Flashcards in Neural mechanisms of attention Deck (21)
top-down v bottom-up attention
Items in the world compete for our attention
Perceptual properties of items determine the strength of competition (bottom up – stimulus driven)
- More similar items - lots of competition – weaker items can lose out
- Less similar items – less competition – easier to pick out relevant items
- Attention can be driven by bottom up factors (stimulus driven)
- But is bottom-up competition enough?
In most situations requiring goal-driven behaviour, bottom-up (stimulus driven) factors are not sufficient to resolve attentional competition
Usually, we have some current goals that define which objects should be preferentially attended to
An attentional ‘template’ is required
The concept of an attentional template forms the cornerstone of the biased competition model.
It essentially says that the attention system needs a template – something to guide it and prioritise processing of task relevant objects in the presence of bottom up competition.
factors determining ease of access into consciousness
These factors are not properties of the stimuli themselves but are defined by the observer’s relationship with the stimulus (‘Top Down’ attention)
neuronal ev for attentional comp in temp cortex
dividual neurons in monkey temporal cortex show ‘preferences’ for particular stimuli
It turns out that individual neurons in monkey temporal cortex show preferences, that is that they respond selectively to a particular type of stimulus.
For example, some neurons might prefer squares, others might prefer triangles.
Here is an example of what the response of such a triangle-preferring neuron might look like.
Chelazzi et al. (1993)
Monkeys made eye movements to a target
- Either a neuron’s preferred stimulus or non-preferred stimulus
They found (top left graph) that the individual neuron’s response initially responded equally strongly regardless of whether the preferred stimulus was the target or the nonpreferred stimulus was the target.
However, about 180 ms after onset of the choice array (to which the monkey has to respond), if the preferred stimulus is the target, the neuron’s firing remains high.
But if the nonpreferred stimulus is the target, the response of the neuron is suppressed.
This clearly shows that neuronal responses in inferior temporal cortex are competitive – if the neuron’s preferred item is the target, then the response remains high and the item wins the competition for attention.
However, if the non-preferred item is the target, neuronal firing drops off.
On the right you can see that the responses of the neuron diverge before the eye movement is made, again demonstrating that visual attention operates independently of the eyes and before a response is carried out
The results suggest that the attentional template is formed by modulation of brain regions that process the relevant object – in this case enhanced neuronal firing in shape-selective cortex.
what does this ev tell us about attention?
Individual neurons show competitive interactions during attentional selection
Competition occurs not in a separate ‘attentional’ brain region but in the brain regions that process the visual features of relevant (and irrelevant) objects
The same neurons that process the visual features of an object (its shape, colour, orientation etc.) are co-opted by the attention system to resolve the competition for selection
imaging ev for attentional selection in early visual cortex
e.g. Event Related Potential (ERP) studies of the Posner cueing effect
ERP ev for early selection - Heinze et al. 1990; Mangun et al. 1993, 1995; Anllo-Vento 1995; Eimer 1997
The earliest studies to look at this using neuroimaging used EEG (ERPs) – this was a suitable method because of its high temporal resolution – you can see what’s happening both early and late in each trial.
Several studies using this methodology confirmed that attention can operate at the earliest levels.
They used ERPs to show that when a target is validly cued, it leads to a greater response in early visual areas than invalidly cued targets.
Note the very early response (P1 is less than 100 ms, N1 just over 100 ms) – this is before subjects make a response, supporting the idea this is an attentional effect rather than a motor response effect.
Note also that these effects occurred over posterior visual cortical areas.
However, using ERPs, we can never be sure of the source of the signals.
ERPs have great termporal resolution, but poor spatial resolution.
The results suggest that when a target appears, if a cue was previously presented pointing to that location, brain activation in early visual cortical areas is higher than when the cue pointed to the other location.
It is almost as if early visual areas that process information in specific locations are ‘primed’ by the cue such that when the target occurs, the activation is higher if the cue and target location are congruent (the same).
neuronal ev for attentional enhancement and suppression
Remember inhibition of return – this is the finding that RTs to cued stimuli are actually slower when there is a longer time between the cue and the target.
It turns out that this pattern occurs also in patterns of brain activation – here measured by EEG.
This study presented targets at either cued or uncued locations.
As you can see, activations in visual cortex were higher initially to cued targets but were subsequently higher to uncued targets.
However, conclusions about the spatial source of EEG signals are limited due to the poor spatial resolution so researchers turned to fMRI to examine precisely where in the brain these effects occurred.
The results suggest that when a target is validly cued, brain activation in early visual cortical areas is higher than when the cue and target occurred in different locations.
Pattern is reversed for inhibition of return
As if early visual areas that process information in specific locations are ‘primed’ by the cue such that when the target occurs, neuronal firing is higher if the cue and target location are congruent (the same).
However, conclusions about the spatial source of EEG signals are limited due to the poor spatial resolution
fMRI ev for comp interactions in V4 - Kastner et al. (1998)
Activation in V4 (colour sensitive visual cortex) lower when items presented simultaneously than sequentially (bottom-up comp)
Tested when stim compete for
When items presented sequentially, responses in visual cortex higher than when simultaneously
When instructed to attend to one of objects, effect disappeared – response as equally high as when presented alone
Demo’s that attention modulates comp interactions at neuronal level in visual cortex
Comp interactions in visual and temporal cortex play role in attentional selection
what has imaging told us about early v late attentional selection
Attention can modulate activation later in processing stream – in V4
attention can also modulate activation in higher level visual cortex - ev from fMRI - Wojciulik et al. (1998)
Is it the object/spatial location that is selected?
Relied on finding that faces and houses activate separate regions of inferior temporal cortex
Perf matching task on 2 faces/houses
Activation in FFA higher when perf task on faces than houses, even though houses in spatial field of vision
Attention can operate on higher level visual processing regions by modulating activation to whole objects
fMRI ev for object-based attention - O'Craven et al. (1999)
Presented subject with overlapping faces and houses
Subjects attended to either moving object, static object/direction of motion
Activation in FFA and PPA depended on which stim attended to
E.g. if attention directed to moving object, activation in FFA higher when face moved
FFA prefers faces and PPA houses
imaging studies shown attentional modulation of neuronal signals throughout the visual processing stream (early and late)
Comp interactions at latest stages of visual processing stream
biased comp model fo attention (Duncan and Desimore)
Many brain areas activated by visual input, and within most of systems activations for diff objects compete – at any level of brain
Comp occurs at level of object properties and whole object – monkey neurophys and human neuroimaging studies demo attentional modulation of neuronal signals throughout visual cortex (ventral visual stream – ‘what’ pathway`)
May be result of attentional signals arriving from elsewhere – what is source of attentional baising signals?
ev that source of attentional signals is frontoparietal cortex - Hopfinger et al. (2000)
Event related fMRI to look at activation time-locked to cue, not target
Dorsal frontoparietal network
Presented with directional cue and asked to make judgement on checkerboards (any grey checks?) but only when appeared in cued location
Found activation across frontal and parietal cortex time-locked to cue – regions activated in prep for upcoming target stim
Could be source of top-down biasing signals
activation time-locked to targets more posterior, in parietal and occipital cortex
Activation to targets more in posterior regions such as parietal and occipital cortices
Shows value of event-related fMRI in being able to differentiate activation to diff timepoints within single trial
how can we be sure frontal regions such as FEF are source of top-down control signals?
1. Activation in frontal regions time-locked to cue
2. When subjects passively attended to cue, frontal activation not observed
fMRI provides correlation data, and can’t infer causation
EEG ev for top-down bias signals originating in front-parietal cortex - effects if TMS over FEF on activation in visual cortex - Taylor et al. (2006)
Causal role of prefrontal cortex activation on attentional signals in visual cortex
Perf Posner cueing task in which central cue pointed left/right and target than appeared on left/right
Used TMS to stimulate FEFs between cue and target
o Responses slower during TMS
o Top = back of brain
- Faster to respond to validly cues targets than invalid
o Normal attention-related neg when no TMS occurs reduced in TMS condition
o Signals over prefrontal cortex have causal effect on signals in visual cortex during attention
but what is the specific neuronal mechanism used to prioritise processing of task-relevant info?
Increased firing rate to attended stim together with reduced firing rate to unattended stim?
But firing rates to single high-contrast stim un receptive field often not increased with attention
Alternative is neuronal oscillation (rhythmic/repetitive patterns of neuronal activity)
o Neuronal signals oscillate at particular freqs
o Diff freqs might correspond to diff functions
o Freq synchronisation between regions might support selective attention – can arise from simple conditions
the 5 freq bands of EEG signals