A universal cue for grammatical categories in the input to children: Frequent frames

Abstract

HighlightsData from typologically diverse languages shows common distributional patterns.Discontinuous repetitive patterns in the input provide cues for category assignment.Morphological frames accurately predict nouns and verbs in the input to children.

Child-directed speech is optimized for syntax-free semantic inference

“…Other causative meanings occur in very similar frames (e.g., “break” as in “mummy would break the hat” and “don’t break the doll”, or “turn” as in “turn the knob” and “shall I turn the page?”). The repetitive nature and frequency of such frames are known to critically help learn categories 31 , 78 . Here, the causative category can be learned from a frequent frame that describes an agent acting on an patient undergoing change.…”

Nouns slow down speech across structurally and culturally diverse languages

“…Second, while speech rate in corpora is mostly studied in terms of the articulation of a word, speech rate variation before words of different types is a measure with great potential to gain insights into the mechanisms of language production. Third, naturalistic corpus studies on widely diverse languages allow detection of signals that do not suffer from the sampling bias in much of current theorizing about language and speech ( 33 , 44 ). Most such work is still largely based on educated speakers of a small number of mostly Western European languages, and it remains unclear whether findings generalize beyond this ( 40 , 45 , 46 ).…”

Lexical category acquisition is facilitated by uncertainty in distributional co-occurrences

“…First, evidence from different languages is required to test the generality of the reported effects beyond English, which was chosen as a target language due to the large availability of corpora and related analyses. Languages with richer morphology can provide different cues to categorization [39], which may then be influenced in other ways or by other factors. Second, the current analysis based on a count-based model that instantiates traditional Hebbian learning with no error feedback [99] should be expanded by also considering other classes of learning mechanisms, such as error-driven learning [100, 101], where the model does not simply store co-occurrences but actively tries to predict what will happen and adjust its beliefs accordingly [102–104].…”

“…Many computational simulations and corpus studies [8–23] as well as behavioral studies [24–32] have corroborated the distributional bootstrapping hypothesis. Moreover, this learning strategy has proven effective for several languages other than English, including French [33], Dutch [34], Spanish [35], German [36, 37], Turkish [37], Chinese [38], and several other typologically diverse languages [39].…”