410Animal Memory and Cognition
In the bisection procedure,two levers may be inserted into
an operant box; the rat is reinforced for pressing the left lever
after a tone of one duration and the right lever after a tone
ofanother duration (Church & Deluty, 1977). In the test con-
dition, the rat is presented with tones of intermediate dura-
tion. Of special interest is the duration at which they choose
each lever half the time (50% responding). It develops that
the interval used by rats is the geometric mean of the two in-
tervals not the arithmetic mean. For example, if the intervals
are 4 s and 16 s, respectively, the arithmetic mean is 10
([4+16]/2), whereas the geometric mean is 8 ( (^4) × 16 ). In
this example, 50% responding would occur closer to 8 s than
to 10 s.
Other experiments indicate that animals time linearly and
that they can start and stop their interval clocks. How animals
time has produced much theorizing. It has been variously
suggested that rats time by employing a pacemaker (e.g.,
Church, Meck, & Gibbon, 1994), by using oscillators (e.g.,
Gallistel, 1994), or by using behaviors that predictably fill
given intervals (e.g., Killeen & Fetterman, 1988).
Gallistel and Gibbon (2000) have presented a timing theory
that is highly ambitious in that it attempts to explain a wide
range of phenomena. The theory has been used to explain phe-
nomena as different as delayed classical conditioning to ex-
tinction following different reward schedules in instrumental
conditioning. Whatever the fate of this theory, it is clear that
animals such as rats and pigeons have well-developed capaci-
ties for timing events. How extensively this timing capacity
enters into learning and cognition appears to be a major issue
that will occupy investigators over the near future.
Memory
Memory is among the most intensively investigated topics in
animal cognition. Animal memory may be studied either in
its own right or as a mechanism controlling learning and
performance. Determining under what conditions forgetting
occurs is an example of the former; examining the capacity of
the memory of nonreward stored on one trial to be retrieved
on the next trial so as to correctly anticipate the reward
outcome on that trial is an example of the latter.
It is now recognized that animals can retain information
over long temporal periods. This was not always known. In
the early days of the investigation of animal learning, labora-
tory data suggested that animals possessed only fleeting
memory. A popular procedure devised by W. S. Hunter
(1913) in the early 1900s is a case in point. In Hunter’s pro-
cedure, called delayed reaction,animals that were retained in
a delay chamber could determine which of three doors lead to
food because a light was flashed in front of the correct door.
After the light went off animals ran from the delay chamber
to the doors after varying retention intervals. Rats failed
at this task with delays of as little as 10 s. Raccoons were able
to respond correctly following a delay of up to 25 s.
Contrast such poor performance with some subsequent
findings obtained under both field conditions and in the labo-
ratory. A certain bird, Clark’s nutcracker, stores the seeds of
pine cones in individual caches in the late summer and early
spring when food is plentiful, recovering the seeds months
later when food is scarce. It is estimated that the bird stores
many thousands of seeds in caches of a few seeds each. A
high percentage of seeds is recovered by the bird. Skinner
(1950) trained pigeons to peck for food at a spot on an illu-
minated key. Following a 4 year retention interval the pi-
geons were tested and immediately pecked the correct key.
Wendt (1937) trained a dog to withdraw its foot at the sound
of a tone paired with shock. After a 30 month retention pe-
riod, foot withdrawal to the tone occurred on 80% of the test
trials, only a slight drop from the prior training session.
A currently debated topic concerns whether memory is a
unitary system or is composed of two or more subsystems.
Some examples of currently postulated subsystems are pro-
cedural versus declarative memory, semantic versus episodic
memory, and long-term versus short-term memory (see, e.g.,
Spear & Riccio, 1994). In the animal area one of the most
popular distinctions is that between working memory and
reference memory. Working memoryis concerned with keep-
ing track of information that may change from one trial to the
next.Reference memoryis concerned with isolating impor-
tant relationships in the situation that are stable over trials.
As an example, consider rats rewarded for a running response
on every other trial, under a single alternation schedule of
rewarded and nonrewarded trials. Rats so trained may even-
tually come to run faster on rewarded than on nonrewarded
trials. In this situation working memory would be used to
determine whether reward or nonreward occurred on the
prior trial and thus whether reward or nonreward is to occur
on the current trial. Reference memory would be employed
to learn that rewards and nonrewards occur according to a
particular rule or schedule—a single alternating one.
Working memory, unlike short-term memory, may be
effective following long retention intervals. In the single-
alternation situation, as indicated earlier, rats have employed
the memory of the reward outcome on the prior trial to antic-
ipate the reward outcome correctly on the current trials when
trials were separated by 24 hr (see, e.g., Capaldi, 1994). The
single-alternation situation is useful for understanding a sec-
ond popular distinction between memories in the animal area
as well as in human memory: that between retrospective and
prospective memory. In the case of the single-alternation