Individual Submission Summary
Share...

Direct link:

When Cues Collide: Parents’ Speech Disambiguates between Conflicting Socio-Perceptual Cues to Reference

Fri, April 9, 3:15 to 4:15pm EDT (3:15 to 4:15pm EDT), Virtual

Abstract

How infants and toddlers learn link words to their referents has long been of interest to developmental scientists (Bloom, 2000). Experimental data suggests that one route through which infants identify the referents of speech is by utilizing a host of perceptual and social cues to reference (e.g., Yurovsky & Frank, 2017). Observational studies of parent-infant interactions suggest however that many of these cues are highly probabilistic, available only some of the time that parents name objects (Frank et al., 2013). In an analysis of a video corpus of parent-infant interactions, we further demonstrate that different cues often conflict, thus leaving word reference unresolved. Our analyses further reveals however that parents’ speech contains certain markers that could help infants disambiguate between conflicting cues to reference.
We analyzed recordings of 27 parent-infant interactions (the 12-month-old Rollins Corpus; Rollins, 2003). We transcribed parents’ speech and identified utterances where an object was named. For each of 1,442 naming events, we coded the status of two commonly-studied cues to reference: the object in the infant’s hands and the object in the parent’s hands (Fig 1; see Frank et al., 2013). Consistent with prior research, both infant and parent were more likely to hold the target object (i.e., named object) than a distractor object (pinfant = 0.04; pparent < 0.001; Fig. 2A). However, on a sizeable portion of naming events, infant and parent holding conflict (M = 0.24, SD = 0.20) where the infant and parent each held a different object (Fig 2B).
How might infants resolve such instances of conflicting referential cues? We examined the availability of lexical cues in parents’ speech that could disambiguate between which cue (i.e., infant vs. parent holding) marks the referent in that speech. Specifically, we examined the presence of Distal (“that”/“those”) and Proximal (“this”/“these”) demonstratives in parents’ speech.
Results revealed that the relationship between the two demonstratives and the two holding cues were systematic and indicative of the referent in parents’ speech. Distal demonstratives more likely occurred in parents’ speech when infants held the named object (M = 0.16, SD = 0.14) than when parents held the named object (M = 0.05, SD = 0,07), t(41) = 3.67, p < 0.001 (see Fig 2B). Proximal demonstratives on the other hand more likely occurred when parents held the named object (M = 0.05, SD = 0.07) than when infants held the named object (M = 0.02, SD = 0.05), t(48).= 2.12, p = 0.04. This interrelation between linguistic markers and socio-perceptual cues did not differ as a function of whether holding patterns conflicted or not, p = .92 (see Fig. 2C).
In sum, although the word learning input contains many perceptual and social cues that may help infants learn the referents of words, these cues are probabilistic and often conflict. We illustrate that parents’ speech contains useful markers that could help infants disambiguate between conflicting cues. Ongoing work in our lab is testing experimentally learners’ sensitivity to these markers and whether they are used in the service of learning.

Authors