Page 173 - Foundations of Cognitive Psychology : Core Readings

Perception 177

Figure 7.35
Droodles. What are these animals? Do you see in (A) an early bird who caught a very strong worm
and in (B) a giraffe’s neck? Each of these figures can be seen as representing something familiar to
you, although this perceptual recognition usually does not occur until some identifying information
is provided.

of laughter. Even so, people rarely realize that there are gaps in the physical
signal they are experiencing. This phenomenon is known as phonemic restoration
(Warren, 1970). Samuel (1981, 1991) has shown that subjects often find it diffi-
cult to tell whether they are hearing a word that has a noise replacing part of
the original speech signal or whether they are hearing a word with a noise just
superimposed on the intact signal (see the top panel of figure 7.36).
The bottom panel of figure 7.36 shows how bottom-up and top-down pro-
cesses could interact to produce phonemic restoration (McClelland & Elman,
1986). Suppose part of what your friend says at a noisy party is obscured so
that the signal that arrives at your ears is ‘‘I have to go home to walk my
(noise)og.’’ If noise covers the /d/, you are likely to think that you actually
heard the full word dog. But why? In figure 7.36, you see two of the types of
information relevant to speech perception. We have the individual sounds that
make up words, and the words themselves. When the sounds /o/ and /g/ ar-
rive in this system, they provide information—in a bottom-up fashion—to the
word level (we have given only a subset of the words in English that end with
/og/). This provides you with a range of candidates for what your friend
might have said. Now top-down processes go to work—the context helps you
select dog as the most likely word to appear in this utterance. When all of this
happens swiftly enough—bottom-up identification of a set of candidate words
and top-down selection of the likely correct candidate—you’ll never know that
the /d/ was missing. Your perceptual processes believe that the word was in-
tact. (You may want to review figure 7.4 to see how everything in this chapter
fits together.)