The aim of this thesis is to understand some of the processes involved in the selection of appropriate tasks, and in the selection of correct responses to those tasks. The main focus of the following experiments will be on how interference and conflict occurs during these processes, either at the perceptual or action selection stages when switching task. It should be possible to demonstrate with increases to the number of attributes, the experience of tasks associated with non-target attributes, and in varying the number of trials before switching tasks, that task selection consists of several stages or processes, and that these processes do not necessarily interact. This will show that the switch cost is not an unitary cost associated with changing a single response-stimulus setting, task parameter, or an overall task set, but a composite of various costs associated with different task selection processes. The experiments in this thesis will use the same-different paradigm, which has shown in previous task
switching studies that any advanced preparation for attention to a relevant stimulus dimension only occurs with the presentation of the stimulus (Meiran & Marciano, 2002; Santee & Egeth, 1980). This will result in an increase in the amount of conflict that occurs between tasks, and between the responses associated with those tasks.
The examination of task set conflict, and the involvement of inhibitory processes
in the selection of correct task sets was examined using the task switching paradigm by Mayr and Keele (2000), where it was found that there was a greater cost on returning to a recent task (CBA) compared with switching to a new task (ABA), this was called backward inhibition. This showed that in task switching in particular, that performance on recent trials can have a significant affect on later trials, and that the costs associated with switching task are not necessarily caused by the same consistent process of reconfiguring the associations and settings associated with the new task. This also led to an initial examination into how experience of a task associated with a particular attribute, which would be present in the stimuli but irrelevant during later trials, would affect new tasks. Particulary during task switch trials, where the recent implementation of new task rules suggest that the task set is more susceptible to interference and
disruption. The findings by Lavie and colleagues (e.g., Lavie, 2005, 1995; Lavie &
Fockert, 2003; Lavie, Hirst, Fockert, & Viding, 2004; Lavie & Tsai, 1994), have also shown that during situations with high attentional load, there is less capacity to process irrelevant distractors. This suggests that as stimulus attributes are increased, and with it the difficulty in focusing on the target attribute, then exogenous stimulus-related interference costs should also be reduced. This may have a profound affect on the congruency effect, which is associated with either the competition or correspondence amongst different tasks to a specific stimulus or attribute (e.g., the responses to the colour and shape of a stimulus). Interestingly, the switch cost should not be affected by
this increase, as this cost is associated either with the item-specific retrieval of
responses to previous trials interfering with the target task (e.g., Allport, Styles, & Hsieh, 1994), or the time needed to reconfigure the new task set (e.g., Meiran, 2000b; Monsell, 1996).
In the first series of experiments (1-3), there was no affect of experience of the
interference task on either the competing task or the irrelevant tasks congruency effect when stimuli only consisted of three attributes. However, when the number of attributes was increased to four, the results showed an increase in the congruency effect for the competing task, with no increase for the interference task following interference task trials. This suggests, at least for the three attribute stimuli, when using the same-different paradigm, where response codes are consistent, that the remaining irrelevant attribute is suppressed, resulting in increased attention to the competing task following processing of the target task. However, with stimuli with four attributes, it became more difficult to attend to the target attribute, until experience of the interference task, after which, the results showed an increased congruency effect
for the competing task, suggesting that the attribute associated with the interference task was now suppressed, resulting in greater attention being paid to the remaining attributes. This shows that not only is the congruency effect mediated by attentional load, but also that the congruency effect can be subject to cognitive control processes, which was previously thought to be an automatic exogenous process (MonseJI, 2003) outside such control. Furthermore, in Experiment 3 it was shown that the number of attributes stimuli had during the previous block of trials would determine how the third attribute of trials with three attribute stimuli would be processed, again showing control parameters affecting the congruency effect could be mediated by experience. However, a possible confound in these experiments is that as the stimulus attributes were increased then so were the stimulus frames. Therefore, any affect, or reduced affect, could be attributed to this increase in stimulus set size. This then led to
the examination of stimulus set size on the switch cost and the congruency effect in Experiments 4 and 5. In a previous paper by Malley and Strayer ( 1995), it was found that the amount of priming from a stimulus was dependent on the size of the stimulus set, and the likelihood of that stimulus being presented in recent trials. Also, in a series by papers by Waszak and colleagues (Waszak, Hommel, & Allport, 2005, 2004, 2003), it was found that switch costs were increased if the stimulus had been previously associated with a competing task, even after an interval of more than I 00 trials, although in this instance stimuli were semantically identifiable drawings (see Figure 14). In Experiment 4, a within-subjects design was used. However, as the increases in stimulus-set size were incremental, due to the inclusion of stimuli from the smaller set sizes in the larger stimulus-sets, there is a serial order effect associated with this design. Therefore, in Experiment 5 a between-subjects design is used, which
controlled for this confound.
The results showed that there was increased switch cost associated with the
smaller eight stimulus set size, and that this was present only during the first block in the between-subjects design. The congruency effect was unaffected by the set size manipulation. This shows, even when using stimuli which were semantically similar, that the likelihood of the previous presentation a specific stimulus and its association with a competing task increases the switch cost, the absence of any effect in the larger set sizes suggest that this cost is affected by working memory capacity. The reduction of this affect in later blocks in the between-subjects design, suggests that once a stimulus has been re-presented to all the tasks over several trials then the priming or activation of the competing task is somewhat reduced relative to the target task compared to the first block trials. Furthermore, this shows a dissociation with the previous results in Experiments 1 to 3, which showed reduced congruency effects with increases to attentional load but did not affect the switch cost.
The next series of experiments (6-12), were prompted by the unusual findings
from task switch studies which have shown an asymmetrical switch cost when
switching between tasks of unequal strength (Allport et al., 1994; Allport & Wylie, 2000; Meuter & Allport, 1999), which compare switch costs for the "stronger" word reading task with the "weaker" colour naming task. The result show an increased switch cost when switching to the "stronger" task, which is attributed to the increased bias in cognitive control needed to switch to the "weaker" task and the need to overcome these settings on the return to the "stronger" task. However, as Yeung and Mansell (2003b) report, following several manipulations, that the asymmetrical switch cost can be reversed under certain conditions. For instance, when black words are delayed by 160 ms and presented on coloured backgrounds, the asymmetric switch cost is reversed, with larger switch costs for the colour naming task.
An interesting point here is that some of the paradigms described use the location of a task stimulus to cue the task, and as such a possible confound with this paradigm could be that location does not prime a task, in comparison to the explicit task-cueing paradigm, which uses a word cue (for a review of the differences between the alternating-runs and explicit-cueing paradigms see Altmann, 2007). Therefore, the processes attributed to studies using the explicit-cueing procedure may almost certainly be different to those in the explicit-cueing procedure.¹
The aim in this series of experiments was to show that by varying the number of
trials before switching task, it was possible to vary task strength (with 8 and 80 trial conditions) and the switch cost following a switch from that task. Also, by varying the number of trials returning to that initial task it was possible to examine factors such as task-set decay and cue-priming, which are contentious issues in the task switching literature. The results showed that the switch cost was determined by the level of competition between the previous task and the target task. The greater the competition, the greater the amount of cognitive control, and consequently a reduction in interference, and as a result a smaller switch cost. In Experiment 12, the task cue was only presented during the intervals between task switches once every eight trials, and as a result the recovery from those earlier control processes was less successful. Therefore, on returning to the initial task cognitive processes were still needed even after the longer 80 trial pre-switch condition. This shows, at least when using this
paradigm, that the switch cost is the result of conflict between task sets, and when that conflict exceeds a certain point then cognitive control intercedes, reducing the switch cost.
The overall impression from the results are that there are several components to
the task switching process, which are differentially affected by changes to the
parameters of the task. Paradoxically, although the task switching paradigm is thought of as a suitable measure of executive functioning, a recent meta-analysis of several neuroimaging studies found not only several regions associated with executive functioning, but also several areas associated with working memory and attention shifting processes (Wager, Jonides, & Reading, 2004). This highlights the diverse number of processes associated with task switching, and this is compounded by the diversity and variation in tasks and paradigms found in the task switching literature.