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

Abstract

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.

Children learn ergative case marking in Hindi using statistical preemption and clause-level semantics (intentionality): evidence from acceptability judgment and elicited production studies with children and adults

“…Semantic accounts hold that learners are guided in part by meaning. For example, the reason we can say Somebody boiled the water but not Somebody danced the boy is that “boiling” but not “dancing” is an activity that an external causer can more-or-less force another entity to undergo ( Ambridge et al , 2020 ; Ambridge et al , 2022 ; Pinker, 1989 ; Shibatani & Pardeshi, 2002 ).…”

“…Indeed, a version of preemption (or learning/reasoning more generally) that did not take into account the “base rate” in this way would be very difficult to support on logical grounds; like arguing that Drug A (5 deaths amongst 50 patients) is safer than Drug B (10 deaths amongst 1,000 patients). For this reason, we follow previous studies that have used the chi-square/Fisher’s exact method for calculating the measure of preemption/competition (e.g., Ambridge et al ., 2018 ; Ambridge et al ., 2020 ; Ambridge et al ., 2022 ; Ambridge et al , 2023 ; Liu & Ambridge, 2021 ; Stefanowitsch, 2008 ; Tatsumi et al ., 2018a ; Tatsumi et al ., 2018b ).…”

“…Unsurprisingly, given its broad linguistic and non-linguistic implications, the question of how to acquire exception-filled generalizations has inspired a great deal of theoretical debate and empirical research. However, with just a handful of exceptions (e.g., Ambridge et al ., 2020 ; Ambridge et al ., 2022 ) research on child language has mainly been restricted to English (e.g., Alishahi & Stevenson, 2008 ; Ambridge et al ., 2008 ; Ambridge et al ., 2009 ; Ambridge, 2013 ; Ambridge & Ambridge, 2020 ; Ambridge & Blything, 2016 ; Ambridge & Brandt, 2013 ; Ambridge et al ., 2011 ; Ambridge et al ., 2012a ; Ambridge et al ., 2012b ; Ambridge et al ., 2013 ; Ambridge et al ., 2014 ; Ambridge et al ., 2015 ; Ambridge et al ., 2018 ; Barak et al ., 2016 ; Bidgood et al ., 2014 ; Blything et al ., 2014 ; Boyd & Goldberg, 2011 ; Brooks & Zizak, 2002 ; Brooks et al ., 1999 ; Goldberg, 2011 ; Gropen et al ., 1991 ; Harmon & Kapatsinski, 2017 ; Hsu & Chater, 2010 ; Irani, 2009 ; Li & MacWhinney, 1996 ; Perek & Goldberg, 2017 ; Perfors et al ., 2010 ; Robenalt & Goldberg, 2015 ; Robenalt & Goldberg, 2016 ; Stefanowitsch, 2008 ; Theakston, 2004 ; Twomey et al ., 2014 ; Twomey et al ., 2016 ; Wonnacott et al ., 2008 ), including two book-length treatments ( Goldberg, 2019 ; Pinker, 1989 ).…”

“…The second possibility – again supported by previous studies of English and a handful of other languages (e.g., Ambridge et al ., 2020 ; Ambridge et al ., 2022 ; Gropen et al ., 1989 ; Gropen et al ., 1991 ) – is that children might be guided by semantics. For example, at least part of the reason that English speakers prefer Somebody made the boy dance over * Somebody danced the boy is that “dancing” is not really an action that is amenable to outside causation: someone can ask a boy to dance, but not force him, he retains some agency.…”

“…For example, at least part of the reason that English speakers prefer Somebody made the boy dance over * Somebody danced the boy is that “dancing” is not really an action that is amenable to outside causation: someone can ask a boy to dance, but not force him, he retains some agency. In contrast, English speakers accept both Somebody made the water boil and Somebody boiled the water because this type of direct, outside causation is possible ( Ambridge et al ., 2020 ; Ambridge et al ., 2022 ; Pinker, 1989 ; Shibatani & Pardeshi, 2002 ). In the domain of Hindi ergative ne marking, effects of semantics might be observed at either or both of two levels: the verb (lexical) and the clause (clausal).…”