Attention Flashcards
What is covert attention
In the current lectures,
I shall not discuss overt attention in which eye or head movements enhance processing of one object
or region. Instead, we shall focus on covert attention, which can arise even when we don’t move our
eyes. In everyday life overt and covert attention are usually select the same objects and regions.
However, we have known since observations of Hermann von Helmholtz in the 1890’s we can
covertly select some objects and ignore others with our attention independently of eye movements.
Which oldest modern research shows attention as a filter？(Cherry 1953）
Modern research on selective attention was only revived following Cherry’s (1953)
observations concerning selective listening in shadowing experiments. Two messages were presented
simultaneously (e.g., by earphones to the two ears: dichotic listening) and participants were instructed
to continuously repeat back one of the messages heard. When one message was presented to one ear
and one to the other, participants had little difficulty in performing this task as they could attend to
one of the ears’ inputs. Little was remembered of the ignored message though the physical
characteristics of the unattended message (pitch, loudness, location) were often noticed.
Filter Attenuation model of Attention
Treisman (1960) found that participants in shadowing
experiments tended to switch channels whenever the attended side was continued briefly on the
unattended side. Cherry’s (1953) finding that a subject’s own name would sometimes be noticed even
when presented in the ignored stream – also suggesting that some quite sophisticated processing must
arise even for ignored information. This led Treisman to propose a Filter Attenuation model of
Attention in which ignored information was attenuated rather than entirely suppressed. Other early
work, for example, Corteen and Wood 1972, also found that words previously associated with
electrical shock gave rise to galvanic skin responses even when presented to the ignored ear
Explain the Spotlight effect (Endogenous cues)
One influential concept of
visual attention was proposed by Posner (1980, JEP: Gen., Vol. 109) to account for findings from his
studies of attention-cueing. In a typical cueing paradigm, cues (e.g., central arrows) are presented
prior to the appearance of a target that could appear to the left or right of a central fixation point. The
cue summons the subject’s attention to either the left or right location. When, on some trials the target
appeared at the location to which attention had been cued (validly-cued trials) participants detected
the target and responded to it more quickly than when neither position was preferentially cued
(neutral trials). Conversely, on trials where the target appeared on the opposite side to the location that was cued (invalidly-cued trials), participants detected the target and responded to it more slowly than
when neither position was preferentially cued (neutral trials). The movement of attention from its
initial central location to the location of the cue was characterized by Posner (and other authors
subsequently) as the movement of a “spotlight” of attention across the visual field. Stimuli falling
within the spotlight were assumed to reach awareness more rapidly and hence responded to more quickly than those outside the spotlight.
Explain the Spotlight effect (Exogenous cues)
Posner studied two types of spatial cueing. Endogenous orienting is elicited by centrally- presented
symbolic cues (e.g., ►). These cues reliably cue attention when they are spatially informative – that
is, when they predict the likely position that the target will appear in. They probably typically reflect
voluntary shifts of attention. However, some types of cue can cue attention even when they hold no
information about the likely position of the target. This type of cueing, associated in particular with
peripheral onsets, elicits exogenous orienting that is fast and reflexive (involuntary). Exogenous cues
can often summon attention to a location even when they are counter-predictive – that is where a cue
on one side will predict that the target is most likely to appear on the other side. Eriksen & Eriksen
(1974) concluded that the minimum size that human observers could set their hypothetical attention
‘spotlight to was 1 degree retinal angle. This conclusion reflected their finding that interfering effects
of task-irrelevant flankers were minimised when they fell more than 0.5 deg retinal angle away from
the target.
Explain the flanker effect
Respond to target letter in the middle
Press left key if H or K, right key if S or C
ignore the flankers!
discuss the effects of attention on
responses of visual cortex to stimuli
(1) single-cell recordings in non-human
primates, which showed that the responses of cells in primate visual cortex are dependent upon the
animal’s attention (see e.g., Moran and Desimone, 1985)
(2)effects are evident in the human
brain, too; this is illustrated in the lecture with regard to a study by Rees et al. (1997) looking at the
effects of attention on human V5/MT coding of motion.
Disorders of attention will disrupt visual perception
(1)– UNILATERAL NEGLECT
Inattention to side of
space opposite lesion
(more severe following
right sided lesions)
¡ Ignore food on left of plate
¡ Groom only right side
¡ Damage to Posterior
Parietal Cortex..
(2)-BALINT’S SYNDROME
Balint, 1909
Parietal lesions to both hemispheres
Fixity of gaze and simultanagnosia
Even when two objects are presented at fixation…
¡ Only one is seen – two cannot be seen simultaneously
¡ Problems shifting attention
(3)-UNILATERAL EXTINCTION
Can detect object or event on either side
¡ Problem arises when LHS and RHS stimuli are both presented
¡ LHS missed - perception of RHS stimulus ‘extinguishes’ perception of LHS
What does attention do?
(1)cueing attention to the
location of a subsequent target could speed up responses to that target, providing an object
performance measure of attention shifting around a computer display.
(2)attention affects perception itself besides simple influencing the preparation of responses–
(3)to combine representations of elementary perceptual features into multi-feature objects.
In Conjunction Search, features have to be combined in order to detect a target, requiring the serial adjustment of the attentional spotlight to potential
target locations.
Feature Integration Theory (FIT) assumes that attention provides the ‘glue’ that combines separate features in an integrated, multidimensional percept.
Treisman and Gelade (1980) measured RT in
search tasks where the number of distractors was varied and found that for Feature Search,
RT is largely unaffected by display size (at least for ‘target present’ responses), whereas for
Conjunction Search, RTs increased linearly with display size, with a 2:1 slope ratio for targetabsent and target-present trials. This pattern suggests that visual search is parallel for feature search, but serial and self-terminating for conjunction search.
FIT (see Treisman, 1988) explains these results within a stage model of visual information
processing. Initially, single features within a display are analysed and coded separately
within specialized feature modules. Although spatial information may be preserved in these
modules, only the presence (but not the location) of specific features is signalled to
subsequent stages. Spatial position is represented in a master map of locations. Selective
attention operates on the location master map by selecting a specific location (and can thus
be described as a ‘spotlight’ moving in visual space). When the attentional spotlight is
applied to a location, the features that are present within this area (as detected within the
feature maps) are selected simultaneously, and entered into an object file where the
different stimulus features are combined. Identification takes place when the object files
are compared to stored descriptions within a recognition network. FIT assumes that the
attentional spotlight is applied serially to different locations. In Feature Search, targets can
be detected on the basis of the parallel coding of single features in feature modules, thus
search can proceed in parallel. In Conjunction Search, features have to be combined in order
to detect a target, requiring the serial adjustment of the attentional spotlight to potential
target locations.
What does attention select?
(1)pioneering studies of
attention research had used spotlight and zoom-lens metaphors for visual attention,
suggesting that attention selects a circular region of space, irrespective of the objects
located at that region.
(2)subsequent studies have demonstrated that attention is affected by object perception.
- I. In one example using neurologically- intact participants, Egly et al., 1984 measured responses to targets following a cue located at one end of an object, comparing responses to targets presented at the
other end of the same object as the cue versus to targets appearing equally far away from
the cue but on a different object.
Spotlight and Zoom-Lens metaphors for attention would predict equal performance in these two cases as the two types of target were equidistant
from the cue. Conversely, if attention selects a whole object once it is cued to part of that
object, responses to targets appearing on the same object as the cue should be faster than
those to targets on another object.
This latter pattern of results was found supporting the
notion that attention is ‘object-based’, selecting objects rather than circular regions of space
- II. The flankers’ effect-Attention selects objects
How much can attention select?
Reviewing performance in
Visual Short-Term Memory tasks,
in
attentional Multiple Object Tracking
and in
rapid enumeration experiments,
we see that 3 to 4 simple items may be processed
efficiently and rapidly in parallel but that performance quickly deteriorates for more than
this number of objects. I suggest, this different findings may converge to indicate the
effective capacity of attention (see Cowan, 2001).
When does attention select?
Early selection views are consistent with
the assumption that a primary role of attention is to prevent perceptual systems becoming
overloaded. They are supported by FMRI and single-unit recordings in non-human support primates which show effects of attention throughout early stages of cortical processing in vision.
These theories draw support from the face that attention seems to select items more efficiently on the basis of relatively simple features coded at early stages of sensory processing than on the basis of more sophisticated characteristics (such as semantic properties or familiarity). After selection, any further processing of unselected inputs is, on these views, greatly attenuated.
(2)In contrast to these views, ‘late-selection’ views of attention typically hold that all items in a scene are processed to very high-levels and one object is then selected for the purpose of controlling action. If human attention were always to select at an early stage of vision, this would present a key limitation. How can attention be guided to the most important information in a natural scene if only very basic features have been processed? The brain would only be able to guide attention randomly around a scene, or to the most salient items. If, instead, attention were to select after much more parallel processing of all objects in a scene, more sophisticated features might be used to guide attention. I review some early work that shows at least four items can be processed in parallel as efficiently as just one or two. However, only one of these items can then be efficiently selected to control action (Duncan, 1980). While these findings do not offer unequivocal support for the original claim of late selection theories that all items undergo high-level processing, they do suggest that attention can select at multiple stages of perception and for multiple reasons.
basic role that attention plays in
control of action
a key reason for attention to select some information, and reject other information, was in order to
ensure that only task-relevant features of our perceptual environments control our behaviour.
For example, in reaching tasks, if attention does not adequately suppress processing of salient irrelevant objects, reach trajectories and saccades will veer toward irrelevant objects during execution. I also present examples in which attention is found to be closely coupled to action, especially the targets of saccadic eye movements. The possibility is raised that, as Rizzolatti’s Premotor Theory maintains, covert attention is a function of saccade or reach programming. I then briefly review evidence that contradicts this assumption, before summarising the lecture’s main points.
