Janet L. McDonald

Language learning: Cues or rules?

Child language researchers have often taken gender and case paradigms to be interesting test cases for theories of language learning. In this paper we develop a computational model of the acquisition of the gender, number, and case paradigm for the German definite article. The computational formalism used is a connectionist algorithm developed by Rumelhart, Hinton, and Williams (1986. In D. Rumelhart & J. McClelland (Eds.), Parallel Distributed Processing; Explorations in the Microstructure of Cognition. Cambridge, MA: MIT Press). Three models are developed. In the first two, various cues to gender studied by Kopcke and Zubin (1983, Zeitschrift fur germanistiche Linguistik, 11, 166–182; 1984, Linguistiche Berichte, 93, 26–50) are entered by hand. In the third, the simulation is given only the raw phonological features of the stem. Despite the elimination of the hand-crafting of the units, the third model outperformed the first two in both training and generalization. All three models showed a good match to the developmental data of Mills (1986, The acquisition of gender: a study of English and German. Berlin: Springer-Verlag) and MacWhinney (1978, Monographs of the Society for Research in Child Development, 43, whole No. 1). Advantages of a connectionist approach over older theories are discussed.

Grammaticality judgments in a second language: Influences of age of acquisition and native language

Native Spanish early and late acquirers of English as well as native Vietnamese early and child acquirers of English made grammaticality judgments of sentences in their second language. Native Spanish early acquirers were not distinguishable from native English speakers, whereas native Spanish late acquirers had difficulty with all aspects of the grammar tested except word order. Native Vietnamese early acquirers had difficulty with those aspects of English that differ markedly from Vietnamese. Native Vietnamese child acquirers had more generalized problems, similar to those of native Spanish late acquirers. Thus, native language appeared to make a difference for early acquirers, whereas a later age of acquisition caused a more general problem. A processing-based model focusing on the difficulty non-native language learners have in rapidly decoding surface form is offered as a possible explanation for both effects.

Sentence Interpretation in Bilingual Speakers of English and Dutch.

Speakers of English and Dutch vary in how strongly they use various syntactic (e.g., word order, prepositions, case inflection) and semantic (e.g., noun animacy) cues to interpret native language sentences. For example, in simple NVN sentences, English speakers rely heavily on word order, while Dutch speakers rely on case inflection. This paper compares the cue usage of English/Dutch and Dutch/English bilinguals with varying amounts of second language exposure to that of native speaker control groups. For all constructions tested, dative constructions, simple NVN sentences, and relative clauses, it was found that with increasing exposure, cue usage in the second language gradually shifts from that appropriate to the first language to that appropriate to the second. A model of cue learning originally proposed to account for monolingual data is found to be compatible with the learning pattern exhibited by bilinguals.

Grammaticality judgments in children: the role of age, working memory and phonological ability.

This paper examines the role of age, working memory span and phonological ability in the mastery of ten different grammatical constructions. Six- through eleven-year-old children (n=68) and adults (n=19) performed a grammaticality judgment task as well as tests of working memory capacity and receptive phonological ability. Children showed early mastery of some grammatical structures (e.g. word order, article omissions) while even the oldest children differed from adults on others (e.g. past tense, third person singular agreement). Working memory capacity and phonological ability accounted for variance in grammaticality judgments above and beyond age effects. In particular, working memory capacity correlated with structures involving verb morphology and word order; phonological ability was important for structures with low phonetic substance. Childrens relative difficulty with the different constructions showed parallels to adult performance under memory load stress, indicating working memory capacity may be a limiting factor in their performance. Implications for performance by memory and phonologically impaired populations are discussed.

Levels of learning: A comparison of concept formation and language acquisition

Abstract The Competition Model of Bates and MacWhinney explains how multiple cues may be acquired and used in assigning linguistic roles in natural language sentences. This paper extends the domain of this model to the nonlinguistic realm by examining the acquisition of categories in a concept learning task. As in the linguistic domain, classification in this particular concept learning task is determined by multiple probabilistic cues. On any particular instance, a cue may or may not be present. Moreover, if a cue is in conflict with another, stronger cue, it may not indicate the correct classification. Error rates and reaction times on this type of concept learning task show a two stage pattern of development. People first rely on cues that most often give the correct classification over all the instances seen. When errors persist, people adjust the strengths of the cues to reflect the relative strengths cues have in conflict situations. The results from this laboratory concept learning task mirror those found in the natural language domain, underscoring the generality of the learning mechanism postulated in the Competition Model

Past tense grammaticality judgment and production in non-native and stressed native English speakers

This paper explores whether the poor mastery of morphosyntax exhibited by second language (L2) learners can be tied to difficulties with non-syntactic processing. Specifically, we examine whether problems with English regular and irregular past tense are related to poor L2 phonological ability and lexical access, respectively. In Experiment 1, L2 learners showed poorer past tense mastery than native English speakers in grammaticality judgment and production tasks. L2 phonological ability was positively correlated with correct performance on regular verbs and negatively with unmarked production. L2 lexical access was positively correlated with correct performance on irregular verbs, and negatively with overregularization production. Experiment 2 simulated these difficulties in native English speakers by placing them under phonological processing (noise) or lexical access (deadline) stress. Noise selectively impacted regular verbs in grammaticality judgment but impacted all verb types in production. Deadline pressure impacted irregular verbs while sparing regular verbs across both tasks. Thus, non-syntactic processing difficulties can have specific impacts on morphosyntactic performance in both non-native and native English speakers.

Sentence Recall by Children With SLI Across Two Nonmainstream Dialects of English

PURPOSE The inability to accurately recall sentences has proven to be a clinical marker of specific language impairment (SLI); this task yields moderate-to-high levels of sensitivity and specificity. However, it is not yet known if these results hold for speakers of dialects whose nonmainstream grammatical productions overlap with those that are produced at high rates by children with SLI. METHOD Using matched groups of 70 African American English speakers and 36 Southern White English speakers and dialect-strategic scoring, we examined childrens sentence recall abilities as a function of their dialect and clinical status (SLI vs. typically developing [TD]). RESULTS For both dialects, the SLI group earned lower sentence recall scores than the TD group with sensitivity and specificity values ranging from .80 to .94, depending on the analysis. Children with SLI, as compared with TD controls, manifested lower levels of verbatim recall, more ungrammatical recalls when the recall was not exact, and higher levels of error on targeted functional categories, especially those marking tense. CONCLUSION When matched groups are examined and dialect-strategic scoring is used, sentence recall yields moderate-to-high levels of diagnostic accuracy to identify SLI within speakers of nonmainstream dialects of English.

Differences in the cognitive demands of word order, plural, and subject-verb agreement constructions

The syntactic devices of subject-verb-object word order, regular plurals, and subject-verb agreement differ in age of acquisition and susceptibility to error within language-disordered populations. In the present article, the performance of adults on a grammaticality judgment task is used to explore whether such differences are related to working memory (both in terms of an externally imposed load and individual differences in capacity) and phonological ability. The results show that word order, the earliest acquired and most resilient device, is not affected by load, memory span, or phonological ability. Plurals are affected marginally by load and significantly by phonological ability. Agreement, the last acquired and least resilient device, is affected by load, memory span, and phonological ability. Thus, consistent with a processing-based explanation, later acquired and less resilient devices have higher working memory and phonological demands.

Retroactive interference from translation equivalents: Implications for first language forgetting

First language vocabulary is vulnerable to forgetting after massive exposure to a second language. Two possible factors responsible for the forgetting are degree of semantic overlap between concepts in the two languages and amount of second language exposure. In a laboratory simulation of the language forgetting situation, participants received 10 exposures to a list of words in a foreign language, followed by 2, 5, 10, or 15 exposures to a list in a second foreign language. The second list consisted of either translation equivalents or new concepts. Participants were then tested for retention of the first list. More retroactive interference was found for translation equivalents than for new concepts and for higher degrees of exposure to the second list. When retention of the first list was broken down in terms of gains and losses, effects of both similarity of the second list to the first and amount of exposure to the second list were found only for losses—a fact that points to lack of discriminability as one of the underlying causes of forgetting. Overall, the experimental paradigm proved useful for exploring and developing theories about the causes of first language forgetting.

Changes in Sentence Processing as Second Language Proficiency Increases

Abstract Due to differences between the grammars of English and French, native speakers of these languages rely differently on the cues of word order, verb agreement and noun animacy in assigning the actor role. In this chapter we examine how native English speakers gain mastery over appropriate French strategies with increasing second language proficiency. The time course of this mastery includes initial abandonment of English word order strategies followed by later development of appropriate French word order strategies, and even later strengthening of an appropriate verb agreement strategy. We show how this sequence of development is tied to the properties of French input these learners receive.