2 contrasting models of working memory capacity give details of the constraints acting upon how participants can memorise and recall items that have been stored in short-term visual memory. These two models will inform the methodological memory protocols employed for the duration of the thesis.
1.5.1 Discrete Slot Model
The Discrete Slot Model (Awh, Barton & Vogel, 2007; Luck & Vogel, 1997) conceptualised working memory capacity in terms of the number of items that can be stored. The model, as seen in Figure 1.9. suggests that working memory contains ‘slots’ which are filled when items are stored in working memory. Luck and Vogel (1997) give a capacity limit of 3-4 items, suggesting that when the capacity limit is reached, no further items can be stored and remembered. This is in line with the ‘magical number 4’ suggested by Cowan (2000). The Discrete Slot Model focuses upon large categorical changes (quantitative changes) in stimuli, for example changing a blue square to a red square. Luck and Vogel (1997) suggested that the Discrete Slot Model discusses information storage as whole or integrated object items instead of individual features. For example, a person could store a square and its orientation
19 and colour in one slot instead of having a separate slot per feature. This essentially suggests that the maximum of 3-4 items could hold many features. If each object has four features, then as well as the four different items being held, sixteen features can also be held.
Rouder, Morey, Cowan, Zwilling, Morey and Pratte (2008) provided evidence for the slot model in suggesting that visual working memory does indeed have a fixed number of slots. In this experiment, array sizes 2, 5 and 8 were manipulated from Luck and Vogel’s (1997) original paradigm. Receiver Operating Characteristics curves/lines were demonstrated to be at 1 for all set sizes, suggesting a linear decrease in performance of the visual memory task as set size increased. Researchers here suggested that the slot model could work just because of its simplicity although further research was suggested in an attempt to incorporate the potential encoding individual differences.
1.5.2 Shared Resource Model
The Shared Resource Model (Bays, Catalo & Husain, 2009, Bays and Husain, 2008) focused upon the resource trade-off between quantitative change detection task demands and qualitative change detection demands (small continuous changes). In stimuli, e.g. the finite or continuous coloured hue characteristics of a square. In this way, researchers can understand how the items are stored in memory, and in this case, the model suggests resource capacity may be best considered in the light of a trade-off between the number of items in an array and in terms of their resolution and precision of the items, therefore, not just simply how many items are stored.
Bays et al. (2009) suggested that one resource is shared equally amongst all items in a visual array meaning that each item can be stored. Within their research, array sizes of 1-6 were used and the precision of the items declined even between 1 and 2 items in the array. In
20 comparison to the Discrete Slot Model, the Shared Resource Model would enable an array size of 6 to be stored successfully within memory. Precision would simply be less accurate than with a lower array size. As the Discrete Slot Model has a capacity limit of 4, an array size of 6 would simply not be encoded within memory due to the limited slots available for storage. As there was no fixed upper limit suggested, this indicates that all items in an array can be stored. Where more items are in the array to be store, less resource is allocated to each item due to it being distributed to more items in memory.
Figure 1.9. An example of the types of representations stored by the Shared Resource Model (centre) and Discrete Slot Model (right) taken from Luck and Vogel (2013).
Similarly, to Bays et al. (2009), Zhang and Luck (2008) created a model which used dynamically distributed resources. However, researches here combined both the idea of slots and a shared resource within the same model, naming this the ‘Slots+Resource’ model. The
‘Slots+Resource’ model suggested that people do store items within working memory slots, and as with the shared resource account, therefore are no upper limits. However, instead of the model simply counting the number of slots free, it was suggested that a resource is shared between each slot so that fine detailed slots could be created. Using a colour wheel methodology (to be critiqued in the following chapter) to show the precision within working
You see this Do you remember Or this?
Discrete Slots Continuous
21 memory capacity, it was suggested that the combination of slots and resources gave a more flexible account than the discrete slot model of Luck and Vogel (1997). However, although a limited amount of slots and a limited amount of resource was suggested, a concrete number was not applied to this model, creating its flexibility.
1.5.3 Contrasting View – Information Limited Model
Initial work from Brady and Alvarez (2011) suggested that the different objects in an array could influence the working memory capacity of an individual. However, this research used stimuli only created for their work (different coloured circles), therefore Brady and Alvarez continued their investigations using very familiar change detection paradigms.
In their more recent work, Brady and Alvarez (2015) discussed a model which suggests that there was no evidence of a fixed slot approach and suggested that instead of a fixed slot approach, working memory storage consisted of the storage of whole complex objects. In their research, an ‘Information Limited Model’ was used to explain working memory capacity. This Information Limited Model proposed that participants can only store 1-2 complex items in an array, unlike the previously proposed 3-4 items as suggested by Luck and Vogel (1997).
A series of three experiments were conducted, it aimed to replicate the results of Awh, Barton and Vogel (2007), Luck and Vogel (1997) and Alvarez and Cavanagh (2004) who all demonstrated the capacity limit of approximately 4 items. These included items with a range of level of complexity (Chinese characters, cubes, grey polygons and Snodgrass object).
Brady and Alvarez (2015) demonstrated that participants could in fact only store one or two items within working memory and not the 4 items as previously suggested. These items, however, were not the simple square items as used by Luck and Vogel (1997). The items were very complex items such as the Chinese characters and cubes as used by Awh et al.
22 (2007) and Alvarez and Cavanagh (2004). Using both small changes in stimuli, known as within-category changes, and larger changes in stimuli, known as cross-category changes, it was shown that the cross-category changes were affected by the distribution of the other items in an array. The dispersion of the items in the arrays was positively correlated with the cross-category change performance. One reason for this correlation was suggested as the influence of the other items in the visual array. When clusters of different items were presented, improvements with cross-category changes were found. This was explained using a phrase known as ‘spatial ensemble representation’ and suggests that the storage of the clustered items allowed participants to store one key feature of a single items separately to the rest of the cluster. This key feature, for example, the darker shade of a cube, would mean that participants could easily recall the feature and therefore the object in the cross-category condition. As the within-category condition had no one feature which stood out to participants all items were stored as complex items making it more difficult to recall one individual items.
1.6 Relationship Between the Two Models (Number/Quantity versus