Copyright 2003 Psychonomic Society, Inc. 730
Psychonomic Bulletin & Review
2003, 10 (3), 730-737
Most people have had the Aha! experience of insight
many times for trivial problems such as riddles or crossword
clues or when, at long last, they achieve a deep understanding
of a vexing problem. However, because it depends
so heavily on subjective experience, a deeper understanding
of the processes underlying insight has been elusive.
Researchers agree that when trying to solve an insight problem
solvers (1) come to an impasse, perhaps because they
are misled by ambiguous information in the problem (Dominowski
& Dallob, 1995; Smith 1995); (2) often cannot report
the processing that enables them to overcome this impasse
(Gick & Lockhart, 1995; Ohlsson, 1992; Schooler &
Melcher, 1995); and (3) experience their solutions as sudden
and surprising (Bowden, 1997; Davidson, 1995; Metcalfe,
1986a, 1986b; Metcalfe & Wiebe, 1987; Schooler,
Ohlsson, & Brooks, 1993).
The experience of insight has been examined with
feeling-of-knowing ratings, warmth ratings, and intuitions
(Bowden, 1997; Davidson, 1995; Dorfman, Shames, &
Kihlstrom, 1996; Metcalfe, 1986a, 1986b; Metcalfe &
Wiebe, 1987; Seifert, Meyer, Davidson, Patalano, & Yaniv,
1995). These approaches have helped characterize the insight
experience, yet each has the shortcoming of relying
solely on participants’ subjective reports. The present experiment
links participants’ subjective judgments of insight
with an objective measure of processing: specifically,
priming for the solution.
Subjective Experience of Insight
When people attempt to solve noninsight problems,
they generally give gradually increasing warmth ratings,
indicating that they believe they are approaching a solution
(warmth is a measure of how close one feels one is to
reaching a solution); in contrast, when people attempt to
solve insight problems, they report no change in warmth
until immediately before solving the problem, when feelings
of warmth increase dramatically (Metcalfe, 1986a).
This pattern may be interpreted as supporting the ideas
that solving insight problems requires reinterpretation of
these problems and that solving such problems can sometimes
(but not always) be relatively easy once the correct
solution path is chosen (Schooler, Fallshore, & Fiore, 1995).
Furthermore, solvers are better at predicting whether they
will eventually achieve solutions of noninsight problems
than they are at predicting solutions of insight problems
(Metcalfe & Wiebe, 1987), possibly due to initial misinterpretation
of the problem.
Why do problem solvers sometimes have the Aha! experience
of insight upon reaching a solution? Solvers seem
to experience insight when they suddenly overcome an impasse
as a result of some unconscious processing, often
termed
restructuring(for a review, see Dominowski & Dallob,
1995), but this begs the question of how restructuring
occurs. Verbalizing solution attempts can actually impede
Research and writing were supported by NIDCD/NIH Grants R29 DC
02160 and R01 DC 04052 to M.J.B. Data were collected while both authors
were at Rush Medical College, Chicago, IL. Correspondence concerning
this article should be addressed to E. M. Bowden or M. Jung-
Beeman, Department of Psychology, Northwestern University, 2029
Sheridan Rd., Evanston, IL 60208-2710 (e-mail: e-bowden@northwestern.
edu or [email protected]).
Aha! Insight experience correlates with solution
activation in the right hemisphere
EDWARD M. BOWDEN and MARK JUNG-BEEMAN
Northwestern University, Evanston, Illinois
In one experiment, we tested for an association between semantic activation in the right hemisphere
(RH) and left hemisphere (LH) and the Aha! experience when people recognize solutions to insight-like
problems. The compound remote associate problems used in this experiment sometimes evoke an Aha!
experience and sometimes do not. On each trial, participants (
N = 44) attempted to solve these problems
and, after 7 sec, named a target word, made a solution decision, and rated their insight experience
of recognizing the solution. As in prior studies, the participants demonstrated more solution priming
for solutions presented to the left visual field-RH (lvf-RH) than for solutions presented to the right visual
field-LH (rvf-LH). As was predicted, following unsolved problems the participants showed greater
priming for solutions that they rated as evoking an insight experience on the subsequent solution decision
than for solutions that did not evoke an insight experience. This association was stronger for solutions
presented to the lvf-RH than for those presented to the rvf-LH. These results tie the subjective
experience of insight to an objective measure—semantic priming—and suggest that people have an
Aha! experience in part because they already had semantic activation that could lead them to recognize
the solution quickly. We believe semantic activation in both hemispheres cooperatively contributes
to problem solving, but weak solution activation that contributes to the Aha! experience is more likely
to occur in the RH than in the LH.
ACTIVATION AND INSIGHT 731
successful solution of insight problems (see, e.g., Schooler
et al., 1993), perhaps because it causes solvers to focus on
their initial representation (Schooler et al., 1995). If difficulty
in solving insight problems stems from failure to recognize
solution-relevant features, the experience of insight might
arise when those features are suddenly recognized—that is,
when activation of such features suddenly surpasses a conscious
threshold.
In Maier’s (1931) classic study, when people solved insight
problems after an indirect hint was provided, solvers
who reported experiencing their solutions as a sudden insight
did not report any awareness of the hint, whereas
solvers who produced their solutions piecemeal uniformly
reported using the hint to reach the solution. Similarly,
when trying to solve anagrams, solvers’ speed, accuracy,
and subjective experience of solution are all affected by
hint words presented too briefly to be identified, but that
apparently elicit some semantic activation (Bowden, 1997).
Furthermore, incorrect solution attempts are often semantically
related to the correct solution (Bowers, Regehr,
Balthazard, & Parker, 1990), suggesting that solutionrelevant
information was influencing solution attempts.
People might have the Aha! experience when they have
solution-relevant activation below the threshold of awareness
prior to producing the solution.
Solvers appear to have solution-related semantic activation
for insight-like problems they have not solved
(Beeman & Bowden, 2000; Bowden & Beeman, 1998).
Problem solvers manifest solution priming for yet-to-besolved
compound remote associate problems (similar to
some items on the remote associates test; Mednick, 1962),
in which solvers must produce a solution word (e.g.,
sweet)
that can form compounds with each of three problem
words (e.g.,
tooth, potato, and heart). After working on
these problems but failing to solve them, solvers read aloud
solution target words faster than unrelated target words
(Beeman & Bowden, 2000; Bowden & Beeman, 1998). A
somewhat similar approach has shown that feeling-ofknowing
for unanswered fact questions is associated with
subsequent solution priming (Yaniv & Meyer, 1987; but
cf. Connor, Balota, & Neely, 1992).
We propose that people have an Aha! experience when
they
suddenly recognize that some information, which
they have already semantically activated, either is the solution
or points to the solution path. The suddenness suggests
that the solution-related activation was previously
below the threshold of awareness, perhaps overshadowed
by other activation not related to the solution. This account
jibes with the consensus view that insight problems
misdirect
solvers to consider unhelpful information or solution
paths: Some misdirected activation may be stronger
than activation of solution-related concepts. Only when
strong misdirected activation subsides can solutionrelated
activation surpass the threshold of consciousness
and be recognized. In contrast, for noninsight problems,
solvers’ strongest initial activation is probably solutionrelated,
and they simply need to carry out operations to
achieve solution (Schooler et al., 1995).
Most previous studies of the insight experience have involved
insight problems typically fitting the criteria outlined
above and fundamentally defined by the fact that
they evoke an Aha! experience. In these studies, it is then
described how solvers subjectively experience such problems
differently than they would noninsight problems
(i.e., problems that do not evoke insight). We used problems
that sometimes evoke insight and sometimes do not.
We then used insight ratings to assess how prior processing
(i.e., semantic activation) differs for problems that
evoke feelings of insight versus those that do not.
Hemispheric Differences
An intriguing question regarding insight solutions is
whether hemispheric differences in patterns of semantic
activation might interact to foster solution. We are not implying
that either hemisphere alone is responsible for solving
insight problems; rather, we propose that the right
hemisphere (RH) engages in cognitive processes that
specifically facilitate solving such problems.
The literature on hemispheric differences in semantic
priming to linguistic stimuli, and the theory that the RH
engages in relatively coarse semantic coding whereas the
left hemisphere (LH) engages in fine semantic coding,
lead us to predict an RH advantage in solution priming for
insight problems. We provide synopses of the theory and
the evidence below; in-depth reviews are available elsewhere
(e.g., Beeman, 1998; Beeman, Bowden, & Gernsbacher,
2000; Beeman & Chiarello, 1998; Beeman et al.,
1994; Burgess & Simpson, 1988; Chiarello, 1998; Chiarello,
Burgess, Richards, & Pollock, 1990; M. Faust & Chiarello,
1998; Koivisto, 1997; Titone, 1998).
According to the RH coarse semantic coding theory,
soon after encountering a word the RH engages in coarse
semantic coding, weakly and diffusely activating alternative
meanings and more distant associates, whereas the
LH engages in relatively fine semantic coding, strongly
focusing activation on a single interpretation of a word and
a few close or contextually appropriate associates (Beeman
et al., 1994; Burgess & Simpson, 1988; Chiarello et al., 1990;
M. Faust & Chiarello, 1998; M. E. Faust & Gernsbacher,
1996; Koivisto, 1997; Nakagawa, 1991; Titone, 1998).
For comprehension of most direct language, LH fine semantic
coding has a clear advantage. For comprehension
of indirect language such as jokes, metaphors, and inferences,
additional semantic coding by the RH may be necessary
(for a review, see Beeman, 1998). This may account
for RH advantages in priming for target words related to
predictive inferences (Beeman et al., 2000) or related to contextually
inappropriate meanings of ambiguous words in
sentences (Titone, 1998). RH coarse semantic coding may
also account for increased neuroimaging signal in the RH
for comprehension of metaphoric sentences over literal
sentences (Bottini et al., 1994), processing of connected
over unconnected discourse (Robertson et al., 2000), and
integration of themes in untitled texts (St. George, Kutas,
Martinez, & Sereno, 1999). Damage to the RH system may
account for the difficulties that RH-damaged patients
732 BOWDEN AND JUNG-BEEMAN
have in understanding jokes (Bihrle, Brownell, Powelson,
& Gardner, 1986; Brownell, Michel, Powelson, & Gardner,
1983), metaphors (Winner & Gardner, 1977), and connotative
meanings (Brownell, Potter, Michelow, & Gardner,
1984), or in drawing inferences (Beeman, 1993; Brownell,
Potter, Bihrle, & Gardner, 1986).
Similarly, solving insight problems often requires secondary
or tertiary interpretations of words and concepts or
information that initially seems only distantly related to
the problem, and might benefit from RH coarse semantic
processing (Beeman & Bowden, 2000; Bowden & Beeman,
1998; Fiore & Schooler, 1998).
In our previous experiments (Beeman & Bowden, 2000;
Bowden & Beeman, 1998), solvers working on compound
remote associate problems revealed different patterns of
semantic activation in the LH and RH. Solvers initially
show solution priming for targets presented to either visual
hemifield, but after they work on the problem for 7 or
15 sec, solution priming is strongly maintained for lvf-RH
target words but fades for rvf-LH target words. Moreover,
when making solution decisions after 7 or 15 sec of solving
effort, solvers respond more quickly to lvf-RH target
words than to rvf-LH target words. This is a striking result,
given the ubiquitous rvf-LH advantage for responding to
words, and suggests that, for insight problems, solutionrelated
activation in the RH is useful at least for recognizing
solutions, and may play a role in generating them.
EXPERIMENT
The following experiment tests for a relation between
solution-related semantic activation, as indexed by priming,
and feelings of insight that accompany solution
recognition. The participants attempted compound remote
associate problems, named solution (or unrelated) words,
made solution decisions, and then rated the insight experience
of their solution recognition. If solvers experience
insight because existing semantic activation is suddenly
recognized as pointing toward the solution, then they
should show greater solution priming on trials for which
they experience insight for their subsequent solution decisions.
Also, as in prior experiments (Beeman & Bowden,
2000; Bowden & Beeman, 1998), solvers should
manifest stronger solution priming and make faster solution
decisions for lvf-RH target words than for rvf-LH target
words.
Method
Participants
. Forty-four students (31 women and 13 men) at the
University of Wisconsin at Parkside participated in the experiment
for partial course credit. All the participants were strongly righthanded
according to a brief handedness questionnaire and were native
speakers of American English.
Materials
. The problems were 144 compound remote associate
problems (Beeman & Bowden, 2000; Bowden & Beeman, 1998;
Bowden & Jung-Beeman, in press) patterned after some items in the
remote associates test (Mednick, 1962), which have been used to examine
creativity, problem solving, and insight (see Dorfman et al.,
1996). Test problems contained three words, each of which could
form a compound word or phrase with the solution word (e.g.,
palm
/shoe/house—TREE). These problems sometimes evoke feelings
of insight and sometimes do not.
Procedure
. Trials began with a central fixation cross presented
on a computer screen, followed by three problem words presented simultaneously
in horizontal orientation above, at, and below fixation.
The participants tried to produce the solution within 7 sec; after the
time limit, or sooner if they stated the solution, the problem words
were erased, a tone sounded for 250 msec, and the fixation cross
reappeared for 500 msec (total stimulus onset asynchrony [SOA]
5
7,750 msec). Then, a target word was presented horizontally for
180 msec, with the inner edge 1.5º of visual angle from fixation.
1
The target word was replaced by a letter-fragment pattern mask
that remained on the screen for 120 msec. The participants had 3 sec
from the offset of the mask to name (i.e., read aloud) the target word.
Half of the target words were solution words and half were unrelated
words; half of each of these groups was presented to the left visual
hemifield, and the other half to the right. The unrelated target words
were the solutions to problems 72 trials away (e.g., the Problem 1 occurred
with either its own solution or the solution to Problem 73).
The participants saw each target word only once over the course of
the experiment. Across participants, condition of target words was
rotated so that all target words occurred equally often as both solutions
and unrelated words and in both visual hemifields.
After the participants had named the target word, the experimenter
pressed a key to record whether the response was correct. Immediately
following the experimenter’s response, the word
SOLUTION?
was presented, and the participants indicated whether the target
word was indeed the solution (15 participants responded verbally,
14 responded with a right-hand buttonpress, and 15 responded with
a left-hand buttonpress; there were no effects of response mode). Response
time (RT) from the onset of the event until the participant responded
was recorded. Finally, the word
RATING? was presented with
the rating scale of 1–5 underneath.
2 Ample time and care were taken
to describe appropriate ratings to the participants. Examples of the
Aha! experience were described. Awareness of decision processes
(such as using a strategy) was emphasized as a criterion for lowinsight
ratings, and lack of awareness (“I just knew, I don’t know how
I knew”) was emphasized as the criterion for high-insight ratings.
The ratings were further described roughly along the following lines:
A rating of 1 means that at first, you didn’t know whether the word was
the answer, but after thinking about it strategically (for example, trying
to combine the single word with each of the three problem words) you
figured out that it was the answer. A rating of 3 means that you didn’t
immediately know the word was the answer, but you didn’t have to think
about it much either. A rating of 5 means that when you saw the word
you suddenly knew that it was the answer (“It popped into my head”; “Of
course!” “That’s so obvious”; “It felt like I was already thinking that”).
Ratings of 2 and 4 indicate feelings somewhere in between. It is up to
you to decide what rating to give each of your responses. There are no
right or wrong answers.
Twenty-six participants used the above scale, and 18 used the reverse
scale; there were no effects of scale direction. The participants
were tested individually, with their chins positioned in a chinrest so
that their heads were held steady and at a constant distance from the
computer monitor. They were given five practice problems with target
words.
Results
On average, the participants solved 19.4% (
SD 5 6.0)
of the problems within the 7-sec time limit and correctly
named 90.2% (
SD 5 6.8) of the brief, lateralized target
words. The data from 6 participants were replaced: 4 solved
few problems (
.2.5 SDs below the mean—i.e., fewer than
4.75%), 1 had few response latencies recorded (
.2.5 SDs
ACTIVATION AND INSIGHT 733
below the mean—i.e., RTs recorded on fewer than 73% of
the trials), and 1 showed very slow naming latencies and
hyperpriming (
.300 msec of priming, .2.5 SDs above
the mean), suggesting strategic naming. The main effects
and effects replicating earlier results are discussed after
the novel analysis of insight rating data.
Insight ratings and the relation to solution priming
.
The main focus of this experiment was to determine
whether the participants’ feelings of insight, as indexed by
the rating scale, had to do with solution-related activation,
as indexed by solution priming (naming solutions faster
than naming unrelated target words) for problems they
had not yet solved. All trials on which the problem was
solved during the 7-sec time limit were excluded from the
analysis.
3 We examined the relation between insight ratings
that the participants gave for their correct solution decisions
(i.e., hits) and the solution priming manifest when
they named targets on those trials. Only ratings by hemifield
cells for which a participant had two or more valid latencies
were considered. Data from 2 additional participants
were removed because they had fewer than five hits,
and data from 2 others were removed because they assigned
a single rating to nearly all the trials. Figure 1 shows
solution priming in each hemisphere for trials assigned
each insight rating by the remaining 40 participants.
To compare the insight-priming relation across hemispheres,
a Pearson correlation between insight rating
scores and priming
4 was calculated for each hemifield for
each participant. This gave a measure, for each participant,
of whether his or her priming scores in each hemifield
were related to his or her idiosyncratic use of the rating
scale. Correlations were predicted to be low, because
there were few observations per hemifield
3 rating condition
cell and because semantic activation is just one
component contributing to both the RT and the insight rating;
moreover, semantic activation in either hemisphere
could presumably affect insight ratings, whereas priming
was assessed in only one hemifield on each trial. These
correlation coefficients were then
z9-transformed (Cohen
& Cohen, 1983). A positive correlation would reveal that
the participants showed more priming for solution words
that subsequently elicited a feeling of insight for their solution
decisions. The participants’
z9-transformed correlation
coefficients were entered into a
t test, which revealed
that, on average, the participants’ insight ratings
correlated better with solution priming for lvf-RH targets
(average
z95.178, SE50.040) than for rvf-LH targets [average
z
9 5 .048, SE 50.041, t(39) 5 2.3, p , .03].
Because only 8 participants used every insight rating
often enough to have more than two observations per
hemifield
3 rating condition cell, an overall analysis of
variance (ANOVA) was untenable. Given the reliable
hemispheric difference in priming–insight correlations
and a priori hypotheses, paired
t tests were used to compare
rvf-LH and lvf-RH priming at each rating level (for
participants who had data in both cells, leaving
ns of 11,
23, 37, 37, and 27 for ratings of 1, 2, 3, 4, and 5, respectively).
These contrasts revealed a reliable lvf-RH advantage
in solution priming only on trials for which the participants
rated their solution decisions as most insightful
[rating of 5: 40-msec lvf-RH advantage,
t(26) 5 5.2, p 5
.03; rating of 4: 63-msec lvf-RH advantage,
t(36) 5 3.6,
p
, .07; at all other ratings, ts , 1].
Data Replicating Results
of Previous Experiments
Naming latency
. When the participants failed to solve
problems within 7 sec, they named target words presented
to the rvf-LH 21 msec more quickly than target words presented
to the lvf-RH [
F(1,43)5 4.0, p5.05]. See Table 1
for mean naming latencies of the 44 participants. The participants
also showed priming, naming solution target
words 55 msec more quickly than they named unrelated
target words [
F(1,43) 5 45.9, p , .0001]. The participants
showed reliable priming (70 msec) for lvf-RH solution
words [
F(1,43) 5 48.4, p , .0001] and for rvf-LH
solution words [39 msec;
F(1,43)5 11.1, p, .002]. Most
importantly, a reliable target type
3 hemifield3 relatedness
interaction reflected a 33-msec RH advantage in solution
priming [
F(1,43) 5 4.4, p 5 .04].
When the participants solved problems, they named
rvf-LH target words 44 msec more quickly than lvf-RH
Figure 1. Mean priming (naming latency for solution words minus latency for unrelated words, in msec)
by insight rating and hemifield of target word.
734 BOWDEN AND JUNG-BEEMAN
target words [
F(1,43)5 9.8, p , .005; see Table 1]. They
also showed priming, naming solution target words 79msec
more quickly than they named unrelated target words
[
F(1,43)5 18.6, p , .0001]. The participants showed reliable
solution priming (106 msec) for lvf-RH target
words [
F(1,43) 5 19.0, p , .0001] and for rvf-LH target
words [48 msec;
F(1,43)5 5.1, p , .03]. A target type3
hemifield of presentation
3 relatedness interaction reflected
a 66-msec priming RH advantage for solution
priming [
F(1,43) 5 4.7, p , .04].
Solution decision latency
. We examined solution decision
latencies only for trials on which the participants
had already correctly named the target words, and only
following unsolved problems. See Table 2 for mean decision
latencies from 42 participants (data from the 2 participants
with fewer than five hits were removed, as has
been noted above). There was a main effect of response
type: The participants made hit responses (responding
“yes” when the target word was the solution) 245msec more
quickly than they made correct rejections [responding
“no” when the target word was not the solution;
F(1,43),
5.7,
p , .03]. The participants responded 67 msec more
quickly to words presented to the lvf-RH than they did to
words presented to the rvf-LH [
F(1,43)5 2.1 , .16]. Response
type and hemifield of presentation did not interact
[
F(1,43) , 1]. The RH advantage in solution decision latencies,
reliable in previous experiments, was not reliable
here, perhaps because decisions were delayed until after
solvers named the target words and the experimenter
scored their responses.
Solution decision accuracy
. When the participants
correctly named the target words following unsolved
problems, they made subsequent solution decisions as accurately
for lvf-RH targets (85.8%) as for rvf-LH targets
(86.4%;
t, 1; see Table 2). Moreover, a sensitivity analysis
(
d9) revealed that the participants were equally sensitive
for their solution decisions on lvf-RH target words
(
d9 5 2.32, SD 5 .84), and on rvf-LH target words (d9 5
2.37,
SD 5 .81, t , 1). In prior studies (Beeman & Bowden,
2000; Bowden & Beeman, 1998), there was a slight
rvf-LH advantage in decision accuracy. Because the participants
did not first name the target words, those earlier
decision data likely included more lvf-RH trials than rvf-
LH trials on which the participants failed to identify the
target words, given that people typically are better able to
read words presented to the rvf-LH than words presented
to the lvf-RH.
DISCUSSION
After attempting to solve compound remote associate
problems for 7 sec, the participants rated their solution decisions
as more insightful when they had prior semantic
activation of the solution, as indexed by solution priming.
Interestingly, this association between feelings of insight
and solution activation was stronger in the RH than in the
LH. Across insight ratings, the participants showed more
solution priming for lvf-RH than for rvf-LH target words
and made solution decisions more quickly for lvf-RH than
for rvf-LH target words, replicating earlier results (Beeman
& Bowden, 2000; Bowden & Beeman, 1998). In addition,
the current results showed that, on trials in which the
target word was identified, the participants made solution
decisions in the lvf-RH just as accurately as they did in
the rvf-LH.
The current paradigm differs from others employed to
study the subjective experience of insight (Davidson, 1995;
Metcalfe, 1986a, 1986b; Metcalfe & Wiebe, 1987) in a
number of ways. Most importantly, this experiment used
individual ratings of insight experience on a trial-by-trial
basis. This is possible with compound remote associate
problems, which sometimes evoke feelings of insight and
Table 1
Mean Naming Latencies (in Milliseconds) and Standard Errors (
SEs)
Following Solved and Unsolved Problems
Solved Unsolved
rvf-LH lvf-RH rvf-LH lvf-RH
Target Type Mean
SE Mean SE Mean SE Mean SE
Unrelated 730 20 803 27 784 25 821 27
Solution 682 25 697 30 745 21 751 24
Priming 48 106 39 70
Table 2
Mean Solution Decision Latency (and Standard Errors,
SEs) (in Milliseconds) After Naming Target Word
and Accuracy Following Unsolved Problems
rvf-LH lvf-RH
Correct False Correct False
Rejections Alarm Hit Miss Rejections Alarm Hit Miss
Dependent Variable
M SE M SE M SE M SE M SE M SE M SE M SE
Latency 2,152 150 2,778 246 1,942 119 2,630 181 2,114 135 2,907 239 1,833 113 2,821 198
Accuracy (%) 92.3 1.5 80.8 1.5 90.3 1.4 82.0 2.1
ACTIVATION AND INSIGHT 735
sometimes do not. Most importantly, even if the subjective
ratings system employed here is flawed, “soft,” and possibly
inconsistent across participants, these results link the
subjective ratings to an independent measure (priming) of
cognitive processing (semantic activation).
The fact that solvers had insight-like experiences on the
same trials for which they manifested solution priming
supports the position that the Aha! experience reflects, in
part, prior subthreshold activation related to the solution.
That is, solvers feel the Aha! experience of insight when
they suddenly recognize that an already activated concept
is the solution or, in more complex problems, points to the
solution path. For some insight problems, solvers easily
achieve solution once they recognize the solution path; for
other problems, further noninsight processing is necessary
after the critical insight (Dominowski & Dallob, 1995;
Schooler et al., 1995).
In our experiment, the fact that solvers manifest priming
indicates that they had solution-related activation; the
fact that they had not solved the problem indicates that
such activation was below the threshold of awareness.
When the participants saw the solution word, it was immediately
recognized as the solution, leading to an Aha!
experience. Note that the solution priming was measured
when solvers named the solution words (a measure of lexical
activation, although expectancies can also play a role
at this long SOA; see Neely, 1991), so it is not the case
that the insight ratings merely reflected quick solution decisions;
rather, they were associated with an independent
measure of semantic activation for that solution, prior to
solution decision. Solution-related activation previously
below the threshold of awareness could surpass that
threshold, because strong misdirected activation (perhaps
in the LH) may subside. Alternatively, solvers could increase
solution-related activation because they may reinterpret
a problem word or encounter hints or environmental
cues (see, e.g., Bowden, 1997; Maier, 1931).
The results also indicate that subthreshold solution activation
more often occurs in the RH than in the LH, and
that such RH activation is more strongly associated with
the Aha! experience than is LH semantic activation. There
are a few possible interpretations of this hemispheric difference
in the priming–insight association. First, it is possible
that subthreshold solution activation in the LH simply
does not generate a feeling of insight when solvers
achieve or recognize solutions. Second, it is possible that
subthreshold solution activation in either hemisphere can
create a feeling of insight, but that such activation is more
likely to occur in the RH than in the LH. The observed
lack of association between LH solution priming and insight
ratings in this case is probabilistic: In our study, insight
ratings reflected activation in either hemisphere, but
hemifield priming assessed activation primarily in one
hemisphere. If solution activation occurred more frequently
in the RH, as the RH advantage in solution priming
suggests, then on some trials the participants would
experience insight due to RH solution activation but not
show priming for rvf-LH solution targets.
It is hypothetically possible that participants may consciously
generate the solution in either hemisphere (although
we maintain that both hemispheres normally contribute
to solution generation) and take longer to verify
solutions generated by the RH than by the LH. If so, RHgenerated
solutions have a longer window of time during
which participants are aware of the solution prior to responding.
If the solution appears as a target word during
this window, participants may provide a high insight rating
because they were “already thinking that,” partially
fulfilling one of several criteria for assigning an insight
rating. However, it should be noted that the insight ratings
stressed many features of the Aha! experience, and the
general tone implied a lack of awareness prior to the instant
of recognition. Furthermore, we maintain that once
solution activation rises to the point that solvers are aware
of and/or attempting to verify solution candidates, this activation
spreads to both hemispheres.
Thus, we conclude that the RH advantages in solution
priming,
5 solution decision latency, and priming–insight
association all suggest that the RH plays a special role in
solving insight problems and in feelings of insight. These
results are consistent with the fact that hints to insight problems
are more effective when presented to the lvf-RH than
when presented to the rvf-LH (Fiore & Schooler, 1998).
Both the size and the strength of activated semantic
fields may contribute to these phenomena. Large semantic
fields activated by RH coarse semantic coding seem
better able to detect semantic overlap, as has been discussed
elsewhere (for a review, see Beeman, 1998). For instance,
semantic activation in the RH is more sensitive than semantic
activation in the LH to potential connective inferences
(Beeman et al., 2000) and to multiple weakly related
primes (Beeman et al., 1994). For compound remote associate
problems, the LH may activate the solution in connection
with only one problem word and fail to detect the
semantic overlap necessary to achieve solution. Then, this
nonconverging activation decays in the LH, perhaps being
out-competed by misdirected activation, whereas converging
activation in the RH is maintained for a longer
time (Beeman & Bowden, 2000).
Solution-related activation in the RH may easily remain
subthreshold, because large, diffusely activated semantic
fields poorly support selection into awareness. In contrast,
solution-related activation in the LH is likely to exceed
threshold and reach awareness (a process we roughly refer
to as
selection), perhaps with a boost from attention. According
to our theory, this tendency emerges due to small
but strongly activated semantic fields arising from LH
fine semantic coding. The LH could be as adept as the RH
at activating solution-related information, but LH activation
leads to immediate solution, whereas RH activation
does not. Indeed, selecting a solution concept seems quite
important in discovering insight solutions (Davidson,
1995), suggesting that the LH may be necessary for solution
generation. However, in all five of our experiments in
which the participants had 7 or 15 sec to solve compound
remote associate problems, the participants demonstrated
736 BOWDEN AND JUNG-BEEMAN
a reliable (or nearly reliable) RH advantage in solution
priming or in raw solution decision latency after generating
solutions (Beeman & Bowden, 2000; Bowden & Beeman,
1998; present experiment).
We are not arguing that the LH solves noninsight problems
and the RH solves insight problems, or that the LH
is conscious and the RH unconscious. Rather, we argue
that people make conscious decisions influenced by partially
independent activation in each hemisphere. We posit
that action and awareness are supported by population
coding—that is, the summed distributed activity of many
thousands of neurons, without the need of an executive, homunculus,
or grandmother cell. This population is divided
over both hemispheres, and is therefore the sum of two
processing styles arising, perhaps, from slightly asymmetric
neural substrates: relatively fine semantic coding in the
LH and relatively coarse semantic coding in the RH. If the
LH strongly activates a narrow field of information, most
of which will also be weakly activated in the RH, the information
activated in the LH will tend to dominate the
population code underlying selection of information for
consciousness and responses. Contributions from the RH
seem particularly important when people comprehend
discourse in which initially unimportant information becomes
important, such as when they draw some inferences
or understand jokes (for a review, see Beeman, 1998). Because
information can be shared between the hemispheres,
these complementary processes are not strictly isolated
from each other. However, analyzing humans’ integrated
awareness and action into hemispheric components can illuminate
the critical processes and factors of complex behaviors
such as problem solving.
REFERENCES
Beeman, M.
(1993). Semantic processing in the right hemisphere may
contribute to drawing inferences from discourse.
Brain & Language,
44
, 80-120.
Beeman, M.
(1998). Coarse semantic coding and discourse comprehension.
In M. Beeman & C. Chiarello (Eds.),
Right hemisphere language
comprehension: Perspectives from cognitive neuroscience
Beeman, M. J., & Bowden, E. M.
(2000). The right hemisphere maintains
solution-related activation for yet-to-be-solved insight problems.
Memory & Cognition
, 28, 1231-1241.
Beeman, M. J., Bowden, E.M., & Gernsbacher, M. A.
(2000). Right
and left hemisphere cooperation for drawing predictive and coherence
inferences during normal story comprehension.
Brain & Language,
71
, 310-336.
Beeman, M., & Chiarello, C.
(1998). Right hemisphere language
comprehension: Perspectives from cognitive neuroscience
Beeman, M., Friedman, R. B., Grafman, J., Perez, E., Diamond, S.,
& Lindsay, M. B.
(1994). Summation priming and coarse semantic
coding in the right hemisphere.
Journal of Cognitive Neuroscience, 6,
26-45.
Bihrle, A. M., Brownell, H. H., Powelson, J. A., & Gardner, H.
(1986). Comprehension of humorous and non-humorous materials by
left and right brain-damaged patients.
Brain & Cognition, 5, 399-411.
Bottini, G., Corcoran, R., Sterzi, R., Paulescu, E., Schenone, P.,
Scarpa, P., Frackowiak, R. S. J., & Frith, C. D.
(1994). The role of
the right hemisphere in the interpretation of figurative aspects of language:
A positron emission tomography activation study.
Brain, 117,
1241-1253.
Bowden, E. M.
(1997). The effect of reportable and unreportable hints
on anagram solution and the Aha! experience.
Consciousness & Cognition,
6
, 545-573.
Bowden, E.M., & Beeman, M. J.
(1998). Getting the right idea: Semantic
activation in the right hemisphere may help solve insight problems.
Psychological Science
, 9, 435-440.
Bowden, E. M., & Jung-Beeman, M.
(in press). Compound remote associate
problems.
Behavior Research Methods, Instruments, & Computers.
Bowers, K. S., Regehr, G., Balthazard, C., & Parker, K.
(1990). Intuition
in the context of discovery.
Cognitive Psychology, 22, 72-110.
Brownell, H.H., Michel, D., Powelson, J.A., & Gardner, H.
(1983).
Surprise but not coherence: Sensitivity to verbal humor in right hemisphere
patients.
Brain & Language, 18, 20-27.
Brownell, H. H., Potter, H. H., Bihrle, A. M., & Gardner, H.
(1986). Inference deficits in right brain-damaged patients.
Brain &
Language
Brownell, H.H., Potter, H.H., Michelow, D., & Gardner, H.
(1984).
Sensitivity to lexical denotation and connotation in brain-damaged patients:
A double dissociation?
Brain & Language, 22, 310-321.
Burgess, C., & Simpson, G.
(1988). Hemispheric processing of ambiguous
words.
Brain & Language, 33, 86-104.
Chiarello, C.
(1998). On codes of meaning and the meaning of codes:
Semantic access and retrieval within and between hemispheres. In
M. Beeman & C. Chiarello (Eds.),
Right hemisphere language comprehension:
Perspectives from cognitive neuroscience
Chiarello, C., Burgess, C., Richards, L., & Pollock, A.
(1990). Semantic
and associative priming in the cerebral hemispheres: Some
words do, some words don’t, . . . sometimes, some places.
Brain & Language,
38
, 75-104.
Cohen, J., & Cohen, P.
(1983). Applied multiple regression/correlation
analysis for the behavioral sciences
Connor, L. T., Balota, D. A., & Neely, J. H.
(1992). On the relation
between feeling of knowing and lexical decision: Persistent subthreshold
activation or topic familiarity?
Journal of Experimental Psychology:
Learning, Memory, & Cognition
Davidson, J. E.
(1995). The suddenness of insight. In R. J. Sternberg &
J. E. Davidson (Eds.),
The nature of insight (pp. 125-156). Cambridge,
MA: MIT Press.
Dominowski, R. L., & Dallob, P. I.
(1995). Insight and problem solving.
In R. J. Sternberg & J. E. Davidson (Eds.),
The nature of insight
(pp. 33-62). Cambridge, MA: MIT Press.
Dorfman, J., Shames, V. A., & Kilhstrom, J. F.
(1996). Intuition, incubation,
and insight: Implicit cognition in problem solving. In G. Underwood
(Ed.),
Implicit cognition (pp. 257-296). Oxford: Oxford University
Press.
Faust, M., & Chiarello, C.
(1998). Sentence context and lexical ambiguity
resolution by the two hemispheres.
Neuropsychologia, 36, 827-
836.
Faust, M., Kravetz, S., & Babkoff, H.
(1993a). Hemisphericity and
top-down processing of language.
Brain & Language, 44, 1-18.
Faust, M., Kravetz, S., & Babkoff, H.
(1993b). Hemispheric specialization
or reading habits: Evidence from lexical decision research with
Hebrew words and sentences.
Brain & Language, 44, 254-263.
Faust, M. E., & Gernsbacher, M.A.
(1996). Cerebral mechanisms for
suppression of inappropriate information during sentence comprehension.
Brain & Language
, 53, 234-259.
Fiore, S. M., & Schooler, J. W.
(1998). Right hemisphere contributions
to creative problem solving: Converging evidence for divergent
thinking. In M. Beeman & C. Chiarello (Eds.),
Right hemisphere language
comprehension: Perspectives from cognitive neuroscience
(pp. 349-371). Mahwah, NJ: Erlbaum.
Gick, M. L., & Lockhart, R. S.
(1995). Cognitive and affective components
of insight. In R. J. Sternberg & J. E. Davidson (Eds.),
The nature
of insight
Koivisto, M.
(1997). Time course of semantic activation in the cerebral
hemispheres.
Neuropsychologia, 35, 497-504.
Maier, N.R. F.
(1931). Reasoning in humans II: The solution of a problem
and its appearance in consciousness.
Journal of Comparative Psychology,
12
, 181-194.
Mednick, S. A.
(1962). The associative basis of the creative process.
Psychological Review
, 69, 220-232.
ACTIVATION AND INSIGHT 737
Metcalfe, J.
(1986a). Feeling of knowing in memory and problem solving.
Journal of Experimental Psychology: Learning, Memory, & Cognition
,
12
, 288-294.
Metcalfe, J.
(1986b). Premonitions of insight predict impending error.
Journal of Experimental Psychology: Learning, Memory, & Cognition
,
12
, 623-634.
Metcalfe, J. & Wiebe, D.
(1987). Intuition in insight and noninsight
problem solving.
Memory & Cognition, 15, 238-246.
Nakagawa, A.
(1991). Role of anterior and posterior attention networks
in hemispheric asymmetries during lexical decisions.
Journal of Cognitive
Neuroscience
Neely, J. H.
(1991). Semantic priming effects in visual word recognition:
A selective review of current findings and theories. In D. Besner
& J. W. Humphreys (Eds.),
Basic processes in reading: Visual word
recognition
Ohlsson, S.
(1992). Information-processing explanations of insight and
related phenomena. In M. Keane & K. Gilhooley (Eds.),
Advances in
the psychology of thinking (
Robertson, D.A., Gernsbacher, M.A., Guidotti, S. J., Robertson,
R.R.W., Irwin, W., Mock, B. J., & Campana, M.E.
(2000). Functional
neuroanatomy of the cognitive process of mapping during discourse
comprehension.
Psychological Science, 11, 255-260.
Schooler, J. W., Fallshore, M., & Fiore, S. M.
(1995). Epilogue:
Putting insight into perspective. In R. J. Sternberg & J. E. Davidson
(Eds.),
The nature of insight(pp.559-587). Cambridge, MA: MIT Press.
Schooler, J.W., & Melcher, J. (
1995). The ineffability of insight. In
S. M. Smith & T. B. Ward (Eds.),
The creative cognition approach
(pp. 97-133). Cambridge, MA: MIT Press.
Schooler, J.W., Ohlsson, S., & Brooks, K.
(1993). Thoughts beyond
words: When language overshadows insight.
Journal of Experimental
Psychology: General
Seifert, C. M., Meyer, D. E., Davidson, N., Patalano, A. L., &
Yaniv, I.
(1995). Demystification of cognitive insight: Opportunistic
assimilation and the prepared-mind perspective. In R. J. Sternberg &
J. E. Davidson (Eds.),
The nature of insight (pp. 65-124). Cambridge,
MA: MIT Press.
Smith, S. M.
(1995). Getting into and out of mental ruts: A theory of
fixation, incubation, and insight. In R. J. Sternberg & J. E. Davidson
(Eds.),
The nature of insight (pp. 229-252). Cambridge, MA: MIT
Press.
St. George, M. Kutas, M., Martinez, A., & Sereno, M. I.
(1999).
Semantic integration in reading: Engagement of the right hemisphere
during discourse processing.
Brain, 122, 1317-1325.
Titone, D.
(1998). Hemispheric differences in context sensitivity during
lexical ambiguity resolution.
Brain & Language, 65, 361-394.
Winner, E., & Gardner, H.
(1977). The comprehension of metaphor
in brain-damaged patients.
Brain, 100, 719-727.
Yaniv, I., & Meyer, D. E.
(1987). Activation and metacognition of inaccessible
stored information: Potential bases for incubation effects in
problem solving.
Journal of Experimental Psychology: Learning,
Memory, & Cognition
NOTES
1. Hemifield effects with Hebrew stimuli demonstrate that direction
of reading affects neither the basic rvf-LH advantage in reading words
(M. Faust, Kravetz, & Babkoff, 1993b) nor the patterns of semantic
priming across the hemifields/hemispheres (M. Faust, Kravetz, &
Babkoff, 1993a).
2. The rating scale appeared on all trials regardless of whether the participants
indicated that the target word was the solution or was not the solution.
This was done so that the appearance of the rating scale did not
give the participants feedback about the accuracy of their solution decisions.
Only ratings and data from trials on which the participants made
hits (correctly accepting solutions) after failing to solve the problems
were analyzed.
3. It is possible that some participants solved the problem in the interval
between the offset of the problem triad and the onset of the target
word. However, on average, the participants solved 2.1 problems in the
final 750 msec of the 7-sec time limit. Only half of these (i.e., 1 problem)
should be followed by a solution target, and this could not have a
great influence on the average RT. Moreover, such solutions should
occur with equal frequency prior to lvf-RH and rvf-LH target words, and
so cannot explain differential priming in the hemifields.
4. Priming scores at all rating points were derived using the common
baseline of latencies for
all unrelated words within each hemifield, because
insight ratings for unrelated words would not be meaningful.
5. A “ceiling effect” is unlikely to be limiting priming for rvf-LH targets,
given that (1) the participants responded equally fast to solution
words presented in both hemifields, (2) naming times were relatively slow
in comparison with single-word priming paradigms, and (3) rvf-LH advantages
in priming have been documented with other prime types.
(Manuscript received May 2, 2000;
revision accepted for publication July 10, 2002.)
ساحة النقاش