How words can and cannot be learned by observation.

Area of Science:

• Cognitive Psychology
• Developmental Psychology
• Linguistics

Background:

• Understanding how children acquire new vocabulary is crucial for developmental psychology.
• Existing models of cross-situational word learning often assume complex statistical analysis of multiple word-meaning hypotheses.
• Investigating naturalistic learning scenarios provides insight into real-world language acquisition.

Purpose of the Study:

• To investigate how individuals learn words from hearing them in varied contexts.
• To determine the informativeness of naturalistic word learning instances.
• To explore the role of cross-situational information in word acquisition and test existing learning models.

Main Methods:

• Adult participants watched videotaped vignettes of parents speaking target words to infants, with the words replaced by beeps or nonsense words.
• Participants attempted to identify the "mystery words" to assess learning informativeness.
• Experiments manipulated the presentation of highly informative vignettes across different contexts to examine cross-situational learning mechanisms.

Main Results:

• Most natural word learning instances (90%) were found to be uninformative, while a small percentage (7%) were highly informative, correlating with identification accuracy.
• Preschoolers demonstrated similar sensitivity to informative instances.
• Learners utilized a single meaning hypothesis per word, retaining it unless disconfirmed, rather than storing multiple hypotheses.

Conclusions:

• Word learning from hearing words across contexts appears to follow a one-trial "fast-mapping" procedure, even with referential uncertainty.
• Findings challenge models that propose storage and statistical tabulation of multiple meaning hypotheses.
• Learners do not retain alternative hypothesized meanings or details of past learning situations, favoring a simpler learning strategy.