INFANT LEARNING LAB

Where are the boundaries between words?

One line of research pursued in the Infant Learning Lab concerns the problem of word segmentation. As adult listeners, we perceive word boundaries when listening to a familiar language. However, these boundaries disappear when we hear a foreign language. This phenomenon occurs because speakers do not consistently pause between words; in fact, silent points in phrases often occur in the middle of a word. In many cases, acoustics are poor cues to word boundaries.

We have proposed that learners, including infants, may detect word boundaries in part by tracking the statistical properties of the sound combinations that they hear. In these experiments, we expose infant learners to sound sequences generated from made-up languages.

Using the Head Turn method of testing infants, we can then see whether infants can detect and use the probabilities with which sounds co-occur to detect word boundaries. Previous research (Saffran, Aslin, and Newport, 1996) has shown that after listening to two minutes of four nonsense words babies prefer to listen to an unfamiliar word more than a familiar word. The idea is that infants become bored of the words they have been listening to for a while and are not interested in listening to them when they hear them again in the test trials, whereas the unfamiliar words grab the infant's attention, prompting them to listen longer. These findings suggest that infants can track the statistics of sound sequences to find word boundaries. Ongoing research explores the kinds of cues that infants use, in combination with sequential statistics, to locate the boundaries between words, including the ways in which syllables are stressed and the pitch patterns characteristic of infant-directed speech.

Can infants learn grammatical patterns?

We are currently exploring what kinds of grammatical regularities infants are capable of learning, and whether they learn syntax in the same way as adults do. All languages are governed by grammatical regularities called “predictive regularities.” That is, elements in the grammar predict other elements (for example, in English, “the” predicts the presence of a noun, such as “dog”). Adults can learn grammars with predictive regularities, but have difficulty with grammars that do not contain such regularities. In studies conducted in this lab, we are exposing infants to grammars in order to explore the limits of infant grammatical learning. We can then compare these results to how non-human animals are able to learn the same kinds of patterns, to determine what makes human infants “special” with respect to language learning.

Is it a word?

In other research, we are trying to see if babies understand at some level that new patterns of sounds can be a word (Saffran, 2001). One way we do this is in our experiments with toddlers. In these experiments, toddlers hear fluent speech like that described above, which contains a series of nonsense words. We then try to teach toddlers the names of new objects. Some of these objects have names that were words in the language they heard, while other objects do not. If children think that the sounds they are hearing are potential words, they should have more success learning the names of objects that were words in the language.

How much do infants know about the sound of words?

In languages, small differences in sound have big differences in meaning. For example, in English, “dock” and “talk” mean something different, even though the only difference between them is /d/ vs. /t/. Infants must learn which differences are important, and what they mean. To understand this process, we have recently begun trying to teach infants the names of new objects and to see how much attention they pay to the sounds of those names.

In these experiments, infants see an object on a monitor and hear the name of the object repeated for as long as they are willing to look at it. At first, this is quite interesting, but after a while infants get bored with experiencing the same word/object pair repeatedly. This is their way of telling us that they have learned the pairing. When infants have learned the pairing, the test trials begin. In the test trials, half of the time the object is paired with the same word it was previously (for example, “bih”), and half of the time it is paired with a different word (such as “dih”). We want to know if infants pay attention to this difference in the sound of the word, and whether they think both of those words can refer to the object.

Can infants learn patterns?

Another line of our research looks at whether infants can learn patterns. For example, in one series of studies we asked whether listening to a list of nonsense words that follow a pattern like consonant-vowel-consonant-vowel (e.g., gola) will help babies pick out new words that follow the same pattern. After only 1 minute of exposure, infants were able to pick up such a pattern and use it to help them find words in fluent speech (Saffran & Thiessen, 2003). We also study how infants learn musical patterns, as well as how they begin to combine speech sounds and melodies in learning sung sequences.

LEARNING LANGUAGES ADULT LAB

In the Learning Languages Adult Lab, we investigate the strategies that adults and older children use to learn languages. We are interested in the extent to which language-learning tools may be useful in learning the regularities of other language-like systems, including visual patterns and musical sequences. In our most recent work, we have examined the ability of adults to learn the regularities in a complex musical system through a brief exposure period.