Language-dependent Aspects of Structural Priming Hiroko Yamashita1, Franklin Chang2, & Yuki Hirose3 2 Rochester Institute of Technology, Max Planck Institute for Evolutionary Anthropology, 3 University of Electrocommunication
1
Abstract
Questions
Speakers must choose between different alternative sentence structures in speaking. Studies in English (Bock & Loebell, 1990; Pickering & Branigan, 1998) have shown that this choice can be explained by representations which make use of SYNTACTIC CATEGORIES (e.g. NP-PP , NP-NP, etc.). The equivalent structural choices in Japanese are made up of the same syntactic category (i.e. NP). Using a technique called structural priming, we examined whether Japanese speaker’s dative structural choices may be primed without syntactic category information.
• Can Japanese speakers show structural priming without differences in syntactic categories? • Can the same theory explain priming in English and Japanese? • Does language learning play a part in determining the organization of the sentence production system?
Japanese speakers were more likely to use the dative sentence in the order that we primed, even though the two syntactic structures did not differ in the order of syntactic categories. We also found that Japanese speakers do not relate dative and formally similar transitive structures in the same way that English speakers do.
Methodology: RSVP Structural Priming
The results suggest that Japanese speakers can make use of representations that go beyond syntactic categories. The current results and English data may be better accounted for by a theory in which language learning plays an important part in organizing the processing system (Chang, et al. 2003, March).
Results
(1) Priming of the mapping from meaning into structures The mapping of meaning into structures is needed to explain spray-load priming (Chang, et al. in press). A similar mechanism can be used to explain the differences between English and Japanese dative priming.
This procedure induced changes from the given structure, and these changes allow us to see priming.
mask
In Japanese, adjuncts are treated as obliques (OBQ), hence there is a different mapping to syntactic functions for LOC-ACC than DAT-ACC.
number comparison task
kaita
English Syntactic Alternations Involve Differences in Syntactic Categories. Japanese does not.
Rather they differ in syntactic functions and thematic roles. (DAT-ACC) OBJ2 OBJ Goal Patient (ACC-DAT) OBJ OBJ2 Patient Goal
sentence recall
47139 san?
doozo
•The Japanese dative alternation does not involve a difference in syntactic categories (all are NPs) because the arguments marked with casemarkers -ga, -ni, or -o are standardly treated as noun phrases (Kuno, 1973; Tsujimura, 1996). (DAT-ACC) NP-NP boy-ga [girl-ni]NP [dress-o] NP gave. (ACC-DAT) NP-NP boy-ga [dress-o] NP [girl-ni]NP gave.
feedback
###########
• The English dative alternation involves a difference in syntactic category (NP vs.. PP). Prepositional Dative (PD) NP-PP The boy gave [a dress] NP [to a girl]PP. Double Object Dative (DO) NP-NP The boy gave [a girl] NP [a dress] NP.
LOC-ACC did not prime datives in Japanese. Therefore English and Japanese relate these structures differently. In a mapping account, LOCATION is treated as a type of GOAL (endpoint of a transitive event). The English mapping for the Prep. Dative and Prep. Locative are the same. English Meaning-Form Mappings: Double object Dative SUB(Agent) OBJ(Goal) OBJ2(Patient) Prepositional Dative SUB(Agent) OBJ(Patient) OBQ(Goal) Prepositional Locative SUB(Agent) OBJ(Patient) OBQ(Goal)
sentence (250 msecs per word)
Akiko-wa
Sensei-ni
• Unlike in English, the formally similar locative structure (LOC-ACC) did not prime the DAT-ACC order.
Accounts of Priming in English and Japanese
- Examines the effect of a prime sentence on production of target. - Rapid Serial Visual Presentation (Potter & Lombardi, 1998) - Phrase by phrase presentation (see chart below).
Tegami-o
• Speakers were biased toward the primed structure, even without a difference in syntactic categories or animacy (ACC-DAT is different from DAT-ACC).
Experiment We replicated Bock & Loebell (1990, Exp 1) in Japanese. Double Object: The wealthy widow sold the church an old Mercedes. Prep. Dative: The wealthy widow gave an old Mercedes to the church. Prep. Locative: The wealthy widow drove an old Mercedes to the church.
Japanese Meaning-Form Mappings: ACC-DAT SUB(Agent) OBJ(Patient) OBJ2(Goal) DAT-ACC SUB(Agent) OBJ2(Goal) OBJ(Patient) LOC-ACC SUB(Agent) OBQ(Goal) OBJ(Patient) Meaning is needed to distinguish the formally similar ni-marked phrases OBJ2 and OBQ. Mapping representations allow meaning to influence structure selection in a limited manner. Japanese and English speakers have to learn different mappings, and hence they show different priming patterns. (2) Learning relative strength among influencing factors Another possible account is that the relative importance of syntactic categories and thematic roles in production is determined by the language that is learned. In this approach, English speakers mainly use syntactic category and mapping representations, while Japanese speakers mainly use the order of thematic roles. More work is needed to distinguish these two accounts.
Priming of PD: Prep. Locatives = Prep. Datives > Double Object Datives Results can be accounted for by surface syntactic configuration. Japanese has a similar relationship between Locatives and Datives, and so we can see if priming in Japanese is due to surface syntactic configuration.
Conclusions Using Structural Priming to Examine Syntactic Representations (Bock & Loebell, 1990) Two structures with approximately the same meaning. Prep. Dative (PD): The boy gave a dress to a girl. Double Object Dative (DO): The boy gave a girl a dress. By first presenting a prime sentence in one structure with a different meaning (e.g. PD: The rancher showed the horse to the sheriff), we can bias the speaker’s preference for the PD structure. This biasing can be thought of as priming of structure.
Factors that Influence English Structural Choices Most English studies report structural choices are influenced by syntactic categories or structures, and not by order of thematic roles (Bock & Loebell, 1990; Bock et al., 1992; Pickering, et al., 2002). However, a recent study (Chang, Bock, & Goldberg, in press) found that English speakers can be sensitive to the order of thematic roles, when the alternation does not involve a change in syntactic categories (NP-PP). THEME-LOCATION LOCATION-THEME
The man sprayed [water]NP [onto the wall]PP. The man sprayed [the wall]NP [with water]PP.
Since these two structures share the same syntactic categories (NP-PP), priming here cannot just be due to syntactic categories. Since thematic roles, by themselves, do not seem to strongly guide priming, they argued that what was primed was the mapping of roles into structures.
Stimuli Prime Sentence: ACC-DAT 社は ワゴン車を 老人ホームに 寄贈した。 (dative verb) CEO-TOPIC wagon-ACC retirement-home-DAT presented TOPIC OBJ OBJ2 Agent Patient Goal NOUN-wa NOUN-o NOUN-ni The CEO presented the wagon to the retirement home. DAT-ACC 社は 老人ホームに ワゴン車を 寄贈した。 (dative verb) CEO-TOPIC retirement-home-DAT wagon-ACC presented TOPIC OBJ2 OBJ Agent Goal Patient NOUN-wa NOUN-ni NOUN-o The CEO presented the wagon to the retirement home. LOC-ACC 社は 老人ホームに ワゴン車を 駐車した。 (transitive verb) CEO-TOPIC retirement-home-LOC wagon-ACC parked TOPIC OBQ OBJ Agent Locative Patient NOUN-wa NOUN-ni NOUN-o The CEO parked the wagon at the retirement home. Target Sentence: ACC-DAT おばあさんは 借金を 米屋に 払った。 (dative verb) Grandmother-TOPIC debt-ACC rice-shop-DAT paid TOPIC OBJ OBJ2 Agent Patient Goal NOUN-wa NOUN-o NOUN-ni The grandmother paid the debt to the rice shop.
Targets changed to DAT-ACC order Two Production Approaches Architecture-based approach The architecture determines the type of representations that are used universally. Prediction: Priming depends on syntactic category differences.
Learning-based approach Language learning plays an important role in structuring the architecture of sentence production (Chang et al, 2003, March). Prediction: Priming depends on the representations that have been learned for a particular language.
Proportion DAT-ACC order
0.5 0.45 0.4 0.35 0.3 0.25 0.2 0.15 0.1
• Structural priming was observed without differences in the order of syntactic categories. Results are incongruent with theories that argue that syntactic categories are architecturally privileged in structural priming (Pickering & Branigan, 1998). • Differences in priming between English and Japanese suggest that language learning can structure language processing (Chang, et al. 2003, March). • One possibility is speakers learn how to map meanings into forms in their particular language, and the mappings that they discover can then be primed. Another possible account is that the relative strength of syntactic categories and thematic roles in production is determined by the particular language that is learned.
References Bock, K., & Loebell, H. (1990). Framing sentences. Cognition, 35(1), 1-39. Bock, K., Loebell, H., & Morey, R. (1992). From conceptual roles to structural relations: Bridging the syntactic cleft. Psychological Review, 99(1), 150-171. Chang, F., Bock, J. K., & Goldberg, A. E. (in press). Do thematic roles leave traces of their places? Chang, F., Dell, G. S., & Bock, K. (2003, March). Comprehending structural priming: A connectionist learning account. Poster presented at the CUNY Sentence Processing Conference, Boston, MA. Kuno, S. (1996). The structure of the Japanese language. Cambridge, MA: The MIT Press. Pickering, M. J., & Branigan, H. P. (1998). The representation of verbs: Evidence from syntactic priming in language production. Journal of Memory and Language, 39, 633-651. Pickering, M. J., Branigan, H. P., & McLean, J. F. (2002). Constituent structure is formulated in one stage. Journal of Memory & Language, 46(3), 586-605. Potter, M. C., & Lombardi, L. (1998). Syntactic priming in immediate recall of sentences. Journal of Memory and Language, 38(3), 265-282. Tsujimura, N. (1996). An Introduction to Japanese Linguistics. Cambridge, MA: Blackwell Publishers Inc.
0.05 0
ACC-DAT
DAT-ACC Prime Sentence Type
LOC-ACC (transitive) Presented at the Architecture and Mechanism of Language Processing, 2003, Glasgow, UK