Cognitive Connie
Understanding baddeley & hitch — working memory
By the early 1970s, the multi-store model proposed that a single short-term store was the obligatory gateway to long-term memory and reasoning. If that were true, filling STM completely with to-be-remembered digits should cripple any concurrent cognitive task. Alan Baddeley and Graham Hitch set out to test this directly.
Defining features
Key figures
Alan D. Baddeley
1934–British cognitive psychologist who developed the working memory model with Hitch (1974), extended it with detailed accounts of the slave systems (1986), and added the episodic buffer (2000). Has remained the model's chief architect and advocate across five decades, making it the dominant framework for research on short-term, active memory.
Graham J. Hitch
1946–Co-designed and ran the original 1974 dual-task experiments with Baddeley. Continued to develop evidence for the phonological loop and working memory capacity across his career at the University of Lancaster and later York.
Methods
The dual-task paradigm
The logic is simple: if two tasks use the same resource, performing them together should hurt performance proportionally to the shared load. If they use separate resources, they can run in parallel with little interference.
Digit preload manipulation
Participants were given a sequence of digits to hold in memory throughout a concurrent task. Preload sizes ranged from 0 (control), to 3 digits (moderate), to 6 digits (approaching or at STM span). If STM were a unitary gateway shared by all cognitive tasks, the 6-digit load should have virtually paralysed concurrent performance.
Concurrent reasoning task
The primary concurrent task required participants to verify grammatical sentences such as "B is followed by A — BA" (true) or "A is not preceded by B — AB" (false). This task requires active reasoning and language comprehension — processes that, on the unitary STM account, should compete directly for STM resources.
Concurrent reading and free recall
In additional experiments, the concurrent task was prose comprehension or free recall of word lists. These varied the type of cognitive demand placed alongside digit preloading, allowing Baddeley and Hitch to assess whether interference was domain-specific or general.
Digit recall at end of trial
At the end of each trial, participants reproduced the digit string. Digit recall was near-perfect across all conditions, confirming that participants genuinely maintained the preload — they were not simply ignoring it to perform better on the concurrent task.
Findings
What the dual-task data showed
Modest interference even at full digit span
Holding 6 digits (at or near STM capacity) slowed reasoning by only ~10% and increased errors modestly compared to the zero-load baseline. This is far less disruption than a unitary gateway account predicts — if all cognitive tasks flowed through one STM, fully occupying it should have stopped reasoning cold.
Separate verbal storage and reasoning systems
The data showed that verbal storage (holding digits) and reasoning/comprehension draw on partially separate resources. Interference was graded and modest rather than catastrophic — consistent with separate subsystems with some central resource overlap (the executive) rather than a single shared gateway.
Articulatory suppression selectively disrupts phonological tasks
Further experiments showed that articulatory suppression (saying "the-the-the" aloud) selectively impaired memory for acoustically similar word lists but not visuospatial tasks — demonstrating that the verbal slave system (later called the phonological loop) is distinct from the visuospatial system.
Criticism
Limitations and ongoing debates
Central executive is underspecified
Baddeley himself acknowledged that the central executive was the most important but least understood component — sometimes called a "ragbag" of attentional functions. Its architecture remains disputed: Miyake et al. (2000) proposed three separable executive functions (inhibition, updating, shifting), but the precise relationship between these and the original Baddeley-Hitch executive is still debated.
The model is difficult to falsify
When new phenomena are discovered, the model can accommodate them by positing new components or sub-processes (as with the episodic buffer added in 2000). This flexibility has been criticised as making the model unfalsifiable — any data can be post hoc fitted by adding complexity.
Embedded-processes and other competing models
Cowan's (1995) embedded-processes model and Oberauer's (2002) three-level model offer alternative architectures that do not require separate slave systems, instead framing working memory as activated long-term memory with a focus of attention. These models handle many of the same phenomena with arguably simpler architecture.
Relevance for today
Why it still matters
Dual-task methodology as a cognitive research tool
The dual-task paradigm introduced by Baddeley and Hitch is now a standard tool across cognitive and clinical neuropsychology. Whenever researchers want to know whether two cognitive processes share resources, they use variants of the digit-preload-plus-secondary-task design — directly inheriting from this paper. It is the primary method for dissociating the components of working memory and mapping their neurological substrates.
Working memory capacity and individual differences
Working memory capacity (WMC) — measured by complex span tasks that require simultaneous storage and processing — is among the strongest predictors of fluid intelligence, reading comprehension, and academic achievement. This entire research programme grew directly from Baddeley and Hitch's insight that STM is not a passive store but an active workspace for ongoing cognition.
Clinical applications in ADHD, aging, and TBI
The working memory model provides the theoretical framework for assessing executive and phonological loop deficits in ADHD, age-related cognitive decline, and acquired brain injury. The central executive maps onto prefrontal function; the phonological loop onto perisylvian language areas; the visuospatial sketchpad onto parieto-occipital regions — giving the model neurobiological traction alongside its cognitive architecture.
Test your knowledge
Frequently asked questions
What did Baddeley and Hitch prove with the dual-task study?+
That verbal storage and reasoning draw on largely separate cognitive resources. Holding a full digit span in memory slowed reasoning by only ~10% — far less disruption than a unitary STM gateway would predict. This proved that STM is not a single bottleneck through which all thought must pass, and motivated the multicomponent working memory model.
What is the difference between the multi-store model's STM and Baddeley's working memory?+
Atkinson and Shiffrin's STM is a passive temporary storage buffer — a single box. Baddeley's working memory replaces it with an active, multicomponent system: a central executive (attentional controller), a phonological loop (verbal storage and rehearsal), a visuospatial sketchpad (visual/spatial storage), and an episodic buffer (multimodal integration). The key difference is that working memory actively processes information rather than merely holding it.
Sources
Last reviewed August 2026- 1.
Baddeley, A. D., & Hitch, G. (1974). Working memory. In G. H. Bower (Ed.), The Psychology of Learning and Motivation (Vol. 8, pp. 47–89). Academic Press.
+About this source
The foundational paper introducing the multicomponent working memory model and reporting the dual-task experiments that motivated it.
- 2.
Baddeley, A. (2000). The episodic buffer: A new component of working memory? Trends in Cognitive Sciences, 4(11), 417–423. https://doi.org/10.1016/S1364-6613(00)01538-2
+About this source
Introduces the episodic buffer as a fourth component bridging the slave systems and long-term memory.
- 3.
Miyake, A., Friedman, N. P., Emerson, M. J., Witzki, A. H., Howerter, A., & Wager, T. D. (2000). The unity and diversity of executive functions and their contributions to complex "frontal lobe" tasks: A latent variable analysis. Cognitive Psychology, 41(1), 49–100. https://doi.org/10.1006/cogp.1999.0734
+About this source
Latent variable analysis proposing three separable executive functions (inhibition, updating, shifting) that partly decompose the underdefined central executive.