Greg Detre
magnocellular
larger receptive fields
transient responses (fast-adapting)
short latency
non-linear (on + off don't cancel each other exactly)
necessary for detecting motion (in or after V1, the M cells are direction/motion-selective)
parvocellular
smaller receptive fields
longer latency (response time)
slow-adapting
carry on firing throughout duration of stimulus � less good at detecting change, good for objects you keep looking at
linear (less direction-selective, through some are (especially ventral stream cells)
�
(cross-talk processing of random dots insufficient to activate the parvo)
V5/MT � global motion
grating moving across the visual field
V1 can only respond to small receptive field, local edges/motion (�the aperture problem)
if two gratings � plaid pattern (needs big receptive fields to see)
damage to MT � selective blindness to motion but not form/colour
V4 � colour constancy subtracts the average of the colour in the surround to compensate the central
also involved in form-processing between V2 and infero-temporal
crucial for object recognition � otherwise agnosias
Zeki over-emphasised colour
V1 � two separate visual areas
blob + inter-blob
parvo � blobs (colour not orientation)
all these areas are 6 layers (like the rest of the neocortex)
M � dorsal
P - ventral
Retinal processing
Visual pathways
simple � edges/bars, orientation-tuned, position-sensitive
complex � same, but it matters where in the receptive field the edge is, so it can't be mapped with a spot of light
end-stopped (simple/complex) � the bar/edge has to terminate in the receptive field
motion/direction-sensitivity
at V1, primates do motion/direction in the M-system, but not in the retina
but the rabbit is pre-specialised (and so throws away a lot of information early) and detects direction at a retinal level)
W-cells � similar to the K system
posture - � superior colliculus (eye movements)
organisation of V1 with respect to each eye
chiasm � 50% of fibres cross
no communication between both eyes before the primary visual cortex
so no disparity in LGN, only after V1
depth perception � dorsal and � parietal (concerned with actions and how far away things are)
individual neurons are tuned to different disparities in V1 (heropta Windike-Muller sector(???))
iterative to pair up the two eyes inputs by neighbouring areas
cortical layers communication with brain
most inputs from preceding area � 4 (stellate cells)
(striate cortex � so many coming from the eyes � big split because of the fibres = t �stripe�)
4 � pyramidal cells (spuerficial � top layer) in layer 2
some input � superficial pyramidal in 2 bypassing the layer 4 stellate cells
output from the superficial pyramidal cells � layer 4 of the next cortical area
deep pyramidal cells in layers 5 + 6
these have inputs from other cells in other layers
outputs descend to BG + superior colliculus
also project backwards to the preceding cortical area
end in layer 1 (apical dendrites of those superficial pyramidal cells)
there are just as many back as forward projections
1 million from LGN � V1
100m from V1 � LGN (Rolls: just because deep pyramidal cells are programmed to give back-projections)
face cell is not a grandmother cell � because does not code for a unique face � distributed coding
superior colliculus � does not code for retinal position but vector (direction the eyes have to move)
how fall out of love?
optimal strategy different for male/female genes
maybe men only love for a few years � mix your genes a couple of times
women should invest in their offspring because they can't spread their genes so easily
women � greater propensity to infidelity
one for genes, one for upbringing
mid-life crisis
everyone has to be programmed to die
natural selection makes progress through change
death after reproductive years � variation
every time it splits, you get some extra noise
an ageing gene might just control other genes
turn them off in rats � age 4 times slower
if you don't die of one thing, die of another
eat less: wear + tear on metabolism