The effect of formant manipulation on the perception of Dutch front vowels

The purpose of this paper is to evaluate the formant suppression effect on the perception of naturally-produced rounded and unrounded front vowels. In an open form response task, participants were asked to identify high and mid Dutch front vowels: the unrounded vowels /i,e/ and the rounded vowels /y,ø/. Vowels were presented either as complete signals or as incomplete signals where the second formant was suppressed. All complete vowels and incomplete unrounded vowels were perceived correctly. The incomplete rounded vowels, however, were more often perceived as their unrounded counterparts. This asymmetry can be explained by: (1) a formant representation where the third formant replaces the suppressed second formant; (2) a mismatch of formant
amplitude for the case of incomplete rounded vowels, and (3) an educated guess based on the frequency of occurrence of the target vowels. Our findings show that the effects in natural vowels are similar to effects found in synthesized vowels. Vowel quality perception is based on formant frequency as primary cue and formant amplitude as secondary cue. Furthermore, the effect of frequency of occurrence provides evidence that vowel categories are not represented with equal strength in the cognitive system.