Distraction and action slips in an everyday task: Evidence for a dynamic representation of task context

Matthew Botvinick, Lauren M. BylsmaPsychonomic 2005 · DOI

Abstract

Contrary to traditional theories suggesting that distractions are most dangerous at critical decision points, this study demonstrates that routine "action slips" are significantly more likely when interruptions occur in the middle of a subtask. Using a naturalistic coffee-making task, we provide evidence for a connectionist model of behavior where mental representations of task context are dynamic and fragile. Our findings suggest that the internal signal tracking "what has been done" is weakest and most susceptible to noise when it is not immediately required for the current motor action.

Introduction: The Mystery of Action Slips

We have all experienced "action slips"—mundane errors like putting milk in the pantry or attempting to start a car that is already running. For decades, the dominant psychological theories, pioneered by researchers like Norman and Reason, posited that these errors occur primarily at "decision points." These theories argued that humans utilize "attentional checkpoints" to consciously verify progress. Under this framework, a distraction is most hazardous if it occurs exactly at a checkpoint, causing the individual to skip the necessary verification and proceed incorrectly.

However, recent computational work using recurrent neural networks suggests a different mechanism. Rather than a static checklist, our internal sense of context—the memory of what we have already accomplished—may be a dynamic signal that fluctuates in strength throughout a task sequence.

Theoretical Framework: The Dynamic Context Signal

The core hypothesis of this study is that the mental representation of task context is not uniform. It is strongest at decision points, where it is vital for selecting the next discrete step in a sequence, and weakest during the execution of a routine subtask (such as pouring or stirring), where the immediate action is the same regardless of the broader progress.

Figure 2: The dynamic signal of task context. The 'distance' metric reflects the strength and distinctness of the internal representation, which reaches its nadir during the middle of a subtask.

As shown in Figure 2, the internal representation of the task undergoes a "U-shaped" fluctuation. Because the context signal is attenuated or "quieter" during the middle of a subtask, it becomes significantly easier for the cognitive "noise" introduced by a distraction to overwhelm and erase that representation. This "wiping of the mental whiteboard" means that when the individual eventually reaches the next decision point, they no longer possess the internal data required to know which steps have been completed.

Methods

To test this, 22 participants were asked to prepare 50 cups of instant coffee following strict procedural rules (e.g., adding coffee first, then cream and sugar in any order, with stirring required after each ingredient).

On 80% of the trials, participants were interrupted with a "Freeze!" command and required to solve a mental subtraction problem (e.g., calculating change from a $10.00 bill). The researchers manipulated the timing of these interruptions into two categories:

  1. Mid-subtask: Immediately after pouring an ingredient, while the participant still needed to stir before reaching a decision point.
  2. End-subtask: Immediately after stirring, just before the next decision point.

Performance was tracked via video coding to identify overt errors (repetitions or omissions) and "microslips" (aborted incorrect reaches).

Key Results

The experimental results provided striking support for the connectionist model over traditional attentional-check theories. Interruptions occurring mid-subtask led to a 4.0% error rate, whereas interruptions at the end of a subtask (the supposed "decision point") resulted in only a 0.9% error rate. Notably, no errors occurred during uninterrupted control trials.

Figure 3: Survival plot of error-free trials. Distractions occurring mid-subtask (black circles) lead to a significant drop in performance compared to end-subtask distractions (open circles).

As illustrated in Figure 3, the drop in performance is specifically tied to the mid-subtask distraction. This effect remained significant even after filtering for minor execution errors (like forgetting to stir), confirming that the vulnerability specifically impacts the high-level tracking of task progress.

Discussion and Conclusion

These findings suggest that action slips are not merely failures of attention at key moments, but rather the result of a fundamental degradation of the background information the brain uses to track state. A distraction in the middle of a subtask has delayed consequences: it doesn't necessarily cause an error in the current motor act, but it destroys the context needed for the next decision.

While this study focused on the mundane task of coffee making, the implications are broad. If our cognitive representations are most fragile when they are "offline" during routine actions, this U-shaped vulnerability may exist in high-stakes environments like surgical theaters or cockpits. Future research utilizing Transcranial Magnetic Stimulation (TMS) could further explore the neural basis of this vulnerability by targeting the prefrontal cortex at specific intervals during sequential tasks.

Summary updated

BibTeX
@inproceedings{botvinick2005distraction,
  author = {Matthew Botvinick and Lauren M. Bylsma},
  title = {{Distraction and action slips in an everyday task: Evidence for a dynamic representation of task context}},
  booktitle = {Psychonomic},
  year = {2005},
  doi = {10.3758/bf03206436},
}
Figure