Running head: CATEGORY SPECIFIC SPATIAL FREQUENCY 1

advertisement
Running head: CATEGORY SPECIFIC SPATIAL FREQUENCY
Category Specific Spatial Frequency Adaptation with Faces and Cars
Yoseph Ali Beki
Thesis completed in partial fulfillment of the requirements of the
Honors Program in Psychological Sciences
Under the Direction of Dr. Isabel Gauthier, PhD and Olivia Cheung, Graduate Student
Vanderbilt University
April 6, 2011
Approved
Date
_____________________________ _______________
1
CATEGORY SPECIFIC SPATIAL FREQUENCY
2
Abstract
This experiment establishes new grounds in the theoretical debate of object perception—between
the domain-specific hypothesis and the expertise hypothesis. Using spatial frequency adaptation,
the results from this study will provide more evidence in support of the expertise hypothesis. The
results will test whether expertise processing (here, face processing) is significantly different
from object processing. Participants were adapted to a range of spatial frequencies (high/low)
and an image category (face/car) to determine if adaptation effect would remain within the image
category. Also, bird and car expertise was measured to analyze the interaction between expertise
and category specific spatial frequency adaptation. While the correlations between adaptation
and expertise were not significant, some adaptation effects were significant and support the need
for further research.
CATEGORY SPECIFIC SPATIAL FREQUENCY
3
Category Specific Spatial Frequency Adaptation with Faces and Cars
Over the past few decades, vision research, in particular the study of face processing, has
grown in sophistication and complexity. Scientists first studied how faces are processed in
comparison to other “non-face” objects, such as tools or houses. They noticed that people
differentiate and recognize individual faces even when faces are similar. All faces have two eyes,
a nose, and a mouth, yet this uniformity does not hinder the identification of a specific face
amongst others. Also, faces are remembered very well in comparison to other objects. With these
observations in mind, psychologists focused on how faces were processed as opposed to other
objects. Through experimentation, they revealed holistic processing, where one processes an
image wholly as opposed to processing it in parts, and the face inversion effect, where
recognition times of studied faces decrease when they are inverted while recognition times of
studied objects are not as affected by inversion (Davidoff & Donnelley, 1990; Tanaka & Farah,
1993; Valentine, 1988; Yin 1969; Young, Hellawell, & Hay, 1987; Yovel, Paller, & Levy, 2005).
As technology advanced to the point of analyzing functions of the brain, the fusiform face area
(FFA), within the fusiform gyrus, was discovered. Through brain imaging, such as fMRI, this
area of the brain was found to react more to faces than to other objects (Kanwisher, McCermott,
& Chun, 1997; Tong, Mascovitch, Weinrib, & Kanwisher, 2000).
With so many differences between face processing and object processing, psychologists
began to create theories to explain how faces and objects are processed in two distinct ways.
CATEGORY SPECIFIC SPATIAL FREQUENCY
4
There were two theories that emerged over others. The first is the domain-specific hypothesis,
which states that faces are processed by a specific area in the brain (the FFA), set apart from
areas that process objects (McKone, Kanwisher, & Duchaine, 2006; Kanwisher, 2000). The
hypothesis proposes that there is an area that is specialized for processing faces that is separate
from areas that process objects. Within this hypothesis, different categories of visual stimuli are
processed in separate areas of the brain. Supporters of this hypothesis used the phenomenon of
holistic processing as proof for their hypothesis. Holistic processing is the idea that one
processes an image as a whole as opposed to processing it in parts. Studies showed that when
trying to focus on one part of an image, participants were better at ignoring other parts of objects
than other parts of faces (Tanaka & Farah, 1993; Davidoff & Donnelley, 1990; Young,
Hellawell, & Hay, 1987; Yovel, Paller, & Levy, 2005). The argument used to support the domain
specific hypothesis that arises from this observation is that holistic processing is specific to
upright faces, so a specific face module must exist within the brain.
The domain-specific supporters also used the face inversion effect as evidence. The face
inversion effect is a phenomenon in which previously studied faces are recognized faster and
more accurately than previously studied objects, but when the same faces and objects are
inverted during testing, objects are recognized faster and more accurately than faces (Yin, 1969;
Kanwisher, Tong, & Nakayama, 1998; Valentine, 1988). The difference in results because of
inversion proved the domain-specific hypothesis to its supporters. They believe that the
particular mode of face processing was hindered by inversion, while the particular mode of
object processing was not affected.
Support for the domain-specific hypothesis was also found using brain imaging to test
whether the brain processes faces separately from objects. Experimenters found that a part of the
CATEGORY SPECIFIC SPATIAL FREQUENCY
5
fusiform gyrus activated when participants were presented with a face and not when they were
presented with an object (Kanwisher, McCermott, & Chun, 1997). The discovery of the FFA in
the fusiform gyrus gave domain-specific supporters an exclusive area in the brain for processing
faces that could prove their hypothesis.
Along with showing activation in the FFA, domain-specific supporters used brain
imaging to show that prosopagnosics, people who cannot recognize faces or are poorly able to
identify faces, have damage to the FFA (Farah, Wilson, Drain, & Tanaka, 1995). They propose
that because of the brain damage or deterioration of the FFA, these patients were not able to
recognize faces, which further supported the exclusive use of the FFA for faces and the domainspecific hypothesis.
The second face processing hypothesis that is currently supported by psychologists is the
expertise hypothesis, which proposes that people have become experts at identifying faces, and
because of this level of expertise, faces are processed in a different way than other objects. To
prove this hypothesis, experimenters have focused on achieving effects exclusive to face
processing for other objects, such as birds and cars. They believe that if face expertise causes
phenomena such as holistic processing, the face inversion effect, and activation of the FFA, then
other types of expertise (bird or car expertise) would cause these phenomena as well.
In support of the expertise hypothesis, dog experts were shown to hold inversion effects
with dogs in comparison to dog novices, so the faces inversion effect was proved not to be only
for faces, but for objects of expertise (Diamond & Carey, 1986). Also, researchers showed that
the FFA was activated in bird and car experts when they were presented with objects of their
expertise (Gauthier, Skudlarski, Gore, & Anderson, 2000a). With these results, the FFA was
CATEGORY SPECIFIC SPATIAL FREQUENCY
6
found not to be exclusive for face processing; the area was used for object processing as well, as
long as the participant was an expert with the given object.
Furthermore, experiments with novel objects (Greebles) found that with recognition
training, participants were able to show activation in the FFA (Gauthier & Tarr, 1997; Gauthier,
Williams, Tarr, & Tanaka, 1998). Novices who had never seen a Greeble before the experiment
showed activation of the FFA when presented with a Greeble after they had gained a level of
expertise with them through training. Along with FFA activation, these Greeble experts also
showed holistic processing and inversion effects with the novel objects (Gauthier & Tarr, 1997;
Gauthier, Williams, Tarr, & Tanaka, 1998). This study along with other similar studies
demonstrated that when people became experts with a certain object, they will process them in
the same way as they process faces (Gauthier & Tarr, 1997; Gauthier et al., 2000b; Tanaka &
Taylor, 1991).
With two prominent hypotheses fighting for validity in the field of face processing,
debates between supporters of each side and critiques of each hypothesis were presented.
Proponents of the domain-specific hypothesis state that there have not been any significant
behavioral effects or imaging results that prove the expertise hypothesis (Kanwisher, 2000;
McKone, Kanwisher, & Duchaine, 2006; McKone & Kanwisher, 2005; Robbins & McKone
2006). Furthermore, they point out that there have been expertise experiments that have failed
(McKone & Kanwisher, 2005; Tanaka et al., 1996). Also, they were able to show that Greeble
training was able to increase performance in categorization and naming tasks in a
prosopagnosiac, which led to the conclusion that expertise with faces did not correlate with
expertise with objects (Duchaine, Dingle, Butterworth, & Nakayama, 2004). Domain-specific
CATEGORY SPECIFIC SPATIAL FREQUENCY
7
supporters used the separation between Greeble/novel object expertise and face expertise to
nullify the experiments using Greeble training and to support their hypothesis.
However, supporters of the expertise hypothesis suggest that their data give validity to
their hypothesis. They also argue that the experiments run by domain-specific hypothesis
supporters are biased toward domain specificity (Gauthier, Anderson, Tarr, Skudlarski, & Gore,
1997). These biased studies do not take the level of categorization into account when ruling out
object expertise.
These experiments use the effects of objects in basic categories such as car or dog to
contrast against the effects of individual faces of Tim or Bob. Cars and dogs are more distinct
from each other than two individual faces. A car has four wheels and a trunk, and a dog has four
legs and a tail. These objects are differentiated from one another by recognizing specific parts or
configuration of parts within them. The difference in these parts leads to basic categorization.
(Gauthier, Behrmann, & Tarr, 1999).
Faces, however, are more closely related to each other. They have the same parts in
them, and the parts are configured in the same manner. All faces have two eyes, a nose and a
mouth, and they are located on faces in a uniform manner. To differentiate between faces,
viewers must use a higher, or subordinate, level of categorization, and they must use more subtle
visual cues such as color, size, and shape (Bruce & Humphreys, 1994; Diamond & Carey, 1986;
Rhodes, 1988; Tanaka & Taylor, 1991). In comparison to basic level categorization that novices
are able to accomplish, subordinate level categorization requires a level of mastery, experience,
or expertise to accomplish.
Therefore, to compare objects and faces, objects must be recognized on the same,
subordinate level as faces. More specific, subordinate categories such as Honda Civic or
CATEGORY SPECIFIC SPATIAL FREQUENCY
8
Dalmatian need to be used. Expertise hypothesis supporters design their experiments to target
subordinate categorization, so that both faces and objects are processed on the same categorical
level.
Also, in regards to the significance of the experiments ran that suggest expertise,
proponents of the hypothesis state that the effects of object expertise is smaller than the effects
of faces due to the level of exposure participants have to their object of expertise (Gauthier &
Bukach, 2007). In relation to the level of exposure to faces, one sees objects of expertise less
than faces. Bird watchers see more people in their lifetime than birds. Therefore, their level of
expertise for faces will be higher than their level of expertise for birds, and the effects from
viewing faces will be greater than the effects from viewing birds.
In response to the supporters of the domain-specific hypothesis that point out failed
expertise experiments, expertise supporters stated that, while some experiments fail, they do not
rule out the rest of the experiments that succeed (Gauthier & Bukach, 2007). These failed
experiments do not nullify the rest of the literature that has shown expertise effects; they only
call attention to situations where expertise effects do not happen and promote further inquiry into
the area.
In light of the existing controversy between the two sides of face perception research,
new forms of testing each hypothesis need to be developed. Supporters of each view of face
perception use similar experiments and analyze their results in a way to favor their hypothesis.
To open a new path of discovery within the field, more innovative experiments and methods
should be used. This experiment will apply spatial frequency and adaptation methods in an
attempt to find a new facet in the face processing theoretical debate.
CATEGORY SPECIFIC SPATIAL FREQUENCY
9
Adaptation methods in visual processing can use visual stimuli containing different
properties that are separated in the visual system. An example of adaptation using visual stimuli
is red-green phenomenon in color perception, where one views a red stimulus for a period of
time and sees green when the red stimulus is removed. Another example is motion aftereffects in
motion perception, in which a person perceives upward motion in a still object after viewing
downward motion for a period of time (Addams, 1834).
In adaptation experiments, a person is exposed to a type of stimulus for a long period of
time, which causes one’s perception to be biased against that stimulus. Adaptation effects occur
because neurons that respond to a type of stimulus fatigue after prolonged exposure to the
stimulus, and neurons that usually compete with the fatigued neurons activate with ease. For
example, the red-green phenomenon occurs because neurons that respond to the color green
compete with neurons that respond to the color red, and draining the red neurons of energy by
looking at the red stimulus causes one to see green (Cohen, 1946).
While the examples of visual adaptation discussed above use low-level visual properties,
higher-level visual adaptations are seen with face identity or category judgments (Webster,
Kaping, Mizokami, & Duhamel, 2004). For example, face adaptation studies have been effective
when using higher-level categorizations, such as gender, race, and emotion. The connection
between these high-level adaptation studies and the low-level adaptation studies discussed earlier
is that the stimuli within each experiment are similar—they are colors, movement, or faces.
Adaptation studies that cross categories, such as a study with faces and cars, have not been
attempted. The problem with cross-category adaptation is that the shared properties of stimuli
decrease as one moves from within-category to cross-category. Therefore, adaptation
CATEGORY SPECIFIC SPATIAL FREQUENCY
10
experiments that attempt to use multiple categories should use properties that remain shared
across categories.
A property that is shared across higher-level categories that can be used in cross-category
adaptation studies is the separation of high and low spatial frequencies throughout the visual
system. All images contain a spectrum of spatial frequencies. Experimenters can filter out
sections of this spectrum from images, leaving certain parts of the spectrum intact within the
image. For example, one can filter the middle and high spatial frequencies out of an image to
retain only low spatial frequencies (LSFs), which contain very coarse information of an image
that give one basic properties of an image. With faces images, LSFs give one the general shape
of the face. An image containing LSFs looks as if it is blurred (Figure 1). One can also retain
only high spatial frequencies (HSFs), which look like sketches of images and very detailed
(Figure 2). With HSFs, one can determine the detailed information of the face, such as wrinkles
and scars (Schyns & Oliva, 1999; Gauthier, Curby, Skudlarski, & Epstein, 2005). Schyns &
Oliva (1999) found that face processing and categorical judgments, such as gender and
expression identification, depend on different ranges of spatial frequencies. These studies show
that a low-level property that is shared across categories, such as spatial frequencies, could be an
effective property to use in cross-category adaptation studies.
Within the experiment, participants were adapted to images of faces or cars in different
spatial frequencies to see if aftereffects of adaptation cross between the different object
categories. The aftereffects were retrieved by having participants judge hybrid images before and
after adaptation. Hybrids that contained two faces with different spatial frequencies were judged
as male or female, and hybrid that contained two cars with different spatial frequencies were
judged as sedan or convertible. When a participant is adapted to a certain stimulus, the neural
CATEGORY SPECIFIC SPATIAL FREQUENCY
11
network that processes that stimulus will be fatigued and will not react to the stimulus as much
as it does before adaptation. The decrease in stimulation is known as an aftereffect. If different
types of images (faces and cars) do not create aftereffects for one another, then the neurons that
process each type of image should be different. With proof that this method shows differentiation
between neural networks, researchers can replicate the study with experts of cars or birds to see
if they process these objects through the face, or expertise, processing network or the object
processing network. Participants were also given an expertise test consisting of same/different
tasks with bird pairs and car pairs to retrieve a d-prime for their selections, a measure for
expertise. Measuring expertise makes sense, because if expertise processing uses the same areas
as face processing, then car experts viewing cars may use the same neural networks as they use
when viewing faces. When adapted to either faces or cars, car experts would fatigue neurons that
are shared between car and face processing, which would cause cross-adaptation. Therefore,
expertise was measured to account for possible cross-adaptation. Furthermore, both car expertise
and bird expertise were measured to see if there is an effect of overall expertise instead of an
effect of car expertise. By accounting for other types of expertise, the effect of overall expertise
can be partialled out during analysis, leaving only the effect of car expertise.
By using known spatial frequency and adaptation methods, new, more definitive results
can be made in order to propel one hypothesis over the other. With the current knowledge in
vision research, there is an abundance of different tasks that can be applied to experiments within
this field, and this test can be the deciding factor to which hypothesis will overcome the other in
this debate.
Method
Participants
CATEGORY SPECIFIC SPATIAL FREQUENCY
12
This experiment included 106 healthy participants from Vanderbilt University and the
surrounding community with normal or corrected vision. These participants were recruited
through Vanderbilt’s research sign-up system, SONA. Twenty participants were excluded due to
being over the age of 39, not completing the experiment, and/or having a d-prime below -0.5 for
bird expertise. D-prime values below this value, according to signal detection theory, were far
below chance on a forced-choice task, so data from participants who scored below this value
were excluded from analysis. The mean age of the eligible participants was 21.8 years old. Fortyone of these participants were male; forty-five were female. Participants were either paid for
their participation or given credit to use to fulfill experimental requirements for undergraduate
psychology courses. They were randomly assigned to one of four groups that differed by
adaptation condition, which created a between-subjects design. The adaptation conditions were
created by making a 2x2 cross with spatial frequency and image type as factors. The four
adaptation conditions (number of participants) were: high spatial frequency faces (22), low
spatial frequency faces (22), high spatial frequency cars (22), and low spatial frequency cars
(21).
Apparatus or Measures
Images – The experiment used pictures of faces and cars. They were grayscaled and sized
with Adobe Photoshop CS5. They were manipulated into low and high spatial frequency images
(<8 cycles per image and >32 cycles per image), being consistent with previous spatial frequency
filtering studies (Gauthier, Curby, Skudlarski, & Epstein, 2005; Schyns & Oliva, 1999). The
images consisted of 40 face pictures and 40 car pictures. The face pictures consisted of 40
individuals (20 males and 20 females), which were selected from the Karolinska face database
(Lundqvist, Flykt, & Öhman, 1998). The car pictures consisted of 40 models (20 sedans and 20
CATEGORY SPECIFIC SPATIAL FREQUENCY
13
convertibles) viewed from the side. After these images were split into two images (HSF and
LSF), 20 male faces were paired with 20 female faces and 20 sedans were paired with 20
convertibles to make LSF/HSF hybrid images that contained both options within a category
(female/male, sedan/convertible), totaling to 40 hybrids within each category (Figure 3 and 4).
The face images were chosen through piloting to see which faces were easily categorized as a
certain sex. The car images were not piloted because they were shown as reliable stimuli in
previous experiments. The other 40 images (20 LSF, 20 HSF) within each category will be used
for adaptation stimuli.
In the expertise test, there were 224 greyscaled bird images and 224 greyscaled car
images (Figures 5 and 6). 112 bird images were separated into 56 pairs of images, each pair
showing a species of birds. The images in each pair show the species of birds at a different
perspective. The other 112 bird images are separated into 56 pairs, each pair showing two
different species of birds. 112 car images are separated into 56 pairs of images, each pair
showing a car model. The images in each pair show the model at a different perspective and can
differ by year of production. The other 112 car images are separated into 56 pairs, each pair
showing two different car models. Lastly, a grayscale mask containing the whole spatial
frequency spectrum (Figure 7) was used during testing.
Design
Participants were randomly placed into one of four groups: High spatial frequency face
adaptation, low spatial frequency face adaptation, high spatial frequency car adaptation, and low
spatial frequency car adaptation. Depending on their group, participants were adapted to a certain
type of image. There were 5 phases to the adaptation experiment: Originals phase, Practice
phase, Pre-adaptation phase, Adaptation phase, and Post-adaptation phase. The pre-adaptation
CATEGORY SPECIFIC SPATIAL FREQUENCY
14
choices for face and car hybrids were compared with the post-adaptation choices to show
differences caused by adaptation. Lastly, participants were given an expertise test that consisted
of same/different judgment of bird pairs and car pairs. Expertise ratings for birds and cars were
used to make correlations between expertise and adaptation effects.
Procedures
After giving informed consent, participants went through the five phases of the
experiment. In the originals phase, participants saw the images similar to what they would see
during testing. They saw 10 unfiltered images of cars and 10 unfiltered images of faces. This
phase allowed participants to become familiar with the stimuli they would be judging. The
images were presented in a steady, continuous stream that was separated by blocks of five
images. Each block contained only one type of image (for example, female faces) and began
with a notification of what would be seen during the block. Participants did not have to make
judgments on these images.
In the practice phase, the participants saw 5 spatial frequency filtered faces and 5 spatial
frequency cars, and they had to decide if they were male faces, female faces, convertibles, or
sedans. This phase accustomed the participants to the judgment task. The image categories
(faces/cars) were separated by block, randomized within each block, and there were four blocks.
Each image was shown twice during this phase. The keys used in this phase (as well as in the
other phases with judgments—pre- and post-adaptation) were “n” for sedans, “m” for
convertibles, “z” female, and “x” for male.
In the pre-adaptation phase, participants saw all of the face hybrids and car hybrids (80
total) which were presented, and they had to decide if they saw male faces, female faces,
convertibles, or sedans. The image categories were separated by block, randomized within each
CATEGORY SPECIFIC SPATIAL FREQUENCY
15
block, and there were four blocks. Each hybrid image was presented twice. A fixation point was
presented before each trial and followed by a 50 ms presentation of a hybrid image which was
then covered with a mask for 50 ms. Participants had an indefinite amount of time to judge the
faces
In the adaptation phase, participants were shown images in regards to their adaptation
condition. They were prompted to pay close attention to the images, and they were shown all of
the adapting stimuli for their adaptation condition. The images were presented continuously in
one block. They were randomized and presented for 2 s each. There was a 200 ms interval
between each stimulus as well.
In the post-adaption phase, they saw all of the face hybrids and car hybrids, and they had
to decide if they saw male faces, female faces, convertibles, or sedans—just as they did in the
pre-adaptation phase. However, between the fixation cross and each trial image, adapting images
within a participant’s adaptation condition were presented for six seconds as a form of readaptation. Within each re-adaptation, four images were presented in the fashion of the
adaptation phase, lasting about six seconds. Each trial was started by the participant pressing the
space bar.
The expertise test asked participants to judge bird image pairs as the same species or not
and to judge car image pairs as the same model or not, irrespective of the production year. The
test consisted of 8 blocks (4 with bird pairs, 4 with car pairs), each containing 28 pairs of images.
The sequence of the image pairs was randomized. Starting with a fixation cross, the images
within a pair were presented sequentially, and participants made a same/different judgment on
the image pair. They used the “s” key to respond “same” and the “d” key to respond “different”.
Results
CATEGORY SPECIFIC SPATIAL FREQUENCY
16
Data Analysis
Each participant’s pre- and post-adaptation phases were compared to test whether
adaptation affected the perception of spatial frequency. The data were reduced to averages across
participants, using the change in LSF responses from pre- to post-adaptation both within their
adaptation image category and outside their adaptation image category. The pre-adaptation
averages were analyzed to determine if there were any pre-existing spatial frequency biases
within image type. The adaptation data (post-adaptation changes from baseline) were compared
between and across groups. Also, the change in LSF response for each participant was matched
with the participant’s results from the expertise test to make scatter plots connecting expertise to
adaptation effects. Both car expertise and bird expertise was measured and analyzed with the
participant’s results using zero-order Pearson-Product correlations and multiple regressions.
Results
The pre-test responses were near chance (as shown in Figures 8 and 9). People adapted to
HSF faces, when tested with faces responded “LSF” 57.4% of the time (SD = 7.3%). When
tested with cars, they responded “LSF” 39.5% of the time (SD = 7.85%). People adapted to LSF
faces, when tested with faces responded “LSF” 54.0% of the time (SD = 8.44%). When tested
with cars, they responded “LSF” 39.9% of the time (SD = 9.00%). People adapted to HSF cars,
when tested with faces responded “LSF” 56.4% of the time (SD = 6.40%). When tested with
cars, they responded “LSF” 37.4% of the time (SD = 11.5%). People adapted to LSF cars, when
tested with faces responded “LSF” 55.0% of the time (SD = 9.19%). When tested with cars, they
responded “LSF” 41.1% of the time (SD = 8.98%).
The differences between the pre-test and the post-test are shown in Figures10 and 11.
Using a single sample t-test for each combination of adaptation condition and test condition (for
CATEGORY SPECIFIC SPATIAL FREQUENCY
17
example, adaptation to HSF faces and test faces), two cases were found to be significant. There
was a significant difference from zero when participants were adapted to HSF faces and tested
with face hybrids, t(21) = -3.175, p = 0.005. There was also a significant difference from zero
when participants were adapted to HSF cars and tested with car hybrids, t(20) = 2.777, p = 0.01.
The t-test data for the rest of the combinations of adaptation conditions and test conditions are
shown in Table 1.
Lastly, when these data are combined with expertise measures, the correlations are not
significant (shown in Table 1). Even after partialling out bird expertise in order to rule out effect
of overall expertise, the results from the adaptation experiment did not hold significant
correlations with car expertise.
Discussion
In this study, there was not adaptation across image type in some of the adaptation
conditions of the experiment. Some differences in LSF responses changed in the intended
direction, but most were not significant changes. Two test conditions produced significant
results. The first significant result occurred when participants were adapted to HSF cars and
tested on car hybrids. These participants reported seeing more LSF cars in the post-test. When
the participants within this adaptation condition were tested with faces, there was no significant
change between pre- and post-test. These results suggest that spatial frequency adaptation was
category specific, because there was an adaptation effect within for cars when adapted to cars,
and this effect did not cross over to faces. The adaptation effect remained within the image
category to which participants were adapted.
In other situations, there was evidence for cross-adapation, but the evidence was not
significant. One case is when participants were adapted to LSF faces and tested on car hybrids.
CATEGORY SPECIFIC SPATIAL FREQUENCY
18
They reported seeing fewer LSF cars in the post-test, which is something that would happen due
to cross-adaptation.
Lastly, in some conditions (one being the second significant result), the opposite of what
was expected occurred. For example, when participants were adapted to HSF faces and tested
with face hybrids, they saw fewer LSF faces during the post-test. The expected result in this
adaptation condition was an increase in the reporting of LSF faces after adaptation, and the
actual result showed the opposite. The same participants did not show adaptation effects for cars,
so even though the adaptation effect was in the opposite direction of what was expected, there
was no adaptation that crossed over image category.
The two significant results show there is no cross-category spatial frequency adaptation.
However, the adaptation effects went in both directions. This conflict of results suggests that cars
and faces may be processed differently. For example, while being adapted to HSF cars causes an
effect in the predicted direction (making more LSF responses when viewing car hybrids), being
adapted to HSF faces causes an effect in the opposite direction. Therefore, while car processing
(or object processing in general) follows traditional adaptation effects, face processing does not.
This difference may be due to levels of expertise. Changes from object processing to expertise
processing has been shown with other effects, such as the inversion effect and holistic
processing, and adaptation may be another case where faces are processed differently from other
non-expertise objects. This difference in processing may cause the non-congruent adaptation
effect for faces and cars.
Along with most of the changes in LSF responses by adaptation condition, the
correlations found from the data were not significant. Therefore, expertise did not significantly
affect the results found in the adaptation experiment. Going along with the earlier suggestion that
CATEGORY SPECIFIC SPATIAL FREQUENCY
19
faces are processed differently than cars (or non-expertise objects), the correlations show
insignificant trends that connect car expertise with different adaptation effects. For example,
when participants are adapted to LSF cars and tested with car hybrids, the opposite of an
adaptation effect occurs with increasing levels of expertise. This suggests that car experts process
LSF cars as they do faces, which leads to the opposite of an adaptation effect. Also, when
participants are adapted to faces in either SF, adaptation effects increase as car expertise
increases. Therefore, when one is a car expert, one experiences a traditional adaptation effect
with faces, which is opposite from what was discussed earlier.
While insignificant, these results suggest that when one is an expert with an object, the
object is processed differently, and if expertise processing is within the FFA, then there may be
competition between different objects of expertise, causing, for example, faces to be processed as
non-expertise objects in car experts. These suggestions are speculative, in that they assume that
behavioral results parallel brain activity, but the results highlight a trend that future studies
should follow.
While some significant results did arise from this study, future experiments of this nature
should revise the design in a few ways. First, the adapting stimuli should include noise from the
other SF. For example, if an adapting stimulus is a HSF face, there should be LSF noise added to
it. This addition would rule out the argument that only lower-level visual areas are being adapted
in this design, meaning that the adaptation effects never reach the category selective areas. By
adding the other SF, participants will have the lower level areas adapted to both SFs while the
higher level area will be adapted to one SF.
Another way to improve this design is to test for face expertise. By testing expertise for
faces, experimenters can gauge participants against each other and make correlations similar to
CATEGORY SPECIFIC SPATIAL FREQUENCY
20
what was found with car expertise. If correlations were found using face expertise, they may
coincide with the speculations about SF biases in expertise, as discussed earlier. Furthermore,
testing expertise for multiple categories of objects may be beneficial. In this study, bird expertise
was used to rule out overall expertise in the correlations, and one may find interactions between
different types of expertise if one tests for expertise in multiple object categories. Continuing
research in this path may further differentiate between the domain-specific hypothesis and the
expertise.
CATEGORY SPECIFIC SPATIAL FREQUENCY
21
References
Addams, R. (1834). An account of a peculiar optical phaenomenon seen after having looked at a
moving body. London and Edinburgh Philosophical Magazine and Journal of Science, 5,
373-374.
Bruce, V, & Humphreys, G. (1994). Recognizing objects and faces. Visual Cognition, 1, 141180.
Cohen, J. (1946). Color adaptation of the human eye. The American Journal of Psychology, 59,
84-110.
Davidoff, J. & Donnelley, N. (1990). Object superiority: A comparison of complete and part
probes. Acta Psychologica, 73, 225-243.
Diamond, R. & Carey, S. (1986). Why faces are and are not special: An effect of expertise.
Journal of Experimental Psychology, 115, 107-117.
Duchaine, B. C., Dingle, K., Butterworth, E., & Nakayama, K. (2004). Normal greeble learning
in a severe case of developmental prosopagnosia, Neuron. 2004, 43(4):469-73.
Farah, M.J., Wilson, K.D., Drain, H.M., & Tanaka, J.R. (1995). The inverted face inversion
effect in prosopagnosia: Evidence for mandatory face-specific perceptual mechanisms.
Vision Research, 35, 2089-2093.
Gauthier, I., Anderson, A. W., Tarr, M. J., Skudlarski, P., & Gore, J. C. (1997). Levels of
categorization in visual recognition studied using functional magnetic resonance imaging.
Current Biology, 7, 645-651.
Gauthier, I., Behrmann, M., & Tarr, M. J. (1999). Can face recognition really be dissociated
from object recognition? Journal of Cognitive Neuroscience, 11, 349-370.
Gauthier, I. & Bukach, C. (2007). Should we reject the expertise hypothesis? Cognition, 103,
CATEGORY SPECIFIC SPATIAL FREQUENCY
22
322-330.
Gauthier, I., Curby, K. M., Skudlarski, P., & Epstein, R. (2005). Activity in the human fusiform
gyrus reflects separate spatial frequency channels. Cognitive, Affective and Behavioral
Neuroscience, 5, 222-234.
Gauthier, I., Skudlarski, P., Gore, J.C., Anderson, A.W. (2000a). Expertise for cars and birds
recruits brain areas involved in face recognition. Nature America, 3, 191-197.
Gauthier, I. & Tarr, M.J. (1997) Becoming a ‘Greeble’ expert: Exploring the face recognition
mechanisms. Vision Research, 37, 1673–1682.
Gauthier, I., Tarr, M.J., Moylan, J., Anderson, A.W., Skudlarski, P., & Gore, J.C. (2000b). Does
visual subordinate-level categorization engage the functionally defined fusiform face
area? Cognitive Neuropsychology, 17, 143-163.
Gauthier, I., Williams, P., Tarr, M. J. & Tanaka, J. W. (1998). Training ‘Greeble’ experts: A
framework for studying expert object recognition processes. Vision Research, 38, 2401–
2428.
Kanwisher, N. (2000). Domain specificity in face perception. Nature, 3, 759-763.
Kanwisher, N., McCermott, J., & Chun, M.M. (1997). The fusiform face area: A module in
human extrastriate cortex specialized for face perception. The Journal of Neuroscience,
17, 4302-4311.
Kanwisher, N., Tong, F., & Nakayama, K. (1998). The effect of face inversion on the human
fusiform face area. Cognition, 68, 1-11.
Lundqvist, D., Flykt, A., & Öhman, A. (1998). The Karolinska Directed Emotional Faces –
KDEF, CD ROM from Department of Clinical Neuroscience, Psychology section,
Karolinska Institutet, ISBN 91-630-7164-9.
CATEGORY SPECIFIC SPATIAL FREQUENCY
23
McKone, E. & Kanwisher, N. (2005). Does the human brain process objects of expertise like
faces? A review of the evidence. In: Dehaene S, Duhamel JR, Hauser M, and Rizzolatti,
editors: From Monkey Brain to Human Brain. MIT Press.
McKone, E., Kanwisher, N., & Duchaine, B.C. (2006). Can generic expertise explain special
processing for faces? Trends in Cognitive Sciences, 11, 8-15.
Rhodes, G. (1988). Looking at faces: First-order and second-order features as determinants of
facial appearance. Perception, 17, 43–63.
Robbins, R. A., & McKone, E. (2007). No face-like processing for objects-of-expertise in three
behavioural tasks. Cognition, 103, 322-330.
Schyns, P., & Oliva, A. (1999). Dr. Angry and Mr. Smile: when categorization flexibly modifies
the perception of faces in rapid visual presentations. Cognition, 69, 243-265.
Tanaka, J.W., & Farah, M.J. (1993). Parts and wholes in face recognition. The Quarterly Journal
of Experimental Psychology, 46, 225-245.
Tanaka, J. W., Kay, J. B., Grinnell, E., Stansfield, B., & Szechter, L. (1996). Face recognition in
young children: When the whole is greater than the sum of its parts. Visual Cognition, 5,
479-496.
Tanaka, J.W., & Taylor, M. (1991). Object categories and expertise: Is the basic level in the eye
of the beholder? Cognitive Psychology, 23, 457-482.
Tong, F., Nakayama, K., Moscovitch, M., Weinrib, O., & Kanwisher, N. (2000). Response
properties of the human fusiform face area. Cognitive Neuropsychology, 17, 257-279.
Valentine, T. (1988). Upside-down faces: A review of the effect of inversion upon face
recognition. British Journal of Psychology, 79, 471-491.
Webster, M. A., Kaping, D., Mizokami, Y., & Duhamel, P. (2004). Adaptation to natural face
CATEGORY SPECIFIC SPATIAL FREQUENCY
24
categories. Nature, 428, 557-561.
Yin, Robert K. (1969). Looking at upside-down faces. Journal of Experimental Psychology, 81,
141-145.
Young, A.W., Hellawell, D., & Hay, D.C. (1987). Configurational information in face
perception. Perception, 16, 747-759.
Yovel, G., Paller, K.A., Levy, J. (2005). A whole face is more than the sum of its halves:
Interactive processing in face perception. Visual Cognition, 12, 337-352.
CATEGORY SPECIFIC SPATIAL FREQUENCY
25
Tables
Adapt
Image
Adapt
SF
Test
Type
T-test
(df)
T-test
p-value
Zeroorder r
Partial
r
Partial
p-value
-0.147
Zeroorder
p-value
0.52
Car
HSF
Car
0.01
Car
HSF
Face
Car
LSF
Car
Car
LSF
Face
Face
HSF
Car
Face
HSF
Face
Face
LSF
Car
Face
LSF
Face
(20)
2.777
(20)
-0.434
(20)
0.239
(20)
0.998
(21)
-0.108
(21)
-3.175
(21)
-1.05
(21)
0.69
-0.07
0.76
0.669
-0.186
0.42
-0.131
0.57
0.81
0.202
0.38
0.231
0.32
0.33
0.074
0.749
0.076
0.749
0.92
-0.181
0.42
-0.132
0.56
0.005
0.232
0.299
0.243
0.276
0.31
0.208
0.35
0.167
0.458
0.51
-0.01
0.96
-0.214
0.338
Table 1
CATEGORY SPECIFIC SPATIAL FREQUENCY
26
Figure Captions
Figure 1 – Example of a low spatial frequency image
Figure 2 – Example of a high spatial frequency image
Figure 3 – Example of a face hybrid
Figure 4 – Example of a car hybrid
Figure 5 – Example of a grayscale bird
Figure 6 – Example of a grayscale car
Figure 7 – Mask used during testing
Figure 8 – Column graph of % LSF face responses by adaptation condition in the pre-test
(includes standard deviation)
Figure 9 – Column graph of % LSF car responses by adaptation condition in the pre-test
(includes standard deviation)
Figure 10 –Column graph of % change in LSF face responses by adaptation condition in the pretest (includes standard error)
Figure 11 – Column graph of % change in LSF car responses by adaptation condition in the pretest (includes standard error)
CATEGORY SPECIFIC SPATIAL FREQUENCY
Figures
Figure 1
27
CATEGORY SPECIFIC SPATIAL FREQUENCY
Figure 2
28
CATEGORY SPECIFIC SPATIAL FREQUENCY
Figure 3
29
CATEGORY SPECIFIC SPATIAL FREQUENCY
Figure 4
30
CATEGORY SPECIFIC SPATIAL FREQUENCY
Figure 5
31
CATEGORY SPECIFIC SPATIAL FREQUENCY
Figure 6
32
CATEGORY SPECIFIC SPATIAL FREQUENCY
Figure 7
33
CATEGORY SPECIFIC SPATIAL FREQUENCY
34
% LSF Face Response by Adaptation Condition
during Pre-Test
65%
60%
% LSF Reponse
55%
50%
45%
40%
35%
30%
HSF Car
LSF Car
HSF Face
Adaptation Condition
Figure 8
LSF Face
CATEGORY SPECIFIC SPATIAL FREQUENCY
35
% LSF Car Response by Adaptation Condition
during Pre-Test
65%
60%
% LSF Response
55%
50%
45%
40%
35%
30%
25%
20%
HSF Car
LSF Car
HSF Face
Adaptation Condition
Figure 9
LSF Face
CATEGORY SPECIFIC SPATIAL FREQUENCY
36
% Change in LSF Face Response by Adaptation
Condition
10%
% Change LSF Face Response
8%
6%
4%
2%
0%
-2%
-4%
-6%
-8%
-10%
HSF Car
LSF Car
HSF Face
Adaptation Condition
Figure 10
LSF Face
CATEGORY SPECIFIC SPATIAL FREQUENCY
37
% Change in LSF Car Response by Adaptation
Condition
10%
8%
% Change LSF Car Reponse
6%
4%
2%
0%
-2%
-4%
-6%
-8%
-10%
HSF Car
LSF Car
HSF Face
Adaptation Condition
Figure 11
LSF Face
Download