Linguistics Department

GURT Practice Talks - Cynthia P. Blanco (Linguistics) and Libby Gertken & Laura Ambrosetti (French and Italian)

Wed, March 7, 2012 | PAR 302

12:00 PM - 1:00 PM

This Wednesday, March 7th, at noon, Libby Gertken (Department of French & Italian), Laura Ambrosetti (Department of French & Italian), and Cynthia Blanco will be giving practice presentations in preparation for the Georgetown University Roundtable on Language and Linguistics (GURT), which will be taking place this weekend in Washington, DC. We'd appreciate your feedback as we continue polishing our presentations! Please find our abstracts below.

Cynthia P. Blanco - Corpus evidence for the effects of lemma and phonology on the Spanish imperfect subjunctive alternation

In Spanish the imperfect subjunctive can be conjugated in either of two ways, ending in –ra or –se (e.g., hablara and hablase ‘(that) (I/he) would speak’). While –ra forms are far more common than their –se counterparts, the nature of and motivation for their distribution has largely focused on dialectal variation, typically pitting Peninsular Spanish, known for high proportions of –se usage [1,2], against Latin American dialects, some of which (e.g. Mexican Spanish) have been reported to seldom use –se if at all [2,3]. Some studies [1,3,4] have also incorporated linguistic factors, such as the four most frequent verbs, clause polarity, or clause type, but no study has either considered all these factors at once or looked at verb frequency or lemma beyond the four most frequent verbs. In the studies that have systematically explored the influence of some of these factors using corpora, tokens have been few and collection methods have varied widely. The current project provides a unified, comprehensive corpus analysis including lemma and frequency using the Spanish Wikicorpus [5], which is composed of 2006 Spanish-language Wikipedia articles representing the entire Spanish-speaking world in a single corpus.

The Wikicorpus yielded 96,085 imperfect subjunctive tokens, and each token was coded for its conjugation (-ra or –se), as well as morphological, lexical, and syntactic factors. Overall, over 22% of tokens in the corpus were conjugated with–se, a total higher than any reported in previous dialectal studies, including Spain [1,3]. Partially confirming earlier results, the three most predictive factors in a logistic regression model were lemma, person-number inflection, and idiomatic usage. The present study looked at lemmas beyond the four most frequent verbs, showing that –se is significantly more common in certain lemmas than in others. Furthermore, the impact of lemma may be at least partly explained by phonology, a variable absent from previous work. This study offers evidence of dissimilation such that verbs whose roots end with the sound of one of the imperfect subjunctive endings are significantly more likely to be conjugated using the more dissimilar ending. The person-number factor added significant predictive power, also possibly due to phonology of these morphological inflections. Unlike the overall model, models for the previously-studied frequent verbs that included the syntactic factors polarity and clause type predicted the –ra/–se distribution better than models without syntactic factors, reinforcing a difference in behavior based on frequency and lemma.

[1] Asratián, A. (2007). Variación –ra/-se en el español hablado en Caracas. Boletín de Lingüística 19 (27), 5-41.

[2] Dale, G.I. (1925). The Imperfect Subjunctive. Hispania 8 (2), 127-129.

[3] DeMello, G. (1993). -Ra vs. -Se Subjunctive: A New Look at an Old Topic. Hispania 76 (2), 235-244.

[4] Fernández Ulloa, T., & Portillo Mayorga, R. (2000). Uso del imperfecto y pluscuamperfecto de subjuntivo entre estudiantes universitarios de Bilbao (Pais Vasco) y Santander (Cantabria). Onomázein 5, 167-187.

[5] Reese, S., Boleda, G., Cuadros, M., Padró, L., & Rigau, G. (2010). Wikicorpus: A Word-Sense Disambiguated Multilingual Wikipedia Corpus. In Proceedings of 7th Language Resources and Evaluation Conference (LREC'10), La Valleta, Malta.

Libby Gertken & Laura Ambrosetti (French and Italian) - Good Enough Processing in French as a First and Second Language

Recent evidence suggests that second language (L2) sentence processing is driven largely by lexical and pragmatic cues [3], whereas native speaker (NS) processing elicits structurally detailed representations. However, several studies [1, 2, 4, 5] have documented that monolingual processing does not consistently entail syntactic detail and accuracy, resulting instead in underspecified representations that are nevertheless systematic. For example, participants in [4] showed surprising inaccuracy when asked to identify the agent or patient of the following sentence: The dog was bitten by the man. The authors concluded that interpretations reflected a probabilistic reading of the content words of the sentence more than its syntactically derived meaning.

In Experiment 1, we analyzed French speakers’ interpretations of active and passive sentences that were either plausible or implausible with the aim of assessing to what extent French speakers use probabilistic parsing heuristics. To our knowledge, these heuristics have not been explored in non-English languages within a Good Enough processing framework [6]. Participants read 72 experimental sentences (e.g. le chien a été mordu par l’homme “the dog was bitten by the man”) displayed on a computer screen and then identified either the agent or patient of the action. Consistent with previous findings for English, French participants were less accurate in thematic role identification for passive than for active structures, and when processing passives, they were less accurate with implausible sentences. These results provide evidence of Good Enough processing among French speakers, whether due to a word-order heuristic whereby noun-verb-noun (NVN) strings are mapped onto agent-verb-patient thematic role sequences or to a plausibility heuristic that results in the processor adopting an analysis consistent with real-world plausibility [4].

In a second experiment currently underway, we compare NS and L2 processing of actives and passives, as well as other non-SVO constructions (i.e. relativized object constructions, right- and left-dislocations) to determine whether English-French bilinguals have a greater tendency to construct superficial representations compared with native French speakers. Response times to agent/patient questions are used as a secondary metric of processing. Since French has less consistent argument ordering patterns than English [7], this study also seeks to draw conclusions about the relevance of an NVN heuristic to French and the possibility of L1-L2 transfer of heuristic routes among English-French bilinguals. We discuss our findings with respect to Good Enough processing approaches and recent claims about fundamental differences between L2 and NS processing.

[1] Christianson, K., Luke, S. G., & Ferreira, F. (2010). Effects of plausibility on structural priming. Journal of Experimental Psychology: Learning, Memory, and Cognition, 36, 538-544.
[2] Christianson, K., Hollingworth, A., Halliwell, J. F., & Ferreira, F. (2001). Thematic-roles assigned along the garden path linger. Cognitive Psychology, 42, 368-407.
[3] Clahsen, H., & Felser, C. (2006). Grammatical processing in language learners. (Keynote article). Applied Psycholinguistics, 27, 3–42.
[4] Ferreira, F. (2003). The misinterpretation of noncanonical sentences. Cognitive Psychology, 47, 164-203.
[5] Ferreira, F., Ferraro, V., & Bailey, K.G.D. (2002). Good-enough representations in language comprehension. Current Directions in Psychological Science, 11, 11-15.
[6] Ferreira, F., & Patson, N. (2007). The good enough approach to language comprehension. Language and Linguistics Compass, 1, pp 71-83.
[7] Lambrecht, K. (1994). Information Structure and Sentence Form. Topic, Focus, and the Mental Representations of Discourse Referents.
Cambridge: Cambridge University Press.

Bookmark and Share