Lewkowicz, David J.

Lewkowicz & Hansen-Tift found that when 4-month-old infants see and hear a person
talking, they look more at her eyes but that 8- and 10-mo infants look more at her mouth. The
developmental attentional shift to the mouth reflects infants’ growing interest in speech.
Attention to the mouth enables infants to gain access to redundant and maximally salient
audiovisual cues which then facilitate speech and language acquisition.
We investigated the separate role of mouth movement and vocalization cues in the attentional
shift from a talker’s eyes to the talker’s mouth. In 3 experiments, we used an eye-tracker to
measure the proportion of attention infants, 4-, 8-, and 10-mo, allocate to the eyes and mouth of a
static/silent face, a static/talking face, and a silently talking face. We found that when infants see
a static person, they attend to the eyes. Lewkowicz & Hansen-Tift found that when infants see
and hear a person talking, 4-mos look at the eyes whereas 8- and 10-mos look at the mouth.
When infants see a silently talking person, only 10-mos look at the mouth. These findings
demonstrate that the shift from the eyes to the mouth is mediated by three factors: dynamic
visual speech cues, an emerging interest in speech, and the redundancy of audiovisual speech.
Thus, younger infants are not interested in speech so they focus on the eyes, whereas older
infants become interested in speech, shifting their focus to the mouth, but initially at 8 m, this
shift requires that speech be multisensory.

Previous research has demonstrated that sensitivity to unimodal nonnative speech
contrasts generally narrows during the first year. Although other work has demonstrated
a processing advantage for multimodal stimuli, research on infants' responsiveness to
nonnative contrasts so far has not examined whether concurrent auditory and visual
speech information can modulate perceptual narrowing. Thus, the current study
investigated the influence ofbimodally specified speech sounds on infants' sensitivity to
a nonnative phonemic contrast. Six-month-old and 10- to 12-month-old infants were
tested in a habituation/test procedure for discrimination of an audiovisual nonnative
speech contrast (Hindi /tal dental vs. /Tal retroflex stop). Findings showed that infants at
both ages exhibited evidence of discrimination following habituation to one of the speech
sounds. These findings suggest that the usually observed decline in responsiveness to
nonnative speech contrasts is limited to audibly specified contrasts and that concurrent
visual speech information enhances the discriminability of such contrasts.

Non-adjacent statistical relations are an important class of sequential structure because they aid in the acquisition of syntax and, thus, language. Previous work has demonstrated that 15-month-old infants are sensitive to distant sequential relations but that these types of relations are difficult to learn. Importantly, it is not known whether the ability to learn non-adjacent statistical relations is based on a domain-specific or domain-general pattern-learning mechanism. We examined the domain-generality of this ability in separate groups of 10- and 12-month-old infants in two experiments utilizing the habituation/test procedure.
<br>
<br>Experiment 1 habituated infants to sequences of five moving/sounding arbitrary shapes and sounds. The sequences contained two target elements that were always separated by a non-target element. Results indicated that neither age group displayed response recovery when the target elements were switched. Experiment 2 simplified the task by using sequences that were three elements in length e.g., ABC and DBE. During the test trials, the last element from the two unique pairings was again switched e.g., ABE and DBC. Results indicated that only the 12-month-olds detected a change in the sequence [t 48 1.76, p 0.05].
<br>
<br>These results indicate that infants’ sensitivity to multisensory non-adjacent statistical dependencies is limited to simple 3 element sequences rather than complex 5 element sequences. Our findings also indicate that infants as young as 12 months of age can learn non-adjacent sequential relations embedded within arbitrary audiovisual sequences, suggesting that this critical ability is domain-general in nature.
<br>