WASHINGTON -- When do babies start to understand words as words? A series of eight experiments with infants has provided evidence that even at eight-and-a-half months, they seem sensitive to word boundaries. The experiments are described in the June issue of the Journal of Experimental Psychology: Human Perception and Performance, published by the American Psychological Association (APA).

Psychologists Sven Mattys, Ph.D., and Peter Jusczyk, Ph.D., of the departments of psychology and cognitive science at Johns Hopkins University investigated how babies find the beginnings and ends of words in utterances. Specifically, would infants sometimes incorrectly group the end of one word with the beginning of another word? For example, would babies respond to the sound of "dice" within "cold ice" or "red ice" in the same way they would respond to "dice" in "two dice"?

Mattys and Jusczyk used the widely validated "head turn preference procedure," in which infants sit in a three-sided booth on their caretaker's lap. A green light flashes in front when they look ahead in "rest" position. To start the experiment, a computer-controlled red light flashes on either the left or right side of the booth, drawing the infant's attention. A concealed loudspeaker behind that light plays the experimental word or passage. A hidden observer watches the infant through a peephole, recording for how long the infant listens to the sample (in other words, looks at the red light associated with the loudspeaker).

Mattys and Jusczyk tested two dozen infants in each of their experiments. The youngest infants were about eight-and-a-half months old. First they "familiarized" an infant to certain words by having the infant listen to at least 30 seconds of a female sing-song voice repeating those words over and over as the infant watched the flashing red light. This entered the word in the infant's memory. Second, in the test phase, the researchers played one of three types of recorded passages. "Target-present" passages contained the actual target word. "Misparsed" passages contained the same sound sequences as the targets, but the sounds occurred between two successive words. For example, infants familiarized to the word "dice" might then get a passage of either, "Two dice can be rolled without difficulty" (target present) or "Weird ice no longer surprises anyone" (misparsed). Control group ("target absent") passages included completely unrelated sounds -- to continue the example above, "Crib oats were rather mysterious until recently."

The researchers compared how long the infants paid attention to the different types of samples. "Infants seem to be more interested when they can pick up something they recognize as familiar amidst the new words of the passages," says Mattys. "It's as if you heard your name in a conversation at a table next to yours." The infants showed a listening preference for passages with the target present ("two dice") passages, but not for the misparsed ("weird ice") passages as compared with the target-absent ("crib oats") passages. This result indicates that they were sensitive to word-boundary cues.

In their article, Mattys and Jusczyk also discuss the various speech cues that allow infants to know when words begin and end, such as rhythmic cues (where the accent falls in a word) and allophonic cues (the way that particular sounds are pronounced when they occur in different positions of a word; for example, a "t" at the beginning of a word is pronounced differently from a "t" at the end of a word). They found that English-learning infants were considerably delayed in their ability to segment words that start with vowels instead of consonants, indicating that although word-segmentation capacities start relatively early, the full development of these capacities is a gradual process extending well into the second year. The infants studied failed to segment words starting with vowels until 16 months of age. Fewer spoken words start with vowels, which provide more subtle acoustic cues than the more explosive consonant sounds. Mattys and Jusczyk state that the full development of word-segmentation capacities may start relatively young but they require well into the second year to develop.

Article: "Do Infants Segment Words or Recurring Contiguous Patterns?", Sven L. Mattys and Peter W. Jusczyk, Johns Hopkins University, Baltimore, Md.; Journal of Experimental Psychology: Human Perception and Performance, 2001, Vol. 27, No. 3.

Full text of the article is available from the APA Public Affairs Office and after June 13 at http://www.apa.org/journals/xhp/xhp273644.html

---American Psychological Association

Back to The Science of Mental Health