Task-Switching and Working Memory

The ability to switch flexibly between tasks allows people to adapt to changing demands in the environment. However, task-switching takes time and produces

interference, as is evident in a variety of procedures that compare performance when tasks change, with performance when tasks remain constant (Allport, et al., 1994; Gopher et al., 2000; Logan & Bundesen, 2003; Rogers & Monsell, 1995). Several accounts of task-switching suggest that switch-costs are associated with changing set which can be explained by working memory processes (Baddeley et al., 2001; Mayer & Kliegl, 2003; Rubenstein et al., 2001). Alternatively, switching can also be seen as a process within working memory (Barrouillet et al., 2004; Cowan, 2005). Task-switching in the current experiment was determined by attempting to

discriminate switch costs from both recall and performance of tasks. Switch costs were calculated by adding reaction times from the single task condition to the

memory span task condition (referred to as the combined condition) and comparing it to reaction times from the perform task condition (as per Logan, 2004).

Based on Logan’s (2004) findings, it was predicted that reaction times in the perform condition would be significantly slower than the combined condition.

Furthermore, these additional reaction time costs would be attributed to task-

switching. The results of the current study were not consistent with Logan’s findings such that reaction times in the combined condition were significantly slower than the perform condition. This finding was the opposite of what was expected and indicate that there were no switch costs present between tasks.

The current results may have differed from Logan’s (2004) due to the

difference in the conceptualisation of switch costs between the two studies. Logan’s combination of the single and span task was based on list position, for example, single task for position three in a sequence was combined with memory span in

position three. Logan did not account for reaction time differences as a result of variations in task difficulty and in his work the single task conditions were equated in terms of list position. Conversely, the current study combined the reaction times from the memory span task and the single task based on the type of task being performed. Thus, although list position was taken into account in the span condition and perform condition, the single task condition was not added to the memory span condition based on list position. Instead, the single task reaction time was determined by the task completed and consisted of the mean reaction time for each participant

performing the task as determined by the single task condition. This allowed the comparison of the span and perform conditions in a way that accounted for the differences in reaction time seen in the pilot study and single task condition. This alternative conceptualisation suggests that switch costs are not present when task difficulty is taken into account in the task-span procedure. This finding means that the current method of measuring switch costs (Logan, 2004) may require modification to specifically take into account difficulty of a task within a span.

In his theoretical paper, Monsell (2003) suggested that three factors were associated with task switch costs: task-set reconfiguration costs, transient task-set inertia, and associative retrieval. In the current study as with Logan (2004),

participants were required to learn a sequence of tasks in the study phase and then recall or recall and use these tasks in a later test stage. Given that task sequence was presented to the participants prior to the presentation of the target stimulus and was not stimulus driven, it is likely that participants in the current study were aware of the upcoming task prior to the presentation of the target stimulus. Consequently, task switch costs, as defined by Logan (2004) may not have been captured in the current

study using this methodology. This possibility will be discussed in more depth in Experiments 2 and 3.

Some researchers have suggested that working memory has an inherent role in task-switching, specifically, that task reconfiguration involves processes that

operate in working memory. Changing the contents of working memory is sufficient to change performance (i.e. changing goals and changing stimulus-response mapping rules; Mayr & Kliegl, 2000; Rubinstein et al., 2001; Sohn & Anderson, 2001). Others argue that reconfiguration involves processes outside of working memory as well as processes that operate on working memory; merely changing goals is not sufficient. The subordinate processes must also be reprogrammed (Logan & Gordon, 2001; Meiran, 2000). Still others argue that switch costs only reflect long-term memory processes that are outside of working memory. Switch costs reflect interference from past task sets or past associations to the stimuli under different task sets (Allport et al., 1994; Allport & Wylie, 2000; Meiran, & Kessler, 2008). The findings of the current experiment support theories suggesting that task-switching, specifically task set reconfiguration involves processes that occur outside of working memory. This proposition will be considered more in the second experiment.

Overall, the results of Experiment 1 found little support for switch costs above and beyond the time required to perform the task. While Logan (2004) found that reaction times in the perform condition were slower than the combined condition, this was not supported by the current research. Instead, the present findings indicate that there is no relationship between task-switching and working memory. This may in part be due to a difference in conceptualisation of the combined condition in the current study

and raises concerns about the validity of Logan’s methods for determining switch costs.