Download DanielBearMCB105 Research Proposal

Daniel Bear
MCB 105
Research Proposal
Background and Significance
Binocular rivalry occurs in conscious humans and monkeys when different
images are simultaneously presented to each eye (1, 2). If the distinct images cover a
relatively small portion of the visual field, humans will usually see one image for several
seconds, then the other, switching back and forth (1). This perceptual phenomenon was
long thought to involve reciprocal inhibition between neurons with monocular visual
receptive fields, and recent studies have found monocular neurons in the human lateral
geniculate nucleus that only fire when the subject perceives one eye’s image (3);
however, binocular rivalry must also involve image dominance and suppression at higher,
eye-independent levels of perception, as switching the images displayed between the two
eyes several times a second does not change the pattern of each image’s dominance and
suppression times (4). Moreover, some of the binocular neurons in V1/V2, V4, and V5
(MT) fire only when a macaque monkey reports seeing one of the two rivaling images,
and nearly all neurons recorded from in the inferior temporal cortex and superior
temporal sulcus fire in response to a particular percept (2, 5). Thus, as early in the visual
pathway as V1, neurons encode one psychophysical representation of visual space, not
just the physical qualities of the two rivaling visual stimuli.
MT, one of the areas with percept-selective neurons during binocular rivalry, can
also influence the behavioral response to a visual stimulus. Some neurons in MT fire
only when their receptive fields contain motion in a particular direction; when small
groups of cells corresponding to a single direction are stimulated with an electrode,
monkeys report seeing motion in that direction, even if a visual stimulus contains no
overall directional motion (6). This suggests that even cells at this level of the visual
system can characterize a perception enough to drive behavior. If the experience of
binocular rivalry were driven by competition between such perception-encoding neurons,
then these cells would be a likely target for other neural pathways that increase or
decrease the dominance of one of the rivaling images.
The predominance of an image is a measure of the relative amount of time the
subject perceives that image alone during binocular rivalry. With two neutral, equivalent
competing stimuli, such as right-slanted black lines and left-slanted black lines, the
predominance of each image is zero as expected (i.e. neither image is perceived a larger
percentage of the time). Many physical characteristics of the rivaling images, including
contrast, color, and surrounding image, can affect the ratio of times spent perceiving each
image (1). For example, using the images above, as the contrast of one set of lines
increases, this image will be suppressed for less time relative to the fixed-contrast lines
(4). While this shows that an image’s intrinsic qualities can increase its predominance
during ambiguous visual stimulation, relatively few studies have carefully measured the
roles of psychological factors and conditioning on an individual’s perception during
binocular rivalry (7). Unlike selective attention paradigms, which show similar neuronal
firing patterns to binocular rivalry, humans seem to have almost no voluntary control
over the predominance of an image (5); whereas a human can, with practice, spend more
time paying attention to one of two unambiguous stimuli, one cannot usually affect the
predominance of a rivaling image by any behavioral trick (1).
This raises the question of whether involuntary psychological factors can alter the
perceptual pattern of binocular rivalry. Although it is implicitly assumed that state of
Daniel Bear
MCB 105
Research Proposal
mind influences perception, the binocular rivalry paradigm offers a way to directly and
quantitatively measure a controlled effect on well-characterized ambiguous stimuli. Note
that the conflict in binocular rivalry occurs not only at the level of perception and
interpretation, as in Gestalt figures, but also at the level of primary physical stimulus.
While state of mind and prior learning clearly influence the perception of Gestalt figures,
binocular rivalry appears to resist conscious efforts to see one image on command or
indefinitely (1). If conditioning were able to affect the predominance of an image in
binocular rivalry, it might allow an individual to alter his perception of information in
visual and other sensory modalities or access stimuli that had been suppressed by a
lifetime of unintentional conditioning. This might also allow researchers to investigate
the neural mechanisms that strengthen and weaken percept-specific circuits in the visual
system, such as the previously discussed neurons in MT.
Classical and operant conditioning, as two of the simplest and most powerful
mechanisms of altering behavior, may be able to affect the predominance of an image in
binocular rivalry. In this scenario, perceiving one image (e.g. right-slanted black lines)
would be the conditioned response or the reinforced “behavior,” in the cases of classical
and operant conditioning, respectively. Some evidence suggests that when one of the two
rivaling images is more emotional or natural than the other, such as a fearful or rightside-up face versus a neutral or upside-down face, the more “significant” one
predominates (7). These studies may be confounded by the physical nature of the
images, however, since the two faces might not have the same contrast or edge makeup,
factors which can directly affect binocular rivalry. One recent study claims to have
replicated this finding using faces with the same physical qualities (8); nevertheless,
emotional faces have become strong classically conditioned stimuli well before the
experiment, making it difficult to induce changes in binocular rivalry due to conditioning
at the behavioral or neuronal level.
Another study from the same group attempted to circumvent this problem by
conditioning a completely neutral stimulus (vertical or horizontal black lines) with
simultaneous shocks to the arm whenever it appeared (7). This technique very modestly
but statistically significantly increased the predominance of the conditioned image, as
well as its probability of being the first percept during rivaling stimulation. The
methodology of this experiment, however, may not optimize predominance: this study
treated the shock as unconditioned stimulus (UCS) and displaying one image as the
paired conditioned stimulus (CS), so that the real unconditioned response and conditioned
response (CR) should be “fear of pain” rather than “perception of the conditioned image.”
As such, the slightly increased predominance may be the result of emotional priming
rather than true classical conditioning. A more logical experiment would reverse the
UCS and CS, so that displaying one image alone is the UCS, and a conditioned stimulus
such as a shock predicts this image. Then, during binocular rivalry, a shock (the CS)
might evoke a conditioned response, the perception of the selected image. (This will be
discussed more in the proposed methods section; here it points out the flaw in the
previous study and the need for more rational investigation of classical conditioning in
binocular rivalry.)
Daniel Bear
MCB 105
Research Proposal
Specific Aims
The overall goals of the proposed research, then, are to clarify the effects of
classical conditioning on image predominance and to explore the completely
uncharacterized effects of operant conditioning in binocular rivalry. This research would
also search for means to change and measure image predominance at the level of percept
reporting (in humans) and neural activity (in macaque monkeys) as ways of studying the
psychology and neurophysiology of perception. These goals are based on the observation
that emotionally and behaviorally significant images tend to override less significant
visual information (7, 8) and the hypothesis that percept-responsive neurons like those
found in MT are susceptible to behavioral modulation. Because the primate brain can
process and act on only a small fraction of available sensory input, conditioning at the
level of perception itself may be a way for the brain to efficiently reconcile ambiguous
stimuli. The proposed research would therefore be aimed toward answering the
following questions:
1. Humans and monkeys—Can pairing a previously unrelated stimulus with a particular
image cause that image to predominate over a physically equivalent image during
binocular rivalry (classical conditioning)? As discussed, previous studies either have
relied on the lifetime conditioning of specific emotional faces or have used questionable
conditioning paradigms (7, 8). Since the goal is to induce measurable changes in
binocular rivalry through classical conditioning, the proposed research would require the
newly introduced stimuli to be associated with initially neutral images. Slanted lines with
opposite direction would be ideal, even though classical conditioning in general may
have a preference for evolutionarily significant stimuli (7). Since UCR = CR in the
classical conditioning paradigm, the protocol should make “perception of left/right
slanted lines” the conditioned response. This would make presentation of that image in a
non-rivaling setting the unconditioned stimulus, for perception of this image is the natural
response. During training, the UCS would be preceded by what would become the CS.
If the conditioning were effective, presentation of the CS before or during binocular
rivalry would cause the CR image to predominate. This approach has more natural logic
than making the image itself the CS, since it normally evokes a perceptual without any
conditioning. These experiments would determine how much prior pairing of stimuli,
either intentionally or unintentionally, can bias perception of a truly ambiguous stimulus
(i.e. one where there is no natural advantage of one percept over the other.)
2. Humans and monkeys—Is there a way to selective reinforce one percept during
binocular rivalry so that it, like a behavior, begins to predominate over inconsequential
percepts (operant conditioning)? Operant conditioning normally occurs when an
individual learns that a specific behavior predicts, and therefore causes, a reward or a
cessation of punishment. The subject will then perform the behavior more often than
before conditioning. Binocular rivalry does not involve any motor or cognitive behavior;
in fact, perceptual switching seems to be almost completely out of voluntary control (1).
Nevertheless, if one percept were to predict a reward more often than the other, the brain
might reinforce perception itself, without involving a traditional behavior. To test this, a
Daniel Bear
MCB 105
Research Proposal
transient reward, such as a pleasing tactile stimulus, would be given to the subject only
while he reported seeing one of the two images during binocular rivalry. Despite the lack
of voluntary control over perception, this reward would signal to the brain that, when the
eyes see two possible images, only one has meaningful consequences. This method
would require rewards during binocular rivalry, as rewarding the same image under
unambiguous conditions is not true operant conditioning: at the behavioral level,
perceiving one of two possible images is different than simply looking at a fixed image;
and at the cellular level, the two situations evoke different neuronal responses, with some
neurons active when an image is displayed unambiguously but silent during rivaling
perception of that same image (2). Moreover, rewarding an image first under
unambiguous conditions is the same as the “classical conditioning” paradigm of (7), but
using rewards instead of punishments. The hypothesis predicts that if one percept is
rewarded during binocular rivalry, this image will increasingly predominate. (A special
technique, to be discussed in the methods section, would be needed to assure that
perception, not the motor method of reporting, were reinforced in monkeys.) This
operant conditioning is predicted to have a more powerful effect than classical
conditioning, since the former may not rely on evolutionarily relevant stimuli. A positive
result would suggest that an ambiguous image can drive its own preferred perception if
one image is somehow deemed important, even if this involves suppressing another entire
set of visual information.
3. Monkeys only—Where in the visual system are conditioned changes in image
predominance detectable by electrophysiological recording? If the proposed classical
and operant conditioning paradigms cause one image to predominate over the other, this
reinforcement must be encoded somewhere in the visual or perceptual systems. Since
V1, V2, V4, MT, and other temporal areas contain percept-responsive neurons, these cell
populations would be recorded by electrode before and after each type of conditioning.
The goal would be to detect the earliest point at which conditioning had modulated
activity in a percept-dependent fashion, as all downstream regions would be expected to
follow similar patterns. This experiment would hint at where emotional and behavioral
information directly influences the visual system to drive perceptual changes. It would
also provide a list of potential neural targets to stimulate as a way to further condition a
perception or alter a behavioral response. Data could help determine what is truly
suppressed in binocular rivalry: percept-specific neurons, feature specific neurons, or
some combination of both.
The specific aims of the proposed research would elucidate the roles of learning
and emotional memory in perceiving an ambiguous stimulus. While it is interesting that
prior experience can influence perception of Gestalt figures, where the physical stimulus
is unambiguous, it would be even more remarkable if learning could cause the brain to
block out or promote entire channels of incoming sensory information, not just their
higher-order representations. These experiments would also emphasize controlled
changes in perception of neutral stimuli, thereby excluding a lifetime’s worth of
emotional and physical confounding variables.
Daniel Bear
MCB 105
Research Proposal
Research Design
Aim 1: Determining the effect of classical conditioning on binocular rivalry.
Humans: Participants with normal vision and psychological make-up will give informed
consent to experiencing potentially aversive sounds, but no physical pain. Since prior
mindset may affect the strength of conditioning, participants initially will be told only
that they are involved in tests of visual concentration; the full details of the experiment
will be disclosed afterward. A total of six participants will be tested: two unconditioned
controls, two yoked controls, and two for the conditioning experiment.
The rivaling stimuli for these experiments will be right-slanted and left-slanted
black lines, which are assumed to be emotionally neutral. Participants will view a
monitor through a mirror stereoscope as in (7) that will isolate the images presented to
each eye. They will have 20 training trials, in which they will learn to press one key if
they see right-slanted lines and another if they see left-slanted lines; during training, the
same image will be presented to both eyes and the two images will alternate
stochastically every few seconds so as to mimic the experience of binocular rivalry, for a
total of 8 seconds.
Participants will then proceed to the conditioning experiment. 10 initial 8 second
rivalry trials will present one image to the left eye and the other to the right eye to induce
binocular rivalry (the images will alternate eyes every trial). Participants will report
percepts in order to calculate any baseline predominance of one image, given by (time
spent perceiving RIGHT LINES – time spent perceiving LEFT LINES) / (total time
perceiving a single image). During this and subsequent tests, the initial percept will also
be recorded. After the baseline trials, the two unconditioned controls will have another
10 identical trials. The two participants designated for conditioning will have 10 trials
for which the same images are presented to each eye in a pattern identical to the training
trials. In these trials, however, the computer will be programmed to emit a short buzzer
sound every time immediately before one of the two images (the designated UCS)
appears. For one participant, the UCS will be right-slanted lines, for the other left-slanted
lines. The buzzer will therefore act as the CS. The yoked controls will have 10 similar
trials except that the computer will buzz at random times throughout each trial,
uncorrelated with which image is about to appear.
Finally, participants will be given 10 test trials during which rivaling images are
presented to each eye, as in the baseline trials. Immediately before each trial begins, the
computer will buzz (give the CS). Initial percepts and total time spent seeing each image
will be recorded and calculated as predominance. The set of conditioning and test trials
(20 total) will be repeated four times for a total of five experimental blocks. Participants
will be debriefed and predominance analyzed. The hypothesized results are statistically
significant increases in both initial percept and predominance of the conditioned image
during the test trials for the conditioned group, but not the yoked group (which should not
have either image associated with the buzzer). This paradigm is designed to maximize
the CS-CR association during binocular rivalry, unlike the technique in (7) which treats
the image itself as a conditioned stimulus and does not present the associated stimulus (an
electric shock) during binocular rivalry.
Daniel Bear
MCB 105
Research Proposal
Monkeys: Macaque monkeys will be trained to report each image with levers
and tested for accurate reporting with catch trials as in (2). Three monkeys will be
trained, one for each of the conditions in the human experiments. The protocol for
conditioning and testing monkeys will be identical to the one used for humans, using a
monkey-adapted apparatus. The same setup used in (2) will also allow extracellular
recording from individual neurons in V1/V2, V4, MT, and other cortical areas before,
during, and after conditioning (see design for Aim 3).
Aim 2: Finding an effective operant conditioning paradigm in binocular rivalry.
Humans: Six participants will give informed consent as in Aim 1, except that there will
not be any aversive sounds. Two participants will be assigned to each of the control,
yoked, and conditioning groups. The rivaling stimuli, training sessions, and baseline
predominance trials will be identical to those in Aim 1. Every participant will then
choose from a list of sustainable computer-produced sounds the one that is most
rewarding or pleasing to him. Participants will then view rivaling stimuli in blocks of 20
8-second trials. The first 10 trials will differ between groups: for the control group, these
trials will be identical to baseline trials in order to measure any effect of hearing a
pleasant sound before binocular rivalry; for the conditioning group, the computer will
emit the chosen sound with a probability of .5 whenever the participants report seeing
one of the two images (the conditioned percept = right-slanted lines for one person, leftslanted for the other), for as long as this percept lasts; and for the yoked group, the
computer will emit the sound with probability .25 whenever the participants report seeing
either percept. The .5 probability in the conditioning group is thought to maximize the
effect of operant conditioning (variable ratio reinforcement). The last 10 trials of each
block will be identical to the baseline trials for all experimental groups. There will be 5
identical test blocks for a total of 100 trials, with image predominance being measured
separately in the first 10 versus the last 10 trials of each block. The first 10 trials of each
block will determine whether reinforcement during binocular rivalry can skew the
predominance of each image, while the last 10 trials will measure any lasting perceptual
“learning,” even in the absence of the original reinforcer. The hypothesized result is a
significant increase in predominance of the reinforced image for the conditioned group,
especially during the first 10 trials when this image is being actively reinforced.
Monkeys: Monkeys will require a slightly different protocol from humans, as a
reward during binocular rivalry might cause a monkey to errantly report the reinforced
image more often, even if he does not actually perceive it. As in Aim 1, three monkeys
will first be trained to report each of the two rivaling images. The monkeys will also
undergo a second round of training in which they learn not to be distracted by sounds
during binocular rivalry. This will be achieved by ceasing training for good whenever
the monkey is distracted by background sounds, and rewarding the monkey with juice
whenever he completes a “noisy” training session. Monkeys must further be tested to
make sure that they are reporting accurately when there is noise present. This will be
done using catch trials (as in (2)) with noise present and by comparing monkeys’ report
to electrophysiological recording. Once cells are located in V4 and MT that respond
binocularly to a particular image, experimenters will compare the cells’ firing pattern
during binocular rivalry to the monkeys’ report. Since many cells are known to vary with
perceptual patterns (2), these recordings can further validate accurate reporting when
Daniel Bear
MCB 105
Research Proposal
sound is present. The monkeys will then choose a pleasing sound from several choices
by pressing a different button to initiate each sound. A button pressed significantly more
often than the others will be deemed pleasing. This sound will then be used in operant
conditioning trials identical to the ones given to humans, with one monkey assigned to
each of the control, conditioned, and yoked groups. Resulting image predominance will
be calculated in the same way.
Aim 3: Detecting electrophysiological changes in monkeys due to the
reinforcement of a percept. Using the electrophysiological setup of (2), individual
neurons in V1/V2, V4, MT, the inferior temporal lobe, and the superior temporal sulcus
will be recorded from with extracellular electrodes while monkeys view unambiguous
right- or left-slanted lines, one eye at a time. When a cell is found that shows a firing
preference for one image over the other, the electrode will be held in place while the
monkey undergoes either the classical or the operant conditioning protocol. During
binocular rivalry, the electrode and resulting data analysis will measure this neuron’s
correlation with reported percept, mean firing duration, and total percentage of the trial
spent firing. The hypothesis is that if either classical or operant conditioning is
successful at reinforcing one percept, the firing of some image-specific neurons must be
potentiated. The neurons predicted to increase their firing rate and percentage are those
which respond binocularly to a given percept, i.e. they fire regardless of which eye
actually sees the image. These neurons are presumably at a higher level in visual
processing than the monocular channels or percept-independent cells, and they represent
the supposed cellular substrate for behavioral modification of perception. Changes in
firing patterns will be measured in cells at all locations in the visual cortex.
Further Directions
Because previous research has shown that emotional images or patterns associated
with aversive stimuli predominate in binocular rivalry, it would be surprising if this new
classical conditioning paradigm had no effect (7, 8). The next question, then, would be
how much conditioning can influence perception when the brain is presented with
physically equivalent stimuli. With the proposed technique, the conditioned stimulus, a
buzzer, might be substituted with a more evolutionarily relevant aversive stimulus, such
as pain (as in (7)), a human voice, or an image of a predator preceding binocular rivalry.
Even more natural would be pairing an evolutionary CS, such as a snake, with a logical
image in binocular rivalry, such as a fearful face. Since one group has used faces with
equivalent characteristics to study binocular rivalry, this would be possible without
confounding physical variables (8). The next step would be to see if classical
conditioning could actually override visual qualities that normally bias binocular rivalry
toward one image, such as higher contrast (2).
Since the proposed operant conditioning protocol is novel to the study of
binocular rivalry, it is possible that a reinforcing stimulus would either have no effect on
image predominance or would distract from image viewing altogether. If this were the
case, the proposed research’s focus would shift to the classical conditioning experiments,
possibly replacing aversive stimuli with rewards for perceiving an image before the onset
Daniel Bear
MCB 105
Research Proposal
of binocular rivalry. Since the brain is normally capable of integrating many sensory
stimuli, however, it seems more likely that a reward would potentiate relevant visual
information. Other rewards could be tested, including pleasing tactile stimuli, food
rewards, and even psychological encouragement (“you’re doing very well”).
Punishments could also be tested using the same protocol for their ability to suppress one
percept. The overall goal would be to study whether the brain treats involuntary
perception as it would behavior, even when it means ignoring a coherent visual stimulus.
Since both images are processed in binocular rivalry, the brain might alter perception at
relatively early stages of the visual pathway under conditions where one image has
proven itself important. If the electrophysiological experiments detected activity changes
in V1/V2, for instance, they could help explain why image switching in binocular rivalry
is completely involuntary, as opposed to selective attention or viewing Gestalt figures:
the physical information of one image might be suppressed too early in visual processing
for voluntary cognitive processes to affect it.
Finally, if changes in V1/V2, V4, and MT neuronal activity accompany
conditioned changes in perception as expected, the next step would be to find out which
other brain areas are modulating the visual cortex. The amygdala has known inputs to the
visual regions of the brain that would likely mediate the classical conditioning and
emotional effects on binocular rivalry, so it would be important to determine whether this
pathway is active at all times or only when the conditioned image is perceived during
binocular rivalry. If it is active whenever the emotion-associated image is presented (i.e.
always during binocular rivalry), the overall parameters of percept switching might be
expected to change without one image predominating strongly; for instance, percept
duration might increase for both competing images. On the other hand, if the emotional
pathway only acts when that image is being perceived or only influences percept-specific
neurons, that image might predominate dramatically as seen in (8). The neural pathways
that could influence the proposed operant conditioning in perception are unknown; follow
up experiments would test the midbrain dopamine system for any influence, possibly by
direct electrical stimulation. This pathway is largely responsible for traditional operant
conditioning, so it might act by a parallel mechanism on perception of ambiguous visual
stimuli.
References
1. R. Blake, N.K. Logothetis, Nat. Rev. Neurosci. 3, 13 (2002).
2. D.A. Leopold, N.K. Logothetis, Nature 379, 549 (1996).
3. J.D. Haynes, R. Deichmann, G. Rees, Nature 438, 496 (2005).
4. N.K. Logothetis, D.A. Leopold, D.L. Sheinberg, Nature 380, 621 (1996).
5. D.L. Sheinberg, N.K. Logothetis, Proc. Natl. Acad. Sci. 94, 3408 (1997).
6. C.D. Salzman, K.H. Britten, W.T. Newsome, Nature 346, 174 (1990).
7. G.W. Alpers et al., Int. J. Psychophys. 57, 25 (2005).
8. G.W. Alpers, A.B. Gerdes, Emotion 7, 495 (2007).