Familiar units prevail over statistical cues in word segmentation.

Area of Science:

• Cognitive Science
• Psycholinguistics
• Language Acquisition

Background:

• Listeners typically use statistical cues, like transitional probabilities between syllables, to identify words in continuous speech.
• However, other cues also play a role in word discovery, and understanding their relative importance is key to comprehending speech segmentation processes.

Purpose of the Study:

• To investigate whether adult learners prioritize familiar units or statistical cues when segmenting continuous speech.
• To dissociate the influence of statistical cues and familiar units in speech segmentation.

Main Methods:

• Participants were familiarized with part-words from an artificial language before an exposure phase.
• Experiment 1 used a two-alternative forced choice (2AFC) task comparing segmentation of words versus part-words.
• Experiment 2 employed a syllable-detection task to mitigate memory and strategy effects.

Main Results:

• Adult learners preferentially segmented familiar part-words, even when these were statistically less cohesive and frequent than actual words.
• An unfamiliarized control group demonstrated typical word segmentation based on statistical cues.
• Results from the syllable-detection task corroborated the preference for familiar units.

Conclusions:

• Familiar units appear to override statistical cues, such as transitional probabilities, in speech segmentation for adult learners.
• This challenges the prevailing view that transitional probability computation is the primary mechanism for word discovery in natural language.