Whether pattern recognition 
is based upon parallel 
processing or upon serial 
processing is still in dispute. 
In serial processing, one item 
is processed at a time. In 
parallel processing, all items 
are processed simultaneously. 
There are difficulties with 
each possibility. Serial 
processing would take too long, 
even if each comparison of the 
percept with a different trace 
required only a few 
milliseconds. After all, there 
are almost an infinite number 
of memories, while 
recognition takes only a 
fraction of a second. Yet the 
results of an experiment 
conducted by Saul Sternberg of 
Bell Laboratories favors serial 
comparison. Subjects were first 
given a list of items to 
memorize. Shortly thereafter 
they were shown an "old" or a 
"new" item and were asked to 
indicate whether or not it had 
been on the list. Sternberg 
discovered that observers' 
reaction time increased by a 
constant amount for each 
additional item on the 
memorized list. This suggests 
that the list had been 
"searched" serially. On the 
other hand, parallel 
processing would take little 
time, but then what 
mechanism would permit 
matching a percept to all 
possible traces at once? 
 
    What properties of an object's 
shape make it seem more or 
less similar to other shapes? 
The answer to this question is 
of direct relevance to our 
understanding of recognition 
as well as form perception. In 
the case of recognition, such 
properties would provide the 
basis for the search of one's 
memories. If we wanted to 
build a machine that could 
"recognize" figures—–for 
example, one that could 
"read"—–we must know what 
properties of the object now 
detected govern the search for 
similar memories, or 
prototypes, stored in the 
machine's memory.