Acquisitional patterns of Spanish anticausative se: The end of the road Inmaculada Gómez Soler University of Memphis Abstract The acquisition of the Spanish morpheme se has proved to be problematic for L2 learners both because of its polyfunctionality and because of the restrictions regarding the types of predicates with which it can combine. This paper sheds light on this problem by focusing on a specific type of se (anticausative se; e.g., El jarrón se rompió ‘The vase broke’) and exploring its acquisition across four proficiency levels. Results of a scalar grammaticality judgment task indicate that lower-proficiency participants’ performance is in line with previous research, which claims that this aspect of Spanish grammar is particularly challenging for L2 learners (as reflected in omission and overgeneralization errors). However, the near-native group shows sensitivity to the abstract features that uniquely characterize verbs that undergo the causative/inchoative alternation. Thus, the current findings suggest that L2 learners manage to overcome the problems experienced at lower levels and, in fact, do succeed at the level of ultimate attainment. 1. Introduction One of the indisputable differences between first and second language acquisition is the success rate experienced by learners; that is, while L1 learners always develop native command of all language properties (except in the case of a language disorder), such a state is not guaranteed for L2 learners (VanPatten & Williams, 2008, among others). On the contrary, L2 learners tend to show different degrees of proficiency at the state of ultimate attainment (e.g., Birdsong, 2006). For those looking at this issue from a
48
Embed
Acquisitional patterns of Spanish anticausative se: The end of the road
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
Acquisitional patterns of Spanish anticausative se: The end of the road
Inmaculada Gómez Soler
University of Memphis Abstract
The acquisition of the Spanish morpheme se has proved to be problematic for L2 learners
both because of its polyfunctionality and because of the restrictions regarding the types of
predicates with which it can combine. This paper sheds light on this problem by focusing
on a specific type of se (anticausative se; e.g., El jarrón se rompió ‘The vase broke’) and
exploring its acquisition across four proficiency levels. Results of a scalar grammaticality
judgment task indicate that lower-proficiency participants’ performance is in line with
previous research, which claims that this aspect of Spanish grammar is particularly
challenging for L2 learners (as reflected in omission and overgeneralization errors).
However, the near-native group shows sensitivity to the abstract features that uniquely
characterize verbs that undergo the causative/inchoative alternation. Thus, the current
findings suggest that L2 learners manage to overcome the problems experienced at lower
levels and, in fact, do succeed at the level of ultimate attainment.
1. Introduction
One of the indisputable differences between first and second language acquisition
is the success rate experienced by learners; that is, while L1 learners always develop
native command of all language properties (except in the case of a language disorder),
such a state is not guaranteed for L2 learners (VanPatten & Williams, 2008, among
others). On the contrary, L2 learners tend to show different degrees of proficiency at the
state of ultimate attainment (e.g., Birdsong, 2006). For those looking at this issue from a
generative perspective, the problem does not lie on L2 learners’ performance at this
advanced level but on how linguistic properties are represented at this stage, whether
these representations do in fact correspond to native representations and whether they can
be considered to be UG-constrained. The study of near-native speakers is particularly
relevant for generativists because, as Sorace (2003, p.130) claims “the competence of
near-native speakers is […] more revealing of UG constraints on L2 acquisition than
those of other stages.” Given that these learners have reached a steady state, any violation
of UG needs to be considered permanent and can not be attributed to the need of further
development, more input or more time of exposure. However, the issue of competence at
the level of ultimate attainment is a matter of debate. For Sorace (2003) optionality is a
characteristic of non-native grammars that can in fact persist to the level of ultimate
attainment (i.e., residual optionality). Optionality surfaces when two forms exist as
alternatives in the non-native grammar. However, these options are not in free variation,
that is, the target option is preferred but the non-target option can still surface
sporadically. On the other hand, there is also evidence that advanced learners can acquire
properties to a native-level, even those deemed more prone to optionality like properties
at the syntax-pragmatics interface (e.g., Rothman, 2009).
This article explores the issue of ultimate attainment by studying anticausative se in
an understudied population (i.e., near-native speakers). Se is one of the most
polyfunctional morphemes in Spanish since it encodes multiple meanings: reflexive (1),
reciprocal (2), passive, and impersonal, among others.
(1) Julia se lava
Julia se wash-3sg.
Julia washed herself
(2) Julia y Alejandro se abrazan
Julia and Alejandro se hug-3pl.
Julia and Alejandro hug each other
This article focuses on a specific use of se, its anticausative interpretation, and its
distribution with unaccusative (e.g., llegar ‘arrive’) and unergative (e.g., hablar ‘to talk’)
predicates. Anticausative se is a valency-reducing morpheme, which has the function of
suppressing the external argument (Burzio, 1986; Cinque, 1988; Grimshaw, 1990;
Reinhart & Siloni, 2005; Koontz-Garboden, 2009). This morpheme participates in the
causative/inchoative alternation by deriving the intransitive version (3) of a transitive
construction (4):
(3) El jarrón se rompió
The vase se broke
The vase broke
(4) María rompió el jarrón
María broke-3sg. the vase
María broke the vase
However, this morpheme cannot co-occur with certain types of unaccusatives (5) or
unergatives (6):
(5) *Mi hermana se llegó
Mi sister se arrived-3sg.
Mi sister arrived
(6) *María se caminó
María se walked
María walked
Although there are universal semantic restrictions on the causative/inchoative
alternation (see section 2.2.1 for a detailed explanation), the way it is encoded varies
from language to language both in the morphology used to express the distinction (e.g.,
the inchoative construction needs the use of se in Spanish while in English it does not
necessitate any additional morphology) and in the group of predicates that participate in it
(e.g., manner-of-motion verbs in English can have a causative interpretation (7) while
they cannot in Spanish (8), examples taken from Montrul (2001b, p. 174)).
(7) English:
a. The soldiers marched
b. The captain marched the soldiers to the tent
(8) Spanish
a. Los soldados marcharon
The soldiers marched-3pl.
The soldiers marched
b. *El capitán marchó a los soldados hasta el campamento
The captain marched-3sg. to the soldiers to the tents
The captain marched to the soldiers to the tent
Thus, acquiring the distribution of anticausative se with unergative and unaccusative
predicates is a complex puzzle in which the learner needs to (i) exhibit sensitivity to the
semantic nuances of verbs that are relevant for the distinction (i.e., only change-of-state
unaccusatives alternate in transitivity, section 2.2) and (ii) realize how the distinction is
marked morphologically in Spanish (se). In order to use se in a native-like manner, the
L2 learner needs to eventually develop a rule for se as an argument absorber whose
interpretation relies on the verbs’ meaning components.
This learning puzzle represents a poverty-of-the-stimulus problem (Chomsky, 1965,
1981; Pinker, 1994; Schwartz, 1998; Rothman & Iverson, 2008) since what the learner
needs to learn is underdetermined in the input for two reasons: (i) there is a one-to-many
correspondence between the morpheme se and its multiple meanings, (ii) there is
insufficient evidence to distinguish between unergatives and unaccusatives given that
there are few reliable tests for unaccusativity in Spanish (Contreras, 1978; Torrego, 1989;
de Miguel, 1992; Aranovich, 2000) and that the use of se actually only applies to a
restricted set of unaccusatives.
Previous studies have found both errors of omission of se with change-of-state
unaccusatives and overextension of se to either unaccusatives, unergatives or both. This
study explores these issues including an understudied group of learners, near-native
speakers. This way, this article will add to the literature by exploring the developmental
patterns found in the acquisition of anticausative se while, at the same time, exploring the
issue of representation in end-state grammars.
This article has the following structure. Section 2 offers background information on
the topics of unaccusativity, transitivity alternations in Spanish and the role of se in this
type of construction. It also presents a review of the L2 literature on transitivity
alternations in Spanish. Section 3 presents the motivation for the study and the research
questions. Section 4 includes a description of the participants, methodology and data
analysis. Section 5 provides a detailed discussion of the results of the study in relation to
the findings of previous work on transitivity alternations. The final section offers some
concluding remarks.
2. Background
2.1 The Unaccusative Hypothesis
According to the Unaccusative Hypothesis (Perlmutter, 1978) there are two types of
intransitive verbs: unergatives and unaccusatives. Unergatives take a theta-marked deep-
structure subject. This NP subject causes the action denoted by the verb (9).
(9) yo [VPcorro]
I run-1sg.
I run
Unaccusatives take a theta-marked deep-structure object. Unlike the NP argument of
unergatives, the NP argument of unaccusatives undergoes the action denoted by the verb
(10).
(10) [VP llego yo]
I arrive-1sg.
I arrive
The exact nature of this distinction is a matter of debate. For some, the distinction is
purely semantic (Dowty, 1991; Van Valin, 1990) and is related to notions of agentivity
and telicity: unergatives are agentive and atelic (activities) while unaccusatives are non-
agentive and telic (achievements). For others, the distinction is syntactic in nature
(Burzio, 1986; Rosen, 1984) and they highlight the fact that unergatives have an external
argument while the sole argument of an unaccusative predicate is internal. Yet there are
others that believe that this distinction cannot be appropriately accounted for without an
understanding of the interplay between syntax and semantics within these predicates
(Levin & Rappaport Hovav, 1995; Sorace, 2000) since the linking rules that map
arguments into the syntax rely on semantic notions like change (mapped to an internal
argument) or immediate cause (mapped to an external argument).
Following this idea of the interface between syntax and semantics in the
unergative/unaccusative distinction, Sorace (2000) proposed the Unaccusative Hierarchy1
(11).
According to Sorace, unergative and unaccusative predicates do not always have a
consistent syntactic behavior. On the contrary, certain types (core, those in the extremes
of the hierarchy) present a constant behavior while the peripheral ones (those in the
middle of the hierarchy) vary both within a specific language and crosslinguistically.
Peripheral predicates are sensitive to the aspectual elements of the sentence in which they 1 The Unaccusative Hierarchy only includes monadic verbs. Dyadic verbs such as paired/alternating unaccusatives do not appear in the hierarchy. There are two types of paired unaccusatives (Sorace, 1993), those which have a transitive alternant (‘to break’) and those that have an unergative alternant (‘to run’ since it can indicate both change of location and controlled process (motional)). This article focuses on the first class, paired unaccusatives with a transitive alternant. According to Sorace (2003), the behavior of this class is controversial because, although it indicates a change of state (telicity), the notion of causation is also relevant. Consequently, these predicates exhibit unstable behavior within particular languages and also cross-linguistically.
(11) Change of location [directed motion]
Unaccusative (least variation) selects BE
Change of state Uncontrolled process Continuation of a preexisting
state
Existence of a state variable behavior Uncontrolled process
[emission] [involuntary reaction]
Controlled processes (motional)
Controlled processes (nonmotional)
Unergative (least variation) selects HAVE
appear (such as notions of telicity and agentivity), which explains their variable behavior.
In particular, Sorace (2000) studied the behavior of unergative and unaccusative verbs
with respect to auxiliary selection in Italian for the formation of the present perfect. What
she found is that the selection of be and have as the preferred auxiliary is clear for core
unaccusatives and unergatives; with core unaccusatives always selecting be and core
unergatives always selecting have. However, auxiliary selection is considerably more
variable for those predicates towards the middle of the hierarchy, with the class
‘existence of a state’ exhibiting the most variable behavior of all. This seems to be the
case also cross-linguistically (e.g., German, Dutch, French).
2.2. Transitivity alternations in Spanish
The specific property tested in this study is the distribution of anticausative se with
unaccusative and unergative predicates. Se is a valency-reducing morpheme, which has
the function of suppressing the external argument. Although se has a multiplicity of
functions (section 2.3) here I focus on its role in the transitive-inchoative alternation. Not
all verbs alternate in transitivity. Specifically, only verbs that express change of state or
location participate in this alternation. In Spanish, the transitive version is not marked by
any morpheme, that is, it exhibits zero morphology (12). However, the inchoative
meaning is derived through the addition of the anticausative morpheme se (13).
(12) María rompió el vaso
María broke-3sg. the glass
María broke the glass
(13) El vaso se rompió
The glass se broke-3sg.
The glass broke
This morpheme, however, cannot appear with an inchoative meaning with non-
alternating unaccusatives such as verbs of directed motion (14) or verbs of appearance
(15).
(14) *Mi hermana se llegó
Mi sister se arrived-3sg.
Mi sister arrived
(15) *El terremoto se ocurrió a las 6
The earthquake se happened-3sg. at 6
The earthquake happened at 6
In this type of sentence, there is not an external cause that can be suppressed by se as
in the case of alternating unaccusatives like in (13) where se can be said to replace María,
the external causer. In the same way, these sentences do not have a transitive counterpart
(16-17).
(16) *Mi hermana llegó a María
Mi sister arrived-3sg. to María
(intended meaning: My sister caused María to arrive)
(17) *La naturaleza ocurrió el terremoto
The nature happened -3sg.the earthquake.
(intended meaning: Nature caused the earthquake to happen)
Additionally, unergatives cannot co-occur with se with an inchoative meaning
(18). And, as was the case for non-alternating unaccusatives, they also lack a transitive
version (19). This is because, although unergatives include a concept of causation, their
cause is internal rather than external like for alternating unaccusatives.
(18) *María se caminó
María se walked
María walked
(19) *María caminó al perro al parque
María walked to the dog to the park
María walked the dog to the park
2.2.1 Theoretical assumptions
In order to operationalize these concepts, I follow Reinhart’s (2000, 2002) and
Reinhart & Siloni’s (2005) feature-based approach to the theta system and
decausativization proposal. Reinhart (2002, p. 229) defines the theta system as “the
system enabling the interface between the systems of concepts and the computational
system (syntax) and, indirectly (via the syntactic representations), with the semantic
inference systems.” The theta system has two basic binary features: +/- c=Cause change
and +/-m=Mental state. The first one indicates agentivity and causation while the second
determines whether an argument is animate or inanimate. The combination of these
features creates the feature clusters presented in Table 1, which broadly correspond to
traditional labels for theta roles. It is important to notice that there is not a one-to-one
correspondence between feature clusters and θ-roles but some clusters have varying
interpretation depending on context. This is crucial for the phenomenon under
investigation in this study (the causative/inchoative alternation).
Additionally, the combination of different clusters determines different verb
classes that exhibit distinct syntactic behavior as presented in Table 2.
Table 2. Verb classes as described by Reinhart’s (2002) Verb classes Argument structure
Non-alternating transitive Eat
V ([+c+m], [-c+m])
Alternating transitive break
V ([+c], [-c-m])
Experiencer Worry
V ([+c], [-c+m])
Unaccusative break/arrive
V [-c-m]
Unergative Walk
V [+c+m]
The key difference between transitive verbs with an inchoative version (break, worry)
and those which lack this version (eat) is that, while for the former the external argument
is unspecified (it could be an unspecified cause, an agent, or an instrument) (20-21), for
the latter the external argument has a fixed interpretation (agent) (22). Thus, the
difference relies on the fact that verbs that alternate in transitivity have an external
argument that is unspecified with respect to the mental state of the argument ([+c], they
do not have a +/-m feature), which is what allows the verb to have a cause, agent or
instrument as the subject. On the other hand, verbs that do not alternate in transitivity
have an external argument ([+c, +m]) that needs to be human/animate given its +m
feature.
(20) The wind /Max /the key opened the door
(21) Max / the noise / the gun scared Lucie
(22) The baby/ *the spoon /* hunger ate the soup
The next issue that needs to be resolved is how the theta system maps from the
lexicon to the syntax. In (23-25) I present Reinhart’s (2002) operationalization of this
matter. One thing to take into account is that, following Williams (1981), she assumes
that the mapping instructions are built into the lexical entry by indices (1 marks an
external role, 2 an internal one).
(23) Lexicon marking:
Given an n-place verb-entry, n>1,
a. Mark a [-] cluster with index 2.
b. Mark a [+] cluster with index 1.
c. If the entry includes both a [+] cluster and a fully specified cluster [/α,/-c],
mark the verb with the ACC feature.
(24) Relevant generalizations of lexical operations:
a. Saturation and reduction apply to the marked entry (i.e., after marking).
b. Reduction eliminates the accusative feature of the verb (fully or partially).
(25) CS merging instructions.
a. When nothing rules this out, merge externally.
b. An argument realizing a cluster marked 2 merges internally; An argument
with a cluster marked 1 merges externally.
First, I present how unergatives are mapped onto the syntax following Reinhart’s
(2002). According to this system, the sole argument of an unergative verb [+c+m] would
merge externally following (25a) given that (23) only applies to entries with more than
one argument. Second, I present how the transitive/inchoative alternation is
operationalized within this system. Following Chierchia (1989), Reinhart (2000, 2002)
argues that the inchoative version is derived from the transitive one. The derivation starts
with a lexical entry such as (26a) which corresponds to a verb that alternates in
transitivity (e.g., break, V ([+c], [-c-m])). According to (23) the [+] cluster ([+c]) is
marked with a 1 and the [-] cluster ([-c-m]) with a 2. Then, the inchoative version is
derived through a reduction operation (expletivization (Reinhart, 2002), renamed
decausativization in Reinhart & Siloni (2005)) by which the external role gets suppressed
(26b). For Reinhart (2002) and Reinhart & Siloni (2005), the eliminated role is not
available in the syntactic structure or interpretation of the derived inchoative version (cf.
Koontz-Garboden (2009) for an opposing view).
(26) Expletivization: Reduction of an external [+c] role (semantically null
function)
a. Vacc (θ1[+c], θ2)à Re(V) (θ2)
b. Re(V) (θ2) = V(θ2)
This operation takes places in all verb classes with a [+c] argument. Thus, it applies
to transitive alternators like break as well as experiencer verbs. However, the derived
entries are quite different, (27) being unaccusative and (28) being unergative.
(27) openacc ([+c], [-c-m]à Re(open)[-c-m]
a. Max opened the door
b. The door opened
(28) worryacc ([+c], [-c+m] à Re(worry)[-c+m]
a. The doctor worried Max
b. Max worried
The remaining question is how basic unaccusatives (non-alternating unaccusatives,
e.g., arrive) are mapped into the syntax in this system. Reinhart (2000, p. 23) assumes,
following Chierchia (1989), that unaccusatives without a transitive counterpart are
derived from some abstract transitive verb.2 The crucial evidence for Reinhart’s (2000, p.
22-23) approach comes from the existence of causative versions of these verbs
crosslinguistically:
come and die don't have alternates in English. However, the Hebrew verb for bring is the
transitive alternate of come, with the same stem but a different verbal morphology (hevi
(brought)/ ba (come)). Same is true for die (met (died) /hemit (killed)). Chierchia notes
that grow, which in English has both entries, has only the unaccusative entry in Italian
(crescere).
In this section I have shown that the feature [+c] is key for transitivity alternations
cross-linguistically. Consequently, access to this feature will define L2 learner success in 2 Levin & Rappaport-Hovav’s (1995) solution to this problem is quite different: they claim this set of verbs (basic unaccusatives) is listed in the lexicon since their lexical entry is frozen as a reduced form.
the learning task as outlined in the next section.
2.3. The learnability problem
The problem faced by the L2 learner when trying to acquire anticausative se is
extremely complex for several reasons. First, the learner needs to exhibit sensitivity to the
semantic nuances of verbs that are relevant for the distinction. In particular, as shown in
the previous section, learners need to restrict transitivity alternations only to those verbs
with a [+c] argument. Extracting this type of information from the input seems unlikely
since learners would have to notice the lack of a transitive alternative for basic
unaccusatives (29) and unergatives (30). Thus, the learner would not count with positive
L2 evidence in order to restrict his grammar. Instead, he will have to rely on noticing the
absence of an ungrammatical form, in this case, the transitive version of unergatives and
unaccusatives.
(29) *Mi hermana llegó a María
Mi sister arrived-3sg. to María
(intended meaning: My sister caused María to arrive)
(30) *María caminó al perro al parque
María walked to the dog to the park
María walked the dog to the park
Second, even when the learner is sensitive to the abstract [+c] feature that
distinguishes predicates that alternate in transitivity, he still needs to ascertain how this
distinction is marked morphologically in the L2. This poses an additional complication
having to do with the multiplicity of functions of se in Spanish (Kempchinsky, 2004;
King & Suñer, 2007). Se is a morpheme whose function is related to argument structure
alternations. However, se’s specific meaning and role changes depending on the type of
predicate that it modifies and the thematic roles involved in the construction. In (31-36) I
present some of the functions of se in Spanish as described by Toth (2000, p. 179-180)
(examples are my own):
(31) Reflexive se (can be used with transitive verbs and alternators)
Ana se lava el pelo por la mañana
Ana se washes the hair for the morning
Ana washes her hair in the morning
(32) Reciprocal se (can be used with transitive verbs and alternators)
Ana y Sofía se saludaron durante el banquete
Ana and Sofía se greeted during the banquet
Ana and Sofía greeted each other during the banquet
(33) Passive se (can be used with transitive verbs and alternators)
Se alquilan apartamentos en la playa
Se rent-3pl. apartments in the beach
Apartments are rented at the beach
(34) Impersonal se (can be used with all types of verbs)
Se vive mejor en España
Se live-3sg. better in Spain
One lives better in Spain
(35) Anticausative se (can be used with alternators)
Se rompió el vaso
Se broke-3sg. the glass
The glass broke
(36) Verb of emotion se (can be used with causative psych-verbs)
Marina se enfadó con su hermana
Marina se got mad with her sister
Marina got mad at her sister
Zyzik (2006) includes some additional functions, which are somewhat more restricted
in scope:
(37) Intransitive dynamic se (to indicate a change of meaning with certain
intransitives)
El coche se salió de la carretera
The car se go out of the road
The car went off the road
(38) Aspectual se (to indicate telicity with transitive verbs)
Se tomaron todo el vino
Se take-3pl. all the wine
They drank up all the wine
(39) Inherent (with a small class of inherent reflexive verbs)
Luis se queja de la tarea
Luis se complain-3sg. about the homework
Luis complains about the homework
The problem for the L2 learner is that, although non-alternating unaccusatives and
unergatives cannot co-occur with anticausative se, they can in fact co-occur with other
types of se. Example (40) illustrates an unergative predicate with reflexive se while (41-
42) are examples of an unergative and an unaccusative predicate co-occurring with
impersonal se.
(31) Se habló a sí misma mirándose al espejo
Se talked-3sg. to herself looking-herself at the mirror
She talked to herself while looking at herself in the mirror
(32) No se habla mientras se come
No se talk-3sg. while se eat-3sg.
One doesn't talk when one eats
(33) A esta clase no se llega tarde
To this class no se arrive-3sg. late
One does not arrive late to this class
Consequently, a surface string [se+VP] corresponds to multiple meanings in
Spanish. If the L2 learner were in fact using pattern learning to deduce a rule for se,
forms like (40-42) could guide him into the incorrect assumption that se can freely co-
occur with unergatives and unaccusatives as well as alternators (however, (5-6) indicate
anticausative se cannot co-occur with basic unaccusatives and unergatives). This could
lead to an overgeneralization problem. But how would the learner recover from this
overgeneralization stage? Were the learner not guided by some universal semantic
restrictions (in this case, sensitivity to the [+c] argument), recovering from this type of
phase would be extremely challenging. This is because, as mentioned, the learner would
need to use negative evidence (i.e., absence of a transitive alternative for basic
unaccusatives and unergatives) to deduce a rule for anticausative se. White, (1991, 2003)
argues that in these types of scenarios in which positive evidence is not available to
restructure a lexical entry, direct negative evidence might be necessary.
2.4. L2 studies on transitivity alternations
There are several studies that focus on transitivity alternations in L2 Spanish from
different perspectives (generative, constructivist, pedagogical) that highlight certain
trends pertinent to the current project. First of all, L2 learners showed a clear tendency to
overgeneralize se to contexts in which it is disallowed. Example (34) represents an
overextension of se with non-alternating unaccusatives and (35) illustrates an
overextension with unergatives.
(34) Julia se llegó
Julia se arrived-3sg.
Julia arrived
(35) Julia se nadó
Julia se nadó-3sg
Julia swam
Both Toth (2000) and Toth & Guijarro-Fuentes (2013), in two pedagogical-
intervention studies of se for Anglophone learners, found that when L2 learners
overgeneralized se, they significantly favored unaccusatives over unergatives. This
showed that learners were taking se as a marker of non-agentive subjects. Consequently,
because se is a marker of non-agentivity for these learners, they did not overextend se to
unergatives. They concluded then that L2 learners did not present a wild grammar but
that their overgeneralization was the product of an incorrect analysis of the function of se.
Therefore, they took this data to indicate that L2 learners are sensitive to different classes
of predicates, which supports the view of the facilitating role of UG in L2 acquisition. On
the contrary, Montrul (1999, 2000, 2001a, 2001b) in her series of seminal studies on
English, Spanish and Turkish, found an overgeneralization trend (by English learners of
Spanish) consisting of overusing se with both non-alternating unaccusatives and
unergatives. She took this to mean that learners interpreted se as a marker of
intransitivity. As in the studies just mentioned, she also argued this was a sign of a UG-
constrained interlanguage grammar. Additionally, Cabrera and Zubizarreta (2003)
distinguished two types of learners: conservative and creative. For them, conservative
learners are those who do not overgeneralize. Conversely, creative learners
overgeneralize either to just non-alternating unaccusatives or to both non-alternating
unaccusatives and unergatives. Notice that no learner overgeneralized just to unergatives.
Also, within the learners who overgeneralized with both verb classes, those who
overgeneralized more with unaccusatives outnumbered those who overgeneralized more
with unergatives.
Secondly, another pattern reported in previous L2 studies is the omission of se in
inchoative constructions (35).
(35) *El jarrón rompió
The vase broke-3sg.
The vase broke
Montrul (1999, 2000, 2001a, 2001b) considered this phenomenon a natural
consequence of the influence of L1 morphological patterns. She argued that L2 learners
whose L1s have zero-derived morphology will reject L2 patterns that have an overt
morphological marker (e.g., English learners of Spanish omitting se with inchoative
constructions). Toth (2000), Toth and Guijarro-Fuentes (2013), and Zyzik (2006) found
evidence for the omission of se in inchoative constructions, which they also ascribed to
L1 transfer effects.
Finally, there are different hypotheses as to how learners start recovering from
these overgeneralization and omission problems and how developmental advances take
place. A combination of learner-external factors (form-focused communicative
2013; Toth, 1999; Tremblay, 2006). In the current study, the lower-level participants
were taking an advanced grammar and composition course, which was a sixth/seventh
semester course for majors and minors; the most proficient learners were Spanish
instructors and teaching assistants, many of whom classified as near-native speakers.
Thus, this study allows us to uncover questions left unanswered in the previous literature
by looking at end-state grammars.
In particular, I explore the following research questions in the current study:
RQ1: Do L2 learners distinguish sentences with se and without se with respect to
alternating unaccusatives, non-alternating unaccusatives and unergatives
respectively?
RQ2: Do L2 learners make overgeneralization errors? This question is divided
into two sub-questions:
a. Do they incorrectly include se with non-alternating unaccusatives and/or
unergatives?
b. If they do, is overgeneralization only restricted to one class (only with
non-alternating unaccusatives or only with unergatives)?
RQ3: Do L2 learners make errors of omission with se? Do they incorrectly omit
se with alternating unaccusatives?
4. The study
4.1. Participants
Thirty-six monolingual Spanish speakers from Spain volunteered as the control
group. They were all college educated and had only been in contact with English
sporadically during vacation. The experimental group (N=65) was composed of college
students who were tested in a large public research university in the Southeast of the
United States. All the participants in the experimental group had been exposed to Spanish
after puberty (average14-years-old) and had learned Spanish in a classroom setting.
Additionally, some of them had spent some time in Spanish-speaking countries through
study-abroad programs. Only speakers whose L1 was English were considered for the
study. L2 learners were classified into 4 proficiency groups according to their score in a
subsection of the DELE (Diploma de Español como Lengua Extranjera/Diploma of
Spanish as a Foreign Language).3 The test included both a multiple-choice vocabulary
section and a fill-in-the-blank grammar section. Table 3 summarizes information about
the participants.
Table 3. Participants’ profiles Proficiency Number
Description
Control 36 Native speakers Near-natives
16
Spanish instructors and teaching assistants
Advanced 21 Spanish instructors and teaching assistants Intermediate
16
6/7th semester course college students
3 This measure has been used extensively by other generative L2 researchers (see the work of Cuza, Rothman, or Montrul among others).
Low
12
6/7th semester course college students
4.2. Methodology
This study consisted of a scalar grammaticality judgment task (GJT) and it was
conducted in PsyScope (Cohen, MacWhinney, Flatt, & Provost, 19934). It is part of a
series of 5 experimental tasks that were conducted together (the results of which have
been reported in other articles). The order of the 5 tasks was randomized.
Each participant received a set of instructions before starting the task and
conducted a training trial. Judgments were based on a 5-point Likert scale to determine
how natural each sentence sounded to participants (with 1 being the most unnatural and 5
being the most natural). Participants provided judgments to 24 experimental sentences
and 24 distractors. All items were randomized. Out of the 24 critical items, 12 contained
alternating unaccusatives, 6 included non-alternating unaccusatives and 6 had unergative
predicates. Half of the sentences in each category were grammatical and half were
ungrammatical. Each participant judged all of the experimental items (both grammatical
and ungrammatical) and saw each sentence once. Each verb was seen twice one in an
ungrammatical sentence and one in its grammatical counterpart but, since the items were
randomized, the grammatical and ungrammatical versions were not presented together.
The task proceeded as follows: a brief paragraph serving as context showed up in
the computer screen. The subject had to read the paragraph and press any key to make it 4 “PsyScope is a program to design and run psychological experiments, used by many experimental labs. It runs on Apple Macintosh computers. It has been developed at Carnegie Mellon by Jonathan Cohen, Matthew Flatt, Brian MacWhinney and Jefferson Provost for Mac OS 9 in the '90s. Thanks to its creators, its code has been made public, under the GNU GPL license. It has been ported to OS X thanks to a collective effort to which several labs kindly contributed. It is now being developed by the SISSA Language, Cognition and Development Lab at Sissa, the RICO group at the Universitat Pompeu Fabra, and many volunteers.” (http://psy.ck.sissa.it)
disappear. Each context was followed by a single sentence, (although the context for the
grammatical and the ungrammatical versions of each sentence pair was the same). The
software did not allow subjects to go back or change their answers.
Example (36) illustrates two of the critical items in the experiment with an
alternating unaccusative predicate. Example (36a) represents the incorrect version
without se and (36b) is the correct version with se.
(36) Mi hijo estaba jugando al fútbol en el jardín. Cuando tiró la pelota a la pared, vi
la pelota yendo directamente a la ventana
My son was playing soccer in the backyard. When he threw the ball to the wall, I
saw the ball going straight to the window
a. La ventana rompió
The window broke-3sg.
b. La ventana se rompió
The window se broke-3sg.
The windows broke
The rationale for choosing the specific verbs used in this task is based on the fact
that not all unergative and unaccusative verbs are created equal, since, as I have
illustrated, some show more consistent behavior than others (Sorace, 1993, 2000).
Consequently, I chose one core unaccusative, one less core unaccusative and one
peripheral unaccusative as well as one core, one less core and one peripheral unergative.
This was done to ensure that there was a balance of categories.5 The exact classification
per category is presented in Table 4.
Table 4. Unaccusative and unergative predicates tested in the GJT 5 Other elements such as sentence length or predicate frequency were not controlled.
p=.009). and between alternating unaccusatives and unergatives respectively since both
contrasts are significant (control: p=.000; near-native: p=.000; advanced: p=.000;
intermediate: p=.000; low: p=.043). This is what we expect since sentences with se are
grammatical with alternating unaccusatives and ungrammatical with the other two
classes. However, what we also see is that, even if the contrast is significant between
these classes, the ratings of intermediate and low proficiency speakers for non-alternating
unaccusatives and unergatives are quite high (slightly below or above 3). This means that
they do not show a complete rejection of se with these classes, which could indicate some
overextension.7
7 As a reviewer points out, this could suggest a lack of confidence in their ability to judge something as ungrammatical. The reviewer argues that this kind of behavior is typical of beginning and intermediate
Second, to determine whether each individual group overextends se to different
degrees with unergatives and unaccusatives, I contrasted participants’ ratings of
sentences with se with these two types of predicates. In theory, we would expect L2
learners to make no significant distinctions since anticausative se is equally
ungrammatical with unergatives and non-alternating unaccusatives. However, there are
two significant contrasts, one for the low-proficiency group and one for the advanced
group (control: p=.088; near-native: p=.058; advanced: p=.034; intermediate: p=.192;
low: p=.043). The low-proficiency group showed a preference towards unaccusatives
while the advanced learners showed a slight preference towards unergatives. The
relevance of this finding will be examined in the Discussion section.
Finally, in order to answer research question 3 (Do L2 learners make errors of
omission with se? Do they incorrectly omit se with change-of-state unaccusatives?) I
contrasted subject ratings of sentences without se with alternating unaccusatives to their
ratings of the same type of sentences with non-alternating unaccusatives and unergatives
respectively. LSD post hocs indicate that the contrast between alternating unaccusatives
and non-alternating unaccusatives is significant for all groups (control: p=.000; near-
native: p=.000; advanced: p=.000; intermediate: p=.000; low: p=.001) as is the contrast
between alternating unaccusatives and unergatives (control=.000; near-native=.000;
advanced=.000; intermediate=.000; low=.000). Nevertheless, again, we see that
intermediate and low-proficiency learners have a high rate of acceptance of alternating
unaccusatives without se, which is completely ungrammatical. This fact suggests transfer
from English, which has zero morphology for the inchoative construction.
learners who only weakly reject things they are nor terribly sure of. Data from a production task from the same learners would be needed to confirm this fact.
5. Discussion
This study examined how L2 learners across four proficiency levels acquired
se. I focused on three main research questions, which I answer by analyzing the results of
the current study while comparing them with the findings of previous researchers. The
research questions are repeated below for convenience:
RQ1: Do L2 learners distinguish sentences with se and without se with respect to
alternating unaccusatives, non-alternating unaccusatives and unergatives
respectively?
RQ2: Do L2 learners make overgeneralization errors? This question is divided
into two sub-questions:
a. Do they incorrectly include se with non-alternating unaccusatives and/or
unergatives?
b. If they do, is overgeneralization only restricted to one class (only with
non-alternating unaccusatives or only with unergatives)?
RQ3: Do L2 learners make errors of omission with se? Do they incorrectly omit
se with change-of-state unaccusatives?
With respect to research question 1, it is clear from the results that learners across all
proficiency levels performed similarly to the native speakers by making significant
distinctions between sentences with se and sentences without se in the three predicate
classes (refer back to Tables 5-7). What is also true is that intermediate and low-
proficiency learners showed distinctions that, although statistically significant, were
much less clear-cut than those of the more advanced learners. This shows that the least
proficient groups have more indeterminate judgments than the most proficient ones.
In order to answer research question 2, we need to look back to Figure 1/Tables 6-7.
All proficiency groups made a statistically significant distinction between non-alternating
unaccusatives and alternating unaccusatives, and between non-alternating unaccusatives
and unergatives. This indicates that they treat them as different classes of verbs and
respect their morphosyntactic make-up. However, it is important to highlight the behavior
of lower proficiency learners. Although these learners made statistical significant
distinctions, they gave mean ratings of about 3 to sentences with se with non-alternating
unaccusatives (El tren se llegó ‘The train arrived’) and unergatives (María se lloró
delante de todos ‘María cried in front of everyone’). This means that, for some learners,
these ungrammatical sentences are indeed grammatical. Therefore, we see a certain
overextension of se.
When the overgeneralization pattern was explored further and participants’ ratings of
sentences with se with unergatives and unaccusatives were contrasted; we found that
while intermediate, near-natives and native speakers made no significant distinctions
between these classes; low-proficiency and advanced learners did. In the case of the
lower-proficiency group, what we see is that they overextended se significantly more
with unaccusatives than with unergatives. This replicates Toth’s (1999, 2000) and Toth &
Guijarro-Fuentes’s (2013) findings. This pattern indicates that the grammars of the low-
proficiency group, although deviant from native grammars, are still UG-constrained:
given that alternating and non-alternating unaccusatives belong to the same semantic
class, these participants could be using se as a marker of non-agentive subjects. This
would explain why alternating and non-alternating unaccusatives patterned together
while unergatives, which have agentive subjects, did not. The behavior of the advanced
learners is undoubtedly more surprising since they exhibited a preference for unergatives.
However, even if the contrast between unergatives and unaccusatives in the se condition
is significant, the means for these categories are remarkably similar (unergatives M=1.9;
non-alternating unaccusatives M=1.61). What is important to notice is that, while the
contrast is significant, still both verb types are regarded as ungrammatical with se since
the means are considerably below 3. Additionally, it needs to be highlighted that no
group overgeneralized se just to unergative predicates.
As previously discussed, the learners at the intermediate and low-proficiency level of
this study are more proficient than most of the learners tested in the previous studies on
transitivity alternations (e.g., Toth, 2000) and at a similar level to Zyzik’s (2006) most
advanced group, among others. Hence, this finding confirms the claims of previous
researchers with regard to the difficulty of acquiring the morpheme se (particularly,
anticausative se) at lower levels of language development. The findings of this study also
answer questions left unanswered by previous research with respect to the behavior of
more proficient learners. What we see here is that the more advanced L2 learners (i.e.,
near-native group) do eventually overcome the problems with anticausative se and
respect the distribution of this morpheme according to predicate type since they do not
exhibit problems of overgeneralization.
Finally, in order to answer research question 3, I refer the reader to Figure 2/Table 5,
which show that all L2 learner groups distinguished between alternating unaccusatives
and non-alternating unaccusatives, and between alternating unaccusatives and
unergatives. However, again we see that, at a descriptive level, intermediate and low-
proficiency speakers differ from the other groups. They gave alternating unaccusatives in
sentences without anticausative se (e.g., La ventana rompió ‘The window broke’) an
approximate average rating of 3, which means that some learners incorrectly accepted
these sentences as grammatical. These learners seemed to be following L1 patterns of
zero derivation; hence, this represents a case of L1 transfer since in English the
inchoative form is not marked with a specific morpheme. This confirms Montrul’s ideas
(2001a) about the role of the L1 in this type of constructions; namely, that L2 learners
whose L1s lack an overt morpheme to express a certain meaning will tend to accept
sentences that show zero morphology and reject those with overt morphology. We see
these trends but only to a certain extent with the least proficient learners.
This article also taps on the topic of the representation of grammar at the level of
ultimate attainment, which is a matter of interest in generative L2 acquisition (e.g.,
Sorace, 2003; Valenzuela, 2006). In particular, Sorace (1993) has proposed three
different states for near-native grammars:
• native-like: when learner representations converge with native representations;
• divergent: when the learner representations diverge from native representations,
learners have determinate judgments but they go in a different direction from
native speaker judgments (optionality is a subtype here);
• incomplete: when learners show indeterminate intuitions.
The near-native group in this study behaves at a native-like level with respect to the
distribution of anticausative se with unergative and unaccusative predicates: not only do
they make the same distinctions than native speakers but they also make them with the
same degree of definiteness. Therefore, their judgments are determinate and in the same
direction than native speaker judgments, which indicates that these learners have
representations convergent with native speakers. This contrasts with the findings from the
advanced learners in Toth (1999) and Tremblay (2006) who still exhibited certain
difficulties with se. However, as argued earlier, these learners’ grammars were not
representative of the steady state because, although the learners were highly proficient,
their interlanguages were still in development.8
I contend that the behavior of near-natives in the current study cannot be a purely
surface phenomenon, that is, near-native performance cannot be the result of replication
of input patterns, but it necessarily has to be an indication of their underlying linguistic
representations. As claimed in Reinhart (2002) and Reinhart and Siloni (2005), verbs that
undergo decausativization need to have a [+c] argument. As described in section 2.2.1,
this applies to verbs like break and psychological verbs like worry (37-38):
(37) María preocupó a Juan
María worried-3sg. to Juan
María worried Juan
(38) Juan se preocupó
Juan se worried-3sg.
Juan worried
Interestingly, not only did the participants in this experiment show sensitivity to the
[+c] argument of alternators such as romper ‘to break’ as shown in their ratings in the
current GJT, but the same group of participants also showed sensitivity to [+c] with
8 In these studies, no particular proficiency measures were used to classify participants. They were placed into a specific proficiency group depending on the Spanish class on which they were enrolled at the time of testing. This approach is problematic because of the varying levels of proficiency we can encounter in students enrolled in the same class.
psychological verbs in a different task reported in Gómez Soler (2013, 2014). This
provides evidence that access to this abstract feature helped these L2 learners understand
decausativization productively with different types of predicates (i.e., alternators and
psych-verbs).
Advocates of connectionist and usage-based models (Ellis & Larsen-Freeman, 2009;
Ellis, O'Donnell & Römer, 2013) could argue that acquisition of se takes place with the
learner tallying inductively examples of sentences with and without se from the input.
Conversely, I put forth the argument that input is indeed insufficient for the acquisition of
this complex phenomenon and that access to the abstract features that compose the
thematic structure of verbs is a necessary requirement for successful acquisition. Given
that unergatives and non-alternating unaccusatives cannot co-occur with anticausative se
but can in fact co-occur with other types of se such as impersonal se (A esta clase no se
llega tarde ‘One does not arrive late to this class’), the input could lead to
overgeneralization patterns. In order for the learner to recover from this
overgeneralization, he would need to notice the absence of transitive alternatives for non-
alternating unaccusatives and unergatives in the input. It would be these structures the
ones that would allow him to discover that alternating unaccusatives do alternate in
transitivity while other types of intransitives do not. As argued by White (1991, 2003), it
is highly unlikely that the L2 learner is able to restructure his grammar based on this type
of indirect negative evidence were he guided by input alone. Another argument to support
L2 learners’ access to abstract semantic features as guiding forces in the acquisition
process of anticausative se is the fact that L2 learners do not exhibit a wild grammar.
That is, L2 learners’ linguistic representations are UG-constrained since they do not
violate UG principles. We can see this in their overgeneralization patterns: the lower-
proficiency group overextended se significantly more with unaccusatives than with
unergatives. As discussed, this suggests that participants take se as a marker of non-
agentivity. Thus, this is an indication that participants are sensitive to the semantic
similarities between alternating and non-alternating unaccusatives. On the other hand, we
would not expect unergatives to be the sole target of overgeneralization since they do not
share semantic aspects with alternating unaccusatives. In fact, we do not see this in any
group, that is, no group overgeneralized just to unergatives (as I argued, even if the
advanced speakers showed a significant distinction between unergatives and
unaccusatives, they still rated both classes below 2 and they did not isolate unergatives).
What is also evidenced in the data from near-natives is that transfer of L1
morphological properties (i.e., alternating unaccusatives without se) has been overcome
at the level of near- nativeness. Toth (1999) reported that his advanced participants were
unable to expunge this specific L1 pattern from their grammars. However, given that
Toth’s participants were not at the stage of ultimate attainment, what the current study
indicates is that highly proficient learners are able to make use of lexical properties of
verbs and are sensitive to the distribution of se with these different classes of predicates,
regardless of the hindering effect that L1 transfer could have at lower levels of
proficiency.
Consequently, this area of Spanish grammar, the acquisition of anticausative se, does
not seem to be subject to fossilization at the near-native level. On the contrary, near-
natives exhibited native command of this construction and converged with native
speakers in their judgments. This entails that they (i) exhibited sensitivity to the semantic
nuances of verbs that are relevant for the distinction ([+c]) and (ii) realized how the
distinction is marked morphologically in Spanish (se). Thus, UG helped the leaner reduce
the hypothesis space by allowing him to group predicates according to their nuanced
semantic and morphosyntactic behavior.
For Reinhart & Siloni (2005, p. 391), the process examined in this article (i.e.
decausativization) is related to a parameter having to do with arity operations:9
(39) The lex-syn parameter: Universal Grammar allows thematic arity
operations to apply in the lexicon or in the syntax.
They argue that English sets to “lexicon” while Spanish sets to “syntax”. However,
decausativization always takes place in the lexicon,10 while reflexivization and other
operations can take place in the lexicon or the syntax. An analysis of L2 learners’
understanding/use of reflexivization or other arity operations (e.g., saturation,
reciprocalization) in conjunction with the findings of this study could shed light on the
lex-syn parameter in L2 acquisition. This is an endeavor for future research.
Finally, this study has certain limitations such as the low number of experimental
items and the lack of another experimental task that could complement the results of the
GJT. Further research will need to address these limitations in order to confirm the
findings of this study. Additionally, exploring participants’ intuitions and use of
unergative and unaccusative predicates with other types of se (e.g., reflexive, impersonal)
will broaden our understanding of L2 learners’ knowledge of Spanish argument structure.
9 Arity operations are those that that affect the syntactic valence of a predicate (Reinhart & Siloni, 2005, p. 390) 10 Following the lexicon interface guideline: The syntactic component cannot manipulate θ-grids: elimination, modification, and addition of a θ-role are illicit in the syntax.
6. Conclusion
This article examined developmental patterns of the acquisition of Spanish
anticausative se across four proficiency levels ranging from low/intermediate to near-
native. The L2 learners in this study were sensitive to the semantic nuances that
distinguish predicates that do and do not alternate in transitivity. However, low-
proficiency learners experienced some problems with morphology related to L1 transfer
and the polyfunctionality of Spanish se. The main contribution of this paper to the study
of transitivity alternations is that it suggests that, at least for the acquisition of
anticausative se in Spanish and despite difficulties at lower levels of proficiency, L2
learners manage to succeed at the end of the road.
Acknowledgments
I am deeply grateful to Misha Becker for her useful feedback on earlier versions of this
manuscript. I also thank the anonymous reviewers of RESLA/SJAL. All remaining errors
are my own.
References
Aranovich, R. (2000). Spanish inchoatives and the semantics of split intransitivity. Paper
presented at the Annual Meeting of the Linguistic Society of America, Chicago.
Birdsong, D. (2006). Age and second language acquisition and processing: A selective
overview. Language Learning, 56, 9-49.
Bruhn de Garavito, J. (1999). The se constructions in Spanish and near-native