Modeling the role of salience in the allocation of overt visual attention

Parkhurst, Derrick; Law, Klinton; Niebur, Ernst

doi:10.1016/s0042-6989(01)00250-4

Cited by 1,204 publications

(1,179 citation statements)

References 54 publications

Supporting

Mentioning

1,075

Contrasting

Unclassified

Order By: Relevance

“…These findings are broadly in line with accounts that posit an important involvement of low-level guidance in scene processing (e.g., Borji et al, 2013a;Coco et al, 2013;Itti & Koch, 2000;Latif et al, 2014;Parkhurst et al, 2002;Pringle et al, 2011;Spotorno & Faure, 2011;Underwood & Foulsham, 2006). That said, clear differences in how perceptual and semantic factors interacted emerged across the two paradigms we used, highlighting the influence of the type of task, and of the task's cognitive load, on how observers utilise these objects' properties in order to guide attentional allocation and information representation.…”

Section: Semantic and Perceptual Prioritisation In Scene Processingsupporting

confidence: 90%

The elephant in the room: Inconsistency in scene viewing and representation.

Spotorno

Tatler

2017

Journal of Experimental Psychology: Human Perception and Perfor

View full text Add to dashboard Cite

We examined the extent to which semantic informativeness, consistency with expectations and perceptual salience contribute to object prioritisation in scene viewing and representation. In scene viewing (Experiments 1-2), semantic guidance overshadowed perceptual guidance in determining fixation order, with the greatest prioritisation for objects that were diagnostic of the scene's depicted event. Perceptual properties affected selection of consistent objects (regardless of their informativeness) but not of inconsistent objects. Semantic and perceptual properties also interacted in influencing foveal inspection, as inconsistent objects were fixated longer than low but not high salience diagnostic objects. While not studied in direct competition with each other (each studied in competition with diagnostic objects), we found that inconsistent objects were fixated earlier and for longer than consistent but marginally informative objects. In change detection (Experiment 3), perceptual guidance overshadowed semantic guidance, promoting detection of highly salient changes. A residual advantage for diagnosticity over inconsistency emerged only when selection prioritisation could not be based on low-level features. Overall these findings show that semantic inconsistency is not prioritised within a scene when competing with other relevant information that is essential to scene understanding and respects observers' expectations. Moreover, they reveal that the relative dominance of semantic or perceptual properties during selection depends on ongoing task requirements. Keywords:Semantic Consistency, Perceptual Salience, Scene Viewing, Change Detection, Eye Movements THE ELEPHANT IN THE ROOM 3 Public significance statement There has been long-standing debate about whether we look sooner at objects that are unexpected for the scene (i.e., semantically inconsistent objects). The present study shows that they are not prioritised over the most expected and informative objects for the event depicted in the scene (i.e., diagnostic objects), especially when these objects also stand out visually from their surroundings. Unexpected objects are looked at later, and when they are perceptually salient they slow down how quickly we select the most expected and informative objects in the scene.This study also shows that the impacts of object perceptual salience and object-scene semantic associations are task-dependent: semantics appear the most important source of guidance when viewers explore the scene for memorisation, whereas perceptual salience has a greater impact when changes have to be found. These findings provide new key insights into the roles of highand low-level factors in how we view and remember scenes. THE ELEPHANT IN THE ROOM 4The Elephant in the Room: Inconsistency in Scene Viewing and Representation Information selection during both online processing and memory representation is one of our fundamental and most striking abilities. Fundamental, because selection of a few, key aspects enables us to adapt and act fle...

show abstract

Section: Semantic and Perceptual Prioritisation In Scene Processingsupporting

confidence: 90%

The elephant in the room: Inconsistency in scene viewing and representation.

Spotorno

Tatler

2017

Journal of Experimental Psychology: Human Perception and Perfor

View full text Add to dashboard Cite

show abstract

“…Of course, while the method in our static scene experiment was not unusual for this topic (e.g., Foulsham, 2015;Foulsham & Underwood, 2008;Parkhurst et al, 2002), there were many differences between stimuli and procedure compared to the real world walking data.…”

Section: Discussionmentioning

confidence: 99%

Are fixations in static natural scenes a useful predictor of attention in the real world?

Foulsham

Kingstone

2017

Canadian Journal of Experimental Psychology / Revue canadienne

View full text Add to dashboard Cite

Word count (main body): 7275Unmasked Manuscript / Manuscrit non anonyme Click here to download Unmasked Manuscript / Manuscrit non anonyme StaticWalkView_Revision2.0.docx -2- AbstractResearch investigating scene perception normally proceeds in laboratory experiments using static images. Much has been learned about how observers look at pictures of the real world and the attentional mechanisms underlying this behaviour.However, the use of static, isolated pictures as a proxy for studying everyday attention in real environments has led to the criticism that such experiments are artificial. We report a new study that tests the extent to which the real world can be reduced to simpler laboratory stimuli. We recorded the gaze of participants walking on a university campus with a mobile eye tracker, and then showed static frames from this walk to new participants, in either a random or sequential order. The aim was to compare the gaze of participants walking in the real environment with fixations on pictures of the same scene.The data show that picture order affects inter-observer fixation consistency and changes looking patterns. Critically, while fixations on the static images overlapped significantly with the actual real-world eye movements, they did so no more than a model that assumed a general bias to the centre. Remarkably, a model that simply takes into account where the eyes are normally positioned in the head-independent of what is actually in the scene-does far better than any other model. These data reveal that viewing patterns to static scenes are a relatively poor proxy for predicting real world eye movement behaviour, while raising intriguing possibilities for how to best measure attention in everyday life.

show abstract

“…It is argued that in general, the initial saccade to an image is driven predominantly by 'bottom-up' process or external factors such as local image contrast (Parkhurst et al 2002;Peters et al 2005). From this perspective, the saliency of the eye region at the earliest stage of face inspection could be attributed to its local structure properties (i.e.…”

Section: Discussionmentioning

confidence: 99%

“…Internal factors, such as an individual's attentional state, expectation, experience and memory, are top-down and task-dependent (Noton and Stark 1971;Mannan et al 1997;Henderson 2003). It is argued that in general, the initial saccade to an image is driven predominantly by external factors (Parkhurst et al 2002;Peters et al 2005), but can also be biased by internal factors (Henderson 2003).…”

Section: Introductionmentioning

confidence: 99%

Initial fixation placement in face images is driven by top–down guidance

Guo

2007

Exp Brain Res

View full text Add to dashboard Cite

The eyes are often inspected first and for longer period during face exploration. To examine whether this saliency of the eye region at the early stage of face inspection is attributed to its local structure properties or to the knowledge of its essence in facial communication, in this study we investigated the pattern of eye movements produced by rhesus monkeys (Macaca mulatta) as they free viewed images of monkey faces.Eye positions were recorded accurately using implanted eye coils, while images of original faces, faces with scrambled eyes, and scrambled faces except for the eyes were presented on a computer screen. The eye region in the scrambled faces attracted the same proportion of viewing time and fixations as it did in the original faces, even the scrambled eyes attracted substantial proportion of viewing time and fixations. Furthermore, the monkeys often made the first saccade towards to the location of the eyes regardless of image content. Our results suggest that the initial fixation placement in faces is driven predominantly by 'top-down' or internal factors, such as the prior knowledge of the location of "eyes" within the context of a face.

show abstract

Modeling the role of salience in the allocation of overt visual attention

Cited by 1,204 publications

References 54 publications

The elephant in the room: Inconsistency in scene viewing and representation.

The elephant in the room: Inconsistency in scene viewing and representation.

Are fixations in static natural scenes a useful predictor of attention in the real world?

Initial fixation placement in face images is driven by top–down guidance

Contact Info

Product

Resources

About