Systemic Mo,va,on of 'Anomalies' in GrammarFarrell Ackerman University of California at San Diego
1
Understudied languages can contribute to the development of linguis=c theory;
Linguis=c theory can contribute to the understanding of understudied languages.
The last sentence in Dryer’s presenta=on yesterday.
HPSG 2014 -‐ University of Buffalo
Typology, descrip,on, and theory: Charles J. Fillmore
The difference between “bu0erfly collectors” and “systema9c observers”:
Rejec9ng the dissectors bench, the morgue like character of natural history museums, and academic zoology in general, these fieldworkers thrived outdoors. Furthermore, unlike most field naturalists before them, they went out into nature not as specimen collectors, but rather as animal watchers. Burkhardt (2005: 69)
This new type of field naturalist transformed the study of animal behavior into the science of compara9ve psychology.
Charles J. Fillmore’s nuanced insights about the nature of grammar systems helped to transform the theore9cal study of language: they have led many researchers over many years to explore construc9on-‐theore9c approaches to linguis9c analysis.
He was a singular naturalist of language and a keen grammar watcher.
2
Typology, descrip,on, and theory: Ivan Sag
Beyond his well-‐recognized theore9cal contribu9ons, an under-‐appreciated dimension of Ivan was his extensive and rich understanding of the history of linguis9cs and the manner in which this informed his views about formal theory and language varia9on.
“…[T]he goal of our enterprise…is to provide a basis for the descrip9on of all human languages…” Sag, Boas, and Kay (2012:70)
“Perhaps the most important goal of Sign-‐Based-‐Construc9on Grammar, which has emerged in the Formal Grammar community, is to provide a formalized framework in which Typological researchers can develop their ideas.” Sag, Boas, and Kay (2012:3)
There were convergences (intuiAons about the systemic nature of grammar organizaAon and the importance of grammaAcal construcAons) and divergences (intuiAons about appropriate formalizaAon) between Chuck and Ivan.
But mostly there was a love of language phenomena and the joyful appreciaAon of parAcipaAng in an adventure that began before they appeared and would be changed for new generaAons because their presence had changed it.
3
Grounding specula,ons: Convergences
Construc9on-‐theore9c approaches to morphology and syntax par9cipate in a reconceptualiza9on of grammar analysis which views grammar systems as complex (adap9ve) systems.
Encourages the use of many of the assump9ons and toolkits that guide the ongoing reconceptualiza9on of complex structures in
Developmental Compara9ve Psychology and Psychobiology (D. Lehrman 1953, 1970; Y-‐Z Kuo, 1960; G. Go0lieb 1970, 1992, 1997; Karmiloff-‐ Smith 1992; E. Thelen & L.B Smith 2003; G. Michel & C. Moore 1995; J. Elman et. al. 1997; J. S9les 2008;; M. Bornstein & M. E. Lamb; 2010; K. Hood et.al. 2010, among others)
(Ecological) Developmental Evolu9onary Biology (Ludwig von Bertalanffy 1933/1962; Paul Weiss 1939/196); Sewall Wright 1968; S. Gilbert et. al. 1996; Lewon9n 2001, S. B. Camazine et. al. 2003; S. Oyama et. al. 2003; M. J. West-‐ Eberhard, 2003; Amundson 2005; S.Gilbert & D. Epel 2009; A. Wallace 2011; Bateson & Gluckman 2011, among others.) A deeply different concep9on of the rela9onship between perspec9ves on analysis and relevant methods of analysis in the developmental sciences and language than found in conven9onal Mainstream Genera9ve research programs e.g., Biolinguis9cs. (Burraco & Longa 2010; Bateson & Gluckman 2011, among others.)
4
General Observa,ons
What is the relevance of the developmental sciences for language analysis?
What I am sugges=ng:
1. Developmental sciences (both psychological and biological) offer conceptual insights about how to approach and explain the nature of objects that arise within complex adap9ve systems.
2. Developmental sciences offer toolkits (computa9onal simula9ons, quan9ta9ve analysis (probablity, sta9s9cs, informa9on theory) for analyzing such objects.
3. When grammar is conceptualized as a complex adap9ve system, then the relevance of the developmental sciences in their focus on systemic organiza9on and its mul9-‐causal explana9ons becomes evident.
5
Gramma,cal analysis and the developmental sciences
What I am not sugges=ng:
1. Grammar derives from biologically pre-‐specified representa9ons reflec9ng domain-‐ specific and species-‐specific proper9es only relevant for human language, i.e. the Biolinguis9c gambit on Universal Grammar.
\
Everything that will be is there in more schema9c form at the beginning.
The adult steady-‐state grammar is there in inchoate form in the infant startstate.
Grammar modularity is there in the beginning, rather than emerging over 9me in the adult. (Karmiloff-‐Smith 1995)
Gramma,cal analysis and the developmental sciences
Suspicions about “language ins9ncts” are bolstered by the fact that it is a common and successful strategy in the developmental sciences to recognize that looking for ins,ncts for X, e.g., a language ins9nct, is more oqen an obstacle to understanding traits and behaviors than it is a help:
The use of “explanatory” categories such as “innate” and “gene,cally fixed” obscures the necessity of inves,ga,ng developmental processes in order to gain insight into the actual mechanisms of behavior and their inter-‐rela,ons. The problem of development is the problem of the development of new structures and ac,vity paLerns from the resolu,on of the interac,on of exis,ng structures and paLerns, within the organism and its internal environment, and between the organism and its outer environment.” In Oyama et. al. 2001:31
Posi9ng ins9ncts and innate representa9ons forecloses asking important ques9ons about why what occurs where it occurs and in the way it does so over ontogene9c and phylogene9c 9me. (Y-‐Z Kuo 1932; Schneirla 1966; Lehrman 1953, 1960, 1970; Go0lieb 1997; Amundson 2005; Blumberg 2009; Jablonka & Lamb 2005; Gilbert and Epel 2009, Hood et. al. 2010, among others.)
“The ins,nct psychologist ends his inves,ga,on where the non-‐ins,nct psychologist begins.” Y-‐Z Kuo 1922
7
Today: Trying to resolve a perennial challenge
The fact about science is that everyone who has a made a serious contribu9on to it is aware, or very strongly suspects, that the world is not only queerer than anyone has imagined, but queerer than anyone can imagine. This is a most disturbing thought, and one flees from it by sta,ng the exact opposite. J. S. Haldane as cited R. G. Reid Biological Emergences: Evolu9on by Natural Experiment 2007:4311.
“...individuals are quite stupid compared to the complexity of the problems we aspire to solve... All anyone can hope to do is to make canny simplifica,ons that do minimum damage to understanding.” P. J. Richerson & R. Boyd Not by genes alone: How culture transformed human evoluAon 2005:248
A ques,on: What do you do if you constantly encounter phenomena that are unexpected or precluded by the standard canny simplifica9ons in linguis9c theory, i.e,. rendered anomalous, which themselves appear to do more than minimum damage to understanding, but don’t want to flee?
8
This Tundra Nenets data and analysis is based on collaborative work with Irina Nikolaeva. Primary reseach on Tundra Nenets was generously supported by a Hans Rausing Language Documentation Grant 2003-2006 with Irina Nikolaeva and Tapani Salminen, We thank our primary consultants Galina Koreneva and Anna Lambdo.
Gramma,cal analysis and the developmental sciences
An answer: Look to successful strategies in the developmental sciences for the analysis of complex adap9ve systems as instruc9ve for understanding grammar organiza9on.
Addi9onal research with James P. Blevins, Olivier Bonami and Rob Malouf on Word and Paradigm (Pa0ern) morphology.
9
Different bets
Two ques9ons:
Q1: Why does what recurs cross-‐linguisAcally, recur?
Q2: Why do the variaAons in encodings of what recurs look the specific way they do?
Two responses: (simplis9cally stated)
R1: Genera9ve proposals focus on answering Q1 (UG) and bet that the answer to Q2 derives from this: the language faculty sets the representa9onal constraints and constraints on basic representa9ons yield observed varia9on.
R2: Construc9on-‐theore9c proposals focus on answering Q2 by appealing to systemic mo9va9ons within par9cular grammar systems for construc9onal varia9on and bet that systemic explana9ons at a broader level can be used to answer Q1.
What are the analy9c intui9ons and representa9onal resources to address Q1 and Q2?
10
Amplifying on Q2:
a. How does varia9on in the cross-‐linguis9c encoding of gramma9cal phenomena instruct us about the nature of human language and the constructs required in linguis9c theories to iden9fy explana9ons/mo9va9ons for the pa0erns we find?
b. What role should linguis9c “rari9es” or “anomalies” play in the development of linguis9c theories?
Rarum = def “a trait...which is so uncommon across languages as not even to occur in all members of a single...family or diffusion area... Diachronically speaking, a rarum is a trait which has only been retained, or only been innovated, in a few members of a single family or sprachbund or of a few of them.” from Plank online Raritätenkabine0 as cited in Cysouw & Wolhgemuth.
11
Rara
Useful dis9nc9on: A rarum is a datum, an anomaly is a phenomenon that is counter to expecta9on.
"The great field for new discoveries," said a scien,fic friend to me the other day, "is always the unclassified residuum." Round about the accredited and orderly facts of every science there ever floats a sort of dust-‐cloud of excep,onal observa,ons, of occurrences minute and irregular and seldom met with, which it always proves more easy to ignore than to aLend to...Any one will renovate his science who will steadily look a]er the irregular phenomena. And when the science is renewed, its new formulas o]en have more of the voice of the excep,ons in them than of what were supposed to be the rules.
James, William The Will to Believe : and Other Essays in Popular Philosophy 1896: 135. Kindle Edi9on.
12
The value of “anomalies” in biology
Despite their characteriza,on as errors of nature, the anomalous, when properly considered, force us to confront and correct those errors in our thinking that o]en impede scien,fic insight and progress.” M. Blumberg Freaks of Nature: What anomalies tell us about development and evoluAon. Oxford University Press 2009:13
“We do not treasure our excep,ons, as William Bateson urged us to do, we autoclave [sterilize FA] them. S. Gilbert The decline of soG inheritance in Gissis and Jablonka eds. Tranforma=ons of Lamarkism. 2011:123
“Anomalies” extend our no9ons of what is possible, and hence, natural: they, accordingly, alter our concep9on of what theory should look like
13
Recombinant poten,al in biological forms
Platypus (monotreme): A furry, egg-‐laying, duck-‐billed, echoloca9ng, venomous (when masculine) creature.
Two odd ques9ons:
1. Is the platypus more or less natural than the duck or the beaver (or bat)? 2. Is the platypus a departure from an essen=al duck or beaver, or is it the other way around?
While all of the listed platypus proper9es are familiar, their alignments produce an evidently “anomalous”, historically con9ngent, but ecologically viable pa0ern (see discussion in The platypus and the mermaid and other figments of the classifying imaginaAon, H. Ritva 1997.)
The pa0ern is new, but built from the arrangement of familiar pieces.
14
The value of “anomalies” in language
“Anomaly” doesn’t exist in language, rather, lurking behind it are anomalous presupposi,ons and convic,ons that obtain in linguis,c theory. A. Kibrik 2003:304
Linguis9c anomaly tells us what is possible in natural language and that, as in biological systems, oddity, or what appears to be anomaly, is oqen explicable in terms of the complex system in which the phenomenon occurs.
Gramma9cal platypuses: reuses of old pieces in new configura9ons for newpurposes constrained by the pathways defined by interdependencies among many contribu9ng factors ( = systemic mo9va9on).
15
Four “capsule descrip,ons”1 of puzzling “rara”: Mul,ple construals in mul,ple object construc,ons What is common: Many languages have valence-‐changing opera9ons that add beneficiary/recipient arguments
What is less common: Some languages have symmetrical objects (Kimenyi 1980, Hyman & Duran9 1982, Bresnan & Moshi 1990, McGinnis 2008, Pylkkänen 2008, Travis 2010, among others)
What is even less common: Some languages permit mul9ple seman9c role construals for (all) mul9ple objects: (McKay & Trechsel 2008, among others)
Symmetrical object construc9ons in Moro, Heiban: Ackerman, Malouf, and Moore (2013, submi0ed)
1. í-‐ɡ-‐ʌ-‐nʌʤ-‐ət̪-‐ú aljásər-‐o kúku-‐ŋ ŋállo-‐ŋ 1sg.sm-‐clɡ-‐main-‐give-‐appl-‐pfv clɡ.Elyasir-‐acc clɡ.Kuku-‐acc clɡ.Ngallo-‐acc a. ‘I gave Elyasir to Kuku for Ngallo.‘ b. ‘I gave Elyasir to Ngallo for Kuku.‘ c. ‘I gave Kuku to Elyasir for Ngallo.‘ d. ‘I gave Kuku to Ngallo for Elyasir.‘ e. ‘I gave Ngallo to Kuku for Elyasir.‘ f. ‘I gave Ngallo to Elyasir for Kuku.‘ 1. D. Pesetsky (2009) commentary on Evans and Levinson (2009)
16
Four “capsule descrip,ons” of puzzling “rara”: Beneficiary/Recipient Construc,on -‐ another varia,onWhat is common in Ditransi9ve Construc9ons: expression of an independent beneficiary/recipient argument licensed by a simple Verb or applica9ve morphological marking on a Verb.
What is less common: NP internal expression for beneficiary/recipient argument on another argument of the Verb (“indirect object lowering” Malchukov ms.):
NP-‐”BEN/REC”-‐POSS.PRO
“Designa9ve” construc9on in Nanai, Tungusic -‐ (Malchukov 2010:148) -‐ similar to the “predes9na9ve” construc9on in Samoyedic -‐ Tundra Nenets (Nikolaeva 2006, ms.), Forest Enets (Siegl 2013) and Ewe ditransi9ve construc9ons reported in Creissels 2005:70.
2. Hin turki-‐ga-‐s e-‐mur-‐em your sledge-‐DES-‐2SG bring-‐NON-‐FUT-‐1SG `I brought the sledge for you’
3. Hin turki-‐wu-‐s e-‐mur-‐em your sledge-‐ACC-‐2SG bring-‐NON-‐FUT-‐1SG `I brought your sledge’
17
Four “capsule descrip,ons”1 of puzzling “rara”: use of applica,ve to encode Compara,ve Construc,ons What is common across Compara9ve Construc9ons:
COMPAREE COMPARATIVE PRED STANDARD OF COMPARISON Murry is hungrier than Sally
What is less common: an APPL(ICATIVE) marker used polyfunc9onally as a compara9ve marker
Compara9ve construc9ons in Moro, Heiban: (Gibbard et.al. 2010, Schadeberg 2013; Kertz, Ackerman and Rose forthcoming)
4. Kúku g-‐ʌ-‐tʃuɲ-‐t̺-‐ú ŋá́lo-‐ŋ Kuku CL-‐MAIN-‐hungry-‐APPL-‐PFV ŋalo-‐ACC `Kuku is hungrier than ŋalo’
5. Kúku g-‐ʌ-‐d-‐ə t̺-‐ú ŋá́lo-‐ŋ éləŋ Kuku CL-‐MAIN-‐be-‐APPL-‐PFV ŋalo-‐ACC rich.person `Kuku was richer than ŋalo
6. Kúku g-‐ʌləŋ-‐ə t̺-‐ú úm:iə ŋó́péa Kuku CL-‐sing-‐APPL-‐PFV boy well `Kuku sang be0er than the boy’ or `Kuku sang well for the boy’
18
Four “capsule descrip,ons” of puzzling “rara”: Possessive Rela,ve Clauses
Many languages have externally headed prenominal NON-‐SUBJECT rela9ves:
[[ øGAP1 … VMC ]LOCAL DOMAIN NPHN1 ]EXTERNAL DOMAIN
NON-‐SUBJ NON-‐SUBJ
built house
`the built house’
Schema 1
1. The rela9ve func9ons as the modifier of the rela9vized head nominal (HN)2. The local domain headed by the verbal mixed category (MC) is a full clause3. The rela9vized nominal bears a NON-‐SUBJECT, (OBJ, ADJUNCT...) rela9on to the gap4. Gap simply a conven9on for indica9ng that something is missing in the local domain that bears a syntac9c & seman9c rela9on to the VMC.
Q: How is a pronominal SUBJ expressed?
19
What’s common: Construc,on Type 1
M(ixed)C(ategory)-‐inflected rela9ve: person-‐number markers (PNM) express SUBJ pronominal on the VMIXED CATEGORY
[[ øGAP1 ... VMC-‐PNMSUBJ ]LOCAL DOMAIN HN1 ]EXTERNAL DOMAIN
Schema 2
Eastern Armenian (IE):
7. [[ (ma) øGAP1 … gnac’-‐əs-‐ ]LOCAL DOMAIN hovanocə1 ]EXTERNAL DOMAIN
1SG.GEN buy.PERF.PART-‐1SG.SUBJ umbrella-‐DEF ` the umbrella I bought’
Observa,on: The PNM is local to the domain defined by the verbal modifier.
20
What is less common: Construc,on Type 2
Possessive relaAve -‐ person-‐number marking (PNM) expresses SUBJ pronominal on the HN: (Ackerman, Malouf & Nikolaeva 2004) [[ øGAP1 ... VMC ]LOCAL DOMAIN HN1-‐PNMSUBJ ]EXTERNAL DOMAINN
Western Armenian, Indo-‐European:
8. [[ (im) øGAP1 … koʁtsadz ]LOCAL DOMAIN təram-‐əs1 ]EXTERNAL DOMAIN 1SG.GEN stole-‐PERF.PART money-‐1SG ‘the money I stole ’
Observa,on 1: The PNM seems to be in the wrong place, i.e., it bears a SUBJ rela9on to the VMC heading the modifying clause. (runs afoul of the locality evident in Pa0ern 1)
Nominal Possessive ConstrucAons: Head-‐marking strategy
9. (im) hin naver-‐əs 1SG.GEN old le0er-‐1SG `my old le0er’
Q: Is the resemblance between these independent construc,ons fortuitous? 21
Summary of PRCs
1. PRCs are externally headed prenominal clauses
2. Their subjects are expressed by (either lexical NPs) or incorporated pronominals on their modified head.
3. As subject pronominals the PNMs should be local, like incorporated pros or gramma9cal agreement, but they are external to the VMC which heads the rela9ve.
Cool data, but so what?
Pesetsky (2009:464 ) expresses appropriate skep9cism about similar “cabinets of curiosi9es” iden9fied in Evans and Levinson (2009), since the importance of unusual variants of familiar construc9ons w/o analysis is hard to evaluate.
So, what sort of analysis can be developed?
23
Organiza,on of the talk
1. An instruc9ve approach to a rarum: Udi endoclisis
2. Possessive rela9ves in Tundra Nenets
3. Conclusions
24
1. An instruc,ve approach to a rarum: Udi endoclisis
25
An instruc,ve explora,on of a puzzling rarum: Udi endoclisis (Harris 2007) Some9mes in the course of examining lesser studied languages an encoding strategy for
a familiar empirical phenomenon seems surprising, even puzzling, given expecta9ons developed on the basis of previous descrip9ve and theore9cal research.
Harris 2007 iden9fies an unusual phenomena: Udi endoclisis, i.e., the posi9oning of person/number markers internal to complex verbal stems.
Consider the Udi verbal template in (10a) and its surface expression (10b):
10a. Incorporated element – person marker – light verb – tense/aspect/mood
b. xoyš -‐ ne -‐ b -‐ sa request 3SG do PRES
`He begs’
the person/number marker ne - appears intermorphemically, internal to a complex, but morphophonologically cohesive single verbal word.
Udi endoclisis
More spectacularly, in (11), the person number marker is interposed intramorphemically between the two segments that comprise the root `to drink’ uɣ :
Harris 2007 iden9fies an unusual phenomena: Udi endoclisis, i.e., the posi9oning of person/number markers internal to complex verbal stems.
Consider the Udi verbal template in (11a) and its surface expression (11b):
11a. ROOT1 – person marker – ROOT2 -‐ tense/aspect/mood
b. u ne ɣ sa drink 3SG drink PRES
`He drinks’
Possible explana,ons (A. Harris 2007)
Why are construc9ons such as these rare?
H1: such cli9cs make the host morpheme difficult to understand (to process)
H2: our innate language capacity makes intra-‐morphemic cli9cs difficult/ “expensive”
H3: this system is too complex to func9on well and is doomed for a short life
H4: this system is too complex to be acquired easily by children.
Problems with these hypotheses
(i) no (direct) evidence exists to support any of these alterna9ve views.
(ii) All four views are based on the fact that languages lack word-‐internal cli9cs, and the reasoning would thus be circular.
(iii) This system has endured for at least 1600 years.
(iv) None of these views provides an explana9on of why Udi (and presumably a few other languages) deviate from the common pa0erns modeled by theories and does so in the par9cular way it does.
Construc,on-‐theore,c (paLern-‐theore,c) approach
Basic General Strategy: Provide detailed descrip9ons of cross-‐linguis9cally recurrent gramma9cal phenomena in all of their variety (without arbitrarily privileging any par9cular encoding).
Familiar Large Ques,on: What are the bounds of variability and what constrainsit?
Uniquely human pa0ern forming capaci9es does not entail that any par9cular pa0ern is en9tled to privileged or universal analy9c status.
Hypothesis: This strategy turns apparently unruly rarity in grammar into instruc9ve guidance about the nature of adequate linguis9c architectures for both syntax and morphology.
Recombina9on of individual elements and ensembles of elements found in independent construc9ons are systemic redeployments of old elements within new configura9ons constrained by con9ngent factors and shaped by the uniquely pa0ern crea9ng capaci9es, as well as cogni9ve and perceptual capaci9es of humans.
29
Poten,a,ng the possible (A. Harris 2007) Harris’ Hypothesis: Udi possesses an unusual construc9on because its historical
development presents an unusual combina9on of circumstances and events that probabilis,cally poten,ate endoclisis.1
Linguis9c “anomaly” tells us what is possible in natural language and that, as in biological systems, rarity, or what appears to be anomaly, is explicable in terms of the dynamics of pa0erns in the system within which the phenomenon occurs;
the “anomalous” may not simply be an oddball encoding of the familiar and reducible to it;
it may actually be different and instruc9ve about how to analyze the more familiar encodings.
What’s clear: Udi endoclisis is composed of cross-‐linguis9cally recurrent proper9es configured in a striking way: pronoun incorpora9on, gramma9caliza9on of auxiliaries, variable morphotac9cs, syntac9c focus construc9ons...
What’s neither clear nor necessary: Udi endoclisis represents a variant of one or another (of a selected structural encoding) of these proper9es, rather than a language par9cular encoding of them all.
1. See Yu on the diachronic pathways permi�ng infixa9on in Miestamo and Walchli eds. New Challenges in Typology, Berlin: Mouton DeGruyter 2007.
“Anomaly”
What is not anomalous in Udi (and follows from widely a0ested diachronic processes) : associa9on of SUBJ status with incorporated pro(nominal)s.
What is anomalous in Udi: the actual pa0erns of distribu9on and surface encodings for the incorporated SUBJ pro.
2. Possessive rela,ves in Tundra Nenets
32
Construc,on Theore,c bet: Quan,ta,ve Distribu,onal Typology (Bickel 2013)
1. What linguis9c structures are there in human languages, and how can we compare them?
2. Where do we find these structures, i.e. are they areally or genealogically restricted, or are they universally preferred or dispreferred?
3. Why do we find the structures where they are?
Construing these ques9ons for our purposes:
These ques9ons bear on iden9fying (and represen9ng) a taxonomy of the basic units of linguis9c analysis, their distribu9ons, and exploring mo9va9ons for why they are the way they are.
This is a strategy for resolving our difficult challenge and this is what I informally explore in the rest of the talk. (Keller 1995 on the dynamics of language change)
33
Taxonomy of Possessive Rela,ve clauses
Typology Ques,on 2: Where do these rela9ves occur?
Observa,on: They seem to appear only in gene9cally related and unrelated languages in Eurasia. (note: E. Armenian/Northern Ostyak have MC-‐inflected rela9ves, but W. Armenian/W. Ostyak have Possessive rela9ves.)
Uralic Ob-‐Ugric (W. Ostyak, Vogul) Samoyed (Tundra Nenets, Enets, Nganasan), Udmurt, Komi, Mari
Mongolic Buryat, Dagur, Kalmyk, Khalkha Mongolian
Turkic Southwestern (Azerbaijani, Turkmen, Gagau, Southeastern (Uzbek, Uighur), Northeastern (Sakha, Shor, Tuva, Khakas, Altai), Northwestern (Kazakh, Kirghiz, Tatar, Bashkir, Karakalpak), Chuvash
Tungus Evenki, Even, Orok
Palaeosiberian isolate Kolyma Yukaghir, Tundra Yukaghir
Indo-‐European Western Armenian
34
Possessive rela,ve clauses
Typology Ques,on 3: Do these distribu9ons follow from any theory, i.e., are they predicted?
Observa=on: Every theory can deploy its tools to redescribe these distribu9ons, e.g. raising, movement and their analogues in lexicalist theories: how organic are the modifica9ons necessary to accommodate them within a theory.
Observa=on: This, of course, is interpretable as responsiveness to new data.
Basic Challenge: Is there a way to moAvate/explain why the Possessive RelaAve looks the way it does and is reliably idenAcal to nominal possessive construcAons wherever it occurs?
A relevant early insight:
In diverse languages considered separately, each for itself and in its own func,oning, the analysis of the rela,ve clause shows a formal structure ordered by a certain func,on that is not always visible. The problem is to uncover that func,on. This can be arrived at by observing that the rela,ve clause o]en has, in a given linguis,c system, the same formal marks as another syntagm of a denomina,on so en,rely different that no one would think that they could be related. Guided by this formal analogy, the interpreta9on of the rela9ve clause becomes possible in terms of func9on. It is the internal rela9onship which we propose to bring to light first. Emile Benveniste 1971
35
Construc,on-‐theore,c gambit
Guiding intui,on: Cross-‐linguis9cally, the language systems containing Possessive Rela9ves (PRC) contain the same four independent contribu9ng construc9ons with language specific encodings. (Fillmore, Kay & O’Connor 1988; Fillmore & Kay 1999, Malouf 2000, Kathol 2004, among others)
Morphological construc9ons: Word and Pa0ern based morphology (Hocke0 1987; Ma0hews 1991; Bochner 1993; Gurevich 2006, Ackerman et. al. 2009, Booij 2010, Ackerman & Malouf 2013, Bonami 2013, Stump & Finkal 2013, Blevins 2014, among others)
Syntac9c construc9ons: Word combining pa0erns (Fillmore, Kay, & O’Connor 1987;, Goldberg 1995, Culicover 1999; Goldberg & Jackendoff 2004, Culicover and Jackendoff 2005, Mueller 2006; Sag 2007 Boas & Sag 2012, Lee 2013, among others)36
Construc,on-‐theore,c gambit
Recombina9on of various elements found in independent morphological and syntac9c construc9ons cooperate to (probabilis9cally) yield a systemic redeployment of a new configura9on, the prenominal Possessive Rela9ve.
P(c1, c2, c3, c4 | PRC ) ≈ 1 If a language has PRC, it is poten9ated by c1-‐c4.
P(PRC | c1, c2, c3, c4) = ? If a language has c1-‐c4, can’t predict presence of PRC, since there are numerous languages that have many or all of c1-‐c4, but don’t have prenominal Possessive rela9ve.
The challenge for the generaliza9on: Try to locate data that disconfirm an explana9on in terms of contribu9ng construc9ons, since more confirming data doesn’t help to establish it’s viability.
37
Systemic explana,on of the possible: Poten,a,ng the possible (Ackerman & Nikolaeva Descriptive Typology and Grammatical Theory To appear)
Explore the hypothesis that grammar is a complex adap9ve system in which interac9ons between its many dimensions and their ingredients produce a canalizing or direc9ng influence concerning what sorts of grammar proper9es and construc9ons may arise over 9me.
Interpreted as a poten9a9ng influence rather than a determinis9c one since in many instances structures permi0ed by par9cular interac9ons simply do not occur, though they could have, given different con9ngent condi9ons.
Certain structures possess an exceedingly low probability of arising, since systemic interac9ons are unlikely to produce them.
Recalls Harris’ main line of inquiry and her argument that systemic proper9es of the grammars license ‘odd’ construc9ons.
In our case the ante is raised: rather than looking at a single construc9on in a lone language, we analyze an ‘odd’ construc9on type that appears in numerous related and unrelated languages in Eurasia.
Moreover, rela9ve clauses are undoubtedly a core linguis9c construc9on according to any criteria, given that they are so pervasive in the languages of the world and that they have been analyzed in every linguis9c theory.
Tundra Nenets: Possessive rela,ves
Anna Lamdo and Maria Barmič
39
Relevant gramma9cal features
Morphological:
1. Largely agglu9na9ve with some cumula9ve markers2. Polyfunc9onal set of PERSON NUMBER MARKERS (PNMs) signaling two-‐place rela9ons: SUBJ/OBJ, POSS-‐OR/POSS-‐ED, LOCATION/LOCATUM3. Morphotac9cs: N-‐CM-‐PX 4. 3 PERS; 3 NUM (SG, DU, Pl)5. 7 nominal CASES
Syntac9c:
1. SOP(REDICATE)2. Numerous PNM inflected non-‐finite verbs
Tundra Nenets: clauses headed by verbal modifiers
Possessive relaAve -‐ person-‐number marking (PNM) expresses SUBJ pronominal on the HN: [[ øGAP ... VMC ]LOCAL DOMAIN HN-‐PNMSUBJ ]EXTERNAL DOMAINN
Tundra Nenets (Uralic):
[[ øGAP … ta-‐wi° ]LOCAL DOMAIN te-‐da ]EXTERNAL DOMAIN giveMC reindeer-‐3SG ‘the reindeer he/she gave’
Observa,on 1: The PNM seems to be in the wrong place, i.e., it bears a SUBJ rela9on to the VMC heading the modifying clause. (runs afoul of locality)
Nominal Possessive ConstrucAons: Head-‐marking strategy
serako te-‐da white reindeer-‐3sg `his/her reindeer’
Q: Is the resemblance between these independent construc9ons fortuitous?
40
Construc,on-‐theore,c gambit
Guiding intui,on: Cross-‐linguis9cally, the language systems containing Possessive Rela9ves (PRC) contain the same four independent contribu9ng construc9ons with language specific encodings. (Fillmore, Kay & O’Connor 1988; Fillmore & Kay 1999, Malouf 2000, Kathol 2004, among others)
Morphological construc,ons: Word and Pa0ern based morphology (Hocke0 1987; Ma0hews 1991; Bochner 1993; Gurevich 2006, Ackerman et. al. 2009, Booij 2010, Ackerman & Malouf 2013, Bonami 2013, Stump & Finkal 2013, Blevins 2014, among others)
Syntac,c construc,ons: Word combining pa0erns (Fillmore, Kay, & O’Connor 1987;, Goldberg 1995, Culicover 1999; Goldberg & Jackendoff 2004, Culicover and Jackendoff 2005, Mueller 2006; Sag 2007 Boas & Sag 2012, Lee 2013, among others)41
Formal MORPHOLOGICAL CXs: PNM as pro on Inflectable Non-‐Finite Words and Possessed nouns
42
Verbal ac9on nominal inflected word: `I am standing’...
Polyfunc,onality of PNM:Person/Number Marker: func9ons as incorporated SUBJ pro on all inflectable non-‐finiteverbs, except PRCs.
Possessed noun word:`my reindeer’...
Polyfunc,onality of PNM:Person/Number Marker:func9ons as incorporated possessor pro on all possessednouns.
nú-‐ `stand’ nú-‐ `stand’ �-‐ `reindeer’ �-‐ `reindeer’
Func,onal MORPHOLOGICAL CX: SUBJ status of PNMs in non-‐finite clauses 1. SUBJ control of inflec,ng adverbs by inflected modified nouns:
12. ŋəmca-m warʹ′exə-nʹ′i nik°lŋa-w° FINITE CLAUSE meat-acc with.difficulty-1sg bite-obj.sg.1sg. ‘I bit the meat with difficulty.’ (T 321)
13. [warʹ′exə-nʹ′i tolabə-wa-mʹ′i] yexara°-da ADJUNCT CLAUSE with.difficulty-1sg read-impf.an-acc.1sg ignore-obj.sg.3sg ‘He didn’t know that I read with difficulty.’
14a. [warʹ′exe-nʹ′i xada-wi°] te-mʹ′i wæwa POSSESSIVE RELATIVE with.difficulty-1sg kill-perf.part reindeer-1sg bad ‘The reindeer I killed with difficulty is bad’
b. (pidər°) nʹ′a-r° warʹ′exe-ta / *warʹ′exe-t° * POSS CONTROL you brother-2sg with.difficulty-3sg/with.difficulty-2sg ŋəno-m taxabta-da-sʹ′° boat-acc destroy-obj.sg.3sg-past ‘Your brother destroyed the boat with difficulty.’
43
The SUBJ status of PNMs in non-‐finite clauses: PRC behaves like a typical non-‐finite clause 2. Differen,al behavioral effects concerning agreement within two types of coordinate construc,ons
Type 1 coordina,on: Y [X-acc nʹ′acʹ′°] V-sg/*du
Finite clauses:
15. Wasʹ′a nʹ′a-mta nʹ′acʹ′° to° / *toŋa-x°h Wasya friend-acc.3sg with come.3sg /come-3du ‘Wasya came with a friend.‘ Subj agreement
16. *Maša-h Wasʹ′a-m nʹ′acʹ′° nʹ′abako-m menʹ′eə-d°m Masha-gen Wasya-acc with sister-acc love-1sg ‘I love [Masha and Wasya]’s sister.‘ *Obj agreement
44
The SUBJ status of PNMs in non-‐finite clauses
Type 2 coordina,on: Y(-du) [X-gen nʹ′ah ] V-sg/du
Finites clauses:
17. Wasʹ′a nʹ′a-nta nʹ′ah toŋa-x°h/to° Wasya friend-gen.3sg with come-3du/come.3sg ‘Wasya and his friend came / Wasya came with his friend.‘ Subj agreement
18. Petʹ′a-m Wanʹ′a-nta nʹ′ah adorŋa-x°yu-da-sʹ′°/*ladorŋa-da-sʹ′° Petya-acc Wanya-gen.3sg with beat-obj.du-past.3sg/beat-obj.sg.3sg-past ‘He beat up Petya and Vanya.‘ Obj agreement
45
The SUBJ status of PNMs in non-‐finite clauses
Possessive rela9ves:
Type 1 coordina,on: Y [X-acc nʹ′acʹ′°] V-sg/*du
19. Petʹ′a-m nʹ′acʹ′° lad°-wi° wenʹ′ako-r°/*wenʹ′ako-rʹ′ih Petya-acc with hit-perf.part dog-2sg/dog-2du ‘the dog you (sg) and Peter hit’
Type 2 coordina,on: Y(-du) [X-gen nʹ′ah ] V-sg/du
20. [Maša-h nʹ′abako-nta nʹ′ah sed°-wi°] səwa-da / səwa-dʹ′ih Masha-gen sister-gen.3sg with sew-perf.part hat-3sg / hat-3du ‘the hat that Masha and her sister made’
46
Interac,on: Referen,al adverbs and coordina,on
Type 1 coordina,on: Y [X-acc nʹ′acʹ′°] V-sg/*du
21. [(pidar°) Petʹ′a-h nʹ′acʹ′° warʹ′exe-t° / *warʹ′exe-tʹ′ih you Petya-gen with with.difficulty-2sg / with.difficulty-2du
xada-wi°] te-r° / *te-rʹ′ih kill-perf.part reindeer-2sg / *reindeer-2du ‘the reindeer which you killed with Petya’
Type 2 coordina,on: Y(-du) [X-gen nʹ′ah ] V-sg/du
22. [(pidar°) Petʹ′a-h n°ah warʹ′exe-tʹ′ih / warʹ′exe-t° you Petya-gen with with.difficulty-2du/with.difficulty-2sg
xada-wi°] te-rʹ′ih / te-r° kill-perf.part reindeer-2du / reindeer-2sg ‘the reindeer which you and Petya killed’
47
Poten,a,ng possessive rela,ve clauses: why do PRCs look like possessive construc,ons?
In Tundra Nenets, as elsewhere in Eurasia, “possessive” construc9ons, are used to express a wide range of rela9ons between two nominals.
“. . . a nominal in the geni9ve case used in the expression of adnominal determina9on designates not only possession in the substan9ve sense of this word, but also a rela9on (relevance), concerning a characteris9c of one en9ty with respect to another en9ty. -‐ Tereshchenko 1956:64
23a. te-‐w° b. �-‐n° reindeer-‐1SG reindeer-‐Pl.1SG ‘my (one) reindeer‘ ‘my (many) reindeer‘ (the one(s) I own, sit on..)
24. 9-‐h ya reindeer-‐GEN soup `deer soup’ (for the reindeer, made of reindeer...)
25. Wata-‐h ya Wata-‐gen soup `Wata’s soup’ (the soup he cooked/eats/like...)
“Possessive construc=ons” represent a vague two place associa=ve rela=on ℜ between a ‘possessor’ and a `possessed’ arg(ument)
48
The seman,cs of possession
In EXTRINSIC POSSESSION (Barker, 1995; Jackendoff, 1977; Partee, 1997; Partee and Borschev, 2003) the precise nature of the associa9ve rela9on is determined pragma9cally or contextually.
Given a classifica9on of two types of nominals,
possessed inflected words provide a rela9on designated by ℜ which is contextually specified, while the related absolute form of the word does not:
49
Morphological Relatedness
Morphological Noun Subtypes
A schema,c example
The vague value of ℜ in the nominal type Noun-‐Possessed is contextually constrained by by discourse considera9ons.
Hypothesis: The Possessive Rela9ve construc9on exploits the vagueness of the ℜ associa9ve rela9on in noun-‐possessed type nominals, restric9ng its seman9cs and syntax to the seman9cs and syntax on the non-‐finite modifier in the rela9ve construc9on.
Noun -‐ Absolute Type 9 `deer’
Noun -‐ Possessed Type
�-‐w° < NP POSS-‐ED, NPPOSS-‐OR.1SG.PRO, >
Contextually determined 2-‐place ℜ rela9on between N (reindeer) and 1st person singular pronominal possessor
50
Pu�ng words together into phrases: Modifier Construc,ons
Modifiers obligatorily agree with the head for number (singular/dual/plural)
Modifiers op9onally agree with the case of the head and the person/number of the possessor (see Nikolaeva 2008, 2014).
26. (pidəro) serako(ro) te-‐ro you.SG white-‐2sg reindeer-‐2sg `your (sg) white reinder’
27. (pidəro) serako-‐q/serako-‐do �-‐do you.sg white-‐pl/white-‐pl.2sg reinder-‐pl.2sg. `your (sg) white deer (pl)’
Non-‐finite modifiers in PRCS show op9onal concord, just like simple adjec9val modifiers.
28. yəda-‐we-‐(m´i) ŋaqŋo-‐m´i shoot-‐PERF.PART-‐1SG duck-‐1SG ‘the duck I shot’
51
Pu�ng words together into phrases: simple Modifier-‐Head Construc,ons
52
!
your white reindeer
Specifying the associa,ve rela,on in PRCs
[[ øGAP … ta-‐wi°(-‐da) ]LOCAL DOMAIN te-‐da ]EXTERNAL DOMAIN giveMC(-‐3SG) reindeer-‐3SG.PRO ‘the reindeer he/she gave’
The possessed nominal `deer’ is associated with a rela9on ℜ whose value is determined by the meaning of the non-‐finite verbform `give’, i.e., ℜ is lexically restricted within the possessive rela9ve construc9on.
PNM interpreted as bearing the seman9c role “giver” and the gramma9cal rela9on SUBJ. (oaen an implicature that the modified N is “possessed” by the SUBJ.)
The conven9onally vague seman9cs of the possessive nominal provides the opportunity for the rela9ve construc9on to specify a lexical meaning supplied by the non-‐finite verb.
The POSSESSIVE RELATIVE strategy exploits the vagueness of the ℜ rela9on of possessives, restric9ng its seman9cs and syntax to the seman9cs and syntax of the non-‐finite verbal modifier.
53
Pu�ng the pieces together: Modifica,on and PRCs
54
the reindeer I killed
3. Conclusions
55
Conclusions
Guiding intui,on: Cross-‐linguis9cally, languages with Possessive Rela9ves (PRC) contain the same four independent contribu9ng construc9ons with language specific encodings.
Morphological construc,ons
Syntac,c construc,ons
56
!
Conclusions
A ques9on: What do you do if you constantly encounter phenomena that are unexpected or precluded by the standard canny simplifica9ons in linguis9c theory, i.e,. rendered anomalous, which themselves appear to do more than minimum damage to understanding, but don’t want to flee?
An answer: Look to how independent construc9on types redeploy their pieces in new combina9ons within par9cular (classes of) gramma9cal systems,
borrowing ideas and tools from developmental sciences in which complex objects are intepretable as the dynamic reuse of old pieces for new purposes.
Benveniste’s analy9c hunch: “the rela,ve clause o]en has, in a given linguis,c system, the same formal marks as another syntagm of a denomina,on so en,rely different that no one would think that they could be related,”
reflects the intui9on guiding a construc9on-‐theore9c approach.
57
Thanks for listening
58