Search
Browse By Day
Browse By Time
Browse By Panel
Browse By Session Type
Browse By Topic Area
Browse Posters
Search Tips
Register for SRCD23
Personal Schedule
Welcome Letter
Program Guide
Change Preferences / Time Zone
Sign In
To learn a language, learners often must infer underlying structural relations from items that surface in linearly non-adjacent positions (e.g., the cat IS chasING the mouse). Infants showed successful learning of such nonadjacent dependencies (NADs) in artificial languages at 15 months, but not 12 months (Gómez & Maye, 2005). Although Marchetto and Bonatti (2015) reported evidence of NAD learning in 12-month-olds, participants could have relied on cues that were confounded with NADs in that study. Yet 12-month-olds are sensitive to distributional-grammatical patterns in their maternal language, some involving non-adjacent relationships (Geffen & Mintz, 2015; Mintz, 2006). Therefore, further direct investigations into 12-month-old’s ability to learn NADs is needed.
Conducted online using Zoom and Pyhab, this study investigates NAD learning in 12-month-olds. We used nonce words presented at a natural speech rate of 3Hz, which supports NAD learning in adults (Wang et al., 2019). Infants listened to pseudo-random sequences of 9 NAD trigrams presented at a rate of 3 words per second. The 9 NAD trigrams consisted of 3 NAD relations, each with 3 unique interveners (Table 1). A habituation trial started when the infant oriented to the screen and ended when the infant looked away from the video for 2 consecutive seconds, or when the trial reached its maximum length of 30 seconds. Then a new trial started with a different pseudo-random sequence. The habituation phase ended when the average looking time of the last three trials dropped to 50% of the average of the first three. During the speech stream, an animated baby remained on the screen to maintain participants' attention.
Infants then heard 3 grammatical and 3 ungrammatical test trials while viewing an animated baby. Grammatical trigrams combined a familiarized NAD relation (e.g., poog__des in Grammar 1) with one of the interveners from other familiarized NAD relations (poog-gop-des). Ungrammatical trigrams combined first and final words from two distinct NADs (poog__foom) with an intervener that did not occur in a familiarized triplet with either word (poog-gop-foom). Thus, both grammatical and ungrammatical trigrams had zero adjacent transitional probabilities within words and therefore were both novel trigrams. If infants learned the NADs during habituation, we expect them to look longer at the ungrammatical than grammatical trials.
Data collection is ongoing. Due to the small sample size (N=13, 5 females), we ran a Bayesian estimation t-test on the differences between the log-transformed looking times to the ungrammatical and grammatical trials (Kruschke, 2013). We set the priors to be a normal distribution with a mean of 0 and a standard deviation of log(15) to cover the entire range of possible differences (each trial lasted 1-15 seconds). Overall, 12-month-olds looked significantly longer to the ungrammatical trials than the grammatical trials (Figure 1). The model predicts the mean of the credible values for the difference to be 0.176, with a 95% HDI of [0.0194, 0.337] (classic t-test: t(12) = 2.60, p = .012). This suggests that 12-month-old infants can learn NADs from artificial speech streams, and natural speech rates may facilitate the learning process.