In this Work Package we developed a new cross-linguistic model of the pattern of verb-marking error in typically developing children and children with Developmental Language Disorder (DLD). Children with DLD show different patterns of error depending on the language they are learning. Developing a model of DLD that works across languages is therefore a good way to identify the underlying problem that results in these errors, and to work out why it shows up in different ways in different languages. It is also a good way to develop evidence-based interventions that can support recovery from error.

We developed the model by taking a computer model from our previous research (MOSAIC) and adding a new mechanism that substituted the most frequent form of the verb for forms of the verb that were particularly rare in child-directed speech. Our simulations showed that the new model (MOSAIC+) was able to capture the higher rate of bare stem errors (e.g. ‘That go there’) in English-speaking children with DLD and the higher rate of agreement errors (equivalent to ‘I goes home’) in Dutch-, German- and Spanish-speaking children. It was also able to capture the higher rate of word-order errors in Dutch and German, where some verb forms need to go before their complements and others need to go after their complements. These results suggest that children with DLD have a particular problem acquiring low frequency forms and are particularly prone to replacing them with forms that are more strongly represented in their systems. A paper describing this work was published in Language Learning (Freudenthal, Gobet & Pine, 2024).

In addition to developing MOSAIC+, we also developed a further model to look more closely at the difference between English and Spanish. This model used a modified version of the classic Rescorla-Wagner model to simulate differences in the ease with which both TD children and children with DLD master English and Spanish verb inflection. The simulations show that these differences can be understood in terms of the interaction between a deficit in sequential learning ability and the structure and statistics of English – in which whether the verb is inflected or not depends on what happens earlier in the sentence – and Spanish – in which it does not. In short, the English version of the model tended to master verb inflection more slowly than the Spanish version because, in addition to learning the relevant verb endings, it also needed to learn when to apply them (i.e., in statements such as ‘That goes there’, but not in questions such as ‘Does that go there?’). A paper describing this work was published in Cognitive Science (Freudenthal, Ramscar, Leonard, & Pine, 2021) and Daniel Freudenthal wrote a blog describing these findings for the LuCiD website: https://www.lucid.ac.uk/2022/03/16/why-verb-endings-are-harder-to-learn-in-english-than-in-spanish/

Overall, these findings suggest two specific weaknesses in the language learning abilities of children with DLD. The first is a particular difficulty in learning low-frequency forms like English third person singular present tense -s forms, which results in errors in which these forms are replaced by more frequent and hence more strongly represented bare stem forms. The second is a particular difficulty in learning non-adjacent dependencies (e.g., does that go versus that goes) that makes it difficult to distinguish the contexts in which inflected and non-inflected forms of the verb are required. Both weaknesses have potential implications for therapy. The first implies the need to support the learning of low frequency inflected forms by artificially increasing their frequency in the input. The second implies the need to support the learning of non-adjacent dependencies either by drawing the child’s attention to the different way in which inflection is marked in statements and questions in English, or by exposing the child to a wider range of instances of the relevant dependency to promote learning.

In our current work we are building on these insights by developing a Deep Learning Long Short Term Memory model of the development of verb inflection in English. This model works in a similar way to the model described in our 2021 paper, but learns to perform a more challenging task (predicting the next word or inflection in the utterance) on the basis of more realistic input (a large corpus of child-directed speech). A paper describing a preliminary version of the model was published in the Proceedings of the 2023 Annual Meeting of the Cognitive Science Society (Freudenthal, Pine & Bannard, 2023) and we are currently in the process of developing the model further with a view to publishing a longer journal article and applying for funding to use the model both to simulate data from children with DLD in more detail and to explore the potential efficacy of different input-based interventions in helping English-speaking children with DLD overcome their difficulties with verb marking.

Project Team 

Daniel Freudenthal, Fernand Gobet, Larry Leonard and Julian Pine (Lead) 

Duration 

4 years 

Project Number 

1.3 

Key Outputs 

Freudenthal, D., Gobet, F. & Pine, J. M. (2024). MOSAIC+: a cross-linguistic model of verb-marking error in typically developing children and children with Developmental Language Disorder. Language Learning, 74(1), 111-145. 

Freudenthal, D., Ramscar, M., Leonard, L. B. & Pine, J. M. (2021). Simulating the acquisition of verb inflection in typically developing children and children with Developmental Language Disorder in English and Spanish. Cognitive Science, 45(3), e12945. 

Additional Outputs 

Freudenthal, D., Pine, J. M. & Bannard, C. (2022). Modelling Children’s Sentence Recall using an Encoder-Decoder Network. J. Culbertson, A. Perfors, H. Rabagliati & V. Ramenzoni (Eds.) Proceedings of the 44th Annual Conference of the Cognitive Science Society (pp. 2576-2582). Toronto, ON: Cognitive Science Society. 

Freudenthal, D., Pine, J. & Bannard, C., (2023). Simulating children’s verb inflection errors in English using an LSTM language model. Proceedings of the 45th Annual Conference of the Cognitive Science Society (pp. 2602-2608). 

Pine, J. M., Freudenthal, D. & Gobet, F. (2023). Building a unified model of the Optional Infinitive Stage: Simulating the cross-linguistic pattern of verb-marking error in typically developing children and children with Developmental Language Disorder. Journal of Child Language.1-17. 

References 

Bedore, L. M., & Leonard, L. B. (2005). Verb inflections and Noun Phrase morphology in the spontaneous speech of Spanish-speaking children with Specific Language Impairment. Applied Psycholinguistics, 26(2),195-225. 

Conti‐Ramsden, G., Botting, N., & Faragher, B. (2001). Psycholinguistic Markers for Specific Language Impairment (SLI). Journal of Child Psychology and Psychiatry, 42(6), 741-748. 

Freudenthal, D., Pine, J. M., Aguado‐Orea, J., & Gobet, F. (2007). Modeling the developmental patterning of finiteness marking in English, Dutch, German, and Spanish using MOSAIC. Cognitive Science, 31(2), 311-341. 

Freudenthal, D., Pine, J. M., Jones, G. & Gobet, F. (2015). Simulating the cross-linguistic pattern of Optional Infinitive errors in children’s declaratives and Wh- questions. Cognition, 143(1), 61-76. 

Hsu, H. J., Tomblin, J. B., & Christiansen, M. H. (2014). Impaired statistical learning of non-adjacent dependencies in adolescents with Specific Language Impairment. Frontiers in Psychology, 5, 175. 

Kueser, J. B., Leonard L. B., & Deevy P. (2018). Third person singular -s in typical development and Specific Language Impairment: Input and neighbourhood density. Clinical Linguistics & Phonetics. 32(3), 232-248. 

Leonard, L. B., Fey, M., Deevy, P., & Bredin-Oja, S. (2015). Input sources of third person singular –s inconsistency in children with and without Specific Language Impairment. Journal of Child Language, 42(4), 786-820. 

Leonard, L. B. (2014). Children with Specific Language Impairment (Second Edition). Cambridge, Massachusetts, London: The MIT Press. 

Räsänen, S. H. M., Ambridge, B., & Pine, J. M. (2014). Infinitives or bare stems? Are English-speaking children defaulting to the highest-frequency form? Journal of Child Language, 41(4), 756-779. 

Rice, M. L., Noll, K. R., & Grimm, H. (1997). An Extended Optional Infinitive Stage in German-speaking children with Specific Language Impairment. Language Acquisition, 6(4), 255-295. 

Rice, M., Wexler, K., & Hershberger, S. (1998). Tense over time: the longitudinal course of tense acquisition in children with Specific Language Impairment. Journal of Speech, Language, and Hearing Research, 41(6), 1412-1431. 

Theakston, A. L., Lieven, E. V. M., & Tomasello, M. (2003). Journal of Speech, Language, and Hearing Research, 46(4), 863-877. 

Tomblin, J. B., Records, N. L., Buckwalter, P., Zhang, X., Smith, E., & O’Brien, M. (1997). Prevalence of Specific Language Impairment in kindergarten children. Journal of Speech, Language, and Hearing Research, 40(6), 1245-1260. 

Related