Presented at the biennial meeting of the Society for Research in Child Development (SRCD), Tampa, FL, April 2003

Alycia Cummings* and Anne Fernald
[*Honors student at Stanford University; now a graduate student at U. C. San Diego]

When addressing a young child, adults in diverse languages frequently use utterances consisting of single words. There is debate whether this characteristic of ID-speech could be useful to infants learning new words. Some argue it is easier to segment and identify new words presented in isolation than in continuous speech because the boundaries are clearly demarcated (e.g., Peters, Brent & Siskind). Others claim that isolated words do not occur consistently enough to play an important role in segmentation (e.g., Woodward & Aslin). Here we present the first experimental evidence testing this hypothesis, investigating whether 18-month-olds are more successful in learning a new word when it is first presented in isolation than when presented as either a medial or final word within a multi-word utterance.

Subjects were 18-month-old infants (n=80), tested in a looking-while-listening procedure in which infants looked at pictures while listening to recorded speech naming one of the pictures. By monitoring infants' eye movements, we obtained a precise record of the time course of word recognition. Infants were randomly assigned to one of four word-teaching conditions:

Isolated - Repeated
Look at that. Toma. Toma. Do you see it?
Isolated - Not Repeated
Look at that. Toma. Do you see it? Toma.
Multiword - Final
Look at the toma. See the toma?
Multiword - Medial
Look at the toma over there. That's a toma there.

On teaching trials subjects were shown one of two novel objects; one was randomly assigned as the toma, or target, the other as the distracter. Infants in the Isolated conditions heard the new word presented in a single-word utterance while infants in the Multiword conditions the novel word was presented in medial or final position in continuous speech. When the distracter object was shown, infants heard attentional utterances with no object label (e.g., Wow. See that? Do you like it?), comparable in intonation to the speech on toma trials. Target and distracter objects were presented three times each during the teaching phase. On six test trials, both novel objects were presented together as the child heard Where's the toma? Control trials containing familiar words were interspersed among test trials. Dependent measures were the speed and accuracy of infants' responses to the named object.

Contrary to predictions that a novel word introduced in a single-word utterance would be learned faster and more reliably than in a multi-word utterance, we found no advantage for words first encountered in isolation. Only infants in the Final condition showed evidence of word learning. While words spoken in isolation do have the benefit of having clearly marked word boundaries, hearing a word presented by itself may have disadvantages for the inexperienced listener, who is unable to anticipate the new information in the speech stream. These results are consistent with other findings from our lab demonstrating that by 18 months infants are attending to the entire carrier phrase, "listening ahead" to predict the upcoming object label.