Academic journal article Memory & Cognition

How Task Errors Affect Subsequent Behavior: Evidence from Distributional Analyses of Task-Switching Effects

Academic journal article Memory & Cognition

How Task Errors Affect Subsequent Behavior: Evidence from Distributional Analyses of Task-Switching Effects

Article excerpt

Switch costs in task switching are often assumed to reflect the strengthening of task-related associations. Recently, we provided evidence that committing an error leads to the strengthening of the wrong task (Steinhauser & Hübner, 2006). In the present study, we report how error-induced control compensates for the effects of error strengthening. We hypothesized that, although error detection cannot prevent an error from being strengthened, it initiates processes that suppress the negative consequences of error strengthening on subsequent trials. Because error-induced control is known to operate slowly, we predicted that a compensatory effect should be observable only for slow responses. In three task-switching experiments, these predictions were confirmed by distributional analyses. The results extend our understanding of the interplay between procedural learning and executive control.

Endogenous control is a prerequisite of goal-directed performance. When control fails, behavior is no longer driven by goals, but rather by external stimuli. This can be demonstrated not only in patients with frontal brain lesions (Lhermitte, 1983; Norman & Shallice, 1986), but also in healthy persons under time pressure or distraction (Monsell, 1996; Norman, 1981; Reason, 1990). For instance, so-called capture errors occur when stimuli evoke an associated but unintended behavior. This is nicely illustrated by the classic example, described by William James, in which an absent-minded person goes to the bedroom to get dressed for dinner. After a while, however, he finds himself lying in bed dressed for sleep (James, 1890).

The aim of the present study was to investigate the consequences of such capture errors for subsequent performance. A correct response to a stimulus usually leads to an incidental learning of the current processing episode (Logan, 1988) or to a strengthening of the applied task (Meiran, 2000). In a recent study using the task-switching paradigm, we observed that, when an error occurs because the wrong task was accidentally applied, the wrong task is strengthened, which then impairs the execution of the originally intended task on a subsequent trial (Steinhauser & Hübner, 2006). In the present study, we addressed the question of whether some processes can counteract these negative consequences of incidental error learning.

Task Switching and Erroneous Task Strengthening

The task-switching paradigm was introduced in order to examine the control processes involved in switching between task sets (Allport, Styles, & Hsieh, 1994; Meiran, 1996; Rogers & Monsell, 1995). The paradigm requires that two or more tasks be performed in alternation. If the performance on task-switch trials is compared with that on task-repetition trials, so-called switch costs are usually observed, which were initially thought to reflect top-down control involved in task-set reconfiguration (Meiran, 1996; Rogers & Monsell, 1995; Steinhauser, Maier, & Hübner, 2007). However, it soon became obvious that the switch costs are also related to memory processes-that is, to priming or the associative strengthening of tasks. According to some accounts (e.g., Meiran, 2000; Schuch & Koch, 2003; Steinhauser & Hübner, 2006), task execution implies that task rules (such as "press left key when the stimulus is an even number") of the current task are strengthened, whereas rules of competing tasks are weakened. Switch costs emerge because repeating the previously strengthened task is beneficial, whereas switching toward the previously weakened task is costly. From this perspective, switch costs can be viewed as an indicator of procedural learning at the level of tasks.1

To investigate the strengthening mechanisms in more detail, we recently analyzed the switch effects on trials following errors (Steinhauser & Hübner, 2006). We found that after an error, the usually observed switch costs turned into switch benefits-that is, performance was improved on task-switch trials relative to task-repetition trials. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.