Chapter 3 Flashcards
what is the dominating sense in humans? why?
vision - more brain area is devoted to it and we trust it when it conflicts with other senses
define form perception
The process through which you manage to see the basic shape and size of an object
Define Object recognition
Process through which you identify what the object is
why is object recognition crucial
It’s essential to apply your knowledge to the world and is crucial for learning
What was the gestault psychologist’s main claim
the perceptual whole is different from the sum of it’s parts because the ways that we perceive a stimulus differs and goes beyond the stimulus itself!
What is the necker cube and what does it prove
A reversible cube that can be perceived in two ways proving that perception goes beyond stimulus by specifying an arrangement in depth although the drawing is neutral in regard to perceptual organization
What are other examples of reversible/bistable images?
The vase/profiles figure and the schroeder staircase
What is figure/ground organization
determination of what is the figure and what is the background
What causes reversable figures to flip back and fourth?
The perceiver changing how they’re organizing/percieving the stimulus
Are reversible figures the only visual stimuli with multiple interpretations?
NO there are many ambiguous figures in the world in need of interpretation
ex. fruit bowl
What are the gestalt principles?
Proximity: elements that are close to eachother are percieve as being the same object
Similarity: elements that are similar to eachother as percieved as partsof the same whole
Continnuation: we see objects continuing even when our vision is blocked
Closure: we perceive closed figures rather than iincomplete ones
Simplicity: We interpret forms in the simplest way possible
What comes first, features or interpretation
Neither, they work together - because features of an object depend on how the form is organized by the viewer and if features are missing we provide them
What are the 2 types of influences when identifying objects?
- Stimulus driven or bottom up influences: come directly from features that are visble in the stimulus 2. knowledge/expectation driven or top down influences: come from the perceiver
What supports the view that detection of simple visual features is the first step in object identification
It’s much easier to search for a target with one simple feature rather than a target defined by a combination of features - what we’d expect if feature analysis is first
what evidence do brain damage studies provide for the feature first claim?
- damage to the parietal lobe can cause integrative agnosia where patients can detect single features but can’t judge how the features are bound to form complex objects
How has TMS research supported the figure first claim?
By disrupting the parietal lobe function and slowing the performance of subjects searching for a target defined by combinations of features
What is a tachistoscopic presentation
An experiment in which a stimulus is shown for a breif duration (20-30ms) followed directly by a mask - used to be performed with a machine called a tachistoscope
What factors affect whether subjects can recognize stimuli presented in a tachistoscopic presentation
- Familiarity: people recognize twice as many words when they are frequent vs. infrequent words
- Recency of view: if the target word is primed by the same word people are more likely to recognize it the second time around (called repetition priming)
What is the word superiority effect?
Words themselves are easier to identify than isolated letters
What is the “two alternative, forced choice” procedure and what does it prove
Either a letter (K) or a word containing that letter (DARK) is presented followed by a mask then a question asking which of two letters (K or E) was in the target. People are better at identifying letters when they are in a word
Do Pseudowords or letter strings produce context benefits
Pseudowords (LARE) are more easily recognized while random strings of letters (JPSR) have no benefit compared to isolated letters
What are the two approaches to explaining the context benefit (as in context of the word not a sentence) for word recognition?
- Pronouncability: easily pronouncable strings provide a context benefit
- Probability/well-formedness: The more frequent a combination of letters is in english the easier it will be to recognize in a psudoword - the more english like the string the easier it is to recognize
What are the systemic errors observed in brief exposure to pseudo words?
People have a tendency to misread less common letter sequences as more common patterns - people perceieve input as being more regular than it is
ex. TPUM is misinterpreted as TRUM or even DRUM
Describe a simple feature net
A network of detectors organized in layers with each subsequent layer being concerned with more complex larger scale objects. The bottom layer is concerned with features, then letters then words.
What is an activation level
The status of a detector at any given moment - how energized it is.
What is a response threshold
The amount of activation required for the detector to fire and send it’s signal to the other detectors to which it is connected
What 2 factors affect activation level?
- Recency: detectors that have fired more recently will have a higher activation level
- Frequency: Detectors that have fired frequently in the past have a higher activation level
Why are frequent words easier to recognize than rare words?
because they have a higher activation from having been frequently fired in the past meaning even a degraded presentation (breif/dim) is enough to reach the response threshold
How can we explain repetition priming
Activation levels are temporarily lifted due to recency of firing and even degraded input will cause the word to reach it’s response threshold
How can we account for the easier recognition of words that are more english like using the feature net?
We have to add another layer for bigram detectors (combinations of letters) between letter detectors and word detectors. Therefore english like combinations have higher activation levels
How does the Feature Net recover from confusion caused by partial letters?
Partial info at the feature level causes confusion at the letter level (too many possible letters activated weakly). This is sorted out at the bigram level because the weak input will be enough to fire the most primed (common) combination and avoid an error
How does the Feature networks bias to regular combinations cause errors
Infrequent words or non words are incorrectly perceived as the more common interpretation
How is knowledge of common spelling patterns organized in the Feature Net?
It is NOT locally represented (stored in a particular location) instead it is distributed knowledge, each detector’s activity level is locally determined and we can only infer knowledge if we consider how the network functions as a whole
How does the McClelland and Rumelhart model differ from the classic feature net?
- absence of bigram detectors
- activation can travel in both directions and across levels
- There are inhibitory connectors that when fired inhibit the activation of another (ex, firing of a G would deactive the TRIP detector)
Why is the McClelland and Rumelhart model perfered over the classic feature net?
Because it reflects the biological organization of neurons and the two-way communication present in visual processing
Explain the recognition by compnonents (RBC) model
A hierarchy of detectors used to identify objects. The lowest level is features, then geons, then geon combinations/ geon assemblies then the object model
What are the advantages of the geon model?
- Viewpoint-independent: You can identify an object from any angle
- Partially hidden objects can still be recognized from just the few visible geons
- Geons play a crucial role in identification, to recognize an object you must first recognize it’s geons
What is the multiple views / template proposal?
People have stored in memory different views of an object (limit = 6) and if the current view doesn’t match one from memory you must perform mental rotation until it does. This means speed of recognition is viewpoint-dependent where some angles will be faster than others
How exactly does recognition proceed according to the multiple viewpoints model
A hierarchy of detectors, the lowest level responds to lines, the higher level; corners and notches and the top level fires to whole objects. These representations are supported by the what pathway (in the inferotemporal cortex) with object-specific neurons that are view-tuned
Which proposals about object recognition involve a hierarchical network?
All proposals involve some kind of network model
define agnosia and prosopagnosia
Agnosia: inability to recognize certain stimuli
Prosopagnosia: Inability to recognize faces ( and in some cases other stimuli as well - birds and cars)
What does Prosopagnosia suggest about face recognition?
It implies a special neural structure exclusively involved in recognizing faces which is damaged in victims of prosopagnosia
What evidence is there to support that face recognition is strongly dependent on orientation?
- When common objects and faces are turned upside down performance suffers in both cases but much larger in the recognition of faces
- A distorted face appears less disfigured when viewed upside down
What evidence suggests that the specialized system for recognition of faces isn’t entirely unique to faces?
- The fusiform face area (specifically responsive to faces) is also activated in subtle distinctions among birds and cars
- The upside down effect can also be demonstrated in other categories of stimuli such as dogs, but only if the subjects are Dog experts
What are the 2 characteristics required to involve the use of the specialized recognition system which is sensitive to orientation?
- The task must involve recognizing specific individuals in a category
- The category must be extremely familiar (faces for humans, dog for dog breeders, birds for bird watchers)
Explain the holistic perception of faces
Recognition of faces depends on the complex relationships between features not an inventory of the faces parts
What is the composite effect?
When given the top half of one face and the bottom half of another subjects have difficulty identifying only the top half because they view the top as part of the whole. However when the faces are misaligned the task becomes much easier
What is the problem with looking at object/word recognition in isolation as a self contained process
Top-down Priming that provides a context can make the stimulus easier to recognize and this draws on knowledge outside of object recognition.
What are the 4 perceptual consancies?
- Size constancy – as objects move closer or farther away the size changes on our retina but we know they don’t actually change size in real life
Ex. Ebbinghaus illusion – something surrounded by bigger things is perceived as smaller - Shape constancy – we perceive things as being the same regardless of the angle at which they hit our retina
- Brightness constancy – brightness of an object… a certain object surrounded by darkness looks lighter and vice versa – lateral inhibition coming into play
- Color constancy – color of an object is perceived relative to it’s background
Munker-White effect:
WHat are the 3 theories for how we recognize objects?
- Features
- Templates
- Geons
What physiological evidence supports the feature theory?
Recordings from neurons: cells respond to very specific simple features - edge detectors, color detectors etc.
What cognitive evidence supports the feature theory?
- conjuction search - harder when there are 2 features
2. Search asymmetry - Certain features are easier to single out thatn others
What are the 2 problems with template theory?
- There aren’t enough templates for every view so some aren’t going to match perfectly (thus we must employ transformations) = viewpoint dependent
- Template matching isn’t strong enough from general object recognition
Whats the problem with feature theory
It looks only at simple features not the spatial relations between them
How does structural theory (RBC) solve the problem of feature theory
it incorporates informations about components AND spatial relations
What features aren’t necessary in object identification according to RBC
Color, texture, shape(?)
What evidence supports geon theory?
- easier to identify incomplete objects when the geons are still recognizable
- Priming only occurs from complementary pictures if the geons are intact
What are the 2 problems with geon theory?
- Some things are difficult to distinguish based on geons (dog/wolf)
- Some things require very idiosyncratic geons to be posited (bread)
Which theory of object recognition is correct? Template, feature or geons?
They all explain some of the data - so we use a featuer network?
What 4 things makes word recognition easier or harder?
- frequency
- repetition
- context
- well formedness
WHat must be explained by our model of feature detection?
- frequency
- repetition priming
- word superiority
- well formedness
- overregularization
- context effects
How did simons and lewis demonstrate change blindness?
the person getting directions got switched behind a caried door
WHat did the strayer and johnston study reveal?
When speaking on the phone people ran more lights
What is apperceptive vs associative agnosia?
Apperceptive (perceptual representation) : Fails to recognize objects due to functional impairment of occipital lobe
Associative (identification/meaning) : Impairment to assigning meaning to an accurately perceived stimulus