Week 4 - Visual Attention Flashcards
(36 cards)
1
Q
Describe the retina
A
- Light rays must pass through several layers of cells before it reaches the photoreceptors
- There are three types of cones each responding most strongly to a specific wavelength:
- Any colour in visible range has a distinct response pattern for the three cone types
- Missing of one cone type causes colour vision defiency. An Ishihara test can detect this.
- Cones and rods have different distributions on the retina:
- More rods than cones
- No photoreceptors in the blind spot
- Fovea is full of cones only – best visual acuity
- Cones are colour sensitive but not light sensitive
- Rods are not colour sensitive but light sensitive. Only one type of rod.
- Retina is part of the central nervous system.
- Retina translates photons to neural language (translates light into brain signals)
2
Q
Describe processing at the retina level
A
- Receptive field – the part of a visual field in which a stimulus can modify the firing of a neuron.
- Rods and cones are in the receptive field of ganglion cells.
- Different types of ganglion cells have different receptive fields. The most basic receptive fields are:
- Excitatory-centre inhibitory surround – increase in firing rate if shone in the middle
- Inhibitory-centre, exhibitory surround – decrease in firing rate if shone in the middle
- There are also receptive fields for colour sensitive ganglions - complimentary colours
3
Q
How does convergence affect sensitivity and acuity?
A
- Convergence and sensitivity:
- Rods are more sensitive to light than cones because they have greater convergence – which is why we use rods in dim light conditions to detect faint stimuli.
- Rods have greater convergence which results in summation of the inputs of many rods into ganglion cells increasing the likelihood of response
- Convergence and acuity:
- Cones have greater acuity due to less convergence and smaller receptive fields
4
Q
What are the two types of bipolar cells? and how do they respond to glutamate?
A
- There are two types of bipolar cells – on and off:
- They respond differently to glutamate
- Glutamate – receptor released by the photoreceptors
- If glutamate hits an on bipolar cell then it will depolarise = increase in neurotransmitter
- If glutamate hits an off bipolar cell then it will hyperpolarise = decrease in neurotransmitter
- No action potential in photoreceptors/bipolar cells – only ganglion cells are the first place you can see an action potential. Rest have graded potentials.
5
Q
How does glutamate release change when a cone is in light or when a cone is in dark?
A
- Cone in the light:
- Cone hyperpolarises
- Less glutamate released:
- On bipolar cell depolarises = more transmitters released
- Off bipolar cell hyperpolarises = fewer transmitters released
- On ganglion cell = higher firing rate
- Off ganglion cell = lower firing rate
- Cone in the dark:
- Cone depolarises
- More glutamate released:
- On bipolar cell hyperpolarises = fewer transmitters released
- Off bipolar cell depolarises = more transmitters released
- On ganglion cell = lower firing rate
- Off ganglion cell = higher firing rate
6
Q
How do neighbouring cells affect the activity of a neuron?
A
- Horizontal cells allow lateral inhibition - the capacity of an excited neuron to reduce the activity of its neighbours
- Neighbouring cone in the dark area is depolarised:
- Connected to a horizontal cell that is activated due to the depolarisation
- Gaba is released
- This further decreases the amount of Glutamate released
- Process gets stronger
- Neighbouring cone in the light is hyperpolarised:
- Connected to a horizontal cell that is activated due to the hyperpolarisation
- Gaba is not released
- This further increased the amount of Glutamate released
- Process gets stronger
7
Q
Describe the M and P pathways in the retina
A
- Midget bipolar cells – small receptive field, mostly cones, and makes about 80% of the ganglion cells in the retina. Midget cells feed into parvocellular layer of LGN.
- Parasol bipolar cells – large receptive field, input from a lot of receptors, forward their information to the LGN. Parasol bipolar cell feeds into magnocellular layer of LGN.
8
Q
Describe the Lateral Geniculate Nucleus (LGN)
A
- Six different layers in the LGN. This is where information from both eyes is forwarded to.
- Parvocellular cells:
- Slow
- Precise
- Details, colour
- Magnocellular cells:
- Fast
- Imprecise
- Motion
- Input LGN:
- Receptive fields of LGN cell (P and M) identical to ganglion cell that “feeds“ it
- Input from one hemifield only
- Input from both eyes (but kept separately for contra- and ipsilateral eye)
- 80% of input to LGN comes from V1 – feedback connections
- Function of the LGN:
- Exact function unknown
- Relay center or switchboard?
- Simple computations
- Input from cortex (feedback?)
- Attention? Weighting of input?
- Control of vergence and focus of eyes
9
Q
Describe the visual cortex
A
- 30% of the cortex is visual cortex (vs. 3% hearing)
- Primary visual cortex - first cortical processing of vision:
- Striate cortex (V1) and extrastriate cortex (V2-V6)
- Striate cortex - stripes because of layer organisation
- V1 - 50% of area input from fovea although fovea is just 0.01% of retina area
- First binocular cells (input from both eyes)
10
Q
Describe Hubel and Wiesel’s 1968 experiment on cats
A
- Hubel and Wiesel, 1968 – single cell recordings with cats. Found orientation columns exist in the primary visual cortex.
- Orientation columns – respond best to specific orientation
- Ocular dominance columns - Neurons in primary visual cortex respond preferentially to one eye (left or right)
- Hypercolumn – a location column with all of its orientation columns
- Retinotopic organisation - nearby hypercolumns code visual input from nearby locations in visual field
- One layer in the column (layer 4) is very thick as it has different functions.
- Different LGN cells feed forward to different layers.
- Only the P cells feed forward to colour blocks
11
Q
What are the three different feature detector cells in the visual cortex?
A
- Simple cells
- Complex cells
- End-stopped (hypercomplex) cells
12
Q
Describe simple cells
A
- The simplest category of visual neurons in the cortex are called SIMPLE CELLS.
- Four of the receptive fields in the LGN forward to one receptive field in V1.
- As was the case with neurons in the retinal and the LGN, simple cells have well-defined on and off regions (accounts for their response to bars and edges) that can be mapped by the technique of shining small spots of light and noting any changes in the firing rate of the neuron.
- Receptive fields in V1 still have excitatory and inhibitory zones in receptive fields.
- In contrast to retinal and LGN cells which have circularly symmetric receptive fields, simple cells have RFs that are elongated along a particular orientation
- Position and orientation selectivity:
- As can be seen, this cell responds very poorly, or may even be inhibited, if the bar of light is placed at the wrong position or at the wrong orientation within the cells RF. Thus, we say that simple cells are selective for position and orientation of bars and edges.
- Decrease in firing as you change orientations – not a huge decrease in firing for a single degree change.
- Simple cells are predominant in hypercolumns.
13
Q
Describe complex cells
A
- Complex cells are the second category of visual neurons in the primary visual cortex.
- Complex cells do not have well-defined on and off regions in their RFs. These cells give little or no response to simple spots of light. However, they do respond to more complex stimuli.
- The RF of a complex cell is simply defined as the region of the retina in which some pattern of light is able to affect the firing rate of the cell.
- Found primarily in V1 and V2.
- Have larger receptive fields than simple cells
- Don‘t have distinct zones of excitation or inhibition
- Respond to stimuli of specific orientation regardless of location (within receptive field)
- Are particularly responsive to motion
- Position selectivity - NO
- Orientation selectivity - YES
14
Q
Describe end-stopped cells
A
- Cells increase their firing rate as bar length increases to fill up its receptive field
- Decrease their firing rate as the bar is lengthened further
- Are in the visual cortex outside of V1 (V2 and up)
- Have receptive and antagonistic receptive fields – if the line crosses the border of the receptive field of antagonistic complex cell, the firing rate of the neuron decreases.
- Are selective for certain orientation, motion, direction, length
- Get information from complex cells
- Hypercomplex cells can code corners, curvature, shape etc
15
Q
What are the two processing streams in the visual cortex?
A
- From V2 on – there are two main visual streams:
1. Dorsal “where path” - parietal lobe: - Input from magnocellular path (m cells)
- Location, direction (V2) & motion (V3&V5)
- Spatial & temporal frequency (V5)
- Ventral “what path” - temporal lobe:
- Input from parvocellular path (p cells)
- Edges, illusory edges (V2), angles (V4)
- Colour (V4)
- Curvature (V4), shapes (TEO)
16
Q
What are the processing steps in the visual cortex?
A
- The major principles of visual information processing:
- Retinotopic organisation - from retina, over LGN, V1 and higher cortex areas: what is near in space is near in representation
- Modularity - neurons that represent specific features of visual stimuli are organised in functional networks (e.g. orientation columns)
- Specialisation - complexity of stimuli that can trigger maximum response increases with processing step
- Complexity increase from V1 to V4
- Size of receptive fields increases, overlap more and are generally less spatially organised (less spatial resolution) with processing
17
Q
What is attention?
A
- Attention is the taking possession of the mind, in clear and vivid form, of one out of what seem several simultaneously possible objects or trains of thought. Focalisations, concentration of consciousness are of its essence. It implies withdrawal from some things in order to deal effectively with others - James, 1890
- Selection of relevant information and withdrawing irrelevant information.
- Attention is a brain mechanism to select information
- It is a bottleneck – we can’t process all the information available
- Attention is a filtering mechanism
18
Q
How do you measure attention? which paradigm?
A
- Hermann von Helmholtz‘s covert attention paradigm (ca. 1860):
- First experimental paradigm to test spatial attention
- Prior to presentation of letter table: covertly direct attention to location
- Very short presentation time from a spark
- Impression of only letters in attended region
- Covert spatial attention: no eye movements
Overt vs. Covert Attention Deployment:
- Overt attention deployment - changing the physical input to the retina by moving eyes
- Covert attention deployment:
- Maintaining the physical input to the retina as identical.
- Prioritizing a part of the visual field
19
Q
What is the neural evidence for cohort attention?
A
- Desimone and Duncan, 1985; Moran and Desimone, 1985:
- Receptive field of neuron sensitive to red
- Same physical input to retina, but response varies as function of which object is attended
20
Q
What is the spatial cueing paradigm?
A
- Posner (1980):
- Participants have to detect dots
- Press a button when there is a dot in any of the boxes
- Spatial cue is the arrow
- Results:
- Cue results in faster responses if it is valid (benefit)
- Cue results in slower responses if invalid (cost)
- Suggest - shift of attention towards cued location
- Enhanced processing of stimulus at attended location
- Only works if cue is valid most of the time → volutional control. Otherwise participants would ignore the arrow.
21
Q
What is the Attentional Spotlight?
A
- Derived from the Cueing paradigm (Posner, 1980)
- The spotlight moves through the visual field - things in the spotlight can be processed more efficiently
- Attention = limited resource (border of spotlight)
- Cues - deliberate or automatic movement of spotlight
- More than one spotlight?
- Attention can be diverted to two distinct regions (Awh & Pashler, 2000)
22
Q
Outline the Spatial Cueing Paradigm using peripheral cues and gaze cues (Posner, 1980)
A
- Peripheral cues:
- flashes in the periphery at same or different location
- Gaze cues:
- face in middle looking to one of the boxes
- Results:
- Both gaze and peripheral cues induce similar cueing effects to arrows (Frischen et al., 2007)
- However, they also work if the cue validity is low (e.g., cue indicating correct location in 50% only)
- May indicate an automatic process compared to the voluntary process displayed by the arrow
23
Q
What is top-down and bottom-up attention deployment?
A
- Top-down attention:
- Intention
- Goals
- Voluntary
- Bottom-up attention:
- Physical features
- Salience
- Automatic/reflexive
24
Q
What happens at the cued location when it is measured using an EEG signal?
A
- Heinze et al., (1990):
- Looked at centre of screen and told to attend left or right hemifield
- Target was presented in intended region or non-intended region
- Measured P1 and N1
- If target is presented at the cued location, there is a larger N1 and P1 component
- P1/N1 reflect early sensory processing:
- P1 and N1 enhanced for stimulus at attended location:
- → Resources dedicated to small region
- → Sensory gain at attended location
- → Better performance
25
Outline the Attentional Blink Paradigm (Raymond et al., 1992)
* Participants looked at stream of letters and digits all presented centrally for a short time
* Task - “Report both letters!“ (two targets: T1 and T2)
* Experimental manipulation - lag between T1 and T2
* In a critical time window (during the “attentional blink”) after T1, performance for T2 is worse
* Dip in performance only if T1 was attended (i.e., not a perceptual effect) – if participants were told to ignore T1 and report T2, there was no dip in performance. Suggests it is an attentional effect.
* Why is this happening?
* TMS showed dorsal stream (task relevant) is busy from T1; TMS impulse signal over IPS region makes dorsal stream recover more quickly
26
What is the visual search task?
* Visual search task:
* Predefined target (e.g., red T) has to be found among distractors
* Different types of tasks:
1. Detection task - is target present?
2. Localisation task - is target in left or right hemifield?
3. Discrimination task - report identity of target!
* Treisman and Gelade (1980):
* Single feature search – shows a pop-out effect. Increasing number of distractors does not affect search performance = parallel process
* Conjunction search - linear increase of response times with increasing number of distractors (shape and colour) = serial process
27
Outline the Feature Integration Theory (Treisman, Geladel 1980)
* If target does not share features with distractors: pop-out – no attention is required in feature search. An automatic process. A parallel process.
* If target shares features with distractors, attention is glue to bring features together (e.g., colour and shape) to build conjunctions. Serial deployment of attention.
* If we are engaging in a feature search, no attention is required as we don't need all the feature maps as we only need colour.
* Illusory conjunctions - if there is not enough time for attention to be deployed to location, features of two items may be combined falsely
* The perception of a red N and a green S may be conjoint to a green N
* Evidence for Feature Integration Theory:
* If you don’t have enough time to serially deploy attention to the relevant spots, you have free floating features that are not bound together.
28
What are the challenges to the Feature Integration Theory (Wolfe, Cave, & Franzel, 1989)?
* Participants have to find the green horizontal line
* Presenting a similar task with more stimuli is not seen as that much harder:
* Results:
* Slope - increase in RT with increasing set size
* Some subjects showed a very shallow slope
* Slopes were not linear → This cannot be explained with the Feature Integration Theory
* Doubling the set size should double the reaction time with the FIT
29
Outline Wolfe's own model of attention (1994; 2007)
* Guided Search:
* Attention is still directed serially, but not random
* Order of attention deployment is determined by summed salience
* No qualitative difference between feature and conjunction search – a continuum between the two
* Slopes vary from very flat to very steep due to signal-to-noise ratio
* Guided Search:
* When target is singleton, it pops out because it is more likely showing the highest activation level on attention map
30
What model was derived from Wolfe’s Guided Search model?
* Itti & Koch, 2001 – Computational Salience Model:
* Items from the visual field are represented on topographic maps
* Features like colour (feature map for each colour that are competing), orientation or shape are represented on distinct maps
* Attention is first deployed to the item with the highest activity
* Goals and intentions result in some feature maps having a higher weight
31
Where in the brain is the saliency or priority map?
* Not clear where exactly saliency and/or priority maps are represented, but likely large network with many feedforward and feedback connections
* LGN?
* Receives much feedback from cortical areas which could allow for weighting of features (Koch & Ullman, 1987)
* Primary visual cortex?
* Activity of V1 neurons varies with saliency (Li, 2002)
* Ventral stream/V4?
* Bottom-up and Top-down signals merge here (Mazer & Gallant, 2003)
* Oculomotor network?
* Eye movements are closely related to saliency/priority (Fecteau & Munoz, 2006)
32
How can we measure attention neutrally? – Contra/ipsi-lateral approach
Luck and Hillyard, 1994:
* Participants had to either look for a colour singleton or orientation singleton and say whether they were present or not
* Targets are presented laterally – this allows comparing contra and ipsilateral activity in EEG signal
* Results:
* Found that if you are instructed to find a red target (and the red target is there), you see a different between contra and ipsi.
* Items elicit an N2pc when they are task-relevant - cannot be a perceptual effect as N2pc only present when target is
* N2pc = negativity contralateral to the attented hemifield:
1. time range of the N2 (second negative peak)
2. posterior electrodes
3. contralateral to the attended hemifield N
Feldmann-Wustefeld et al., (2011) found that there is an N2pc for angry and happy faces but the N2pc for angry faces comes first --\> more and earlier attention deployment towards targets which may indicate that threat-relevant stimuli get attentional priority
33
Outline the Additional Singleton Paradigm (Theeuwes, 1992; 2010) - Suppression and Attention
* Find diamond and report line orientation:
* Some trials he would have an additional singleton (colour singleton)
* Results showed that this addition impaired response times – shows that there is some involuntary capture of attention by irrelevant information
* Top-down processing: Goal is to find diamond
* Bottom-up processing: Salient distractor captures attention → Coordination of top-down and bottom-up processing is crucial
34
How can we measure attentional capture and suppression?
This is hard to measure. For example, if we measure the N2pc and find there is a larger N2pc for targets, does this mean there is more attention towards target or less attention towards distractor?
Systematic Lateralisation Approach (Hickey et al., 2009) allows us to tell the different between the two:
* They presented the salient distractor on the vertical midline – only the target was presented laterally
* This allowed them to measure a ‘pure’ N2Pc because the salient distractor cannot affect the lateralised EEG signal
* We can switch the positions of the two items and now the signal should not be affected:
* Distractor-N2pc - because target is on vertical midline, it cannot affect the lateralised EEG signal!
* PD = measure of suppression of irrelevant items
* Results:
* Target lateral, distractor on vertical midline:
* Target-N2pc („pure“ N2pc): Reflects attention deployment towards the target
* Distractor lateral, target on vertical midline:
* First Distractor N2pc (reflects attention capture)
* PD (reflects suppression)
* Distractor N2pc is sometimes not present if the task is easy and attention isn’t necessary
35
Does suppression actually help us to focus on the targets?
* Gaspar & McDonald, 2014:
* participants had to find yellow target and ignore red target
* If distractor is closer to target, it interferes with it more – lateral inhibition
* Also found a clear PD component in the fast trials and less in slow trials – suggests only when they were efficiently suppressing the irrelevant information could they focus on the relevant information and be quick at responding
36
What is the overview of the different types of cues used in the spatial cueing task?
1. Exogenous cues:
* Peripheral sudden onsets
* Not affected by cue validity
* Automatic
* Short latency (~50 ms) but transient (~200 ms) cueing effect
1. Endogenous cues:
* Central ”interpreted“ cues (e.g., arrows)
* Only work for high cue validity
* Under voluntary control
* Longer latency (~200 ms) but more sustained (\>500 ms) cueing effect
1. Gaze cues:
* Central cues
* Not affected by cue validity
* Automatic
* Medium latency of cueing effect
