Validating explicit rating tasks for measuring pronunciation biases: A case study of ING variation

Area of Science:

• Psycholinguistics
• Sociolinguistics
• Corpus Linguistics

Background:

• Spoken language exhibits significant variability due to pronunciation variants.
• Psycholinguistic research often uses rating tasks to gauge user biases toward these variants.
• Empirical validation of whether these elicited biases reflect actual usage is lacking.

Purpose of the Study:

• To empirically validate if experimentally elicited user estimates of pronunciation variants accurately reflect real-world usage patterns.
• To investigate the correlation between user estimates and conversational speech data for English ING pronunciations.
• To identify potential asymmetries in the relationship between elicited and corpus word biases.

Main Methods:

• Correlating user estimates from explicit rating tasks with conversational speech data.
• Analyzing English variable ING pronunciations under different experimental prompts.
• Comparing experimentally elicited word biases with corpus word biases.

Main Results:

• Explicit rating tasks yield word biases that significantly correlate with corpus word biases.
• The observed correlations between elicited and corpus word biases are modest.
• Asymmetries exist in the relationship between elicited word biases and corpus word biases.

Conclusions:

• While rating tasks offer insights into word biases, they only partially capture real-world usage.
• Future research should integrate word biases into the study of sociolinguistic variation and language processing.
• The findings highlight the need for careful interpretation of experimentally elicited biases in relation to actual language use.