Features, costs and perception

Here we make some observations about die nature of features and how they are employed in both the target and file join functions. Firstly, it is important to note that the term perceptual is often used in a very general manner, which can lead to confusion over what we are actually talking about. The main point is not to confuse any idea of auditory perception with linguistic perception. It is well known that the ear cleverly processes the signals it receives and has a non-linear response over the spectrum. While tiiis can, and perhaps should, be modelled, it is important to note that this will go only so far, since on top of this sits an independent linguistic perceptual space, which behaves in a quite different manner. 

Rgure 16.12 The effect of category boundaries on perception. The acoustic distance between units a and b and between c and d is relatively high, and that between units b and c is relatively low. However, because units b and c belong to different linguistic categories they will be perceived to have the largest perceptual distance. 

Anodier angle on the problem of why acoustic measures and perception correlate so badly is that the cost function as normally portrayed contradicts basic properties of human perception. As we saw in Section 16.5.6, it is good policy to ignore any objective cost differences below a certain value because humans will not perceive such small differences. Additionally, above a certain value, the costs are so high as to make any distinction between one high cost and another meaningless. 

---