ASPECTOS GENERALES DE LA FINCA

Poulton (1952) appears to have been the first to systematically investigate whether the development of anticipation in a tracking task that involves a repetition depends on the perception of the movement of the target. He tested the differences in proficiency associated with pursuit versus compensatory tracking. In pursuit tracking, both the perceptual input and the participants' movement output are displayed. In compensatory tracking, only the difference between the input and the participants' output is available (as tracking error). Pursuit tracking resulted in smaller tracking errors than compensatory tracking because it was only in this case that the participants were able to perceive the action of the target directly, and thus were able to leam the regularities of the target sequence. With compensatory tracking, the action of the target was masked by the actions of the control movements. This interfered with learning of regularities. Therefore, learning of the regularities in the movement of the target was faciHtated by direct perception of the movement.

Stadler (1989) was the first to directly test whether learning of the structured sequence is based on the anticipation of the appropriate motor responses or on the anticipation of the location of the next target. He used the visual search task introduced by Lewicki, Czyzewska, and Hoffman (1987; see chapter 2). To test the perceptual learning hypothesis, he transferred participants to a task that involved the same responses but different perceptual displays; stimuli were moved into the inner comers of the appropriate quadrants. An increase in response times was found with perceptual transfer and suggested that learning is dependent on the perceptual characteristics of the task.

Chapter 3 Mechanisms, Models and Locus of Learning Page 97 to a different response apparatus. The new apparatus kept the same square spatial arrangement of the response keys, but increased the distance between them. Also, participants were required to use the index and middle fingers of the right hand during training but were required to use the middle and index fingers of both right and left hands during transfer. Stadler (1989) reported a decrease in response time when participants were transferred to the new response requirements which suggested that learning was not based on response anticipation. However, although the above finding suggests that sequence learning is independent of the actual motor responses, response times in the transfer phase may have been speeded up by a factor that is unrelated to sequence learning. There is considerable evidence to suggest that choice reaction times are faster when two alternative responses are executed by fingers from different hands rather than fingers from the same hand (Komblum, 1965; Reeve & Proctor, 1988; Alain, Taktak & Buckoltz, 1993). This may account for response time speed-up found by Stadler (1989).

Cohen et al. (1990) also tested whether the locus of learning in a dual-task situation is on the motor responses performed. A group of participants practised a structured sequence in a three choice reaction time task using the three fingers of one hand, each finger corresponding to a different stimulus location. Movement requirements were changed for half of the participants who were then asked to perform the task using one finger for all stimulus locations. The remaining participants were transferred to a random sequence while keeping the same response requirements. No deterioration in performance was found for the group that was transferred to different movement requirements, while performance of the group that was switched to the random sequence declined. This indicated that learning was not tied to the effector system used for the response during practice.

A slightly different approach was employed by Howard, Mutter and Howard (1992), who tested whether performance of the motor responses is necessary for the development of response time speed-up. They compared a "response" group that performed a serial response time task and an "observation" group that responded only to the first repetition of the presented sequence in each one of the eight blocks of practice and just observed the following repetitions. In a second experiment the observation group just observed the sequence during the first three blocks and responded to the initial trials of blocks four to eight. Both response and observation groups exhibited a similar degree of learning during the response time task, suggesting that responding to all trials is not a necessary condition for learning.

In a generation-with-correction task, the observation group achieved greater accuracy during the first cycles of sequence generation. The response group, however, showed a greater degree of improvement during the generation task and reached the same overall level of performance as the observation group by the last sequence presentation. It appeared that the requirements of the response time task for the response group hampered performance in the initial trials of the generation task. As Howard et al. (1992) state "Continued learning for the response group during the generation phase may reflect the need for an additional recoding of pattern knowledge into a form that could better support task demands" (p.l035). Howard et al. (1992) concluded that overt responses are not necessary for serial pattern learning to occur and that therefore learning may be based on the perceptual aspect of the repetition.

However, it may also be argued that the enhanced abUity of the observation group to predict the sequence in the generation task indicates that this group may have been engaged in a more analytical mode of processing. This was in a way dictated by the requirements of the practice task. Participants were

Chapter 3 Mechanisms, Models and Locus of Learning Page 99 instructed to "carefully observe" a sequence of stimuli and they had full attentional resources available. They still had to respond to this pattern for a few trials, which probably motivated them to attend to specific attributes of this sequence that would allow them to speed-up their responses.

On the other hand the response group did not have any reason to overtly analyse the spatial properties of the sequence of stimuli because their main task was to respond as fast as possible to each one of these stimuli. A representation of the sequence could only develop within the boundaries of the attentional resources left by the task requirements of overt responding. It may be the case that performance on the response time task is affected by both the development of fast selection and execution of the response movements and the development of sequential knowledge that allows preparation for following events. Consequently, Howard et al.'s (1992) results could be fully explained by a differential influence of these two processes on performance levels of the two groups rather than the account proposed by them based on a purely perceptual basis of learning in the response time task.

One study, however, that reported persuasive evidence to suggest that a repetition of responses is not required for sequence learning was carried out by Ziepier (1994). In one experiment, participants had to respond to five different letters by pressing five corresponding keys in a matrix-scanning task. Each target letter was presented together with distractor letters, in a five by five matrix. The location of the target on the next trial was determined by the identity of the target letter in the previous trial according to some simple rules. Therefore, a target letter could follow any other target letter but it appeared in a predictable spatial location. Practice with this task did not involve a repetition of a series of motor responses as motor responses were determined by the letter identity which was random. Ziepier (1994) reported that a group that practised

with the sequential dependencies between letter identities and the location of the next target showed a performance speed-up compared to a control group that practised with a random sequence of target locations. Therefore, Ziepler (1994) has shown that sequential dependency learning can occur without a repetition of motor responses.