Professional Documents
Culture Documents
As genetics needs its model orga a running computer program, mates is to examine first those
nisms, its Drosophila and Neuros- MATER, and subjected to addi moves that permit the opponent
pora, so psychology needs standard tional empirical testing (3). the fewest replies. A comparison of
task environments around which the MATER program with think-
knowledge and understanding can The MATER theory is an applica ing-aloud protocols from human
cumulate. Chess has proved to be tion to the chess environment of a chess players confirms the impor
an excellent model environment for more general theory of problem tance of heuristic search as a basic
this purpose. About a decade ago in solving that employs heuristic underlying process.
the pages of this journal, one of us, search as its core element (4). The
with Alien Newell, described the MATER theory postulates that While the MATER theory was suc
progress that had been made up to problem solving in the chess envi cessful in accounting for much of
that time in using information-pro ronment, as in other well-struc what was known about chess think
cessing models and the techniques tured task environments, involves a ing in mating situations, some im
of computer simulation to explain highly selective heuristic search portant empirical phenomena—
human problem-solving processes through a vast maze of possibilities. some of them known when the
(1). A part of our article was devot Normally, when a chess player is theory was formulated, some of
ed to a theory of the processes that trying to select his next move, he is them discovered subsequently—
expert chess players use in discov faced with an exponential explosion eluded the theory's grasp. In this
ering checkmating combinations of alternatives. For example, sup paper, after describing the phenom
(2), a theory that was subsequently pose he considers only ten moves ena, we should like to tell the story
developed further, embodied in for the current position; each of of a ten-year effort to account for
these moves in turn breeds ten new the recalcitrant facts.
moves, and so on. Searching to a
Herbert A. Simon took his bachelor's and depth of six plies (three moves by An important by-product of this ef
doctor's degrees at the University of Chi- White and three by Black) will al fort has been to bring about a con
cago, the latter in 1943. He has served on ready have generated a search vergence of the theory of problem
the faculties of the University of Califor- space with a million paths. Hence, solving with theories that have
nia, Berkeley, and Illinois Institute of if every legal move is considered (as been developed to explain quite dif
Technology, and, since 1949, on the facul-
ty of Carnegie-Mellon University, where would be the case in an exhaustive ferent phenomena, which psycholo
he is Richard King Mellon Professor of search), an enormous search space gists label "perception," "rote
Computer Science and Psychology. Begin- would be generated. Such a search learning," and "memory." In the
ning with an interest in decision-making
in organizations, Professor Simon has been
is beyond the capacity of the past, both theorizing and experi
led during the past fifteen years into re- human player, as well as present- mentation relating to these differ
tearch on human performance in complex day computers. Humans seldom ent kinds of tasks—problem solv
tasks, using the computer to simulate cog- search more than a hundred paths ing, perceiving, learning by rote,
nitive processes. With Alien Newell, he is in choosing a move or finding a and remembering—have tended to
coauthor of a recent book, Human Prob checkmate, and they seldom con go their separate ways. In the
lem Solving.
William G. Chase is an Associate Professor sider more than two or three possi course of our story we will see how
of Psychology at Carnegie-Mellon Univer- ble moves per position. these theories come together to ex
sity, where he has served since receiving plain chess skill; we will see the im
his Ph.D. from the University of Wisconsin
in 1969. His research has concentrated on
The MATER theory postulates portant constraint that a limited-
the elementary information processes un- that humans don't consider moves capacity short-term memory im
derlying cognition. He has edited a recent at random. Rather, they use infor poses on problem solving in chess
book, Visual Information Processing. mation from a position and apply and how this limit can be bypassed
This research was supported by Public some general rules (heuristics) to by specific perceptual knowledge
Health Service Research Grant MH-
07722, from the National Institute of Men-
select a small subset of the legal acquired through long experience,
tal Health. Address of both authors: De- moves for further consideration. For stored in long-term memory, and
partment of Psychology, Carnegie-Mellon example, one powerful heuristic accessed by perceptual discrimina
University, Pittsburgh, PA 15213. that MATER uses in finding check tion processes.
394 American Scientist, Volume 61
The phenomena the previous experiment, but now esting and relevant to find out how
constructed random positions with the human eye extracts information
In Amsterdam, Adriaan de Groot, them. Under the same conditions, from a complex visual display like a
who was the first psychologist to all players, from master to novice, chess position and to see whether
carry out extensive experiments on recalled only about three or four this extraction process is compat
problem solving using chess as the pieces on the average—performing ible with the assumptions of the
task, also initially formulated his significantly more poorly here than heuristic search theories.
theory in terms of heuristic search the novice did on the real positions.
(5). His subjects ranged from quite (The same result was obtained by A pair of Russian psychologists,
ordinary players to some of the W. Lemmens and R. W. Jongman Tichomirov and Poznyanskaya,
strongest chess grandmasters in the in the Amsterdam laboratory, but placed an expert before a chess po
world, including several former their data have never been pub sition with instructions to find the
world champions. He was puzzled lished, 8.) best move, and they observed his
by one thing: none of the statistics eye movements during the first 5
he computed to characterize his In sum, these experiments show seconds of the task (11). The eye
subjects' search processes—number that chess skill cannot be detected movements were inconsistent with
of moves examined, depth of from the gross characteristics of the the hypothesis that the subject,
search, speed of search—distin search processes of chess players during these 5 seconds, was search
guished the grandmasters from the but can be detected easily using a ing through a tree of possible moves
ordinary players. He could only perceptual task with meaningful and their replies.
separate them by the fact that the chess content. The experiment with
grandmasters usually chose the random boards shows that the mas To describe further what Tichomi
strongest move in the position, ters' superior performance in the rov and Poznyanskaya found, we
while ordinary players often chose meaningful task cannot be ex must say a word about how the eye
weaker moves. Why were the plained in terms of any general su operates. The eye has a central re
grandmasters able to do this? periority in visual imagery. The gion of high resolution, the fovea
Wherein lay their chess skill? perceptual skill is chess-specific. (about 1° in radius), surrounded by
Moreover, a theory of problem solv a periphery of decreasingly lower
The perceptual basis of chess mas ing in chess that does not include resolution. Most information about
tery. One clue to this riddle came perceptual processes cannot be an visual patterns is acquired while
when de Groot repeated and ex adequate theory—cannot explain the fovea is fixated on them; and
tended an experiment that had the superior ability of the strong the eye moves abruptly, in so-called
been performed earlier in the USSR player to choose the right moves. saccadic movements, from one
(6). He displayed a chess position point of fixation to the next. There
to his subjects for a very brief peri Eye movements at the chess board. are at most about four or five sac
od of time (2 to 10 seconds) and The second set of phenomena we cadic movements per second.
then asked them to reconstruct the must consider are also perceptual,
position from memory. These posi but of a more recent discovery. Ex In Tichomirov and Poznyanskaya's
tions were from actual master planations in terms of heuristic record of the first 5 seconds of their
games, but games unknown to his search postulate that problem solv subject's eye movements, there
subjects. The results were dramat ing, and cognition generally, is a were about 20 fixations.
ic. Grandmasters and masters were serial, one-thing-at-a-time process. these centered on squaresMost of
of the
able to reproduce, with almost per (We are oversimplifying matters to board occupied by pieces that any
fect accuracy (about 93% correct), make the issue clear, but the over chess player would consider to be of
positions containing about 25 piec simplification will suffice for the importance to the position. There
es. There was a quite sharp drop-off present.) Many psychologists have were few fixations at the edges or
in performance somewhere near the found this postulate implausible corners of the board or on empty
boundary between players classified and have sought for evidence that squares. Moreover, a large number
as masters, who did nearly as well the human organism engages in ex of the saccades moved from one
as grandmasters, and players clas tensive parallel processing (9). The piece to another, where the former
sified as experts, who did signifi intuitive feeling that much infor piece stood in a "chess" relation—
cantly worse (about 72%). Good mation can be "acquired at a that is, an attack or defense rela
amateurs (Class A players in the glance" argues for a parallel proces tion—to the latter. For example,
American rating scheme) could re sor. Of course, the correctness of the eye would move frequently from
place only about half the pieces in the intuition depends both on the a pawn to a Knight that attacked
the same positions, and novice amount of information that can ac it, or to a Knight that defended it,
players (from our own experiments) tually be acquired and upon what or from a Queen to a pawn it at
could recall only about eight pieces is meant by a "glance." If a glance tacked.
(about 33%). There is a quite nice means a single eye fixation (lasting
gradation on this perceptual task as anywhere from a fifth of a second to It is important to note that the sac
a function of chess skill, and we a half-second or longer), then we cadic movements were not random
have verified this in our own exper know that there are high-speed se —therefore, that some information
iments (7). rial processes (e.g. short-term must have been acquired peripher
memory search, visual scanning) ally about the target square before
We went one step further: we took that operate within this time range the saccade began. From other evi
the same pieces that were used in (10). Thus, it is certainly inter dence, we know that a strong chess