Periodic Reporting for period 1 - SASC (Synergies in the Acquisition of Syntactic Categories)
Reporting period: 2018-09-03 to 2020-09-02
Context of the project:
My current work focuses on the synergies between lexical and syntactic acquisition. Several studies have shown that children can learn the meaning of new words (word semantics) by relying on the other words in the sentence (the syntactic context), a mechanism called syntactic bootstrapping (e.g. Gleitman, 1990). For instance, when hearing “It’s a pratch!” 18-month-olds can infer that pratch refers to an object, while upon hearing “It’s pratching!,” they can infer that it refers to an action. The question of how infants learn which syntactic contexts correspond to which semantic features is still unresolved (e.g. verb = action). Surely, infants’ ability to track statistical regularities must play some role. After all, syntactic categories are essentially distributional categories that are formed based on the occurrence of a set of words in the same context. Previous attempts to model distributional learning of categories have proven this to be feasible (e.g. Mintz, 2003). Preverbal infants possess another powerful tool for the initiation of syntactic categories: they have under their belt a handful of known words. Early understanding of the meaning of frequent concrete words starts around 6 months of age (e.g. “mommy”, “hand”), while the formation of semantic classes is thought to happen before the age of one (e.g. “ball” as object vs “eat” as action). Based on the semantic seed hypothesis (Christophe, Dautriche, Carvalho, & Brusini, 2016), infants could exploit this handful of known words as a seed for creating syntactic categories: they would track the syntactic contexts in which known words occur (e.g. objects such as ‘ball’ and ‘car’ appear after “This is a”), then infer that a novel word appearing in the same context is likely to share some semantic properties with the seed (e.g. “This is a” + “dax” -> dax = object).
The SASC project investigates the impact of a small developing lexicon on syntactic acquisition, as well as the strengths and the limits of the learning mechanism behind it. At this moment, there is a gap in the experimental literature, where theoretical frameworks based on computational models need to be tested in order to integrate new insights and to gain a complete picture of the mechanism that enables the acquisition of word meaning and the creation of syntactic categories. These insights will not only be used to interpret experimental data and to make new predictions for future studies, but also to understand morphosyntactic patterns in world’s languages. Inspection of cross-linguistic universals reveals that across the world’s languages, certain conceptual categories are often marked in morphosyntax (eg. object/action, animate/inanimate), while others are not. Hence, the SASC project also aims at giving us insight on the cause of the cross-linguistic regularity, which might be that since morphosyntax helps infants to learn word meanings, only those morphosyntactic regularities that mark conceptual distinctions that are noticeable by young infants would be selected for human languages. This project is building a bridge between researchers studying language acquisition, conceptual development, and language evolution.
Overall objectives:
Line 1 explores infants’ early sensitivity to the syntactic contexts of common words (nouns, verbs). Line 2 tests young children’s ability to use known and newly learned function words (i.e. pronouns, articles) to interpret the meaning of co-occurring novel words as either objects (nouns) or actions (verbs). Line 3 investigates young children’s ability to predict word semantics from new function words. This is done by introducing a new distinction that is found in natural languages, but absent from their own language (i.e. syntactic cues for animacy). Future ongoing work aims at testing the limits of this syntax-based inference to see whether languages’ morphosyntax is biased by «core cognition » distinctions.
My current work focuses on the synergies between lexical and syntactic acquisition. Several studies have shown that children can learn the meaning of new words (word semantics) by relying on the other words in the sentence (the syntactic context), a mechanism called syntactic bootstrapping (e.g. Gleitman, 1990). For instance, when hearing “It’s a pratch!” 18-month-olds can infer that pratch refers to an object, while upon hearing “It’s pratching!,” they can infer that it refers to an action. The question of how infants learn which syntactic contexts correspond to which semantic features is still unresolved (e.g. verb = action). Surely, infants’ ability to track statistical regularities must play some role. After all, syntactic categories are essentially distributional categories that are formed based on the occurrence of a set of words in the same context. Previous attempts to model distributional learning of categories have proven this to be feasible (e.g. Mintz, 2003). Preverbal infants possess another powerful tool for the initiation of syntactic categories: they have under their belt a handful of known words. Early understanding of the meaning of frequent concrete words starts around 6 months of age (e.g. “mommy”, “hand”), while the formation of semantic classes is thought to happen before the age of one (e.g. “ball” as object vs “eat” as action). Based on the semantic seed hypothesis (Christophe, Dautriche, Carvalho, & Brusini, 2016), infants could exploit this handful of known words as a seed for creating syntactic categories: they would track the syntactic contexts in which known words occur (e.g. objects such as ‘ball’ and ‘car’ appear after “This is a”), then infer that a novel word appearing in the same context is likely to share some semantic properties with the seed (e.g. “This is a” + “dax” -> dax = object).
The SASC project investigates the impact of a small developing lexicon on syntactic acquisition, as well as the strengths and the limits of the learning mechanism behind it. At this moment, there is a gap in the experimental literature, where theoretical frameworks based on computational models need to be tested in order to integrate new insights and to gain a complete picture of the mechanism that enables the acquisition of word meaning and the creation of syntactic categories. These insights will not only be used to interpret experimental data and to make new predictions for future studies, but also to understand morphosyntactic patterns in world’s languages. Inspection of cross-linguistic universals reveals that across the world’s languages, certain conceptual categories are often marked in morphosyntax (eg. object/action, animate/inanimate), while others are not. Hence, the SASC project also aims at giving us insight on the cause of the cross-linguistic regularity, which might be that since morphosyntax helps infants to learn word meanings, only those morphosyntactic regularities that mark conceptual distinctions that are noticeable by young infants would be selected for human languages. This project is building a bridge between researchers studying language acquisition, conceptual development, and language evolution.
Overall objectives:
Line 1 explores infants’ early sensitivity to the syntactic contexts of common words (nouns, verbs). Line 2 tests young children’s ability to use known and newly learned function words (i.e. pronouns, articles) to interpret the meaning of co-occurring novel words as either objects (nouns) or actions (verbs). Line 3 investigates young children’s ability to predict word semantics from new function words. This is done by introducing a new distinction that is found in natural languages, but absent from their own language (i.e. syntactic cues for animacy). Future ongoing work aims at testing the limits of this syntax-based inference to see whether languages’ morphosyntax is biased by «core cognition » distinctions.
#1 In Babineau & Christophe (in prep), we found that 11-month-olds are already sensitive to the functional context surrounding verbs and nouns (for example, verbs such as mange ‘eat’ can occur after the pronoun tu ‘you’, but not after the determiner des ‘some’). Moreover, French-learning 14-month-olds can assign novel words to syntactic categories based on co-occurring function words: a novel word that co-occurs with determiners is expected to be used with other determiners, not pronouns; Babineau, Shi & Christophe, 2020).
#2 In Babineau, de Carvalho, Trueswell & Christophe (2021), after watching a 5-minute training video in which a novel function word “ko” either replaced pronouns (verb condition) or articles (noun condition), 3-to-4-year-olds (and adults) were able to infer the meaning of novel content words that co-occurred with the newly-learned function word “ko” (Oh look, ko dase!), interpreting them as either referring to novel actions or novel objects.
#3 In Barbir, Babineau, Fiévet, & Christophe, in prep), we tested French-learning 20-month-old infants’ ability to learn that a novel determiner “ko” was restricted to occurring before animates, while “ka” appeared only before inanimates, after encountering the two forms alongside words they already know (ko chien, ka poussette, “ko dog”, “ka stroller”). At test, they looked more towards a novel animate when hearing “ko,” and towards a novel inanimate toy when hearing “ka” (“Oh look! Ko bamoule”; where bamoule is a novel noun). In a follow-up, we will compare toddlers’ ability to learn a classification that is not found in natural languages. Due to covid-19, this last study has been put on hold for the moment.
#2 In Babineau, de Carvalho, Trueswell & Christophe (2021), after watching a 5-minute training video in which a novel function word “ko” either replaced pronouns (verb condition) or articles (noun condition), 3-to-4-year-olds (and adults) were able to infer the meaning of novel content words that co-occurred with the newly-learned function word “ko” (Oh look, ko dase!), interpreting them as either referring to novel actions or novel objects.
#3 In Barbir, Babineau, Fiévet, & Christophe, in prep), we tested French-learning 20-month-old infants’ ability to learn that a novel determiner “ko” was restricted to occurring before animates, while “ka” appeared only before inanimates, after encountering the two forms alongside words they already know (ko chien, ka poussette, “ko dog”, “ka stroller”). At test, they looked more towards a novel animate when hearing “ko,” and towards a novel inanimate toy when hearing “ka” (“Oh look! Ko bamoule”; where bamoule is a novel noun). In a follow-up, we will compare toddlers’ ability to learn a classification that is not found in natural languages. Due to covid-19, this last study has been put on hold for the moment.
The main objective of the SASC project is to shed light on the learning mechanism used during language acquisition. More precisely, it tests some of the field proposals on the creation of syntactic categories, as well as the strengths and the limits of the learning mechanism involved. To do so, I use cutting-edge experimental techniques, such as eye-tracking, animation of talking animals, and innovative designs involving naturalistic training videos. With the use of an eye-tracker to measure infants’ gaze trajectory during the presentation of audio/visual stimuli, I can measure their online inferences about the meaning of a new word (i.e. choice of one of the two videos). These inferences are linked to their ability to abstract patterns, track certain conceptual distinctions, and to generalize a newly acquired rule. In the training videos used for research line 2 and 3, a woman is looking directly to the camera, producing simple sentences with a child-directed speech, and playing with a few toys at a time. Interestingly, children who participated to the studies found those naturalistic videos very engaging, which enabled them to stay attentive for an extensive period of time. Following these training videos, we showed that infants successfully associated a new syntactic context to a semantic category. Hence, there is an immediate potential for concrete applications. That is, on-screen education has potential societal relevance as a complement to live education promoting language development.