People often do multiple things at the same time. We can talk on the phone while stacking the dishwasher, and some university students seemingly know how to type messages on their cell phones while listening to a lecture. There are many occasions, however, when such multi-tasking breaks down or becomes dangerous. For example, notwithstanding how we might feel about it, there is strong evidence that use of your cellphone while you are driving is dangerous. People cannot focus on driving and talking on the phone at the same time without adverse consequences to performance—indeed, distracted driving may be more dangerous than drunk driving.
When two tasks cannot be conducted in parallel, it is common to invoke a “bottleneck” of processing, defined as a stage of processing that cannot be shared by a second concurrent task. In consequence, preparation of a second action is delayed until preparation of the first action is completed. This is obviously problematic if the first action involves a conversation on your cell phone and the second one requires application of the brake because the traffic light ahead just turned red. This phenomenon, that a task must await completion of another one, is known as the Psychological Refractory Period (PRP), which we have discussed here before.
There has been a plethora of research on the PRP, and much is now known about the parameters driving the effect. However, to date there has been no agreement on the exact locus of the bottleneck within the flow of information between stimuli and responses. Specifically, the bottleneck could conceivably impact perception of the stimulus (light turning red), selection of a response (press accelerator vs. brake), the initiation of a response (prepare to move foot to brake), or overt production of the response (press brake pedal).
A recent article in the Psychonomic Bulletin & Review provided a review of the existing evidence to pinpoint that locus. Researchers Stuart Klapp, Dana Maslovat and Richard Jagacinski came to the conclusion that the bottleneck “is due to a process that programs the timing of response initiation, and which must be completed immediately prior to responding”—in other words, the bottleneck occurs quite late in the sequence of information processing.
It is informative to flesh out how Klapp and colleagues came to this conclusion.
The figure below shows the workhorse methodology of research into the PRP together with several possible models of the effect. In all cases there are two stimuli, S1 and S2, that are separated by the stimulus onset asynchrony (SOA). The SOA is a primary variable that is manipulated in PRP experiments. Participants make two separate responses, R1 and R2, one for each stimulus. By methodological necessity, the response to S1 must be completed before the response to S2 can be made, but other than that the various components of processing are, in principle, free to overlap.
For example, in panel (b) there is no overlapping processing at all between the two stimuli—in consequence, the presentation of S2 is followed by a long delay interval before any cognitive processing commences. In panel (c), it is assumed that people can perceive the second stimulus immediately upon presentation, even if they are still busy with response selection for the first stimulus. Finally, in panel (d) most processing of S2 can occur concurrently with processing of S1, with the exception of programming the timing of response initiation.
How can experiments differentiate between the various models in the above figure?
The obvious implication of the model in panel (b) is that once SOA exceeds a certain threshold, the response time to S2 (RT2) should remain constant because all of the processing of S1 has been completed and adding more delay does not have any further effect (and RT2 is the same as it would be if in a single-task situation). Conversely, if SOA is shortened below that threshold, then the processing of S2 is postponed until all of S1-processing has been completed. In other words, by systematically reducing SOA until RT2 increases, the bottleneck can be identified.
But how would we know whether certain aspects of processing are occurring in parallel? How would the data differ if the model in panel (c) were true rather than that in panel (b)?
These two models can be disentangled by manipulating the perceptual difficulty of S1. Suppose one makes S1 very difficult to detect, so responding (RT1) is delayed. If the bottleneck blocks perception of S2 while S1 is being perceived, then the response time to S2 would be delayed (so provided SOA is sufficiently short, RT2 should increase together with RT1 when the first stimulus is harder to perceive). However, as noted by Klapp and colleagues, this result is not observed in the literature—instead, making S1 more difficult (and hence delaying R1) has no effect on RT2.
Finally, how would we know whether response selection is postponed by the bottleneck (as in panel (c)) or can occur in parallel (panel (d))?
The interpretative logic is similar to that involving perception: if the difficulty of response selection for S2 is increased (and hence slowed), then this manipulation should have less impact on RT2 when SOA is reduced if two response selections can occur in parallel. Conversely, if increased difficulty of response selection delays R2 irrespective of SOA, then clearly the bottleneck involves response selection. Klapp and colleagues suggest that the literature shows no evidence for the involvement of a bottleneck in response selection.
Putting it all together, after review of all available evidence, Klapp and colleagues conclude that:
“The PRP effect is attributable to a bottleneck that delays processing for a second response until processing related to a first response has been completed. The bottleneck does not include overt responding. Further evidence indicates that neither stimulus perception nor response selection produces this bottleneck.”
Instead, based on further review and analysis of findings from studies outside the PRP arena, Klapp and colleagues argue that the bottleneck results from programming the timing of response initiation. On this account, the PRP is seen to arise from a fundamental restriction on the timing and initiation of responses. Apparently we can detect stimuli in parallel and we can even figure out how to respond to each stimulus in parallel—however, we can only program the timing of one response at a time.
Psychonomics article highlighted in this post:
Klapp, S. T., Maslovat, D., & Jagacinski, R. J. (2019). The bottleneck of the psychological refractory period effect involves timing of response initiation rather than response selection. Psychonomic Bulletin & Review, 26, 29-47.