Comprehending Spoken Words Without Hearing Phonemes: A Case Study

Charlotte Jacquemot, Emmanuel Dupoux, Christophe Pallier, Anne‐Catherine Bachoud‐LéviCortex 2002 · DOI

TL;DR

This paper presents a fascinating case study of a patient with aphasia who exhibits a paradoxical dissociation: severe difficulty in discriminating basic speech sounds (phonemes) yet relatively intact comprehension of spoken words. This challenges traditional models of speech processing that assume a strict sequential dependency between phonemic analysis and word recognition.

What Problem Does This Paper Address?

Understanding spoken words involves processing a continuous acoustic signal and breaking it down into discrete linguistic units, primarily phonemes. Models of spoken word comprehension typically propose two main levels:

  1. Sub-lexical level: This level is responsible for transforming the acoustic signal into phonological units like phonemes or syllables.
  2. Lexical level: This level involves matching these phonological units to stored word representations in our mental lexicon and retrieving their meaning.

While many studies have documented patients with lexical impairments (difficulty recognizing words despite intact phoneme processing), cases of sub-lexical impairment without significant comprehension deficits are rare. This is counterintuitive because if the initial phonemic analysis is flawed, it should logically lead to problems in identifying words. This paper investigates a patient who presents the opposite of the expected pattern: a severe deficit at the sub-lexical level but only mild impairment at the lexical level, prompting a re-evaluation of the necessity and nature of sub-lexical processing for word comprehension.

Methods

The researchers employed a series of psycholinguistic tasks to assess the patient's speech processing abilities, focusing on the dissociation between sub-lexical and lexical levels.

  • AX Discrimination Task: This is a standard task for assessing phonemic processing. Participants hear two stimuli (words or nonwords) and must decide if they are identical or different. In this study, pairs of C₁VC₂VC₃ structured words and nonwords were used, differing by a single consonant. The task was performed with stimuli spoken by the same speaker or by two different speakers to probe sensitivity to acoustic variations.
  • Lexical Decision Task: This task assesses lexical access. Participants hear a stimulus and must decide if it is a real word or a nonword. The stimuli used were words and nonwords that differed by only one phoneme from a word. The researchers also analyzed performance based on word frequency (high vs. low) and concreteness (concrete vs. abstract).
  • Auditory-Written Word Matching: This task probes the precision of phonological representations used for word recognition. The patient listened to a word and had to select its written form from three options (target word, phonologically related word, phonologically related nonword). This task focused on high-frequency concrete words.
  • Auditory Word-Picture Matching: Similar to the above, this task assesses the link between auditory input and semantic/lexical representations. The patient heard an auditory stimulus (either the correct name of a picture or a phonologically similar distractor) and had to decide if it matched a presented picture. This task also focused on high-frequency concrete words.

Datasets

The study did not introduce new datasets but utilized existing stimuli and methodologies:

  • Stimuli Construction: Words and nonwords were constructed with specific phonological structures (C₁VC₂VC₃) and varied by a single consonant for the AX discrimination task. For the lexical decision task, words and nonwords differing by one phoneme were used, categorized by frequency and concreteness.
  • Word Frequency Matching: Words were matched for frequency based on the norms provided by Content et al. (1990).

Key Insights and Ideas

  • Paradoxical Dissociation: The most striking finding is the patient's severe impairment in discriminating phonemes (performing at chance level) while showing significantly better, albeit not perfect, performance in tasks requiring word recognition. This directly challenges models that posit a strict, sequential dependency where flawless sub-lexical processing is a prerequisite for lexical access.
  • Lexical Processing Can Be Spared Despite Sub-lexical Deficits: The patient's ability to perform lexical decision and word-picture/word-matching tasks suggests that some form of lexical access and representation can occur even with severely degraded phonemic input.
  • Heterogeneity within Lexical Impairment: The patient's performance in the lexical decision task was not uniform. They performed better with high-frequency and concrete words, indicating that the lexical impairment, while present, was not global and was influenced by word properties.
  • Potential for Direct Access or Alternative Routes: The findings open the door to explanations involving "direct access" models, where acoustic patterns might directly access lexical representations without full phonemic decomposition, or models where phonological information is processed differently for recognition versus explicit judgment tasks.

Key Results

  • AX Discrimination: The patient performed at chance level (53% correct for same speaker, 52% for different speakers) on the AX discrimination task, indicating a severe deficit in phonemic discrimination.
  • Lexical Decision: The patient performed significantly above chance (68% correct) on the lexical decision task. Performance was better for high-frequency words (70.3% correct) than low-frequency words (53.1% correct) and for concrete words (69.8% correct) than abstract words (37.5% correct).
  • Auditory-Written Word Matching: The patient achieved 90.1% correct, significantly above chance, indicating a strong ability to match spoken words to their written forms, particularly for high-frequency concrete words.
  • Auditory Word-Picture Matching: The patient performed very well at 89.8% correct, suggesting intact ability to link spoken words to their corresponding pictures.

Limitations and Future Work

  • Nature of the Sub-lexical Deficit: The paper acknowledges that the exact nature of the sub-lexical deficit is unclear. While the AX task suggests a phonemic discrimination issue, the authors propose it might be a deficit in "reading out" phonological information rather than a complete lack of processing. The possibility of an impaired acoustic level itself is also raised but considered less likely.
  • Role of Acoustic vs. Phonological Processing: The authors discuss the difficulty in disentangling acoustic processing from phonological processing, especially in the AX task where identical acoustic tokens were presented.
  • Future Directions: The paper suggests two main theoretical avenues:
  • Direct Access Models: These models propose that words can be accessed directly from the acoustic signal without mandatory phonemic analysis, potentially explaining the patient's spared lexical comprehension.
  • Dual-Route Models (e.g., Shortlist): This approach suggests that different phonological representations might be used for explicit judgment tasks (like AX discrimination) versus implicit word recognition. The patient might have impaired conscious access to phonological details for judgment but retain the ability to use them for recognition. This is likened to blindsight phenomena.

Further research could explore the patient's ability to process other phonological features, the neural correlates of this dissociation, and how different types of aphasia might manifest such unusual patterns.

Summary updated

BibTeX
@inproceedings{jacquemot2002comprehending,
  author = {Charlotte Jacquemot and Emmanuel Dupoux and Christophe Pallier and Anne‐Catherine Bachoud‐Lévi},
  title = {{Comprehending Spoken Words Without Hearing Phonemes: A Case Study}},
  booktitle = {Cortex},
  year = {2002},
  doi = {10.1016/s0010-9452(08)70059-5},
}
Figure