Rapid successive prompts queue in the central system rather than being processed in parallel

What it is

The psychological refractory period shows that when two stimuli requiring a decision appear in rapid succession, the central system handles them by queuing — processing them one after another, serially — rather than in parallel, handling both decisions at the same time. Even when the two tasks route through entirely different sensory-input and motor-output channels (one seen, one heard; one answered with a hand, one with a foot), the queuing effect still shows up, which indicates the bottleneck is not in the peripheral sensory or motor channels but in some shared central processing stage.

Why it happens

The bottleneck can be pinned to the central system rather than the periphery precisely because, if the delay were only caused by congestion in peripheral channels, giving the two tasks entirely independent input-output channels should let them be processed in parallel without interfering. But experiments repeatedly show that the queuing effect persists at roughly the same magnitude even when the channels are fully independent — pointing to a single shared resource, indifferent to which specific sensory or motor channel is involved, dedicated to the step of "making a decision / selecting a response." Whatever body parts the two tasks use, they both ultimately have to pass through this same stage.

Studying it

The key design for confirming "queuing rather than parallel" is to route the two tasks through different sensory channels and different responding limbs as much as possible; if the delay still follows the same inverse relationship with SOA — shorter interval, bigger delay — that rules out peripheral channel-sharing as the cause and pins it down to the central system. Some studies go further, using techniques like EEG to localize the processing stage that this shared bottleneck corresponds to, trying to determine whether it is response selection itself or an earlier stimulus-classification stage that is sharing the resource.

Where it stops holding

Not every dual-task pairing shows the classic serial-queuing pattern. When the second task is simple enough that it barely involves a genuine decision (close to a reflex), or when the two tasks have become integrated through extensive practice into a coordinated routine (a highly trained bimanual skill), the observed delay can be much smaller than the typical PRP prediction. These exceptions are often used to discuss whether the bottleneck can be partly bypassed under certain conditions, but there is no settled conclusion that the bottleneck can be eliminated entirely.

Applying it

Do not expect that putting two decision-requiring prompts on different sensory channels (one visual, one auditory) or requiring different response modes (one a tap, one spoken) will achieve "parallel processing" and shorten overall wait time — as long as both require the user to judge and choose, the central queuing limit still applies, and separating the channels does not deliver genuine parallel processing. The approach that actually reduces total time is cutting the number of tasks requiring an independent decision at the same moment, rather than spreading decision tasks across different channels. How to check: design a two-task test with fully separated channels and compare whether its reaction-time pattern still follows the queuing rule of shorter SOA meaning bigger delay; if it still does, channel separation did not deliver the expected parallel benefit.

Related

• Same group: A8.04.1 When the first response isn't finished, processing of the next stimulus is postponed · A8.04.2 The shorter the interval between two stimuli, the greater the delay in the second response · A8.04.4 This limit cannot be trained away
• Nearby: A8.01 The structure of reaction time
• Search terms: central bottleneck theory · serial processing · dual-task interference