Perceptual bias in speech error data collection: insights from Spanish speech errors

Area of Science:

• Linguistics
• Psycholinguistics
• Speech Production

Background:

• Naturalistic speech errors are a valuable tool for studying language production.
• Potential biases in data collection can affect the reliability of speech error analysis.
• Previous research primarily focused on Germanic languages, with limited data from Romance languages.

Purpose of the Study:

• To assess the reliability and validity of naturalistic speech errors in language production research.
• To identify and mitigate potential biases in collecting speech error data.
• To compare speech error patterns in Spanish with those reported for Germanic languages.

Main Methods:

• Analysis of a new corpus of Spanish speech errors collected by multiple, theoretically naïve observers.
• Comparison of findings with existing corpora of speech errors from English, German, Dutch, and Spanish.
• Identification of specific patterns in Spanish speech errors, such as phonological error location, consonant position, syllable stress, error directionality (perseveration vs. anticipation), and lexical bias.

Main Results:

• Spanish speech errors exhibit patterns contrary to predictions based on bias, including more errors between words, less involvement of word-initial consonants, equal error likelihood in stressed/unstressed syllables, more perseverations than anticipations, and no lexical bias.
• Results from the new Spanish corpus were consistent with previous reports, even with different collection methods.
• Despite potential susceptibility to bias, collecting speech errors in the wild with multiple collectors appears reasonably unbiased.

Conclusions:

• Naturalistic speech error collection is a largely unbiased method for language production research.
• Observed differences in speech error patterns between Spanish and Germanic languages represent genuine cross-linguistic variations.
• The findings support the use of naturalistic speech errors for robust cross-linguistic studies in language production.