If you have ever reached the end of a paragraph and realised you have no idea what it said, you have experienced working memory failure in reading. The words were decoded. Your eyes moved correctly. But the sentence-level and passage-level meaning never formed — because working memory ran out of capacity before the integration happened.
Working memory is the cognitive system that holds information active while you process new information. In reading, it does several jobs simultaneously: it holds the beginning of a sentence while you reach the end, tracks the identity of pronouns as they shift across paragraphs, maintains the logical structure of an argument while processing the next step. The capacity of this system is one of the most reliable predictors of reading comprehension in the research literature.
The landmark Daneman and Carpenter study
The foundational research comes from Meredith Daneman and Patricia Carpenter's 1980 study, which introduced the reading span test. Unlike simple memory measures (digit span, word span), the reading span test requires participants to do two things simultaneously: read sentences aloud for comprehension, while also remembering the final word of each sentence. At the end of a set of sentences, participants recall the final words.
This dual-task design captures the essential demand of reading — you must store information (words already processed) while simultaneously processing new information (the next sentence). Reading span scores predicted SAT verbal comprehension scores, performance on pronoun resolution tasks, and the ability to integrate information across sentences significantly better than simpler memory measures (Daneman & Carpenter, 1980).
The finding has been replicated hundreds of times across languages, age groups, and reading contexts. Working memory capacity is one of the strongest predictors of individual differences in reading comprehension (Cain, Oakhill, & Bryant, 2004).
Why working memory matters more for complex sentences
The relationship between working memory and reading is not uniform across all text types. Simple, short sentences with familiar words make minimal demands on working memory. Complex sentences — those with embedded clauses, passive constructions, or long-range dependencies — make much higher demands.
Consider the difference:
- "The dog bit the man." (minimal working memory load)
- "The dog that the man whom the children like fed bit." (high working memory load — a centre-embedded relative clause)
When reading the second sentence, you must keep "the dog" active in working memory while processing two embedded clauses before you reach the main verb "bit." Readers with lower working memory span struggle disproportionately with these constructions (Just & Carpenter, 1992).
This is why reading dense philosophical or legal texts feels exhausting in a way that narrative fiction does not. The syntax of academic writing makes systematically higher demands on working memory.
The vocabulary connection
Here is a counterintuitive finding from working memory research: one of the best ways to effectively increase your available working memory for reading is not to train memory directly, but to build vocabulary.
When you encounter a familiar word, lexical access — retrieving the word's meaning — happens rapidly and automatically, consuming minimal working memory. When you encounter an unfamiliar word, you must consciously search for meaning, deploy inference strategies, or guess from context. This lexical struggle consumes working memory resources that would otherwise be available for higher-level comprehension.
Brysbaert, Mandera, and Keuleers (2018) found that readers with larger vocabularies are faster at lexical decisions and show less vulnerability to word frequency effects — their lexical access is more robust and less effortful. This efficiency translates directly into more working memory capacity available for integrating meaning.
The practical implication: reading widely in a domain is one of the most effective ways to free up working memory for that domain's texts. The first time you read Dostoevsky, the 19th-century Russian names and the dense psychological prose consume considerable working memory just to decode. The second time, much of this is automatic — freeing your working memory for the deeper structural and thematic work. This is part of why re-reading classic novels at higher speeds often preserves comprehension well.
Can you train working memory to read better?
This is a live area of debate. Working memory training programmes (like dual n-back tasks) do produce improvements on the trained tasks. The question is whether these gains transfer to reading comprehension.
A 2024 meta-analysis found that working memory training led to significant improvements in reading comprehension, particularly for readers with specific learning disabilities — but the effect sizes for typical readers were smaller and less consistent (Charqaoui & El Haddadi, 2024). The gains tend to be more robust when the training involves language-like stimuli rather than purely spatial or numerical tasks.
The current scientific consensus is cautious: working memory training can help, but it is not a reliable fast-track to reading faster with better comprehension for most readers. The more consistently effective route is reducing the working memory cost of decoding through vocabulary growth and domain familiarity.
Subvocalisation and working memory
The subvocalisation debate intersects interestingly with working memory research. Subvocalisation — the inner speech you hear when reading silently — appears to be partly a working memory strategy. The phonological loop (the component of working memory that rehearses verbal information) is actively engaged in silent reading, even though no external sound is produced.
Suppressing subvocalisation at high RSVP speeds may be partly an effect of working memory overload rather than a deliberate technique. When the phonological loop cannot keep up with the presentation rate, inner speech naturally diminishes. This is one mechanism by which RSVP at high speeds reduces comprehension — the phonological rehearsal that normally supports sentence integration becomes impossible.
At moderate RSVP speeds (300–400 WPM), most readers retain some inner speech and maintain sentence-level comprehension. The warpread.app RSVP reader lets you set your WPM precisely so you can find the speed at which phonological processing remains intact.
What this means practically
Working memory research gives you three concrete levers:
1. Build domain vocabulary before tackling dense texts. Before reading a dense academic text in an unfamiliar field, spend time with introductory material in the same domain. This reduces the per-word working memory cost and leaves more capacity for comprehension.
2. Adjust WPM to the text's working memory demands. A thriller novel makes minimal demands on working memory — high WPM is appropriate. A legal document or philosophical argument makes heavy demands — slow down to give working memory time to do its work. See our guide to reading speed for different content types.
3. Use active reading techniques to offload working memory. Note-taking, annotation, and summarising are essentially working memory externalisation strategies — they let you write down what you have processed so that working memory can process the next segment without having to hold the previous one. This is documented in detail in our post on active reading techniques.
Reading speed and comprehension are not simply a matter of how quickly your eyes move. They are a matter of how efficiently your cognitive architecture processes the information your eyes deliver. Working memory sits near the centre of that architecture.
Practice reading at your optimal speed on warpread.app — free RSVP reader
References
- Daneman, M., & Carpenter, P.A. (1980). Individual differences in working memory and reading. Journal of Verbal Learning and Verbal Behavior, 19(4), 450–466.
- Just, M.A., & Carpenter, P.A. (1992). A capacity theory of comprehension: Individual differences in working memory. Psychological Review, 99(1), 122–149.
- Cain, K., Oakhill, J., & Bryant, P. (2004). Children's reading comprehension ability: Concurrent prediction by working memory, verbal ability, and component skills. Journal of Educational Psychology, 96(1), 31–42.
- Brysbaert, M., Mandera, P., & Keuleers, E. (2018). The word frequency effect in word processing: An updated review. Current Directions in Psychological Science, 27(1), 45–50.
- Charqaoui, L., & El Haddadi, A. (2024). The Impact of Working Memory Training on Reading Performance. Science Step Journal, 6th issue.
Benchmark your reading performance
Turn the cognitive science into practice — take the free WPM speed test, then work through the Speed Reading Fundamentals course to build your technique.