Accident - execute: increased activation in nonnative listening

Mirjam BroersmaInterspeech 2006

TL;DR

This study reveals that non-native English listeners experience heightened "lexical competition" when processing words with similar beginnings but different vowels, leading to slower and less accurate recognition compared to native speakers. This suggests that the challenges of distinguishing subtle phonetic differences in a second language can significantly impair spoken word recognition.

What Problem Does This Paper Address?

Understanding spoken language relies on activating multiple potential word candidates (lexical representations) and then selecting the correct one. This process involves "lexical competition," where activated words vie for recognition. While this competition is generally beneficial for native speakers, it can become problematic when many words sound similar to the input. This paper investigates whether this problem is amplified for non-native listeners. Specifically, it explores how the perception of English words with partially overlapping onsets (like "accident" and "execute," which differ only in the vowel of the first syllable) affects lexical activation in both native English speakers and proficient Dutch learners of English. The core question is whether non-native listeners, who may struggle with fine phonetic distinctions, experience greater lexical competition than native speakers.

Methods

The researchers employed a two-pronged approach:

  • Cross-Modal Fragment Priming Task: Participants heard the initial part of an English word (the "prime") and then saw a full English word or non-word on a screen (the "target"). They had to decide if the target was a real English word. The key manipulation was the relationship between the prime and the target:
  • Identity Prime: The prime was the beginning of the target word itself (e.g., "acci" for "accident"). This should facilitate recognition.
  • Mismatch Prime: The prime was the beginning of a word that differed from the target only in the critical vowel (e.g., "acci" for "execute"). This is where increased competition was expected for non-native listeners.
  • Control Prime: The prime was from an unrelated word. The researchers measured reaction times and accuracy to assess how much the primes facilitated or hindered target recognition.

  • Phonetic Categorization Task: This task was designed to assess the participants' ability to distinguish between the English vowels /æ/ (as in "cat") and /ɛ/ (as in "bed"), which were the source of the difference in the "accident"/"execute" pair. Participants heard auditory stimuli and had to categorize them as either /æ/ or /ɛ/. This was crucial for understanding the non-native listeners' perceptual capabilities.

Datasets

Not explicitly detailed as separate datasets. The study used:

  • 24 pairs of trisyllabic English words with stress on the first syllable, differing in the /æ/-/ɛ/ contrast in the first syllable (e.g., "accident" - "execute").
  • Associated phonologically and semantically unrelated trisyllabic words for control primes.
  • Filler words and non-words were also constructed to ensure a robust experimental design.
  • All auditory stimuli were recorded by a male native English speaker.

Key Insights and Ideas

  • Non-native listeners struggle more with phonetic ambiguity: The core finding is that the perceptual difficulty of distinguishing subtle phonetic contrasts (like /æ/ vs. /ɛ/) in a second language leads to greater activation of competing words.
  • Lexical competition is amplified in non-native listening: When non-native listeners misperceive a sound, it can activate a wider range of lexical candidates for a longer duration, making it harder to pinpoint the intended word.
  • The "accident" vs. "execute" contrast is a good probe: This specific pair, differing only in a vowel that is problematic for Dutch speakers, effectively highlighted differences in processing between native and non-native listeners.
  • Phonetic categorization ability predicts priming effects: The study confirmed that Dutch listeners, who performed worse on the /æ/-/ɛ/ categorization task, also showed greater effects of lexical competition in the priming task.

Key Results

  • Cross-Modal Priming:
    • Native English listeners showed clear facilitation from identity primes and no significant facilitation (or even slight inhibition) from mismatch primes.
  • Dutch listeners showed facilitation from both identity and mismatch primes, indicating that the mismatch primes were still activating words that were perceived as plausible, leading to increased competition.
  • Dutch listeners made significantly more errors overall than English listeners.
  • Phonetic Categorization:
    • English listeners correctly categorized the /æ/-/ɛ/ contrast with high accuracy (85.6% for /æ/, 91.7% for /ɛ/).
  • Dutch listeners performed significantly worse (54.6% for /æ/, 66.9% for /ɛ/), with their performance on /æ/ being only slightly above chance.
  • Both groups showed a bias towards perceiving /ɛ/.

Limitations and Future Work

  • Limited scope of phonetic contrast: The study focused on a single phonetic contrast (/æ/-/ɛ/). The authors suggest that non-native listeners encounter many such problematic contrasts, and the cumulative effect could be substantial.
  • Proficiency level: While Dutch participants were described as having a high level of proficiency, the study did not explore the effects of varying proficiency levels.
  • Future Work: The authors propose that the increased lexical activation in non-native listening, stemming from perceptual ambiguities, can significantly complicate spoken word recognition. Future research could explore the extent to which this effect is exacerbated by multiple phonetic ambiguities within a single word and investigate strategies listeners might use to overcome these challenges.

Summary updated

BibTeX
@inproceedings{broersma2006accident,
  author = {Mirjam Broersma},
  title = {{Accident - execute: increased activation in nonnative listening}},
  booktitle = {Interspeech},
  year = {2006},
}
Figure