Testing a computational model of causative overgeneralizations: Child judgment and production data from English, Hebrew, Hindi, Japanese and K’iche’

How do language learners avoid the production of verb argument structure overgeneralization errors (*The clown laughed the man c.f. The clown made the man laugh), while retaining the ability to apply such generalizations productively when appropriate? This question has long been seen as one that is both particularly central to acquisition research and particularly challenging. Focussing on causative overgeneralization errors of this type, a previous study reported a computational model that learns, on the basis of corpus data and human-derived verb-semantic-feature ratings, to predict adults’ by-verb preferences for less- versus more-transparent causative forms (e.g., * The clown laughed the man vs The clown made the man laugh) across English, Hebrew, Hindi, Japanese and K’iche Mayan. Here, we tested the ability of this model (and an expanded version with multiple hidden layers) to explain binary grammaticality judgment data from children aged 4;0-5;0, and elicited-production data from children aged 4;0-5;0 and 5;6-6;6 (N=48 per language). In general, the model successfully simulated both children’s judgment and production data, with correlations of r=0.5-0.6 and r=0.75-0.85, respectively, and also generalized to unseen verbs. Importantly, learners of all five languages showed some evidence of making the types of overgeneralization errors – in both judgments and production – previously observed in naturalistic studies of English (e.g., *I’m dancing it). Together with previous findings, the present study demonstrates that a simple learning model can explain (a) adults’ continuous judgment data, (b) children’s binary judgment data and (c) children’s production data (with no training of these datasets), and therefore constitutes a plausible mechanistic account of the acquisition of verbs’ argument structure restrictions.

Maîtrise des clitiques datifs dans les structures bitransitives en français L2 par des apprenants anglophones : influence de la structure argumentale de la L1

The objective of this study is to measure the influence of L1 verb argument structure, as well as verb meaning, on the mastery of dative clitics in French as a second language for a group of Anglophone learners. More specifically, we focus on ditransitive structures. While French and English share the V NP PP structure, English also has a double-object structure, V NP NP, for a subset of verbs. The results of our study show that L1 argument structure influences the mastery of dative clitics in French, especially for verbs that only accept the double-object structure in English. Further, the behaviour of our participants with verbs that accept the dative alternation led us to conduct a follow-up study. The findings show that verb meaning also influences performance with dative clitics, but this effect cannot be explained by L1 influence.

Word Recognition and Word Prediction in Preschoolers With (a Suspicion of) a Developmental Language Disorder: Evidence From Eye Tracking

Purpose This study compares online word recognition and prediction in preschoolers with (a suspicion of) a developmental language disorder (DLD) and typically developing (TD) controls. Furthermore, it investigates correlations between these measures and the link between online and off-line language scores in the DLD group. Method Using the visual world paradigm, Dutch children ages 3;6 (years;months) with (a suspicion of) DLD ( n = 51) and TD peers ( n = 31) listened to utterances such as, “Kijk, een hoed!” ( Look, a hat! ) in a word recognition task, and sentences such as, “Hé, hij leest gewoon een boek” (literal translation: Hey, he reads just a book ) in a word prediction task, while watching a target and distractor picture. Results Both groups demonstrated a significant word recognition effect that looked similar directly after target onset. However, the DLD group looked longer at the target than the TD group and shifted slower from the distractor to target pictures. Within the DLD group, word recognition was linked to off-line expressive language scores. For word prediction, the DLD group showed a smaller effect and slower shifts from verb onset compared to the TD group. Interestingly, within the DLD group, prediction behavior varied considerably, and was linked to receptive and expressive language scores. Finally, slower shifts in word recognition were related to smaller prediction effects. Conclusions While the groups' word recognition abilities looked similar, and only differed in processing speed and dwell time, the DLD group showed atypical verb-based prediction behavior. This may be due to limitations in their processing capacity and/or their linguistic knowledge, in particular of verb argument structure.

Verb argument structure overgeneralisations for the English intransitive and transitive constructions: grammaticality judgments and production priming

We used a multi-method approach to investigate how children avoid (or retreat from) argument structure overgeneralisation errors (e.g., *You giggled me). Experiment 1 investigated how semantic and statistical constraints (preemption and entrenchment) influence children’s and adults’ judgments of the grammatical acceptability of 120 verbs in transitive and intransitive sentences. Experiment 2 used syntactic priming to elicit overgeneralisation errors from children (aged 5–6) to investigate whether the same constraints operate in production. For judgments, the data showed effects of preemption, entrenchment, and semantics for all ages. For production, only an effect of preemption was observed, and only for transitivisation errors with intransitive-only verbs (e.g., *The man laughed the girl). We conclude that preemption, entrenchment, and semantic effects are real, but are obscured by particular features of the present production task.

The wrong horse was bet on – The Effects of Argument Structure versus Argument Adjacency on the Processing of Idiomatic Sentences

Psycholinguistc research remains puzzled by the question under what circurmstances syntactically transformed idioms keep their figurative meaning. In this study we examined the effects of verb argument structure and argument adjacency on the processing of idiomatic and literal sentences in German. In two sentence-completion experiments, participants listened to idiomatic and literal sentences, both in active and passive voice, without the sentence-final verb. They indicated via button-press, which of three visually presented verbs best completed the sentence.In both experiments, idiomatic sentences were processed faster than literal ones, and active sentences faster than passive ones. In passivized sentences, the patterns of argument structure and argument adjacency reversed across experiments: In Experiment 1, sentences with ditransitive verbs were processed faster than sentences with transitive verbs, and vice versa in Experiment 2. This pattern corresponds to faster processing of adjacent than of nonadjacent arguments and thus points to the dominating role of argument adjacency rather than argumentstructure in the processing of passivized sentences. With respect to idiom processing, we conclude that the adjacency of the verb and its arguments determines whether passivized idioms keep their figurative meaning.

Testing a computational model of causative overgeneralizations: Child judgment and production data from English, Hebrew, Hindi, Japanese and K’iche’

How do language learners avoid the production of verb argument structure overgeneralization errors (*The clown laughed the man c.f. The clown made the man laugh), while retaining the ability to apply such generalizations productively when appropriate? This question has long been seen as one that is both particularly central to acquisition research and particularly challenging. Focussing on causative overgeneralization errors of this type, a previous study reported a computational model that learns, on the basis of corpus data and human-derived verb-semantic-feature ratings, to predict adults’ by-verb preferences for less- versus more-transparent causative forms (e.g., *The clown laughed the man vs The clown made the man laugh) across English, Hebrew, Hindi, Japanese and K’iche Mayan. Here, we tested the ability of this model to explain binary grammaticality judgment data from children aged 4;0-5;0, and elicited-production data from children aged 4;0-5;0 and 5;6-6;6 (N=48 per language). In general, the model successfully simulated both children’s judgment and production data, with correlations of r=0.5-0.6 and r=0.75-0.85, respectively, and also generalized to unseen verbs. Importantly, learners of all five languages showed some evidence of making the types of overgeneralization errors – in both judgments and production – previously observed in naturalistic studies of English (e.g., *I’m dancing it). Together with previous findings, the present study demonstrates that a simple discriminative learning model can explain (a) adults’ continuous judgment data, (b) children’s binary judgment data and (c) children’s production data (with no training of these datasets), and therefore constitutes a plausible mechanistic account of the retreat from overgeneralization.

Verbal Argument Structure in the War Scroll

This essay presents the results of an extended study of verbal argument structure in the War Scroll (1QM). I first establish a method based in generative linguistic theory. I then illustrate this method with a discussion of the argument structure of Qal ‮יצא‬‎ in 1QM and other Dead Sea Scrolls. Following this case study, I present the data from 1QM on verb argument structure—specifically, instances where 1QM adds evidence that is not covered in previous studies of the Dead Sea Scrolls. 1QM presents few developments from earlier Hebrew; I argue that such continuity is significant. I conclude with reflections on the implications of argument structure in 1QM for the study of ancient Hebrew.

Automation of the Northwestern Narrative Language Analysis System

Purpose Analysis of spontaneous speech samples is important for determining patterns of language production in people with aphasia. To accomplish this, researchers and clinicians can use either hand coding or computer-automated methods. In a comparison of the two methods using the hand-coding NNLA (Northwestern Narrative Language Analysis) and automatic transcript analysis by CLAN (Computerized Language Analysis), Hsu and Thompson (2018) found good agreement for 32 of 51 linguistic variables. The comparison showed little difference between the two methods for coding most general (i.e., utterance length, rate of speech production), lexical, and morphological measures. However, the NNLA system coded grammatical measures (i.e., sentence and verb argument structure) that CLAN did not. Because of the importance of quantifying these aspects of language, the current study sought to implement a new, single, composite CLAN command for the full set of 51 NNLA codes and to evaluate its reliability for coding aphasic language samples. Method Eighteen manually coded NNLA transcripts from eight people with aphasia and 10 controls were converted into CHAT (Codes for the Human Analysis of Talk) files for compatibility with CLAN commands. Rules from the NNLA manual were translated into programmed rules for CLAN computation of lexical, morphological, utterance-level, sentence-level, and verb argument structure measures. Results The new C-NNLA (CLAN command to compute the full set of NNLA measures) program automatically computes 50 of the 51 NNLA measures and generates the results in a summary spreadsheet. The only measure it does not compute is the number of verb particles. Statistical tests revealed no significant difference between C-NNLA results and those generated by manual coding for 44 of the 50 measures. C-NNLA results were not comparable to manual coding for the six verb argument measures. Conclusion Clinicians and researchers can use the automatic C-NNLA to analyze important variables required for quantification of grammatical deficits in aphasia in a way that is fast, replicable, and accessible without extensive linguistic knowledge and training.