Perceptual Organisation Flashcards
Describe the binding problem
Perceptual organisation- combining distant elements and features into a coherent ‘whole;, for example how do we separate objects and background
how can we make sense of visual information even when there is minimal input? (dog)
What does the Dalmatian dog “illusion” (?) demonstrate about perceptual organisation
People will always notice this Dalmatian when they see it again
Once an organised percept, always an organised percept: effect of experience
What is the relevance of biological motion?
We have high level perceptual information regarding movement with minimal information
Also the effect of ‘knowledge’ or memory: we are very often exposed to moving bodies
What two psychological mechanisms of perceptual organisation are mentioned?
Gestalt laws
Texture segregation
What four neural mechanisms of perceptual organisation are mentioned?
- selective deficits
- grandmother cells vs assembly coding
- Anatomical connections
- Contextual modulation
What art style emphasises this perceptual organisation?
pointillism- objects are defined by the grouping of individual dots of colour
Why do these dots for objects?
Name the four ‘classic’ Gestalt laws
1) Proximity: parts are grouped when close together
2) Similarity: parts are grouped when similar (e.g colour, contrast, illuminance, shape)
3) Connectedness: parts are grouped when connected
4) Common fate: when objects move together
Name a criticism of the law of 1) connectedness and 2) common fate
1) Isn’t connectedness just proximity?
2) isn’t common fate just similarity?
Name something which demonstrates the gestalt law of common fate
Random dot kinematograms use similarity of motion to define objects by motion alone
What is meant by the Gestalt principle of closure?
Convex (closed) shapes are preferred over concave
e.g ( ) ( ) ( ) ( )
What other Gestalt law does the previous example go against? What does this demonstrate?
proximity: the convex brackets are further than the concave brackets
This demonstrates competition between Gestalt laws, in this situation closure ‘wins’ over proximity
What is meant by the Gestalt law of collinearity?
also known as good continuation.
We tend to group elements together that are in the same direction/ orientation
e.g > < vs >
An example was shown in the lecture where white objects were shown against black and the same black objects against white. In both scenarios the same objects were perceived but in different colours.
What allowed these shapes to have been seen as the ‘objects’ rather than the background?
The gestalt law of symmetry: contours that form a symmetrical shape are grouped over contours that form an asymmetrical shape
What is meant by the gestalt law of pregnanz?
Also known as the law of good forms (pregnanz is german)
Preference for or reduction to the simplest forms
(olympic rings)
What is a criticism of the gestalt law of pregnanz?
It is v similar to good continuation and closure
What is meant by the Gestalt law of figure-ground?
Some region is figure, other is background that ‘continues’ behind figure
What is meant by texture segregation?
Regarding gestalt principle of similarity, wanted to see how much features could differ before they were seen as separate objects
name some ways in which elements could differ in order for it to be perceived differently
orientation, colour, size
sometimes features differ but there is no immediately obvious segregation, what is this type of segregation referred to as?
pre attentive segregation
What is useful about these ‘not segregating texture’?
They point out what features are and are not useful for segregation
Would these symbols from the lecture be segregating textures?
⊿ ⇲
Why or why not
No, Both elements consist of the same orientations of line segments, only different in their relative position (‘relative phase’ in Fourier terms) > orientation matters, phase does not
the lecture then displayed what appeared to be standing and lying 10s and 2s and five made up of lines. What rule of segregation did this demonstrate?
Both sets of elements (standing and lying 10’s vs 2&5’s) consist of the same orientations of line segments, only different in their relative position (relative phase in Fourier terms) > orientation group statistics matter
It looks at the horizontal vs vertical ‘energy’ (consistency, frequency(?))
Through experimentation it has been found that what 6 features matter for segregation in regards to similarity?
- Orientation
- Direction of Motion
- Disparity (Depth)
- Spatial Frequency (Size)
- Luminance
- Color
What disorder exists regarding this perceptual organisation?
Apperceptive/ Integrative agnosia
Patients see individual features, such as color, orientation or motion, but cannot integrate this into a whole, failure of perceptual organisation (can know what an object is through touching it but not through sight)
What damage in the brain is responsible for apperceptive agnosia?
Right Hemisphere occipital / temporal / parietal lesions usually quite diffuse (e.g. CO damage). Not one area but likely communication between visual areas
What five tasks can demonstrate whether someone has apperceptive agnosia?
Degraded letter task, Gollin picture task (perceptual organisation images like degraded letters and incomplete drawings)
Copying is also bad
Unusual views/ shading task (large shadows) is also bad
What usually causes this type of brain damage?
Carbon monoxide poisoning because it can cause oxygen shortening. The posterior of your brain is more vulnerable to oxygen shortage than the rest of the brain
What is the difference between apperceptive and integrative agnosia?
Integrative is more mild. Patients can copy but piece by piece. Unusual views test can vary.
What task is often suitable for integrative agnosia
Overlapping objects are not seen individually
Can patients with apperceptive and integrative agnosia recognise faces?
Yuh!
What is meant by feedforward processing?
The information goes from lower regions to higher regions.
V1 is the starting point of projection to the other visual areas and the rest of the brain.
From there, activity spreads to the other visual areas.
at what speed does this feedforward sweep occur?
Speeds through the visual system within 80 ms by means of feedforward connections from
What information does this fast feedforward sweep provide? (3)
Provides the neurons with their receptive field tuning properties
•Provides fast detection of ‘hardwired’ features and feature constellations
•In that sense provides a set of grouping operations
What unconscious process does this fast feedforward sweep then allow outside of the visual cortex?
This then enables ‘intelligent’ sensorimotor reflexes
How often is there a “spike” and why is this significant regarding the fast feedforward sweep?
There’s only one spike per 10ms. The V1 cortex is activated around 40ms, other areas such as 7a, V3, MST etc is activated around 50ms and a range of other areas such as V2, V4 and even prefrontal cortex are activated around 65ms.
This means these areas are reached after only around 2/3 spikes.
What trends are generally seen in the cells further on in the feedforward convergence? (2)
During this feedforward convergence, cells get increasingly larger receptive fields, and get progressively more complex tuning properties
Why may neurons in V4 and other areas have larger receptive fields and more complex characteristics?
There are horizontal and feedback connections
Therefore ________ are mediated by _________
RF tuning properties are mediated by the feedforward sweep
How does this feedforward sweep seem to function in apperceptive and integrative agnosia patients?
In the apperceptive / integrative agnosia patient, this ‘feedforward sweep’ seems to function normally: he sees features like orientation, color, motion, and even faces. He just can’t make sense of the combination of these.
How can these patients make sense of faces when they are made up of seemingly indefinite shapes and orientations?
apparently, to the brain, a ‘face’ is just as primary a feature as orientation, colour or motion is. Its detection is ‘hardwired’ into the feedforward convergence of visual information processing
How would the fastforward sweep help you if someone threw a punch at you
The fast feedforward sweep enables the rapid extraction of ‘hardwired’ features and feature constellations, enabling swift visually guided action.
It forms a sort of ‘intelligent reflex arch’, enabling the execution of well trained, automatic, visually driven motor programs
This would mean that you might automatically go to dodge or block the punch.
What also occurs in the ventral stream
Also in the ventral stream, fast recognition of important feature constellation occurs, enabling rapid extraction of (e.g. emotionally) relevant information
Describe two paths the signals may go from the area V1 when translating visual input to action. How long does this take to occur approximately?
V1 > MT > Par > Mot
V1> V2 > V4 > TE > Mot
Mot > Action
Full sensory to motor cascade within ~ 120 ms
How can these horizontals connections be observed?
Anterograde neuro-anatomical tracers (in this case Pha-L) are injected in a specific site, are taken up by cells, and via axonal transport reach all their axon terminals
To what extent are cells connected to nearby cells,
Cells are mainly connected with nearby cells (short range horizontal connections) but also - more selectively- with more distant ones.(long range horizontal connections)
What are meant by boutons?
Boutons indicate synaptic contact
What did the tracing of horizontal connections combined with the visualisation of orientation columns (optical imaging) reveal
It turned out that long range horizontal fibers in V1 (layers 2/3, 5) selectively interconnect cells with the same orientation preference
What gestalt law does this finding regarding orientations relate to?
This seems highly relevant for the observation that elements with the same orientation tend to group together perceptually (grouping by similarity)
Horizontal connections are the neural ‘substrate’ for the Gestalt law of grouping by similarity
What further finding was there regarding these horizontal connections between orientation cells?
Further studies revealed that the horizontal connections in V1 tend to spread along a ‘line’
This line coincides with the orientation preference of the injection site, so that the horizontal connections seem to connect cells with the same orientation and lying along a line in visual space
What other gestalt law does this finding apply to?
This type or collinear spread seems to be the substrate for the Gestalt law of ‘good continuation’ or ‘collinearity’, studied in contour grouping (also proximity)
What detail given in the gestalt law of collinearity also coincided with findings regarding the orientation cells?
That this grouping perception is less effective when greater than around 30 degrees
These connections existing are one thing, but how are these expressed in the activity of the cells?
Local gestalt grouping: V1 cells respond more strongly when orientations in the surround
Where have similar findings to this been found and what do they consist of
Similar findings have been found in area MT, where most neurons are direction selective. Here, patchy horizontal connections connect cells with similar direction preference
What gestalt law do the findings in area MT therefore correlate to
Gestalt law of ‘common fate’
What is the role of feedback signals in V1?
Contextual modulation of responses; integrating information from the various cells to establish grouping perception, background and figure etc
Describe in relation to time the information that is processed in area V1 from when the information enters the brain
40- 100ms: isolated information, cells only look at what happens within their RF (which is identical)
100 + ms: integrated information, V1 activity is modulated by global visual context
Where can you look to see the modulation which reflects figure ground segregation?
The google doc of digital diagrams cuh, on the gang bruh
How is it shown that contextual modulation reflects figure-ground organisation?
recording responses and having the RF go over a small square box against a background of another shade.
The responses will be stronger as the RF goes over the object, particularly the edges
Describe what happens in the feedforward process at 50 ms, 70 ms and 100ms
50ms- Orientations processes by V1 (Low level feature detectors e.g orientation or motion, depth, colour etc)
70ms- Orientation info reaches extrastriate areas (High level feature detectors e.g convex shapes, squares etc)
100ms- Recurrent processing yields figure-ground segregation (through feedback into posterior areas) (low and high level feature information combines to form integrated percept of square made of orientated line segment)
What proof is there for this feedback mediating V1 integration ?
Extrastriate area lesions abolish contextual modulation
In regards to the processing mechanisms in the visual hierarchy what functions do a) feedforward connections b) Horizontal connections and c) feedback connections carry out?
Feedforward connections- Carrying orientation information
Horizontal information- Lateral inhibition between neurons tuned to the same orientation. Present at every level in the hierarchey
Feedback connections- These connections do a retinotopic feedback,
The resulting activity in V1 therefore looks like this (kinda)
__/’—-‘\___
Explain in relation to these connections hhow this activity is recorded
The lateral inhibition between neurons tuned to the same orientation in horizontal connections creates the bumps. The retinotopic feedback of the feedback connections progressively fills in the region between the bumps.
There were then studies which examined whether you could selectively inactivation these feedback connections while leaving the feedforward connections intact.
Describe these and the subsequent results
It proved difficult but it was found that these feedback and recurrent processing connections used a different neurotransmitter. When the NMDA receptor blocker, APV was applied seemed to restrict these processes while maintaining the feedforward connections.
How were bottom up visual signals reduced?
By AMPA receptor blocker CNQX, but RP is not
How are these findings relevant to conducting visual neuroscientific research on animals?
The NMDA receptors are much more sensitive to anaesthesia than AMPA receptors. Therefore you selectively lose the feedforward connections
To what extent do the combinations of different feature contrasts combine to add to contextual modulation?
RECURRENT SIGNALS IN V1 REFLECT PERCEPTUAL ORGANIZATION OUTCOME Contextual modulation is present for every possible feature contrast, combination does not add
sorry for shouting
How were these findings reflected in human studies?
(obviously couldn’t conduct the same experiments but…)
It was shown through either applying or not applying TMS at 100ms, 160ms or 240ms that the perception of Kanizsa squares ( four 3/4 circles depicting a square) requires feedback from LO to V1.
It was found that if you applied it at 100ms in the higher level LO people would lose the percept to an extent. This feedback then goes back to the earlier areas about 60 ms later and if you disrupt V1/ V2 then, you also lose the percept. If it was a constant feedforward movement, then you would expect the opposite result.