Test 2 Flashcards
Invariance in object recognition
Object recognition tolerates substantial differences in retinal images (i.e., object constancy)
Template matching - Objects recognition
Object recognition based on fit with existing template in long-term memory (may use normalisation tricks like rotation).
Template alone doesn’t seem like the way to go, visual system has to do more
Feature analysis - Objects recognition
Attempts to obtain object constancy by having multiple feature detectors working in parallel
Image -> Feature demons -> Cognitive demons -> Decision demon
Structural Description - Objects recognition
Recognition model that describes objects by the structural organisation of parts/components (object skeletons)
Recognition by components (geon model) - Objects recognition
Identifies simple geometric components that make up complex objects (called shape primitives or “geons”). Much harder to recognise objects when geons are difficult to figure out due to occlusion or deletion.
Example of changing geon type or size
Equal amount of metric manipulation impacts object recognition more when the change involves geon types
View-based model - Objects recognition
Train participants to recognise novel objects at several 2D views then test with trained/untrained views. Participants perform better with trained views. Priming effects are translational invariant: moving object laterally b/w prime and test has no effect on performance
Object recognition is…
High level
Rapid
Object recognition and segmentation
Object recognition can influence figure/ground assignment in bistable displays. Experiment showed equal times of detection and categorization. Thus recognition can happen as fast as figure/ground segmentation but NOT within class identification
How rapid is object recognition?
between 100 - 150ms. This suggests mostly feed-forward mechanisms. This helps narrow down time for object recognition.
Define Apperceptive visual agnosia
Impaired visual perception of form/shape of objects, despite normal elementary vision (acuity, brightness, colour, etc)
Define Associative visual agnosia
A selective impairment in visual recognition of objects despite apparently adequate visual perception of them
what is the Lateral occipital complex
A set of ventral visual regions that responds more strongly to everyday objects than scrambled images of those objects. Responds to object format more than object cues.
Selective responses along ventral visual pathway
Some ventral regions show highly selective responses to particular object types (faces, bodies, scenes, words, etc)
Big debate about category-selective areas
Are these areas dedicated for recognising only preferred objects (modular view), or do they all work together in recognising all objects (distributed or many-to-many view)
Alternative hypotheses Prosopagnosia - Object agnosia hypothesis
Prosopagnosia is the most visible symptom of visual object agnosia. But there can be Prosopagnosia without object agnosia.
Alternative hypotheses Prosopagnosia - Within-class hypothesis
Prosopagnosia impairs within-class recognition of any objects. Testing within class -> worse for within faces than within other things.
Alternative hypotheses Prosopagnosia - Visual similarity hypothesis
Prosopagnosia impairs recognition of visually similar exemplars of an objects (upgraded within class). When testing dissimilarity scale of cars tracked at the same rate as other participants.
Alternative hypotheses Prosopagnosia - Expertise hypothesis
Prosopagnosia impairs recognition of objects with which we have obtained expertise. Predict they could not learn greebles, BUT they learned greebles similar to other participant success rates.
is there a double dissociation between face and object processing
Can be. One subject saw all the faces in the illusion and in silhouette could only see face not hand.
Define Face composite effect
Harder to discriminate face-halves aligned than misaligned because aligned halves are perceptually integrated
Face part-whole effect
Easier to discriminate features in whole face than in isolation because faces are represented as wholes.
Define the concept of Face Space
A multidimensional perceptual space in which the dimensions are features we use to discriminate faces (eye height, nose size, etc.). Individual faces are mapped according to their values on the dimensions
In relation to face space define prototype face and anti-face
Norm = A face that sits at the center of face space because it has the average value on all dimensions. Norm face doesn’t exist in real world, rather it’s a statistical average of all faces one has ever seen. Anti-face = A face that lies on the same trajectory as the original face but on the other side of the norm. An anti-face contains all the opposite features of its original face.
Norm based coding of faces
Adaptation to matched anti-face facilitates identification but adaptation to non-matched anti-face doesn’t. This shows that identity trajectories from norm face exist.
Feature coding in posterior face patch
Face cells in ML are tuned to simple attributes/dimensions of faces (hair thibkness, inter-eye distance, etc.)
Face space coding in posterior face patch
Face cells in ML track feature values across multiple attributes/dimensions in face space
Identity-invarient coding in anterior face patch
Face cells in AL are tolerant to image-level changes and show robust reponse to specific identities
Hierarchical organization of face selective areas
Spatial information is integrated across increasingly larger regions of faces, eventually allowing for holistic processing
Face-selective cells causes face perception
Stimulation of face cells in FFA causes direct changes in the perceptual experience of seeing a face
Adult-like category-selective map in children/infants
Cortical regions selective for faces, places, and objects show adult-like spatial organization early in development. Perceptual hallmarks of holistic processing (inversion, composite, part-whole effects) are present by 4-5 yrs.
Developmental prosopagnosia
Lifelong faceblindness despite normal vision, intellect, and lack of brain damage that tends to run in families. Generally face recognition skills are more similar in identical than in fraternal twins, indicating genetic contributions.
Objects vs scenes, what is the difference?
Objects are spatially compact entities we act upon; scenes are spatially distributed entities we act within
Scene recognition
Scenes can be recognised by signature objects (toaster > kitchen) or by overall spatial layout (openness > desert)
Coarse to fine hypothesis
Scene recognition by global spatial layout begins with rapid processing of low spatial frequency information
Scene-selective areas
Secondary visual regions that are highly responsive to spatial layout of scenes but not to objects or faces
Functional organisation of scene-selective areas
PPA damage impairs place/landmark recognition; RSC damage impairs ability to orient relative to landmark
What is a cognitive map and what is it made of?
A mental representation of local visual environment that results from learning without obvious reward.
Place cells- respond to a particular place in the environment of an animal
Head direction cells - activate when head in particular direction
Grid cells - Neurons whose multiple firing locations define a periodic
triangular array covering the entire local environment
David Marr’s tri-level framework - what is it?
the three levels at which any
machine carrying out an information-processing task must be understood
David Marr’s tri-level framework - what makes it?
- Computation. What is the goal of the process, why is it appropriate, and
what is the strategy by which it can be carried out? - Representation/algorithm. How can this computation be carried out? What is the
representation for the input and output, and what is the
algorithm for the transformation? - Implementation. How can the representation and algorithm be implemented
physically?
Expertise - Three types of names
Basic names - guitar
Superordinate (very general) - instrument
Subordinate (Very specific) - bass guitar
Cognitive Economy
Your brain is lazy and will always choose the quick easy way to do it
We use basic names because we have just enough information to get the task done
Often subordinate info is not necessary
Expert vs. Novice Categorisation Dog experts (judges) and bird experts (Bird watchers)
When responding to novice categories mostly basic
When it is their expert category they use more of the subordinate naming
In expert domain the labels for subordinate are the same as in the basic condition
Is expertise transferrable?
As you move further away from your area of expertise the expertise decreases. Expertise facilitates learning of new items related to our field of experience but transfer is limited.
Expert levels in the Gray Owl Example
Basic: owl
Subordinate: great grey owl
Superordinate: bird
objects and objects of expertise
Use similar mechanisms in the brain
Neural Mechanisms of Expertise
We focus on features that are diagnostic. Identifying objects of expertise results in greater activation of the LOC.
Multisensory integration
Perception allows us to create an internal representation of the external world, helps us understand the external world rather than an exact representation
Visual representations combine information to create a single perceptual experience
McGurk effect - what is it?
When you play a short audio recording with different videos of people moving their lips it changes your auditory perception
McGurk effect - What does it show? When does this happen? How does this happen?
- Visual information alters subjective auditory experience
- Low signal-to-noise ratio
- Audio-visual integration and Converging pathways
Altering visual perception: Playing beats with disc flash/es
As you increase beats, people perceive more flashes. Fairly robust, even though you know you still observe it. Only exception is two flashes and one beat, does not work.
Altering visual perception: faces on continuum with emotive sounds
With no sound it should be pretty accurate. Should not be totally clear around the middle. We judge the person to have the emotion which matches the noise.Second study asked for what the face emotion was, NOT what emotion the person had. The effect does get a lot smaller BUT there is still a significant effect.
Cat experiment for visual perception
Visual and auditory on separate timers. Response in neurons was greater than the sum of the two if they occur in the same time/space.
Left superior temporal sulcus: multisensory
When they occur together the firing is greater than the sum of the two firings alone. Subadditive response when it is incongruent.
Multisensory integration allows us to combine information from different perceptual domain
Perceptual illusions
Improve signal-to-noise ratio
Senses may provide stronger signals for particular judgements
Vision informative about space
Audition informative about time
Perception is modular, but integrated
Convergence in heteromodal brain regions
Feedback to the primary sensory cortices
Crosstalk between different sensory cortices
Infant perception: What’s built in and what’s learned?
Visual perception - not much to see before birth
Auditory perception - auditory learning begins prenatal
Preferential looking procedure
Give two stimuli, see which is preferred. BUT difficult to interpret
Habituation procedure
Wait for baby to get bored then change stimulus
Visual perception: Acuity
Acuity develops. The image does not focus on the retina early on, not until toddler.
Feedback loops
The outputs of a process feed back into the process as regulating inputs
Critical periods
Time periods in which specific experiences are necessary for typical development to occur
Visual perception: colour
Harder to fixate on target when it was in the same colour category as the background, similar between adults and infants
Therefore categorical perception of colour is likely not dependent on language or teaching
Visual perception: Face discrimination
Newborns: no preference for own-race or other-race faces; no preference for human or monkey faces
3-month-olds: prefer caregiver-race faces over other-race faces; prefer human faces over monkey faces
Perceptual narrowing
Increase in the precision of perceptual processing in one category, at the expense of perceptual processing outside that category
Auditory Localization: Where are sounds coming from?
They get better but then there is a dip where it is worse… U shaped curve.
Owl glasses example: younger owls change more and adapt faster.
High-amplitude sucking method
Testing preferences: Infants learn that if they suck in a certain way they can hear a certain sound; test whether infants suck to produce that sound
Auditory perception: What sounds do babies like to hear?
Newborns prefer mom’s voice over a stranger’s.
Newborns prefer hearing a story they heard in the womb (3rd trimester)
Newborns prefer speech sounds to nonspeech sounds
Newborns prefer hearing their mothers’ native tongue
High-amplitude sucking method
Testing preferences
Infants learn that if they suck in a certain way they can hear a certain sound; test whether infants suck to produce that sound
High-amplitude sucking method
Testing discrimination
Infants suck while listening to one sound until they are habituated (sucking rate decreases to a certain level); test whether infants dishabituate (increase their sucking rate) when a new or new-category sound is played, in contrast to when an old or same-category sound is played