Which speech-recognition materials yield the steepest performance-intensity function?

Which speech-recognition materials yield the steepest performance-intensity function?

• A. Synthetic sentences
• B. Nonsense syllables
• C. CID W‑1
• D. CID W‑22

Answer: (A)

The key idea is that the steepness of a performance-intensity curve reflects how much a listener must rely on the audible phonetic details alone, with little help from context or lexical knowledge. Synthetic sentences are designed so that meaningful meaning is not available from the words themselves, while still maintaining a grammatical structure. That means recognition depends mainly on hearing the phonetic cues clearly. When the speech is barely audible, those cues are hard to extract and accuracy is low; as intensity increases even a little, those phonetic cues become audible and identifying them becomes much easier, so performance improves rapidly. That rapid gain with small increases in intensity yields the steepest PI function among the options. In contrast, word lists like CID W-1 or CID W-22 provide real lexical items and benefit from top-down context and lexical predictability, so listeners can use prior knowledge to help recognition even when some phonetic detail is masked. This context softens the rise in performance as intensity grows, producing a shallower slope. Natural or more context-rich materials tend to show even more redundancy, further flattening the slope.

The key idea is that the steepness of a performance-intensity curve reflects how much a listener must rely on the audible phonetic details alone, with little help from context or lexical knowledge. Synthetic sentences are designed so that meaningful meaning is not available from the words themselves, while still maintaining a grammatical structure. That means recognition depends mainly on hearing the phonetic cues clearly. When the speech is barely audible, those cues are hard to extract and accuracy is low; as intensity increases even a little, those phonetic cues become audible and identifying them becomes much easier, so performance improves rapidly. That rapid gain with small increases in intensity yields the steepest PI function among the options.

In contrast, word lists like CID W-1 or CID W-22 provide real lexical items and benefit from top-down context and lexical predictability, so listeners can use prior knowledge to help recognition even when some phonetic detail is masked. This context softens the rise in performance as intensity grows, producing a shallower slope. Natural or more context-rich materials tend to show even more redundancy, further flattening the slope.