How many chunks can people hold in working memory, and how much does this differ across individuals and tasks?

Why "how many chunks" has no clean answer

A chunk is a familiar unit that bundles smaller pieces of information together, like a word made of letters or a familiar fruit made of a color and a shape. The appeal of chunking is that it seems to let people bypass working memory's limits by storing one compact unit instead of many separate items [2]. But that immediately makes the question "how many chunks?" slippery: if a chunk can be almost any size, then a capacity measured in chunks is only as meaningful as the definition of chunk being used.

Experiments on chunking found that the benefit of chunking held even for other, non-chunked material held at the same time, which supports the idea that chunking genuinely reduces the load on working memory rather than just changing how items are counted [2]. Crucially, though, the benefit was independent of chunk size only when each chunk was made of unique elements that could be replaced by its first element; when chunks overlapped in their elements, that shortcut failed and chunk size mattered again [2]. The authors conclude that working memory capacity is not limited to a fixed number of chunks regardless of their size [2].

The task and the material change the answer

Whether features get bundled into chunks depends on what is being remembered. When people memorized drawings of fruits with colors and shapes, memory was best when the color matched the fruit's natural appearance, such as a red apple, but response times in that condition were faster, not slower [3]. If the features had been stored as a single chunk requiring an extra decoding step, the congruent condition should have been slower; instead the pattern led the authors to conclude that visual features of real-world objects are not stored in a chunk [3]. So chunking is not a universal compression trick that applies equally to all kinds of material.

Verbal memory shows its own task-dependent wrinkles. In monolingual Spanish speakers, longer words led to higher capacity estimates during the storage phase of a digit span task, a reversal of the usual word length effect, which the authors attributed to higher-order strategies such as chunking overriding the expected penalty [4]. During the manipulation phase, by contrast, longer words led to lower capacity estimates, the more traditional word length effect [4]. The same people, on the same kind of verbal task, showed opposite directions of the word length effect depending on whether they were simply storing or also manipulating the material.

Where individual differences actually show up

The clearest evidence for individual differences comes from comparing groups with different expertise. A meta-analysis of interpreters and interpreting students versus non-interpreters found a significant working memory advantage of medium size for interpreters, but only on tasks built from verbal stimuli, not on tasks built from non-verbal stimuli [1]. The gap also grew larger when the difference in interpreting expertise between the two groups was wider [1]. That pattern points to experience-linked differences in verbal working memory performance rather than a universal chunk-count that some people simply have more of.

Other work points toward strategy differences between individuals. High-span individuals, for example, notice their own names in an ignored channel during dichotic listening much less often than low-span individuals, a difference the paper discusses in the context of working memory chunking [5]. That is a difference in what gets through, not a direct count of chunks, and it fits the broader picture that people differ in how they use working memory rather than in a single fixed slot count.

What this means if you want to remember more

The practical lesson is that capacity is better thought of as flexible than as a hard number. Chunking can free up room for additional material, and the freeing effect appears to work by retrieving a compact chunk representation from long-term memory that replaces the individual elements, leaving capacity for what comes next [2]. That also explains why the position of a chunk matters: chunks early in a list improved recall of other material, while chunks at the end of the list did not [2].

So the honest answer to "how many chunks" is that it depends on how the chunks are built, what the task asks you to do with them, and how much relevant experience you bring. The evidence supports chunking as a real way to reduce working memory load [2], shows that it does not apply equally to all visual material [3], and shows that expertise is linked to measurable verbal working memory advantages [1]. None of these studies pins down a single universal chunk capacity, and the differences they document are about performance on particular tasks rather than a fixed internal container size.

Sources used in this answer

Related research questions