Peer-reviewed research
Speech recognition with primarily temporal cues
Robert V. Shannon et al., Science, 1995 · DOI 10.1126/science.270.5234.303
Read the originalRead on October 17, 2026
- What it is
- A 1995 study from the House Ear Institute in Los Angeles that tested listeners on speech whose spectral detail was largely removed. Envelopes from a few broad frequency bands modulated noise of the same bandwidths, and listeners still identified consonants, vowels and words.
- How much weight it carries
- Peer-reviewed and widely cited. The test used simulated (vocoded) speech and simple sentences, so the result shows what temporal cues can carry, not how well recognizers perform on everyday recordings.
Facts taken from this source
- Shannon and colleagues report that increasing the number of bands to beyond three or four gave no further improvement in their results.
“further improvement as the number of bands increased to three or four” (Results, voicing and manner)