Brainscape's Knowledge GenomeTM

Browse over 1 million classes created by top students, professors, publishers, and experts.


See full index

PSYC 440 > Face Recognition and Modularity > Flashcards

Face Recognition and Modularity Flashcards

1
Q

How is perception an inference and how does it connect to the concept of modularity?

A

An example:
you illusory percept of a and b as differing in luminance
1. depends on the inferences that are rapidly and automatically generated
2. cannot be undone by simply knowing and believing that the percept is illusory

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
2
Q

What is modularity?

A

arises from the idea that the brain can function as a computer
-a module is a function that you can call over and over again to perform a certain set of tasks

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
3
Q

What does luminance perception appear to assume?

A

-there is a single light source
-light is uniformly distributed
-solid objects cast shadows

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
4
Q

What are the assumptions of luminance built into and what do they affect?

A

these assumptions are built into the way the perceptual system makes inferences about luminance and reflectance when given luminance

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
5
Q

What do the assumptions made about luminance constitute?

A

domain specific knowledge and that is applied to these computations

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
6
Q

What happens when light hits your eye a certain way?

A

there is an inferential process and the light is hitting your eye in a certain way and it is hitting a system which has assumptions built into it about light sources and shadows and light distribution

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
7
Q

What did fodor advocate for in modularity of mind?

A

that these inferences are being carried out by modules
-that cognitive functions are independent of one another

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
8
Q

What are some key features of modules?

A

-fast and automatic
-domain specific
-informationally encapsulated - cannot think your way out of a module and information cannot link between modules
-innate

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
9
Q

What is a recurring debate in cognitive neuroscience about functional modules?

A

whether functional modules are a good descriptions of the functions of anatomically defines areas
-can we take functions and impose them onto anatomy

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
10
Q

What are the three major points to consider when considering the debate over face perception and modularity?

A
  1. The FFA, domain specificity, perceptual expertise, and interactive specialization
  2. viewpoint invariant identification of faces from face patches in macaques
  3. computational models of face representations
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
11
Q

What is interactive specialization in face perception?

A

the fact tat faces are special and one of the few objects you can tell thousands apart - so maybe the expertise from experience is important npt the domain specialization

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
12
Q

What are the evolutionary arguments for face processing?

A

-recognizing conspecifics (kin, in group members, etc) is important for survival in social species
-faces convey emotional information
-other primates seem to process faces in a way that is different from the way they process objects

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
13
Q

What is the computational argument for why there might be a module for face processing?

A

-faces cannot be recognized in the way other objects seem to be - when we see a face there is a processed triggered where we know who that is in parallel and can recognize them whereas with objects we look at the parts of an object and then determine what that object is rather than the whole object coming together and meaning something or someone like a face does

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
14
Q

What evidence is there for a module for face processing?

A

is face idenitfication fast an automatic - yes

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
15
Q

How quickly can we recongize individuals from images?

A

within 75 ms

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
16
Q

Is face identification informationally encapsulated?

A

this is a harder questions with less data but knowing about an optical illusion does not help you unsee it which means you need to switch in and out of different processing modes

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
17
Q

Does face identification depend on domain specific knowledge?

A

-a very controversial question - we process face stimuli differently when they are upside down than right side up
-critically if you model face recognition as recognition by components it does not work well
-the role of configural information in face processing and how it might be related to other parts

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
18
Q

In the left hemisphere there is an analogous area to the FFA for recognizing text so when you flip words upside down what happens?

A

you have to go letter by letter and slow you down when reading it which makes you memorize it cause the harder the stimuli the more likely you will memorize it

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
19
Q

If you do a lexical decision task and you increase the length of the words what happens to the speed at which you process stimuli?

A

there is no relation to length meaning you process stimuli of varying lenght the same meaning you do not go letter by letter when reocngizing words

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
20
Q

If you do a lexical decision task and you increase the length of the words and happen to flip the orientation of the letters what happens to the speed at which you process stimuli?

A

then a linear process arises where it takes longer for longer words which means you do end up going letter by letter

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
21
Q

What part of the brain is thought to be important for word recongition?

A

left fusiform gyrus

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
22
Q

Where would a module for face processing live in the brain anatomically speaking?

A

-the candidate face area is part of the ventral visual stream
- a portion of the fusiform gyrus specifically in the right hemipshere is sometimes called as the fusiform face area or ffa
-ventral poriton of the temporal cortex right before you transition from temporal to occipital lobe

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
23
Q

What is the double dissociation between the dorsal and ventral streams in the visual system?

A

patients with damage to the dorsal stream like RV have little difficulty in identifying objects but when asked to grasp or manipulate them they show deficits in using an objects shape to generate a motor plan

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
24
Q

Where did initial evidence that the right fusiform plays an important role in face perception come from?

A

-patients who exhibit prosopagnosia
-these patients have impaired face recognition but not object recognition (i.e. the man who mistook his wife for a hat)
-they had a lesion in the ffa or fusiform gyrus

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
25
Q

What is an example of the double dissociation between face and object recognition?

A

-patients who exhibit prosopagnosia
-these patients have impaired face recognition but not object recognition
-but agnosics have difficulty with objects but not faces (i.e. DF had object recognition issues but could recognize faces)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
26
Q

What is the hughlings Jackson problem and how did it affect how we evaluate the role of the FFA in speech?

A

hughlings jackson - to locate the damage which destroys speech and to locate speech are two different things
-just because damage to the right fusiform disrupts face processing does not mean it is responsible for face recognition it could simply be a chain in the face recongition process - every method used to investigate human condition has some very substantial limits so start to believe a story when you see evidence confirming from neuropsychology in agreement with fmri data or or behavioral data

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
27
Q

What did the imaging evidence say about the FFA in face recognition?

A

see blob in right fusiform gyrus for face processing
-like faces more than objects
-like intact faces more than parts of faces
-need the whole face configuration to activate this area
-matters for the active and passive tasks - the active task where you need to tell if you have seen a face before or not causes greater activation
-used block design
-people are not good at recognition of faces of different species

28
Q

Is there a statistical comparison that can be done between the activation for faces versus objects in different areas of the brain?

A

yes and it was revealed that for faces the anterior right fusiform was actuvated while for objects the medial posterior parahippocampal region
-both of these are along the ventral stream

29
Q

What is the N170?

A

when they used erg measurements from an eeg and found a difference in the evoked response to pictures of faces compared to othe robjects that peaks around 170 ms
-this activity seems to be generated by the fusiform gyrus
-face processing from brain to retina takes 80ms so somehting else is happening here
-the dipoles are coming from are ambiguous but using a forward model we can say that if we assume the dipoles are from the fusiform it matches the dipoles that are observed

30
Q

What are the arguments for the modularity of the ffa?

A
  1. appears to be highly domain specific - we can think of it as a visual area that has a receptive filed defined by some set of features that allows it to infer the presence of a face (only wants to see intact face)
  2. its responses are rapid and automatic
    -within 200ms responses in the right fusiform distinguish faces from a wide range of other objects

-these responses are evoked during passive stimuli meaning which means the brain does it already

31
Q

What is an alternative account that the fusiform area is generally recruited for perceptual visual expertise?

A

-maybe what is special about faces is that we are experts at distinguishing them
-part of this is that we have very extensive and meaningful experience looking at other peoples faces
-is thats all there to the specialization of the fusiform we would expect FFA activity in car and bird experts

32
Q

Where was activity in the fusiform found to colocalize along with faces?

A

objects with expertise - it is the perceptual expertise itslef and not just the stimulus itself so we have made ourselves perceptual experts at recognizing faces - importnat to note that the n is small for bird and car experts (but n was still okay to get published)

33
Q

Can we turn people into perceptual experts?

A

-participants were trained on a set of computer generated 3d models called greebles
-they learned to identify them by family gender and at the individual levels
-expertise is operationalized as being as fast to name the individual as they are to name any of the other category labels

-greeble experts were found to show de novo preference for greebles in the FFA; could also still recongize faces via the ffa

34
Q

Why could some people not reach the criteria for greeble expertise?

A

-cause the names were non english like so the linguistic information could not stick

35
Q

What type of domain contribution does the mid fusiform provide in regards to perception?

A

domain general
-the debate has not been settled if the mid fusiform is domain specific for faces or not

36
Q

What has the concept of domain specificity shaped in cognitive neuroscience?

A

the study of structure function relationships

37
Q

Is the modularity of face processing innate?

A

the data suggessts this is not plausible

38
Q

Why is the debate between innateness and experience dependency never all or none?

A

because for instance if you take conrad lorenz and his duck imprinting is innate but experience is important because the ducks need to see conrad to imprint on him

39
Q

What is innate about face recongition?

A

infants seem to prefer humans faces to wide rnage of other stimuli
-could it be that the domain specific configural processing stratgey is somehow hard wired
-or is there another explanation for this preference

40
Q

Infants prefer to look at conspecifics (members of the same species or people) but this preference tends to decline early in development, why is the basis of this preference nt the same thing that makes faces speacil to adults?

A

infants like the face like configuration over non face like but if we change this with top heavy image or bottom heavy image or other orientations - infants lilke top heavy

THIS MEANS THAT THERE MIGHT BE SOME INTERATUVE SPECIALIZATION THAT WITH THIS BIAS FOR TOP HEAVY IMAGES COUPLED WITH SOCIAL INTERACTION WE THEN GET THE MODULE LIKE PROCESSING FOR FACES

41
Q

Since the fusiform gyrus is not necessary for infants responses what is the other area involved?

A

superior colliculus - important for stimuus trackong that allows you to lock in on faces early in development

42
Q

How does face specific activity in the ffa grow up between childhood and adolescence?

A

there might be an innately specified process that guides the modularization of the fusiform for something like face specific processing but it does not appear that kikely that the fusiform carries an innate function from early in development

expect by five years of age that the fusiform should have preference for faces than adults - adolescents - have nice big effects from design experiment and see building and navigation and objects and faces - get big orbits response sin adolescents and the blue is important for object rcongiotn and in green have the parahipocampal place area is important for navigation - faces atcuavte teh fusiform and places important for social like superior temporal sulcus and the superior and anterior temporal cortex is important for faces and - for kids more cortex is atcuavted for the objects and buildings and for faces there is not a lot of fusiform activity which means face processing is not mature in 5-8 year olds

43
Q

Ideas that bigoted are genetic and infants prefer their own ethnicity than another is all fake ties that face recognition ties to childhood - do this thing in Minnesota and see a homogenous population and white kids and they ask how well they can recognize the sam person they have white faces and Japanese faces - adults are better at white faces than Japanese faces and that emerges very very late so these are grade levels so from five to 12 years old do not see race effect until later in development - looking at 5 year old brains and see they cannot tell faces apart so develops very gradually - what did this show?

A

-the ability to recognize people from faces we have just met seems to develop well into adolescence
-whats notable here is not just the increase in accuracy (memory test get better at the cause of schooling with age increasing)
-whats notable is the very gradual perceptual narrowing as evidenced by the late emergence of the other race effect

44
Q

What did Tsao and Freiwald try to show with the macaque face patches?

A

-how do macaques solve the problem of knowing that is the same person in different orientations
-the way we talk about face processing hinges on studies where we do not deal with this invariance problem at all
-did monkeys cause electriphysiology is in monkey cortex

45
Q

What did Tsao and Freiwald find in fmri in the macaques?

A

the same experimental technique was used to idneify face specific processes in humans and it reveales a number of patches along the ventral stream in macaques
-found face seletuve patches because they do not respond to objects as strongly as faces but they respond to objects stronger than scrambled faces

46
Q

What were the patches identified for face recognition by Tsao and Freiwald?

A

ML and MF are posteiror and AL and AF are anteior to ML and MF and AM is the most anterior to them all
-individual neurons in these patches generally repsonses more stringly to faces than any other visual stimuli
-they also appear to be hierarchically organized - the more anterior you get the more transformed the stimulus gets which it should have directly received from the primary visual cortex posteriorly

47
Q

What did the three cells in ML and MF (most posterior) lowest part in hierarchy show?

A

-the top panel shows responses to a bunch of face stimuli relative to a set of object and artificial control stimuli
-the bottom panel shows responses to many different peoples faces grouped by orientation

-cell 2 at the top panel has very low background level of response and responses are a lot stronger 200ms after the face is presented which is evidence for face selectivity
-cell 1 has greater baseline activity but the activity increases after seeing a face - the activity is suppressed for non faces at 200ms and then baseline activity resumes
-cell 3 has no response to faces cell 3 is responsive to faces in only a certain orientation

48
Q

What did the three cells in AM (most anterior) show?

A

cell 3 - a cell in AM that is generally suppressed by faces but responds strongly to a particular individual
-this response is viewpoint invariant - aka responds to same face of same individual regardless of orientation

49
Q

Why is viewpoint invariance important?

A

-in the early stages of vision receptive fields are location and orientation specific’
-we can make a curve detector or a corner detector from these but they wont help us identify a front view and left and right profile of the same person as the same
-ideniity is invraiant to rotation because rotation is an accident of our spatial relationship to a person
-recall that cell in ML and MF responded to a particular profile

50
Q

How did Tsao and Freiwald further demonstrate the identity specificity of responses in AM?

A

they looked at the similarity of population responses
-we can look at the correlation of activity levels in an area and then make a heat map based on similarity for responses to different images
-in this way we can observe how the similarity space defined by neural activity is organized

-essentially provided an alternative to the subtraction method and used representational similarity approach in terms of neural activity - neural activity representation is a distributed pattern of activity over a bunch of cells - some cells respond strongly and some weakly because they are inhibited due to the stimulus and you can compare the pattern or representation to one evoked by another neural image - can look at a pattern of images and see what other images allow for that same abstraction

51
Q

When we look at the representational similarity approach used by Tsao and Freiwald what do we see and what experimental design was used?

A

experimental design - stimuli were presented in blocks by orientation and individuals were presented in the same sequence for all blocks

-these are idealized responses for three different kinds of requirements

-the dark squares in i (ML and MF cell patch) indicate a high degree of similaity in responses to different individuals with the same orientation

-the dark squares in ii (AL and AF patch) indicate a high degree of similarity in respones to different individuals with the same orientation and the mirror image of that orientation (like right and left profile are mirror images of each other so the neurons responded to both)

-the dark sqaures in iii (AM patch) indicate a high dgree of similairity in responses to different images of the same person viewed from different orientations or agles

52
Q

How could Tsao and Freiwald representational similairty approach findings come to be on a theoretical level?

A

-they provide a simple hierarchical model of the system
-the receptive field in layer i encode face features in a location specific way
-similar looking individuals from the same viewpoint will evoke similar activity patterns but the same individual from different viewpoints will activate very different patterns

-in layer ii receptive fields are defined by over responses in layer 1
-layer i cells that fire maximally to the same combination of face features in mirror symmetric orientations are connected to the same layer ii cell
(there is also lateral inhibition within this layer that sharpens the individual identity; the stringers neuron gets stringer and surpasses everything else)
-in this way layer ii cells are partially viewpoint indepndent

-in layer iii receptive fields are defined over responses in layer ii
-by taking input only from those cells that respond to the same features across a range of mirror symmetrical viewpoints these cells now become identity selective and viewpoint invariant

53
Q

What are some limitations in the Tsao and Freiwald study?

A

why did they have the monkeys respond to individual specific responses for people
-these are also people the monkeys have never seen before so how do they have individual responses for people
-the first layer in the network (ML and MF patches) do not have stored representations of these people they have never seen before

54
Q

Since the first layer in the network (ML and MF patches) do not have stored representations of these people they have never seen before what could they have instead?

A

feature detectors - when we talk about this we mean the relative configural properties of the face like overall shape of hair or eyebrow slant or intereye distance- some neurons have a best face so they will max out when a face they are most sensitive to arises

55
Q

When freiwald looked at MF responses to cartoon faces what did they find?

A

some neurons seemed to code for both part based and configural features of the stimuli

56
Q

How are feature detectors good evidence for domain specificity?

A

-some features like inter eye distance eyebrow slant and iris size are good candidates for a domain specific representation of faces
-since humans and monkeys both have eyes eyebrows irises etc these could be tuned up by experience with monkeys and less extensively with human handlers in a way that generalizes across species

57
Q

What did macaques raised in isolation from other monkeys and without exposure to human faces show about the innateness of face modularity?

A

-deprived animals preferred looking at monkey or human faces than to looking at pictures of other objects (but we do not know how they would respond to top heavy bottom heavy stimuli)
-unlike typically reared monkeys they did not prefer monkey faces to human faces

58
Q

How do face patches especially AM function as a module for face recognition?

A

-responses to faces are fast and automatic elicited with passive viewing at <200ms
-individual recognition seems to depend on domain specific representation
-what is innate about these features is a hard question but we at least know that a behavioral preference for faces doe not depend on early experience

59
Q

Is the FFA a human homologue to the AM face patch?

A

no - AM is at the very anterior edge of the inferior temporal lobe in macaques (anterior temporal regions are used to decode individual identity but FFA cannot) (the other face patches are distributed along the superior temporal sulcus whereas FFA is on the ventral surface of the temporal lobe)
-the homolog in humans is the inferior temporal cortex

60
Q

What can the FFA do based on similairty analyses?

A

-can distinguish faces from houses but does not provide info about individual identities

61
Q

What information do anterior temporal regions provide?

A

-anterior temporal regions contain information about individual identity but don’t distinguish stimulus classes

62
Q

What information does the early visual cortex have?

A

response similairty is driven by the overall similarity of the image

63
Q

Why is the FFA so heavily discussed despite the negative findings?

A

-it was an impressive disocvery for the time
-methods to decode the information in brain areas were developed much later than intial FFA arguments
-early fmri techniques were not appropriate for observing anterior poles
people did not study the inferior temporal cortex because of the temporal pole being susceptible to blowout artifact for technical reasons did not see inferior and anterior temporal cortex in humans cause it was difficult to study and it was impressive and a what u would expect based on the neurpsych literature - recognizes that there is a face in ffa but is not where the faces are being pulled part from one another

64
Q

What can we say about the function of the FFA?

A

responds rapidly and selective to faces but also to other objects of visual expertise
-face selectivity in ffa develops gradually until late in childhood
-damage to the ffa disrupts the ability to identify faces
-but it does not appear to encode information for individual recognition (nor is it homologous with regions that do in an evolutionary relative)

65
Q

The notion of modularity has been important in guiding research on structure function relationships in the brain but what is difficult about it?

A

cannot generate consensus that a particular region is a module or whats its function is
-aka it is a good way to try to think about how the brain works but does not actually translate

66
Q
A

PSYC 440 (13 decks)

Brainscape helps you reach your goals faster, through stronger study habits.
© 2024 Bold Learning Solutions. Terms and Conditions