Carry-over of attentional settings between distinct tasks: A transient effect independent of top-down contextual biases

https://doi.org/10.1016/j.concog.2021.103104Get rights and content

Highlights

  • Attentional settings in a task persist to a second unrelated task, affecting search.

  • Top-down biases within scenes have no impact on the persistence of attentional set.

  • The effect of a previously relevant attentional set on eye movements is brief.

Abstract

Top-down attentional settings can persist between two unrelated tasks, influencing visual attention and performance. This study investigated whether top-down contextual information in a second task could moderate this “attentional inertia” effect. Forty participants searched through letter strings arranged horizontally, vertically, or randomly and then made a judgement about road, nature, or fractal images. Eye movements were recorded to the picture search and findings showed greater horizontal search in the pictures following horizontal letter strings and narrower horizontal search following vertical letter strings, but only in the first 1000 ms. This shows a brief persistence of attentional settings, consistent with past findings. Crucially, attentional inertia did not vary according to image type. This indicates that top-down contextual biases within a scene have limited impact on the persistence of previously relevant, but now irrelevant, attentional settings.

Introduction

To allow for effective task completion an individual’s limited cognitive resources will be biased towards relevant information and locations and away from irrelevant information and locations. This is achieved via the adoption of a top-down attentional set (e.g., Egeth et al., 1984, Folk et al., 1992, Leber and Egeth, 2006). The top-down set supports selective attention (Johnston and Dark, 1986, Schneider and Shiffrin, 1977, Theeuwes, 1993) by prioritising specific stimuli and areas of space on the basis of task demands. It therefore dictates which stimuli will be selected and which will be ignored. Findings show that these ‘attentional control settings’ (Folk et al., 1992) influence the allocation of attention and therefore performance in a task (e.g., Kaptein et al., 1995, Rossi and Paradiso, 1995).

When a task changes the attentional settings should be updated to reflect this, however this is not always the case. Studies have shown that after responding to specific targets over a large number of trials, participants are unable to inhibit these targets when they subsequently become irrelevant (e.g., Leber and Egeth, 2006, Thompson et al., 2007). This is argued to be due to the persistence of the top-down attentional set; the stimuli that were previously task-relevant continue to match the attentional settings and so resources are allocated to them. Leber and Egeth (2006) suggested that the failure to change set is due to the fact that the costs associated with switching set (greater investment of cognitive resources in order to inhibit the old settings) outweighs the benefit to performance of switching (in their study this benefit was the reduced contingent capture by irrelevant peripheral distracters). They supported this proposal by showing that an attentional set would persist from one task to a second task following 320 trials in the first task, but not following only 40 trials. The increased experience serves to consolidate the set, making it more difficult to update in line with new task demands, requiring more resources to inhibit.

The findings of Leber and Egeth (2006) suggested that perseveration of an attentional set would only occur following significant investment in the initial set. This would indicate that in more dynamic tasks, persistence of top-down settings would not pose a problem. However, newer findings show that this is not the case. For example, Wendt, Kähler, Luna-Rodriguez, and Jacobsen (2017) have demonstrated the persistence of spatial visual attention between two tasks of short duration. In an initial task participants were presented with three letters and were asked to identify the central target, therefore adopting a narrow focus of attention due to the need to inhibit flanking distracters, or they were asked to judge whether the three letters were the same or different, therefore adopting a wide focus of attention to allocate resources to all items. Following this first task participants were given a ‘probe’ task in which three digits were presented to the same locations as the letters in the preceding task and participants had to identify a particular target that was located in the centre or the periphery of the digit string. They found that central targets were identified faster than peripheral targets when the probe task was preceded by the flanker task, but the difference between response times to central and peripheral targets was much smaller following the same/different task. This shows that the spatial spread of attention adopted for one task can persist to a second task, even when this leads to a detriment to performance.

The study by Wendt et al. (2017) shows even limited exposure to an initial task can result in the persistence of attentional settings to a second task, yet their paradigm incorporated relatively simple stimuli presented to the same spatial location across all trials. Other work has adopted methods that are better able to reflect dynamic, real-world situations. Thompson and Crundall (2011) developed a set-switching paradigm in which participants searched through letter strings presented horizontally, vertically, and randomly across the screen and then searched a real-world image. To measure the impact of the letter strings on the allocation of attention to the images eye movements were recorded and results showed that spatial visual attention in this second task was influenced by the orientation of the preceding letter search. Vertical search in the picture task increased following a vertical letter search and decreased following a horizontal letter search.

These studies measure attentional set switching, yet this often occurs in tandem with task switching because when a task changes this requires an update in the internal rules used to respond to specific stimuli (an intentional set) and an update in the rules used to select specific stimuli for processing (an attentional set) (Rushworth, Passingham, & Nobre, 2002). In a standard task switching paradigm participants are shown stimuli and are asked to use one of two rules in order to respond to the stimuli. For example, when presented with a blue letter, participants respond to whether it is an uppercase or lowercase letter, when presented with a red letter, participants respond to whether it is a vowel or a consonant. In ‘no-switch’ trials the colour will be consistent (e.g. a red letter followed by another red letter), in ‘switch’ trials the colour will be different (e.g. a red letter followed by a blue letter). Numerous studies have shown that response times are longer on switch trials compared to no-switch trials and the difference between the two trial types is referred to as a switch cost (e.g., Meiran, 1996, Rogers and Monsell, 1995, Wylie and Allport, 2000). Allport, Styles, and Hsieh (1994) accounted for switch costs using the task-set inertia hypothesis which states that the old set will persist to a new task and the resources required to inhibit the set means there are fewer resources available for the new task.

It is notable that attentional set switching leads to performance switch costs in a similar way to task-switching. This is seen in the study by Wendt et al. (2017) but it has also been found in experiments utilising the paradigm developed by Thompson and Crundall (2011). For instance, when participants were asked to detect hazards in driving stimuli, not only did the orientation of the preceding letter search influence eye movements to the driving stimuli, but response times were slower following a vertical letter search compared to following a horizontal letter search (Hills et al., 2018, Thompson and Crundall, 2011). A vertical letter search has additionally been found to negatively impact the ability to recognise upright faces (Hills, Mileva, Thompson, & Pake, 2017). It may be argued that this impairment to performance reflects an effect of fatigue rather than attentional shifting, on the basis that a vertical search is more difficult, takes longer, and thus leads to performance costs in a subsequent task. However, if that was the case it would also be expected that carry-over would be greater with increased repetition of the letter search task, yet studies have consistently shown that the persistence of attentional set to the second task does not vary according to whether participants complete one or three letter searches (Hills et al., 2017, Hills et al., 2018, Thompson and Crundall, 2011).

The influence of persisting attentional settings on eye movements has also been found by Longman, Lavric, and Monsell (2013). They presented participants with faces that each contained a letter on the forehead. In each trial participants were cued to identify the letter or the face and their results showed that on switch trials participants were more likely to fixate the previously-relevant region. They referred to this effect as “attentional inertia”. Longman, Lavric, and Monsell (2017) propose that attentional inertia is a component of task-set inertia but the two can be separated; attentional inertia being the persistence of where/what the attentional resources should be allocated towards, and task-set inertia being the persistence of the rules for how to respond to the attended information. They used a gaze-contingent paradigm in which participants moved their eyes from a fixation cross to a stimulus that appeared in different areas of the screen. The stimulus only appeared once participants moved their eyes from fixation and they were more likely to move towards areas that previously contained target stimuli, leading to slower response times to targets appearing in a different location to that of the previous trial, compared to when target location was repeated across trials. However, they also found that when participants were given an unlimited amount of time to prepare for a trial (i.e. they remained fixated until they were ready) there was no evidence of attentional inertia and participants did not re-fixate previously relevant areas, yet the performance switch costs still remained. It is therefore worth investigating attentional inertia as an effect independent to task-set inertia.

Within the task switching literature there is reference to switching involving activation of the new set and disengagement from and inhibition of the old set (e.g. Rogers & Monsell, 1995). Studies showing that carry-over increases due to investment in the initial set (e.g., Leber and Egeth, 2006, Thompson et al., 2015) reveal that difficulty in inhibiting an old set plays a specific role in the persistence of top-down settings, but so far it is unclear whether the activation of the new set has any influence on attentional inertia. There is some support for this influence, notably within the variation in findings regarding attentional inertia. In the studies completed by Thompson and Crundall (2011) the carry-over of search behaviour between the letter and picture search tasks only influenced vertical spread of search. Thompson et al. (2015) argued that this was because the images used in those experiments were road scenes which evoke a horizontal spread of search (e.g., Crundall and Underwood, 1998, Konstantopoulos et al., 2010) and the biases already present in the images would interact with the carry-over from the letter search. Specifically, due to the familiar context of road images observers would quickly extract the scene ‘gist’ (Friedman, 1979, Oliva, 2005). This would then activate an attentional set suitable for such a scene (allocating resources towards the horizontal axis due to the statistical likelihood of relevant objects occurring in this area) and this set would moderate the impact of the attentional settings from the preceding task.

Scene gist is the basic information that allows a scene to be categorised (e.g., Oliva, 2005; Oliva and Torralba, 2001, Wu et al., 2014), it can be extracted in less than 100 ms, and it has a key influence on the guidance of attention and eye movements in natural scenes (e.g. Torralba, Oliva, Castelhano, & Henderson, 2006). There is substantial evidence that the semantic context of a scene can bias attention (and visual search). Foulsham, Kingstone, and Underwood (2008) demonstrated that in natural, outdoor scenes participants had a tendency to make more horizontal eye movements and this bias towards the horizontal axis shifted as the images were rotated. This contrasted with indoor scenes for which there was little difference between the proportion of horizontal and vertical eye movements. These findings support the notion that different stimuli incorporate different top-down information, and this biases attention to certain locations and objects. This aligns with work showing that, on the basis of past experience, attention is prioritised to areas that are more likely to contain task-relevant information (e.g., Brockmole and Henderson, 2006, Hills et al., 2012, Shinoda et al., 2001).

The main aim of the present work was to test the influence of existing top-down semantic (contextual) signals on the activation of an attentional set. There is now a body of evidence showing that top-down attentional settings can persist from a task in which they are relevant to a task in which they are no longer relevant, and it has been argued that attentional set switching should be explored separately to task switching. Yet, unlike task switching, the mechanisms involved in attentional set switching are not fully understood, beyond knowing that switching is more effortful when more resources have been invested in the initial set (i.e., Leber and Egeth, 2006, Thompson et al., 2015) and that preparation can support set switching (Longman et al., 2017). Such findings emphasise the role of inhibition in successful set switching. Building on past work and taking advantage of findings showing that attention varies due to contextual information, the present study will explore whether, like task switching, attentional set switching is impacted by the activation of a new set, in addition to the inhibition of an old set. This will show the relative importance of activation and inhibition within attentional set switching and will also indicate whether some contexts are more at risk from attentional inertia than others.

The paradigm of Thompson and Crundall (2011) was utilised but the type of images used in the second task was manipulated. In every trial participants were presented with nine letters arranged horizontally, vertically, or randomly across the screen and were asked to search the letters to determine if there were three or four vowels presented. They completed either a single letter search or three letter searches (different letters but presented in the same ‘orientation’ within each trial) before being shown an image for 4000 ms. They were asked to view this image to make a judgement about how complex they found it and the images shown were road scenes, nature scenes, or fractal images. The use of fractals followed the work of Foulsham and Kingstone (2010) who found that participants made more horizontal eye movements on fractal images and made more vertical eye movements on nature images. However, whilst search in nature scenes varied according to the angle with which the scene was rotated, eye movements to fractals remained constant across all angles of rotation. This indicates that fractals do not contain semantic information to guide attention and so they are free of top-down influences (the consistent biasing of attention to the horizontal axis is instead representative of oculomotor behavioural biases, e.g. Tatler & Vincent, 2009). Given the argument that attentional set switching involves activation of the new set (in addition to inhibition of the old set) it was predicted that the attentional inertia effect would be smaller for images that contain more top-down information (i.e. road images) compared to images that contain no semantic information (fractals). This is because the contextual guidance in the road images would allow for quick activation of a new attentional set overriding any persisting influence from the previous task. The lack of semantic context in fractal images will mean that an attentional set is not activated, allowing for greater carry-over from the letter search task.

An additional aim of the current work was to further explore the duration of the carry-over effect by comparing eye movements in the first 1000 ms of the picture search task with those in the second 1000 ms. Thompson et al. (2015) found that whilst orientation of letter strings influenced initial eye movements to a picture search, this carry-over was short-lived and did not last beyond 1000 ms. The present experiment will provide further evidence for the time course of attentional inertia and will also demonstrate whether the duration of the effect is impacted by the top-down information in the second task. Three measures of carry-over were used, horizontal and vertical spread of search and saccade direction. Spread of search was calculated as the standard deviation of the horizontal and vertical positions of each fixation and saccade direction was the percentage of saccades made in an upwards, downwards, leftwards, or rightwards direction. The measures were compared for each level of orientation of the letter search and for the first 1000 ms and the second 1000 ms in the picture search. In line with previous findings, it was predicted that attentional inertia would be more apparent in the first 1000 ms of the picture search task. However, the effect would last beyond 1000 ms when participants are searching the fractal images due to the lack of contextual information to guide attention and override the previously relevant settings.

Section snippets

Design

The experiment was completed using a 3 × 3 × 2 within-participants design. The first independent variable was orientation of the stimuli presented in the letter search task (horizontal, vertical, and random). The second independent variable was the image type presented after the letter search (road scenes, nature scenes, and fractal images). The third independent variable was time. Eye movements in the picture search task were separated into two time epochs of the first 1000 ms and the second

Results

Analysis of the carry-over task was conducted on eye-movements to the picture search. Eye movements were recorded for the full 4000 ms that each image was presented, however analysis was restricted to eye-movements made in the first 2000 ms and these eye movements were separated into two epochs (first 1000 ms and second 1000 ms) to allow an investigation of the duration of the carry-over effect. Consistent with past research, the first full fixation and saccade made to the picture in each trial

Discussion

Research has shown that top-down settings which guide attentional resources in one task can persist to a second task, influencing performance (e.g., Leber and Egeth, 2006, Longman et al., 2013, Thompson et al., 2007) and the spatial allocation of attention in the second task (e.g., Hills et al., 2016, Longman et al., 2017, Thompson and Crundall, 2011, Thompson et al., 2015, Wendt et al., 2017). The current study explored the underlying mechanisms associated with this “attentional inertia”

CRediT authorship contribution statement

Catherine Thompson: Conceptualization, Methodology, Formal analysis, Writing - original draft, Writing - review & editing. Alessia Pasquini: Methodology, Investigation, Formal analysis. Peter J. Hills: Conceptualization, Writing - original draft.

References (41)

  • D.A. Allport et al.

    Shifting intentional set: Exploring the dynamic control of tasks

  • J.R. Brockmole et al.

    Using real-world scenes as contextual cues for search

    Visual Cognition

    (2006)
  • D.E. Crundall et al.

    Effects of experience and processing demands on visual information acquisition in drivers

    Ergonomics

    (1998)
  • I. Dombrowe et al.

    The costs of switching attentional sets

    Attention, Perception, & Psychophysics

    (2011)
  • H.E. Egeth et al.

    Searching for conjunctively defined targets

    Journal of Experimental Psychology: Human Perception and Performance

    (1984)
  • C.L. Folk et al.

    Involuntary covert orienting is contingent on attentional control settings

    Journal of Experimental Psychology: Human Perception and Performance

    (1992)
  • A. Friedman

    Framing pictures: The role of knowledge in automatized encoding and memory for gist

    Journal of Experimental Psychology: General

    (1979)
  • P.J. Hills et al.

    Carryover of scanning behaviour affects upright face recognition differently to inverted face recognition

    Visual Cognition

    (2017)
  • P.J. Hills et al.

    Abberant first fixations when looking at inverted faces in various poses: The results of the centre-of-gravity effect?

    British Journal of Psychology

    (2012)
  • B. Hommel et al.

    Theoretical issues in stimulus-response compatibility

    (1997)
  • Cited by (4)

    View full text