Chapter 3 Chapter Summary

Chapter Overview:

This chapter focuses on the models of short-term memory (STM) and working memory. Most early models viewed STM as a static, limited-capacity, short-duration store of the information we are currently thinking about. Later models added the manipulation of the items in STM and renamed it working memory.

The first model of STM, fueled by H.M.’s case, was proposed by Donald Broadbent (1958). In this model, primary memory (his term for STM) was composed of two sub-stores: the S-system and the P-system. The S-system was a forerunner of sensory memory discussed in the previous chapter—a large capacity, pre-categorical store of sensory information. The P-system held conscious thought in a phonological code for a short period of time. Rehearsal was needed to maintain information in the P-system and sometimes led to the information travelling to secondary memory (long-term memory).

George Miller’s (1956) research on the capacity of primary memory suggests we can hold seven plus or minus two items. These items could be “chunked” into larger, meaningful units of information, thus increasing the capacity. Note though that later research (e.g., Simon, 1974) found that we can maintain fewer items in STM when the chunks are larger or more complex.

Other research in this era focused on the duration of primary or STM memory. In the well-known Brown-Peterson paradigm, participants had to remember three consonants (well below capacity). To prevent rehearsal, participants also had to count backwards by three or four from a specified number for three to 18 seconds. Peterson and Peterson (1959) found that recall accuracy decreased significantly with the delay and suggested that the duration of a trace in primary memory was about 20 seconds.

Atkinson and Shiffrin’s modal model of memory built on Broadbent’s model by including the processes by which information travels from one memory store to another. Information comes into a modality-specific sensory register. If it is attended to, it goes into STM, a small-capacity, short-duration store. Material in the STM can be encoded into the long-term memory store (LTS), where it can be held indefinitely.

Serial-position curve studies offer popular support for the modal model. Recency effects in a free recall task are thought to be the result of material still being held in STM. Primacy effects are thought to be the result of the material being encoded into LTM because of repeated rehearsal. However, more recent research was able to vary primacy and recency effects under different conditions and thereby calls into question this assumption.

The most common model of working memory is the multicomponent model, originally proposed by Baddeley and Hitch (1974). It contained three components: the phonological loop, the visuospatial sketchpad, and the central executive.

The phonological loop is the component that stores speech-like code. It also has an auditory rehearsal process (e.g., saying a list of words over and over to maintain them in working memory). This component can account for the several robust findings, including phonological-similarity effect (verbal recall is poorer when to-be-remembered items sound similar), the effect of articulatory suppression (verbal recall is poorer when participants are concurrently engaging in a task involving speech), the word-length effect (fewer words can be recalled if they take longer to say), and the irrelevant speech effect (recall was poorer when participants listen to speech sounds when learning verbal information).

The visuospatial sketchpad is the limited-capacity component in Baddeley and Hitch’s model that holds visual and spatial information. Research from mental scanning (e.g., Kosslyn, 1976) and mental rotation (e.g., Shepard & Metzler, 1971) revealed that information in the sketchpad are isomorphic representations (meaning they are life-like). Other research (e.g. Pearson et al., 1999) has shown that the visuospatial sketchpad does not use the same cognitive resources as the phonological loop but can be used cooperatively in the same task.

The third component of the Baddeley and Fitch model is the central executive, which supervises attentional resources to determine what information will enter working memory.

While Baddeley and Fitch’s model was well received, a number of problems with the model became apparent and Baddeley revised the model in 2000. This 2000 model introduced a new component: the episodic buffer. The episodic buffer can hold, integrate and manipulate information from the visuospatial sketchpad, the phonological loop, and LTM for a short period of time.

The Baddeley model has been studied and critiqued since 2000 and a new and improved version was published in 2012. No new components were added; instead changes in this version focused more on the flow of information in working memory. There are two main changes in this version. The first, items in both the phonological loop and the visuospatial sketchpad are now believed to flow directly into LTM. Secondly, the episodic buffer must not only contain multidimensional information but that the binding of modality-specific material must also occur here, thus making the episodic buffer the epicentre of working memory.

Other working memory models have also been proposed, including Cowan’s embedded-process model of working memory (1999). In this model, working memory is not an entity in itself but a set of processes that allows information from LTM and from sensory memory to be activated (paid attention to). Activated items are brought into our consciousness and, as in Baddeley’s (2012) model, can be manipulated.

The neuroscientific evidence supports the idea that STM and working memory are different entities. Different parts of the brain are active when we engage in STM tasks than when we manipulate the material. Scientific enquiries on the neurobiology of STM go back to the 1930s, originally suggesting that STM is localized in the prefrontal cortex. Later research questioned this assumption and the sensory-recruitment model has since become the dominant model. This model suggests that STM is not localized to one area per se but in the areas of brain responsible for the perceptual processing of the stimulus in STM.

Specifically, researchers such as Paulesu et al (1993) studied the phonological loop. They examined the activity in the brain during tasks that utilized just the phonological store and activities that just used the articulatory loop. The consensus now is that the phonological store is linked to area Brodmann area (BA) 40 in the parietal lobe and the articulatory loop in linked to Broca’s area. Other research has focused on the visuospatial sketchpad and has found that area 19 in the right occipital lobe and area 40 in the right parietal are associated with maintaining visual images in memory.

Working memory tasks, on the other hand, need attentional resources as well as the ability to integrate various modalities. Areas of the brain together called the frontoparietal cognitive control network are involved in the attentional component of working memory tasks. Since the episodic buffer binds information from multiple modalities, it is not surprising that tasks employing this component activate the neurological areas associated with the items being processed.

Research has suggested that the frontoparietal cognitive control network is also involved in emotional regulation. An increased ability to regulate our emotions is associated with superior performance on working memory tasks but not STM tasks.

Working memory skills are vital in academic learning and impaired performance in working memory tasks has been associated with lower academic performance. “Brain training” to improve working memory performance show promise as an intervention tool to improve academic success.

Learning Objectives:

Having read this chapter, you will be able to do the following:

1. Differentiate between the concepts of short-term memory and working memory.
2. Explain how the case of H.M. influenced early models of memory.
3. Sketch Broadbent’s model.
4. Critique Broadbent’s model.
5. Differentiate between Broadbent’s model and the modal model.
6. Sketch the modal model.
7. Critique the modal model.
8. Sketch Baddeley and Hitch’s (1974) multicomponent model of working memory.
9. Describe the central executive, the phonological loop, and the visuospatial sketchpad.
10. Describe experimental research that led to the development of the multicomponent model put forth by Baddeley (2000).
11. Describe the episodic buffer.
12. Sketch Baddeley’s (2000) multicomponent model of working memory.
13. Discuss experimental research that led to the development of the multicomponent model put forth by Baddeley (2012).
14. Sketch Baddeley’s (2012) multicomponent model of working memory.
15. Review early research that looked for a link between STM and the prefrontal cortex.
16. Explain the sensory recruitment model and provide evidence in support of this model from research using both verbal and visual stimuli.
17. Describe the neurological regions that are central to working memory.
18. Compare and contrast the brain regions involved in STM with those involved in working memory.