After completing study 2a using a 900 millisecond maintenance interval, it was found that participant could detect size changes as small as a 5% increase or decrease in the size of a shape. It was thought that the qualitative change detection task sizes of 25%, and to a less extent, 20%, in study 2a may be seen as not cognitively challenging enough; therefore, the 25% changed will be eliminated from the qualitative change detection tasks of the remaining doctoral studies to increase task difficulty. As the qualitative change detection task in study 4 will essentially use a 4000 millisecond maintenance interval to allow an inclusion of an interference task, current researchers decided to increase the maintenance interval in the current task so that study 2b was created. It was found that there was no difference between the percentage changes during the 4000 millisecond and 900 millisecond conditions.
Ideally, the study aimed to achieve a 75%-80% performance level. Clearly, the 25% change performance level was too high in both maintenance conditions and had to be taken out. It is proposed that if the 25% change had been used in the next investigation, the interference tasks would not be impacting upon a high resolution qualitative change detection task process as participants would more likely be using categorical comparison processes during the probe phases
97 4.10 Chapter Summary
This chapter piloted the stimuli and protocol which will be used during some of the subsequent experiments of the current doctoral thesis. Researchers used small changes in stimuli, of as little as 5% bigger and smaller, to demonstrate that participants could detect small changes within a qualitative change detection task. Combining the 900 millisecond and 4000 millisecond data demonstrated significant differences in the performance levels of the 25% and 5% changes in shape size.
Results have led researchers to disregard the 25% changes for the next experiment in the current doctoral thesis due to the ceiling effects presented with this percentage change. The use of the 4000 millisecond maintenance interval will continue within the dual task procedures of study 4 as this interval produced the most appropriate performance levels (75-81%) which demonstrated that the task was not too easy for participants.
The next experiment in the current doctoral thesis will look at the nature of the quantitative change detection task in more detail, focussing upon the potential inclusion of verbal representations in the task. The study will investigate whether a more multicomponent or domain specific approach can be taken with regards to representation use in the task.
98 CHAPTER 5
Quantitative Dual Task Experiment 5 Quantitative Dual Task Experiment 5.1 Chapter Overview
Chapters 3 and 4 piloted the qualitative and quantitative change detection tasks to be used in the remainder of this thesis, giving details of the most appropriate retrieval context to use and the most appropriate size changes to use with a young adult sample. This chapter of the doctoral thesis now introduces a dual task methodology to investigate the representations used within the quantitative change detection task. The aim of this study is to try and understand the working memory architecture at use during the completion of the quantitative visual memory task and to discover if the task demands are as visual domain-specific as suggested by Baddeley (2012) and Shah and Miyake (1996) and required by Luck and Vogel (2013). It has been suggested that visual tasks may use verbal representations and domain general attentional resources (Brown & Wesley, 2013; Shipstead & Yonehiro, 2016;
Vergauwe et al., 2009) and the current dual task methodology also aims to investigate this.
An introduction to the dual task methodology background literature is given with the results of study 3 of the current doctoral thesis.
5.2 Background
In chapter 1, details of the two working memory models were given.
Logie (2011) created a multicomponent approach to working memory, suggesting that the phonological stores and Visual Cache remain as two individual components. However, at a closer inspection of the model, Logie suggested that visual information is inputted through an Episodic Buffer like component. This component allows the integration of visual and verbal material within memory and because of this; Logie’s approach can be seen as a more multicomponent approach to visual working memory.
99 When considering a more domain specific approach, the model created by Baddeley (2012) suggests that visual information may be stored on a more perceptual level and that it is directed straight into the Visuospatial Sketchpad (VSSP). This perspective is seen as a gateway account as there are no suggested influences of long term semantic involvement or executive resource processes. The domain specific approach leads researchers to conclude no direct interaction between peripheral working memory components, such as the Phonological Loop or the VSSP.
Luck and Vogel (2013) supported the views of Baddeley (2012) and defined visual working memory tasks in a recent review, suggesting that working memory tasks use visual components only. Researchers’ defined visual working memory tasks as using visual representations only, and that if other types of representations were used (semantic, verbal) then this is not seen to be a visual memory task.
Shah and Miyake (1996) also proposed a more domain specific approach to working memory with their investigation looking at the separability of working memory resources for both spatial thinking and language processing tasks. In their investigation, using an interference protocols, researchers demonstrated how a spatial span task could correlate well with a spatial visualisation task (for example, a Spatial Relations Task) but not with any verbal ability measures. Similarly, the reading span task correlated well with the verbal ability measures (Verbal SAT Scores) but not with any spatial tasks. Showing how domain specific tasks could predict similar types of task (spatial tasks predicting the performance on other types of spatial tasks or verbal tasks predicting other types of verbal tasks), Shah and Miyake (1996) proposed a working memory approach whereby there are two systems within memory – one which utilised verbal information and another which utilised visuospatial information.
100 Brown and Wesley (2013) contradicted ideas from Luck and Vogel (2013) suggesting that visual tasks can incorporate aspects of mixed strategy use such as the use of semantics and verbal information. In their experiment, the VPT was investigated and it was found that participants could make use of semantics and verbal information when being presented with a matrix pattern. Brown and Wesley (2013) suggested that visual working memory capacity can be improved when visual tasks allow for the use of verbal coding and semantics as well as the initial use of visual information, meaning that the visual specific accounts may not be appropriate to discuss visual memory tasks. In addition, the Luck and Vogel (1997) visual working memory change detection task has come under criticism in the suggestion that semantics could be recruited when carrying out the task. Hartshorne (2008) argued that the change detection protocol could suffer from proactive interference, interference coming from several trials earlier. This would imply that the representation for the trial information is not so transient and potentially has a stable component underpinned by LTM (See Phillips &
Christie, 1977; and also Lin & Luck, 2012). Should this be the case then there is scope for LTM semantics recruitment. In addition, Brady and Alvarez (2011, 2015) suggested that when the memory arrays (and probe array) affords the opportunity for hierarchical organisation then participants will make use of a more abstracted representation of the visual information in order to enhance task performance. Thus, Shipstead and Yonehiro (2016) have argued that the relationship between change detection performance and complex cognition could results from domain specific and domain general resources demanded in visual change detection protocol demands.
One conventional method employed to look at the demands a task makes upon the working memory architecture is to use a dual task paradigm. In this case, two tasks are completed
101 simultaneously, with one task (Secondary Task) alongside (e.g. in the retention interval) of another (Primary Task).
Rudkin, Pearson and Logie (2007) used a dual task paradigm to investigate executive processes in both visual and spatial working memory tasks. The Corsi Blocks Task and the VPT were the primary tasks in this experiment with a random digit generation task being the secondary task. As both the Corsi Blocks Task and the VPT were initially not seen to involve any of verbal processing, it was suggested that the verbal secondary task would have no effect on these. However, results of the investigation demonstrated that both the Corsi Blocks Task and the VPT (in study 1) showed a decrease in performance when the random number generation was used, suggesting the use of potential generic executive/attentional resources and not just visual specific information. The Corsi Blocks dual task deficit was much greater, suggesting that this task may place a higher load upon the generic executive resources compared to the more visually related VPT.
Hamilton, Heffernan and Coates (2003) also used an interference paradigm to look at the developmental changes of the visual working memory architecture in children in relation to both visual (matrix patterns task) and spatial span tasks (Corsi Blocks Task). Four secondary tasks were used in this case - visual mask, spatial tapping, speech articulation and a verbal fluency task. Results demonstrated that working memory architecture does change between adults and children. The verbal fluency task was shown to have an effect on the visual span task (matrix patterns task) in both children and adults, suggesting an overlap in the use of both visual and executive components in working memory for all ages. However, speech articulation only affected spatial span in the children, suggesting that adults may use less phonological specific and more executive resources (or an Episodic Buffer component) within their working memory architecture for visual span tasks. As speech articulation has
102 been suggested to only hit the Phonological Loop within memory, this could suggest that the visual span task may use a form of higher level representations bound by Long Term Memory and this may be starting to develop within the memory structure of a child. An Episodic Buffer component, instead, is one which could be used to target such higher level integrated representations within a visual span task and could also be an explanation as to why the articulatory suppression had no effect upon the higher level visual span task. The use of an Episodic Buffer component will be detailed in a later section of this chapter when work of Brown and Wesley (2013) is discussed.
Vergauwe, Barrouillet and Camos (2009) investigated visual and spatial working memory dissociations, using both spatial and visual interference dual tasks. In a series of smaller experiments, participants took parts in tasks which assessed spatial and visual processing and storage abilities.
Results from this investigation suggested a more multicomponent approach to the working memory architecture with both visual and spatial secondary tasks hitting both visual and spatial primary tasks. In particular, it was demonstrated that the spatial interference task (Bar Fit task) also affected a visual matrix task presented, indicating that this task may use recruit visual and spatial resources. Although, the spatial interference task (bar fit task) affected the results of the spatial processing task (ball movements) at a higher level than the visual processing task, visual secondary tasks also affected the spatial processing task. Vergauwe et al. (2009), however, did not look at any use of secondary task explicitly demanding verbal resources, therefore the current study aims to do this using two visual change detection tasks.
For the purposes of the current study, the Bar Fit Task (Vergauwe et al., 2009) will be used as the visual secondary task. This task was demonstrated to show interference in spatial and visual working memory; therefore, should provide current researchers with a clearer idea of
103 which working memory components are at use when completing the current quantitative change-detection task which has been created. As visual and spatial interference effects are associated with the VSSP (Baddeley, 2012), the Visual Cache and the Inner Scribe (Logie, 2011), interference effects would show the use of any specific components during the tasks.
Interference effects are not just visual specific. Lépine, Bernardin and Barrouillet (2005) used a verbal parity task to demonstrate verbal interference effects. It was found that the administration of a simple verbal task (stating if a number is odd or even) had a detrimental effect on memory span (reading span). This task disrupted the memory of the working memory span task, leading researchers to conclude that it may not be the task itself which causes interference but the fact that attention is directed away from the primary task.
However, from their study, Lépine et al. (2005) concluded that any phonological task may have had the ability to cause disruption as interference could be concerned with the allocation of attention to each task as well as disrupting the specific working memory components involved in processing and storage. This is one question that will be addressed within the current experiment as the fundamental aim is to identify whether the verbal parity task (as used by Lépine et al., 2005) does cause interference in a visual memory paradigm through its generic attentional demands (see Vergauwe et al., 2009 above).
In contrast to Lépine et al. (2005), Jones Madden and Miles (1992) suggested that irrelevant phonological speech could not disrupt any visual information stored. In their research, the repetition of the letters F, K, L, M, P, Q and R. were shown to cause no effects to the primary visual memory task. One reason for the lack of interference could be due to the components at use during the secondary irrelevant speech task. The earlier work of Jones et al. (1992) suggested that the irrelevant speech interference made demands upon the Phonological Loop alone and therefore would not disrupt any visual task which made use of the non-verbal
104 components. This account contrasts with that of Lépine et al. (2005). If the interference was due to the allocation of attention alone then the research of Jones et al. (1992) would show effects with the visual task. As this was not the case, the use of the verbal parity task in the research of Lépine et al. (2005) may be more verbal specific. Articulatory suppression alone may not be expected to have a cross modal effect if its generic attentional demand is low. The verbal parity task, however, demands Long Term Memory access and this demand upon attentional resources (see Unsworth & Engle, 2007; Unsworth & Spillers, 2010; Unsworth et al., 2014) may be sufficient to impact upon the primary change detection tasks in the current change detection protocols.