Speech Perception by 6- to 8-Month-Olds in the Presence of Distracting Sounds

Abstract

The role of selective attention in infant phonetic perception was examined using a distraction masker paradigm. We compared perception of /bu/ versus /gu/ in 6- to 8-month-olds using a visual fixation procedure. Infants were habituated to multiple natural productions of 1 syllable type and then presented 4 test trials (old–new–old–new). Perception of the new syllable (indexed as novelty preference) was compared across 3 groups: habituated and tested on syllables in quiet (Group 1), habituated and tested on syllables mixed with a nonspeech signal (Group 2), and habituated with syllables mixed with a non-speech signal and tested on syllables in quiet (Group 3). In Groups 2 and 3, each syllable was mixed with a segment spliced from a recording of bird and cricket songs. This nonspeech signal has no overlapping frequencies with the syllable; it is not expected to alter the sensory structure or perceptual coherence of the syllable. Perception was negatively affected by the presence of the auditory distracter during habituation; individual performance levels also varied more in these groups. The findings show that perceiving speech in the presence of irrelevant sounds poses a cognitive challenge for young infants. We conclude that selective attention is an important skill that supports speech perception in infants; the significance of this skill for language learning during infancy deserves investigation.

Notes

¹Although the dBA scale weights frequencies to better equate for differential sensitivities across the spectrum, the adjustments are focused on the low-frequency end of the spectrum. The A scale weights energy in midfrequencies (speech range) and the high frequencies (5–10Khz) similarly, even though phone curves show that the ear is less sensitive to sound energy in this higher frequency range (Lipscomb, 1978.) This explains why we find some difference in our sound level measures when the syllables and nonspeech signals are presented at comparable loudness levels.

²Although the perceptual grouping of the syllables and nonspeech segments into separate streams was very compelling for adults, we have no direct evidence that this was so for the infants. The synchronous onset of the two signals and the arrival of the two signals from the same spatial location could prompt infants to integrate the speech and distracter as a single auditory stream. However, even in this context, substantial cues to support stream segregation remain. The syllables and distracter segments are natural, coherent signals emanating from markedly dissimilar sound sources. The distracter segment was also constantly changing and, postonset, these two signals diverge completely in temporal, spectral, and amplitude structure. Delivering the signals from different locations was unlikely to promote better stream segregation unless the physical separation was large because spatial separation is not a strong cue in source segregation for infants (Ashmead, Clifton, & Perris, 1987).

³In Table 1, the standard deviation for preference ratio scores is only slightly lower in the quiet group compared to the distracter group and the DQ group; this is due to an obvious outlier in the quiet group (see Figure 3). Variability differences across groups are bigger when variance is indexed using the semi-interquartile range, which is less affected by extreme scores (quiet group = .04, distracter group = .12, DQ group = .08).

⁴Infant preference has been shown to gradually shift from familiarity to novelty preference as processing progresses such that familiarity preference emerges in the initial processing of a stimulus and novelty preference appears later as infants habituate and their interest in a stimulus begins to decline (Hunter & Ames, 1988). Some infants with very low preference ratio scores may display a reliable (familiarity) preference for the habituation syllable. These infants are showing some ability to discriminate the syllables but they have not encoded the habituation syllable as well as the infants who show a reliable novelty preference.

⁵Infants with preference ratio scores above versus below the preference ratio criterion were collapsed across the distracter and DQ groups. Infants in the above preference ratio criterion subgroup required significantly less listening time to reach the habituation criterion (M = 93.6 sec, SD = 40.6) compared to the infants in the below preference ratio criterion subgroup (M = 126 sec, SD = 31.7), t(18) = 7.07, p > .009, partial η² = .174. Thus, the good syllable encoders took less time to habituate than the poor syllable encoders.