ΓΝΩΣΤΟΝ ΜΑΖΕΥΤΗΚΑΜΕ ΓΙΑ ΝΑ ΦΩΝΑΞΟΥΜΕ
ΣΤΟΥΣ ΚΟΥΦΟΥΣ /ΒΟΛΕΦΤΕΣ/ ΑΛΛΑ ΑΥΤΟΙ ΤΟ ΚΛΕΙΣΑΝ ΤΟ ΜΑΓΑΖΙ
Anti-causative: όταν βράζουν οι φακές, παραµένουν... ολόκληρες όπως είναι ωµές ή
... mporei na … http://www.radicio.com/forum/lofiversion/index.php/t1126.html
Non- Active Voice morphology:
Anti-causative: Σε όλα τα µέρη που τρώνε φάβα ανοίγουν ένα λάκκο και ρίχνουν
µέσα λάδι, γιατί η φάβα βράζεται µόνο µε το νερό της.
http://www.agioskosmas.gr/periodiko.asp?sitema=1&sitem=11&isue=28&artid=150 Passive: α) Συνταγές από βότανα (2-14 βότανα µαζί), τα οποία βράζονται από τον
Turning to a discussion of the PPs found, note that although PPs among ACT anti-
causatives were less frequent in the ILSP corpus than the Web when an animate
subject was involved, they were more frequent with inanimate subjects (total PPs
among anti-causatives with anim (ILSP vs Web): χ2=20.775, p<.001; among anti-
causatives with inanim (ILSP vs Web): χ2=6.818, p=.009). Thus, a large difference in
the frequency of PPs was attested between anti-causatives with animate and inanimate
subject in the Web (anim (66) vs inanim (9): χ2=43.320, p<.001), while no such
difference was attested in the ILSP (anim (23) vs inanim (24): χ2=.021, p=.884). Note
also that PPs among passive readings with inanimate subjects were attested only in
the Web, including only apo-agent and me- instrument phrases, the remaining
readings did not provide us with relevant evidence.
23 We cite the example: "Για να µπω στο σπίτι µου", λέει ένας κάτοικος του Ναυπλίου, το οποίο "βράζει" από ξένες το καλοκαίρι (το "καµάκι" είναι εποχιακό επάγγελµα), κι όπου εκτυλίσσεται ο "µύθος", "πρέπει να περάσω µέσα από υπνόσακους. (ILSP: 328253)
CORPUS ANALYSIS CHAPTER 3
88
3. The verb jerni (lean)
The distribution of the various readings that the verb received in the two corpora
examined is shown in Table 12:
Table 12: the verb jerni (lean) (ACT-NACT)
verb
morphology
interpretation [+/-animacy] of syntactic subject
animate inanimate
ILSP Web total ILSP Web Total
active transitive 25 241 266 18 139 157
% 27.47 56.04 51.06 14.4 16.81 16.49
unergative 57 127 184 4 63 67
% 62.63 29.53 35.32 3.2 7.62 7.04
anti-causative 9 62 71 103 625 728
% 9.9 14.43 13.62 82.4 75.57 76.47
non-active reflexive 0 2 2 0 0 0
% 40 40
anti-causative 0 3 3 0 1 1
% 60 60 33.34 33.33
passive 0 0 0 0 2 2
% 66.66 66.67
We next cite examples that illustrate the various readings, drawn from the corpora
examined.
Animate subject:
Active Voice Morphology
Transitive: Oι ψηφοφόροι των σοσιαλιστών αγνόησαν µαζικά την επίσηµη γραµµή
του κόµµατος, προτίµησαν το «όχι» και έγειραν την πλάστιγγα.
Reflexive: Μα πάλι, ως είχε θυµηθεί, πού γέρνεται , πού µένει για λόγου του µια του
Κερά ακριβαναθρεµµένη, εγρίκα µέσα στην καρδιά µαχαίρι, και πληγώνει, ...
el.wikisource.org/wiki/.../Ε'
Anti-causative: Στα 70 γέρνεται και ως τυφλα τυφλαινεται. Στα ογδοηντα δεν φελα
µονο το ψωµι χαλα. eletg.wordpress.com/.../4/
Inanimate subject:
Active Voice Morphology
Transitive: ... Οι επτά πόντοι της διαφοράς (52‐45) έγειραν την πλάστιγγα
υπέρ τους κι έφεραν το άγχος στο ... http://bhma.net/archive/2002/0329/20.asp Unergative: Από το 6,5 χλµ., ο δρόµος φαρδαίνει και τα πλατάνια γέρνουν να τον
κοιτάξουν χαρίζοντάς µας ευεργετική δροσιά τις ώρες που ο ήλιος «χτυπάει». ...
Turning now to a discussion of the PPs that were found, note that only apo-cause
phrases were frequent, among active anti-causatives with inanimate subjects, only in
the Web. Furthermore, the Web included a larger variety of PPs, self and apo-
instrument among them, as in the above mentioned examples.
CORPUS ANALYSIS CHAPTER 3
91
4. The verb sapizi (rot)
The distribution of the various readings that the verb received in the two corpora
examined is shown in Table 14:
Table 14: the verb sapizi (rot) (ACT-NACT)
verb
morphology
interpretation [+/-animacy] of syntactic subject
animate inanimate
ILSP Web total ILSP Web total
active transitive 2 182 184 8 69 77
% 3.78 38.73 35.18 6.16 5.41 5.48
anti-causative 51 288* 339 122 1206 1328
% 96.22 61.27 64.82 93.84 94.59 94.52
non-active anti-causative 0 0 0 0 10 10
% 90.90 90.9
passive 0 10 10 0 1 1
% 100 100 9.1 9.10
*Since there no instances of unergative uses (ACT) with animate subjects but for an insignificant 0.02% we included the 13 examples attested in the count of anti-causatives. No further statistics were conducted for this interpretation.
The examples below illustrate the various readings of the verb in active and non-
active forms:
Animate subject:
Active Voice Morphology
Transitive: Τον τύπο απ' ότι είπε τον σάπισε στο ξύλο. Απλά της έκλεψε την τσάντα
Web apo-agent 92 41 apo-cause 2 6 2 89 92 2 2 apo-instr 67 21 16 17 me-instr 2 2 2 32 17 self 1 1 1 other pp (locative) 21 54 24 4 12 14 *Unergative readings co-occurred with me –instrument (with animate (n=1) and inanimate subjects
(n= 3), while unergative readings with inanimate subjects were mostly followed by other PPs like ‘sto mati’ (in the eye) (n=41) only in the Web corpus.
**Also among reciprocal readings with animate subjects we found apo-cause (n=1) or me-instrument (n=4) and the expression ‘metaxi tus’ (each other) (n=22) only in the Web corpus.
*** We cite the example: Με ελαφρά εγκαύµατα µεταφέρθηκε σε κλινική του Μονάχου, στη Γερµανία, ο (κάποτε µαύρος) τραγουδιστής Μάικλ Τζάκσον, 41, ο οποίος χτυπήθηκε από φωτοβολίδες κατά τη διάρκεια ενός κονσέρτου, από ενθουσιώδεις οπαδούς του. (ILSP: 1547748)
Turning next to an analysis of the various PPs we note that locative phrases with
reference to the object caused hitting were the most frequent among active anti-
causatives, both in the ILSP and Web corpus, while some apo-cause phrases were
also found (with both animate and inanimate subjects in the ILSP, while only with
inanimate in the Web corpus). Also, few PP-instruments are attested only with
animate subjects (apo- in the ILSP and me- in the Web corpus) and self only in the
Web. Among non-active verb forms, note that apo-cause were found among anti-
causative readings in both corpora, more frequently with inanimate than animate
subjects (ILSP: χ2=3.130, p=.077, and Web: χ2=0.050, p=.824). Also, apo-
instruments were significantly more frequent in the Web than the ILSP corpus (Web
(5) vs ILSP (88): χ2=74.075, p<.001) where me-instrument phrases were also attested.
CORPUS ANALYSIS CHAPTER 3
132
5. The verb tripai (pierce)
The verb tripai (pierce) is more frequent in the active (86.34%) than the non-
active verb form (13.66%) (act: 1201 vs nact: 190; χ2=734.810, p<.01). The Table
below illustrates the various readings that the verb tripai (pierce) receives with regard
to voice morphology and subject animacy, as indexed in the corpora examined.
Table 31: the verb tripai (pierce ACT-NACT)
verb morphology
interpretation [+/-animacy] of syntactic subject
animate inanimate
ILSP Web total ILSP Web total
active transitive 52 359 411 58 564 622
% 100 99.17 99.28 90.62 78 79.01
anti-causative 0 0 0 6 159 165
% 9.38 22 20.99
unergative 0 3 3 0 0 0
% 0.83 0.72
non-active reflexive* 2 49 51 0 0 0
% 33.33 66.21 63.75
anti-causative 1 17 18 1 14 15
% 16.67 22.97 22.5 16.67 13.47 13.64
passive 3 8 11 5 90 95
% 50 10.82 13.75 83.33 86.53 86.36
*There is also one example of reciprocal reading
We next cite some examples of the various readings with respect to voice morphology
and subject animacy, as found in the corpora:
CORPUS ANALYSIS CHAPTER 3
133
Animate subject:
Active morphology
Transitive: Ο Χριστός πέθανε πάνω στο Σταυρό. Ο θάνατός Του βεβαιώθηκε από
τους Ρωµαίους στρατιώτες, που τρύπησαν µε τη λόγχη τους την πλευρά του, για να
ξεµατώσει και ... http://www.neostypos.gr/opsis_article.php?sub=4&art=05043013
Anti-causative: ... και πάλι αέρα 14 πόντων στο 34’ (60-74), µε Σέκουλιτς και
* Also reflexive readings with me-instrument (n=6) were attested
Turning next to an analysis of the various PPs that follow the various readings of
‘tripai’ (pierce) note that in the ILSP corpus PPs were less frequent (n=6) than in the
Web corpus (n=58) (χ2=42.250, p<.001). Note however that apo-causes and
instruments were found among active anti-causatives with inanimate subject in both
corpora (only one example in each case in the ILSP). Also, in both corpora we found
apo-instruments with non-active anti-causatives and passives with animate subject
CORPUS ANALYSIS CHAPTER 3
136
(although ILSP examples were always very few) while only in the Web we found
apo-instrument among non-active anti-causatives and passives with inanimate subject.
Also note the existence of me-instrument in the Web corpus, which was furthermore
mostly frequent among passives with inanimate subject (55.5%) (pass-inanim (n=20)
vs all other readings (n=16).
6. The verb berdevi (mingle)
The verb berdevi (mingle) is more frequent in the active (53.89%) than the
non-active (46.11%), but not significantly so (χ2=.640, p=.424). This pattern was
followed in both corpora, despite their significant difference in size. Note that the
Web is larger in size than the ILSP both in ACT (ILSP vs Web: χ2=1034.143, p<.001)
and NACT forms (ILSP vs Web: χ2=868.233, p<.001).
It is worth noting at this point that in the presence of an animate subject the
active verb form used transitively (causatively) is used either to declare that animate
provoked animate to get mixed up, or that animate mixed something or someone with
someone or something else.
The Table below illustrates the various readings that the verb berdevi (mingle)
receives with regard to voice morphology and subject animacy, as indexed in the
corpora examined.
Table 33: the verb berdevi (mingle ACT-NACT)
verb
morphology
interpretation [+/-animacy] of syntactic subject
animate inanimate
ILSP Web total ILSP Web Total
active transitive 178 851 1029 60 761 821
% 100 100 100 96.77 97.56 97.5
anti-causative 0 0 0 2 19 21
% 3.23 2.44 2.5
CORPUS ANALYSIS CHAPTER 3
137
non-active reflexive* 17 26 43 0 2 2
% 16.67 3.33 4.86 0.33 0.28
anti-
causative**
85 757 842 109 605 714
% 83.33 96.67 95.14 100 99.67 99.72
∗ We also found reciprocal readings both in the presence of an animate and an inanimate subject (one example in each case), which we cite below:
--Ο "γηραιός" είχε πρόβληµα στο χώρο του κέντρου, όπου ο Γκόνιας και ο Γιοβάνοβιτς µπερδεύονταν και έτσι σ' όλο το πρώτο ηµίχρονο ο Σαπάνης είχε πάρει µόλις τρεις πάσες στα πόδια του! (ILSP: 522684)
--Κορµιά που µπερδεύονταν µεταξύ τους, και µε τα σεντόνια, και µε τα τούλια, µαλλιά κατάµαυρα, ανάκατα, και αισθάνθηκε το στόµα του να γεµίζει µε σάλιο χλιαρό και πηχτό σαν σπέρµα. (ILSP: 1747153)
** We also found one example of passive reading in the nact-inanim condition, cited here: --Η αρωµατική νάφθα δεν πρέπει να µπερδεύεται µε την αλειφατική νάφθα. (ILSP: 371277)
We next cite some examples of the various readings with respect to voice morphology
and subject animacy, as found in the corpora:
Animate subject:
Active morphology
Transitive: Νοµίζω οτι µπερδέψανε το άρθρο για τον παραπάνω κύριο µε το άρθρο
για τον θάνατο..........γύρω -γύρω το φέρνουν , εκεί το καταλήγουν...χα!χα!χα! ...
• me-phrases introducing the second complement were often among the sentences indexed: in the nact-anim condition for the reflexive reading (n=4) and for the anti-causative (n=5) as well as in the nact-inanim condition for the reciprocal reading (n=1), for the anti-causative (n=32) and for the passive (n=1). Also me-other was frequent: in the act-inanim condition for the anti-causative reading (n=10), in the act-anim condition for the anti-causative reading (n=35), in the nact-inanim condition for the anti-causative reading (n=122),
• Note also that the active causative verbs with animate subject either declare that animate provoked animate to ‘get mixed up’ with a me-instrument (n=15), or that animate ‘mixed’ something or someone ‘with’ someone or something else (me-other: n=341). The active causative verbs with inanimate subject are less frequent (me-instrument (n=12) and me-other (n=10)).
Turning to a discussion of the PPs that are included in the sentences indexed in the
corpora for this specific verb, we should note that in the ILSP corpus while there were
no active anti-causatives with such structures, there were non-active ones. Among
non-active anti-causative with animate subjects PPs were more frequent and show a
larger variety than PPs among non-active anti-causatives with inanimate subjects. In
the Web corpus, PPs were more frequent and show variation, especially among non-
active anti-causatives. Note furthermore that PPs were significantly more frequent
with animate (n=224) than inanimate subjects (n=90) (χ2=57.185, p<.001), while even
apo-agent and self were attested.
CORPUS ANALYSIS CHAPTER 3
140
7. The verb leroni (spill)
Finally, the verb ‘leroni’ (spill) is more frequent in the active (72.27%) than
the non-active (27.73%) voice morphology (act: 563 vs nact: 216; χ2=154.569, p<.01).
The Table below illustrates the various readings that the verb leroni (spill)
receives with regard to voice morphology and subject animacy, as indexed in the
corpora examined.
Table 35: the verb leroni (spill ACT-NACT)
verb
morphology
interpretation [+/-animacy] of syntactic subject
animate inanimate
ILSP Web total ILSP Web total
active transitive 25 269 294 9 227 236
% 100 99.26 99.32 90 87.64 87.73
anti-causative 0 2* 2 1 32 33
% 0.74 0.68 10 12.36 12.27
non-active reflexive 0 3 3 0 0 0
% 6.67 5.77
anti-causative 7 42 49 4 160 164
% 100 93.33 94.23 100 100 100
*We cite the two examples: --Εν πάση περιπτώσει «λέρωσαν» στις στολές. Επιστρατεύθηκαν λοιπόν δύο εκπαιδευµένα γεράκια από το Πεντάγωνο, τα οποία ανέλαβαν την περιφρούρηση του εναέριου ... http://tovima.dolnet.gr/print_article.php?e=B&f=14108&m=A48&aa=1
--Ο Τόνυ Πράις, 33 ετών, διευθυντής της βρετανικής εταιρείας WStore βρήκε κολληµένη τσίχλα κάτω από ένα γραφείο και φρίκαρε επιπλέον επειδή λέρωσε από τσίχλα ... http://athens.indymedia.org/articles.php?type=Other%20Press&offset=990&maxRecords=10&sort_mode=publishDate_desc&type=Other%20Press
We next cite some examples of the various readings with respect to voice morphology and subject animacy, as found in the corpora:
CORPUS ANALYSIS CHAPTER 3
141
Animate subject:
Active morphology
Transitive: Από τις 25 του µήνα, κάθε φορά που θα εντοπίζεται σκύλος να λερώνει
τον δρόµο, οι ιδιοκτήτες θα τιµωρούνται µε πρόστιµο 20 ευρώ. ...
* Other PPs are also found among the reflexive readings with inanimate subjects (n=2) and the reciprocal readings with animate subjects as ‘metaxi tus’ (each other) (n=77)
CORPUS ANALYSIS CHAPTER 3
164
6. The verb metaferi (transfer)
Turning to an examination of the verb metaferi (transfer ACT-NACT), the
Table below illustrates the distribution of the various readings obtained from each
corpus examined.
Table 49: the verb metaferi (transfer) (ACT-NACT)
interpretation [+/-animacy] of syntactic subject
animate inanimate
ILSP Web total ILSP Web total
ACT transitive 2028 1715 3743 1014 837 1851
% 100 100 100 100 100 100
NACT reflexive 25 98 123 0 2 2
% 4.18 24.68 11.03 0.13 0.07
anti-causative 2 9 11 328 304 632
% 0.23 2.28 0.98 33.07 18.84 24.26
passive 570 290 860 664 1307 1971
% 95.47 73.04 77.13 66.93 81.03 75.66
Various examples that illustrate the readings that the verb metaferi (transfer ACT-
NACT) receive are listed below:
Animate subject:
Active
Transitive: ... απαιτούµενη ποσότητα. Νταλίκες το µεταφέρανε από δω κι από κει.
Με εξήντα ανθρώπους να περιµένουν εντολές http://www.mic.gr/cinema.asp?id=8427
CORPUS ANALYSIS CHAPTER 3
165
Non-active morphology
Reflexive: ... Από τα µέσα στα έξω το καλοκαίρι, καθώς µε λίγες εξαιρέσεις όλες οι
φυλές των clubbers, µεταφέρονται στον Πλατανιά. Και όταν µιλάµε για Χανιά,
εννοούµε κατά
http://stigmes.gr/gr/grpages/kritever/hania.htm
Anti-causative: Γιατί πολλές φορές οι νέοι ενώ ονειρεύονται για την σχέση τους,
όταν φτάνουν στο γάµο παραιτούνται εύκολα και µεταφέρονται στο µοντέλο που
*apo-other include mostly apo-complement and apo-locative phrases as well as apo- being part of complex locative expressions. **Other PPs are also attested among reflexive readings with inanimate subject (n=18)
Turning to a discussion of the various PPs in the corpora, we should note that both in
the ILSP and the Web corpus apo-phrases were frequent among the reflexive and
anti-causative readings with animate subjects (but anti-causatives with inanimate
subjects too in the Web) and among the passive readings with inanimate subjects (and
animate ones in the ILSP). Also, other PPs are highly frequent with reflexive, anti-
causative and passive readings both with animate and inanimate subject (except for
the case of passives). Apo-agent phrases are attested only among passives with
inanimate subject (more frequently in the Web than the ILSP: χ2=3.769, p=.052).
With respect to apo-cause note that in the ILSP corpus the only structures it co-
occurred with were passives with animate subject, while in the Web we also found
apo-cause among anti-causatives with inanimate subject and reflexives with animate
subject.
CORPUS ANALYSIS CHAPTER 3
171
4.3.4. Interaction of Voice Morphology and [+/- animacy] of syntactic subject
with Verb Readings across Verb Classes
The purpose of this section is to address the question whether it is not only Voice
morphology (ACT-NACT) and the subject animacy that interact with the distribution
of the various readings that the examined verbs received in the two corpora, but also
verb classification. We remind the reader that the analyses performed so far, revealed
a different distribution of the various interpretations that the verbs received in total,
not only with regard to Voice morphology (see Tables 4 & 5), but also with regard to
the interaction of Voice morphology with subject animacy (Table 6). For ease of
exposure we repeat the results obtained below, in Graph 13.
Graph 13: The distribution of interpretations with respect to Voice Morphology
vs the Interaction of Voice morphology with subject animacy (ACT/NACT vs
Animate/Inanimate)
CORPUS ANALYSIS CHAPTER 3
172
In this Graph we present the (%) distribution of ACT/NACT, of act-anim & nact-
inanim and of act-inanim & nact-inanim in the two corpora. It is thus understood (as
shown in Graph 13) that in the total of ACT utterances transitive is the most frequent
interpretation in both ILSP and Web, while when we consider subject animacy the
distribution changes as follows: in ACT forms with animate subjects the most
frequent reading is the transitive one, while in ACT with inanimate subjects there is
ambiguity between transitive and anti-causative readings (the latter reading being
though preferred in the Web Corpus).
Turning to NACT forms, passive is the most frequent reading in both corpora,
while in the ILSP the next preferred one is the anti-causative (followed by the
reflexive) and in the Web the reverse pattern is attested, the reflexive being more
frequent than the anti-causative. Few reciprocal uses are found in both corpora. Note
that the mean of the total frequencies from both corpora, showed an ambiguity
between the reflexive (27.85%) and the anti-causative readings (23.56%), due to the
large difference in the frequency of reflexives between the two corpora. Results from
the combination of NACT with animate subjects (nact-anim) show a different
distribution in the two corpora: in the ILSP we note an ambiguity between passive and
reflexive readings (the former slightly more frequent than the latter), while in the Web
the most frequent reading by far is the reflexive, followed by the passive one. Anti-
causative and reciprocal reading follow. Let us also recall here that the mean of the
total frequencies (of both corpora) showed that reflexive is the most frequent reading,
illustrating mostly the tendencies presented in the Web, due to the significant
difference between passive and reflexive readings in this corpus and due to the
different size of the two corpora. Finally, in both corpora the most frequent reading
with regard to NACT with inanimate subjects (the nact-inanim combination) was the
passive, followed by the anti-causative, although the difference between the two was
larger in the Web than in the ILSP. Reflexive readings are very rare and so are
reciprocal ones, both being more evident in the Web than the ILSP corpus.
Let us now consider whether there are differences in the frequencies of the
various interpretations the verbs of different classes receive with respect to Voice
55.26%), lijizi (bend) (antic: 66.92%). When we compare the distribution of the
readings attested in each verb we note that there are no large differences in the way
they are used. Jerni (lean) is the only verb that differs with respect to its use in the
two corpora since it is mostly unergative in the ILSP while it is transitive in the Web.
The remaining verbs are used almost in a similar way in the two corpora: klini (close),
vrazi (boil) and stegnoni (dry) are transitives in both corpora. The results of the
CORPUS ANALYSIS CHAPTER 3
178
following three verbs show some differences in the attested readings between the two
corpora; hence, ljoni (melt) is ambiguous between transitive and anti-causative
readings, but transitive uses are slighlty preferred over the anti-causative ones in the
ILSP, while the reverse pattern is attested in the Web; sapizi (rot) is anti-causative in
both corpora, but transitive uses are more frequent in the Web than in the ILSP;
finally, lijizi (bend) is anti-causative in both corpora, but in the ILSP no significant
difference is attested between anti-causative and transitive uses, while in the Web the
difference between the two readings is significant.
ACT forms with inanimate subjects of all the verbs favour the anti-causative
reading when the syntactic subject is inanimate, except for the verb ljoni (melt) which
is more frequently used as transitive. In fact, the distribution of five verbs, namely
klini (close), vrazi (boil), jerni (lean), sapizi (rot) and stegnoni (dry), does not differ in
the two corpora, while lijizi (bend) is also anti-causative in both corpora, the only
difference being that anti-causatives are more frequent in the Web than the ILSP
corpus. Ljoni (melt), on the other hand, is anti-causative in the ILSP, while transitive
in the Web. Thus, due to the large size of the Web, the mean total of frequencies from
both corpora, illustrates the tendency presented in the Web.
On the other hand, while these verbs are not supposed to allow for voice
alternation, we found instances of verbs with non-active voice morphology, which
were (quite) few and presented highly individual differences. Moreover, only klini
(close) was found in both corpora, hence, results from the remaining verbs come only
from the Web. More specifically NACT forms with animate subjects from the verb
klini (close) received reflexive (69.59%) and passive readings (28.44%). Lijizi (lean)
was exclusively used as reflexive, while passive readings were significanlty preferred
over reflexives in the other verbs, except for the rarely used in the nact-anim
condition verb jerni (lean), which favoured an anti-causative reading instead (60%).
NACT forms with inanimate subjects were also attested only in the Web, except
for klini (close)24 which was frequent in both corpora. The combinations of NACT
verbs with inanimate subjects were highly used with a passive reading except for
sapizi (rot), which was perceived as anti-causative instead. A final note with respect 24 Also, the following NACT occurrences were attested in the ILSP: 7 NACT with inanimate subjects in vrazi (boil) perceived as passives and 2 NACT with inanimate subjects in stegnoni (dry) with anti-causative readings.
CORPUS ANALYSIS CHAPTER 3
179
to the two corpora: while these verbs were not at all (or rarely) used in the ILSP with
NACT morphology, they were quite frequent in the Web: some examples in the nact-
inanim combination with passive readings are stegnoni-NACT (dry) (ILSP:0-
Web:137), ljoni-NACT (melt) (ILSP:0-Web:50) and vrazi-NACT (boil) (ILSP:7-
Web:293), while klini-NACT (close) presents a similar pattern too, although it was
also used in the ILSP (ILSP: 60 - Web: 587).
Voice Alternating Anti-causatives verbs (Class II), when active with animate
subjects, they were in their majority transitive in the ILSP and Web corpus, no
significant differences between them; very few anti-causative and unergative
interpretations were also attested. ACT forms with inanimate subjects were
ambiguous between transitive and anti-causative readings, the former being preferred
over the latter in both corpora: the only difference between the corpora lies in that
anti-causative readings were more frequent in the ILSP than the Web but not
significantly so.
When non-active verb forms are examined, the distribution of interpretations
differs in the two corpora: NACT forms with animate subjects in the ILSP are
ambiguous between passive and anti-causative readings, reflexive ones being the least
preferred ones. This same combination (nact-anim) was instead perceived ambiguous
between anti-causative and reflexive readings, passive ones following in the Web. A
very small rate of reciprocal interpretations was also attested in both corpora. The
mean of the total frequencies from both corpora showed ambiguity between, anti-
causative and reflexive readings, passive being the least preferred, probable due to the
larger size of the Web. NACT with inanimate subjects were perceived as passives
more frequently than anti-causatives in the ILSP while they were ambiguous between
anti-causative and passive readings in the Web, which is also what the mean of the
total results from the two corpora shows.
Turning to a discussion of the potential different distribution of individual verbs
in the two corpora, note that all the verbs behaved similarly in the ILSP and Web
corpora with respect to ACT forms with animate subjects, they are mostly used as
transitives. ACT forms with inanimate subjects differed with respect to eachother and
with respect to the distribution of some in the two corpora. The verbs berdevi
(mingle), leroni (spill) and tripai (pierce) were similarly used in the two corpora as
CORPUS ANALYSIS CHAPTER 3
180
transitives (only in the latter anti-causative readings are more frequent in the Web
than the ILSP) katharizi (clean) and htipai (hit) are ambiguous between anti-causative
and transitive readings in both corpora. The remaining verbs differed in the two
corpora: tendoni (stretch) was transitive in the ILSP, but ambiguous between anti-
causative (preferred) and transitive uses in the Web; tsalakoni (crumple) was
transitive in the ILSP, but ambiguous between transitive (preferred) and anti-causative
readings in the Web. In all, among ACT forms with inanimate subjects, transitive uses
remain the most frequent, except for htipai (hit), in which the anti-causative reading is
preferred over the transitive but not significantly so. Note finally that, according to the
mean of the total results from the two corpora, katharizi (clean) and tsalakoni
(crumple) are ambiguous between transitive (preferred) and anti-causative readings.
Data from NACT forms of these verbs with animate subjects also revealed
individual verb variation with respect to the corpora examined: katharizi (clean) and
was passive in the ILSP, while anti-causative in the Web; htipai (hit) was preferably
perceived as passive in the ILSP and ambiguous between reflexive (preferred) and
passive in the Web, tripai (pierce) was also passive in the ILSP but reflexive in the
Web. A similar pattern in the two corpora was attested in tendoni (stretch) (reflexive),
in berdevi (mingle) & leroni (spill) (anti-causatives) and in tsalakoni (crumple)
(passive) (but the rate of passives was significantly more elevated in the Web than the
ILSP). In all, tendoni (stretch) and tripai (pierce) were mostly used as reflexives;
katharizi (clean), tsalakoni (crumple), berdevi (mingle) and leroni (spill) were mostly
used as anti-causatives, while htipai (hit) yielded equivalent percentages among the
various readings (refl: 37.68%; antic: 20.12%; pass: 35.03%). NACT forms with
inanimate subjects behaved similarly in the two corpora very often: tendoni (stretch),
berdevi (mingle) and leroni (spill) were anti-causatives, while katharizi (clean) and
tripai (pierce) were passives. Tsalakoni (crumple) was passive in the ILSP, while anti-
causative in the Web and htipai (hit) was also passive in the ILSP, but ambiguous
between passive and anti-causative readings in the Web. In all, Class II NACT verbs
with inanimate subjects are distinguished between those which favoured an anti-
*Note that few occurrences of NACT forms among the Voice (Non)-Alternating Anti-causatives were also attested. Very few examples were attested only in the Web corpus: these were apo-cause (anim: 0; inanim: 3), me-cause (anim: 0; inanim: 2), me-instr (anim: 1; inanim: 0).
As shown in the Table the anti-causative readings that non-active verb forms receive,
were most frequently followed by apo-cause in the ILSP corpus except for Class III
verbs with animate subjects, mostly followed by other PPs (84.61%): Class II: (anim)
50%; and (inanim) 70%; Class III (inanim) 100%. In the Web corpus, NACT anti-
causatives of Class II verbs with animate subjects co-occurred with me-instrument
(58.68%), of Class III verbs with other PPs (47.63%), while NACT anti-causative
readings of Class II verbs with inanimate subjects co-occurred with apo-cause
(32.33%) and other PPs (36.14%) (no significant differences between them) and with
inanimate subjects with other PPs (48.5%). The apo-agent found among Class II verbs
are considered as an animate cause as in the example below:
Ο συνάδελφος απέναντι είπε ότι πιθανόν να µπερδεύτηκαν από τους άλλους
που ήταν πολιτικώς ενάγοντες.
http://www.eksegersi.gr/efeteio/praktika/22_2.htm
Table 56: The distribution of PPs among verbs with passive readings
*Note that few occurrences of NACT forms among the Voice (Non)-Alternating Anti-causatives were also attested. They were followed by various PPs: in the ILSP corpus we found apo agent (anim: 2; inanim: 5), apo-cause (anim: 1; inanim: 1). In the Web corpus we found apo-agent (anim: 15; inanim: 26), apo-cause (anim: 0; inanim: 3), others pp-cause (anim: 0; inanim: 18), apo-instr (anim: 1; inanim: 3), me-instr (anim: 3; inanim: 51)
As shown in the Table the passive readings that receive the non-active forms of the
verbs examined are most frequently followed by apo-agent, apo-instrument and me-
instrument phrases. More specifically Class II verbs with animate subjects are
followed in the ILSP by apo-agent (41.30%) and apo-instrument (36.95%), with no
significant differences between the frequencies, while in the Web they are mostly
followed by apo-agent (57.76%). In sentences with inanimate subjects the most
frequently attested PPs were apo- and me- instrument (30.13% and 34.24%
respectively) in the ILSP, apo-agent (27.67%) and me-instrument (37.42%) in the
Web. Class III verbs with animate subjects were followed by other apo-phrases
(70.66%) in the ILSP while by apo-agent (48%) in the Web; in sentences with
inanimate subjects they were followed by apo-agent (25.24%), apo-instrument
(30.1%) and apo-other (32.04%) in the ILSP, while mostly by me-instrument
(47.81%) in the Web. Note also that Class I verbs were also found in NACT forms co-
occurring with passive readings, mostly with apo-agent and apo-cause PPs.
Let us finally see how the PP-agent, cause and instrument were distributed
with respect to verb reading and subject animacy in each verb class.
CORPUS ANALYSIS CHAPTER 3
192
b. The apo-agent
Graph 19: Total distribution of the apo-agent phrases used among (ACT-NACT)
anti-causative and passives (Verb-Classes, ILSP-Web Corpus)
The apo-agent co-occurred mostly with passive readings. Note however, that it was
not very frequent with respect to the total number of passive readings attested in both
corpora, but significantly more frequent in the ILSP than the Web Corpus (χ2=18.253,
p<.001), since it was found in almost half of the passive occurrences in the ILSP
(1097/1966; 55.57%), while in less than one fourth in the Web (1268/6779; 18.7%).
This is consistent with previous assumptions (Laskaratou & Philippaki-Warburton
1984, Joseph & Philippaki-Warburton 1987, a.o.), which suggest that the use of an
overt apo-phrase is considered marked in Greek. Also, in the ILSP apo-agent phrases
were more frequently attested among Class III verbs with animate (Class III: 300 vs
other Classes: 49 (χ2=180.519, p<.001)) and with inanimate subjects (Class III: 206 vs
other Classes: 79 (χ2=56.593, p<.001). In the Web corpus apo-agent phrases were
more frequent among Class III verbs with inanimate subjects (Class III: 594 vs Class
II: 318: (χ2=83.526, p<.001) and Class III: 594 vs Class I: 101 (χ2=349.711, p<.001),
while when we compared sentences with animate subjects, apo-agent were more
frequent among Class II verbs (Class II: 161 vs Class III and I: 75 +19 (χ2=17.604,
p<.001). Note however that some apo-agent phrases were attested also among active
anti-causatives (Class I) in both corpora, irrespective of subject animacy and very few
CORPUS ANALYSIS CHAPTER 3
193
occurrences are attested among Class II verbs with animate subjects receiving anti-
causative readings in the Web.
c. The apo- and other- cause phrases
Graph 20: Total distribution of PP-cause used among (ACT-NACT) anti-
causative and passives (Verb-Classes, ILSP-Web Corpus)
The apo-cause is usually found among (ACT-NACT) anti-causative readings, while
total frequences differ with respect to the corpora, due to the small size of the ILSP
corpus. Apo-cause phrases co-occur more frequently with anti-causative readings
among Class I verbs with inanimate significantly less than with animate subjects both
in the ILSP (anim:19 vs inanim:38) (χ2=6.333, p=.012) and the Web (anim: 67 vs
inanim: 141) (χ2=26.327, p<.001). They are also highly frequent among NACT Class
II verbs with anti-causative readings; note however that in the ILSP apo-cause phrases
are more frequent with inanimate subjects (anim: 21 vs inanim: 35), while in the Web
with animate (anim: 121 vs inanim: 114), but no significant differences are attested.
CORPUS ANALYSIS CHAPTER 3
194
Note also few examples of apo-cause phrases among anti-causative readings of Class
III verbs with inanimate subjects and among passive readings of all verb classes.
The remaining PP-causes drawn from the corpora were more frequent in the
Web than the ILSP corpus in Class I (8 vs 278: χ2=254.895, p<.001) and Class II
verbs (11 vs 67: χ2=40.205, p<.001), while few examples of Class III were attested
only in the Web. More specifically, in the ILSP corpus they were marginally used
among Class I verbs in ACT receiving anti-causative readings and among Class II
verbs in NACT also receiving anti-causative readings, while no significant differences
were attested with respect to subject animacy. In the Web corpus other PP-causes
among active anti-causative readings of Class I verbs were significantly more
frequent with inanimate than animate subjects (anim: 88 vs inanim: 190) (χ2=37.424,
p<.001). Among NACT anti-causative readings of Class II verbs other PP-causes
were more frequent with animate than inanimate subjects (anim: 48 vs inanim: 19)
(χ2=12.552, p<.001), but in a much smaller scale than Class I verbs.
Turning to a comparison between apo- and other PP- causes the former are
more frequent than the latter in all cases examined except for the anti-causative
readings of Class I verbs in ACT in the Web corpus, where they are equally used with
animate subjects (67 vs 88: χ2=2.845, p=.092), while other PP-causes are more
frequent than apo-cause with inanimate subjects (141 vs 190: χ2=7.254, p=.007).
CORPUS ANALYSIS CHAPTER 3
195
d. The apo- and me- instrument phrases
Graph 21: Total distribution of apo- and me- instrument phrases used among
(ACT-NACT) anti-causative and passives (Verb-Classes, ILSP-Web Corpus)
Apo-instrument phrases were mostly frequent among Class II verbs in both corpora
(ILSP: 47/60 and Web: 171/211), while few examples were also attested to co-occur
with anti-causative readings among Class I verbs with inanimate subjects in the ILSP
(10/60) and the Web corpus (5/211) and some other examples were attested to co-
occur with passive readings among Class III verbs with inanimate subjects
specifically in the Web corpus (27/211). With respect to Class II verbs note that apo-
instrument phrases co-occurred with passive readings in the two corpora both with
animate (ILSP: 17 and Web: 17) and inanimate subjects (ILSP: 22 and Web: 34).
However, they co-occurred on a larger scale with anti-causative readings specifically
in the Web corpus. More specifically, apo-instrument was significantly more frequent
among NACT Class II verbs with animate than inanimate subjects (78 vs 36:
χ2=15.474, p<.001), finding which can be attributed to semantic reasons. Their
frequency rates differed significantly from passives when animate subjects were
involved in the sentence (78 vs 17: χ2=39.168, p<.001) while they did not differ when
inanimate subjects were involved (36 vs 34: χ2=.057, p=.811). Few occurrences of
CORPUS ANALYSIS CHAPTER 3
196
apo-instruments were also attested among Class II verbs in ACT with inanimate
subjects receiving anti-causative readings.
Turning to me-instrument phrases they were mostly frequent among Class III
verbs (ILSP: 131 and Web: 351), while some were also attested among Class II verbs
(ILSP: 39 and Web: 252) and few among Class I verbs (ILSP: 6 and Web: 178).
Starting with a discussion concerning the environments where they were mostly
frequent, i.e Class III verbs, passive readings favoured the presence of PP-instrument
both with animate and inanimate subjects in the ILSP corpus (69 vs 62), but more
frequently with inanimate than animate subjects in the Web corpus (284 vs 26:
χ2=214.723, p<.001). With respect to Class II verbs, PP-instrument were attested with
passive (anim: 10 and inanim: 25) and anti-causative readings with NACT verb forms
(anim: 4) in the ILSP corpus and with passive (anim: 35 and inanim: 119) and anti-
causative readings both with ACT (anim: 2 and inanim: 36) and NACT forms (anim:
35 and inanim: 35) in the Web corpus.
Overall me-instrument phrases were used instead of apo-instrument more
frequently among passive readings of Class III verbs (32 vs 490: χ2=401.847, p<.001).
On the other hand, me- and apo-instrument phrases are equally used with Class II
verbs (218 vs 291: χ2=1.960, p=.162). Finally, me-instrument phrases are
significantly more frequent than apo-instrument phrases among Class I verbs (21 vs
184: χ2=129.605, p<.001).
Summary
The use of apo-phrase is rather infrequent: apo-agent phrases are attested only in
4.10% (98/1966) of passive interpretations in the ILSP corpus and 5.85% (397/6779)
in the Web corpus. Apo-cause phrases are attested in 5.66% (57/1006) of anti-
causative interpretations in the ILSP corpus and 6.8% (249/3661) in the Web corpus
with NACT verbs and 2.24% (62/2765) in the ILSP and 2.62% (225/8569) in the Web
corpus with ACT verbs. On the other hand, apo-phrases are used for other purposes
(to express location, time, directionality a.o.) which represent the 66.86% of the total
apo-phrases attested in the two corpora.
CORPUS ANALYSIS CHAPTER 3
197
On the other hand, different structures under ‘other PP’ are the most frequent
phrases that serve to disambiguate the verb reading. In other words, context is
frequently supportive to the verb interpretation in adult Greek language: hence, a
large percentage of other PPs used with prepositions such as me (with) or se (in) tend
to be often used: among ACT verb forms 4.88% (135/2765) in the ILSP and 9.11%
(781/8569) in the Web and among NACT verb forms 3.97% (40/1006) in the ILSP
and 17.48% (640/3661) in the Web denoting an anti-causative reading and among
NACT verb forms 10% (197/1966) in the ILSP and 10.62% (720/6779) in the Web
denoting a passive reading.
5. Discussion
5.1. The size of the samples
Overall, we should note that the results analysed come from the Web Corpus, since it
represents the 70.92% of the total data. In fact, the small size of the ‘ILSP corpus’ and
the lack of a sample with colloquial speech were the reasons why we enlarged the
research database and created a new one, the ‘Web corpus’ with sentences drawn
from the Internet (Google search machine).
Frequencies drawn from the two corpora present a similar pattern in the use of
active and non-active verbs irrespective of verb class, i.e. more active (68.70%) than
non-active verb forms (31.30%) are attested with respect to the total of Class I, II and
III verbs. However, when we consider the frequencies obtained in each verb class
examined, we distinguish differences between the two corpora. More specifically,
while in both corpora ACT forms of Class I verbs (Voice non-alternating anti-
causatives) were more frequent than NACT ones, the NACT forms were significantly
more frequent in the Web than the ILSP corpus. Also, ACT forms of Class I verbs
were rare in the ILSP corpus, except for klini (close) which was more frequent in the
ILSP than the Web and stegnoni (dry) for which the frequencies obtained in the two
corpora did not differ significantly. Moreover, non-active verb forms of Class I
remain rather infrequent in the Web as in the ILSP Corpus, given the fact that they
involve neologisms. With respect to Class II (Voice alternating anti-causatives) we
note that ACT forms are more frequent than NACT ones in both corpora, except for
CORPUS ANALYSIS CHAPTER 3
198
berdevi (mingle) which is similarly used in both corpora. However, the verbs tendoni
(stretch) and leroni (spill) present low frequency in the ILSP corpus, while tsalakoni
(crumple) is infrequent in both corpora. Class III verbs (activity predicates) present a
different distribution in the two corpora. In the ILSP the verbs htenizi (comb), pleni
(wash) and dini (dress) are infrequent in ACT and not used at all in NACT. Only the
NACT occurrences of metaferi (transfer) and krivi (hide) are frequent in the ILSP. On
the other hand, no such problems were evidenced in the Web. It is worth noting also,
that, among ‘reflexives’, dini (dress) was more frequent in NACT than ACT in both
corpora, while pleni (wash) was more frequent in ACT than NACT in the ILSP, but
more frequent in NACT than ACT in the Web; htenizi (comb) was found only in
NACT in the ILSP, while it was more frequent in ACT than NACT in the Web, a
pattern attested in the remaining verbs of this class in both corpora.
The fact that the examined verbs did not present the same frequency of use in
the two corpora is probably due to the difference of genre of the texts found in the
corpora: the ILSP corpus comprises texts drawn from literature sources, from daily
newspapers, from other scientific and rather formal texts. The Web Corpus comprises
mostly texts drawn from various informal sources (blogs), from chat pages and other
messages produced by non-professional writers, cited in various written texts, which
are likely to illustrate a quasi-oral informal speech.
5.2. The interpretations
With regard to the interpretations they received, the active verb forms were found to
co-occur with transitive (72.77%) and anti-causative readings (25.42%), while the
non-active verb forms were mostly used with passive readings (44.15%), anti-
causative (23.56%) and reflexive (27.85%) ones being also available. The two corpora
did not differ with respect to the distribution of the various readings, but for the
reflexive interpretation which was more frequent in the Web than the ILSP. The fact
that ACT forms were mostly used as transitives while the NACT forms as passives,
shows that Agent-Theme structures are ‘prototypical’, hence more frequent.
When we consider the combination of the [+/-animacy] of the syntactic
subject with the [+/-active] verb morphology, the frequencies of the readings attested
are distributed as follows: the active forms in sentences with animate subjects are
used transitively, while in sentences with inanimate subjects they co-occur both with
CORPUS ANALYSIS CHAPTER 3
199
anti-causative (preferred) and transitive readings. However, in the act-inanim
condition the use of anti-causative readings is higher in the Web Corpus, since the
difference between anti-causative and transitive uses is not significant in the ILSP
while it is significant in the Web. The non-active verb forms with animate subjects
receive more frequently reflexive than passive/anti-causative readings in total, an
effect of the Web corpus, larger in size, given that in the ILSP, this combination is
mostly passive (again, probably an effect of speech genre). In NACT verbs with
inanimate subjects passive readings are preferred over anti-causative ones, in both
corpora.
In other words, when an animate subject is involved, active verbs are mostly
used transitively. Change-of-state verbs with internal cause, such as sapizi (rot), are
also attested with a transitive/causative reading in the presence of an animate subject
(cf. also Theophanopoulou-Kontou, 2001; Roussou & Tsimpli, 2007), which however
is commonly a non-literal one, contrary to the fact that they do not appear to have a
transitive version (Levin & Rappaport, 1995). Anti-causative readings seem to be
blocked, as also argued in the literature (Anagnostopoulou & Alexiadou, 2004,
Tsimpli, 2005, 2006). However, some instances of anti-causative readings found in
sentences involving an animate subject are to be further discussed. In fact, they
mostly come from examples with non-literal readings of Voice Non-alternating Anti-
causatives (Class I), drawn mostly from the Web corpus. More specifically, the anti-
causative readings attested among Class I verbs are distributed as follows: klini
2009, for English; Peristeri, Tsimpli & Tsapkini, submitted, for Greek)25. However,
the results are not interpreted in the same way. We thus proceed with a brief overview
of their findings.
Data from Burkhardt et al.’s study (2003), though limited to unergative and
unaccusative verbs participating in the causative/inchoative alternation26, supported
the Split Intransitivity Hypothesis (Perlmutter, 1978; Levin & Rappaport-Hovav,
1995 a.o.). More specifically, data from control subjects (9 adults) showed that a
statistically significant priming effect appears right after the verb for unergatives,
while later on for unaccusatives. Authors attributed the effect right after the 25 In these researches Cross Modal Lexical Priming (CMLP) tasks were administered to agrammatic (Burkhardt, Piñango and Wung, 2003 for English and Peristeri, Tsimpli & Tsapkini, submitted, for Greek) and normally developing English speaking adults (Friedmann, Taranto, Shapiro & Swinney, 2008). In these CMLP taks sentences were presented orally at a normal speaking rate and, at some point during each sentence a letter sequence (word or non-word) was briefly visually displayed on a screen. The participants were asked to make a lexical decision on the letter sequence. With respect to traces created under movement, a priming effect on the trace position means that the moved element has been reactivated in this position, facilitating Response Time. 26 Authors claim to have limited data since no clear pattern of the non-alternating unaccusatives subtype was evidenced in the performance of Broca’s aphasics (2003:17).
ON-LINE PROCESSING CHAPTER 4 OF TRANSITIVITY ALTERNATIOSN
208
unergative verb to the existence of a trace of the subject base-generated in Spec-VP
(according to the VP-Internal Hypothesis (Koopman & Sportiche, 1991 a.o.)), the
trace of the moved NP in unaccusatives being located lower in the structure.
In Friedmann et al. (2008), data from 120 English speaking adults showed
that subjects of (non-alternating)27 unaccusatives reactivate after the verb, while
subjects of unergatives do not, also supporting the Split Intransitivity Hypothesis
(Perlmutter, 1978; Levin & Rappaport-Hovav, 1995 a.o.). However, a mixed pattern
attested in the reactivation of alternating unaccusatives (see fn.27) was attributed to
the different distribution of the most frequent uses of the verbs examined. A search in
the Brown Corpus revealed that the ones which behaved like unaccusatives were
frequently found with an affected subject (in Dowty’s terms, 1991), while the ones
which behaved like unergatives were often found with a subject ‘volitionally engaged
in the activity denoted by the verb’ (Friedmann et al., 2008:18), supporting the claim
that alternating unaccusatives have their single argument base-generated in subject
position (Haegeman 1994, Belletti 1988). However, evidence from neuroimaging
(Shetreet et al., 2009) suggests that activation of the middle temporal gyrus reveals a
lexical operation that derives unaccusative verbs, while activation of the inferior
frontal gyrus may be involved with the execution of the syntactic operation in the
case of unergatives.
Peristeri et al. (submitted) studied Greek agrammatic patients and 15
unimpaired controls. Voice morphology and [+/-animacy] of the syntactic subject
were manipulated, factors which have been proved to determine Greek speakers’
choices (cf. Tsimpli, 2006) and are also used in our research. There was a priming
effect right after the verb in the unergative condition (attested only in one patient and
three controls), which was accounted for in terms of the productivity of the
transitivization of unergative verbs in Greek (see also Roussou & Tsimpli, 2007) and
not in terms of the VP-internal subject hypothesis (as suggested in Burkhardt et al.,
2003). Furthermore, what is interesting for Greek, given the availability of ACT
unergatives and anti-causatives/unaccusatives as well as anti-causatives/
unaccusatives with both ACT/NACT morphology, is that data from patients revealed
27 In the sense of Haegeman (1994), i.e. not having a transitive/causative counterpart (eg. vanish) vs alternating, i.e. inchoative that have a transitive/causative counterpart (eg break); see also 1.1
ON-LINE PROCESSING CHAPTER 4 OF TRANSITIVITY ALTERNATIOSN
209
‘limited access to the lexicon-filtered information constraining ACT verbs as
unaccusatives rather than unergatives’ (Peristeri et al., submitted: 37). Finally, an
animacy effect interacting with voice morphology was registered in both controls and
aphasics. More specifically, with respect to NACT unaccusatives, priming was
exhibited only with inanimate subjects, suggesting that NACT unaccusatives with
animate subjects were perceived as reflexives, the level of interpretational ambiguity
being raised. In ACT, controls were not influenced by the subject animacy, a finding
which shows that they were lexically driven, in contrast to aphasics whose
performance was affected by animacy. Although aphasics are suggested not to be
able to establish the A-chain between the antecedent and the post-verbal gap for ACT
verb types of both voice ‘alternating’ and ‘non-alternating’ verbs, the availability of
NACT functioning as a marker of transitivity is suggested to facilitate processing of
the former case. This sensitivity to optional voice alternation in aphasics shows
access to a distinct morpho-phonological component (something that reminds us of
the independent availability of a morpho-phonological component in Borer’s system
(2004) raising implications for child developing grammars in language acquisition).
On-line research studies have also been conducted on other structures
involving NP movement, such as the passive in English, aiming to examine whether
there is a processing load in non-canonical constructions. Ferreira (2003) using biased
reversible (The dog bit the man), non-reversible (The mouse ate the cheese) and
reversible-symmetrical (The woman visited the man) sentences, and also
manipulating plausibility (The dog bit the man vs. The man bit the dog), asked
participants (English-speaking adults) to identify the thematic roles in the sentence
orally presented (e.g., Who was the do-er?). She suggested that participants
misinterpreted passives, especially when they expressed implausible ideas. In
addition, RTs to actives were significantly shorter - 1899 ms - than to passives - 2156
milliseconds, reflecting that passives pose greater processing load than actives.
Nevertheless, surface frequency of a syntactic form did not seem to determine ease of
processing, as active sentences and subject-clefts were comprehended equally easily
despite the rareness of the latter type. Contrary to that, in Rohde & Gibson (2003), the
processing of actives was not easier than the processing of passives when the word-
by-word self-paced reading task adult participants encountered, contained active
ON-LINE PROCESSING CHAPTER 4 OF TRANSITIVITY ALTERNATIOSN
210
sentences involving subject-extraction (passive sentence: The reporter that was
attacked by the senator ignored the president vs active sentence: The reporter that
attacked the senator ignored the president). Authors noticed that reading rates were
in line with frequency data taken from counts in the Penn Treebank (Brown and WSJ)
and suggested that the online sentence reading control mechanism is not necessarily
sensitive to processing difficulty at the thematic level. Finally, in Stromswold et al.
(2002), a study on English L1 acquisition, investigating the processing of actives and
passives in real-time using eye-tracking in a sentence-picture matching task, adult
controls were attested to make use of acoustic, morphological, and semantic cues and
decide on-line, i.e. “at or before the verb stem (i.e., before participle)” for the active
sentences and “at the past participle” for the passive sentences (Stromswold et al.,
2002).
In Greek, a morphologically rich language, NACT voice morphology marks
transitivity alternations. In previous research on subject-object ambiguity resolution
(Papadopoulou & Tsimpli, 2005), it has been suggested that Greek speakers are
sensitive to morphological cues during processing. Thus, it is interesting to examine
the degree of sensitivity to morphological marking. Moreover, results suggesting an
interaction of voice morphology and subject animacy (Peristeri et al., submitted)
receive support from the present study, as will be shown next. Finally, we hope that
the comparison of the data obtained in our SPR task to frequencies drawn from adult
written corpora will shed some light on the question regarding the possibility of a
statistically driven parser, as implied in previous research (eg. Friedmann et al., 2008;
Rohde & Gibson, 2003) and the degree of the interaction of Voice morphology and
subject animacy with these ‘statistical records’ with respect to voice ‘alternating’ and
‘non-alternating’ anti-causatives in Greek (Chapter 6).
ON-LINE PROCESSING CHAPTER 4 OF TRANSITIVITY ALTERNATIOSN
211
2.2. Specific rationale
Based on the research aims stated above, the effect of the morpho-phonological
properties of Greek active and non-active voice, of the animacy of the syntactic
subject and their potential interaction are to be investigated. More specifically, the
SPR task was designed to determine Greek native speakers’ perception of the role of
these features in sentences where Voice (Non) - alternating anti-causative verbs are
always in the anti-causative structure, i.e. no object is present.
All the verbs examined have been argued to belong to the lexically
determined verb class of ‘anti-causatives/ergatives’ (cf. Anagnostopoulou &
Alexiadou, 2004; Levin & Rappaport Hovav, 1995). Alternatively, it has been
suggested that specific readings assigned to these verbs are the result of pragmatic or
encyclopaedic information attached to verb/conceptual entries in the adult grammar
as well as the interaction of this information with Voice marking and animacy of the
subject (Tsimpli, 2006). The implication is that while NS’ acceptability judgments
will exhibit preferences according to verb classes, reaction times in sentence
processing, as well as on the decision making segment will vary when voice
morphology, animacy of the syntactic subject and choice of verb are controlled for.
ACT/NACT voice morphology is of primary importance in verb processing
since active morphology does not show unaccusativity in a transparent way and is
typically associated with unergative and transitive structures as well (Tsimpli, 2006).
Non-active morphology too can give rise to reflexive or non-reflexive (passive, anti-
causative) readings. Animacy is relevant as a semantic feature which ‘affects’
syntactic derivations, in that it affects the preference for one of the grammatically
available syntactic derivations: In combination to active verbs it strongly favours a
causative structure since its presence blocks the ‘alternating’ status of voice
alternating anti-causatives (Alexiadou & Anagnostopoulou, 2004), although there are
metaphorical uses, some quite common. Inanimate subjects with active verbs co-
occur both with causative and anti-causative readings. In combination to non-active
verbs, animacy allows for all available readings, i.e. reflexive and non-reflexive, the
former involving an external argument while the latter a derived subject (Tsimpli,
2006). Inanimate subjects with non-active verbs co-occur with passive and anti-
ON-LINE PROCESSING CHAPTER 4 OF TRANSITIVITY ALTERNATIOSN
212
causative readings distinguished at an interface level where the cause or agent
interpretation of the external argument becomes relevant, but there are also
metaphorical reflexive uses, some quite common (see Chapter 3). Verb Classes are
also relevant at an interface level, since they are assumed to determine NS’
preferences with respect to Voice (Non)-Alternating Anti-causatives.
Predictions for the acceptability judgment (AJ) task
With regard to acceptability judgments, given that all the sentences presented are
unaccusative structures (no object is present), participants are expected to consider
acceptable all active forms with an inanimate subject (both Class I and II verbs), as
well as non-active forms with either animate or inanimate subjects for the Voice
Alternating Anti-causatives (Class II). On the other hand, participants are expected to
consider marginal or unacceptable all the non-active forms of Class I verbs and the
active forms with animate subject of Class II verbs, since animacy has been argued to
‘block’ the anti-causative reading (Alexiadou & Anagnostopoulou, 2004).
Active morphology does not show unaccusativity in a transparent way and is
typically associated with unergative and transitive readings/structures (Tsimpli,
2006), thus, the time needed for the decision making (RTs) should be higher with
ACT than NACT forms, where there is a clear marking of transitivity alternations.
Moreover, the RTs for ACT forms with animate subjects should be higher than the
ones in ACT forms with inanimate subjects: the subject animacy is expected to affect
the processing of unaccusative structures, given that inanimate subjects are associated
with a theme-role (Dowty 1991; 2003), as is the case in these structures. RTs for
NACT forms with inanimate subjects should be faster than NACT with animate
subjects: participants are expected to attempt to resolve the ambiguity between a
reflexive and a non-reflexive reading (a derived and a non-derived structure), in
sentences involving an animate subject, or between a passive and an anti-causative
reading (two derived structures) in the presence of inanimate subjects. Finally, Voice
Non-alternating Anti-causatives (Class I verbs) in NACT morphology are expected to
yield shorter RTs than Class I ACT verbs and Voice Alternating Anti-causatives
(Class II verbs) irrespective of voice morphology (ACT/NACT) because participants
consider the given sentences marginal or unacceptable.
ON-LINE PROCESSING CHAPTER 4 OF TRANSITIVITY ALTERNATIOSN
213
Predictions for the self-paced reading (SPR) task
A common assumption in psycholinguistics is that language processing is sensitive to
morphological marking among other properties. As far as voice morphology is
concerned, we expect that RTs in the processing of the verb segment, i.e. the critical
segment, in the present experiment would provide longer reaction latency in the non-
active voice morphology condition, since NACT is morphologically more marked
than ACT, and NACT always signals transitivity alternations, with the exception of
deponents.
Furthermore, given the variety of interpretations of the voice morpheme
(ACT/NACT), the test was designed to investigate the other syntactic and pragmatic
factors found on later segments, since no disambiguation of the verbal interpretation
was included by the end of the sentences. Animacy is predicted to affect the reaction
times after the verb (critical) segment. This is based on the grounds that animacy,
being a semantic feature, will be relevant to later stages of the processing, supporting
models which suggest the priority of syntax over semantics in the course of sentence
processing (cf. Serial autonomous models eg. Ferreira and Clifton, 1986; Frazier,
1987a, b, 1990; Frazier and Rayner, 1982). If the animacy effect were attested earlier,
it would provide evidence in favour of multiple constraint satisfaction models
(MacDonald, 1997; MacDonald et al., 1994; Spivey-Knowlton, Trueswell &
ON-LINE PROCESSING CHAPTER 4 OF TRANSITIVITY ALTERNATIOSN
216
3. The SPR and AJ tasks
The SPR task in which sentences are presented in a word-by-word or phrase-by-
phrase fashion are supposed to reflect initial parsing choices since the reader cannot
go back to the words or phrases (s)he has previously read and, thus, (s)he is
encouraged to keep-up-to-date with sentence processing. The basic rationale
underlying this task is that increased reading times (RTs) on a particular segment
(compared to the same segment in a control condition) indicate relatively higher
processing difficulty (Just et al., 1982). Moreover, the word-by-word fashion would
serve the present study because the ease or speed of access to a word during sentence
processing has been shown to be affected by several factors: in several studies,
specific lexical items were controlled in terms of their most frequent occurrence in
multiple environments with respect to their morpho-syntactic apparatus, although no
explicit reference of a morpho-syntactic analysis is provided. For example, the more
frequent use of that as a complementizer than a demonstrative determiner is
suggested to affect parsing even in environments that unambiguously favour the
second reading (cf. Tabor et al., 1997). By analogy, we specifically wanted to see if
frequency alone, its combination with morpho-syntactic and pragmatic factors, or
animacy affected the time needed to process the verb segment.
In the present word-by-word SPR task the participants also performed an AJ
task at the end of each sentence. Namely, they were asked to determine the
acceptability of the sentences they were presented with, on a rating scale from ‘1’
(=totally unacceptable) to ‘9’ (=totally acceptable). The choice of a ‘?’ was also
provided in case the participants were not able to judge the acceptability of the
sentence due to non-experimental factors, i.e. they had not paid attention, or they had
forgotten the sentence. They were encouraged to base their judgments on their own
intuition and not on prescriptive rules. The AJ task was to serve a triple role. First, it
would ensure that participants paid attention to the task, since the reader was
encouraged to process sentences in order to comprehend and evaluate them. Second,
the degree of acceptability of the sentences reflects the adults’ preferred
interpretations, so that they could be then compared with the frequency rates found in
the corpora examined (see Chapter 3 for detail). The evaluation of the sentences
ON-LINE PROCESSING CHAPTER 4 OF TRANSITIVITY ALTERNATIOSN
217
could also provide evidence for the adults’ sensitivity to various types of grammatical
and semantic information and the relative processing difficulty in the structures
examined. Thus, a 9-point rating scale was chosen because of the nature of the
sentences under investigation, which may not be categorically considered
grammatical or not. Also, reaction times for the AJ task were recorded, in order to
investigate possible processing difficulties for specific conditions.
3.1. Design and materials
The materials consisted of 56 experimental sentences, all involving unaccusative
structures. There were 14 activity verbs, presented in active and in non-active
morphology and each form was included in sentences with an animate or an
inanimate syntactic subject, giving rise to four experimental conditions (act-anim,
act-inanim, nact-anim, nact-inanim) as shown in (19) to (22) below (see the Appendix
I for the total list of experimental sentences):
(19) Το / αγόρι / τσαλάκωσε / αφού / ήρθαν / οι / καλεσµένοι.
The / boy / crumple ACT.PAST.3Sg / after / come PERF.PAST.3Pl/ the / guests.
“The boy crumpled after the guests had come.”
(20) Το / ύφασµα / τσαλάκωσε / αφού / ήρθαν / οι / καλεσµένοι.
The / tissue / crumple ACT.PAST.3Sg / after / come PERF.PAST.3Pl/ the / guests.
“The tissue crumpled after the guests had come.”
(21) Το / αγόρι / τσαλακώθηκε / αφού / ήρθαν / οι / καλεσµένοι.
The / boy / crumple NACT.PAST.3Sg / after / come PERF.PAST.3Pl/ the / guests.
“The boy got crumpled after the guests had come.”
(22) Το / ύφασµα / τσαλακώθηκε / αφού / ήρθαν / οι / καλεσµένοι.
The / tissue / crumple NACT.PAST.3Sg / after / come PERF.PAST.3Pl/ the / guests.
“The tissue got crumpled after the guests had come.”
ON-LINE PROCESSING CHAPTER 4 OF TRANSITIVITY ALTERNATIOSN
218
Note however that twenty-eight quartets of experimental sentences were created out
of the fourteen verbs to be tested28. These quartets were equally distributed across
four versions, so that in each version there were seven sentences per each
experimental condition (see (19)-(21) and (22)-(25)). Consequently, each participant
saw all conditions but never saw the same item more than once. Moreover, each
participant saw either the two animate or the two inanimate conditions of a specific
verb.
The fourteen verbs examined in this task are all in the 3rd singular simple past
or simple present tense29. The sentences across the four conditions were minimally
different, in the sense that they included the same words and they differed only in the
syntactic subject (animate vs inanimate) and the verb morphology (active vs non-
active).
The following 14 verbs, classified as voice (non)-alternating anti-causatives in
the literature (see Theophanopoulou-Kontou, 2000; Anagnostopoulou & Alexiadou,
2004; Tsimpli, 2005, 2006), are used in the task.
28 Two quartets were designed for each verb, so that participants did not see the same context more than once. In the alternative quartet, which also comprised seven segments, the syntactic subject and the subordinate clause that followed the verb were different as in (22) to (25) below:
(22) Η / κοπέλα / τσαλάκωσε / όταν / έπαιζε / το / ράδιο. The / girl / crumple ACT.PAST.3Sg / while / play IMP.PAST.3Sg/ the / radio. “The girl crumpled while the radio was on.”
(23) Η / φούστα / τσαλάκωσε / όταν / έπαιζε / το / ράδιο. The / skirt / crumple ACT.PAST.3Sg / while / play IMP.PAST.3Sg / the / radio. “The skirt crumpled while the radio was on.”
(24) Η / κοπέλα / τσαλακώθηκε / όταν / έπαιζε / το / ράδιο. The / girl / crumple NACT.PAST.3Sg / while / play IMP.PAST.3Sg / the / radio. “The girl got crumpled while the radio was on.”
(25) Η / φούστα / τσαλακώθηκε / όταν / έπαιζε / το / ράδιο. The / skirt / crumple NACT.PAST.3Sg / while / play IMP.PAST.3Sg/ the / radio. “The skirt got crumpled while the radio was on.”
Thus, participants who were presented with Version 1 of the task saw example (19) for the active-animate condition but example (22) for the non-active - animate, example (21) for the active-inanimate but example (25) for the non-active- inanimate. This way, they saw a total of 14 experimental sentences with seven different verbs but never saw the same context, in order to avoid any repetition effect on the RTs of the SPR or the AJ task, as well as on the degree of acceptability. 29 Tense was not included in the research variables, so verbs were presented in present or past tense, depending on the frequency of their use in corpora. Present tense was used only with neologisms attested in corpora (see Chapter 3), which of course remain not frequently used (Fotiadou, in press).
ON-LINE PROCESSING CHAPTER 4 OF TRANSITIVITY ALTERNATIOSN
We next present the effects attested for each verb of the two verb classes. All the
acceptability rates of Voice Non-alternating Anti-causatives yielded a significant
Voice effect, attributed to the ungrammaticality of NACT forms. A significant
ON-LINE PROCESSING CHAPTER 4 OF TRANSITIVITY ALTERNATIOSN
231
Animacy effect was also found in all verbs but jerni (lean) which was accepted with
both animate and inanimate subjects in ACT forms. Lack of a significant interaction
between the two main effects (Voice and Animacy) in lijizi (bend) and stegnoni (dry)
is attributed to the fact that grammatical ACT and ungrammatical NACT do not differ
significantly with respect to Animacy possibly because anti-causatives are primarily
preferred with inanimates. Significant interaction between the two main effects
(Voice and Animacy) in ljoni (melt), vrazi (boil) and sapizi (rot) are due to the fact
that only NACT with inanimate subjects are considered grammatical, while in klini
(close) it is due to the preference of ACT and NACT verbs with inanimate over
animate subjects as well as the preference of ACT over NACT forms.
More specifically, starting with the verb klini (close) note that acceptability
rates were affected both by voice (F(1,19)=8.717, p=.008) and animacy
(F(1,19)=16.471, p=.001) as well as by the interaction between voice and animacy
(F(1,19)=13.330, p=.002). T-tests showed that participants preferred inanimate to
animate subjects with active verbs (t(19)= -5.571, p<.001) but no significant
differences arose between animate and inanimate subjects with non-active verbs31.
Namely, sentences like ‘To parathiro eklise an ke emis de thelame’ (The window
closed even if we didn’t want to) is more preferred than ‘O jitonas eklise an ke emis
de thelame’ (The neighbor closed even if we didn’t want to), while no such
preference is attested in the presence of non-active verb morphology. Furthermore,
active was more accepted than non-active verb morphology only with inanimate
subjects (t(21)= 4.516, p<.001)32.
Acceptability rates on the verb lijizi (lean) depended on voice
(F(1,20)=47.817, p<.001) and animacy (F(1,20)=10.807, p=.004) but the interaction
was not significant. Overall, active verb morphology is preferred to non-active one
31 Differences between nact-anim vs nact-inanim (3.76 vs 4.23) are not significant because nact morphology is generally not accepted even if largely used (see Corpora). Furthermore, variation among participants in the judgments resulted to non-significant mean differences. 32 Differences between act-anim vs nact-anim (3.50 vs 3.76) are not significant because animate subjects are not preferred in a sentence with this specific verb, though structures cannot be characterised as ungrammatical ones, since no violations are observed: the presence of an object is optional in cases like (she closed the store); moreover the use of these specific structures is attested in corpora but with a non-litteral reading (i.e. I mitera eklise (act) meta to atichima = mother was finished after the accident) or ‘I mitera klistike (nact) meta to atichima = mother was closed to herself after the accident).
ON-LINE PROCESSING CHAPTER 4 OF TRANSITIVITY ALTERNATIOSN
232
(for animate subjects: act (6.00) > nact (1.73) and for inanimate subjects: act (7.48) >
nact (3.24)). The sentences ‘O fandaros / I alisida lijise horis na prospathiso poli’
(The soldier / The chain bent without my trying hard) are preferred to their non-active
counterparts. Note also that inanimate subjects are slightly more accepted in
sentences with both active and non-active verbs.
Acceptability rates for the verb jerni (bend) yielded a significant effect of
voice (F(1,17)=68.338, p<.001). The verb is more accepted in the active voice
morphology both with animate (act-anim: 6 vs nact-anim: 1.73) and inanimate
subjects (act-inanim: 7.48 vs nact-inanim: 3.24). Namely, the sentences ‘To luludi / I
neari ejire prin pesi I nihta’ (The flower / The young girl leaned before the night
arrives) are the most accepted ones.
Acceptability rates on the verb stegnoni (dry) were affected both by voice
(F(1,19)=118.686, p<.001) and animacy (F(1,19)=15.804, p=.001) but interaction
was not significant. In other words, participants scored high sentences where the verb
was in active morphology (as To engoni /To pandeloni stegnose an ke itan poli
vregmeno’ (The grandson / The trousers dried although it was/they were very wet)
over To engoni /To pandeloni stegnothike an ke itan poli vregmeno’ (The grandson /
The trousers was/were dried although it was/they were very wet)) and preferred an
inanimate over an animate subject: i.e. act-anim (5.25) vs act-inanim (7.86); nact-
anim (1.86) vs nact-inanim (2.73). Notice however that low acceptability of the act-
anim condition may be due to the use of the NP to engoni (the grandson).
Acceptability ratings of the verb ljoni (melt) revealed the two main effects
(Voice: F(1,19)=97.675, p<.001; Animacy: F(1,19)=39.457, p<.001) as well as a
significant effect of the interaction between voice and animacy (F(1,19)=6.370,
p=.021). Active morphology was more accepted than non-active both in sentences
involving an animate (t(19)= 2.557, p=.019) or an inanimate subject (t(21)= 8.472,
p<.001), as in the example of ‘To pedi / To keri eljose horis na to katalavo’ (The boy
/ The candle melted without me noticing). Furthermore, the presence of an inanimate
subject was more accepted than the presence of an animate one regardless of voice
once with an animate and once with an inanimate subject, i.e. a total of ten sentences.
Given that the verbs do not belong to either the reflexive or the anti-causative class,
the animate subject sentences would be ambiguous between the reflexive, the passive
ACQUISITION CHAPTER 5 OF TRANSITIVITY ALTERNATIOSN
275
and the anti-causative reading while the inanimate subject sentences would show no
preference between the passive and the anti-causative reading.
Some examples of the test sentences with their preferred reading, according to
verb class, are presented below:
(34) To klidhi spai Anti-causative
the key breaks-ACT
(35) I bala tripai / tripjete Anti-causative
the ball pierces-ACT/pierces-NACT
(36) O Mickey tripjete Anti-causative
the Mickey pierces-NACT
(37) To agori plenete Reflexive
the boy washes-NACT
(38) To agori pleni Transitive
the boy washes-ACT
(39) To pedhi vafete Reflexive/Passive/Anti-caus.
the child paints-NACT
(40) I porta vafete Passive/Anti-causative
the door paints-NACT
There were thirty sets of pictures used to test the critical sentences, some of which
were used twice. Specifically, the five triplets designed for ‘reflexives’ and
‘alternating anti-causatives’ with animate subjects, respectively, were used twice in
order to test voice morphology. The rest of the pictures were seen only once. In order
to avoid repetition, the experiment was administered in two sessions with at least
three weeks distance between them. Participants were presented with twenty
sentences in the first session and the remaining twenty in the next one. Sentences
were distributed so that participants heard a total of ten sentences with animate
subjects (five with verbs in ACT and the other five in NACT morphology) and ten
with inanimate subjects (five with ACT verbs and five with NACT verbs) in each
session and never saw the same picture or heard the same verb during the same
session.
ACQUISITION CHAPTER 5 OF TRANSITIVITY ALTERNATIOSN
276
Each set of pictures was designed on the basis of the following variables: verb
morphology, subject animacy and verb class. In each set of pictures there was a
‘preferred’ picture whose preference status is accounted for by a combination of all of
the above variables36. For each group of verbs, every triplet of pictures included the
following choices:
(a) Class I (Voice Non-alternating Anti-causatives):
The sentences with ‘Non-alternating’ (ACT) anti-causative’ verbs were presented
with inanimate subjects and there was a preferred picture (anti-causative), a dis-
preferred one (transitive)37 and a non-target one (passive), given ACT morphology.
Thus, in a sentence like to klidhi spai (the key breaks-ACT), the ‘preferred’ picture is
the one showing the key breaking due to an external cause, in this example by falling.
One of the other pictures showed the key breaking something else, in this example,
the key hole, (the ‘transitive’ reading). The third picture showed someone (an agent)
breaking the key (the ‘passive’ reading).
(b) Class II (Voice Alternating Anti-causatives):
(a) For sentences with ACT morphology and inanimate subject, there was a preferred
picture (anti-causative), a non-target one (passive) given ACT verb morphology and a
dis-preferred one (transitive). Thus, in a sentence like i bala tripai (the ball pierces-
ACT), the preferred picture is the one showing the ball with a hole with air blowing
out and a nail next to it. One of the other pictures showed the ball piercing something
else; in this example the ball was intact in front of a painting with a hole in it (the
‘transitive’ reading) which is non-target given NACT verb morphology. The third
picture showed someone (an agent) opening a hole to the ball with a nail (the
‘passive’ reading).
36 The term ‘preferred’ is used instead of ‘target’ since the latter refers to grammaticality. Moreover, preference is argued to be determined on the basis of animacy, verb class and voice morphology, of which animacy and verb-class are semantic/pragmatic properties while voice morphology is lexical-syntactic. Since the grammar allows more than one reading in most cases, the grammatical but less preferred reading is referred to as ‘dis-preferred’. 37 The reason why the transitive option is termed dis-preferred for this category as well as category (IIc) verbs is because null objects do not seem to be easily accessible with anti-causatives in ACT morphology (Perez-Leroux, Pirvulescu & Roberge 2008).
ACQUISITION CHAPTER 5 OF TRANSITIVITY ALTERNATIOSN
277
(b) For sentences with NACT morphology and inanimate subject, there was a
preferred picture (anti-causative), a dis-preferred one (passive) and a non-target one
(transitive). The set of pictures used is the one used for (a).
(c) For sentences with NACT morphology and animate subject, there was a preferred
picture (anti-causative) and two dis-preferred ones (passive and reflexive). Thus, in a
sentence like o Mickey tripiete (o Mickey pierces-NACT), the preferred picture is the
one showing Mickey pricked by accident in a cactus. One of the other pictures
showed Mickey pierced by someone else, in the example by a nurse, with a needle
(the ‘passive’ reading). The third picture showed Mickey deliberately piercing himself
with a needle (the ‘reflexive’ reading).
(c) Class III (Reflexives):
(a) For sentences with NACT morphology and animate subject, there was a preferred
picture (reflexive), a dis-preferred one (passive) and a non-target one (transitive).
Thus, in a sentence like to agori plenete (‘the boy washes-NACT’), the ‘preferred’
picture is the one showing the boy washing himself. One of the other pictures showed
the boy being washed by someone else (the ‘passive’ reading) which is allowed by the
grammar but dis-preferred in terms of verb class. The third picture showed the boy
washing someone else (the ‘transitive’ reading) which is non-target given NACT
morphology.
(b) For sentences with ACT morphology and animate subject, there was a preferred
picture (transitive) and two non-target ones (passive and reflexive) given ACT verb
morphology. For both (a) and (b) the same set of pictures was used.
(d) Class IV (Activity predicates):
(a) For sentences with NACT morphology and animate subjects, there was a
reflexive, a passive and an anti-causative event. There is no preference based on verb-
class in this case. Thus, in a sentence like to pedhi krivete (the boy hides-NACT) one
of the pictures showed the boy hiding behind an armchair (the ‘reflexive’ reading),
another picture showed someone else hiding the boy, in the example a woman
pushing the boy behind the armchair (the ‘passive’ reading). The third picture showed
ACQUISITION CHAPTER 5 OF TRANSITIVITY ALTERNATIOSN
278
the boy standing still behind an armchair partly hiding him (the ‘anti-causative’
reading).
(b) For sentences with NACT morphology and inanimate subjects, there were two
pictures, the anti-causative and the passive, both equally preferred, and a non-target
one (transitive) given NACT verb morphology. Thus, in a sentence like to kuti krivete
(the box hides-NACT) one of the pictures showed a big box behind an armchair, only
the edges of the box were free at sight (the anti-causative’ reading) and another
showed a woman pushing a big box behind an armchair in order to hide it (the
‘passive’ reading). The third picture showed a big box in front of a hiding armchair
(the ‘transitive’ reading).
ACQUISITION CHAPTER 5 OF TRANSITIVITY ALTERNATIOSN
279
3.2. Procedure
Participants at the beginning of the session were informed that they will be seeing sets
of three pictures and simultaneously hear a sentence: they were instructed to decide as
soon as possible which picture matched the sentence they heard. Participants were
presented with sets of three pictures by picture cards, such as the ones illustrated
below, thus the three alternative readings were simultaneously presented. At the same
time, the investigator uttered a sentence in isolation, as in the examples also cited
below, corresponding to the sets of triplets demonstrated for the different verb classes
examined as shown in Picture (6) for the sets of triplets illustrating Voice Alternating
Anti-causatives, in Picture (7) for the two sets of triplets illustrating Voice Non-
alternating Anti-causatives, in Picture (8) for the sets of triplets illustrating Reflexives
and in Picture (9) for the sets of triplets illustrating Activity Predicates:
Picture 7: an example of the sentences used and the sets of pictures presented
(Voice Non-alternating Anti-causatives)
To klidi spai
the key breaks-ACT
ACQUISITION CHAPTER 5 OF TRANSITIVITY ALTERNATIOSN
280
Picture 8: an example of the sentences used and the sets of pictures presented
(Voice Alternating Anti-causatives)
(a) I bala tripai / tripjete (b) O Mickey tripjete
the ball pierces-ACT /pierces-NACT the Mickey pierces-NACT
Picture 9: an example of the sentences used and the sets of pictures presented
(Reflexives)
(a) To aghori plenete (b) To aghori pleni
the boy washes-NACT the boy washes-ACT
ACQUISITION CHAPTER 5 OF TRANSITIVITY ALTERNATIOSN
281
Picture 10: an example of the sentences used and the sets of pictures presented
(Activity predicates)
(a) To pedhi krivete (b) To kouti krivete
the kid hides-NACT the box hides-NACT
The total of sentences presented in each session were pseudo-randomized in that
sentences with verbs of the same class and of the same voice morphology were not
sequentially presented to subjects. Furthermore, the triplets used twice and the triplets
designed for the same verb as (a) and (b) above were always presented in separate
sessions.
Sessions were administered individually in dedicated rooms in the children’s
schools and in other isolated areas for adults. Participants saw each triplet of pictures
while at the same time they heard the experimental sentence uttered by the
experimenter at a normal speaking rate. Sessions were video-recorded. Every session
lasted approximately fifteen minutes. Parental and school consent were obtained prior
to the children’s participation in the study.
3.3. Participants
The SPM task includes 100 subjects in total. Participants were 75 children (age range:
3-5;10), divided in three age groups, and 25 adult controls (age range: 20-38). The
relevant information is presented in Table 65.
ACQUISITION CHAPTER 5 OF TRANSITIVITY ALTERNATIOSN
282
Table 65: Participants
Subjects N Age range Mean Age
Group 1 25 3-3;10 3;7
Group 2 25 4-4;10 4;2
Group 3 25 4;11-5;10 5;4
Adult controls 25 20-38 28
Children in all child groups were born in Northern Greece and live in the towns of
Veria and Thessaloniki. At the time of the study they all attended pre-school and
nursery classes. The native language of all children is Greek, their home and school
language. Children were recruited in four nursery schools (Spring in Panorama
Thessaloniki, Smurfs and Little House on the Prairie and the 3rd State Nursery School
of Veria, in Veria).
The task performed 3 more children (age range: 2;9-2;11) who were excluded
because during the task they showed repeatedly the picture positioned in a particular
spot (for example always the central one). Also data from 10 more children were
eliminated, because they were incomplete. No pre-tests were performed to find out if
children are normally developed, but no diagnosed cases were included in the task.
The Adults control group, also live in the areas of Veria and Thessaloniki. All
the participants had higher education but none had studied linguistics and they were
all naïve with respect to the research questions of the study.
ACQUISITION CHAPTER 5 OF TRANSITIVITY ALTERNATIOSN
283
3.4. Research Questions
This interpretation task aimed to identify the preferred interpretation among the
choices of transitive, passive, reflexive and anti-causative readings with verbs marked
with active and non-active voice in sentences with animate or inanimate subjects. The
specific research questions are listed below:
a. Is there evidence for the +/-reflexive difference in adult NS data (Tsimpli, 2006)?
In other words do adult data show a preference of reflexive interpretations in NACT
verbs with animate subject, irrespective of verb class?
b. Do data provide evidence for the existence of a lexical verb class of anti-causatives
(Alexiadou & Anagnostopoulou, 2004)?
c. Does the Grammar (voice morphology and +/- reflexive difference) constrain the
possible readings of a verb in NS Greek speaking children (Tsimpli, 2006)? Is there
more ambiguity in the interpretation of non-active verb forms in Greek L1 children,
compared to adult native controls (eg. Class of reflexive verbs)?
d. Is there evidence for the absence of a syntactic passive in the groups of L1 Greek
children (cf. Borer & Wexler 1987, 1992)?
e. Is there a preference for an implicit agent in passives and anti-causatives alike
regardless of voice morphology in child groups (cf. Verrips, 2000)?
ACQUISITION CHAPTER 5 OF TRANSITIVITY ALTERNATIOSN
284
4. Results
4.1. Distribution of readings with respect to Voice Morphology and Subject
Animacy
Starting with the analysis of the total responses received in the task with regard to
(ACT-NACT) verb forms in combination to animate and inanimate subjects, Graph
37 presents the adult controls’ preferences.
Graph 37: distribution of preferred readings with respect to Voice Morphology
and Subject animacy (Adults)
Before proceeding with a presentation, we specifically need to notice that the
percentages of ACT verb forms with animate subject illustrate the preferred reading
of five activity predicate verbs which are classified as ‘inherently reflexives’ in
NACT (cf. Alexiadou & Anagnostopoulou 2004) (Class III); in that sense there are no
unergative or anti-causative (metaphorical or literal) readings found with ‘Voice
(Non)-Alternating Anti-causatives’ in the corpus (see Chapter 3). ACT verbs with
inanimate subjects illustrate the preferred readings of ten ‘Voice (Non)-Alternating
Anti-causative’ verbs (Classes I and IIa): no ACT forms of ‘activity predicates’ are
included, resulting to a relatively low rate of transitive uses. NACT verb-forms, on the
ACQUISITION CHAPTER 5 OF TRANSITIVITY ALTERNATIOSN
285
other hand, seem more representative: the rates of preferred readings in NACT forms
with animate subjects illustrate results from three verb-classes, namely ‘Voice
Alternating Anti-causatives’ (Class IIc), ‘activity predicates’ (Class IVa) and
‘reflexives’ (Class III), a total of fifteen verbs, while the rates of preferred readings in
NACT forms with inanimate subjects illustrate results from two verb-classes, namely
‘Voice Alternating Anti-causatives’ (ClassIIb) and ‘activity predicates’ (Class IVb), a
total of ten verbs.
More specifically, as shown in the Graph, when ACT verb forms were
presented in sentences with animate subjects the only reading depicted was the
transitive one, while in sentences with inanimate subjects transitive, anti-causative
and passive readings were available. The reader is reminded that the verbs used in this
condition were Voice (Non)-Alternating Anti-causatives, hence the preference of the
anti-causative reading (antic (63.8%) vs other readings (37.2%): χ2=6.760, p=.009)
and the availability of an ungrammatical passive one, given ACT morphology, which
will be discussed in the next section.
When NACT forms of the verbs were presented in sentences with animate
subjects, the reflexive readings were preferred over the anti-causative and passive,
also available (refl (80.54%) vs other readings (19.46%): χ2=38.440, p<.001). The
mostly preferred reflexive reading reinforces the assumption that animacy is a
stronger determinant than verb-class (Fotiadou & Tsimpli, to appear; see also Tsimpli,
2006). In the presence of inanimate subjects, however the preferred passive and anti-
causative readings did not differ (pass (39) vs antic (34): χ2=.342, p=.558). Few
ungrammatical transitive responses with NACT verbs (n=7) are attributed to a
methodological problem with the pictures designed for two of the ‘Voice Alternating
Anti-causatives’ and will be further discussed in the next section.
ACQUISITION CHAPTER 5 OF TRANSITIVITY ALTERNATIOSN
286
Graph 38: distribution of preferred readings with respect to Voice Morphology
and Subject animacy (Child Groups)
All child Groups performed different from the Adult controls. Starting with the non-
target responses, all child groups incorrectly allowed for passive and reflexive
readings of ACT verbs with animate subjects and transitive readings of NACT verbs
with animate subjects. Moreover, due to a strong pragmatic bias, non-target responses
are somehow regulated depending on the knowledge of the world they have
estabished based on their own experience (Fotiadou & Tsimpli, to appear): the non-
target passive reading gradually decreases, given that the action involved in the
‘reflexives’ used in the task do no longer involve an agent other than ‘self’. With
respect to ACT verbs with inanimate subjects all child groups allowed for an incorrect
passive reading, as Adults.
Turning to the distribution of the other available readings with respect to the
research variables note that grammatical transitive responses increase from Group 1 to
Group 3, but not significantly. Nevertheless, if we compare the transitive choice in
ACT and NACT voice in the child data, all within-group differences are significant:
Group 1 (χ2=12.255, p<.001), Group 2 (χ2=12.600, p<.001), Group 3 (χ2=44.085,
p<.001) indicating that knowledge of the effects on transitivity due to voice change is
ACQUISITION CHAPTER 5 OF TRANSITIVITY ALTERNATIOSN
287
aleady part of the child grammar. In ACT verbs with inanimate subjects the rates of
an anti-causative reading also increase (Group 1 vs 3: χ2=,5.018 p=.025), while the
‘ungrammatical’ passive responses decrease (Group 1 vs 3: χ2=10.385, p=.001). With
respect to NACT verb forms with animate subjects the availability of reflexive
readings increases from Group 1 to Group 3 (χ2=6.000, p=.014); in the presence of
inanimate subjects all child groups provided ambiguous (passive and anti-causative)
responses. In all, child data show that there is a tendency that the responses from
Group 3 approximate adult choices, in that ungramatical responses decrease while
grammatical ones increase; however child data differ from adult data depicting not
lack of abstract syntactic knowledge, but insufficient exposure to pragmatic properties
of verbs that would regulate child responses according to verb classification (Tsimpli,
2006; Fotiadou & Tsimpli, to appear). The distribution of the available readings is
next discussed in detail for each verb class used in the task.
Summary
Overall, adults mostly perceived active verbs with inanimate subject as anti-
causatives: their preferred readings seem to rely on verb classification, since these
verbs are labeled as such in the literature. Nevertheless, when adults encountered non-
active verbs with animate subjects they provided more reflexive than passive or anti-
causative responses, irrespective of verb classification. Finally, NACT verbs with
inanimate subjects were considered ambiguous between passive and anti-causative
readings.
Child data differed from the adult controls’ in that incorrect passive and anti-
causative readings were provided for active verb forms and transitive for non-active
verb forms. These findings are discussed in more detail in the next section. With
respect to grammatical responses the pattern attested is in line with the adults’
preferences.
ACQUISITION CHAPTER 5 OF TRANSITIVITY ALTERNATIOSN
288
4.2. Interpretations with respect to Morphology and Animacy per Verb Class
For the reader’s convenience, the presentation of the results will be according to
Voice Morphology and Subject Animacy (the variables tested) in each Verb Class.
The classification of verbs follows the example of the presentation in Chapter 3,
except for Class III of Chapter 3 which is divided here into (III) ‘reflexives’ and (IV)
‘activity predicates’.
4.2.1. ACT verb morphology and Inanimate Subject: Voice (Non)-alternating
Anti-causatives (Classes I and IIa)
Starting with the ten verbs (five voice ‘non-alternating’ and five ‘alternating’ anti-
causatives) and the relevant sentences included in this category, results presented in
Graph show that all groups allow for all readings. Note also that Class IIa differ from
Class I only in the availability of a NACT alternate.
Graph 39: voice (non)-alternating anti-causatives (I)-(IIa): ACT morphology and
Inanimate Subject
ACQUISITION CHAPTER 5 OF TRANSITIVITY ALTERNATIOSN
289
Transitive responses are found in all groups, but the adults provide more transitive
responses for class IIa. Moreover, in both verb classes there is no developmental
pattern attested with respect to the choice of the transitive reading, thus supporting the
claim that null objects are allowed in Greek child grammars (Tsimpli &
Papadopoulou 2006). Transitive responses across age groups may be accounted for as
an effect of ACT morphology. Moreover, ACT voice in combination with verb class
may be responsible for the difference in transitive responses in the adult group
(χ2=24.923, p<.001). Among Voice Non-alternating Anti-causatives (Class I),
transitive responses are found in all groups, but the adults provide fewer transitive
responses than any child group (Group 1 vs Adults: (χ2=5.828, p=.016), Group 2 vs
Adults: (χ2=5.828, p=.016), Group 3 vs Adults: (χ2=4.481, p=.034)). Among Voice
Alternating Anti-causatives (Class IIa) transitive responses are also found in all
groups, but the adults provide more transitive responses than child groups 1 and 2
(Group 1 vs Adults: (χ2=11.951, p=.001), Group 2 vs Adults: (χ2=8.138, p=.004). The
rates of transitive responses in child groups gradually increase (Group 1 vs Group 3:
(χ2=4.083, p=.043)), up to the point that children in Group 3 approximate adult
choices (Group 3 vs Adults: (χ2=2.253, p=113)).
Turning to anti-causative readings note that the same pattern is attested in both
verb classes with respect to child responses, while adults gave significantly more anti-
causative responses for Class I than Class IIa verbs (χ2=8.720, p=.003). More
specifically, anti-causative readings are significantly more frequent than transitive in
all child groups for both voice ‘non-alternating’ (Group 1: (χ2=7.563, p=.006), Group
2, (χ2=18.050, p<.001), Group 3 (χ2=22.827, p<.001)), and ‘alternating’ anti-
causatives (Group 1: (χ2=23.211, p<.001), Group 2, (χ2=19.512, p<.001), Group 3
(χ2=10.894, p=.001)). In Adults, anti-causative readings were significantly more
frequent than transitives only in Voice Non-alternating Anti-causatives (Class I:
χ2=75.438, p<.001; Class IIa: χ2=2.462, p=117).
Finally, all groups incorrectly allowed also for passive readings. In both verb
classes (I and IIc) Group 1 gave more passive than anti-causative readings, but not
significantly so. While the rate of passive readings decreases for both Class I and IIc
verbs, a significant difference that evinces a developmental pattern is attested only in
Class IIc verbs (Group 1 vs Group 3: χ2=8.711, p=.003). With respect to the passive
ACQUISITION CHAPTER 5 OF TRANSITIVITY ALTERNATIOSN
290
(non-target) readings in Class I verbs, their rate decreases between Groups 1 and 3,
but the difference is only significant between each child group and the controls
(Group 1 vs Adults: (χ2=20.753, p<.001), Group 2 vs Adults: (χ2=9.615, p=.002),
Group 3 vs Adults: (χ2=9.00, p=.003)). Also, anti-causative readings are significantly
more frequent than passives in Class I verbs, only in the responses that adults gave
(χ2=38.088, p<.001). In Class IIa, Groups 1 and 2 equally allowed for anti-causative
and incorrect passive readings. The rate of incorrect passive readings decreases
developmentally (Group 1 vs Group 3 (χ2=8.711, p=.003)), up to the point that
passives are significantly less frequent than anti-causatives in the responses that
children of Group 3 (χ2=10.894, p=.001) and adults gave (χ2=17.778, p<.001). The
fact that adults also allow for the (non-target) passive reading is problematic.
Thus, we next present results from each experimental sentence.
Table 66: Non-target passive reading among ACT verbs with inanimate subject
(Classes I and IIa)
Class I Class IIa
S1 S2 S3 S4 S5 Mean
Total S6 S7 S8 S9 S10
Mean
Total
Group 1 9 17 11 15 9 61 12 14 10 11 12 59
Group 2 1 11 13 14 6 45 1 11 13 10 8 43
Group 3 3 9 12 14 6 44 1 7 5 11 6 41
Adults 9 2 2 7 0 20 1 1 12 3 4 21
The passive non-target reading in the adult data with respect to Class I verbs is mostly
found with the sentences to klidhi spai (the key breaks) (S1) and I porta klini (‘the
door closes’) (S4). It should be noted that these two verbs are change-of-state verbs
with external cause; hence contextually the ‘agent’ reading may be more salient than
with the other verbs included in this class. The ‘agent’ reading was salient for child
groups for all the verbs used; child groups do not seem to disallow the passive reading
for ACT anti-causatives, a result which is consistent with Verrips’ (2000) data from
Dutch.
ACQUISITION CHAPTER 5 OF TRANSITIVITY ALTERNATIOSN
291
The passive non-target reading in the adult data with respect to Class IIa verbs
is found with the sentence I karekla dhiploni (S8), less with I kaltsa leroni (S9) and I
bala tripai (S10) and once with to trapezi htipai (S6), to shini tendoni (S7). The
problem was specifically evident with (S8) and the ‘passive’ reading was depicted by
a woman folding the chair. It is possible that this was perceived as a middle structure,
i.e. the chair can fold, i.e. folding is one of its properties. Given that 12/25 responses
from the adults were non-target with this particular sentence, it is possible that the
increased number of passive readings is an artifact. Alternatively, the passive non-
target reading received for these and the rest of the sentences may be attributed to the
middle reading available for all verbs.
Turning to the transitive and anti-causative readings, Table 67 below presents
the distribution of all test sentences in each Verb Class for the child and adult control
groups.
Table 67: Transitive and anti-causative readings with ACT anti-causative verbs
(inanimate subject)
Class I Class IIa
S1 S2 S3 S4 S5 Mean
TotalS6 S7 S8 S9 S10
Mean
Total
Group 1 6 1 5 5 4 21 7 1 3 5 1 17
Group 2 4 4 7 4 2 21 9 5 4 3 0 21
Group 3 2 2 5 7 3 19 16 6 7 2 0 31
trans
itive
Adults 3 1 0 4 0 8 8 5 5 21 5 44
Group 1 10 7 9 5 12 43 6 10 12 9 12 49
Group 2 20 10 5 7 17 59 15 9 8 12 17 61
Group 3 20 14 8 4 16 62 8 12 13 11 19 63
anti-
caus
ativ
e
Adults 13 22 23 14 25 97 16 19 8 1 16 60
The dis-preferred transitive reading in the adult data, with respect to Class I verbs, is
found only with the sentences to klidhi spai (the key breaks) (S1), to dhentro lijizi (the
tree bends) (S2) and I porta klini (the door closes) (S4), while child groups did not
show specific preferences depending on the sentences used. The dis-preferred
ACQUISITION CHAPTER 5 OF TRANSITIVITY ALTERNATIOSN
292
transitive reading in the adult data, with respect to Class IIa verbs, is found with all
the sentences but specifically with I kaltsa leroni (the sock spills/is spilled) (S9) and
To trapezi htipai (the table hits/is hit) (S6). Child groups gave transitive responses
mostly for (S6), while less to no transitive answers is provided for (S9) and I bala
tripai (the ball pierces/is pierced) (S10).
The anti-causative reading in the adult data is the preferred one for all verbs
included for Class I, while among Class IIa it is mostly evidenced in to shini tendoni
(the cord stretches/is stretched) (S7), as well as in to trapezi htipai (the table hits/is
hit) (S6) and I bala tripai (the ball pierces/is pierced) (S10) among Class IIa verbs. In
the child data, the sentences (S1) and (S5) among Class I verbs and (S10) among Class
IIa verbs are mostly anti-causative. In the remaining sentences the passive reading
(preferred over the other readings) gradually decreases and the anti-causative
increases.
4.2.2. NACT verb morphology and Inanimate Subject: Voice Alternating Anti-
causatives and Activities (Classes IIb and IVb)
Turning to the ten verbs (five ‘voice alternating anti-causatives’ and five ‘activities’)
and the relevant sentences included in this category, results presented in Graph show
that all child groups allow for all readings while Adults allow for transitive (incorrect)
readings only with Class IIb verbs. Recall that these verbs may appear with ACT
voice morphology too. The same verbs and relevant sentences included for IIa are
presented here under IIb, the only difference being that the verb is in NACT voice
morphology. Class IVb verbs are activity predicates presented here in NACT with
inanimate subjects.
ACQUISITION CHAPTER 5 OF TRANSITIVITY ALTERNATIOSN
293
Graph 40: voice alternating anti-causatives and activities (IIb)-(IVb): NACT
morphology and Inanimate Subject
The responses provided for the two verb classes do not differ in any age group: For
both verb classes (IIb and IVb) Group 1 children prefer the passive over the anti-
causative interpretation (IIb: χ2=4.00, p=.046; IVb: χ2=7.397, p=.007), while child
groups 2 and 3, as well as Adults do not show significant preference of one reading
over the other, for any of the two verb classes.
With respect to the non-target (transitive) reading, children incorrectly allow it
with both classes of verbs. Nevertheless, the difference between the non-target and the
two target readings is significant for all child groups and for both classes: for Class
IIb: Group 1 (χ2=45.00, p<.001), Group 2 (χ2=49.928, p<.001), Group 3 (χ2=49.928,
p<.001); for Class IVb: Group 1 (χ2=60.552, p<.001), Group 2 (χ2=52.488, p<.001),
Group 3 (χ2=88.200, p<.001). In the ‘voice alternating anti-causative’ class of verbs,
Adults also produced a few non-target responses (7/125). Thus, we next present
results from each experimental sentence.
ACQUISITION CHAPTER 5 OF TRANSITIVITY ALTERNATIOSN
294
Table 68: Non-target transitive reading with NACT Voice alternating anti-
causative verbs (inanimate subject)
Class IIb Class IVb
S11 S12 S13 S14 S15Mean
Total S36 S37 S38 S39 S40
Mean
Total
Group 1 2 3 7 7 6 25 6 0 10 2 1 19
Group 2 3 1 7 4 8 23 6 0 10 3 3 22
Group 3 0 0 5 5 13 23 3 0 6 0 1 10
Adults 0 0 2 2 3 7 0 0 0 0 0 0
The transitive non-target reading in the adult data with respect to Class IIb verbs is
found with the sentences to shini tendonete (the cord is stretched) (S13), I karekla
dhiplonete (the chair is folded) (S14), to trapezi htipiete (the table is hit) (S15). It is
possible that the transitive responses are due to a problem with the ‘triplets’ for these
verbs and more specifically to the pictures depicting the transitive reading not clearly
enough. In (S13) a man was pulling a rope to stretch the sail in a boat, thus the passive
interpretation was possible due to the presence of an agent. In (S14) a folded chair
was closing the table cloth in it, the illustration being also suitable for the description
of a state of affairs, i.e. the anti-causative reading. Finally, in (S15) the illustration of
the transitive reading differed from the anti-causative only in that in the former the
wall touching the table was damaged, while in the latter the table itself. This problem
may also account for the difference in the transitive responses produced by Group 3
which are significantly higher in IIb than IVb (χ2=5.021, p=.024). If this is correct,
then children in Group 3 approximate adult choices with respect to transitive
responses, given that they are found only in these three sentences. Also, transitive
responses of the other two child groups were higher in theses sentences compared to
(S11) and (S12).
The transitive non-target reading in the child data with respect to Class IVb
verbs is mostly found in the sentence I ammos metaferete (the sand is transferred)
(S38) (Group 1: 10/19, Group 2: 10/22 and Group 3: 6/10) where the picture
illustrated a boy seating in the seaside holding shells transferred with the sand because
of the wind. Thus, the transitive interpretation was not easily depicted for pragmatic
ACQUISITION CHAPTER 5 OF TRANSITIVITY ALTERNATIOSN
295
reasons. Also, the sentence I obrela vrehete (the umbrella is wet) (S36) seems to have
caused some interpretational problem specifically in child groups 1 and 2, since the
picture illustrating the transitive reading included a girl opening un umbrella full of
water; the existence of an animate may have disoriented younger children with respect
to the attributed reading.
If we compare the readings of ‘voice alternating anti-causatives’ in (IIa) and
(IIb), where the difference is only voice morphology, adults produce significantly
more transitive responses for ACT (n=44) than NACT (n=7) marked verbs
(χ2=26.843, p<.001). The number of transitive responses produced by children is not
significantly different. This could be due to an overgeneralization that children make
of the ‘alternating’ property of this class of verbs with inanimate subjects. Recall that
voice changes on these verbs when the subject is inanimate do not correspond to
transitivity changes, in that the anti-causative reading is available with either voice
marking. The overgeneralization consists in the children assuming that voice
morphology does not affect the availability of the transitive reading either, which is
nevertheless the least preferred in both ACT and NACT compared to the passive and
the anti-causative.
With regard to the passive and anti-causative readings Table 69 below
presents the distribution of all test sentences in each Verb Class for the child and adult
control groups.
Table 69: passive and anti-causative readings with NACT verbs with inanimate
subjects (Classes IIb and IVb)
Class IIb Class IVb
S11 S12 S13 S14 S15Mean
TotalS36 S37 S38 S39 S40
Mean
Total
Group 1 16 13 12 10 9 60 5 23 9 13 7 67
Group 2 11 11 11 14 3 50 4 23 5 10 16 58
Group 3 13 9 10 16 3 51 2 25 4 14 16 61
Pass
ive
Adults 19 20 3 6 11 59 1 21 9 5 16 52
ausa Group 1 7 9 6 8 10 40 14 2 6 10 7 39
ACQUISITION CHAPTER 5 OF TRANSITIVITY ALTERNATIOSN
296
Group 2 11 13 7 7 14 52 15 4 10 12 6 45
Group 3 12 16 10 4 9 51 20 0 15 11 8 54
Adults 6 5 20 17 11 59 24 4 16 20 9 73
Among Class IIb verbs, with regard to adult data, the passive reading is particularly
evident in the sentences I kaltsa leronete (the sock is spilled) (S11) and I bala tripiete
(the ball is pierced) (S12), the anti-causative is the preferred reading in the sentences
to shini tendonete (the cord is stretched) (S13) and I karekla dhiplonete (the chair is
folded) (S14), while to trapezi htipiete (the table is hit) (S15) is ambiguous between
the two readings. With regard to child data the passive reading is preferred over the
anti-causative for Group 1, while ambiguity between the two readings is attested for
Groups 2 and 3, except for to trapezi htipiete (the table is hit) (S15) for which the anti-
causative reading is referred over the passive one in all child groups. Among Class
IVb verbs, adults and all child groups preferred the passive over the anti-causative
reading for to dhentro stolizete (the tree is decorated) (S37), but the anti-causative for
I obrela vrehete (the umbrella is wet) (S36). With regard to the remaining sentences
adult differ from child data: while to kuti krivete (the box is hidden) (S39) is anti-
causative for adults, it is ambiguous for all child groups; I ammos metaferete (the sand
is transferred) (S38) is anti-causative for adult and child groups 2 and 3, while Group
1 prefer the passive reading; I porta vafete (the door is painted) (S40) is passive for all
but Group 1 who showed ambiguity between passive and anti-causative
interpretations.
ACQUISITION CHAPTER 5 OF TRANSITIVITY ALTERNATIOSN
297
4.2.3. NACT verb morphology and Animate Subject: Voice Alternating Anti-
causatives and Activities (Classes IIc and IVa)
In Graph 41, the same verbs and relevant sentences as in Graph 40 are presented (five
‘voice alternating anti-causatives’ and five ‘activities’), the difference being that the
subject is animate. Given that the passive and the anti-causative readings are assumed
to involve the same derivation and differ only in the ‘agent’ or ‘cause’ implicit
argument, we present these responses separately and jointly in order to compare them
with reflexive readings.
Graph 41: voice alternating anti-causatives and activities (IIc)-(IVa): NACT
morphology and Animate Subject
Overall children consider sentences of both Class (IIc) and (IVa) ambiguous between
the reflexive and the non-reflexive (passive/anti-causative) readings. However,
Groups 1 and 2 gave fewer reflexive than pass/antic responses in class IIc verbs and
the difference is significant (Group 1: χ2=12.168, p<.001; Group 2: χ2=5.000, p=.025).
However, reflexive readings in class IIc gradually increase up to the point that in
Group 3 the difference between reflexive and non-reflexive readings is not significant,
providing evidence for a developmental pattern; the difference between Groups 1 and
ACQUISITION CHAPTER 5 OF TRANSITIVITY ALTERNATIOSN
298
3 is significant (χ2=7.807, p=.005). The difference between reflexive and non-
reflexive readings in class (IVa) verbs was not significant in any of the child groups.
If we compare reflexive with pass/antic responses in both verb classes, Group 1
children produced more pass/antic responses than reflexives overall (χ2=7.744,
p=.005). The difference between the reflexive and the pass/antic. choice for Groups 2
and 3 was not significant for either group jointly for both verb classes, the implication
being that the two ‘classes’ of verbs are not represented as separate in child
grammars, as yet.
Adult controls gave more reflexive than pass/antic responses for both verb
classes and the difference is significant (Class IIc: χ2= 31.752, p<.001 and Class IVa:
χ2=16.200, p<.001). This finding indicates that for adults the reflexive interpretation is
based on the combination of animacy and NACT morphology, and not verb class.
The only verb class effect in the adult responses is that in Class IIc the anti-causative
reading was preferred over the passive (χ2=11.645, p=.001) and the reverse is attested
in Class IVa, i.e. the passive is preferred over the anti-causative (χ2=14.400, p<.001).
Child Groups present ambiguity between passive and anti-causative readings in both
verb classes. However, in Class IIc the anti-causative reading increases gradually but
not significantly, while in Class IVa the passive reading increases gradually up to the
point that Group 3 significantly prefers the passive over the anti-causative reading
(χ2=10.881, p=.001).
We next present results from each experimental sentence. The Table below
presents the distribution of all test sentences in each Verb Class for the child and adult
control groups with respect to reflexive readings.
Table 70: Reflexive readings with NACT anti-causative and activity verbs
(animate subject)
Class IIc Class IVa
S16 S17 S18 S19 S20Mean
Total S31 S32 S33 S34 S35
Mean
Total
Group 1 2 5 19 6 11 43 16 12 16 13 3 60
ACQUISITION CHAPTER 5 OF TRANSITIVITY ALTERNATIOSN
299
Group 2 2 7 21 7 13 50 21 11 16 11 4 63
Group 3 5 10 21 5 16 57 15 16 15 11 1 58
Adults 15 22 23 12 22 94 22 24 20 19 0 85
The reflexive reading is the preferred reading in all adults data, but for the to pedhi
metaferete (the kid is transferred) (S35) of Class IVa verbs. On the other hand, child
Groups 1 and 2 do not show this preference for the verbs of either verb class, except
for to pedhi leronete (the kid spills himself) (S18) in Class IIc verbs. Responses of
Group 3 show a preference of the reflexive reading in all verbs irrespective of verb
class except for (S35) among Class IVa verbs, as adults, as well as the sentences to
koritsaki htipiete (the girl hits herself) (S16) and o Mickey tripiete (Mickey pierces
himself) (S19) among Class IIc.
The Table below presents the distribution of all test sentences in each Verb
Class for the child and adult control groups with respect to non-reflexive
(passive/anti-causative) readings.
Table 71: Passive and anti-causative readings with NACT verbs and animate
subjects (Classes IIc and IVa)
Class IIc Class Iva
S16 S17 S18 S19 S20Mean
TotalS31 S32 S33 S34 S35
Mean
Total
Group 1 8 3 7 10 4 32 5 10 1 2 15 33
Group 2 6 14 2 9 1 32 4 5 5 8 15 37
Group 3 4 13 1 6 0 24 9 1 6 11 20 47 pass
ive
Adults 5 0 0 0 1 6 3 0 1 6 22 32
Group 1 15 13 3 9 10 50 4 8 3 10 7 32
Group 2 17 4 2 9 11 43 0 9 4 6 6 25
Group 3 16 2 3 14 9 44 1 8 4 3 4 20
anti-
caus
ativ
e
Adults 5 3 2 13 12 25 0 1 4 0 3 8
Adult controls attribute a passive reading only in (S16) and (S20) among Class IIc
verbs, while anti-causative responses are attested more in o Mickey tripiete (Mickey is
ACQUISITION CHAPTER 5 OF TRANSITIVITY ALTERNATIOSN
300
pierced) (S19) and to pedhi tendonete (the boy is stretched) (S20) and remain very few
for the remaining sentences. Among Class IVa verbs they choose the passive reading
mostly in to pedhi metaferete (the kid is transferred) (S35) and less in to pedhi krivete
(the boy is hidden) (S31) and I nifi stolizete (the bride is decorated) (S34); the anti-
causative responses are very few and are attested in o kirios vrehete (the man is wet)
(S33), (S35) and to pedhi vafete (the boy is painted) (S32).
Child Groups show a mixed pattern with respect to the readings they attribute
in sentences in either verb class: all child groups prefer a passive reading in (S35),
while Group 1 attributes also a passive reading in (S32), while an anti-causative in
(S34); for Group 2 (S32) is ambiguous between reflexive and anti-causative; for
Group 3 (S34) is ambiguous between reflexive and passive; the remaining sentences
include few responses of either reading.
ACQUISITION CHAPTER 5 OF TRANSITIVITY ALTERNATIOSN
301
4.2.4. NACT-ACT verb morphology and Animate Subject: ‘Reflexive’
We finally turn to the verbs classified as ‘reflexives’ (Class III); participants were
addressed sentences with these verbs in NACT and ACT voice morphology with
animate subjects. The Graph below illustrates the distribution of the readings attested
in each Group.
Graph 42: reflexive verbs (III): NACT-ACT morphology and Animate Subject
Starting with the analysis of Class III verbs with NACT morphology, participants of
all groups allowed for both reflexive and passive readings. More specifically, Group 1
accepts both the reflexive and the passive reading (χ2=2.893, p=.089), whereas
Groups 2, 3 and adults significantly prefer the reflexive over the passive (Group 2:
χ2=9.797, p=.002; Group 3: χ2=54.223, p<.01; Adults: χ2=117.128, p< .01). With
respect to the non-target reading, all three groups of children allow for the transitive
interpretation, which, nevertheless, shows a developmental pattern between Groups 1
and 3 (χ2= 4.765, p=.029).
When these verbs are in ACT form, all child groups incorrectly allow for
passive and reflexive readings, while adults interpret them as transitives only.
Furthermore, passive readings decrease significantly from Group 1 to Group 3
ACQUISITION CHAPTER 5 OF TRANSITIVITY ALTERNATIOSN
302
(χ2=6.149, p=.013), while reflexive remain equally available; transitive readings
increase, but not significantly. It is noteworthy that Group 2 presents a simultaneous
increase of incorrect passive readings and a decrease of correct transitive ones, but the
effect disappears in Group 3. Note also, that if we compare the transitive choice in
NACT and ACT voice in the child data, all within-group differences are significant:
Group 1 (χ2=12.255, p<.001), Group 2 (χ2=12.600, p<.001), Group 3 (χ2=44.085,
p<.001) indicating that knowledge of the effects on transitivity due to voice change is
already part of the child grammar. Thus, the overgeneralization attested in Classes IIa
and IIb (Sections 3.2.2.1. and 3.2.2.2.) are not concluded to indicate lack of Voice in
child grammar, since in the ‘reflexive’ class, voice changes signal significant changes
in the transitive preference.
We next present results from each experimental sentence. With respect to the
passive and reflexive readings Table illustrates the distribution of all test sentences in
NACT voice morphology (IIIa) for the child and adult control groups.
Table 72: Passive and reflexive readings of NACT verbs with animate subjects
(Reflexives (IIIa))
passive reflexive
S21 S22 S23 S24 S25Mean
Total S21 S22 S23 S24 S25
Mean
Total
Group 1 13 8 3 12 11 47 10 11 20 11 13 65
Group 2 7 12 5 10 8 42 16 11 18 14 17 76
Group 3 6 7 2 4 1 20 18 16 22 21 24 101
Adults 0 2 0 0 0 2 25 23 25 25 25 123
With respect to adult controls responses, all verbs were attributed a reflexive reading
except for to pedhi dinete (the kid is dressed) (S22) where passive was also allowed.
With respect to child data, passive readings are attested for all the sentences; they are
more frequent than reflexive in to pedhi skupizete (the kid is wiped) (S21) and to
pedhi plenete (the kid is washed) (S24) for Group 1and in to pedhi dinete (the kid is
dressed) (S22) for Group 2; in the remaining sentences the passive is highly frequent
also.
ACQUISITION CHAPTER 5 OF TRANSITIVITY ALTERNATIOSN
303
With respect to the transitive readings Table illustrates the distribution of all
test sentences of Class III verbs for the child and adult control groups. Results include
non-target responses provided for NACT voice morphology (IIIa) and target
responses provided for ACT voice morphology (IIIb).
Table73: (Non)-target transitive reading of (N)ACT verbs with animate subjects
(Reflexives (IIIa-b))
NACT ACT
S21 S22 S23 S24 S25Mean
Total S26 S27 S28 S29 S30
Mean
Total
Group 1 2 6 2 2 1 13 7 6 9 6 10 38
Group 2 2 2 2 1 0 7 5 7 6 6 4 28
Group 3 1 2 1 0 0 4 10 9 12 11 13 55
Adults 0 0 0 0 0 0 25 25 25 25 25 125
The non-target transitive responses of NACT verb forms are more attested in child
groups 1 and 2; they are found in all sentences but mostly in Group 1 responses for to
pedhi dinete (the kid is dressed) (S22). Given that non-target responses of this
sentence are more than any other, a methodological problem of the illustration of the
mother who looked childish may be to blame. No non-target transitive responses are
attested in adult controls data. Target readings of these same verbs in ACT voice
morphology are attested in all sentences and increase gradually from one group to the
other. Adult responses are target transitives in their majority.
With respect to the non-target passive and reflexive readings Table illustrates
the distribution of all test sentences in ACT voice morphology (IIIb) for the child and
adult control groups.
ACQUISITION CHAPTER 5 OF TRANSITIVITY ALTERNATIOSN
304
Table 74: Non-target passive and reflexive readings of ACT verbs with animate
subjects (Reflexives (IIIb))
passive reflexive
S26 S27 S28 S29 S30Mean
Total S26 S27 S28 S29 S30
Mean
Total
Group 1 9 6 7 3 7 32 9 13 9 16 8 55
Group 2 8 11 10 3 14 46 12 7 9 16 7 51
Group 3 3 4 3 0 5 15 12 12 10 14 7 55
Adults 0 0 0 0 0 0 0 0 0 0 0 0
The non-target passive and reflexive readings with ACT verb forms are attested in all
child groups but not in adult controls. The fact that passive responses increase from
Group 1 to Group 2 is problematic; also while the reflexive decreases, it increases
form Group 2 to Group 3. In other words, while Group 1 shows no preferences, Group
2 prefers the passive interpretation and Group 3 prefers the reflexive even if
ungrammatical, given ACT verb morphology.
Summary
To summarize, adult responses in the SPM task show that ‘non-alternating anti-
causatives’ (I) seem to form an independent verb class for the adult group, while child
data show in general a mixed pattern; nevertheless, results from individual verbs (as
from spai (break) mostly interpreted as anti-causative from all child groups) show that
children were influenced from lexical properties of some verbs too. Also, the
availability of the NACT morpheme in Class IIa verbs may be responsible for the
more frequent transitive responses provided by adults for these verbs compared to
Class I verbs. Child data did not seem to differentiate between preferred readings
according to alternating vs. non-alternating anti-causatives. All groups (children and
adult controls) also allowed some passive responses, strictly speaking ungrammatical
due to ACT voice. These readings are attributed to the salience of the agent reading
possibly due to the semantics of anti-causatives (i.e. change-of-state) in general.
ACQUISITION CHAPTER 5 OF TRANSITIVITY ALTERNATIOSN
305
However, on a more specific level, given that the passive reading with ACT verbs is
elicited with specific sets of pictures and specific verbs, the acceptability of the agent
interpretation could be due to a possible agent depicted with Class I verbs or with the
possible property reading with Class IIa verbs.
NACT verbs of the ‘alternating anti-causative’ class (IIb) and the ‘activities’
class (IVb) with inanimate subjects, show ambiguity between the passive and the anti-
causative reading in adult and child data. This finding supports the assumption that
these readings are not different syntactically. It should be noted however that
preference for the passive over the anti-causative reading or the reverse is found in
adult and child responses for individual verbs.
NACT verbs of the ‘alternating anti-causative’ class (IIc) and the ‘activities’
class (IVa) with animate subjects, show a strong reflexive preference in the adult data
with no corresponding class distinction. Child data show ambiguity between the
reflexive and the non-reflexive readings, once again showing that their preferences are
not based on verb class and that NACT morphology is truly ambiguous. When
individual verbs are examined some exceptions are attested: for example, both child
groups and adult controls interpreted (S35, i.e. the verb metaferi ‘transfer’ in NACT)
as passive.
Finally, reflexives (III) - always presented with animate subjects - are
exclusively interpreted by adults as such when occurring in NACT morphology and as
transitive verbs when in ACT, the implication being that in combination with voice
morphology adult grammars include a reflexive verb class. In contrast, child groups
show ambiguity in the interpretation of this class of verbs as well; they allow both
passive and reflexive readings as well as some ungrammatical transitive readings. The
same finding is attested when these same verbs were used in ACT: child groups
provided some transitive responses but also some ungrammatical reflexive and
passive ones.
ACQUISITION CHAPTER 5 OF TRANSITIVITY ALTERNATIOSN
306
4. 3. Per Verb analysis
The per verb analysis presented below aims to explore the degree of the interaction
between verbal semantics and verb classification with subject animacy and Voice
Attested readings not included in the table: * unergatives (act-anim)= 4 **reciprocals (nact-anim)=1(nact-inanim)= 1 *** unergatives (act-anim)=3 and reciprocals (nact-anim)=1 **** unergatives (act-anim) =248 and (act-inanim)= 12 ***** unergatives (act-anim)=4 (act-inanim)=50 and reciprocals (nact-anim)=83 (nact-inanim)=6
The more frequent occurrence of ACT verb forms should elicit shorter RTs, according
to coarse-grained models of sentence processing. Frequency data from the total of
adult written corpora we examined showed that in a total of 63268 sentences analysed
GENERAL DISCUSSION CHAPTER 6 COMPARING FREQUENCIES TO ADULT & CHILD LANGUAGE DATA
327
ACT (43466) verbs were included in the 68.7% of the sentences, thus clearly
significantly more frequent than NACT (19802) verbs used in 31.3% of the total
number of sentences. Considering the frequency of anti-causative readings among
ACT and NACT verbs, further evidence is provided with regard to the RTs received
in the SPR task: anti-causative readings show similar rates with ACT (25.42%) and
NACT (23.56%) verbs, indicating once again that ACT anti-causatives are more
frequent than NACT anti-causatives. Limiting furthermore the search to verbs
classified in the literature as ‘anti-causatives’, the data in Table 75 show that in the
total number of occurrences in the corpora of all the verbs included in the SPR task
(N=31072) ACT marked verbs (75.42%) are more frequent than NACT ones
(24.57%). Moreover, ACT verbs with animate subjects were mostly used transitively
(8196/10411; 78.72%), while ACT verbs with inanimate subjects were mostly
interpreted as anti-causatives (8857/13285; 66.67%). On the other hand, NACT forms
with inanimate subjects were mostly passive (2258/4003; 56.40%) and anti-causative
(1705/4003; 42.59%), while with animate subjects reflexive (1536/3631; 42.30%) and
anti-causative (1235/3631; 34.01%).
With respect to the processing data, RTs on the verb segment support coarse-
grained models given that a significant voice effect was registered: ACT verb forms
yielded significantly shorter reading times than NACT verb forms. However, this
finding is also consistent with the operation of a grammar-driven parser, since voice
morphology signals transitivity alternations. Moreover, NACT morphology is more
complex than ACT, hence the difference between RTs received.
The distribution of the most frequent readings of ACT and NACT forms per
verb is required in order to evaluate ‘fine-grained’ lexicalist accounts of sentence
processing. The verbs included in the SPR task are more frequently used in ACT than
NACT, and interpreted as transitives or anti-causatives. Few exceptions are attested,
where the frequency of ACT and NACT forms does not differ: specifically, the verbs
berdevi (mingle) and tendoni (stretch) were equally used in ACT and NACT, while
the verb katharizi (clean) was more frequently used in NACT than ACT. The high
availability of both transitive and anti-causative readings in ACT is attributed to the
two verb classes of anti-causatives included: ‘voice non-alternating anti-causatives’
GENERAL DISCUSSION CHAPTER 6 COMPARING FREQUENCIES TO ADULT & CHILD LANGUAGE DATA
328
are interpreted as anti-causatives, while ‘voice alternating anti-causatives’ are
interpreted as transitives instead. This implies that the verb class distinction based on
the possibility of voice alternation on verbs is registered in the lexicon of the native
speaker and regulates frequencies of readings and accessibility thereof, accordingly.
Among ‘voice non-alternating anti-causatives’ (interpreted as anti-causatives
in the corpora) few exceptions are attested: the verbs klini (close) and ljoni (melt)
were equally frequently used as transitives and anti-causatives. The NACT forms of
these verbs were infrequent in the corpora, but with some exceptions: more
specifically, the verb klini (close) was highly used in NACT morphology too, with a
reflexive reading; the verb vrazi (boil), although appearing significantly less
frequently in NACT than ACT, was often used as passive. Among ‘voice alternating
anti-causatives’, the verbs berdevi (mingle) and tendoni (stretch) were used with a
transitive reading when in ACT and with an anti-causative reading when in NACT.
Also, the verb katharizi (clean) was mostly used as passive when in NACT. The
The ‘reflexives’ used in the task are mostly interpreted as such both in the corpora
(98.04%) and the adult data (100%) when in NACT morphology, while they are
interpreted as transitives when in ACT morphology. In child data however along with
reflexive readings for NACT forms, passive interpretations were also very frequent,
though developmentally decreasing. Notice, for example that for the verb pleni (wash)
passive interpretations were highly available in Group 1 (48%) but significantly
reduced in Group 3 (16%); similarly the other two verbs were ambiguous between
passive and reflexive readings for Group 1 (htenizi (comb): 44% and dini (dress):
32%), while reflexive readings were preferred instead by Group 3 for the verbs (pleni
GENERAL DISCUSSION CHAPTER 6 COMPARING FREQUENCIES TO ADULT & CHILD LANGUAGE DATA
346
(wash): 84%, htenizi (comb): 96% and dini (dress): 64%). Some transitive
(ungrammatical) responses also decreased with age. The ACT forms of these verbs
were also incorrectly perceived as passives or reflexives, as well as with the target
transitive reading. Thus, unambiguous input of these verbs does not seem to have
directed children towards an adult-like reading of these verbs.
The two ‘activities’ used in the task received the following interpretations with
animate subjects in the corpora: krivi (hide) was mostly used as reflexive (1088/1137;
95.69%), while metaferi (transfer) as passive (860/994; 86.51%). Adult controls’
preferences mirror these frequency data. Child data also show that the preferred
readings are similar to those of adults, though the other available readings (namely,
passive and anti-causative for krivi (hide) and reflexive and anti-causative for metaferi
(transfer) are also attested at different rates.
NACT forms of krivi (hide) with inanimate subjects were more frequently
used as anti-causative (819/1257; 65.15%) and NACT forms of metaferi (transfer) as
passive (1971/2605; 75.66%). Both adult and child data are not fully consistent with
these frequencies. While adults interpreted krivi (hide) as anti-causative, they
considered metaferi (transfer) ambiguous between passive and anti-causative
readings. Krivi (hide) was ambiguous between anti-causative and passive readings for
all child groups, while for metaferi (transfer) they provided highly divergent
responses across groups: Group 1 gave equal percentages of transitive
(ungrammatical), anti-causative and passive responses; in Group 2 there is a small
decrease of the passive interpretation and an increase of the anti-causative one; Group
3 perceived it more frequently as anti-causative.
Turning to voice non-alternating anti-causatives (see Table 75), recall that in
the SPM task these verbs are used only in ACT voice with inanimate subjects. The
most frequent reading in this condition (act-inanim) is the anti-causative in both
corpora and adults’ responses. Recall however that passive interpretations were also
attested in adult data, though considered ungrammatical due to voice morphology.
Passive interpretations were specifically attested in spai (break) (36%) and klini
(close) (28%). Children also allowed very frequently for passive readings with these
verbs.
GENERAL DISCUSSION CHAPTER 6 COMPARING FREQUENCIES TO ADULT & CHILD LANGUAGE DATA
347
‘Voice alternating anti-causatives’ were included in active forms with
inanimate subjects and in non-active forms both with animate and inanimate subjects.
In the corpora ACT with inanimate subjects were transitive and not anti-causative
(leroni (spill): 236 vs 33; tripai (pierce): 622 vs 165; htipai (hit): 933 vs 973 and
tendoni (stretch): 124 vs 0). Adults instead, perceived these verbs mostly as anti-
causative, except for the verb leroni (spill) for which transitive interpretations were
more frequent (84%). Child group data included many anti-causative readings and
(ungrammatical) passive ones, while few transitives were presented for all the verbs:
the only exception where transitive uses were more frequent than the other
interpretations provided was Group 3’s responses for the verb htipai (hit) (64%).
NACT forms of these same verbs with inanimate subjects were most
frequently used as anti-causatives in the corpora, except for tripai (pierce) which was
found more as passive than as anti-causative (95 vs 15). For adult controls these verbs
were mostly considered as passives, except for tendoni (stretch). The only case where
we could assume that the adults’ responses converge with frequencies is to consider
only Web frequencies for htipai (hit) where passive readings are preferred over the
not significantly fewer anti-causative ones. Child data were ambiguous between
passive and anti-causative readings, while they also exhibited some ungrammatical,
due to NACT morphology, transitives.
Last, the most frequent readings in the corpora with respect to NACT ‘voice
alternating anti-causatives’ with animate subjects were distributed as follows: leroni
(spill) was anti-causative, tendoni (stretch) and tripai (pierce) were reflexives, while
htipai (hit) co-occurred with reflexive (450), passive (416) and anti-causative (239)
readings. Adults considered htipai (hit) more frequently reflexive (60%), while
passive and anti-causative readings were equally distributed; tendoni (stretch) and
leroni (spill) were also mostly reflexives, while tripai (pierce) was anti-causative.
Child groups attributed more frequently an anti-causative reading in htipai (hit), a
reflexive in leroni (spill), while they considered ambiguous between the reflexive and
the anti-causative the verb tendoni (stretch), and between the passive and the anti-
causative the verb tripai (pierce).
GENERAL DISCUSSION CHAPTER 6 COMPARING FREQUENCIES TO ADULT & CHILD LANGUAGE DATA
348
Overall, ‘item-based’ approaches do not seem to be supported by empirical
data of this research. Instead, a consideration of the interaction of subject animacy and
voice morphology is necessitated in order to account for the present findings. But if
we consider data from individual verbs already reported, it seems that the hypothesis
of ‘verb-island’ constructions is not supported either. Findings rather provide
evidence in favour of the hypotheses (iii) and (iv): namely, children seem to be
sensitive to more fine grained morpho-syntactic analyses, providing evidence in
favour of productivity and creativity in child language and evidence for knowledge of
thematic-roles affecting the predicate meaning.
Although children have not mastered the notion of verb-classes, as shown by
their performance with ‘reflexives’ and ‘voice non-alternating anti-causatives’, they
seem to already have part of morpho-syntactic knowledge: they treat NACT
morphology verbs as involving non-canonical argument to position mappings. Their
few ungrammatical (transitive) responses may be attributed to their incomplete
mastery of morphology (cf. Borer, 2004); similarly the non-target passive
interpretations attributed to ‘voice non–alternating anti-causatives’ can be accounted
for by the existence of an implicit external argument in the derivation (cf. Verrips,
2000), while transitive readings with these verbs provide support for the claim that
null objects are allowed in child grammar (cf. Papadopoulou & Tsimpli, 2005).
Attempting an overall evaluation of the hypotheses formulated for language
acquisition, the predictions according to the usage-based approach do not seem to be
valid in relation to this data. In contrast, the data seem to be supportive of the
assumption that children have abstract knowledge of transitivity alternations, but fail
to perform adult-like due to their incomplete mastery of morphological marking on
the verb mapped onto the effects on argument representation in syntax (e.g. transitive
readings for NACT verbs). Moreover, they fail to provide adult-like responses due to
the lack of sufficient exposure to individual verbs used in appropriate pragmatic
contexts, which would enable them to generalize and form verb classes. However, it
is possible that the elevated rate of passive interpretations for ‘reflexive’ verbs is an
effect of personal experience from the child’s perspective: at a young age for
example, children don’t wash, or dress all alone but some other person assists them.
GENERAL DISCUSSION CHAPTER 6 COMPARING FREQUENCIES TO ADULT & CHILD LANGUAGE DATA
349
Moreover, the frequently attributed ‘passive’ interpretation of ACT anti-causatives is
indicative of this same suggestion, i.e. that exposure to language input helps children
generalize abstract knowledge already at place in order to map particular meanings to
grammatical structures and incorporate notions such as verb classification, developed
at later stages (Tsimpli, 2005,2006).
5. Conclusion
This thesis has provided evidence related to ‘experience-based’ models of
sentence processing and ‘usage-based’ approaches to language acquisition with
respect to transitivity alternations in Greek. Morphological marking of Voice (ACT-
NACT) signals transitivity alternations although not in a deterministic way, since
there are ACT anti-causatives lexically-constrained while NACT marking allows for
various readings (passive, reflexive, anti-causative, middle). It is assumed that only
the reflexive derivation is distinct from the others, namely passive, anti-causative and
middle, in that the subject in reflexive structures is non-derived.
Assessing the processing of and the judgements on transitivity alternations
with an on-line SPR task addressed to monolingual Greek adults revealed a significant
voice effect as well as a significant interaction of voice morphology and subject
animacy. It has thus been demonstrated that the parser is sensitive to morphological
cues such as Voice marking on the verb, while semantic factors such as animacy are
integrated in subsequent stages. This conclusion is based on the fact that the voice
effect was evidenced on the 3rd (critical) segment and often remained active on the
segments following the verb, while the animacy effect, or the interaction between
voice morphology and subject animacy was evidenced only on segments following
the verb.
In accordance with ‘coarse-grained’ models of sentence processing, a
frequency effect was found, while predictions in line with more ‘fine-grained’ models
of sentence processing could not be evaluated with respect to frequency alone. The
GENERAL DISCUSSION CHAPTER 6 COMPARING FREQUENCIES TO ADULT & CHILD LANGUAGE DATA
350
present findings have shown that some effect of the input frequency is attested,
without however constituting a deterministic factor of processing transitivity changes.
On the other hand, a comparison of the findings from the acceptability
judgment and the RTs received in the SPR task, the most frequent readings obtained
in the SPM task and the corpora of adult written (formal and informal) speech reveals
that adult NS of Greek do base their final judgment and preferences on the most
frequently used structure. However, reaction times on the judgment itself do not
reveal any facilitation, evidenced, for example, by short RTs, when the structure
involved is a frequently used one. Instead, findings support a morpho-syntactic and
semantic analysis on-line, while delay is attested when pragmatic knowledge and
lexical properties of verbs such as their verb class, is not met with a desired reading.
The child groups who participated in the off-line SPM task have been shown
to provide non-adult like interpretations of the sentences examined with respect to
transitivity alternations. Their responses, however, differ from those of adults not in
terms of non-target performance but mainly due to the incomplete status of notions
such as ‘verb class’ that regulates, to a large extent, adult performance. Children
appear to be sensitive to grammatical constraints with respect to the difference
between reflexive and non-reflexive derivations, and the factors this difference is
based on, namely theta-feature attraction by NACT voice and the consequent (non)-
derived subject. The tendency towards unique, unambiguous readings of NACT
morphology with animate or inanimate subjects in adult performance is attributed to
the integration of lexical, pragmatic, frequency-based as well as syntactic and
semantic features. The last two are also part of child grammars; the former three
however, are strongly dependent on length of exposure and on the ability to rapidly
integrate consolidated lexical and pragmatic knowledge, such as semantic properties
of verb classes as well as felicitous readings in particular contexts.
The assumption that transitivity alternations consist of a lexicon-syntax
interface phenomenon is corroborated by the corpus analysis too. Verbs classified as
‘reflexives’ or ‘anti-causatives’ receive other interpretations too, depending on the
context in which they are embedded and on the speech register. For example, the
reflexive reading is favoured in the presence of an animate subject (or an inanimate
GENERAL DISCUSSION CHAPTER 6 COMPARING FREQUENCIES TO ADULT & CHILD LANGUAGE DATA
351
non-literal one) for the majority of the verbs examined, even though they are not
classified as such. The interaction of voice morphology and animacy of the syntactic
subject has been shown to significantly affect the reading of verbs classified as ‘voice
alternating anti-causatives’: in the availability of both ACT and NACT morphology
speakers tend to use ACT forms to denote transitive readings (specifically in informal
registers, such as that found in the Web corpus) and NACT to denote passive
interpretations (specifically in a formal register, as instantiated in the ILSP corpus).
Moreover, in instances of colloquial speech high productivity of NACT morphology
is observed, even with verbs that traditionally do not allow this morphological
marking (‘voice non-alternating anti-causatives’), specifically to express passive
interpretations. Finally, the use of the apo-phrase is considered a non-reliable criterion
for passivization in Greek, since it has been found to frequently co-occur with passive
readings only in the formal register (ILSP corpus), while it was associated more
frequently with readings other than the agent one.
The above summary of some of the main findings of this research and the
incompatibility between some theoretical accounts on one hand, and NS actual use of
these structures on the other, further support the suggestion that the grammar does not
distinguish between the passive, the middle, the anti-causative readings but leaves
them underspecified, while an interaction of lexico-semantic properties of the verb
and pragmatic information eventually leads to a strong or unique bias towards one of
these readings.
GENERAL DISCUSSION CHAPTER 6 COMPARING FREQUENCIES TO ADULT & CHILD LANGUAGE DATA
352
6. Methodological limitations and Further Research
Research conducted in the field of psycholinguistics aim to indirectly assess
the linguistic structures and processes underlying the human’s ability to speak and
understand language by means of observation of linguistic behaviour. General
conclusions however, should be drawn only following careful experimental design
and sufficient participants who would be assessed in various ways on the targeted
domains of inquiry. Thus, the present study should be carefully approached and
limitations on the methodology used should be taken into account.
Starting with the selection of the child participants, the fact that no pre-test
evaluation of cognitive or linguistic level of development was used to establish that
children are within the normal range is a drawback. Parents and teachers were asked
their opinion on each child’s development, although this information may still be
considered insufficient.
As far as the materials used are concerned, a general limitation involves the
test items, which were often problematic. More specifically, some of the triplets of
pictures used in the SPM task were identified as problematic and the responses
obtained from these pictures may be an artifact, as already noted in Chapter 5.
Experimental sentences used in the SPR task were created in relation to the variables
controlled: thus, in many cases clauses were problematic but this was part of the
experimental manipulation, given that an acceptability judgment task included in the
procedure aimed to identify the degree of their unacceptability. However, the fact that
mean RTs received are extremely high during sentence processing and, at the same
time, with considerable deviations in the per item analysis, might suggest further
research. Also, a potential effect of the verbs’ compatibility with a null object should
also be evaluated. This necessitates a frequency count with regard to transitive and
‘intransitive’ (i.e. with a null object) uses of the verbs, which was not included in the
present research.
Consequently, in order to improve the validity of the tasks and the
generalizability of the statistical results we consider repeating the procedure with
some changes. Also, we consider analyzing the per subject data in order to examine
GENERAL DISCUSSION CHAPTER 6 COMPARING FREQUENCIES TO ADULT & CHILD LANGUAGE DATA
353
whether there is individual variation in the processing of the verbs examined with
respect to the research variables already discussed (i.e. voice morphology and subject
animacy, as well as the frequency of use of the structures with specific
interpretations). These analyses could help us draw some conclusions in relation to
questions such as the differentiation between individual grammars in the native
speakers . The tasks used together with tests on elicited production of NACT and
ACT verbs as well as eye-tracking data would be of primary importance. Also, the
examination of deponents is of interest for future research, given that these verbs
receive transitive or unergative readings although they appear in non-active (NACT)
form.
It should also be noted that this study examined the effects of morpho-
syntactic and semantic properties of transitivity alternations in the language abilities
of monolingual child and adult populations. It did not however look into bilingual
populations and learners of Greek L2. The bilingual individuals, in particular, would
provide an insight into the intrinsically difficult areas of transitivity alternations since
they are expected, as bilinguals, to have better control of ambiguous interpretations
(Bialystok & Viswanathan, 2009).
Following the above, further research focusing on the exploration of the
domain of VP on various populations could effectively shed light on the nature of the
phenomenon and its relation to Interfaces: anti-causative/passive and reflexive
readings attributed derive from a syntactically-constrained derivation interfacing with
the morphological component, lexical-semantic information attached to the verb entry
and extra-clausal pragmatic and encyclopeadic information (Tsimpli, 2006). It thus
opens up a number of distinct investigations with respect to different interfaces
involved and different populations known to show strengths and weaknesses in some
of these interfaces.
REFERENCES
354
References
Akhtar, N., 1999. Acquiring basic word order: Implications for language development in young children. Quartely Journal of Experimental Psychology, 49 A(1). 216-233.
Akhtar, N., and Tomasello, M., 1996. Two-year-olds learn words for absent objects and actions. British Journal of Developmental Psychology, 14. 79-93.
Alexiadou, A., and Anagnostopoulou, E., 1999. Non-active morphology and the direction of transitivity alternations. In: P. Tamanji, M. Hirotani, and N. Hall (Eds.), Proceedings of NELS 29. 40–55.
Alexiadou, A., 2001. Adjective syntax and noun raising: word order asymmetries in the DP as the result of adjective distribution. Studia Linguistica, 55. 148–217.
Alexiadou, A., and E. Anagnostopoulou, 2004. Voice morphology in the causative - inchoative alternation: Evidence for a non-unified structural analysis of unaccusatives. In: A. Alexiadou, E. Anagnostopoulou, M. Everaert (eds.), The Unaccusativity Puzzle: Explorations of the Syntax-Lexicon Interface. Oxford: Oxford University Press. 114–136.
Alexiadou, A., Anagnostopoulou, E., & M. Everaert, (eds.), 2004. The Unaccusativity Puzzle: Explorations of the Syntax-Lexicon Interface. Oxford: Oxford University Press.
Alexiadou A., Anagnostopoulou, E., and Schäfer, F., 2006. The properties of anticausatives crosslinguistically. In M. Frascarelli (ed.), Phases of Interpretation. Berlin: Mouton de Gruyter. 187-211.
Anagnostopoulou, E., and Everaert, M., 1999. Towards a More Complete Typology of Anaphoric Expressions. Linguistic Inquiry, 30. 97-119.
Aston, G., and Burnard, L., 1998. The BNC Handbook: Exploring the British National Corpus with SARA. Edinburgh, Edinburgh University Press.
Babyonyshev, M., Fein, R., Ganger, J., Pesetsky, D., Wexler, K., 2001. The maturation of grammatical principles: Evidence from Russian unaccusatives. Linguistic Inquiry, 32. 1–44.
Baker, M., 1988. Incorporation: A Theory of Grammatical Grammar Changing. Chicago: University of Chicago Press.
Baker, M., Johnson, K., and Roberts, I., 1989. Passive arguments raised. Linguistic Inquiry, 20. 219–251.
Baroni, I.M., and Ueyama M., 2006. Building general- and special-purpose corpora by Web crawling. Proceedings of the 13th NIJL International Symposium, Language Corpora: Their Compilation and Application. Tokyo. NIJL. 31-40.
REFERENCES
355
Baroni M. and Motoko, U., 2006. Building general- and special-purpose corpora by Web crawling. Proceedings of the 13th NIJL International Symposium, Language Corpora: Their Compilation and Application. Tokyo. NIJL. 31-40.
Belletti, A., 1988. The case of unaccusatives. Linguistic Inquiry, 19 (1). 1-34.
Bennis, H., 2004. Unergative Adjectives and Psych Verbs. In: A. Alexiadou, E. Anagnostopoulou, M. Everaert (eds.), The Unaccusativity Puzzle: Explorations of the Syntax-Lexicon Interface. Oxford: Oxford University Press. 84-113.
Bever, T., 1970. The cognitive basis for linguistic structures. In: J. R. Hayes (ed.), Cognition and the Development of Language. New York: Wiley and Sons. 279–362.
Bialystok E., and Viswanathan, M., 2009. Components of executive control with advantages for bilingual children in two cultures. Cognition, Issue 112. 494-500.
Boland, J.E., Tanenhaus, M.K., Carlson, G., Garnsey, S.M., 1990. Lexical projection and the interaction of syntax and semantics in parsing. Journal of Memory and Language, 29. 413–432.
Borer, H., 1994. The projection of arguments. In: E. Benedicto and J. Runner (eds.), Functional Projections. University of Massachusetts Occasional Papers, 17. 19–47.
Borer, H., 1996. Passives without theta grids. MS. University of Massachusetts.
Borer, H., 1998. Deriving passives without theta-grids. In: S. Lapointe, D. Brentari, and P. Farrell (eds.), Morphology and its Relation to Phonology and Syntax (pp. 60–99). Stanford: California, CSLI Publications.
Borer, H., 2004. The Grammar Machine. In: A. Alexiadou, E. Anagnostopoulou, M. Everaert (eds.), The Unaccusativity Puzzle: Explorations of the Syntax-Lexicon Interface. Oxford: Oxford University Press. 288–331
Borer, J., 2005. Structuring Sense: Volume II: The Normal Course of Events. Oxford: Oxford University Press
Borer, H., and Wexler, K., 1987. The maturation of syntax. In: T. Roeper, E. Williams (eds.), Parameter Setting. Dordrecht: Reidel. 123–172.
Borer, H., and Wexler, K., 1992. Bi-unique relations and the maturation of grammatical principles. Natural Language and Linguistic Theory, 10. 147–189.
Brooks, P., and Tomasello, M., 1999. Young children learn to produce passives with nonce verbs. Developmental Psychology 35(1), 29-44
REFERENCES
356
Brysbaert, M. and Mitchell, D.C., 1996. Modifier attachment in sentence parsing: Evidence from Dutch. Quarterly Journal of Experimental Psychology, 49A. 664-695.
Burkhardt, P., Piñango, M.M., and Wong., K., 2003. The Role of the Anterior Left Hemisphere in Real-Time Sentence Comprehension: Evidence from Split Intransitivity. Brain and Language, 86, 1. 9-22.
Burzio, L., 1981. Intransitive verbs and Italian auxiliaries. PhD Dissertation. MIT.
Burzio, L., 1986. Italian syntax. Dordrecht: Kluwer.
ByBee, J.L., 1995. Regular Morphology and the Lexicon. Language and Cognitive Processes, 10 (5). 425-55.
ByBee, J.L., and Hopper, P., (eds.), 2001. Typological Studies in Language, 45: Frequency and the Emergence of linguistic Structure. Amsterdam/Philadelphia: John Benjamins Publishing Company
Campos, H. 1987. Passives in Modern Greek. Lingua, 73. 64-78.
Chierchia, G., 1989. A semantics for unaccusatives and its syntactic consequences. Ms., Cornell University.
Chierchia, G., 2004. A semantics for unaccusatives and its syntactic consequences. In: A. Alexiadou, E. Anagnostopoulou, M. Everaert (eds.), The Unaccusativity Puzzle: Explorations of the Syntax-Lexicon Interface. Oxford: Oxford University Press. 22-59
Chomsky, N., 1965. Aspects of the theory of syntax. Cambridge, MA. MIT Press.
Chomsky, N., 1969. Linguistics and Philosophy. In: S. Hook (ed.), Language and Philosophy. NY: New Yorki University Press.
Chomsky, N., 1972. Language and Mind (enlarged edition), Harcourt Brace Jovanovich, New York.
Chomsky, N., 1981. Lectures on government and binding. Dordrecht: Foris.
Chomsky, N., 1986. Knowledge of language. New York: Praeger.
Chomsky, N., 1995. The Minimalist Program. Cambridge, MA: MIT Press.
Chomsky, Noam. 1993 (1996-3rd Ed). A Minimalist Program for Linguistic Theory. In K. Hale S.J. Keyser (eds) View from Building 20: Essays in Linguistics in honor of Sylvain Bromberger. MIT Press. 1-52.
REFERENCES
357
Chomsky, N., 2000. Minimalist inquiries: the framework. In: R. Martin, D. Michaels, and J. Uriagereka (eds.), Step by Step: Essays on Minimalism Syntax in Honor of Howard Lasnik. Cambridge, Mass.: MIT Press. 89-156.
Chomsky, Noam. 2001. Derivation by Phase. In Michael Kenstowicz (ed.). Ken Hale:
A Life in Language. Cambridge, MA: MIT Press. 1-52.
Chomsky, N., 2004. Beyond Explanatory Adequacy . In A. Beletti (ed.). Structures and Beyond: The Cartography of Syntactic Structures, Vol.3. New York: Oxford University Press. 104-131
Chomsky, N., 2005. Three Factors on language in the Language Design. Linguistic Inquiry, 36. 1-22.
Chomsky, Noam. 2008. On Phases. In: R. Freidin, C. P. Otero and M.L. Zubizarreta, eds., Foundational Issues in Linguistic Theory. Essays in Honor of Jean-Roger Vergnaud. MA: MIT Press. 133-166
Collins, C. T., 2005. A smuggling approach to the passive in English. Syntax, 8 (2). 81-120.
Condoravdi, C., 1989. The Middle: where Semantic and Morphology meet. MIT Working Papers in Linguistics, 11. 16-30.
Cuetos, F., and Mitchell, D., 1988. Cross-linguistic differences inparsing: Restrictions on the use of the Late Closure Strategy in Spanish. Cognition, 30. 73-105
Cuetos, F., Mitchell, D., and Corley, M.M.B., 1996. Parsing in different languages. In: M.Carreiras, J. Garcia-Albea, and N. Sebastian-Galles (eds.), Language Processing in Spanish. Hillsdalle, NJ: Lawrence Erlbaum. 145-187.
Demuth, K., 1989. Maturation and the acquisition of the Sesotho passive. Language, 65. 56–80.
DeVilliers, J., 1985. The acquisition of English. In: D. Slobin (ed.), The Crosslinguistic Study of Language Acquisition. Hillsdale, NJ: Lawrence Erlbaum. 27–139.
Dowty, D. R., 1979. Word Meaning and Montague Grammar. Dordrecht: Reidel.
Dowty, D., 2003. The dual analysis of adjunct complements in Categorical Grammar. In: E. Lang, C. Maienborn, C. Fabricious-Hansen (eds.) Modifying Adjuncts. Berlin: Mouton de Gruyter. 33-66.
Embick, D., 1998. Voice systems and the Syntax/Morphology interface. MIT Working Papers in Linguistics 32: 41–72.
REFERENCES
358
Embick, D., 2004. Unaccusative syntax and verbal alternations. In: A. Alexiadou, E. Anagnostopoulou, M. Everaert (eds.), The Unaccusativity Puzzle: Explorations of the Syntax-Lexicon Interface. Oxford: Oxford University Press. 137–158
Ferreira, F., 2003. The misinterpretation of non-canonical sentences. Cognitive Psychology, 47. 164-20.
Ferreira, F., and Clifton, C., 1986. The independence of syntactic processing. Journal of Memory and Language, 25. 348–368.
Ferreira, F., and Henderson, JM., 1991. Recovery from misanalyses of garden-path sentences. Journal of Memory and Language, 30. 725-742.
Fotiadou, G., 2007. Voice Morphology and the apo-phrase. In: E. Agathopoulou, M. Dimitrakopoulou, D. Papadopoulou (eds.), 17th International Symposium: Selected Papers on Theoretical and Applied Linguistics, Vol. II. Thessaloniki: Monohromia. 378-387.
Fotiadou, G., in press. Productivity and Interpretations of Greek Voice Morphology: A Corpus Based Analysis. Actes du 30e Colloque International de Linguistique Fonctionnelle. Paris: L’Harmattan.
Fotiadou, G., and Tsimpli, I.M., to appear. On the L1 acquisition of passives and reflexives in Greek: Does frequency count?. Lingua
Fox, D., and Grodzinsky, Y., 1998. Children’s passive: A view from the by-phrase. Linguistic Inquiry, 29. 311–332.
Frazier, L., 1978. On Comprehensing Sentences: Syntactic Parsing Strategies. Unpublished doctoral dissertation. University of Connecticut.Strorrs.
Frazier, L., 1987a. Sentence processing: a tutorial review. In: Coltheart, M. (ed.), The Psychology of Reading. Lawrence
Frazier, L., 1987b. Syntactic processing: evidence from Dutch. Natural Language and Linguistic Theory, 5. 519–559.
Frazier, L., 1990. Parsing modifiers: special purpose routines in the human sentence processing mechanism. In: Balota, D.A., D’Arcais, G.B.F., Rayner, K. (eds.), Comprehension Processes in Reading. Lawrence Erlbaum Associates Inc., Hillsdale, NJ. 301–331.
Frazier, L., and Clifton, C., 1996. Construal. MIT Press, Cambridge, MA.
Frazier, L., Fodor, J.D., 1978. The sausage machine: a new two-stage parsing model. Cognition, 6. 1–34.
Frazier, L., Rayner, K., 1982. Making and correcting errors during sentence comprehension: eye movements in the analysis of structurally ambiguous sentences. Cognitive Psychology, 14. 178–210.
REFERENCES
359
Friedmann, N., Taranto, G., Shapiro, L.P., and Swinney, D., 2008. The leaf fell (the leaf): the online processing of unaccusatives. Linguistic Inquiry, 39. 355-377.
Galani, A., 2005.The morphosyntax of verbs in modern Greek. Doctoral dissertation, University of York, U.K.
Garnsey, S.M., Pearlmutter, N.J., Myers, E., and Lotocky, M.A. 1997. The contribution of verb bias and plausibility to the comprehension of temporally ambiguous sentence. Journal of memory and Language, 37. 58-93.
Gerken, L.A., and Shady, M.E., 1996. The picture selection task. In: D. McDaniel, C. McKee and H. Cairns (eds.), Methods for assessing children's syntax. Cambridge, MA: MIT Press.
Gibson, E. Pearlmutter, N., Canseco-Gonzalez, Hicock, G., 1996. Recency preference in the human sentence processing mechanism. Cognition, 59. 23-59.
Gleitman, L., 1990. The structural sources of verb meaning. Language Acquisition, 1. 3– 56.
Gordon, P., and Chafetz, J., 1990. Verb-based versus class-based accounts of actionality effects in children's comprehension of passives. Cognition, 36. 227-254.
Grimshaw, J., 1981. Form, function and the language acquisition device. In: C. L. Baker, J. M. McCarthy (eds.), The Logical Problem of Language Acquisition. Cambridge, MA: MIT Press. 165–182.
Grimshaw, J., 1982. On the Lexical Representation of Romance Reflexive Clitics. In: J. Bresnan (ed.), The Mental Representation of Grammatical Relations. MIT Press, Cambridge, Mass.
Grimshaw, J., 1990. Argument Structure. Cambridge, MA: MIT Press.
Haegeman, L., 1994. Introduction to Government and Binding Theory, 2nd edn, Blackwell, Oxford.
Hale, K., and Keyser, J., 1986. Some transitivity alternations in English. Lexical Project Workshop Paper, 7. Center of Cognitive Science. MIT.
Hale, K., and Keyser, J., 1993. On argument structure and the lexical expression of syntactic relations. In: K. Hale, J. Keyser (eds.), A view from Building 20. Cambridge, MA: MIT Press. 53–109.
Hale, K., and Keyser, J., 2002. Prolegomenon to a Theory of Argument Structure. Cambridge, MA: The MIT Press.
Haspelmath, M., 1993. More on the typology of inchoative/causative verb alternations. In: B. Comrie, M. Polinsky (eds.), Causatives and Transitivity. Amsterdam: John Benjamins. 87-120.
REFERENCES
360
Holton, D., Mackridge, P., Philippaki-Warburton, I., 1997. Greek. A Comprehensive Grammar of the Modern Language. Routledge, London.
Jackendoff, R., 1972. Semantic Interpretation in Generative Grammar. Cambridge, MA: The MIT Press
Jackendoff, R., 1983. Semantics and Cognition. Cambridge. MA: MIT Press.
Jackendoff, R., 1990. Semantic Structure. Cambridge. MA: MIT Press.
Jaeggli, O., 1986. Passive. Linguistic Inquiry, 17. 587-622.
Janda, R., and Joseph, B., 2002. Pseudo-Agglutinativity in Modern Greek Vverb-Inflection and “Elsewhere”. Papers from the 28th Regional Meeting of the Chicago Linguistics Society (1992). Vol., 1. 251-266.
Joseph, B. D., and Philippaki-Warburton, I., 1987. Modern Greek. Croom Helm.
Joseph, B.D., and Smirniotopoulos. J.C., 1993. The Morphosyntax of the Modern Greek Verb as Morphology and Not Syntax. Linguistic Inquiry, 24. 388-397.
Just, M.A., Carpenter, P.A., and Woolley, J.D., 1982. Paradigms and processes and in reading comprehension. Journal of Experimental Psychology: General, 3. 228-38.
Kallulli, D., 2004. De-agentivised causers or non-active causative predications. In the Pre-conference Proceedings of the Workshop "Demoting the Agent: Passive and Other Voice-related Phenomena": 59-66. University of Oslo, November 25-27.
Kallulli, D., 2006a. Unaccusative with dative Causers and Experiencers: A unified account. In: D. Hole, A. Meinunger and W. Abraham (eds.), Datives and Similar Cases. Amsterdam: John Benjamins. 271-300.
Kallulli, D., 2006b. A unified analysis of passives and anticausatives. In O. Bonami and P. Cabredo-Hofherr (eds.), Empirical Issues in Syntax and Semantics 6: 201-225. Available at http://www.cssp.cnrs.fr/eiss6
Katsika, K., 2009a. Exploring the minimal structure in prepositional phrase attachment ambiguities: Evidence from Greek. Lingua, 1474. J. Linguistics (2008), doi:10.1016/j.lingua.2008.04.006
Katsika, K., 2009b. Sentence Processing Strategies in Adults and Children: PP Attachment in Corpora and Psycholinguistic Tasks. Unpublished doctoral dissertation. Aristotle University of Thessaloniki. Thessaloniki.
Kayne, R., 1988. Romance se/si. GLOW newsletter 20.
Klairis, Ch., and Babiniotis, G., 1999. Grammar of Modern Greek: (2). The Verb in Modern Greek: the organization of the message [Γραµµατική της νέας
REFERENCES
361
ελληνικής. 2. Το ρήµα της νέας ελληνικής: η οργάνωση του µυνήµατος] In collaboration with A. Bakakou, A. Mozer, and S. Skopetea. Athens: Ellinika Grammata.
Koopman, H., and Sportiche, D., 1991. The position of subjects. Lingua, 85. 211–258.
Kratzer, A., 1989. The position of subjects. Lingua, 85. 211-258.
Kratzer, A., 1994. The event argument and the semantics of voice. MS. University of Massachusetts.
Kratzer, A., 1996. Severing the external argument from its verb. In: J. Rooryck and L. Zaring (eds.), Phrase Structure and the Lexicon. Dordrecht: Kluwer. 109–137.
Kratzer, A., 2000. Building Statives. In: L.J. Conathan et al. (eds.), Proceedings of the 26th Annual Meeating of the Berkeley \linguistics Society. Berkeley. 385-399. http://semanticsarchive.net/Archive?G15MmI0M/kratzer.building.statives.pdf
Lakoff, G., 1968. Ordering of transformational rules. Bloomington: Indiana University Linguistics Club.
Lakoff, G., 1970. Irregularity in Syntax. New York: Holt, Rinehart and Winston.
Laskaratou, C., 1984. The passive voice in Modern Greek. PhD Dissertation. University of Reading.
Laskaratou, C., and Philippaki-Warburton, I., 1984. Lexical vs transformational passives in M.G. Glossologia, 2–3. 99–109.
Lavidas, N., 2007a. Transitivity Alternations in Greek [Μεταβολές σtη µεταβατικότητα του ρήµατος της Ελληνικής]. Unpublished doctoral dissertation. National and Kapodistrian University of Athens. Athens.
Lavidas, N., 2007b. The diachrony of the Greek anticausative morphology. In: A. Alexiadou (ed.), Studies in the Morpho-Syntax of Greek. Cambridge Scholars Publishing. 106-137.
Levin, B., and M. Rappaport Hovav, 1995. Unaccusativity: At the Syntax – Lexical Semantics Interface. Cambridge, MA: MIT Press.
Levin, B., and M. Rappaport Hovav, 1996. From lexical semantics to argument realization. Cambridge: Cambridge University Press.
Levin, B., and Rappaport Hovav, M., 1999. Two Structures for Compositionally Derived Events. Proceedings of SALT 9. 199-223.
Levin, B., and Rappaport Hovav, M., 2005. Argument Realization, Research Surveys in Linguistics Series, Cambridge University Press, Cambridge, UK.
REFERENCES
362
Lieven, E., Pine, J., and Barnes, H. 1992. Individual differences in early vocabulary development. Journal of Child Laguage, 19. 287-310.
MacDonald, M.C., 1993. The interaction of lexical and syntactic ambiguity. Journal of Memory and Language, 32. 692-715
MacDonald, M.C., 1997. Lexical representations and sentence processing: An Introduction. Language and Cognitive Processes, 9. 195-201
MacDonald, M.C., Pearlmutter, N.J., Seidenberg, M.S., 1994. Syntactic ambiguity resolution as lexical ambiguity resolution. In: Clifton, C., Frazier, L., Rayner, K. (eds.), Perspectives on Sentence Processing. Lawrence Erlbaum, Hillsdale, N.J. 178–269.
MacWhinney, B., 2000. The CHILDES Project: Tools for Analyzing Talk. 3rd Edition. Mahwah, NJ: Lawrence Erlbaum Associates
Manney, L., 1993. Middle Voice in Modern Greek. PhD dissertation. University of san Diego
Manney, L., 2000. Middle Voice in Modern Greek: Meaning and Function o an Inflectional Category. Amsterdam/Philadelphia: Benjamins
Manzini, M. R., and Roussou, A., 2000. A minimalist theory of A-movement and control. Lingua, 110. 409–447.
Marantz, A., 1984. On the Nature of Grammatical Relations. Cambridge, MA: MIT Press.
Maratsos, M., Fox, D.E.C., Becker, J.A., and Chalkley, M.A., 1985. Semantic restrictions on children’s passives. Cognition, 19. 167–191.
Marchman, V., and Bates, E., 1994. Continuity in lexical and morphological development: A test of the critical mass hypothesis. Journal of Child Language, 21. 339-366.
Marinis, T., 2007. On-line Processing of Passives in L1 and L2 Children. In: Belikova, A., Meroni, L. and Umeda, M. (eds.), Proceedings of the 2nd Conference on Generative Approaches to Language Acquisition North America (GALANA). Somerville, MA: Cascadilla Proceedings Project. 265-276.
Mavrogiorgos, M., 2007. The syntax-morphology interface is at the edge:evidence from Greek clitics. In: A. Alexiadou (ed.), Studies in the Morpho-Syntax of Greek. Cambridge Scholars Publishing. 125-221
Mitchell, D.C., and Brysbaert, M., 1998. Challenges to recent theories of cross-linguistic variation in parsing: Evidence from Dutch. In: D. Hillert (ed.),
REFERENCES
363
Sentence Processing: A Crosslinguistic perspective. Syntax and Semantics, vol. 31. San Diego, CA: Academic Press. 313-335.
Mitchell, D.C., Cuetos, F., Corley, M.B., and Brysbaert, M., 1995. Exposure-based models of human parsing: Evidence for the use of coarse-grained (nonlexicalist) statistical records. Journal of Psycholinguistic Research, 24. 469-488.
Müller N., Schmitz, K., Cantone, K., Kupisch, T., 2006. Null arguments in monolingual children: A comparison of Italian and French. In: V. Torrens and L. Escobar (eds.), The Acquisition of Syntax in Romance Language. Amsterdam / Philadelphia: John Benjamins Publishing Company. 69-94
Papadopoulou, D., Clahsen, H., 2003. Parsing strategies in L1 and L2 sentence processing: a study of relative clause attachment in Greek. Studies in Second Language Acquisition, 25. 501–528.
Papadopoulou, D., Clahsen, H., 2006. Ambiguity resolution in sentence processing: the role of lexical and contextual information. Journal of Linguistics 42 (1). 109–138.
Papadopoulou, D., Marinis, T., 2004. Lexical effects in sentence processing: evidence from modifier attachment in Greek. Proceedings of the 6th International Conference on Greek Linguistics, Crete, 2003. http://www.philology.uoc.gr/conferences/6thICGL/ebook/.
Papadopoulou, D., Tsimpli, I., 2005. Morphological cues in children’s processing of ambiguous sentences: a study of subject/object ambiguities in Greek. In: Brugos, A., Clark-Cotton, M.R., Ha, S. (eds.), BUCLD 29: Proceedings of the 29th Annual Boston University Conference on Language Development. Cascadilla Press, Somerville, MA. 471–481.
Papadopoulou, D., 2006. Cross-linguistic Variation in Sentence Processing: Evidence From RC Attachment Preferences in Greek. Springer.
Papangeli, D., 2004. The morphosyntax of argument realization. PhD Dissertation, University of Utrencht (LOT 86)
Papastathi, M., 1999. The Syntax of the Greek Middle: A Minimalist Approach. In: T. Pavlidou and C. Tzitzilis (eds.), Proceedings of the 4th International Conference on Greek Linguistics, University of Cyprus, Nicosia 2000. Thessaloniki. 418–425.
REFERENCES
364
Papastathi, M., 1999. The Syntax of the Greek Middle: a Minimalist Approach. Proceedings of the 4th International Conference on Greek Linguistics, University of Cyprus, Nicosia 2000.
Papastathi, M., and Tsimpli, I., 2004. The interaction of voice morphology with modality: evidence from transitivity alternations in Greek. Paper presented at the LAGB meeting, Surrey.
Perez-Leroux, A.T., Pirvulescu, M., Roberge, Y., 2008. A Syntactic Transitivity Approach to Null Objects in Child Language. Lingua, 118. 370-398.
Peristeri, E., Tsimpli I.M., and Tsapkini, K., submitted. The Processing of Unaccusativity in Agrammatism. Journal of Applied Pycholinguisitics.
Perlmutter, D., 1978. Impersonal passives and the unaccusative hypothesis. Papers from the Annual Meeting of the Berkeley Linguistic Society, 4. 157-189.
Perlmutter, D., and Postal, P., 1984. Impersonal Passives and Some Relational Laws. In: D. Perlmutter, and C. Rosen (eds.), Studies in Relational Grammar, 2. Chicago: The University of Chicago Press. 126-170.
Pesetsky, D., 1995. Zero Syntax. Experiencers and Cascades. Cambridge, MA: MIT Press.
Philippaki-Warburton, I., 1998. Functional Categories and Modern Greek Syntax . The. Linguistic Review, 15. 159-186.
Pinker, S., 1984. Language Learnability and Language Development. Cambridge, MA: MIT Press.
Pritchett, B.L., 1992. Parsing with grammar: Islands, heads, and garden paths. In: H. Goodluck, and M. Rochermont (eds.), Island Constraints: Theory, Acquisition, and Processing. Dordrecht: Kluwer. 321-349.
Ralli, A., 1999. Topics in comparative morphology. Class Notes of Thermi International Summer School in Linguistics (TISSL).
Ralli, A., 2003. Morphology in Greek linguistics.The state of the art. Journal of Greek Linguistics, 4. 77-129
Ralli, A., 2005. Morphology [Μορφολογία]. Athens: Patakis.
Randall, J., van Hout, A., Weissenborn, J., and Baayen, H., 2004. Acquiring Unaccusativity: A cross-linguistic look. In: A. Alexiadou, E. Anagnostopoulou, and M. Everaert (eds.), The Unaccusativity Puzzle. Oxford: Oxford University Press, 332–354.
Rayner, K., Carlson, M., Frazier, L., 1983. The interaction of syntax and semantics during sentence processing: eye movements in the analysis of semantically
REFERENCES
365
biased sentences. Journal of Verbal Learning and Verbal Behavior, 22. 358–374.
Reinhart, T., 1996. Syntactic Effects of lexical Operations: Reflexives and Unaccusatives. OTS, Working Papers in Linguistics.
Reinhart, T., 2000. The Theta System: Syntactic Realization of Verbal Concepts. OTS Working Papers in Linguistics. University of Utrecht, Utrecht: The Netherlands.
Reinhart, T., 2001. A synopsis of the theta system. UiL OTS Working Papers, University of Utrecht.
Reinhart, T., 2003. The Theta System – An Overview. Theoretical Linguistics, 28. 229-290.
Reinhart, T., and T. Siloni, 2004. Against an unaccusative analysis of reflexives. In: A. Alexiadou, E. Anagnostopoulou, and M. Everaert (eds.), The Unaccusativity Puzzle. Oxford: Oxford University Press. 159–180.
Rivero, M. L., 1990. The Location of Nonactive Voice in Albanian and Modern Greek. Linguistic Inquiry, 21. 135-146.
Rivero, M.L., 1992. Adverb incorporation and the syntax of adverbs in Modern Greek. Linguistics and Philosophy, 15. 289–331.
Rizzi, L. 1997. The fine structure of the left periphery. In: L. Haegeman (ed.): Elements of Grammar: Handbook in Generative Syntax. Dordrecht: Kluwer Academic Publishers. 281-337.
Roeper, T., 1987. Implicit arguments and the head-complement relation. Linguistic Inquiry, 18. 267–310.
Rohde, D. and Gibson, E., 2003. The on-line processing of active and passive structures in English. CUNY Conference 2003.
Roussou, A., and Tsimpli, I.M., 2007. Clitics and Transitivity. In: A. Alexiadou (ed.), Studies in the Morpho-Syntax of Greek. Cambridge Scholars Publishing. 138-174.
Roussou, A., 2008. Voice morphology and ergativity in Modern Greek. Πρακτικά του 8ου ∆ιεθνούς Συνεδρίου Ελληνικής Γλωσσολογίας (ICGL8).
Shetreet, E., Friedmann, N., and Hadar, U., 2009. The neural correlates of linguistic distinctions: Unaccusative and unergative verbs. Journal of Cognitive Neuroscience. DOI: 10.1002/hbm.20904
REFERENCES
366
Schmitz, K., and N. Müller, 2008. Strong and clitic pronouns in monolingual and bilingual acquisition of French and Italian. Bilingualism: Language and Cognition, 11 (1). 19-41
Setatos, M., 1997. Diathesis in Koini Neoelliniki [Η ∆ιάθεση στην Κοινή Νεοελληνική]. Studies for the Greek Language, 17. 202-215.
Schneider, W., Eschman, A., and Zuccolotto, A., 2002. E-Prime User’s Guide. Pittsburg: Psychology Software Tools Inc.
Sioupi, A., 1998. The typology of middle constructions, ergative verbs and passive voice: How similar are they after all? In: S. Lambropoulou (ed.), Proceedings of the 12th International Symposium on Theoretical and Applied Linguistics. Thessaloniki: University Studio Press. 159–170
Smirniotopoulos, J. C., 1991. Lexical Passives in Modern Greek. PhD Dissertation. The Ohio State University (ed. 1992, Outstanding Dissertation in Linguistics series, New York: Garland Press)
Sorace, A., 2004. Gradience at the Lexicon-Syntax Interface: Evidence from Auxiliary Selection and Implications for Unaccusativity. In: A. Alexiadou, E. Anagnostopoulou, M. Everaert (eds.), The Unaccusativity Puzzle: Explorations of the Syntax-Lexicon Interface. Oxford: Oxford University Press, 243-268.
Spivey-Knowlton, M., Sedivy, J.C., 1995. Resolving attachment ambiguities with multiple constraints. Cognition, 55. 227–267.
Spivey-Knowlton, M. and Tanenhaus, M.K., 1994. Referental context and syntactic ambiguity resolution. In: C. Clifton, L. Frazier, and K. Rayner (eds.), Perspectives on Sentence Processing. Hillsdale, NJ: Erlbaum. 415-439
Spivey-Knowlton, M.,Trueswell, J., and Tanenhaus, M.K., 1993. Context Effects in syntactic ambiguity resolution: Discourse and semantic influences in parsing reduced relative clauses. Canadian Journal of Experimental Psychology, 47. 276-309.
Stephany, U., 1981. Verbal Grammar in Modern Greek early child language. In: P. Dale, D. Ingram (eds.), Child Language: An International Perspective. Baltimore: University Park Press.
Stephany, U., 1997. The acquisition of Greek. In: D.I. Slobin (ed.), The Crosslinguistic Study of Language Acquisition 4, Hillsdale, N.J.: Erlbaum. 183-333.
Strohner, H., and Nelson, K., 1974. The young child’s development of sentence comprehension: Influence of event probability, nonverbal context, syntactic form and strategies. Child Development, 45. 564–576.
Stromswold, K., Gisenband, J., Norland, E., Ratzan, J., 2002. Tracking the acquisition and processing of English passives: Using acoustic cues to disambiguate
REFERENCES
367
actives and passives. Paper presented at the CUNY conference on sentence processing. New York City.
Tabor, W., Juliano, C., and Tanenhaus, M.K., 1997. Parsing in a Dynamical System: An Attractor-based Account of the Interaction of Lexical and Structural Constraints in Sentence Processing. Language and Cognitive processes, 1997, 12 (2/3). 211-271.
Taraban, R., McClelland, J.L., 1988. Constituents attachment and thematic role assignment in sentence processing: influences of content-based expectations. Journal of Memory and Language, 27. 597–632.
Theakston, A., Lieven, E., Pine, J., and Rowland, C., 2001. The role of performance limitations in the acquisition of verb argument structure. Journal of Child Language, 10. 123-133.
Theophanopoulou-Kontou, D., 1980. Medio reflexives in Modern Greek [Μέσα αυτοπαθή της Νέας Ελληνικής]. Anatypo. Volume in honour of G.I. Kourmoulis, 1-21.
Theophanopoulou-Kontou, D., 1982. Middle verbs of Modern greek: some preliminary observations on the system of diatheses [Τα µέσα ρήµατα της Νέας Ελληνικής: Μερικές προκαταρκτικές παρατηρήσεις στο σύστηµα των διαθέσεων]. Studies in Greek Linguistics, 2. 51-78.
Theophanopoulou-Kontou, D., 1983. Patient vs. non-patient orientation of the action and the voice distinction in Modern Greek. Glossologia, 2-3. 75-90.
Theophanopoulou-Kontou, D., 2000. -O/-me alternations in MG patient oriented constructions: anticausatives and passives. [∆οµές της ΝΕ µε προοπτική του δέκτη και η κατανοµή –ω/-µαι: τα αντιµεταβιβαστικά και τα παθητικά]In: Pavlidou, T. and C. Tzitzilis (eds.), Proceedings of the 20th Annual Meeting of the Linguistics Dept. of the University of Thessaloniki. 146–157
Theophanopoulou-Kontou, D., 2001. Errors in language usage: truth and myth [Λάθη στη χρήση της γλώσσας: Αλήθεια και Μύθος]. In: G.Charis (ed.), Ten Myths about Greek Language. Athens: Patakis. 53–61.
Theophanopoulou-Kontou, D., 2002. Causative structures in Modern Greek. A Productive Scheme [Μεταβιβαστικές δοµές στη Νέα Ελληνική. Ένα Παραγωγικό σχήµα]. Proceedings of International Conference in Greek Linguistics 5 (ICGL 5). Paris: L’ Harmattan.
Theophanopoulou-Kontou, D., 2003. Motion Verbs in Modern Greek and their transitive use [Ρήµατα Κίνησης της Νέας Ελληνικής και η µεταβιβαστική τους Χρήση]. In: D. Theophanopoulou-Kontou, C. Laskaratou, M. Sifianou, M. Georgiafentis, and V. Spyropoulos (eds.). Contemporary Tendencies in Greek Linguistics: Studies in Honour of Irini Filippaki-Warburton [Σύγχρονες Τάσεις στην Ελληνική Γλωσσολογία: Μελέτες αφιερωµένες στην Ειρήνη Φιλιππάκη- Warburton]. Athens: Patakis. 237-255.
REFERENCES
368
Theophanopoulou-Kontou, D., 2004. The structure of VP and the mediopassive morphology. The passives and anticausatives in Modern Greek. Parousia, 15–16. 173– 206.
Thornton, R., Gil M, and MacDonald, M., 1998. Accounting for cross-linguistic variation: A constraint-based perspective. In: D. Hillert (ed.), Sentence Processing: A Crosslinguistic Perspective. Syntax and Semantics, vol. 3. San Diego, CA: Academic Press. 211-223.
Thornton, R., MacDonald, M., and Gil, M., 1999. Pragmatic constraints on the interpretation of complex noun phrases in Spanish and English. Journal of Experimental Psychology: Learning, Memory, and Cognition, 25(6). 1347-1365.
Tomasello, M., 1992. First verbs: a case study of early grammatical development. Cambridge: Cambridge University Press
Tomasello, M., 2003. Constructing a Language: A Usage-based Theory of Language Acquisition. Harvard University Press.
Tomasello, M., and Brooks, P., 1998. Young children’s earliest transitive and intransitive constructions. Cognitive Linguistics, 9. 379-395.
Tomasello, M., Brooks, P., and Stern, E., 1998. Learning to produce passive sentences through discourse. First Language, 18. 223-237.
Traxler, M.J., and Pickering, M.J., 1996. Plausibility and the processing of unbounded dependencies: An eye-tracking study. Journal of Memory and Language, 35. 454-475
Trueswell, J.C., and Tanenhaus, M.K., 1994. Toward a lexicalist framework for constraint-based syntactic ambiguity resolution. In: C. Clifton, L. Frazier, and K. Rayner (eds.), Perspectives on sentence processing. Hillsdale, NJ: Lawrence Erlbaum Associates Inc. 155-179
Trueswell, J.C., Tanenhaus, M.K., Garnsey, S.M., 1994. Semantic influences on parsing: use of thematic role information in syntactic disambiguation. Journal of Memory and Language, 33. 285–318.
Tsimpli, I.M., 1989. On the properties of the passive affix in Modern Greek. UCL Working Papers in Linguistics, 1. 235–260.
Tsimpli, I. M., 2005. Voice in Greek: Description of the system and study of its development in Greek as a native and a second/foreign language [Η Φωνή στην Ελληνική: Περιγραφή του συστήµατος και µελέτη της ανάπτυξής του στην ελληνική ως µητρική και ως δεύτερη/ξένη γλώσσα]. In: Laskaratou, Ch., Georgiafentis, M., Tsimpli, I.M., Philippaki-Warburton, I., Spyropoulos, V., Tsamadou-Jacoberger, I., Cheila-Markopoulou, D. and Tsangalidis, A., (eds.). Syntax in the learning and teaching of Greek as a foreign language [Η σύνταξη
REFERENCES
369
στη µάθηση και στη διδασκαλία της ελληνικής ως ξένης γλώσσας]. Athens: Patakis. 62-116.
Tsimpli, I.M., 2006. The acquisition of voice and transitivity alternations in Greek as native and second language. In: S. Unsworth, T. Parodi, A. Sorace and M. Young-Scholten (eds.), Paths of Development in L1 and L2 acquisition: In honor of Bonnie D. Scwartz. Language Acquisition and Language Disorders, 39. John Benjamins Publishing Company. 15–55.
Tzartzanos, A., 1989. Neoelliniki Syndax (of the Koini Dimotiki). [Νεοελληνική Σύνταξη (της κοινής δηµοτικής)]. Vol. 1. Thessaloniki: Kyriakides
Tyler, L.K., Marslen-Wilson, W.D., 1977. The on-line effects of semantic context on syntactic processing. Journal of Verbal Learning and Verbal Behavior, 16. 683–692.
Valian, V., 1991. Syntactic subjects in the early speech of American and Italian children. Cognition, 40. 21-81
Van Hout, A., 1992. Linking and projection based on event structure. MS. Tilburg: Tilburg University.
Van Hout, A., 1996. Event Semantics of Verb Frame Alternation: A Case Study of Dutchand its Acquisition. Doctoral dissertation, Tilburg University
Van Hout, A., 2000. Event semantics in the lexicon- syntax interface: Verb frame alternation in Dutch and their acquisition. In: C. Tenny and J. Pustejovsky (eds.), Events as Grammatical Objects. Stanford, Calif.:CSLI. 239-82
Van Hout, A., 2004. Unaccusativity as Telicity checking. In: A. Alexiadou, E. Anagnostopoulou, M. Everaert (eds.), The Unaccusativity Puzzle: Explorations of the Syntax-Lexicon Interface. Oxford: Oxford University Press. 60-83
Vassilaki, S., 1988. The morphology of passive voice in Modern Greek [Η µορφολογία της παθητικής φωνής στα Νέα Ελληνικά]. In: Studies on Greek Linguistics. Aristotle University of Thessaloniki. 185-203.
Verrips, M., 2000. Passives and implicit arguments in child language. In: S.C. Howell, S.A. Fish, and T. Keith-Lucas (eds.), Proceedings of the 24th Annual Boston University Conference on Language Development. Somerville, MA: Cascadilla Press. 749–760.
Volk, M., 2002. Using the Web as Corpus for Linguistic Research. In: Renate Pajusalu and Tiit Hennoste (eds.), Tahendusepuudja. Catcher of the Meaning. A Festschrift for Professor Haldur Oim. Publications of the Department of General Linguistics 3. University of Tartu.
Warburton, I., 1970. On the Verb in Modern Greek Language Science Monographs, Volume 4. The Hague: Mouton, Bloomington, Indiana University.
REFERENCES
370
Warburton, I., 1975. The passive in English and Greek, in: Foundations of Language, 13. 563–578.
Wasow, T., 1977. Transformations and the lexicon. In: P. Culicover, T. Wasow, and A. Akmajian (eds.), Formal Syntax. New York: Academic Press. 327–360.
Wehrli, E., 1986. On some properties of French clitic se. In: H. Borer (ed.), The Syntax of Pronominal Clitics. New York: Academic Press. 263-283.
Wexler, K., 1998. Very early parameter setting and the unique checking constraint: A new explanation of the optional infinitive stage. Linqua,106.
White, L., 2009. Grammatical Theory: Interfaces and L2 Knowledge, Chapter 3. In: C. William, T. Ritchie, and K.Bhatia (eds.), The New Handbook of Second Language Acquisition. Emerald. 49-68.
Williams, E., 1981. Argument structure and morphology. The Linguistic Review, 1. 81-114.
Zevgoli, S., 2000. The Parameter of Reflexive Anaphors: A Comparative Study of the English and the Modern Greek Nominal Phrase. Unpublished Doctoral Dissertation, University of Cambridge.
Zombolou, K., 1996. Deponents in Greek: the formation of passive voice and diathesis, thematic roles and cognitive-semantic analysis [Τα αποθετικά ρήµατα της νέας ελληνικής: σχηµατισµός παθητικής φωνής και διάθεσης, θεµατικοί ρόλοι και γνωστική σηµασιολογική ανάλυση]. Studies in Greek Linguistics, 16. 160-171.
Zombolou, K., 1997. In quest of the agent: Deponents vs mediopassive diathesis. A different approach of the Greek voice system [Αναζητώντας το δράστη: Αποθετικά ρήµατα έναντι της µεσοπαθητικής διάθεσης. Μια άλλη προσέγγιση του ελληνικού ρηµατικού συστήµατος]. Studies in Greek Linguistics, 17. 229-242.
Zombolou, K., 2004. Verbal Alternations in Greek: a Semantic Analysis. Unpublished doctoral dissertation, The University of Reading.
Zombolou, K., Varlokosta, S., Alexiadou, A., and Anagnostopoulou, E., 2009. Acquiring Anticausatives vs. passives in Greek. Presentation in BUCLD 36, 6-8 November 2009
Appendices
APPENDICES
372
Appendix I
The experimental sentences of the on-line SPR and AJ Task
(ACT-NACT)
Version 1 Voice Non-Alternating Anti-causatives (Class I)
1. (a) I mathitria eklise afu teljosan I dhiagonismi.
The student close- ACT. 3 Sg. after finish-PAST. 3Pl the exams
“The student closed after the end of the exams.”
(b) I porta eklise afu teljosan I dhiagonismi.
The door close- ACT. 3 Sg. after finish-PAST. 3Pl the exams
“The door closed after the end of the exams.”
(c) I mathitria klistike afu teljosan I dhiagonismi.
The student close- NACT. 3 Sg. after finish-PAST. 3Pl the exams
“The student was closed after the end of the exams.”
(d) I porta klistike afu teljosan I dhiagonismi.
The door close- NACT. 3 Sg. after finish-PAST. 3Pl the exams
“The door was closed after the end of the exams.”
2. (a) I kori lijise prin o iljos dhisi
The daughter bend-ACT. 3 Sg. before the sun set-PRES. 3Sg
“The daughter bent before the sun set.”
(b) To kladhi lijise prin o iljos dhisi
The branch bend- ACT. 3 Sg. before the sun set-PRES. 3Sg
“The branch bent before the sun set.”
(c) I kori lijistike prin o iljos dhisi
The daughter bend- NACT. 3 Sg. before the sun set-PRES. 3Sg
“The daughter was bent before the sun set.”
(d) To kladhi lijistike prin o iljos dhisi
The branch bend- NACT. 3 Sg. before the sun set-PRES. 3Sg
“The branch was bent before the sun set.”
APPENDICES
373
3. (a) To koritsi ejire prin erthi I epitropi.
The girl lean- ACT. 3 Sg. before come-PRES. 3Sg the committee
“The girl leaned before the committee arrived.”
(b) To dhentro ejire prin erthi I epitropi.
The tree lean- ACT. 3 Sg. before come-PRES. 3Sg the committee
“The tree bowed before the committee arrived.”
(c) O koritsi jernete prin erthi I epitropi.
The girl lean- NACT. 3 Sg. before come-PRES. 3Sg the committee
“The girl is being leaned before the committee arrives.”
(d) O dhentro jernete prin erthi I epitropi.
The tree lean- NACT. 3 Sg. before come-PRES. 3Sg the committee
“The tree is being bowed before the committee arrives.”
4. (a) To eggoni stegnose an ke itan poli vregmeno.
The grandchild dry-ACT.3Sg. although be-PAST.3Sg very wet
“The grandchild dried although it was very wet.”
(b) To pandeloni stegnose an ke itan poli vregmeno.
The trousers dry- ACT.3Sg. although be-PAST.3Sg very wet
“The trousers dried although they were very wet.”
(c) To eggoni stegnothike an ke itan poli vregmeno.
The grandchild dry- NACT.3Sg. although be-PAST.3Sg very wet
“The grandchild was dried although it was very wet.”
(d) To pandeloni stegnothike an ke itan poli vregmeno.
The trousers dry- NACT.3Sg. although be-PAST.3Sg very wet
“The trousers were dried although they were very wet.”
5. (a) To pedhi eljose horis na to katalavo
The kid melt-ACT.3Sg without it understand-PRES.1Sg
“The kid melted before I knew.”
(b) To keri eljose horis na to katalavo
The candle-ACT.3Sg without it understand-PRES.1Sg
“The candle melted before I knew.”
(c) To pedhi ljonete horis na to katalavo
APPENDICES
374
The kid melt-NACT.3Sg without it understand-PRES.1Sg
“The kid is being melted before I know.”
(e) To keri ljonete horis na to katalavo
The candle melt-NACT. 3Sg without it understand-PRES. 1Sg
“The candle is being melted before I know.”
6. (a) O ergatis evrase prin pame sto saloni.
The worker boil- NACT.3Sg. before go-PAST.3Pl to the living room
“The worker boiled before we went to the living room.”
(b) O kafes evrase prin pame sto saloni.
The coffee boil- ACT.3Sg. before go-PAST.3Pl to the living room
“The coffee boiled before we went to the living room.”
(c) O ergatis vrazete prin pame sto saloni.
The worker boil- NACT.3Sg. before go-PAST.3Pl to the living room
“The worker is being boiled before we went to the living room.”
(d) O kafes vrazete prin pame sto saloni.
The coffee boil- NACT.3Sg. before go-PAST.3Pl to the living room
“The coffee is being boiled before we went to the living room.”
7. (a) O nearos sapise prin arhisun I zestes.
The young man rot- ACT.3Sg. before start-PRES.3Pl the warmths
“The young man decomposed before the big heat.”
(b) To fruto sapise prin arhisun I zestes.
The fruit got rot - ACT.3Sg. before start-PRES.3Pl the warmths
“The fruit got rot before the big heat.”
(c) O nearos sapizete prin arhisun I zestes.
The young man rot - NACT.3Sg. before star-PRES.3Pl the warmths
“The young man is being decomposed before the big heat.”
(d) To fruto sapizete prin arhisun I zestes.
The fruit rot - NACT.3Sg. before start-PRES.3Pl the warmths
“The fruit is being rotten before the big heat.”
APPENDICES
375
Voice Alternating Anti-causatives (Class II)
8. (a) O komotis berdhepse eno to htenisma telione.
The hairdresser mingle- ACT.3Sg. while the hairdressing finish-PAST.3Sg
“The hairdresser got mixed up while the hairdressing was almost finished.”
(b) O kotsos berdhepse eno to htenisma telione.
The ban mingle- ACT.3Sg. while the hairstyle finish-PAST.3Sg
“The ball got mixed up while the hairdressing was almost finished.”
(c) O komotis berdheftike eno to htenisma telione.
The hairdresser mingle- NACT.3Sg. while the hairstyle finish-PAST.3Sg
“The student got mixed up while the hairdressing was almost finished.”
(d) O kotsos berdheftike eno to htenisma telione.
The ban mingle- NACT.3Sg. while the hairstyle finish-PAST.3Sg
“The ban was mixed up while the hairdressing was almost finished.”
9. (a) To aghori tsalakose afu irthan I kalezmeni.
The boy crumple- ACT.3Sg. after come-PAST.3Pl the guests
“The boy crumpled after the guests arrived.”
(b) To ifasma tsalakose afu irthan I kalezmeni.
The cloth crumple- ACT.3Sg. after come-PAST.3Pl the guests
“The cloth crumpled after the guests arrived.”
(c) O aghori tsalakothike afu irthan I kalezmeni.
The boy crumple- NACT.3Sg. after come-PAST.3Pl the guests
“The boy was crumpled after the guests arrived.”
(d) O ifasma tsalakothike afu irthan I kalezmeni.
The cloth crumple- NACT.3Sg. after come-PAST.3Pl the guests
“The cloth was crumpled after the guests arrived.”
10. (a) I jineka lerose otan I jiorti teliose.
The woman spill- ACT.3Sg. when the party end-PAST.3Sg
“The woman spilled when the party was over.”
(b) To forema lerose otan I jiorti teliose.
The dress spill- ACT.3Sg. when the party end-PAST.3Sg
“The dress got spilled when the party was over.”
APPENDICES
376
(c) I jineka lerothike otan I jiorti teliose.
The woman spill- NACT.3Sg. when the party end-PAST.3Sg
“The woman was/got spilled when the party was over.”
(d) To forema lerothike otan I jiorti teliose.
The dress spill- NACT.3Sg. when the party end-PAST.3Sg
“The dress was spilled when the party was over.”
11. (a) O horeftis tendose an ke itan poli dhiskolo.
The dancer stretch- ACT.3Sg. although be- PAST.3Sg very difficult
“The dancer stretched although it was very difficult.”
(b) To lastiho tendose an ke itan poli dhiskolo.
The cord stretch- ACT.3Sg. although be -PAST.3Sg very difficult
“The cord stretched although it was very difficult.”
(c) O horeftis tendothike an ke itan poli dhiskolo.
The dancer stretch- NACT.3Sg. although be-PAST.3Sg very difficult
“The dancer was stretched / stretched himself although it was very difficult.”
(d) To lastiho tendothike an ke itan poli dhiskolo.
The cord stretch- NACT.3Sg. although be -PAST.3Sg very difficult
“The cord got stretched although it was very difficult.”
12. (a) I modhistra tripise molis stamatisa na kitao.
The dressmaker pierce- ACT.3Sg. as soon as stop- PAST.1Sg to look
“The dressmaker pierced as soon as I stopped looking.”
(b) To ifasma tripise molis stamatisa na kitao.
The cloth pierce- ACT.3Sg. as soon as stop -PAST.1Sg to look
“The cloth (got) pierced as soon as I stopped looking.”
(c) I modhistra tripithike molis stamatisa na kitao.
The dressmaker pierce- NACT.3Sg. as soon as stop-PAST.1Sg to look
“The dressmaker got pierced / pierced herself as soon as I stopped looking.”
(d) To ifasma tripithike molis stamatisa na kitao.
The cloth pierce- NACT.3Sg. as soon as stop-PAST.1Sg to look
“The cloth got/was pierced as soon as I stopped looking.”
APPENDICES
377
13. (a) To aghori katharise an ke itan poli vromiko.
The boy clean- ACT.3Sg. although be-PAST.3Sg very dirty
“The boy cleaned although it was very dirty.”
(b) To pukamiso katharise an ke itan poli vromiko.
The shirt clean- ACT.3Sg. although be-PAST.3Sg very dirty
“The shirt (got) cleaned up although it was very dirty.”
(c) To aghori katharistike an ke itan poli vromiko.
The boy clean- NACT.3Sg. although be-PAST.3Sg very dirty
“The boy was cleaned /cleaned himself up although it was very dirty.”
(d) To pukamiso katharistike an ke itan poli vromiko.
The shirt clean- NACT.3 Sg. although be-PAST.3Sg very dirty
“The shirt got/was cleaned up although it was very dirty.”
14. (a) I beba htipise an ke oli prosehame poli.
The baby girl hit- ACT.3 Sg. although all watch-PAST.1Pl a lot
“The baby girl hit though we were watching (her) carefully.”
(b) I kabana htipise an ke oli prosehame poli.
The bell hit- ACT.3Sg. although all watch-PAST.1Pl a lot
“The bell rung though we were watching it carefully.”
(c) I beba htipithike an ke oli prosehame poli.
The baby girl hit- NACT.3Sg. although all watch-PAST.1Pl a lot
“The baby girl was hurt though we were watching (her) carefully.”
(d) I kabana htipithike an ke oli prosehame poli.
The bell hit- NACT.3Sg. although all watch-PAST.1Pl a lot
“The bell was hit though we were watching it carefully.”
APPENDICES
378
Version 2 Voice Non-Alternating Anti-causatives (Class I)
15. (a) O jitonas eklise an ke emis dhe thelame.
The neighbor close- ACT.3Sg. although want-PAST.1Pl not
“The neighbor closed despite our will.”
(b) To parathiro eklise an ke emis dhe thelame.
The window close- ACT.3Sg. although want-PAST.1Pl not
“The window closed despite our will.”
(c) O jitonas klistike an ke emis dhe thelame.
The neighbor close- NACT.3Sg. although want-PAST.1Pl not
“The neighbor was closed despite our will.”
(d) To parathiro klistike an ke emis dhe thelame.
The window close- NACT.3Sg. although want-PAST.1Pl not
“The window was closed despite our will.”
16. (a) O fandaros lijise horis na prospathiso poli.
The soldier bend- ACT.3Sg. without try-PRES.1Sg much
“The soldier bent no effort needed.”
(b) I alisidha lijise horis na prospathiso poli.
The chain bend- ACT.3Sg. without try-PRES.1Sg much
“The chain bent no effort needed.”
(c) O fandaros lijistike horis na prospathiso poli.
The soldier bent- NACT.3Sg. without try-PRES.1Sg much
“The soldier was bent no effort needed.”
(d) I alisidha lijistike horis na prospathiso poli.
The chain bent- NACT.3Sg. without try-PRES.1Sg much
“The chain was bent no effort needed.”
17. (a) I neari ejire prin pesi I nihta.
The young girl lean- ACT.3Sg. before fall-PRES.3Sg the night
“The girl leaned before the night fell.”
(b) To luludhi ejire prin pesi I nihta..
The flower lean- ACT.3Sg. before fall-PRES.3Sg the night
APPENDICES
379
“The flower bowed before the night fell.”
(c) I neari jernete prin pesi I nihta..
The young girl lean-NACT.3Sg. before fall-PRES.3Sg the night
“The young girl is being leaned before the night falls.”
(d) To luludhi jernete prin pesi I nihta.
The flower lean- NACT.3Sg. before fall-PRES.3Sg the night
“The flower is being bowed before the night falls.”
18. (a) I ghata stegnose prin vji o iljos.
The cat dry-ACT.3Sg. before go-PRES.3Sg out the sun
“The cat dried before the sun came”
(b) I bluza stegnose prin vji o iljos.
The T-shirt dry- ACT.3Sg. before go-PRES.3Sg out the sun
“The T-shirt dried before the sun came”
(c) I ghata stegnothike prin vji o iljos.
The cat dry- NACT.3Sg. before go-PRES.3Sg out the sun
“The cat was dried before the sun came”
(d) I bluza stegnothike prin vji o iljos.
The T-shirt dry- NACT.3Sg. before go-PRES.3Sg out the sun
“The T-shirt was dried before the sun came”
19. (a) To aghori eljose an ke ehi poli krio.
The boy melt- ACT.3Sg. although have-PRES.3Sg very cold
“The boy melted although it is very cold.”
(b) To vutiro eljose an ke ehi poli krio.
The butter melt- ACT.3Sg. although have-PRES.3Sg very cold
“The butter melted although it is very cold.”
(c) To aghori ljonete an ke ehi poli krio.
The boy melt- NACT.3Sg. although have-PRES.3Sg very cold
“The boy is being melted although it is very cold.”
(d) To vutiro ljonete an ke ehi poli krio.
The butter melt- NACT.3Sg. although have-PRES.3Sg very cold
“The butter is being melted although it is very cold.”
APPENDICES
380
20. (a) To pedhi evrase prin ftasi o pateras.
The boy boil- NACT.3Sg. before arrive-PAST.3Sg the dud
“The boy boiled before dud arrived”
(b) To avgho evrase prin ftasi o pateras.
The egg boil- ACT.3Sg. before arrive-PAST.3Sg the dud
“The egg was ready before dud arrived”
(c) To pedhi vrazete prin ftasi o pateras.
The boy boil- NACT.3Sg. before arrive-PAST.3Sg the dud
“The boy is being boiled before dud arrives”
(d) To avgho vrazete prin ftasi o pateras.
The egg boil- NACT.3Sg. before arrive-PAST.3Sg the dud
“The egg is being cooked before dud arrives”
21. (a) O andras sapise prin analithun I eksetasis.
The man rot - ACT.3Sg before resolve-PRES.3Pl the exams
“The man decomposed before we got the results of the exams.”
(b) To kreas sapise prin analithun I eksetasis.
The meat rot - ACT.3Sg before resolve-PRES.3Pl the exams
“The meat got rotten before we got the results of the exams.”
(c) O andras sapizete prin analithun I eksetasis.
The man rot - NACT.3Sg. before resolve-PRES.3Pl the exams
“The man is being decomposed before we get the results of the exams.”
(d) To kreas sapizete prin analithun I eksetasis.
The meat rot - NACT.3Sg before resolve-PRES.3Pl the exams
“The meat is being decomposed before we get the results of the exams.”
APPENDICES
381
Voice Alternating Anti-causatives (Class II)
22. (a) O fititis berdhepse eno imun poli prosektiki.
The student mingle- ACT.3Sg. while be-PAST.1Sg very careful
“The student mixed up while I were very careful.”
(b) To kuvari berdhepse eno imun poli prosektiki.
The ball mingle- ACT.3Sg. while be -PAST.1Sg very careful
“The ball got mixed up while I were very careful.”
(c) O fititis berdheftike eno imun poli prosektiki.
The student mingle- NACT.3Sg. while be -PAST.1Sg very careful
“The student got mixed up while I were very careful.”
(d) To kuvari berdheftike eno imun poli prosektiki.
The ball mingle- NACT.3Sg. while be -PAST.1Sg very careful
“The ball got mixed up while I were very careful.”
23. (a) I kopela tsalakose otan epeze to radhio.
The girl crumple- ACT.3Sg. when play-PAST.3Sg the radio
“The girl crumpled while the radio was on.”
(b) I fusta tsalakose otan epeze to radhio.
The skirt crumple- ACT.3Sg. when play-PAST.3Sg the radio
“The skirt crumpled while the radio was on.”
(c) I kopela tsalakothike otan epeze to radhio.
The girl crumple- NACT.3Sg. when play-PAST.3Sg the radio
“The girl was crumpled while the radio was on.”
(d) I fusta tsalakothike otan epeze to radhio.
The skirt crumple- NACT.3Sg. when play-PAST.3Sg the radio
“The skirt was crumpled while the radio was on.”
24. (a) O athlitis lerose prin teliosi o aghonas.
The sportsman spill- ACT.3Sg. before end-PAST.3Sg the game
“The sportsman dirtied before the end of the game.”
(b) I fanela lerose prin teliosi o aghonas.
The T-shirt spill- ACT.3Sg. before end-PAST.3Sg the game
“The T-shirt got dirty before the end of the game.”
APPENDICES
382
(c) O athlitis lerothike prin teliosi o aghonas.
The sportsman spill- NACT.3Sg. before end-PAST.3Sg the game
“The sportsman got/was dirty before the end of the game.”
(d) I fanela lerothike prin teliosi o aghonas.
The T-shirt spill- NACT.3Sg. before end-PAST.3Sg the game
“The T-shirt got dirty before the end of the game.”
25. (a) O papus tendose otan bike I mitera.
The grandfather stretch- ACT.3Sg. when enter- PAST.3Sg the mother
“The grandfather stretched when mother got into the room.”
(b) To shini tendose otan bike I mitera.
The cord stretch- ACT.3Sg. when enter- PAST.3Sg the mother
“The cord stretched when mother got into the room.”
(c) O papus tendothike otan bike I mitera.
The grandfather stretch- NACT.3Sg. when enter- PAST.3Sg the mother
“The grandfather stretched himself when mother got into the room.”
(d) To shini tendothike otan bike I mitera..
The cord stretch- NACT.3Sg. when enter-PAST.3Sg the mother
“The cord got/was stretched when mother got into the room.”
26. (a) I nosokoma tripise prin jini I eghirisi.
The nurse pierce- ACT.3Sg. before become- PAST.3Sg the surgery
“The nurse pierced before the end of the surgery.”
(b) To sendoni tripise prin jini I eghirisi.
The sheet pierce- ACT.3Sg. before become – PAST.3Sg the surgery
“The sheet (got) pierced before the end of the surgery.”
(c) I nosokoma tripithike prin jini I eghirisi.
The nurse pierce- NACT.3Sg. before become - PAST.3Sg the surgery
“The nurse got pierced/pierced herself before the end of the surgery.”
(d) To sendoni tripithike prin jini I eghirisi.
The sheet pierce- NACT.3Sg. before become - PAST.3Sg the surgery
“The sheet was/got pierced before the end of the surgery.”
APPENDICES
383
27. (a) I kiria katharise kathos itan poli vromiki.
The lady clean- ACT.3Sg. since be-PAST.3Sg very dirty
“The lady cleaned as (she) was very dirty.”
(b) I kurtina katharise kathos itan poli vromiki.
The curtain clean- ACT.3Sg. since be-PAST.3Sg very dirty
“The curtain (got) cleaned up as it was very dirty.”
(c) I kiria katharistike kathos itan poli vromiki.
The lady clean- NACT.3Sg. since be-PAST.3Sg very dirty
“The lady was cleaned / cleaned herself up as (she) was very dirty.”
(d) I kurtina katharistike kathos itan poli vromiki.
The curtain clean- NACT.3Sg. since be-PAST.3Sg very dirty
“The curtain was/got cleaned up as it was very dirty.”
(b) The same set of pictures was used to control an utterance with a verb in ACT
where the only possible reading is the transitive.
(i) Transitive: Χτένιζε τα µαλλιά της αλογοουρά.
‘She combed her hair in ponytail. (ILSP: 1061555)
(IV) NACT activity predicates
(a) (i) Anti-causative: KATΑ ΒΑΘΟς, πίσω από το ατρόµητο, τάχα, προσωπείο του
«αδιάφθορου ερευνητή», κρύβεται ένα πολύ φοβισµένο ανθρωπάκι.
‘DEEP INSIDE, behind the fearless face of the ‘incorruptible researcher’, a
frightened person is hidden’. (ILSP: 1613341)
(ii) Passive: Σε θρίλερ, µε πρωταγωνιστές που δεν κατονοµάστηκαν αλλά
«κρύφτηκαν» πίσω από το γνωστό τίτλο «διαπλεκόµενα», µε καταγγελίες για τη
APPENDICES
391
µετατροπή της Ελλάδας, από την κυβέρνηση, σε χώρα του τζόγου, αλλά και µε
απειλές ακόµη και για παραποµπές στο Ειδικό ∆ικαστήριο, εξελίχθηκε η συζήτηση
διάταξης του φορολογικού νοµοσχεδίου που η Ν.∆. θεώρησε ότι αφορά το
βιντεολόττο και η κυβέρνηση διέρρηξε τα ιµάτιά της περί του αντιθέτου, χωρίς να
απαντήσει στην ουσία του θέµατος.
‘A horror film, with leading actors people who were not named but “hidden”
behind the well known title of “conflicting interests” with charges concerning the
transformation of Greece, by the Government, to a country of gamble, as well as
with threats even of committal to the Special Court, turned out to be the discussion
on the taxation bill, which the Party of ‘Nea Dimokratia’ thought relevant to
‘videolotto’ while the Government fought for the contrary, without really replying
to the matter.
(ILSP: 1664415)
(iii) Reflexive: Με το αυτοκίνητό του, µια µπλε BMW, έκλεισε το δρόµο στο
αυτοκίνητο της Παναγιώτας και κρύφτηκε πίσω από τους θάµνους.
‘He blocked the way to Panayota’s car with his car, a blue BMW, and hid behind
the bushes’. (ILSP: 599998)
(b) (i) Anti-causative: Το µικρό αυτοκίνητο στάθηκε µπροστά στο περίπτερο και
κρύφτηκε εντελώς.
The small car stood in front of the kiosk and got completely hidden’. (ILSP:
1052806)
(ii) Passive: Μήπως λησµόνησε η κυβέρνηση ότι το σκάνδαλο Κοσκωτά κρύφτηκε
πίσω από το διαβόητο απόρρητο των τραπεζικών καταθέσεων;
‘Did the Government forget that the Koskotas’ scandal was hidden behind the
notorious deposits secrecy?’ (ILSP: 1204233)
(iii) Transitive (only in ACT): ∆ιασχίζουµε ανηφορικό παγωµένο τµήµα του, που
κρύβει άπειρες παγίδες.
‘We cross its uphill glassy part, which hides numerous traps.’ (ILSP: 245938)
APPENDICES
392
Appendix IV Some extractions of the ‘Web corpus’ The complete Google corpus used for this research as well as the data from the ILSP corpus are available from the webpage of the Language Development Lab, School of English, Aristotle University of Thessaloniki: www.enl.auth.gr/langlab.