Rapid Statistical Learning Supporting Word Extraction From Continuous Speech

Abstract

The identification of words in continuous speech, known as speech segmentation, is a critical early step in language acquisition. This process is partially supported by statistical learning, the ability to extract patterns from the environment. Given that speech segmentation represents a potential bottleneck for language acquisition, patterns in speech may be extracted very rapidly, without extensive exposure. This hypothesis was examined by exposing participants to continuous speech streams composed of novel repeating nonsense words. Learning was measured on-line using a reaction time task. After merely one exposure to an embedded novel word, learners demonstrated significant learning effects, as revealed by faster responses to predictable than to unpredictable syllables. These results demonstrate that learners gained sensitivity to the statistical structure of unfamiliar speech on a very rapid timescale. This ability may play an essential role in early stages of language acquisition, allowing learners to rapidly identify word candidates and "break in" to an unfamiliar language.

Keywords: language acquisition; open data; open materials; reaction time; speech segmentation; statistical learning.

Fig. 1.

Stimulus design (a) and example trial sequence (b). Each speech stream was composed of a randomly selected subset of 12 syllables drawn without replacement from a pool of 24 syllables. These 12 syllables were then randomly distributed to create four trisyllabic words, which were repeated four times each. The resulting 16 words (48 syllables) were concatenated together in pseudorandom order and presented aurally without any pauses between them. Before each stream was played, participants saw and heard a target syllable and were asked to identify it as quickly as possible every time it occurred in the stream.

Fig. 2.

Reaction time (RT) data as a function of triplet position (first, second, or third syllable in the word) and word presentation (first, second, third, or fourth occurrence of the word in the stream). Overall mean RT is shown in (a). Mean RT predicted by main effects of interest (triplet position, word presentation, and their interaction), controlling for the effect of overall syllable position in the stream, is shown in (b). Error bars represent ±1 SEM.