Topic 3: Visual Object Recognition Flashcards

Question

What are inversion effects?

Answer 1

are said to occur for stimuli that people have difficulty noticing unusual details (anomalies) in which turned upside down yet are easily and rapidly detected when the stimuli is right side up inversion effects are strong for faces, which has been argued to indicate they are processed holistically (rather than in a "piecemeal" or serial way, a little at a time/bit by bit, etc.)

Answer 2

higher-level (superordinate) category e.g., dogs, cats, faces, houses, etc. this is what we generally think of as being affected in visual agnosia

Answer 3

specific examples from a given category, e.g., their dog, my cat, your face, etc. this is what seems to be affected in prosopagnosia

Answer 4

Kanwisher is closely associated with a line of work arguing that the fusiform face area (FFA) is specialized (calibrated, or "tuned") for processing face-like stimuli much of her early work focused on distinguishing processing related to type vs. token identification, which may help explain the perspective she brought to bear on understanding what appears to be the selectivity of the FFA

Answer 5

given that many people demonstrate FFA activation when viewing faces, another question is whether FFA activation causes us to perceive faces, or whether it's activation is merely correlational along similar lines, what other parts of the brain may be important for facial processing? another debate in the literature is whether or not the function of the FFA is "intrinsic" (e.g., coded for in our genes), or dependent on experience

Answer 6

records electrical signals directly from the cortex surface this is in contrast to trying to pick up the remnants of those signals by placing the electrodes directly on the scale, as with EEG ECoG is often used in clinical contexts (e.g., to localize tissue that is especially prone to epilepsy) can provide a unique research opportunity to measure neural activity in a way that isn't usually possible can also be used to electrically stimulate the cortex (which also affords some unique research opportunities)

Answer 7

Parvizi et al. (2012) measured a signal that is thought to be closely related to the OLD signal (oscillatory power in the Gamma band) and, like that signal, reflect higher rates of synaptic activity they found similar results as with fMRI (e.g., elevated activity in the FFA when viewing faces) they were able to directly stimulate the FFA and observe (in real time) what impact that had on the perceptual experience of the patient fitted with the ECoG the patient reported that stimulation of the FFA resulted in a distortion to his perception of faces in the room (but nothing else), supporting the idea this region plays a causal role in the perception of faces

Answer 8

one case study in which the patient could recognize facial expressions but not identify faces (e.g., recognize who they belonged to) another case study in which a patient could identify faces but not facial expressions (would we expect a double dissociation related to this if we obtained the right imaging data)

Answer 9

case study of a birdwatcher who lost the ability to discriminate between different species of birds, without having their perceptual identification of objects from other categories obviously affected

Answer 10

although deficits in face perception may be the most likely/obvious symptom, what we refer to as prosopagnosia may more generally reflect a loss of perpetual expertise, which may take various different forms (even if the most obvious as it applies to facial processing for the average person)

Answer 11

an alternative kind of account to the idea that the FFA is specialized for processing faces is that it is more generally used to process any stimuli that one develops expertise in within the visual domain the average human probably becomes very good at recognizing faces due to the value of being able to do so in a highly social society (remembering who has helped/hurt you in the past etc.)

Answer 12

it seems entirely consistent with the variability we often see in case studies (e.g., bird watcher losing the ability to discriminate amongst different species after FFA damage)

Answer 13

could relate to the way the participants naturally engage with the stimuli if we're shown pictures of faces with minimal instructions, the average person may (relatively automatically/spontaneously) attempt to determine whether they know the person on the other hand, although we could go through the same recognition process when shown other kinds of stimuli if we wanted to, it may be the case that the average person is much less likely to spontaneously make that kind of judgment for other categories of stimuli if so, the way participant approach the task may potentially differ across stimuli types, which could introduce a confound

Answer 14

perhaps the most well-known attempt is often referred to as the "Greebles" study, named after the novel stimuli used that was created specifically for the experiment by creating stimuli that didn't already exist, Gauthier and Tarr (1997) effectively controlled for prior exposure

Answer 15

the Greeble stimuli show inversion effects, which suggest they are processed in a holistic way (just like faces: we don't focus on one part at a time in isolation in a "piecemeal" fashion) they tend to be categorized at the individual/token level, rather than category/type level (once again, like faces) before "training" (to recognize/categorize the Greeble stimuli), the FFA does not respond strongly to Greeble stimuli after many hours of training, the FFA ow has a robust response to those stimuli suggests acquiring perceptual expertise with these stimuli led to recruitment of the FFA

Answer 16

could the fact that they chose a novel stimulus that, although not existing before their study, nevertheless looked a bit like they had faces (and/or "behaved like face stimuli" in various ways) be problematic? once you put a meaning on a abstract thing, it's hard to not think of it that way

Answer 17

interestingly, Bukach et al. (2012) replicated the above experiment with a prosopagnosia patient and found they were eventually able to reach "expert" status (albeit they required more trial than controls to achieve this) they were thought to use alternative strategies to compensate for their deficit, as we've discussed happens in other contexts (e.g., visual agnosia)

Answer 18

while not correcting for multiple comparisons can increase the rate of false positives (Type 1 error), correcting for multiple comparisons can also make the threshold so high for achieving statistical significance that it becomes very likely that you will be unable to reject the null hypothesis even if it is not true (Type 2 error, or a false positive)

Answer 19

at the level of a whole brain, which typically has >20,000 voxels, one would expect that 1000 or more voxels would be falsely identified as being "activated" unless one were to apply a correction the most rigorous is to divide each p value by the number of tests (called Bonferroni correction) for just two voxels, this means that a critical p of 0.025 needs to be achieved; not a big deal; but for 20,000, the critical p would drop to 0.00000025

Answer 20

Kanwisher's work was able to get around the problem of multiple comparisons by using a approach in which, instead of doing many comparisons involving the activity of individual voxels at singular points in times OR correcting for multiple comparisons, she essentially averaged the activity of specific groups of voxels over many trials

Answer 21

if we find that region 1 responds more strongly to object A than object B, would you feel comfortable making the claim that must mean region 1 is specialized for object A? probably not a much stronger argument could be made if we found that region 1 responds more strongly to object A, as compared to a large array of other kinds of objects (object B, object C, object D, etc.) so we can't rule out the idea the region 1 responds to object A and another given object (e.g., object X) until we directly compare activation of region 1 when shown object A, with what happens when shown object X

Answer 22

we can't get close to testing every possible comparison we could think of (with every other kind of object that exists in the world) we therefore have to practical and at some point decide the balance of evidence is enough to support a claim like the FFA is specialized for responding to faces, without going overboard, but with enough data to feel reasonably confident so you can really never be absolutely certain any given region ONLY responds to one type of stimuli and never anything else but it's important to gather as much evidence as you can to at least minimize the possibility that there is some obvious alternative explanation (e.g., another kind of major category of objects that you didn't think to test)

Answer 23

involved a very "low-level" solution, simply bring the experimental participants back for multiple fMRI sessions so they could gather enough data, to rule out enough alternative explanations, to satisfy the scientific community that her argument was at least plausible (albeit not an absolute certainty)

Answer 24

in theory, this allows us to go beyond simply looking for "hot spots" of activation (i.e., what part of the brain become most active when viewing stimuli x) to examine the distributed pattern of activation throughout the whole brain, which is presumably important in many situations

Answer 25

we might think of traditional "univariate" fMRI studies as being about measuring localization of function, while MVPA focused more on distributed representations

Answer 26

e.g., how does the distributed pattern of activation when viewing faces in one subset of the data (even numbered trials) compare to viewing faces in another subset of the data (odd numbered trials)?

Answer 27

e.g., how does the distributed pattern of activation when viewing faces in one subset of the data (even numbered trials) to viewing tools in another subset of the data (odd numbered trials)?

Answer 28

they found a high degree of within-category correlations (and low degree of between-category correlations), for all categories on the surface, this would seem to suggest that there is a great deal of consistency in the activation of a widespread and distributed pattern of activation throughout the brain when viewing stimuli from a specific category in other words, yes the FFA becomes active when you view faces, but so do lots of other areas... and that exact pattern of distributed activity is very consistent across blocks when viewing faces at the same time, that distributed pattern (when viewing faces) seems to be rather different from what occurs when viewing tools

Answer 29

all object-selective cortex: included data from all voxels excluding maximally responsive voxels: excluded voxels with the strongest activation, or in other words, exclude "hot spots" the same general result held, even when excluding the "hotspots" in the second version of their analyses: a high degree of within-category correlations, and low degree of between-category correlations, for all categories

Answer 30

the results of this study suggest that there is much more to facial processing than just the FFA rather, a large/widespread network of regions seems to be recruited when viewing faces along these lines, something similar could be said about the widespread/distributed pattern of activation when viewing objects from other categories however, it is worth noting that simply excluding the "hot spots" does not control for signals those locations may be sending to the rest of the cortex

Answer 31

fusiform face area (FFA): selectively responds to faces (maybe?) parahippocampal place area (PPA): selectively response to places/scenes/landscapes extrastriate body area (EBA): selectively response to body parts

Answer 32

damage to the PPA is associated with difficulty recognizing places/scenes/landscapes (topographic agnosia) providing a parallel of sorts with prosopagnosia