The Influence of Distraction on Speech Processing: How Selective is Selective Attention?

Sandra I. Parhammer, Miriam Ebersberg, Jenny Tippmann, Katja Stärk, Andreas Opitz, Barbara Hinger, Sonja RossiInterspeech 2019

TL;DR

This study investigates whether our brains automatically process grammatical errors in speech, even when we're not paying attention. By using EEG to track brain activity while participants performed either a semantic or a syntactic task, the researchers found that grammatical errors in unattended speech are largely ignored unless attention is specifically directed to syntax, suggesting that morphosyntactic processing is not fully automatic.

What Problem Does This Paper Address?

Understanding spoken language is a complex cognitive feat, requiring us to filter out irrelevant information and focus on what's important. A key question in speech processing research is the extent to which different aspects of language, like grammar, are processed automatically. Specifically, this paper asks: when we're distracted, does our brain still catch grammatical errors in speech, or are these errors only processed when we're actively listening for them? Previous research has shown that syntactic errors can elicit specific brain responses (like the P600 component), but these responses often disappear when participants are distracted. However, the role of error severity and the specific nature of the attentional task have not been fully explored. This study aims to clarify how selective attention influences the processing of morphosyntactic errors in German irregular verbs, considering different degrees of error severity.

Methods

The researchers employed electroencephalography (EEG) to record brain activity from participants while they listened to German sentences. The core of the study involved manipulating irregular verbs within these sentences to create three conditions:

  • Correct: The verb is in its correct past tense form.
  • Incorrect but Attested: The verb uses an irregular past tense form that exists in German, but is not the correct one for the context.
  • Incorrect and Unattested: The verb uses a past tense form that is not a valid irregular form in German.

Two experiments were conducted, differing only in the task participants performed after hearing each sentence:

  • Experiment 1 (Semantic Judgement Task): Participants were shown a picture and had to decide if the object was mentioned in the sentence. This task was designed to direct attention away from the grammatical structure and towards the meaning of the words.
  • Experiment 2 (Syntactic Judgement Task): Participants were shown emoticons and had to decide if the sentence was grammatically correct. This task explicitly directed attention towards the grammatical structure of the sentence.

EEG data was analyzed by segmenting brain activity around the critical verb and calculating grand averages for each condition. Statistical analyses (ANOVAs) were performed on different time windows and brain regions to identify significant differences in neural responses.

Datasets

Not discussed in the paper. The stimulus material was created by the researchers for this study.

Key Insights and Ideas

  • Attention is Crucial for Catching Grammatical Errors: The most striking finding is that when participants were distracted by a semantic task (Experiment 1), their brains showed no significant differences in response to correct vs. incorrect verbs. This strongly suggests that processing these morphosyntactic errors is not an automatic process that happens regardless of attention.
  • Syntactic Focus Unlocks Error Detection: Conversely, when participants were explicitly tasked with judging the grammar (Experiment 2), they did show distinct brain responses to the incorrect verbs, particularly the "incorrect but attested" type. This highlights the power of focused attention in engaging syntactic processing mechanisms.
  • Error Severity Matters (and Interacts with Attention): While the "incorrect and unattested" errors (the most severe) didn't elicit a significant response in either experiment, the "incorrect but attested" errors did show a response in the syntactic task. This suggests that the degree of violation might influence how readily it's processed, but only when attention is directed appropriately.
  • N400-like Response, Not P600: Interestingly, the brain response observed in Experiment 2 for the incorrect attested verbs was a broadly distributed negativity resembling an N400 component, rather than the typical P600 component often associated with syntactic processing. The authors suggest this might be due to the sentence structure requiring both syntactic and lexico-semantic processing, which the N400 is known to reflect.

Key Results

  • Experiment 1 (Semantic Task): No significant differences in ERP components were found between the correct, incorrect attested, and incorrect unattested verb conditions.
  • Experiment 2 (Syntactic Task):
    • A broadly distributed negativity, similar to an N400, was observed between 300-900 ms for the "incorrect attested" condition compared to the "correct" condition.
  • A significant interaction between condition and electrodes was found between 900-1300 ms, with post-hoc tests confirming a negativity for the "incorrect attested" relative to the "correct" condition at the Fz electrode.
  • No significant effects were found for the "incorrect unattested" condition in either time window.

Limitations and Future Work

  • Limited Error Types: The study focused on irregular verb past tense forms. Future work could explore other types of morphosyntactic errors (e.g., subject-verb agreement, gender agreement) and their interaction with attention.
  • N400 vs. P600: The observed N400-like response in the syntactic task, rather than the expected P600, suggests that the processing of these specific errors in naturalistic sentences might involve more than just pure syntactic repair. Further research could disentangle these components.
  • Automaticity Debate: While the study strongly suggests that morphosyntactic processing of these errors is not fully automatic, the exact threshold for automaticity remains an open question.
  • Clinical Implications: The authors suggest that deficits in selective attention could potentially impact language processing abilities, opening avenues for research into populations with attention disorders.

Summary updated

BibTeX
@inproceedings{parhammer2019influence,
  author = {Sandra I. Parhammer and Miriam Ebersberg and Jenny Tippmann and Katja Stärk and Andreas Opitz and Barbara Hinger and Sonja Rossi},
  title = {{The Influence of Distraction on Speech Processing: How Selective is Selective Attention?}},
  booktitle = {Interspeech},
  year = {2019},
}
Figure