The role of linguistic experience in the processing of probabilistic information in production

ABSTRACT

Speakers track the probability that a word will occur in a particular context and utilise this information during phonetic processing. For example, content words that have high probability within a discourse tend to be realised with reduced acoustic/articulatory properties. Such probabilistic information may influence L1 and L2 speech processing in distinct ways (reflecting differences in linguistic experience across groups and the overall difficulty of L2 speech processing). To examine this issue, L1 and L2 speakers performed a referential communication task, describing sequences of simple actions. The two groups of speakers showed similar effects of discourse-dependent probabilistic information on production, suggesting that L2 speakers can successfully track discourse-dependent probabilities and use such information to modulate phonetic processing.

KEYWORDS:

• Language production
• L2 speech
• phonetics
• reduction

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1. A series of analyses considering the influence of these proficiency variables on performance revealed no reliable effects. Given the small sample size of the groups and the substantial variation in the dependent measure we draw no strong conclusions from these null effects.

2. An error in the counterbalancing of one list was discovered during testing. In length 4 trials, all fourth events used the same action (rotate), leading to some repetition of that action within a trial. However, because the fourth event of trials were never analyzed, the general structure of the list was still acceptable for the purposes of the experiment. Therefore, data from this participant was included. The error was rectified for the other two randomized versions of this list.

3. MCMCglmm's default fixed effects prior was utilized. The priors for the random effect covariance matrices and residual covariance matrix were each given by an inverse-Wishart distribution, with an identity scale matrix and the lowest possible degrees of freedom.

4. These were taken from 10 independent chains of 10,000 iterations, thinned so that every 10th sample was used, with 3000 iterations as burn-in. A multivariate potential scale reduction factor (Brooks & Gelman, 1997) of 1 confirmed mixing of these chains.

Additional information

Funding

This research was supported by the National Institutes of Health under [grant number 1R21HD077140]; Eunice Kennedy Shriver National Institute of Child Health and Human Development.