[ CogSci Summaries home | UP | email ]

Charness, N. (1981). Aging and skilled problem solving. Journal of Experimental Psychology: General, 110, 21-38.

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

Charness (1981a, 1981b)


1981b could be;

Charness, N. (1981). Search in chess: Age and skill differences. Journal of Experimental Psychology: Human Performance and Perception, 7, 467-476.

but I'm not sure. --Jim R. Davies, editor

The purposes of this article are to 1) describe the molar-equivalence molecular-decomposition research strategy and 2) examine age-related influences on problem solving. Charness distinguishes between age-deficits attributable to hardware loss and software loss. The former correspond to processing efficiency and structural changes, and the latter correspond to strategy deficiencies. Presumably, hardware loss in the absence of compensatory software leads to performance deficits.


Charness argues that chess is ideal for investigating age and skill for a number of reasons. First, it provides an excellent rating skill that is a reflection of molar-level skill, but not necessarily molecular-level skill. That is, "The rating scale merely guarantees that strengths and weaknesses will balance at the level of molar performance" (p. 23). Second, there have been few changes in the rules of chess. The possibility of historical/cohort confounds is thus minimized.


Charness points out that a cross-sectional design is most appropriate for chess because a change in rating from one age to the next may reflect a change in skill, but it also may reflect the fact that current players are more skilled. But, "Given two players of equal strength (rating), yet unequal ages, we can be fairly confident the players possess comparable chess skill" (p. 24). Based on this consideration, the sample was selected such that older and younger players were matched on skill-level, that is, so that the age-skill correlation was near zero (.09). The tasks included choose-a-move, rapid evaluation of end-game positions, and various memory tasks.


Experiment 1


As expected, Charness found strong relations between skill, but not age, and quality of move in the choose-a-move task. Older players took less time to choose a move than did younger players. An interesting finding was that correct recall in the unexpected recall task was negatively related with age, but positively related to skill. (Recall that the younger and older groups were matched on chess rating and that there was no effect of age on move quality in the choose-a-move task.) Charness explains that the age deficit in the task could be attributable to encoding or retrieval failures, or to both. A retrieval variables was derived by computing the time taken to begin recall, but there were no effects of age or skill. A second measure was mean number of chunks recalled per position. There was a highly significant age x skill interaction: the higher the skill, the fewer the chunks, but less so for older adults than for younger adults. That is, "older players show only a modest increase in chunking efficiency with increasing skill, whereas young players show much greater increases with increasing skill" (p. 30).


The general picture is that older players recalled more chunks with more pieces. As Charness explains, "This peculiar result stands in contradiction to a theory of chess skill that implicates efficient encoding as a determinant of skill. Two players possessing the same rating may vary greatly in chunking pattern if one is young and the other comparatively old . . . " (p. 30).


Experiment 2


An end-game evaluation task (win, lose, or draw?) was administered. Skill level was a significant predictor of accuracy, but age was not. In a finer-grain analysis, the data were broken down into evaluations that occurred within 10 seconds of the stimulus, and those that occurred after 10 seconds. The number of quick evaluations was not related to skill or age, but the proportion correct for quick evaluations was related to skill: "The more skilled you are, the higher the proportion of correct quick evaluations" (p. 34). The proportion correct for slow evaluations was unrelated to skill and age.


Comments and Questions


Charness states that "When the performance of chess players who vary in age and skill is examined, critical aspects of performance do not relate to the playerís age, but only to skill level" (p. 34). Critical aspects of performance were measured with the choose-a-move and end-game evaluation tasks. What is confusing is that these tasks might be considered indices of skill. Therefore, in a sample of younger and older adults matched on chess rating, age-relations with such measures would not be expected. What is the distinction between indices of skill and indices of component processes of that skill?


Charness also found that a memory measure for chess positions was positively related to chess rating, but negatively related to age. Charness attributed the finding to retrieval difficulties for the older adults (what about frequency and recency effects?), and concluded that "Age and skill tend to trade-off in memory performance" (p. 34). How so? One possibility is efficiency of retrieval. However, older adults chose moves quicker, with no loss in accuracy, in the choose-a-move task. Charness suggests that the older adults terminate search sooner "in order to rely more on short-term than on long-term memory" (p. 35). (Do older adults show age-related decline in short-term memory?)


He states, "one possibility is that the compensatory mechanism for older players involves having more refined move generators or possibly more refined evaluations stored in association with the hypothesized large (but equivalent for a given skill level) repertoire of stored patterns" (p. 35). (Charness rules out more refined evaluations because there was no relation of age to static evaluation accuracy.) Overall, Charness concludes that "The compensating mechanism is more efficient search of the problem space, as evidenced by older players taking less time to select an equally good move" (p. 37).


Charnessís conclusions can be challenged on a number of grounds. For example, the older adultsí faster latencies in the choose a move task may not reflect more efficient search per se. Rather, older adults may simply respond as soon as they decide on the best move, whereas younger adults may evaluate other moves even after they have decided on what might be the best move. After analyzing verbal reports from the current study, Charness (1981b) found that older adults consider fewer base moves (first move) in the choose-a-move task, but that search depth was equivalent in younger and older adults. Originally, Charness proposed that older adults have better move generators and are therefore more efficient in search. But, in the verbal reports, older adults mentioned best moves less often than did younger adults. Thus, "A more likely explanation for the age-related search differences is that older players may dispense with the verification phase when they find an adequate candidate move" (p. 474). In other words, if they consider a base move and find it to be reasonable, they make the decision. This reflects a sort of strategy or attitudinal difference: "Perhaps older players have a better sense of what a given position holds (global evaluation) and they may set a more realistic level of aspiration, enabling them to terminate search earlier" (p. 474). (Metaknowledge?)

Back to the Cognitive Science Summaries homepage
Cognitive Science Summaries Webmaster:
JimDavies (jim@jimdavies.org)
Last modified: Tue Mar 7 15:43:58 EST 2000