2018 Friday Session A 1200

Friday, November 2, 2018 | Session A, East Balcony | 12pm

Crying helps, but being sad doesn’t: Verbs, but not adjectives, constrain referent selection for novel nouns through age three
K. Syrett, A. LaTourrette, B. Ferguson, S. Waxman

INTRODUCTION: Semantic information encoded in lexical representations plays a powerful role in language comprehension by constraining the hypothesis space for the meaning of co-occurring words, introducing selectional restrictions relevant to compositionality, and facilitating referent selection [1]. Fluent speakers and listeners actively recruit such semantic information. Previous work has documented that by two, children demonstrate this ability when identifying a potential referent: they use the co-occurrence of known nouns to identify the semantic category of unknown verbs [2]; use verb meaning to predictively parse sentences [3]; and use known verbs to select the referent for a novel noun [4]-[5]. In a preferential looking study, Ferguson, Graf, & Waxman (2014) demonstrated that infants as young as 19 months use the animacy restrictions of known verbs to select the referent of a novel noun. Given that both verbs and adjectives such as those in (1) have similar surface-level distribution, denote properties, encode similar animacy restrictions, and occur in child-directed and child-produced speech at comparable times, we asked whether adjectives could perform a similar function of narrowing the hypothesis space for novel nouns and referent selection.

(1) The dax is [crying/sad, eating/hungry, sleeping/tired, smiling/friendly]

We show (rather surprisingly) that even at age three, adjectives still lag far behind verbs.

EXPERIMENT: Selected adjectives were identified through an independent  animacy  norming study with adults. We then recruited 45 24-month-olds and 48 36-month-olds. In each of six test trials (see Figure 1), children were shown two novel objects (one animate). The objects then disappeared, and children heard either Informative, animacy-restricted sentences (The dax is sad) or Neutral, unrestricted sentences (The dax is wet) featuring a novel noun. We recorded children’s eye movements when they were shown the objects again and asked to identify the noun’s referent (Where’s the dax?).

In contrast to previous findings with known verbs, children did not look more to the animate referent in the Informative condition than in the Neutral condition, ps>.2 (Figure 2). An analysis of the gaze time-course identified a short window, 0-500ms after word onset, in which 36- month-olds showed a marginal preference for the animate referent, p=.067. Thus, 36-month-olds may be on the cusp of successful inference. However, a follow-up forced-choice pointing task with 36-month olds (in progress) reveals that the time pressure associated with the original methodology is not the challenge: children remain at chance (57% correct).

CONCLUSIONS: Even at three, children do not successfully recruit known adjectives to identify an animate referent for a novel noun, whereas they did so with verbs by 19 months. We entertain two hypotheses about why adjectives do not have a similar function as verbs. First, the present progressive aspectual morphosyntax of verbs may be a call to attention to inspect the unfolding event. Second, the presence of a perceptible attribute for verbs/events may have an advantage over these adjectives, which lack a perceptual cue and may require theory of mind to tap into a mental state. Thus, by allowing multiple linguistic and psychological factors to team up, verbs win out over adjectives.