Abstract

When learning a language, it is crucial to know which syllables of a continuous sound string belong together as words. Human infants achieve this by attending to pauses between words or to the co-occurrence of syllables. It is not only humans that can segment a continuous string. Songbirds learning their song tend to copy ‘chunks’ from one or more tutors’ songs and combine these into their own song. In the tutor songs, these chunks are often separated by pauses and a high co-occurrence of elements, suggesting that these features affect chunking and song learning. We examined experimentally whether the presence of pauses and element co-occurrence affect the ability of adult zebra finches to discriminate strings of song elements. Using a go/no-go design, two groups of birds were trained to discriminate between two strings. In one group (Pause-group), pauses were inserted between co-occurring element triplets in the strings, and in the other group (No-pause group), both strings were continuous. After making a correct discrimination, an individual proceeded to a reversal training using string segments. Segments were element triplets consistent in co-occurrence, triplets that were partly consistent in composition and triplets consisting of elements that did not co-occur in the strings. The Pause-group was faster in discriminating between the two strings. This group also responded differently to consistent triplets in the reversal training, compared to inconsistent triplets. The No-pause group did not differentiate among the triplet types. These results indicate that pauses in strings of song elements aid song discrimination and memorization of co-occurring element groups.

Highlights

  • Learning which syllables of a continuous speech stream belong together as words is one of the first challenges that human infants face when they are acquiring a language

  • The results of the string discrimination training indicate that zebra finches discriminated more readily between two strings when there were longer pauses between element triplets

  • It cannot be excluded that this enhanced discrimination is affected by the total duration of the strings, which increases with increased pause length

Read more

Summary

Introduction

Learning which syllables of a continuous speech stream belong together as words is one of the first challenges that human infants face when they are acquiring a language. Pauses are not always reliable and can occur both between and within words. Another way of detecting regularities is by paying attention to the transitions between syllables (Saffran et al 1996a, b; Aslin et al 1998). Syllables that occur together more often and have a higher transitional probability are more likely to form a word. Infants use this feature to correctly segment speech streams. Computational models support the hypothesis that transitional information can be sufficient for correct word segmentation (Brent and Cartwright 1996; Swingley 2005)

Methods
Results
Conclusion
Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call