For example, Ericsson et al.

(1993) compared student

violinists who were the “best” with those who were merely
“good” and with those who were in training to become music
teachers. The best students had accumulated about 7,400 hours
of deliberate practice by the age of 18 compared to 5,300
hours for the good students and 3,400 hours for the teachers-
in-training. Such differences account for why elite performers
so often report having begun their training at an early age. An
early start allows one to begin accumulating the huge number
of practice hours needed to outperform others. Those who
begin at a later age are simply unable to catch up. Because
deliberate practice is so effortful, the amount that can be
tolerated each day is necessarily limited. For this reason, elite
performers often practice about 4 hours per day. Ericsson et al.
(1993), for example, found that the best violin students
engaged in solitary practice (which was judged to be the most
important type of practice) for approximately 3.5 hours per
day, spread out across two to three sessions, each session
lasting an average of 80 minutes. Note that this did not include
time spent receiving instruction, giving performances, or
playing for enjoyment. The students also devoted about 3.5
hours a day to rest and recreation and obtained more than
average amounts of sleep. Top-level performers in intellectual
pursuits display similar characteristics. Novelists typically
write for about 3 to 4 hours each day, usually in the morning.
Eminent scientists likewise write for a few hours each
morning—the writing of articles arguably being the most
important activity determining their success—and then devote
the rest of the day to other duties. Skinner is especially
instructive in this regard. In his later life, he would rise at
midnight and write for 1 hour, then rise again at 5:00 A.M. and
write for another 2 hours. The remainder of the morning was
devoted to correspondence and other professional tasks, while
much of the afternoon was devoted to recreational activities
such as tinkering in his workshop and listening to music. He
deliberately resisted any urge to engage in serious writing at
other times of the day, feeling that this often resulted in poor-
quality writing the next morning. The limited amount of
writing he did each day was more than compensated for by the
consistency with which he wrote, resulting in a steady stream
of infl uential articles and books throughout his career (Bjork,
1993). Skinner (1987) recommended that students adopt a
similar approach to improve the quality of their writing.
Congruent with this, research has shown that effective college
students are more likely to describe themselves as utilizing a
balanced approach to studying, involving regular study
sessions with frequent breaks, than a driven approach,
involving minimal breaks and studying to the point of
exhaustion (Bouvier & Powell, 2008). Of course, Ericsson et
al. (1993) do not completely discount the role of heredity in
expert performance. Heredity might well affect the extent to
which one becomes interested in an endeavor, as well as one’s
ability to endure the years of hard work needed to become an
elite performer. Nevertheless, the obvious importance of
deliberate practice suggests that we should not be too quick to
discount our ability to acquire a certain skill. Although many
of us might not have the desire, time, or resources to become
elite athletes, excellent musicians, or famous scientists, this
does not rule out the possibility of becoming better tennis
players, learning how to play the guitar, or signifi cantly
improving our math skills. After all, the best evidence
available suggests that it is largely a matter of practice. (See
Ericsson, Charness, Feltovich, & Hoffman, 2006, for a recent
comprehensive overview of scientifi c research on expert
performance; see also Starkes & Ericsson, 2003, for an
overview of research on expert performance in sports.)

What did concern him was whether the concept of hunger, as

defi ned in some measurable way (such as number of hours of
food deprivation), was scientifi - cally useful and led to
testable hypotheses. Hull’s theory was also a pure S–R theory
because it assumed that learning consists of the establishment
of connections between specifi c stimuli and specifi c
responses. Thus, like Watson, he viewed behavior in a very
mechanistic fashion. Lest this seem dehumanizing, recognize
that it is not far removed from some modern-day cognitive
approaches, which view humans as analogous to computers
that process bits of information from the environment (input)
to produce responses (output). This is actually quite similar to
Hull’s model of behavior: Specifi c stimuli (input) yield
specifi c responses (output), with certain internal events
mediating the process. In fact, some versions of modern-day
cognitive psychology can even be considered an outgrowth of
Hull’s neobehaviorism.5 Hull was the most infl uential
experimental psychologist of the 1940s and 1950s.
Unfortunately, it turned out that major aspects of his theory
(which are beyond the scope of this text) were very diffi cult
to test. As well, the theory was highly mathematical and grew
increasingly complex as equations were expanded and modifi
ed. Many of these modifi cations were forced on Hull by his
critics, the most famous of whom was Edward C. Tolman.
(For a major overview of Hull’s theory, as well as to gain a
sense of its complexity, see Hull, 1943.) FIGURE 1.2 In Hull’s
neobehaviorism, theorists make use of intervening variables,
in the form of hypothesized physiological processes, to help
explain the relationship between the environment and
behavior. Environmental events Internal events (intervening
variables) Internal physiological processes, such as hunger and
fatigue Observable behavior 5 Interestingly, people seem less
critical of the cognitive information-processing approach to
psychology, which draws an analogy between humans and
computers, than they are of the traditional behavioral
approach, which draws an analogy between humans and
animals such as rats. Perhaps this is because we are impressed
by the ability of computers to perform certain human-like
tasks (e.g., play chess), and we are insulted by the notion that
humans and rats have anything in common. Yet, outside their
specialized abilities, computers are quite inferior to rats.
Imagine, for example, that a man, a rat, and a computer are
washed up on a deserted island. To the extent that the man
emulates the rat (if he is capable of it), he will likely survive;
to the extent that he emulates the computer, he will sit on the
beach and rot. Rats have a marvelous ability to learn and
adapt; present-day computers do not. Fortunately for us,
humans are far more rat-like than computer-like.

Hull’s S-R theory of learning is often categorized as a

“molecular” theory because it viewed behavior as consisting
of a long chain of specifi c responses connected to specifi c
stimuli. Edward Tolman (1886–1959) disagreed with this
approach and believed that it would be more useful to analyze
behavior on a “molar” (i.e., broader) level. For example, he
felt that we can understand a rat’s behavior in a maze more
accurately as a goal-directed attempt to obtain food than as a
long chain of discrete stimulus-response connections that, in
machine-like fashion, lead to food (e.g., Tolman, 1932). This
molar approach to learning is similar to the gestalt approach to
perception (Kohler, 1947), from which Tolman drew much of
his inspiration. To the gestalt psychologists, perception is not
simply the summation of different bits of conscious experience
but is instead a “holistic” process resulting in an organized,
coherent, perceptual experience. We perceive a house as more
than just a combination of bricks and boards; it is bricks and
boards plus something more. As the famous gestalt saying
goes, “the whole is more than the sum of the parts.” Similarly,
for Tolman, behavior was more than just a chain of discrete
responses attached to discrete stimuli. It was instead an overall
pattern of behavior directed toward particular outcomes, and it
can be properly analyzed only on that level. Although Tolman
disagreed with much of Hull’s theorizing, he did agree that
intervening variables may be useful in a theory of learning (in
fact, it was Tolman who fi rst suggested this). However, while
Hull’s intervening variables were physiological-type processes
like hunger and fatigue, Tolman’s were considerably more
mentalistic. The Tolmanian rat, as well as the Tolmanian
person, was not simply motivated by drives and habits but also
had “expectations” and “hypotheses.” Thus, Tolman’s
cognitive behaviorism (sometimes called “purposive
behaviorism”) utilizes intervening variables, usually in the
form of hypothesized cognitive processes, to help explain
behavior (see Figure 1.3).
Tolman’s (1948) most famous intervening variable is the
cognitive map, which is a mental representation of one’s
spatial surroundings. Evidence for this concept was derived
from a study on “latent learning” by Tolman and Honzik
(1930). This experiment was conducted in an attempt to
disprove Hull’s notion that behavior must be rewarded for
learning to take place; that is, in the absence of some type of
reward, nothing can be learned. To test this notion, Tolman
and Honzik trained three groups of rats on a complex maze
task (see Figure 1.4). The rats in a continuous-reward group
always found food when they reached the goal box, but the
rats in the two other groups found no food when they reached
the goal box (they were simply removedfrom the maze and
then fed several hours later). Training proceeded at the rate of
one trial per day for 10 consecutive days. As expected, the
rewarded group learned to run quickly to the goal box,
whereas the two nonrewarded groups took much longer to do
so. Following this initial phase of training, on day 11 of the
study the rats in one of the nonrewarded groups also began
receiving food when they reached the goal box. According to
Hull, the rats in that group should only then have started to
learn their way through the maze, which would have been
demonstrated by a gradual improvement in their performance.
What Tolman and Honzik found instead was a dramatic
improvement in the rats’ performance on the very next trial. In
fact, on day 12 of the study, the newly rewarded group slightly
outperformed the group that had been receiving a reward from
the outset (see Figure 1.5).
Tolman interpreted these results as indicating that the initially
nonrewarded rats had in fact learned the maze during the fi rst
10 trials of the experiment, and that they had learned it at least
as well as the group that had been receiving food. He would
later interpret these fi ndings as indicating the development of
a “cognitive map” (Tolman, 1948), which only became
apparent when the rats fi nally began to receive food. Thus,
this experiment is regarded as a classic demonstration of latent
learning, in which learning occurs despite the absence of any
observable demonstration of learning and only becomes
apparent under a different set of conditions. The experiment
also demonstrates the distinction between learning and
performance, because learning was apparently taking place
even when the subjects showed no evidence of such learning
in their performance at that time. (See, however, Jensen, 2006,
for a more detailed exposition and critique of these fi ndings
and how they have been interpreted.) Although Tolman
believed that it was theoretically useful to incorporate
cognitive variables, he remained, in many other ways, a
standard behaviorist. For example, like Hull and Watson, he
believed that introspective reports of thoughts and feelings are
so unreliable as to be of little scientifi c value. He maintained
that his own theoretical inferences about cognitive processes
were based entirely on direct observations of behavior and
were thus objectively based. Tolman once even apologized for
the “shameful necessity” of having to discuss conscious
experience in a text he was writing (Tolman, 1932)—a refl
ection perhaps of how frustrated psychologists had been by the
old introspectionist approach. Like other behaviorists, Tolman
also believed strongly in the usefulness of animal research for
discovering basic processes of learning, and almost all of his
research was conducted with rats. Much of Tolman’s research
was directly aimed at refuting Hull’s theory of learning. Hull
was able, in increasingly complex ways, to modify his theory
suffi ciently to account for many of Tolman’s fi ndings.6 As a
result, during Tolman’s life, his cognitive approach never
achieved the same popularity as Hull’s neobehavioral
approach. With the advent of the cognitive revolution in
psychology, however, many of Tolman’s research methods and
concepts have been adopted by modern researchers. Cognitive
behaviorism is now a fl ourishing fi eld of study, and the study
of cognitive processes in nonhuman animals is specifi cally
known as “animal cognition” or “comparative cognition.”