[ CogSci Summaries home | UP | email ]

Ackerman, P. L. (1988). Determinants of individual differences during skill acquisition: Cognitive abilities and information processing. Journal of Experimental Psychology: General, 117, 288-318.

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

Ackerman (1988)


Ackerman proposes that different abilities underlie performance at successive stages of skill acquisition. General ability measures (e.g., abstract reasoning) underlie performance in Phase 1. As production systems for the consistent aspects of performance are formed, the influence of these factors declines, and perceptual speed abilities emerge as significant predictors of performance in Phase 2. Finally, in Phase 3, performance is determined primarily by non-cognitive psychomotor abilities.


Complexity and Consistency


Two factors mediate dynamic changes in ability-performance relations: complexity and consistency. Complexity refers to the information processing demands of the task, and can be manipulated by, for example, varying the number of response choices or the stimulus display duration. Increased complexity necessitates greater reliance on general intellectual abilities. Consistency refers to the uniformity of information processing demands across practice, and is typically defined in terms of constancy of stimulus-response relations. On the one hand, in an inconsistent task, complete separation of targets and distractors is maintained in a. That is, stimuli that are targets on one trial are targets on all trials, while stimuli that are distractors on one trial are distractors on all trials. On the other hand, in an inconsistent task, stimulus-response mappings vary so that a stimulus may be either a target or a distractor for a given trial. Progression through three stages of skill acquisition described above is possible only for consistent tasks.


Relations of General and Perceptual Speed Abilities to Performance


The goal of the first six experiments reported was to test predictions about correlations of performance with perceptual speed and general intellectual ability measures. For consistent and moderately complex tasks, performance was expected to be initially more strongly correlated with general ability measures than with perceptual speed measures. With continued practice, the former relation was expected to decline, while the latter relation was expected to increase. That is, with continued practice, it was predicted that perceptual speed would make an increasingly larger relative contribution to performance than general abilities.


Experiment 1


In experiment 1, subjects performed a simple, consistent choice-reaction time task in which the goal was to press a key on a numeric keypad corresponding to appearing on the computer screen. For example, if subjects saw the number 6, their task was simply to press the key for the number 6. In a transfer segment, the complexity of the task was increased by pairing numbers with arbitrary two-letter abbreviations. Additional encoding and translations requirements were thereby imposed.


Ackerman reports that, in the initial training segment, perceptual speed correlations were initially high, and that general ability correlations remained stable. In the subsequent transfer segment, general ability correlations were initially higher than perceptual speed correlations, as would be expected with the increase in task complexity. With continued practice, general ability correlations declined while perceptual speed regained influence. Unfortunately, none of the expected trends was significant. For example, although Ackerman states that "As predicted, during the training segment, Perceptual Speed ability was highly correlated with performance and showed attenuation with practice," the degree of attenuation was non-significant [i.e., R2cub = .88, F(3, 2) = 5.11, p > .05]. Therefore, Ackermanís conclusions regarding dynamic ability-performance relations are weak, if not misleading.


The aim of Experiments 2-5 was to replicate ability-performance relations with more cognitively demanding tasks. In Experiment 2, subjects performed a semantic search task in which they memorized taxonomic category labels (e.g., animal), and then indicated the location of category exemplars (displayed along with distractor exemplars) belonging to target taxonomies. Briefly, general ability correlations declined linearly with practice, while "the Perceptual Speed-performance correlations started low but increased as skills were acquired" (p. 300). Again, however, the latter trend was non-significant.


For Experiment 3, the complexity of the preceding task was manipulated by decreasing the memory load from three to two. As expected, in both memory load conditions, the general ability-performance relations declined significantly with practice. However, perceptual speed-performance correlations did not increase as general ability-performance correlations declined. In fact, there was a significant negative trend. Thus, while the results of Experiment 2 and 3 seem to support the claim that general ability-performance correlations decline with practice on consistent tasks, the contribution of perceptual speed abilities is less clear.


A varied mapping version of the task used in Experiments 2 and 3 was used in Experiment 4, and the expectation was that general ability-performance correlations for this task would show shallower decline than the same correlations for the CM versions of the task. In addition, perceptual speed-performance correlations were expected to peak later for the VM version. General ability-performance correlations were initially higher in the varied mapping condition than in the consistent mapping condition with the same memory load, and although Ackerman makes the claim that "VM performance . . . showed shallower attenuation during task practice . . . " and that "the VM condition had much greater and longer lasting associations with General ability" the slopes of the functions relating practice and general-ability performance correlations are strikingly similar. Perceptual speed-performance correlations did appear to peak later in the VM conditions, but, overall, Ackermanís conclusions are again not supported by the data.


In Experiment 5, a mixed consistent/varied task was created. The general ability-performance correlations declines mirrored those found in VM performance, and the perceptual speed-performance correlations peaked later, though this trend was non-significant. Experiment 5 provides additional support for the finding that general ability-performance correlations decline with practice, but the suggestion that perceptual speed emerges as a significant predictor later in practice is clearly not supported by the data.


The finding that general ability-performance correlations decline with practice was repeated in Experiment 6 with a spatial (as opposed to semantic) task. And furthermore, the intercept for the general ability-performance function was higher for the spatial measure than the semantic task. To the extent that spatial tasks are more complex (i.e., novel) than semantic tasks, this might suggest that complex tasks impose greater demands on general abilities than simpler tasks. Unfortunately, a test of intercept differences is not reported, so no firm conclusions can be made about this possibility. The cubic trend for perceptual-speed and performance was significant, but it appears that the function almost mirrored the general ability-performance function. This is inconsistent with the view that perceptual speed-performance relations peak only after general ability-performance relations have declined.


Psychomotor tasks were introduced in Experiment 7. To review, Ackerman predicts that psychomotor abilities should emerge as the most important predictor of performance in the third phase of skill acquisition. Consistent with earlier results, general ability-performance correlations declined with practice. The perceptual speed-performance trend was again non-significant. However, psychomotor abilities showed increasing correlations.


Generalizability of Results to a Complex Task


The goal of the final experiment was to assess the generalizability of results obtained with simple tasks to a complex task, namely, the Kanfer-Ackerman Air Traffic Controller task. Notwithstanding Ackermanís contention that "the pattern of results is consistent with predictions from the theory," the results do not provide convincing support for the theoretical predictions regarding dynamic changes in ability-performance correlations. For example, although the relationship between performance and general ability is graphically depicted as negative and linear, this trend was non-significant. The quadratic trend relating practice and perceptual speed-performance correlations was also non-significant. The relation between practice and psychomotor ability-performance correlations was in the expected direction.




Ackermanís claim that "By and large, the data are congruent with the theoretical predictions" does not seem to be supported by the data. For example, Nevertheless, the finding that general ability-performance correlations decline linearly with practice was replicated in Experiments 2-7. It therefore seems reasonable to suggest that general abilities become progressively less important with practice. There is also some indication that general ability-performance correlations are higher for more complex tasks, although intercept differences were not tested. By contrast, the data do not consistently support the claim that perceptual speed-performance correlations peak later in practice and after general ability-performance correlations have declined. Significant relations between practice and perceptual speed-performance correlations were found in only three of eight experiments, and it is not clear that the significant trends support Ackermanís theoretical predictions.


Back to the Cognitive Science Summaries homepage
Cognitive Science Summaries Webmaster:
JimDavies (jim@jimdavies.org)
Last modified: Tue May 25 18:35:59 EDT 1999