Lexical and grammatical aspect in language acquisition processing

Lexical and grammatical aspect in language acquisition, processing, and disorders Yasuhiro Shirai Case Western Reserve University (yasuhiro. shirai@case. edu) LABEX Lecture 1 Sorbonne Nouvelle University - Paris 3, March 27, 2018

Overview • • Language bioprogram hypothesis Input distribution as alternative explanation The prototype hypothesis Implications for research on tense-aspect and other areas of language research

Language Bioprogram Hypothesis (Bickerton 1981, 1984, 1999) • Pidgin (simplified contact varieity) -> Creole (pidgin acquired as L 1) • Creole languages develop complex structures not present in pidgin • Commonalties among genetically unrelated creole languages => bioprogram explanation (cited by Pinker, Jackendoff as evidence for innateness hypothesis)

Tense-aspect as one area of commonality: PNPD and SPD Terminology: Four classes of lexical aspect _______ State Activity ~~~~ Accomplishment ~~~~~~x Achievement 　 x love, contain, know, think that. . run, walk, swim, think about. . . paint a picture, build a house fall, drop, win the race (e. g. Vendler 1957)

PNPD (punctual-nonpunctual distinction) SPD (state-process distinction) 1) grammatical encoding in unrelated creole languages 2) additional support from language acquisition data

Punctual Non-Punctual Distinction in L 1 acquisition --> Children mark punctuality by using past tense markers (French: Bronchart & Sinclair 1973, Cognition) (Italian: Antinucci & Miller, 1976, JCL)

State-Process Distinction in L 1 acquisition --> Children never make errors of attaching progressive marker to stative verbs (*I am knowing it. ) (English: Brown 1973)

The Aspect Hypothesis (Shirai 1991, Andersen & Shirai 1994) • State Ach • past/perfective 4 • imperf. past 1 ---> • progressive X Activity Acc <---3 <--- 2 <---1 2 ---> 3 ---> 4 1 ---> 2 ---> 3 (Shirai 1995, BUCLD) • a. k. a: Defective Tense Hypothesis, Aspect Before Tense Hypothesis, Aspect First Hypothesis, Primacy of Aspect Hypothesis, Lexical Aspect Hypothesis

Input distribution as explanation Shirai, Y. (1994). On the overgeneralization of progressive marking on stative verbs: Bioprogram or input? First Language, 14, 67 -82. Shirai, Y. & Andersen, R. W. (1995). The acquisition of tense/aspect morphology: A prototype account. Language, 71, 743 -62. Li, P. & Shirai, Y. (2000). The acquisition of lexical and grammatical aspect. Berlin: Mouton de Gruyter. Shirai, Y. (2009). Temporality in first and second language acquisition. In W. Klein & P. Li (Eds. ) The expression of time. Berlin: Mouton de Gruyter. Chen, J. & Shirai, Y. (2010). The development of aspectual marking in Mandarin Chinese. Applied Psycholinguistics.

• “Note that this does not reflect anything in Italian grammar; all Italian verbs, whether punctual or nonpunctual, activity or change-of-state verbs, have both perfective and imperfective past tenses” (Bickerton 1981, p. 174) • However…. • Distribution of verb morphology is not even in relation to verb classes in native speech

Table 1. Distribution of past forms and inherent aspect in Italian adult-adult speech (based on Leone, 1990) token count (%) telic atelic Perfective past Imperfective past 97 28 type count (%) telic atelic 3 95 5 72 42 58

Table 2. Distribution of past tense morphology and inherent aspect in children's and mothers' speech at the earliest stage (percentage based on token count) State Activity Accomplishment Achievement Children 0 2 0 98 Mothers 17 10 13 60

Development of Progressive Marking --> also driven by skewed distribution in the input

Table 3. Distribution of inherent aspect with progressive inflections in motherese and the children's speech at stage 1 (raw token frequencies in parentheses) • • Adam (2; 3 -2; 4) State Mother 0% (0) Child 0% (0) Activity 51% (26) 58% (11) Accomplishment 14% (7) 10% (2) Achievement 35% (18) 32% (6) • • Eve (1; 6 -1; 7) State Mother 0% (0) Child 0% (0) Activity 53% (38) 75% (9) Accomplishment 8% (6) 0% (0) Achievement 39% (28) 25% (3) • • Naomi (1; 6 -1; 10) State Mother 3% (4) Child 4% (5) Activity 65% (93) 68% (93) Accomplishment 12% (18) 4% (5) Achievement 20% (29) 24% (33)

Can children use progressive marking just like adults do right from the beginning? • NO • Analysis of the first half of stage one reveals that children’s use of progressive is mostly restricted to activities (e. g. singing) and iterative achievements (e. g. jumping) • This restriction is relaxed in the second half of stage 1.

SPD: Overgeneralizaiton of progressive on stative verbs? • Yes, for one child (Naomi) in particular • Naomi’s mother used stative progressives but not the other two mothers • Adam • Eve • Naomi Child 0. 2% (1) 1. 7% (5) 3. 4% (23) Mother 0% (0) 3. 8% (20)

Summary: Tense-aspect • Acquisition patterns of tense/aspect morphology can be explained by input distribution without resorting to a bioprogram. • Context of learning as an alternative explanation to learner-internal constraints on acquisition

• What is the mechanism of such input -based learning?

Distributional learning and prototype-based initial representations Budwig (1996, p. 143) notes that although language acquisition research has shown that children link the use of particular linguistic forms with particular meanings/functions in ways not necessarily identical with the adult model, little investigation has been undertaken regarding the basis for such linkage.

Pronoun case acquisition in English (Budwig 1986) • I ---> low agentivity/control (e. g. I like, I think. . ) • me/my -->high agentivity/control (e. g. Me open that. )

Input distribution (Budwig 1996) 59% of I ---> state verbs (I think, I like. . . ) 61% of I ---> non-control act

Surprising frequency bias in native speech • Hopper (1997): gap between what we think we say and what we actually say. • A) The book that John bought today • B) The man that John met today • B is extremely infrequent in native speech (Ozeki & Shirai 2007, SSLA; Mak et al. 2002, JML)

Relative skewing (e. g. 60%) in the input ==> Absolute form-meaning mapping (almost 100%) - Prototype formation based on the skewed distribution in the input - Connectionist simulation yields similar results (Li & Shirai 2000)

Other grammatical morphemes 1. Japanese causative morphology -sase (Shirai et al. 2000, CLRF) Early restriction to indirect causation (permissive/assistive) -sase in the input to children is almost exclusively used for indirect causation

2. Conditionals in Japanese/Korean (Akatsuka & Clancy 1993, JK 2) - Early acquisition of D-conditionals (permission, prohibition, etc. ) e. g. Tabecha dame "It's no good if (you) eat” - High frequency of D-conditionals in the input as opposed to regular conditionals - Frequency effect : One child whose parents used both with equal frequency acquired both types of conditionals around the same time

3. Nominative Case Omission in Japanese and Korean • Japanese -ga: Miyamoto, Wexler, Aikawa, & Miyagawa (1998, BUCLD) Higher rate of Nominative case omission for the subject NP of unaccusative verbs

Korean data (Lee, Wexler, & Baek, 2000, ICKL) Omission rate of nominative -ka by the child JK Unaccusative Phase 1 (2; 0) 90. 1% Phase 2 (2; 1 -2; 4) 75. 7% Phase 3 (2; 5 -2; 11) 48. 1% Non-unaccusative 83. 3% 36. 8% 14. 7%

Nominative Case Omission in Adult Speech (JK's mother) Child-directed speech 45% (unaccusative) vs. 25. 1% (non-unaccusative) Adult-adult speech 12. 6% (unaccusative) vs. 8. 5% (non-unaccusative)

Objection to the prototype scenario • Is it just the discourse context of spontaneous interaction that inflates the skewed distribution in learner speech? (e. g. Weist 1989, FL)

Experimental studies also suggest children's initial representation is restricted to prototypes. • aspect – (Li & Bowerman 1998, FL, Stoll 1998, FL, Wagner 2000, JCL) • causative -sase – (Morikawa & Ono 1997, SRCD)

Tomasello's Usage based approach Verb Island Hypothesis – Early restriction of grammatical forms are based on item-based learning My proposal – Early restriction results from restricted semantic representation (evidence from overregularization of past tense forms)

So, where does such frequency bias come from? • It is how the particular construction is used in language. (Naturalness of combination, Comrie 1976, Discourse motivation, Andersen & Shirai, 1994, SSLA) • Brown (1973) pointed out that past marking started with "a small set of verbs which name events of such brief duration that the event is almost certain to have ended before one can speak of. . It is reasonable to guess that these forms may have been always or almost always in the past in mother's speech" (p. 334).

Learning context as explanation • Why has input not been given importance in L 1 acquisition research? A) early correlational research (Brown, Newport et al) B) influential theories of acquisition (Chomsky, Slobin) • Comeback of input argument connectionism Tomasello’s usage-based model Language-specificity (Bowerman, Choi, etc. ) Statistics with UG (Yang 2004, 2016)

Semantic bias in sentence processing • Madden & Zwaan (2003, M&C) - perfective facilitation • Ferretti, Kutas & Mc. Rae (2007, JEP) - imperfective facilitation of locations --> They are only concerned with asymmetry regarding grammatical aspect.

However… • Madden & Zwaan only used accomplishment verbs, which is congruent with perfective aspect • Yap et al. (2009, M&C) found: - perfective facilitation with accomplishment - imperfective facilitation with activity Ferretti et al. included activity (atelic) and accomplishment (telic) in a single experiment

Aspect in Cantonese SLI • Low productivity of perfecitve zo 2 for SLI children (token frequency was comparable with age-match, but not type frequency) • SLI children’s use of aspect markers were more restricted to prototypical combination (e. g. zo 2 with Achievement) (Stokes & Fletcher 2000 Linguistics, 2003, JSLHR) (Fletcher, Leonard, Stokes & Wong, 2005, JSLHR)

Tense-aspect in English SLI (Leonard et al. 2007, JSLHR) • No effect of telicity for SLI children! Children not performing at ceiling (i. e. 100% correct) prototype -ed Non-proto -ed SLI 37 33 prototype -ing Non-proto -ing 58 64 　 TD-MLU 83* 69 TD-Age 97* 85 59* 41 81* 65 *significant effect of lexical aspect (p. <. 05)

Conclusion by Leonard et al. • Difficulties … may be compounded in children with SLI if they fail to make use of associations between the lexical aspect of verb prediates and the grammatical function of the accompanying inflections [as the starting point for learning tenseaspect inflections].

Conclusion of this talk • Semantic bias exists in any areas of tenseaspect, whether acquisition or processing, normal or disorder. • In constructing stimulus sentences in experiments, care should be taken so that there is no particular semantic bias in one direction or the other.

• Shirai, Y. (2011). Lexical and grammatical aspect in language acquisition, processing, and disorders. In S. Arita et al. (Eds. ), Studies in Language Sciences, 10 (pp. 41 -57). Tokyo: Kurosio Publishers.