Previous Article | Next Article 
The Journal of Neuroscience, 2000, 20:RC64:1-4
RAPID COMMUNICATION
The Left Hemisphere's Role in Hypothesis Formation
George
Wolford,
Michael B.
Miller, and
Michael
Gazzaniga
Psychology Department, Dartmouth College, Hanover, New Hampshire
03755
 |
ABSTRACT |
In a probability guessing experiment, subjects try to guess which
of two events will occur next. Humans tend to match the frequency of
previous occurrences in their guesses. Animals other than humans tend
to maximize or always choose the option that has occurred the most
frequently in the past. Investigators have argued that frequency
matching results from the attempt of humans to find patterns in
sequences of events even when told the sequences are random. There is
independent evidence that the left hemisphere of humans houses a
cognitive mechanism that tries to make sense of past occurrences. We
performed a probability guessing experiment with two split-brain
patients and found that they approximated frequency matching in their
left hemispheres and approached maximizing in their right hemispheres.
We obtained a conceptual replication of that finding on patients with
unilateral damage to either the left or right hemisphere. We conclude
that the neural processes responsible for searching for patterns in
events are housed in the left hemisphere.
Key words:
split-brain patients; interpreter; decision making; probability matching; maximizing; hypothesis generation
| |
INTRODUCTION |
From
our own experience and from experimental evidence, we know that people
are prone to search for and posit causal relationships among events.
One's relative or friend might remark that she cannot take vitamin X
because it causes a rash. Such relationships are often made on the
basis of scant data and are often false. Yellott (1969)
provided a
particularly striking demonstration of the extent to which people posit
such causal relationships. In his experiment, a light was flashed to
either the left or the right on each trial. Before each trial, subjects
had to predict which of two lights would appear. Subjects participated
in many trials in which the two lights appeared randomly with a
probability that varied across blocks. In most blocks of trials, the
most frequent light appeared with a p = 0.8. Subjects'
predictions matched the frequency of the actual presentations
(frequency matching). In the final block, the experiment changed
without the subjects' knowledge. At that point, the light appeared
wherever the subject predicted it would. In other words, if the subject
predicted the light would appear on the right, it did. If the subject
predicted the light would appear on the left, it did. After 50 trials
of this, Yellott stopped the experiment and asked subjects for their
impressions. Subjects continued to predict the previously most frequent
light 80% of the time during those last 50 trials but overwhelmingly
responded that there was a fixed pattern to the light sequences and
that they had finally figured it out. They proceeded to describe
elaborate and complex sequences of right and left choices that resulted in their responses always being correct. This outcome supports the
contention that subjects had been searching for causal sequences all
along and were fooled into thinking they had succeeded.
In a variety of such guessing experiments, humans typically exhibit
frequency matching. That is if the lights to the two sides are
presented with probabilities p and (1 
p),
the subjects guess the two lights with probabilities p and
(1 p) (Humphreys, 1939; Estes, 1961). The tendency
to match frequency has intrigued investigators because it is a
nonoptimal strategy for this paradigm. Maximizing, or choosing the most
frequent option all of the time, yields more correct guesses than
matching as long as p 
0.5. In other words, if the
red light occurs with a frequency of 70% and a green light occurs with
a frequency of 30%, overall accuracy will be highest if the subject
predicts red all of the time (maximizing). Frequency matching will lead
to correct answers 58% of the time (0.7 * 0.7 + 0.3 * 0.3). Maximizing
will lead to correct answers 70% of the time (0.7 * 1.0 + 0.3 * 0.0).
Interestingly, most other animals maximize in such paradigms (Hinson
and Staddon, 1983). So why do humans choose a less optimal strategy
than rats? Our view is that humans believe there is a pattern, even if
told the sequence is random, and they attempt to figure out the
pattern. Any reasonable pattern hypothesized by the subjects would have
to match frequency if it were to be a correct hypothesis. Perhaps
animals other than humans adopt a more optimal strategy than humans in
this paradigm, because they are not as hindered by the tendency to
search for and posit causal hypotheses.
Gazzaniga (1989, 1995) and Metcalfe et al. (1995) have
hypothesized the existence of an interpreter that plays the role of trying to make sense out of the information that it confronts, in other
words, generating causal hypotheses. Using split-brain patients,
Gazzaniga (1995) provided evidence that this interpreter is
located in the left hemisphere in most individuals. The simultaneous concept test provides an example of the function of the interpreter. In
this task, a split-brain patient is shown a picture exclusively to the
left hemisphere (e.g., a chicken) and another picture exclusively to
the right hemisphere (e.g., a snow scene). The patient is then given an
array of pictures and asked to point to a picture associated with the
presented pictures. In the above example, the left hemisphere chose a
chicken claw, and the right hemisphere chose a shovel. When asked to
explain the choices, the patient responded, "Oh, that's simple. The
chicken claw goes with the chicken, and you need a shovel to clean out
the chicken shed." The right hemisphere is unable to produce speech,
so it cannot explain its selection. The left hemisphere is unaware of
the picture that the right hemisphere is responding to (i.e., the
snow scene), so it must generate its own interpretation of why
the left hand pointed to a shovel. The left hemisphere, observing the
actions of the left hand and right brain, interprets those actions
within the context of what it knows (i.e., a chicken claw) and
generates an explanation for the shovel that is consistent with its
knowledge (Gazzaniga, 1989).
| |
EXPERIMENT 1 |
We hypothesized that the interpreter might be the structure that
underlies the tendency to posit causal explanations and may be
responsible for the frequency matching observed in probability guessing
experiments. To test this, we presented separate probability guessing
experiments to the two hemispheres of two split-brain patients,
predicting frequency matching in the left hemisphere and maximizing in
the right.
| |
Materials and Methods |
J.W. and V.P. were split-brain patients whose corpus callosi had
been severed as treatment for epilepsy. Each participated in five
blocks of 100 trials each. All stimuli were presented and all responses
were collected on a computer. Each trial began with a row of three
arrows (>>>) pointing right or left. The arrows signified which
visual field to make a prediction in. If the arrows pointed to the
right, they were told that either a small green square would be
presented toward the top of the computer screen on the right side or a
small red square would be presented toward the bottom of the screen on
the right side. They were instructed to guess whether the square would
be at the top or the bottom by pressing the appropriate key on the
right side of the keyboard with their right hand. One hundred
milliseconds after their guess, a square was presented 4° to
the right of fixation for 100 msec. The top square was presented with
p = 0.8, and the bottom square was presented with
p = 0.2. If the arrows were pointing to the left,
everything above was reversed. They were instructed to guess whether a
top square or bottom square would be presented in the left visual
field. The probability of a top square in the left visual field was
p = 0.7, and the probability of a bottom square was
p = 0.3. The sequence in the left visual field was
independent of the sequence in the right visual field. All of the
sequences were generated randomly using the random number generator in
the computer language. Both subjects were told to always maintain fixation on the center arrows. V.P.'s eye position was monitored with
an ISCAN tracking system. Feedback on the proportion of correct guesses
was provided at the end of each block of trails.
Results
The results of the first experiment are shown in Figures
1 and 2.
The error bars are based on the assumption of Bernoulli trials. As
predicted, when J.W. and V.P. made guesses about stimuli presented to
the left hemisphere, they came close to matching the frequency of
occurrence of previous stimuli (V.P. undershot frequency matching
slightly). When they made guesses about stimuli presented to the right
hemisphere, they moved steadily toward maximizing, and both were
choosing the more frequent square >20% over frequency matching by the
final block. As an alternative measure of performance, we calculated
the value of criterion from signal detection theory. Taking the most
frequent alternative as the signal, one would expect criterion to be
more liberal as performance approaches maximizing. For J.W., the
criterion in the final block was estimated to be 1.38 in the right
hemisphere with a 95% confidence interval of ±0.065. His criterion in
the left hemisphere was 0.651 ± 0.036. Negative values indicate
a more liberal criterion, so the criterion was substantially more liberal in the right hemisphere. For V.P. the criterion in the right
hemisphere was 1.83 ± 0.257 and in the left hemisphere was
0.31 ± 0.033. Again, V.P.'s criterion was more liberal in the
right hemisphere. The difference between the hemispheres is not evident
immediately but emerges over trials. The likely reasons for this are
presented in Discussion.
View larger version (13K):
[in this window]
[in a new window]
|
Figure 1.
Probability guessing behavior in each hemisphere
of a split-brain patient (J.W.) relative to the past frequency of
presentation. Error bars represent SD based on the assumption of
Bernoulli trials.
|
|
View larger version (13K):
[in this window]
[in a new window]
|
Figure 2.
Probability guessing behavior in each hemisphere
of a split-brain patient (V.P.) relative to the past frequency of
presentation. Error bars indicate SD based on the assumption of
Bernoulli trials.
|
|
| |
EXPERIMENT 2 |
As a conceptual replication to the finding with the split-brain
patients, we carried out a similar paradigm on a series of patients
with unilateral damage to the frontal and prefrontal cortex. Because
the interpreter is assumed to be localized in the frontal and
prefrontal areas of the left hemisphere, we predicted that the patient
with damage to that area would show maximizing as in the right
hemisphere of the split-brain patient. The patients with unilateral
damage to the right hemisphere should have an intact interpreter and
were predicted to show frequency matching.
| |
Materials and Methods |
Five patients participated in a conceptual replication. All five
patients had unilateral, focal lesions to the dorsolateral prefrontal
cortex as a result of stokes, as revealed by high-resolution magnetic
resonance imaging scans (for details on patients' lesions, see Swick
and Knight, 1996). Four of the patients had lesions localized to the
right prefrontal cortex, and one patient had a lesion localized to the
left prefrontal cortex. The ages of the patients ranged from 33 to 79 years old. The one left frontal patient was nonaphasic. The procedure
for these patients was identical in every respect to the procedure used
in Experiment 1.
Results
The results of the replication are shown in Figure
3. The error bars in Figure 3 represent
±2 SD around the mean of the right frontal patients in each block. As
predicted, the four patients with right frontal damage exhibited
choices close to frequency matching, whereas the single left frontal
patient was significantly closer to maximizing. The effect was fairly
clear, because the left frontal patient was closer to maximizing in
every block than any right frontal patient.
View larger version (12K):
[in this window]
[in a new window]
|
Figure 3.
Probability guessing behavior in four patients
with unilateral right hemisphere damage and one patient with unilateral
left hemisphere damage relative to the past frequency of presentation.
Error bars represent ±2 SD around the mean of the right frontal
patients in each block.
|
|
| |
DISCUSSION |
The results from both split-brain patients and from the patients
with unilateral damage to the frontal cortex show approximate frequency
matching in the left hemisphere and movement toward maximizing in the
right hemisphere. The right hemisphere data closely mirror data
produced by animals other than humans. We analyzed the data in terms of
probability relative to frequency matching. Nearly identical patterns
are revealed when using absolute probabilities. In both experiments,
the difference between the hemispheres emerges over trials and is most
evident in the final blocks. This is not surprising, because either
strategy requires information about the nature of the sequences, and it
requires some number of trials to extract that information. The
situation is complicated for the two split-brain patients, because they experience two simultaneous and independent sequences, one in the left
field and one in the right. Furthermore, both are patients with a
history of epilepsy and brain surgery. The frontal patients have
experienced major brain traumas and function reasonably well but with
some loss of cognitive functioning. In both the previous human and
nonhuman animal literature, frequency matching or maximizing emerges
over trials in a relatively gradual fashion.
Why might we try to search for and posit these causal relationships?
Clearly, it would be of great utility to search for causal relationships among events if such relationships existed. From an
evolutionary perspective, finding such relationships may have had
survival value. Those who could uncover simple causal relationships such as determining which caves would stay dry and where the game could
be found might live longer and produce more offspring. Although this
tendency to search for causal relationships has potential benefits, it
can lead to nonoptimal behavior when there is no simple causal
relationship. Some of the common errors in decision making are
consistent with the notion that we are prone to search for and posit
causal relationships even when the evidence is insufficient or even
random. We find that the search for causal explanations appears to be a
left hemisphere activity, consistent with previous research on the interpreter.
| |
FOOTNOTES |
Received Sept. 7, 1999; revised Jan. 3, 2000; accepted Jan. 5, 2000.
This work was supported by National Institute of Neurological Disease
Grants NS 17778-16 and NS 31443-07 and the McDonnell Pew Foundations.
Correspondence should be addressed to George Wolford, Psychology
Department, Dartmouth College, 6207 Gerry Hall, Hanover, NH 03755. E-mail: george.wolford{at}dartmouth.edu.
This article is published in
The Journal of Neuroscience, Rapid Communications Section,
which publishes brief, peer-reviewed papers online, not in print. Rapid
Communications are posted online approximately one month earlier than
they would appear if printed. They are listed in the Table of Contents
of the next open issue of JNeurosci. Cite this article as:
JNeurosci, 2000, 20:RC64 (1-4). The
publication date is the date of posting online at
www.jneurosci.org.
| |
REFERENCES |
-
Estes WK
(1961)
A descriptive approach to the dynamics of choice behavior.
Behav Sci
6:177-184[Medline].
-
Gazzaniga MS
(1989)
Organization of the human brain.
Science
245:947-952[Abstract/Free Full Text].
-
Gazzaniga MS
(1995)
Principles of human brain organization derived from split-brain studies.
Neuron
14:217-228[ISI][Medline].
-
Hinson JM,
Staddon JER
(1983)
Matching, maximizing and hill-climbing.
J Exp Anal Behav
40:321-331[Medline].
-
Humphreys LG
(1939)
Acquisition and extinction of verbal expectations in a situation analogous to conditioning.
J Exp Psychol
25:294-301.
-
Metcalfe J,
Funnell M,
Gazzaniga MS
(1995)
Right-hemisphere superiority: studies of a split-brain patient.
Psychol Sci
6:157-163.
-
Swick D,
Knight RT
(1996)
Is prefrontal cortex involved in cued recall? A neuropsychological test of PET findings.
Neuropsychologica
34:1019-1028.
-
Yellott Jr JI
(1969)
Probability learning with noncontingent success.
J Math Psychol
6:541-575.
Copyright © 2000 Society for Neuroscience 0270-6474/00/$05.00/0
TOP
ABSTRACT
INTRODUCTION
