Non-Local Syntactic Dependencies in Infancy: Wh-Question Representations Develop Between 14- and 18-Months

• In Event: Beyond Categories: Infant Syntax Above the Word Level

Abstract

Infants produce and comprehend wh-questions by 20-months (Stromswold, 1995; Rowland et al., 2003; Seidl et al., 2003; Gagliardi et al., 2016), but it is not known when they represent the structure of these sentences in an adult-like way. Using a listening-preference task (Maye et al. 2002), we find that (1) 18-month-olds represent wh-phrases as non-local arguments in wh-questions, and (2) younger infants do not. These results support the hypothesis that representations of non-local syntactic dependencies in wh-questions develop in infancy (Gagliardi et al., 2016; Perkins & Lidz, 2020).

Infants heard sentences with transitive verbs ("kiss," "hug," "tickle," "bump," "hit," "cover"), paired with an unrelated video of rotating shapes. To facilitate lexical retrieval, we familiarized infants to 72 seconds of these verbs in transitive frames. At test, infants heard 12 trials in two between-subjects conditions (Table 1): either wh-object questions or simple declaratives, in alternating trials with and without local objects after the verb. We recorded looking time towards the screen as a measure of interest in each trial type. If infants represent wh-phrases as non-local arguments, then they should find local objects in wh-questions unacceptable because there are too many arguments. This predicts opposite relative preferences for local objects in wh-questions and declaratives. If infants instead only know that these verbs require objects, then they should be satisfied whenever a local object is present, predicting the same pattern of preference in both conditions.

In four experiments (each n=32), we tested 18-month-olds (18;0-18;31, mean=18;12), 17-month-olds (17;1-17;28, mean=17;14), 15-month-olds (15;0-15;28, mean=15;14), and 14-month-olds (14;0-14;29, mean=14;13). Figure 1 plots infants' listening preferences as differences in total looking time during trials with local objects vs. without. Participants whose preferences were more than 1.5 times the interquartile range above the 75th or below the 25th percentile of the sample were excluded as outliers.

In Study 1, 18-month-olds' preferences were significantly different for declaratives and wh-questions (Welch’s t(27.32)=3.50, p<0.002). 18-month-olds who heard declaratives showed a significant preference for sentences with local objects (t(14)=2.70, p<0.02), but showed the reverse preference for wh-questions (t(15)=-2.23, p<0.04). Thus, 18-month-olds differentiated between declaratives and wh-questions, and preferred grammatical sentences of each type. In Studies 2-4, infants did not show significantly different preferences for declaratives and wh-questions. Across both sentence types, 15-month-olds slightly preferred sentences without local objects (t(29)=-2.28, p<0.03). 14- and 17-month-olds showed no significant preferences. Thus, infants younger than 18-months sometimes responded to locally missing objects, but did not respond differently to declaratives vs. wh-questions.

We find that (1) infants by 18-months represent the non-local dependencies in wh-questions syntactically, and (2) these representations develop before this age. 18-month-olds showed opposite preferences for local objects in declaratives and wh-questions, demonstrating awareness that a wh-phrase acts as a non-local object of the verb. Younger infants did not show this awareness: 15-month-olds listened longer to sentences without local objects, but did so independently of whether the sentence was a declarative or wh-question. These results invite further investigation into infants' syntactic dependency development before 18-months, and the mechanisms that enable this development.

Authors

• Laurel Perkins, University of California - Los Angeles
  Presenting Author

• Jeffrey L Lidz, University of Maryland - College Park
  Non-Presenting Author