Lecture 13- Reading and object recognition II Flashcards
How does visualisation of hypercolumns work?
-can visualize whole column response by changes in blood flow on surface of cortex -stimuli: different views of same object -columns responding to different views of same object (hypercolumns) overlap -multiple views of same object brought together in one area and treated as equivalent for identification? -hypoercolumn provides invariant view? -always get and invariant response -Neighbouring regions of monkey ventral occipito-temporal cortex respond to different views of same object
What did the imaging of columns when shown complex objects?
-complex object may stimulate many columns at once -each column part of a hypercolumn responding to different aspect of object -response may be to very subtle aspect of object -object identified by constellation of simpler elements (no single tiger column)
What is the column activation when looking at a fire extinguisher?
- multiple columns define the objects -the redness activates something
- black and white= a different combination lights up
- so spot 1 excited by protrubances (things sticking out of objects)
- spot 2= curved lines
- spot 3= rectangualar bottom =again have multicolumn response that gives us the understanding of the object
What are the more complex responses in a monkey ventral occipito-temporal cortex?
-some neurons in monkey ventro-occipito-temporal cortex respond to different versions of multiple objects -cell located high up in object recognition hierarchy -not clear what common elements stimulate this cell -uncertainty due to incomplete testing of cell?
What is the neuropsychology of object recognition?
-have looked at some of the neural circuitry underlying object recognition, a lot remains to be understood -now look at at object recognition in a more general way to understand what problems must be overcome -understand what the neural apparatus is trying to do
How does understanding of the visual world work?
-need to rapidly analyse scenes into individual elements (objects) and determine identity, distance and movement -can’t rely on having previous experience of object -object recognition modules parse any scene into elements (objects) -the visual system is very interested in intersection (to identify overlap and distance)
What are intersections important for?
-object recognition is very hard if intersections are not included -it becomes much easier when intersections are included -the concentration on intersections is the basis for some classic visual illusions
What is the visual problem when looking at a whole image?
-1. image clutter: overlapping and partial views of objects (sheer number of objects) -2. object variety: so many different things to handle -3. variable views: similar objects seen from different angles
What are the steps in recognising images?
- Identify edges
- Identify uniform regions bounded by edges
- Identify object and ground
- Assign border ownership
- Group together things with similar properties (single objects?)
- Fill in missing bits (interpolate)
- Identify object (link to name, previous experience, use etc.) -this is not a complex image -first must identify the edges= give you the boundaries
What is the object recognition and invariance?
-invariance strongest feature of object recognition -infinite number of views of infinite number of objects, can’t have all possible views stored in memory -how are models of complex objects stored in the brain to allow comparison with stimuli? -Two theories: a)object based=view independent recognition by components b)image based- view dependant, depends on some prototypical views of object in memory
What is theory 1- object-based object recognition?
- brain is hard wired for approx 30 shapes
- all objects made up of these shapes
- each shape varies along five axes (curvature, co-linearity, symmetry, parallelism and co-termination)
- the brain can recognise 30 shapes, you break down the things you see are made up of the combination of the 30 objects
What is theory 2: Image-based object recognition?
-all possible views of object calculated from handful of different views (side view, head on view, back view) -neither of theories seems compelling evidence is mixed -first couple of times you see an image= you store elemental views, then when see a different view can reconstruct it
What is object agnosia?
-when cannot recognise objects
What is the connection of strokes and object recognition?
-if inferior occipito-temporal lobe is important for object recognition, then strokes and fMRI should affect object recognition tasks -strokes give two extreme types of problems with object recognition -both lead to agnosia (ignorance or absence of knowledge)
What are the two types of agnosia?
-apperceptive and associative
What is the apperceptive agnosia?
-can describe objects and their function from memory -can identify by touch -can’t copy drawing of object -can’t see objects, see assemblage of parts -can’t even see the object, cannot distinguish it from background etc. -because the visual system is damaged in the posterior temporal lobe
What is associative agnosia?
-can describe objects and their functions from memory -can identify by touch -can copy drawing of object -can see object as object but can’t name it -can see the object but the connection to the naming etc is damaged
Where is the damage to when have apperceptive agnosia?
-damage to first stage of object recognition (visualising the object- in posterior temporal lobe)
Where is the damage to when have associative agnosia?
-damage to second stage of object recognition (identifying the object- in anterior temporal lobe)
What is alexia?
-alexia is a form of associative agnosia -the same part of the brain is damaged as the part damaged in alexia
What is face recognition?
-a special case of object recognition
What are the characteristics of face recognition?
-face often used as stimulus for vision tasks -powerful stimulus= faces very important to primates, identify individuals, convey emotion and other information -selectively affected by strokes (prosopagnosia) -can recognise that it is a face, but not who it belongs to, can recognise voice
What is prosopagnosia?
-strokes producing prosopagnosia involve fusiform and lingual gyrus in inferior temporal lobe -often damage bilateral but some cases with right damage only -can also involve other forms of agnosia (alexia, some other object classes reported) -much more likely to occur with right side damaged, cases where only right lobe is damaged, both lobes can recognise faces, but right lobe probably better
What is developmental prosopagnosia?
-some people born with prosopagnosia (genetic lesion that disturbs wiring in the brain) -have completely normal vision -often face recognition only deficit, normal object recognition -a specific type of object recognition with dedicates neural circuitry -both developmental and induced prospagnosia can include general object recognition deficits -prosopagnosia is often correlated with alexia etc, but can have associated object recognition problems= as the parts of the brain that do this are close to one another
Are faces recognised in a different place to letters?
-yes -alternating testing with faces versus words always activates neighbouring areas -faces ventromedial to letters -it’s the same in all 5 subjects -in temporal medial lobe
What do faces activate?
-multiple regions in ventral stream -macaques have five other areas in addition to FFA that respond to faces (humans similar) -damage to any one can cause prosopagnosia -not known if non-faces have multiple areas -multiple places in the ventral visual pathway which react to faces, 5 or 6 places -fusiform face area= this is the one where all the information comes together
What are the role of multiple face areas?
- each face responsive region does a different thing
- brought together the invariant view of a face and face recognition
- why not have just one face area?
- responses are different depending on the view of the face
- identifies “faces” from one direction
- generalises across mirror images (cf “b” and “d”)
- unites various views of individual face (invariance?)
What is the role of hierarchy in object recognition?
-invariance may be sum of activity in sites 1 to 4 -area 5 key to identifying face as individual person -connects to higher centres for naming emotional impact, past history -the substations of going to the fusiform face area that contribite to the ability to identify the face
What is face recognition like with a split brain?
-Joe tested with images JW (self) morphed together with MG (other) -compared left and right hemisphere for ability to identify self and not-self -split brain= -right= better at face recognition -the right hemisphere at 80% joe can tell that it is
What does the right hemisphere do in a split brain in terms of object recognition?
-much better than left at recognising not self faces -right could recognise self but needed a complete image
What does the left hemisphere do in a split brain in terms of object recognition?
- much better than right at recognising self faces
- maybe that left better at working with few clues for self
Do faces share the temporal lobe with other object recognition?
- yes -face, word, everyday object and place
- recognition areas are all present in neighbouring regions of inferior temporal lobe
What is the organisation of object recognition areas visualised with fMRI?
- organisation not necessarily sequential or neat
- multiple areas for each object class can exist
- the same pattern in monkey and a human
- multiple regions that are specialised for obejcts of a particular nature
What are other object recognition areas?
- tend to be in reproducible places
- body parts: extrastriate body area (lateral occipito temporal cortex and fusiform body area)
- places/landmarks
- parahippocampal place area (loss leads to landmark agnosia)
- words: visual words form area, left inferior temporal lobe (loss leads to alexia)
What is the inferior occipito-temporal cortex and object recognition for?
-does the inferior-temporal lobe really have regions specialised for identifying specific classes of objects? -what about train spotters and dog breeders? -could we identify parts of human inferior temporal cortex for train or dog identification? -if we can what is the significance of this (cf the letterbox)? -is the brain plastic to give you what you need?
Does the inferior temporal cortex have a gradient of capability?
- inferior temporal lobe may have rostro-caudal and medio-lateral gradients representing a range of object features
- specific classes of objects ay be best detected by a particular location where cortex sensitivity match object parameters
- results in reproducible locations on cortical surface for best detecting specific classes of objects
- if you treat trains as a special class of object then you will have a train spotting region
- other people will recognise trains using their normal object recognition area
What is the letter box and the visual pathways?
-the letter box lies in the what visual pathway -down to the temporal lobe, ventral pathway -we are pre-adapted to recognise letters and words by the existence of an object recognition system
How are letters objects?
-letters are just objects -analysed by part of the object recognition circuitry in the ventral occipital-temporal cortex -shares this region with other specialised object recognition regions -have written symbols been optimised to match the underlying object detectors of the inferior occipito-temporal cortex -writing systems share many features : high contrast characters, average three strokes per characters, a reduced lexicon of shapes that recur across cultures
What is the hypothetical scheme of word recognition?
-
How is reading object recognition?
-letters as objects are new (approx. 6000 years old) -left ventro-occipital area is object recognition area -now contains letter-box for processing written language -what did this area do more than 6000 years ago -what was recycles to provide space for letters
Who are the studies done on mostly?
-virtually all fMRI studies are done on literate university students -not the condition during our evolution -literacy changes anatomy of the brain
What does literacy change?
-changes corpus callosum -compared corpus callosum size in literates versus non-literates -small change in region where parietal fibres cross -compared corpus callosum size in literates versus non-literates -small change in region where parietal fibres cross -compared corpus callosum size in literates vs non-literates -small change in region where parietal fibres cross
Does literacy changes grey and white matter throughout the brain?
-yes -grey and white matters regions also larger in literates (mainly parietal and occipital) -explains corpus callsoum differences
What is the cortical functions in literates vs illiterates?
-reading tasks in literates activates standard “reading/language” network -silent in illiterates -literacy meant VWFA became highly selective for words
What is lost from VWFA?
-compared images of faces, houses, tools, words, or checkerboards at location of VWFA -in illiterates, similar responses from all -in literates strongest response to words and greater loss (left side) is in response to faces and checkerboards -face response now dominant on right side in literates, but equal bilaterally in illiterates -incidentally showed house response varied with birth location (rural vs city) -so face recognition is lost in left hemisphere, replaced with the letter recognition
Summary:
-1. we have evolved verbal language skills -2. the existence of these circuits has preadapted us to manipulate language using other, non-verbal, learned symbol systems -3.Letter and word recognition depends on existing object recognition areas that are pre-adapted to recognise objects no matter how they are presented -4.The letter box is in left hemisphere, close to language circuits, partially in place of strong left hemisphere face recognition
