Lecture 8 - Object perception + recognition Flashcards
What is recognition?
the identification of a stimulus as an object we are familiar with
What is the problem?
an infinite no. of possible retinal images can correspond to a particular object
Solution?
visual input must be matched with internal object representations held in object stores
Theories of object recognition - structural analysis?
Objects may differ in their 2-dimensional appearance but they become equivalent in 3D space i.e. visual input must be analysed at different levels
Davis Marr 1982?
- structural analysis proceeds through 4 sequential levels until a 3D model is achieved which can be matched with stored object representations
- raw primal sketch?
processing of intensity (brightness) changes across the retina
- full primal sketch?
geometric organisation of these intensity changes i.e. definition of edges and contours
- 2½-D sketch (½ because it includes depth information)
processing of spatial locations of visible surfaces, includes depth but only represents the object from the observers viewpoint, perceived objects cannot be generalised
- 3-D model (not a sketch anymore)
represents the object independently of the observers position, 3D standard form is achieved i.e. real shapes and their relative positions to each other, can be matched with stored object representations
Why are Marr’s hypotheses excellent?
they have inspired a cognitive model
Why is having a cognitive model excellent?
having a cognitive model is excellent as it this explains a cognitive mechanism which can be used to make predictions about human behaviour
Why are predictions excellent?
Predictions are excellent as we can test them and thus verify or falsify the model
A model of object recognition - Ellis & Young 1988?
look at diagram
1. object
2. initial representation (primal sketches)
3. viewed-centred representation (2½-D sketch)
4. object-centred representation (3-D model)
5. object recognition units
- there is a recognition shortcut from the viewer-centred representation for unambiguous viewer-centred representations
6. semantic system = storage of general but also personal facts about objects
7. name retrieval = accessing these facts generates name retrieval
Testing model predictions - Mr S reported by Benson & Greenberg 1969?
- was a young soldier who suffered carbon monoxide poisoning
- neurological tests revealed that he can: maintain fixation, name colours and describe sensations, detect small changes in brightness and wavelength on psychophysical tests and detect movements of small objects
- but he cannot perceive shape or form e.g. cannot copy simple figures, letters, digits or match simple objects
- he also cannot recognise or name objects and pictures of objects (body parts, letters, numbers, faces)
Group of patients with right posterior lesions, reported by Warrington 1982?
- can match objects for size, colour, brightness
- can recognise objects from pictures in usual views
- cannot recognise objects in unusual views
- no impairments with object recognition in everyday life
- can explain use of objects e.g. torch = hand held light
AB reported by Warrington 1975?
- civil servant with progressive cerebral atrophy (ongoing loss of neurons and connections between them)
- patient can: distinguish basic shapes, identify colours, numbers and letters and match pictures of different views (usual and unusual)
- patient cannot: name common objects or famous faces , make substitutions when categorising objects or identify meaningful sounds
Evaluation of Ellis and Young model of object recognition?
- it predicts the behavioural pattern of patients with 2 different types of visual agnosia
-> visual agnosia = inability to recognise seen objects
-> impaired perceptual stages (apperceptive agnosia) e.g. Mr S - cannot copy or match even simple shapes, profound impairment in shape perception
-> impaired semantic stages (associative agnosia) e.g. AB - can copy and match objects but intact percept has no meaning
-> distinction originally suggested by Lissauer 1890 - Patients with right posterior lesions cannot create object-centred representations - they have a perceptual categorisation deficit but don’t have visual agnosia (object recognition is intact)
Conclusion of the model?
- The model is valid as it makes accurate predictions about behavioural impairments
- But single cases may not be valid measures to test models
- But it is valid for now