Exploring cultural impact on word segmentation

Project: Research project (funded)Research

Project Details


British infants, like Dutch and Parisian French infants, fail to segment trained strong-weak words from passages at 7.5 months of age, when North American infants (learning English in the US or French in Montreal) have been found to do so (Nazzi et al., 2006, 2008). It is our hypothesis that differences in the input to American as compared with British infants may lead to these observed differences in experimental findings in the prelinguistic period. This proposal explores ways in which the differences between the infant directed speech (IDS) to which American and British infants are exposed may be linked to differences in their performance on perception experiments (using the Head Turn [HT] paradigm). Our specific objectives are to:

1. Establish the basic group-age profile for infant segmentation and word form recognition in the United Kingdom and the United States.
2. Establish characteristics of IDS as regards quantity, structure and prosody in each country.
3. Explore the link between the differences in performance on the HT task in the two locations and the differential characteristics of IDS.

Layman's description

There is evidence to suggest that infants raised hearing British English may reach some milestones – (a) the ability to identify words within continuous speech and (b) first word production – later than infants raised hearing American English. Our goal was to test for such a lag using the same experimental procedure in both countries, and if a lag was found, to determine whether it could be explained by differences in quantity or quality of parental speech in the two cultures. We collected data on infants’ word recognition (in isolation) and segmentation (in passages) and recorded parents speaking to their infants in both countries.

We have now tested 91 British infants, aged 8, 9, 10.5 months. No group could segment words from passages. Infants in both cultures recognised isolated words at 11 months.
Recruitment in the US proved problematic; we are still testing American infants, for whom final segmentation results are not yet available. Thus we do not yet have conclusive findings as to the existence of a developmental lag between American and British infants.

We have recorded speech from 12 British and 4 American families. We are continuing to recruit US families. Preliminary results show cultural differences: American parents produce more words and American mothers use more exaggerated intonation, make more pauses after words, and produce more words in isolation. These effects may make words in the American input more salient, and could explain the claimed difference in the attainment of linguistic milestones.

Key findings

Evidence that infants exposed to British vs. American English differ in early word production and in their ability to detect words in continuous speech comes from the Communicative Developmental Inventory (Schafer et al., 2000), observational studies (Keren-Portnoy et al., 2000) and experimental studies: A previous study in our lab with 8-month-olds failed to find successful detection of words in running speech, whereas American infants did succeed at this (Jusczyk et al., 1999). By using the same experimental procedure in the US and the UK we hoped to find out whether there is a developmental lag, and to relate these findings to parents' speech to infants in the two communities.

We have now successfully tested 91 British infants, aged 8, 9, 10.5 months. No group could detect words in passages. Previous studies in our lab showed that British infants recognise isolated words at 11 months.

Recruitment in the US proved problematic: 25 American infants have been successfully tested to date; testing continues. Final results for the American infants are not yet available, but the isolated-word recognition findings were replicated in our US lab, at the same age. A younger group in the US did not succeed at an earlier age than do the British infants. We do not yet have conclusive findings as to whether or not there is a developmental lag between American and British infants.

We recorded speech from all 12 British and 4 (out of 12) American families. Two analyses were carried out, one based on automated counts of parental word use in the entire recording, and the other on two tellings by the infants’ mothers of two wordless picture books that we supplied, featuring one object per page. Preliminary results, based on four families from each culture, show the following differences:

• the American parents produce more words per hour, US mean 2711, UK mean 2309 words.
Acoustic analysis of 149 one-syllable words serving as ‘targets’ (i.e., pictured) in the book reading episodes, comparing target words to the speech which preceded them showed:
• higher pitch differences in the American vs. the British input speech
• no difference in the duration ratio (length of the syllable) between the target syllable and the syllables in the preceding speech between US and UK mothers
• No clear picture regarding difference in loudness between the target word and the preceding speech.
• American mothers made longer pauses immediately following the target words than did British mothers.
Analysis of the transcribed book reading episodes based upon 1119 target words showed more isolated words in the US than in the UK.
These characteristics (higher pitch differences, more pauses following the word and more words in isolation) are likely to make words more salient in the American than in the British input, and could explain the claimed difference in the attainment of linguistic milestones.

Effective start/end date1/05/0930/04/10