5.2.1 The importance of a top-down attentional set
It was illustrated in Chapter One that a visual search task has been a popular methodology for modelling selective attention. For instance, a simple search for a unique target among homogenous distracters is thought to occur preattentively, without the need for focused attention, whereas a conjunction search in which the target differs from distracters in more than one feature value is thought to require focused attention (e.g., Treisman & Gelade, 1980). Research in this area initially supported two distinct levels of search efficiency; a simple search is efficient, producing a shallow search slope regardless of display set size, and a conjunction
search is inefficient, resulting in a slope which increases with an increase in set size. Treisman and Gelade (1980) suggested that the steep search slope represents a serial search, in which every item in the display must be searched before the target can be found. This was questioned by studies showing that efficiency of a conjunction search can be improved by selectively searching through a “subset” of features. For example, Egeth et al. (1984) and Kaptein et al. (1995) showed that when searching for a target (e.g., a red circle) which shares one feature with some distracters (e.g., red squares) and another feature with other distracters (e.g., green circles) a feature-based
attentional set will allow observers to selectively attend to one feature (e.g., the colour red) meaning that fewer items need to be searched to find the target (all green shapes will be ignored). Utilising a visual search task therefore allows researchers to
investigate attentional selection, revealing how an observer may ‘parse’ a visual scene to make processing more manageable, and focus on task-relevant stimuli.
These findings regarding the selection of information for further processing have been accompanied by studies which have used visual search tasks to measure the relative influence of bottom-up and top-down control over the allocation of attention. For example Folk and colleagues (Folk et al., 1992; Folk et al., 1994) have made use of a spatial cuing task, and Theeuwes and colleagues (Theeuwes, 1991a, 1991b, 1992, 1994; Theeuwes & Burger, 1998; Theeuwes et al., 2006) have made use of a visual search for a feature singleton. The work by Folk et al. has provided evidence for the involuntary orienting hypothesis, whereby an irrelevant item presented in the visual field will only capture attention when it matches the top-down control settings established for the target search. They argue that this reveals the importance of top- down control over the exogenous capture of attention, and shows that attentional capture cannot occur solely on the basis of bottom-up processes. However the work
by Theeuwes has shown that an irrelevant item can capture attention regardless of any ‘feature relevance’ to the task demands, therefore providing evidence that attentional capture can be driven purely on the basis of bottom-up control. Due to the substantial amount of evidence for both sides of the argument, from the respective paradigms, many researchers have claimed that the selected methodologies of the two research groups may partially account for the variations in findings (this has been discussed in Chapter One of the present thesis, but see also Folk & Remington [2006] for an overview).
5.2.2 Implications of past studies on the present visual search task
As previously mentioned, the focus of this thesis is not to look at the balance between top-down and bottom-up influences. Instead the work concentrates on how attention is allocated using a top-down attentional set, and whether a set has the potential to persist to a second task and influence attention despite being irrelevant for the new task. It is therefore critical to the present methodology that an attentional set is adopted for the visual search task (in order to determine any persistence of this set to the change detection task); and the above research is beneficial as it outlines instances in which the task demands result in extraneous variables that can affect the top-down set.
For instance, to allow any carry-over effect to be measured, the top-down set must be very specific. This rules out a conjunction search (e.g., Treisman & Gelade, 1980) because the task would always involve a search for more than one feature which may compromise set specificity. Although the search will therefore be a ‘simple’ search task, taking into account the findings that a search for a feature singleton among homogenous distracters can be completed preattentively without the
need for focused attention (and without the need for an attentional set; e.g., Treisman & Gelade, 1980), the present task will require a search for more than one target presented among heterogenous distracters. The task instructions must also be such that they encourage the adoption of a specific set (e.g., feature-detection mode) as opposed to a more general set (e.g., singleton-detection mode; Bacon & Egeth, 1994), as this should benefit performance in the search task and prove that the set has indeed been adopted.
One further point to mention is that although the search should be sufficiently demanding to ensure the establishment of a top-down set, there should be no
possibility that a task switch or a switch in attentional set occurs in this initial task. Kumada (2001) has criticised the methodology used by Theeuwes and colleagues because although the task is described as a simple search, it is in fact a ‘compound’ search (Duncan, 1984). In his task Theeuwes (e.g., 1990; 1992) asked participants to search for a target defined by a specific feature (e.g., search for a colour singleton), however they had to report a different feature of this target once found (e.g., the orientation of a symbol inside the singleton). The to-be-reported feature therefore did not match the to-be-searched-for feature meaning that any top-down set based on the target-defining feature would not benefit target identification. If participants have to search for the target using one feature and then identify the target using another feature this involves a switch in the goals of the task. This may influence any carry- over effect and should therefore be avoided in the present work.
5.2.3 Persistence of a visual search strategy
The following experiments completed by the author study the persistence of a visual search strategy from one task to a second task, yet this is not the first example
of a strategy transferring between tasks. Research has shown that when participants are asked to search for a target among distracters, search efficiency is greater if a preview array is shown prior to the target array. This preview array contains a selection of distracters that will be present in the search array, and Watson and Humphreys (1997) suggest that these distracters are ‘marked’ so that when the search array is presented (containing the old distracters, the new distracters, and the target) attention will be inhibited from revisiting the already searched distracters and will instead be confined to the new items. This means that fewer items need to be searched to find the target compared to situations in which no preview is given. In the past this
visual marking effect was only thought to occur for locations (Watson & Humphreys,
1997), however more recently the effect has been found for objects (Watson & Humphreys, 1998) and features (Olivers & Humphreys, 2002; 2003), supporting the findings of Egeth et al. (1984) and Kaptein et al. (1995) that attention can be
selectively guided to a subset of features.
Visual marking is described as an inhibitory strategy which acts in a similar way to the attentional set proposed by Folk et al. (1992). Olivers and Humphreys (2003) state that whilst Folk et al. attribute contingent capture to an ‘excitatory’ attentional set, visual marking is an ‘inhibitory’ attentional set. Negative priming (e.g., Tipper, 1985) is very similar to this and is also indicative of an inhibitory strategy used in selective attention. Negative priming refers to the slow response to targets that have been distracters on previous trials. In line with contingent capture, when completing a search any item matching the target template will receive excitatory feedback, ensuring priority of these items in attentional selection. Houghton and Tipper (1994) also suggest that any item which does not match the target template will receive inhibitory feedback. Findings show that when inhibited
items (distracters) subsequently become targets, responses to the targets are slower and less accurate than responses to new targets that have not been encountered previously.
The evidence outlined above shows the importance of top-down control in guiding attention through the visual field, and biasing attention to the most task- relevant items and areas within a scene. Crucially it also shows that an attentional set (albeit an inhibitory set) can carry-over and influence performance on a later portion of a task. Building on these findings the following set of experiments will explore how an attentional set established to complete one search task may impact upon performance in a subsequent search task. In order to expand upon previous data in the literature (and the findings of the AB experiments reported earlier) the experiments will ensure greater differentiation between the two search tasks.