The two subsystems draw on relatively independent resources, so simultaneous information in different channels doesn't interfere

What it is

The phonological loop and the visuospatial sketchpad don't share a single resource pool that they take turns drawing from — each has a relatively independent pool of its own. Putting one piece of information into each channel at the same time produces far less mutual interference than forcing two pieces of the same kind of information into a single channel. This is a key structural claim of the multi-component model of working memory: the capacity limit isn't a single overall gate, but shows up separately across several relatively independent subsystems.

Why it happens

If the two subsystems really did share one pool of processing resources, adding information to either channel should equally drag down performance in the other. What's actually observed is a fully crossed pattern instead: a manipulation that specifically disrupts verbal processing (articulatory suppression) only drags down verbal tasks, and a manipulation that specifically disrupts spatial processing (spatial tapping) only drags down visuospatial tasks — swap the target and each manipulation barely does anything. This precise, non-overlapping "who disrupts whom" pattern only shows up if two genuinely separate resources sit behind the two channels; if a single shared resource were behind both, interference should be broad and wouldn't land this precisely on just the matching task type.

Studying it

The core paradigm for testing this independence is the crossed dual-task design: have the same participants perform tasks under four combinations — verbal retention plus verbal interference, verbal retention plus spatial interference, spatial retention plus verbal interference, and spatial retention plus spatial interference — and compare the performance drop across the four. The expected pattern is a clear drop for matched combinations (verbal-with-verbal, spatial-with-spatial) and a much smaller drop for crossed combinations (verbal-with-spatial, spatial-with-verbal). This double dissociation is stronger evidence than a single-direction dissociation, because it rules out the simpler explanation that one manipulation is just harder in general and drags down every task regardless of type.

Where it stops holding

The independence being described here is relative, not an absolute separation: if a task actually requires converting or cross-checking between the two kinds of information — translating a remembered spatial path into a verbal description, or checking a spatial position against a verbal instruction — it necessarily engages both channels and forces their contents to interact substantively, and interference reappears, because the part responsible for coordinating and converting between them doesn't belong to either independent storage subsystem — it's a separate, overarching mechanism. Independence mainly holds for "holding two things in parallel," not for scenarios that "require active integration."

Applying it

• If a user needs to remember both a piece of verbal information and a piece of spatial or visual information at the same time through a brief interruption (a spoken confirmation plus a remembered position, say), it's safe to have them hold both in parallel, as long as they aren't asked to actively convert or cross-check one against the other during the interruption — mutual interference will be small.
• Avoid loading two pieces of information that are fundamentally the same type (two pieces of text that both have to be held through silent rehearsal, or two layouts that both have to be held through spatial memory) onto a user at the same time — this same-channel stacking causes far more interference than it might seem to, even when the two pieces look like "two different things."
• The moment a task design requires "convert this information into another form before using it," recognize that this is no longer simple parallel holding, and it needs the extra scrutiny appropriate to a more error-prone integration task.
• Verification: design a comparison test covering the four combinations — verbal-verbal, verbal-spatial, spatial-verbal, spatial-spatial — and check whether same-type combinations show a noticeably larger drop than crossed combinations. If all four drop by similar amounts, the task actually requires cross-channel integration, and the "channels are independent, no interference" assumption doesn't simply apply.

Related

• Same group: A6.12.1 The phonological loop holds verbal and numeric information · A6.12.2 The visuospatial sketchpad holds visual and spatial information · A6.12.4 The central executive coordinates the subsystems and allocates attention, and has its own capacity limit · A6.12.5 Presenting the same information through both visual and verbal channels can raise working memory's effective capacity
• Nearby: A5.02 Divided attention and dual-tasking
• Search terms: double dissociation · dual-task interference · domain-specific working memory