[ CogSci Summaries home | UP | email ]

Larkin, J., McDermott, J., Simon, D. P., & Simon, H. A. (1980). Expert and novice performance in solving physics problems. Science, 208, 1335-1342.

Author of the summary: David Zach Hambrick, 1998, gt8781a@prism.gatech.edu


The magic of words is such that, when we are unable to explain a phenomenon, we sometimes find a name for it—as Moliere’s physician "explained" the effects of opium by its dormitive property. So we "explain" superior problem solving skill by calling it ‘talent,’ ‘intuition,’ ‘judgment,’ and ‘imagination’ (p. 1335).


Expert performance is mysterious in the sense that describing the differences between experts and novices is relatively difficult. What do experts and novices do differently? One of the difficulties is that the processes underlying expert performance are extremely rapid—they unfold in milliseconds. Nevertheless, as Larkin et al. explain, it is possible to construct computer simulations that can account reasonably well for novice-expert differences. The similarity of simulated and behavioral data serves as a check on a theories validity.


Experts have more knowledge than novices. The explanation for superior pattern perception presented in this article is dated, but a couple of interesting points are made. First, referring to chess, "This large set of perceptual patterns serves as an index, or access route, not only to the expert’s factual knowledge but also to his or her information about actions and strategies" (p. 1336). LTM is a set of production rules in which the patterns are the IF portion of the statement. The THEN portion of the statement specifies some action to perform. That is, "Whenever the stimulus to which a person is attending satisfies the conditions of one of his productions (that is, contains a recognizable pattern), the action is immediately evoked (and possibly executed)" (p. 1337). Hence, as Anderson (1995) notes, "masters effectively ‘see’ possibilities for moves; they do not have to think them out" (p. 294).


Research on expertise has revealed a number of differences in the way experts and novices go about problem solving. Not surprisingly, experts solve problems faster than novices. What accounts for this speed difference? In physics, experts use a forward reasoning strategy, whereas novices use a backward strategy. That is, novices start from the unknown and work to the givens, whereas experts work from the givens to the unknowns. Consider the following example from Anderson (1995): "A block is sliding down an inclined plane of length l where x is the angle between the plane and the horizontal. The coefficient of friction is y. The subject’s task is to find the velocity of the block when it reaches the bottom of the plane" (p. 285).


The experts takes the givens and generates the solution. That is, he takes the givens and computes the quantities that are necessary for generating the solution (e.g., acceleration). The production rule might be as follows: "If you know the values of all the independent variables in any equation (condition), try to solve for the dependent variable (action)" (p. 1339). Basically, the idea is that the expert has a production rule for generating a quantity when certain other quantities are given. An example is:


IF the quantities v, v, and t are known,

THEN the acceleration a can be calculated.


The equation is instantiated in the production rule, and the result is obtained when the production rule executes. Consistent with this idea, when solving physics problems, experts mention the quantity that they are substituting into an equation (e.g., a), but not the equation used to generate that quantity. The novice, by contrast, starts with the unknown (v) and comes up with an equation that might be used to solve for v. Then, he or she computes the quantities necessary to generate v. The novice has declarative knowledge of the equation, but it is not yet instantiated in a production rule.


Summary, Comments, and Questions


Knowledge is an "essential prerequisite to expert skill" (p. 1342). This is obvious, but, as Larkin et al. point out, "The expert is not merely an unindexed compendium of facts . . . " (p. 1342). Familiar patterns guide the expert to relevant knowledge about appropriate actions and strategies. I have expanded this notion in terms of Anderson’s production rule system. Patterns are stored as declarative facts. When a familiar pattern is recognized—that is, when a pattern activates its representation in declarative memory—it becomes the activated subset of long-term memory called working memory. If the contents of working memory match the IF portion of a production rule, stored in procedural memory, then that production rule is applied. The product of the production rule, specified in the THEN portion of the statement, then enters working memory—as a thought. This thought can be translated into a physical action—for example, a chess move—but can also serve as the trigger for an additional thought.


As Larkin et al. point out, understanding expert-novice differences does not imply that expertise can be acquired effortlessly because "The extent of the knowledge an expert must be able to call upon is demonstrably large, and everything we know today about human learning processes suggests that, even at their most efficient, those processes must be long exercised" (p. 1342).

Back to the Cognitive Science Summaries homepage
Cognitive Science Summaries Webmaster:
JimDavies (jim@jimdavies.org)
Last modified: Tue Mar 7 14:49:11 EST 2000