Cognition 130 (2014) 174–185

Contents lists available at ScienceDirect

Cognition
journal homepage: www.elsevier.com/locate/COGNIT

Working Wonders? Investigating insight with magic tricks

Amory H. Danek a,⇑, Thomas Fraps b, Albrecht von Müller c, Benedikt Grothe a,
Michael Öllinger c,d
a
Division of Neurobiology, Department Biology II, Ludwig-Maximilians-Universität München, Grosshaderner Straße 2, 82152 Planegg-Martinsried, Germany
b
Trick 17 magic concepts, Neureutherstr. 17, 80799 Munich, Germany
c
Parmenides Foundation, Kirchplatz 1, 82049 Munich, Germany
d
Department Psychology, Ludwig-Maximilians-Universität München, Leopoldstr. 13, 80802 Munich, Germany

a r t i c l e i n f o a b s t r a c t

Article history: We propose a new approach to differentiate between insight and noninsight problem solv-
Received 27 November 2011 ing, by introducing magic tricks as problem solving domain. We argue that magic tricks are
Revised 29 August 2013 ideally suited to investigate representational change, the key mechanism that yields sud-
Accepted 4 November 2013
den insight into the solution of a problem, because in order to gain insight into the magi-
cians’ secret method, observers must overcome implicit constraints and thus change their
problem representation. In Experiment 1, 50 participants were exposed to 34 different
Keywords:
magic tricks, asking them to find out how the trick was accomplished. Upon solving a trick,
Insight problem solving
Aha! experience
participants indicated if they had reached the solution either with or without insight.
Constraint relaxation Insight was reported in 41.1% of solutions. The new task domain revealed differences in
Magic solution accuracy, time course and solution confidence with insight solutions being more
Representational change likely to be true, reached earlier, and obtaining higher confidence ratings. In Experiment
2, we explored which role self-imposed constraints actually play in magic tricks. 62 partic-
ipants were presented with 12 magic tricks. One group received verbal cues, providing
solution relevant information without giving the solution away. The control group received
no informative cue. Experiment 2 showed that participants’ constraints were suggestible to
verbal cues, resulting in higher solution rates. Thus, magic tricks provide more detailed
information about the differences between insightful and noninsightful problem solving,
and the underlying mechanisms that are necessary to have an insight.
Ó 2013 Elsevier B.V. All rights reserved.

1. Introduction nature remains elusive (see Chu & MacGregor, 2011, for a
review).
Sometimes, genius strikes. This moment of sudden The feeling of suddenly knowing the solution to a diffi-
comprehension is known as insight and ‘‘is thought to arise cult problem is generally accompanied by a strong affec-
when a solver breaks free of unwarranted assumptions, or tive response, the so-called Aha! experience, and a high
forms novel, task-related connections between existing confidence that the solution is correct (Sternberg &
concepts’’ (Bowden, Jung-Beeman, Fleck, & Kounios, 2005, Davidson, 1995). Furthermore, insight is thought to be clo-
p. 322). Insightful problem solving is a fundamental think- sely linked to processes that restructure the mental repre-
ing process and nearly one century of psychological re- sentation of a problem (Duncker, 1945; Kaplan & Simon,
search has been dedicated to demystifying it, yet its true 1990; Ohlsson, 1992). More specifically, the representa-
tional change theory (RCT, Knoblich, Ohlsson, Haider, &
Rhenius, 1999; Ohlsson, 1992) assumes that prior knowl-
⇑ Corresponding author. Tel.: +49 89 21086470; fax: +49 89 edge and inappropriate assumptions result in self-imposed
452093511. constraints that establish a biased representation of the
E-mail address: [email protected] (A.H. Danek). problem and thus prevent a solution. The RCT postulates

0010-0277/$ - see front matter Ó 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.cognition.2013.11.003
 A.H. Danek et al. / Cognition 130 (2014) 174–185 175

the process of constraint relaxation as one possibility to (2) Magic tricks activate constraints that determine the
change the biased problem representation, i.e. the over- problem solving process (constraint hypothesis).
constrained assumptions must be relaxed. For example, (3) The time course of the two types of problem solving
in Katona’s Triangle Problem (1940), participants were is different, with insight solutions reached earlier
asked to build four equilateral triangles with only six (time course hypothesis).
matchsticks. The problem cannot be solved if a two- (4) Participants’ confidence in the correctness of their
dimensional representation is used, this being the typical solution differs between insight and noninsight
approach of most problem solvers. It is necessary to over- problem solving (confidence hypothesis).
come the self-imposed ‘‘two-dimension’’ constraint and
search for a three-dimensional solution, that is, by building 1.1. Magic tricks as a new insight task
a tetrahedron.
First empirical evidence for constraint relaxation was The ancient art of conjuring could perhaps be called
provided by Knoblich et al. (1999; see also Knoblich, Ohls- ‘‘applied psychology’’ in the sense that magicians system-
son, & Raney, 2001; Öllinger, Jones, & Knoblich, 2008) who atically exploit the limitations of human visual perception
found that the degree of necessary constraint relaxation and attention. Magicians deliberately evoke inappropriate
was mirrored in the differential difficulty of individual constraints that hinder the observer from seeing through
problems. However, it remains difficult to directly demon- the magic trick. The experiment begins when the curtain
strate that constraints were actually relaxed or even that is raised – and, just as any skilled experimenter, the magi-
they existed before the problem was solved, because the cian keeps improving his methods from performance to
sources of difficulty of a specific problem are often un- performance based on the data (feedback) that is provided
known or highly variegated. The classical 9-dot problem by the audience’s reactions.
is one example for a problem with several different sources Historically, psychologists’ attempts to link magic and
of difficulty (Kershaw & Ohlsson, 2004; Öllinger, Jones, & psychology date as far back as the 19th century (Jastrow,
Knoblich, 2013b). 1888). More recently, it has been suggested that magic
In the past, researchers have confined themselves to techniques could be adopted as research tools for cognitive
investigating insight problem solving mostly in the frame- science and first studies have already been published in the
work of a small set of insight problems. Reviewing the field of visual attention with special magic tricks as stimuli
tasks available so far, MacGregor and Cunningham (2008) (e.g. Kuhn, Kourkoulou, & Leekam, 2010; Kuhn & Land,
identified a need for new sources of insight problems and 2006; Kuhn & Tatler, 2005; Kuhn, Tatler, Findlay, & Cole,
suggested rebus puzzles as one potential addition. Another 2008; Parris, Kuhn, Mizon, Benattayallah, & Hodgson,
relatively new set of problems, already widely used, are 2009). These studies demonstrate how magic tricks can
compound remote associate problems (e.g. Bowden & be utilized to learn more about human visual perception
Jung-Beeman, 2003, Sandkühler & Bhattacharya, 2008, and attention (see Kuhn, Amlani, & Rensink, 2008, for a
adapted from the Remote Associates Test by Mednick, thorough discussion).
1962). However, like so many classical problem solving In the present study, we take this one step further by
tasks, both of these are restricted to verbal material and presenting magic tricks and asking participants to find
rely on access to an answer that is already stored in mem- out how the trick worked, i.e. which method was used by
ory (the solution word) rather than on the generation of a the magician to create the magic effect. We assume that
truly novel solution. In the spatial domain, matchstick if people overcome the over-constrained problem repre-
arithmetic tasks (Knoblich et al., 1999) are an important sentation induced by the magician and find the ‘‘solution’’
and relatively new contribution. Still, although the use of of a magic trick, they will experience insight. We see two
these tasks has brought forward fruitful results (e.g. main reasons for that assumption:
Knoblich et al., 2001; Öllinger, Jones, & Knoblich, 2006), it First, similar to classical insight problems (Weisberg,
seems appropriate to take a more unconventional ap- 1995), the domain of magic tricks activates self-imposed
proach beyond the current problem domains to better constraints (Ohlsson, 1992; Öllinger & Knoblich, 2009). Be-
understand insight problem solving. sides sleight of hand, many magic tricks exploit implicit
Here, we propose a new task domain: Magic tricks. We assumptions of the spectator as part of their methods
suggest that this more applied domain allows new (e.g. if someone performs a throwing motion the spectator
approaches to reveal the underlying mechanisms (e.g. con- expects that he will throw something). The magician ben-
straint relaxation) by manipulating particular knowledge efits from the fact that these constraints are activated
aspects in a broader and more natural way than in artificial highly automatically and that it is almost impossible to
geometrical or verbal puzzles. We assume that this new overcome them (Tamariz, 1988). Consequently, the subjec-
domain provides generalized results that show that tive search space (Kaplan & Simon, 1990; Newell & Simon,
insightful problem solving is a special type of thinking that 1972) for possible explanations of an observed trick is
can be clearly demarcated from other, more analytical ways fairly constrained. In contrast to insight problems, magic
of problem solving (see Weisberg & Alba, 1981; Öllinger & stimuli do not consist of a riddle or a puzzle, but instead
Knoblich, 2009). Specifically, we hypothesize that: the problem is consolidated by the discrepancy between
the observed event with unexpected outcome (Parris
(1) Magic tricks allow to differentiate between insight et al., 2009) and the prior knowledge activated by such
and noninsight problem solving (process an apparently familiar event. This discrepancy often leads
hypothesis). the magician’s audience into an impasse – a state of mind
176 A.H. Danek et al. / Cognition 130 (2014) 174–185

in which people are completely puzzled and have no idea achieve the magic effect might be conceivable (Tamariz,
how this magic effect could possibly have taken place. To 1988). In fact, almost every conjuring effect can be
overcome such an impasse and find the solution, the achieved by several different methods, for example, Fitzkee
over-constrained assumptions must be relaxed (RCT, Fleck compiled a list of possible methods for 19 basic effects that
& Weisberg, 2013; Jones, 2003; Kershaw, Flynn, & Gordon, comprises 300 pages (Fitzkee, 1944, quoted according to
2013; Kershaw & Ohlsson, 2004; Ohlsson, 1992; Thevenot Lamont & Wiseman, 1999, p. 7). Which method the con-
& Oakhill, 2008; Öllinger, Jones, Faber, & Knoblich, 2013a; jurer applies depends on the individual strengths of each
Öllinger, Jones, & Knoblich, 2013b; Öllinger et al., 2008), as method and on the performing situation (e.g. large vs.
outlined before. small audience). Participants might find the true solution,
Second, a magic trick can be considered as a highly but might perhaps also come up with another plausible
intriguing problem, which strongly motivates the observer solution or alternatively, a solution that is actually impos-
to find a solution. Observing something impossible hap- sible (given the information from the video clips), i.e. a
pening right in front of your eyes poses a challenge for your false solution.
rationality, and therefore, after the first sensation of won- An example of a magic trick illustrates our account
der and astonishment has passed, the situation is critically (trick #20, see Appendix A. The full video clip can be found
analysed. Anyone who has ever witnessed a magic perfor- at http://www.youtube.com/watch?v=3B6ZxNROuNw). A
mance, will remember the strong desire to know how the coffee mug and a glass of water are presented to the audi-
magic effect is achieved (the usual response is ‘‘Let me ence. The magician pours water into the mug, as depicted
see that again!’’). Of course, magicians rarely offer such in Fig. 1a. Holding the mug with his arms stretched, the
second chances, but that is exactly what we did in the magician snaps his fingers – then he turns the mug upside
present work. down and a large ice cube drops out (Fig. 1b). In a few sec-
We infer from the first point that it might be possible to onds, the water has turned into ice. How does this work?
gain sudden insight into the inner working of a magic trick Most people react with astonishment and disbelief be-
by relaxing self-imposed constraints (constraint hypothe- cause according to their prior knowledge, this is not possi-
sis). This does not exclude that tricks can also be solved ble (Parris et al., 2009). Water can turn into ice, but not in
in a more analytical and step-wise way, as also discussed such a short time period (at room temperature), and addi-
in classical insight problems (Evans, 2008; Metcalfe, tionally, it does not turn into a perfect ice cube by itself.
1986; Weisberg, 1995), e.g. by systematically thinking Seemingly, causal relationships and laws of nature that
through different solution possibilities. In this case, we as- were acquired through past experience have been violated
sume that the solving process will take longer and that the (Ohlsson, 1992; Parris et al., 2009). An artful magician in-
solution will not be experienced as ‘‘sudden’’ anymore duces the impression that he controls the natural laws in
(time course hypothesis). To differentiate between these a supernatural way and can bend them as he wishes. Be-
two solving processes, we will use the subjective Aha! sides astonishment, the spectator is faced with the open
experience as a classification criterion to differentiate be- question of how the magician did the trick: A problem is
tween insight solutions (solutions accompanied by an consolidated that must be solved. In the subsequent prob-
Aha!) in contrast to noninsight solutions (solutions without lem solving process, the situation is analysed, setting up
Aha!). That is, we adopted the common approach (e.g. the initial problem representation. Due to observers’ prior
Jarosz, Colflesh, & Wiley, 2012; Jung-Beeman et al., 2004; knowledge, this representation is often biased and over-
Kounios et al., 2006, 2008) introduced by Bowden and constrained (Knoblich et al., 1999). Wrong assumptions
Jung-Beeman (2007) and Bowden et al., (2005) of asking turn into constraints that restrict the search space and pre-
participants directly if they had experienced an Aha! or vent a solution. In the example trick, the following
not. In addition, we assessed participants’ feeling of confi- assumptions are skilfully evoked by the magician:
dence for each solution, expecting that insight solutions
would differ from noninsight solutions with regard to
these ratings (confidence hypothesis).
For our experimental rationale, it is important to note
that each magic trick consists of an effect and of a method
(Ortiz, 2006; Tamariz, 1988). The magic effect is what the
observer perceives (e.g. a coin vanishes) and the method
is how the trick works, the secret behind the effect (e.g.
skill, mechanical devices, misdirection). Conjurers employ
a method to produce an effect (e.g. Lamont & Wiseman,
1999). Typically, the magician tries to guide the spectators’
attention away from the method and towards the effect. In
the present study, participants experienced the effect and
were then asked to discover the method.
A second important point to consider is that in contrast
to most verbal puzzles or riddles, magic tricks do not have
one clear unambiguous solution. Of course, for each magic
trick, there exists one true solution, that is, the method that
was actually used by the magician. Still, other methods to Fig. 1a. Screenshot from the beginning of the trick.
 A.H. Danek et al. / Cognition 130 (2014) 174–185 177

2. Experiment 1

2.1. Method

2.1.1. Participants
50 healthy volunteers, most of them students (mean age
24.4 ± 3.3; 16 male), were recruited through announce-
ments at the University of Munich and were paid 32 € for
their participation. None of them had any neurological dis-
eases and all had normal or corrected-to-normal acuity.
Two participants were excluded because they did not solve
any of the presented tricks resulting in a final sample size of
48.

2.1.2. Testing material

With the aid of a professional magician (TF), a careful
Fig. 1b. Screenshot from the end of the trick.
pre-selection of magic tricks was conducted with regard
to sensory as well as cognitive requirements: Only visual
1. The mug and the glass are real, ordinary objects. effects that could be performed in absolute silence, with
2. The water is real water. no other interactive elements necessary (e.g. assistant,
3. The mug is empty. interaction with the audience). We used short tricks, with
4. The water is poured into the mug. only one effect and one method. 40 magic tricks were se-
5. It is a real ice cube. lected and recorded in a standardized setting, again with
6. There is no water left in the mug after the ice cube has the magician TF. We ran three pilot studies on a sample
fallen out. of 50 students to ensure that the tricks were understand-
able, i.e. that participants perceived and were able to re-
Some of these assumptions may be correct, but others port the intended magic effect. Tricks were also rated
are wrong, and these are the crucial assumptions that cre- with regard to the extent of surprise that they caused
ate the magic effect. They have become constraints that re- (see Appendix A). Three tricks that turned out to be not
strict the search space for a solution. The constraints have feasible for a filmed performance were removed, and 17
to be relaxed to attain a broader search space that includes tricks had to be improved in a second recording session
the solution. (e.g. better camera angle). The final number of stimuli
In the present example, only the third assumption is was 34 (plus 3 practice trials). The video clips that ranged
wrong. The ‘‘empty’’ mug is actually filled with a piece of from 6 to 80 s were presented on a 17’’ computer screen
special white napkin, glued to the bottom of the mug, displayed by the PresentationÒ software version 12.1. The
and the ice cube. Because the inner side of the mug is also tricks covered a wide range of different magic effects
white, the observer can neither detect the napkin nor the (e.g. transposition, restoration, vanish) and methods (e.g.
transparent ice cube if the mug is kept in motion while misdirection, gimmicks, optical illusions) and are listed in
casually showing it empty. The water is indeed poured into detail in Appendix A.
the mug, but is fully absorbed by the napkin. And voilà,
only the ice cube falls out when the mug is turned upside 2.1.3. Design and procedure
down – it’s magic! Participants were seated in a distance of 80 cm in front
We argue that if the observer achieves to overcome the of a computer screen. After filling in an informed consent,
initial constraint (empty mug), his search space is restruc- participants were orally instructed by the experimenter.
tured (Kaplan & Simon, 1990; Öllinger et al., 2013b) and Their task was to watch magic tricks and to discover the
new solution possibilities are opened up allowing him to secret method used by the magician. Following Bowden
find the correct solution (napkin) or to think of other pos- and Jung-Beeman’s approach (2007), participants were
sibilities to contain the water (e.g. double bottom). asked to categorize their solution experiences into insight
Taken together, we claim that a magic trick can be re- and noninsight solutions. The instruction for these judge-
garded as a challenging problem, and that the spectator ments (in German) read as follows (adapted from
takes the role of a problem solver who attempts to find Jung-Beeman et al., 2004): ‘‘We would like to know
explanations for the magic effect. whether you experienced a feeling of insight when you
In Experiment 1, we implement the new domain of ma- solved a magic trick. A feeling of insight is a kind of
gic with the aim of differentiating between two types of ‘‘Aha!‘‘ characterized by suddenness and obviousness. Like
problem solving processes, namely insight vs. noninsight an enlightenment. You are relatively confident that your
problem solving (process hypothesis). If magic tricks actu- solution is correct without having to check it. In contrast,
ally trigger these two types of problem solving, we expect you experienced no Aha! if the solution occurs to you
differences between them with regard to time course (in- slowly and stepwise, and if you need to check it by watch-
sight solutions reached earlier, time course hypothesis) ing the clip once more. As an example, imagine a light bulb
and subjective appraisal (confidence hypothesis). that is switched on all at once in contrast to slowly
178 A.H. Danek et al. / Cognition 130 (2014) 174–185

dimming it up. We ask for your subjective rating whether (Danek, Fraps, von Müller, Grothe, & Öllinger, 2013). Fur-
it felt like an Aha! experience or not, there is no right or thermore, in each session, an additional quantitative and
wrong answer. Just follow your intuition.’’ The experi- qualitative assessment of participants’ individual Aha!
menter interacted with participants until they felt pre- experiences was conducted after the end of the experi-
pared to differentiate between these two experiences. ment. This data is reported in Danek, Fraps, von Müller,
After three practice trials, a randomized sequence of 34 Grothe, and Öllinger (submitted for publication), but not
magic tricks was presented. If a trick was solved, partici- relevant for the present analysis and therefore not consid-
pants had to indicate on a trial-by-trial basis whether they ered further.
had experienced an Aha! during the solution. If partici- Trick repetition: In general, magic tricks are fairly diffi-
pants failed to solve the trick, the video clip was repeated cult. To increase solution rates, each trick was repeated up
up to two more times while solving attempts continued. to three times, thereby breaking the old magicians’ rule:
As soon as they had found a potential solution, partici- Never show the same trick twice! For the reader interested
pants were required to press a button. The button press in magic, please consult Lamont, Henderson, and Smith
stopped the video clip and terminated the trial. A dialog (2010) for a critical discussion of that point. First evidence
with the following question appeared: Did you experience that trick repetition increases the likelihood of detecting
an Aha! moment? Participants indicated Yes or No with a the method was provided by Kuhn and Tatler (2005). In a
mouse click. Subsequently, they were prompted to type pilot study, we confirmed this finding and could show that
in their solution on the keyboard and gave a rating of in about 50% of trials, participants were able to detect the
how confident they felt about the correctness of their solu- method after one repetition of the trick.
tion on a scale from 0% to 100%. Fig. 2 illustrates the
procedure. 2.2. Results of Experiment 1
Please note that participants never received any feed-
back about the accuracy of their solutions. To control for 2.2.1. Response coding
familiarity of tricks, at the end of the entire experiment Participants solved magic tricks and categorized their
participants received a questionnaire with screenshots solutions into insight (with Aha!) and noninsight solutions
from all 34 tricks and were asked to indicate whether the (without Aha!). We use this categorization as our indepen-
solution to a trick was previously known to them. These dent variable. The dependent variables are: Solution Rate
tricks were excluded on an individual level and handled (number of solved tricks), Solution Accuracy (true or false)
as missing data. The entire experiment lasted about 2 h. and Number of Presentation (number of times a trick was
Note that there was a second testing session 14 days la- presented until participants solved the trick or until they
ter, in which participants had to perform an unexpected failed after the third presentation). We applied repeated
recall of solutions. These results are reported elsewhere measures analyses of variance (ANOVA) of the mean

Fig. 2. Procedure of one trial. Different phases and timing are displayed. Note that individual tricks vary in length.
 A.H. Danek et al. / Cognition 130 (2014) 174–185 179

number of solved tricks for statistical analyses. All p-values

are Greenhouse-Geisser corrected.
Participants’ solutions were coded off-line as true or
false by two independent raters, Cronbach’s alpha as a
measure of inter-rater reliability was 0.99. True solutions
were identical with the procedure that the magician had
actually used. False solutions consisted of methods that
were impossible with respect to the conditions seen in
the video clip. If no solution at all had been suggested,
the tricks were coded as unsolved.
In some trials (5.4% of all solutions), participants sug-
gested an alternative, but potentially conceivable method
(see introduction). We added those to the true solutions
category. 5.2% of the data points had to be discarded be-
cause the tricks were already familiar to participants.

Fig. 4. Mean number of solved tricks (out of 34) as a function of Solution

2.2.2. Solution rate and accuracy Type and Accuracy. Error bars denote standard errors of the mean. Grey
For this analysis, results were collapsed across repeti- bars indicate true solutions, black bars false solutions. Significant
differences are marked with an asterix.
tions. 45.8% of all trials (34 tricks x 48 participants yielded
a total of 1632 trials) were not solved, i.e. participants
watched the trick three times without suggesting a
(insight vs. noninsight) and Solution Accuracy (true vs.
solution. Those trials were excluded from further analyses
false) was conducted, with the number of solved tricks as
because no insight could occur. In 49%, participants
dependent variable. It revealed a significant main effect
suggested a solution (coded as either true or false). For
of Solution Type (F(1, 47) = 7.18, p < .05, g2partial ¼ :13) with
41.1% of the solved magic tricks, participants had reported
more noninsight than insight solutions and a significant
insight. The remaining 58.9% were classified as noninsight
main effect of Solution Accuracy (F(1, 47) = 37.05, p < .01,
solutions. Fig. 3 shows the percentages of true and false
g2partial ¼ :44) with more true than false solutions. There
insight and noninsight solutions. The ratio of true/false
was also a significant interaction (F(1, 47) = 12.47, p < .01,
solutions clearly varies between the two solution catego-
g2partial ¼ :21).
ries (15.6%/4.6% vs. 16.1%/12.7%), with only few false
Follow-up t-tests showed that there were significantly
solutions for insightful problem solving.
(t(47) = 7.35, p < .01, Cohen’s dz = .98) more true insight
The difference between insight and noninsight problem
solutions (M = 5.29, SD = 3.91, grey bar) than false insight
solving is illustrated more pointedly in Fig. 4, depicting the
solutions (M = 1.56, SD = 1.83, black bar). This is in contrast
mean number of solved tricks for each solution category. A
to noninsight solutions with no significant difference be-
repeated measures ANOVA with the factors Solution Type
tween the number of true (M = 5.48, SD = 3.0) and false
(M = 4.33, SD = 2.73) solutions.
5.2%
Discarded 2.2.3. Number of presentations
15.6% Tricks were presented up to three times. Therefore, a
True insight
solution trick could be solved during the first, second or third pre-
sentation. To test hypothesis 3, we investigated the respec-
tive time course of insight vs. noninsight solutions, again
using the number of solved tricks as dependent variable.
45.8% 4.6%
Not solved False insight Fig. 5 shows differences between insight and noninsight
solution problem solving with regard to time course: Insight solu-
tions occurred most frequently during the second presen-
tation of the trick (bar shaded in light grey) and then
during the third presentation (bar shaded in dark grey).
16.1%
True noninsight For noninsight solutions, this pattern is reversed. Hardly
solution any problems were solved during the first presentation
(black bars).
12.7% We conducted an ANOVA for repeated measures with
False noninsight the factors Solution Type (insight vs. noninsight) and Num-
solution
ber of Presentation (P1, P2, and P3) that revealed signifi-
Fig. 3. Overview on the data obtained. Mean percentages of not solved cant main effects for both factors (Solution Type,
and solved tricks, and their proportion of true and false solutions in the F(1, 47) = 7.12, p < .05, g2partial ¼ :13, and Presentation,
insight and noninsight categories. True insight solution: true or plausible F(2, 94) = 82.42, p < .01, g2partial ¼ :64). For the main effect
solution + reported Aha! experience; False insight solution: impossible
solution + reported Aha! experience; True noninsight solution: true or
of the factor Presentation, follow-up paired t-tests showed
plausible solution, without Aha! experience; False noninsight solution: that significantly less tricks were solved in P1 (M = 1.1,
impossible solution, without Aha! experience. SD = 1.6) than in P2 (M = 7.8, SD = 3.5) with t(47) = 14.16,
180 A.H. Danek et al. / Cognition 130 (2014) 174–185

of using the new domain of magic tricks that allows a de-

gree of ambiguousness that cannot be found in most of the
hitherto used mathematical, verbal or visual-spatial prob-
lems that are characterized by clear and unambiguous
solutions.
As hypothesized, the data obtained provided a further
differentiation between the two solution types, namely
with regard to time course: Insight solutions were
reached earlier than noninsight solutions, occurring more
often during the 2nd presentation (P2) of the video clip
than during the 3rd presentation (P3). For noninsight
solutions, this pattern was reversed (more often in P3
than P2). This data shows that increasing the number of
repetitions reduced insight solutions, and increased ana-
lytical, step-wise problem solving strategies. Watching
Fig. 5. Mean number of solved tricks (out of 34) as a function of Solution
the trick a second time, some participants gained sudden
Type and Number of Presentation. Error bars denote standard errors of
the mean. Black bars depict the number of solutions during the first insight into the trick (insight solution). It seems conceiv-
presentation of the trick, light grey bars depict solutions during the able that if participants did not experience insight during
second presentation and dark grey bars depict solutions during the third P2, they switched into an analytic mode. Consequently,
presentation.
solutions found during P3 were more seldom experienced
as sudden, but based on previous solving attempts and the
p < .01, Cohen’s dz = 1.9 and also significantly less than in systematic exclusion of hypotheses. Therefore, if more
P3 (M = 7.8, SD = 3.3) with t(47) = 11.97, p < .01, Cohen’s repetitions than three were included in an experimental
dz = 1.8. There was no significant difference between P2 paradigm, we would predict even less insight events with
and P3. increasing repetitions. This points also to a methodologi-
The significant interaction, F(2, 94) = 32.31, p < .01, cal problem of our paradigm: The repetition of the video
g2partial ¼ :41, shows the differential time course for the stimuli might determine a search and exclusion strategy
two types of problem solving, with insight solutions that could be different from working on a static problem
reached earlier than noninsight solutions. outlet.
The present experiment reveals a third difference with
2.2.4. Confidence rating regard to participants’ subjective appraisal of their
To test hypothesis 4, we compared the mean confidence performance (confidence hypothesis): The individual
rating of insight solutions (84.62% on a scale from 0% to certainty of having produced a correct response was signif-
100%) to the mean rating of noninsight solutions icantly higher for insight solutions than for noninsight
(63.08%). A significant difference (t(45) = 11.22, p < .01) solutions.
was found, indicating that participants were more confi- Summing up, insight solutions occurred earlier, partici-
dent that their insight solutions were correct than that pants felt more confident about them and in fact, they
their noninsight solutions were correct. were more likely to be true.
The finding of false insights (4.6% of all trials) has inter-
2.3. Discussion of Experiment 1 esting theoretical implications. False insights are trials in
which participants solved the trick, indicated that they
Using magic tricks, we found evidence for two different had experienced insight, but suggested a methodically
types of solving processes, with 41.1% of solutions classi- impossible solution. The existence of false insights has
fied as insight solutions and the remaining 58.9% as nonin- been debated, for example Sandkühler (2008, p. 2) states
sight solutions. The two solution types differed in their that ‘‘a true insight must lead to a correct solution’’. The
accuracy: Insight solutions were significantly more often Gestalt psychologists assumed that insight ‘‘always moves
true than false, whereas noninsight solutions were equally towards a better structural balance’’, believing in a ‘‘certain
true or false. Consequently, having an insight results very infallibility in insight’’ (Ohlsson, 1984, p. 68). This is in
likely in a true solution. This finding strongly provides evi- accordance with the immediate feeling of certainty
dence for the claim: ‘‘To gain insight is to understand (Sternberg & Davidson, 1995) that is often reported after
something more fully’’ (Dominowski & Dallob, 1995, p. insightful solutions (as it was the case in the present study,
37). Solving a problem by insight results in a deeper under- too). In contrast, Sheth et al. report the occurrence of incor-
standing of the problem (Sandkühler & Bhattacharya, rect solutions that were rated as highly insightful by prob-
2008). Without insight, the chances for producing a true lem solvers (2009, p. 1273). Ohlsson (1984, quoted in
or a false solution were nearly even. The differentiation Ohlsson, 1992, p. 3) originally defined insight as ‘‘the sud-
and analysis of true and false solution is, at least to our den appearance in consciousness of the complete and cor-
knowledge, completely new in the realm of insight prob- rect solution’’. Acknowledging the existence of false
lem solving. It provides first evidence that there might be insights, he later stated in a revision of his theory that
a qualitative difference in the comprehension of a problem the criterion of correctness of a solution is not useful for
between step-wise, analytical problem solving, and a definition of insight (Ohlsson, 1992). His conclusion that
problem solving by insight. This differentiation is a result correctness of solution is a ‘‘contingent characteristic
 A.H. Danek et al. / Cognition 130 (2014) 174–185 181

which accompanies some insights but not all’’ (Ohlsson, 3.2. Method
1992, p. 3) is clearly supported by our data. We conclude
that the present findings prove the existence of false in- 3.2.1. Participants
sights, but show that they are far less likely than true 62 students (26.2 ± 6.3; 17 male) participated for 10 € in
insights. the experiment. None of them had any neurological dis-
Our first experiment demonstrated that solving magic eases and all had normal or corrected-to-normal acuity.
tricks via insight differs from noninsight problem solving Participants were randomly assigned to either the experi-
in three important aspects: Accuracy, time course and mental group (informative cues) or the control group (no
solution confidence. But one hypothesis remains to be informative cues), with 31 participants each.
tested: We claimed that magic tricks activate constraints
that determine the problem solving process (constraint
3.2.2. Testing material
hypothesis) and the rationale of Experiment 1 is based on
Magic tricks: A set of 12 tricks was selected from the ori-
the implicit assumption that the subjective Aha! experi-
ginal 34 tricks used in the previous experiment: Tricks # 2,
ence stands for the cognitive process of representational
3, 4, 5, 6, 11, 12, 21, 23, 26, 27, 29 (see Appendix A for a de-
change and that constraint relaxation is the reason why
tailed description). The tricks were filtered out using two
the found differences occurred. This assumption is plausi-
criteria: (a) Tricks were very unlikely to be solved after
ble, and it is supported by the existing literature (e.g. Met-
the first presentation (solving rate after first viewing
calfe & Wiebe, 1987), but asks also for an empirical test
<10%). (b) It was possible and plausible to identify one con-
showing that (1) constraints play a role in magic tricks
straint of each trick that might be the main source of prob-
and (2) those constraints can be changed. We will address
lem difficulty (Kershaw & Ohlsson, 2004; Kershaw et al.,
these questions in Experiment 2.
2013).
Cues: Three different approaches were used to identify
the crucial constraints of each trick: First, the magician
3. Experiment 2
who had performed all tricks (TF) identified the main con-
straint that had to be relaxed in order to ‘‘see through the
3.1. Introduction
trick’’, i.e. the main source of difficulty. Second, one of the
authors (AD) analyzed participants’ written solution at-
In a second experiment, we address the assumption
tempts from the first experiment, especially the false ones,
that magic tricks impose constraints that restrict observ-
to identify the ill-defined assumptions participants used.
ers’ solution search space (constraint hypothesis). Because
Third, a student who was not familiar with the research
the constraints typically encountered by problem solvers
question watched all tricks and indicated which kind of
are known and exploited by the magician, magic tricks
cue would have helped her to find out the solution. Finally,
represent an ideal domain to systematically manipulate
based on this information, we created verbal cues (short
constraints. For example, if the main obstacle in a trick
sentences) that relaxed the putative main constraint, but
consists of the fact that a ball is usually perceived as a
did not give away the solution. For example, in trick #12,
whole (but is in fact a half-ball), we assume that this
a fake throw from one hand to the other is performed,
constraint can be relaxed by cueing the concept of a half-
guiding observers’ attention to the wrong hand and thus
sphere. Therefore, it seems plausible that the imposed con-
allowing a coin to vanish and re-appear under a napkin.
straints determine the problem difficulty of the magic
The main constraint is the fake throw that triggers the
tricks. We hypothesize that such conceptual constraints
wrong assumption that the coin has been transferred to
can be relaxed by verbal cues and, consequently, solution
the other hand. Thus, the cue is ‘‘transfer to other hand’’,
rates will increase in comparison to an uninformed control
inviting participants to think about the alleged transfer.
group.
Constraint relaxation occurs spontaneously, but can
also be triggered by cues, if the constraints are known (Öl- 3.2.3. Design and procedure
linger et al., 2013b). There is evidence that providing cues After providing informed consent, participants were or-
can facilitate solutions to insight problems (Grant & Spivey, ally instructed by the experimenter to watch the magic
2003; Thomas & Lleras, 2009), but there are also conflicting tricks and try to discover the secret method used by the
findings (Chronicle, Ormerod, & MacGregor, 2001; Orm- magician. The experimental group was told that there
erod, MacGregor, & Chronicle, 2002). Recently, we could would be a hint after the first presentation of the trick.
demonstrate that in classical insight problems like the After two practice trials, a randomized sequence of the
nine-dot problem and Katona’s five-square problem, per- 12 magic tricks was presented. The procedure was identi-
ceptual cues are only helpful if they restrict the initial cal to Experiment 1 (see Fig. 2), with the following
search space and the relaxed search space at the same changes: (1) Cues were presented immediately after the
time. That is, cues help to navigate the initial search space first presentation of the trick and before the fixation cross
and increase the likelihood of a representational change. reappeared and lasted for 3000 ms. The control group was
After constraints are relaxed through a representational presented with the word ‘‘Text’’ instead of a cue, also for
change, the search space increases even more and then 3000 ms. 2) Participants were asked only question #2
cues are helpful again in order to restrict the larger search (‘‘Please describe the solution!’’). Familiarity of tricks was
space, so that a solution can be found (Öllinger, Jones, & again checked with a questionnaire. Experiment 2 lasted
Knoblich, in press; Öllinger et al., 2013b). about 30 min.
182 A.H. Danek et al. / Cognition 130 (2014) 174–185

3.3. Results of Experiment 2 restructured, allowing the observer to solve the trick.
Importantly, we found a significant, but only moderate in-
3.3.1. Response coding and data analysis crease in the solution rate when informative cues were
Participants’ solution attempts were coded off-line as provided (33.2% vs. 20.9%). That is, not all participants
solved or unsolved by two independent raters (Cronbach’s could benefit from the provided cue. First, this demon-
alpha as a measure of inter-rater reliability was 0.98). We strates that we did not tell the entire solution, but only a
used the following two solution categories: Solved trials helpful cue. Second, this implicates that we did not resolve
comprised solutions that were identical with the proce- all sources of difficulty, a finding that is in line with the
dure that the magician had actually used or an alternative, multiple-sources of difficulty account of Kershaw and
potentially conceivable method. Unsolved trials consisted Ohlsson (2004; see also Kershaw et al., 2013; Öllinger
of methods that were impossible with respect to the condi- et al., 2013a,b), suggesting that having the right informa-
tions seen in the video clip. If no solution at all had been tion is not sufficient for the solution of a problem if the
suggested, the tricks were coded as unsolved, too. participant does not know how to integrate the informa-
The dependent variable was the participants’ mean per- tion in order to solve the problem (see Öllinger et al.,
centage of solved tricks. It was calculated as follows: For 2013b). It could also be an indicator that some of our cues
each participant individually, the total number of solved did not work in the intended way.
tricks (ranging from 0 to 12) was divided by 12, i.e. the to- In sum, Experiment 2 confirmed our hypothesis that
tal number of presented tricks. When participants already constraints play a role in magic tricks and that those con-
knew a trick, the trick was discarded and the total number straints can be relaxed via cues, facilitating a solution. This
was weighted by 12 minus the number of discarded tricks. finding is in accordance with the concept of constraint
This applied to 1.5% of trials. We analyzed the data with an relaxation postulated by the RCT (Knoblich et al., 1999;
univariate ANOVA, with the between-subjects factor Test- Ohlsson, 1992). Thus, magic tricks represent an ideal do-
group (experimental vs. control). main to systematically manipulate constraints.

3.3.2. Solution rates 4. General discussion

The overall solution rate was 27%. On average, the
experimental group (with informative cues) solved 33.2%, The present work provides evidence that the new task
whereas the control group (not informative cues) solved domain of magic tricks allows to differentiate between in-
20.9%, as depicted in Fig. 6. There was a significant main ef- sight and noninsight problem solving, with insight solu-
fect for the factor Testgroup, F(1, 60) = 11.84, p = .001, tions being more accurate, reached earlier and receiving
g2partial ¼ :17. higher ratings of confidence. Further, we could show that
magic tricks activate constraints and that proper cues can
3.4. Discussion of Experiment 2 help to overcome such constraints.

Experiment 2 investigated the importance of con- 4.1. Theoretical implications

straints for the solution of magic tricks. It was assumed
that if constraints play a role, cues might help to relax These results have three main theoretical implications:
these and consequently would increase the solution rate. First, with regard to the long-standing debate of the nature
It was found that cues increased solution rates in compar- of the insight process (e.g. Weisberg & Alba, 1981; and the
ison to a control group that received no informative cue. reply by Dominowski, 1981), our results support the con-
This might be evidence that constraints in magic tricks ceptualization of insight problem solving as a special pro-
can be relaxed by pointing out the wrong assumptions that cess (Davidson, 1995, 2003) that is qualitatively different
observers have made. It seems plausible to assume that from analytical or re-productive thinking (e.g. Weisberg
subsequently, the biased problem representation is & Alba, 1981). This in accordance with a number of other
recent studies (e.g. Jung-Beeman et al., 2004; Knoblich
et al., 2001; Kounios et al., 2008; Subramaniam, Kounios,
Parrish, & Jung-Beeman, 2009). Differences between in-
sight and noninsight problem solving are not confined to
solution accuracy, time course and solution confidence,
but also extend to differences in memory processes, with
insight solutions remembered better than noninsight solu-
tions, as we could recently demonstrate in a related study
(Danek et al., 2013).
Second, the present findings demonstrate the great po-
tential of using magic tricks as a new insight problem solv-
ing task. That magic tricks can be solved either with or
without insight is advantageous because it allows for a
comparison of different problem solving processes (insight
and noninsight problem solving), without changing the
Fig. 6. Solution rates for the experimental and the control group. type of task used. Our results support the idea that any
 A.H. Danek et al. / Cognition 130 (2014) 174–185 183

given problem may pose representational obstacles for unnoticed events. If participants had implemented such a
some solvers, but not for others (Ash, Cushen, & Wiley, strategy, this would speak more for an elaboration strategy
2009) and therefore may be solved through insightful pro- instead of constraint relaxation as the basis for a solution.
cesses or through more analytical processes or a combina- Elaboration means that the problem representation is
tion of both (Bowden et al., 2005). Another advantage of ‘‘changed by being extended or enriched’’ (Ohlsson, 1992,
magic tricks is that the solutions are largely unknown. p. 13) and has been proposed as another possibility to gain
Only 5.2% (Exp 1) and 1.5% (Exp 2) of all trials had to be insight. Although the effectiveness of the cueing manipula-
discarded in the present work. This is in clear contrast to tion in Experiment 2 strongly supports the constraint
most of the often used classical insight problems (e.g. the relaxation explanation of our results, we cannot fully rule
nine-dot problem) that can be found in psychology text- out the other possibility. Further studies that use e.g. only
books, in online resources or puzzle books, leading to a one presentation of the magic trick, and give participants
substantial number of participants that must be removed enough time to ‘‘mentally simulate’’ the trick could help
in each study due to familiarity. to clarify this question.
Third, the findings from Experiment 2 provide evidence A further limitation is that the magic tricks might not
that constraints play an important role in magic tricks, and have only one single constraint or source of problem diffi-
that they can be manipulated by appropriate cues, result- culty (see Kershaw & Ohlsson, 2004; Kershaw et al., 2013).
ing in higher solution rates. Therefore, constraint relaxa- The solution rates from the cueing condition indicate this,
tion can be regarded as a general mechanism that is since the rates are far from approaching 100%. Our inter-
sufficient to gain insight into the inner working of a magic pretation is that there exist additional constraints that
trick. This finding can be integrated within the representa- were not relaxed by the implemented cues. Successful
tional change theory of insight and expands her explana- solvers might use and integrate additional information. In
tory power. future experiments, the cues could be improved, perhaps
by using pictorial cues instead of verbal ones, and think-
aloud protocols could be used to identify possible addi-
4.2. Limitations tional constraints that prevent insight into the problem.
In sum, we offer a new, feasible approach for investigat-
A problem of our paradigm is the repetition of video ing the complex phenomenon of insight that impacts on
stimuli. This kind of presentation was, at least to our existing theories. In the long run, this work might help to
knowledge, never used before in insight problem solving, further elucidate the process of insight problem solving
and therefore it is difficult to directly compare our findings which is a vital part of human thinking and yet so difficult
with the existing literature that mostly used static problem to grasp.
displays. In general, a magic trick is a very complex stimu-
lus that is made of a stream of actions. We decided to facil-
itate the solution process by introducing up to three Acknowledgements
repetitions of the same video clip to obtain sufficiently
high solving rates. We cannot exclude the possibility that We thank Matus Simkovic for help with programming
participants accumulated additional information during the experiment. We also thank Eline Rimane, Svenja Brodt
these repetitions, perhaps by attending to different parts and Franca Utz for collecting part of the data and rating the
of the visual display and therefore ‘‘discovering’’ previously solutions.

Appendix A. List of magic stimuli

Trick name Magic effect Trick description Solveda Withb Surprisec

Aha!
1 Knives Transposition Two differently colored knives change places 14.58 28.57 2.55
2 Orange Transformation An orange is transformed into an apple 16.67 50.00 3.57
3 Monte Transposition A card swaps places with another one 22.92 27.27 2.82
4 Rope Restoration A rope is cut in two pieces and restored to one 25.00 50.00 2.50
5 Coin Trick 1 Vanish Out of three coins, one vanishes 27.08 23.08 2.91
6 Billiard Balls Appearance A little red ball multiplies 29.17 35.71 2.71
7 Coin Trick 2 Appearance and A coin is held up in the air, vanishes and 31.25 40.00 2.61
Vanish reappears
8 Card Trick 1 Telekinesis Cards turn over by themselves 31.25 33.33 2.54
9 Rubik’s Cube Transformation Rubik’s cube is solved by throwing it up in the air 33.33 50.00 3.17
10 Salt Vanish Salt is poured in the fist from where it disappears 33.33 43.75 3.04
11 KetchupBottle Vanish A ketchup bottle is put in a bag and disappears 35.42 29.41 3.27
12 Coin Trick 3 Transposition A coin wanders from the hand under a napkin 37.50 44.44 2.58

(continued on next page)

184 A.H. Danek et al. / Cognition 130 (2014) 174–185

Appendix A (continued)
Trick name Magic effect Trick description Solveda Withb Surprisec
Aha!
13 Bottled Scarf Vanish A red scarf disappears from a closed bottle 39.58 36.84 3.00
14 Pen Penetration Paper is pierced by a pen, but remains intact 43.75 42.86 2.75
15 Money Transformation Sheets of white paper turn into 50 Euro bills 43.75 28.57 3.00
16 Matchsticks Penetration One matchstick wanders through another one 45.83 50.00 2.59
without breaking it
17 Glass Vanish A champagne glass is covered by cloth and 47.92 39.13 2.26
disappears
18 Red Scarf Appearance A large red scarf appears from nowhere 47.92 60.87 2.73
19 Card Trick 2 Restoration A card is ripped in pieces and restored 50.00 33.33 3.22
20 Ice Cube Transformation Water is poured into a mug and transformed into 50.00 25.00 3.14
an ice cube
21 Coin Trick 4 Penetration A coin penetrates a sealed glass 52.08 20.00 3.00
22 Ball Transformation A ball gets transformed into a cube 52.08 36.00 2.50
23 Card Trick 3 Penetration Cards are chained to each other and unchained 54.17 30.77 3.25
without damage
24 Flying ball Telekinesis A ball is floating between the magician’s hands 54.17 42.31 3.00
(Levitation)
25 Card Trick 4 Transformation Cards in a glass change their colours 58.33 42.86 2.50
26 Coin Trick 5 Transposition 3 coins wander from one hand into the other 62.50 40.00 2.67
27 Salt ‘n Pepper Vanish Salt and pepper are poured into one hand and 64.58 32.26 3.13
the pepper disappears
28 Flying Bun Telekinesis A bun is covered by a napkin and starts to fly 66.67 37.50 2.50
(Levitation)
29 Bouncing Egg Physical A real egg is bounced repeatedly on the floor 72.92 25.71 3.05
impossibility without breaking
30 Scarf to Egg Transformation A scarf turns into an egg 77.08 70.27 2.67
31 Bowling Ball Topological A large bowling ball is carried in a thin suitcase 81.25 28.21 2.83
impossibility
(size)
32 Coat Hanger Topological A coat hanger is pulled from a small purse 83.33 50.00 2.88
impossibility
(size)
33 Cigarette Vanish Cigarette and lighter disappear while the 85.42 53.66 3.17
magician tries to light his cigarette
34 Spoon Transformation A spoon is put into the magician’s mouth and 95.83 65.22 2.88
when removed, it has changed into a fork
Tricks are sorted according to their difficulty (starting from the least solved ones).
a
Percentage of participants who solved the trick (after repeated viewing).
b
Percentage of participants who indicated an Aha! experience (of those participants who had solved it).
c
In a pilot study, 50 participants rated their level of surprise caused by the magic effect from 1 (not at all surprised) to 4 (very much surprised). The mean
rating for each trick is indicated.

References Chu, Y., & MacGregor, J. N. (2011). Human performance on insight pro-
blem solving: A review. The Journal of Problem Solving, 3(2), 119–150.
Danek, A. H., Fraps, T., von Müller, A., Grothe, B., & Öllinger, M. (submitted
Ash, I. K., Cushen, P. J., & Wiley, J. (2009). Obstacles in investigating the
for publication). It’s a kind of magic – new insights into the nature of
role of restructuring in insightful problem solving. The Journal of
Aha!.
Problem Solving, 2(2), 6–41.
Danek, A. H., Fraps, T., von Müller, A., Grothe, B., & Öllinger, M. (2013).
Bowden, E. M., & Jung-Beeman, M. (2003). Normative data for 144
Aha! experiences leave a mark: Facilitated recall of insight solutions.
compound remote associate problems. Behavior Research Methods,
Psychological Research, 77(5), 659–669.
Instruments, & Computers, 35(4), 634–639.
Davidson, J. E. (2003). Insights about insightful problem solving. In J. E.
Bowden, E. M., & Jung-Beeman, M. (2007). Methods for investigating the
Davidson & R. J. Sternberg (Eds.), The psychology of problem solving
neural components of insight. Methods, 42(1), 87–99.
(pp. 149–175). Cambridge: Cambridge University Press.
Bowden, E. M., Jung-Beeman, M., Fleck, J. I., & Kounios, J. (2005). New
Davidson, J. E. (1995). The suddenness of insight. In R. J. Sternberg & J. E.
approaches to demystifying insight. Trends in Cognitive Sciences, 9(7),
Davidson (Eds.), The nature of insight (pp. 125–155). Cambridge, MA:
322–328.
MIT Press.
Chronicle, E. P., Ormerod, T. C., & MacGregor, J. N. (2001). When insight
Dominowski, R. L. (1981). Comment on ‘‘An examination of the alleged
just won’t come: The failure of visual cues in the nine-dot problem.
role of ‘fixation’ in the solution of several ‘insight’ problems’’ by
The Quarterly Journal of Experimental Psychology: Section A, 54(3),
Weisberg and Alba. Journal of Experimental Psychology: General,
903–919.
110(2), 193–198.
 A.H. Danek et al. / Cognition 130 (2014) 174–185 185

Dominowski, R. L., & Dallob, P. (1995). Insight and problem solving. In R. J. Mednick, S. A. (1962). The associative basis of the creative process.
Sternberg & J. E. Davidson (Eds.), The nature of insight (pp. 33–62). Psychological Review, 69, 220–232.
Cambridge, MA: MIT Press. Metcalfe, J. (1986). Feeling of knowing in memory and problem solving.
Duncker, K. (1945). On problem-solving. Psychological Monographs, 58. Journal of Experimental Psychology: Learning, Memory, and Cognition,
Whole No. 270. 12(2), 288–294.
Evans, J. S. B. T. (2008). Dual-processing accounts of reasoning, Metcalfe, J., & Wiebe, D. (1987). Intuition in insight and noninsight
judgment, and social cognition. Annual Review of Psychology, 59, problem solving. Memory and Cognition, 15, 238–246.
255–278. Newell, A., & Simon, H. A. (1972). Human problem solving. Englewood
Fleck, J. I., & Weisberg, R. W. (2013). Insight versus analysis: Evidence for Cliffs, NJ: Prentice Hall.
diverse methods in problem solving. Journal of Cognitive Psychology, Ohlsson, S. (1984). Restructuring revisited: I. Summary and critique of the
25(4), 436–463. Gestalt theory of problem solving. Scandinavian Journal of Psychology,
Grant, E. R., & Spivey, M. J. (2003). Eye movements and problem solving. 25, 65–78.
Psychological Science, 14(5), 462–466. Ohlsson, S. (1992). Information-processing explanations of insight and
Jarosz, A. F., Colflesh, G. J. H., & Wiley, J. (2012). Uncorking the muse: related phenomena. In M. Keane & K. J. Gilhooly (Eds.), Advances in the
Alcohol intoxication facilitates creative problem solving. Conscious- psychology of thinking (pp. 1–44). London: Harvester-Wheatsheaf.
ness and Cognition, 21(1), 487–493. Öllinger, M., Jones, G., & Knoblich, G. (in press). Insight and search in
Jastrow, J. (1888). The psychology of deception. Popular Science Monthly, Katona’s Five-Square problem, Experimental Psychology.
34(10), 145–157. Öllinger, M., Jones, G., Faber, A. H., & Knoblich, G. (2013a). Cognitive
Jones, G. (2003). Testing two cognitive theories of insight. Journal of mechanisms of insight: The role of heuristics and representational
Experimental Psychology: Learning, Memory, and Cognition, 29, change in solving the eight-coin problem. Journal of Experimental
1017–1027. Psychology: Learning, Memory, and Cognition, 39(3), 931–939.
Jung-Beeman, M., Bowden, E. M., Haberman, J., Frymiare, J. L., Arambel- Öllinger, M., Jones, G., & Knoblich, G. (2006). Heuristics and
Liu, S., Greenblatt, R., et al. (2004). Neural activity when people solve representational change in two-move matchstick arithmetic tasks.
verbal problems with insight. PLoS Biology, 2(4), 500–510. Advances in Cognitive Psychology, 2(4), 239–253.
Kaplan, C. A., & Simon, H. A. (1990). In search of insight. Cognitive Öllinger, M., Jones, G., & Knoblich, G. (2008). Investigating the effect of
Psychology, 22(3), 374–419. mental set on insight problem solving. Experimental Psychology, 55(4),
Katona, G. (1940). Organizing and memorizing: Studies in the psychology of 270–282.
learning and teaching. New York: Columbia University Press. Öllinger, M., Jones, G., & Knoblich, G. (2013b). The dynamics of search,
Kershaw, T. C., Flynn, C. K., & Gordon, L. T. (2013). Multiple paths to impasse, and representational change provide a coherent explanation
transfer and constraint relaxation in insight problem solving. Thinking of difficulty in the nine-dot problem. Psychological Research, 1–10.
and Reasoning, 19(1), 96–136. http://dx.doi.org/10.1007/s00426-013-0494-8.
Kershaw, T. C., & Ohlsson, S. (2004). Multiple causes of difficulty in Öllinger, M., & Knoblich, G. (2009). Psychological research on insight
insight: The case of the nine-dot problem. Journal of Experimental problem solving. In H. Atmanspacher & H. Primas (Eds.), Recasting
Psychology: Learning, Memory, and Cognition, 30(1), 3–13. reality: Wolfgang Pauli’s philosophical ideas and contemporary science.
Knoblich, G., Ohlsson, S., Haider, H., & Rhenius, D. (1999). Constraint Berlin: Springer.
relaxation and chunk decomposition in insight problem solving. Ormerod, T. C., MacGregor, J. N., & Chronicle, E. P. (2002). Dynamics and
Journal of Experimental Psychology: Learning, Memory, and Cognition, constraints in insight problem solving. Journal of Experimental
25(6), 1534–1555. Psychology: Learning, Memory, and Cognition, 28(4), 791–799.
Knoblich, G., Ohlsson, S., & Raney, G. E. (2001). An eye movement study of Ortiz, D. (2006). Designing miracles. El Dorado Hills: A-1 MagicalMedia.
insight problem solving. Memory and Cognition, 29(7), 1000–1009. Parris, B. A., Kuhn, G., Mizon, G. A., Benattayallah, A., & Hodgson, T. L.
Kounios, J., Fleck, J. I., Green, D. L., Payne, L., Stevenson, J. L., Bowden, E. M., (2009). Imaging the impossible: An fMRI study of impossible causal
et al. (2008). The origins of insight in resting-state brain activity. relationships in magic tricks. NeuroImage, 45(3), 1033–1039.
Neuropsychologia, 46(1), 281–291. Sandkühler, S., & Bhattacharya, J. (2008). Deconstructing insight: EEG
Kounios, J., Frymiare, J. L., Bowden, E. M., Fleck, J. I., Subramaniam, K., correlates of insightful problem solving. PLoS One, 3(1), 1–12.
Parrish, T. B., et al. (2006). The prepared mind. Psychological Science, Sheth, B. R., Sandkühler, S., & Bhattacharya, J. (2009). Posterior beta and
17(10), 882–890. anterior gamma oscillations predict cognitive insight. Journal of
Kuhn, G., Amlani, A. A., & Rensink, R. A. (2008). Towards a science of Cognitive Neuroscience, 21(7), 1269–1279.
magic. Trends in Cognitive Sciences, 12(9), 349–354. Sternberg, R. J., & Davidson, J. E. (1995). The nature of insight. Cambridge,
Kuhn, G., Kourkoulou, A., & Leekam, S. R. (2010). How magic changes MA: MIT Press.
our expectations about autism. Psychological Science, 21(10), Subramaniam, K., Kounios, J., Parrish, T. B., & Jung-Beeman, M. (2009). A
1487–1493. brain mechanism for facilitation of insight by positive affect. Journal
Kuhn, G., & Land, M. F. (2006). There’s more to magic than meets the eye. of Cognitive Neuroscience, 21(3), 415–432.
Current Biology, 16(22), 950–951. Tamariz, J. (1988). The magic way. Madrid: Frakson Books.
Kuhn, G., & Tatler, B. W. (2005). Magic and fixation: Now you don’t see it, Thevenot, C., & Oakhill, J. (2008). A generalization of the representational
now you do. Perception, 34(9), 1155–1161. change theory from insight to non-insight problems: The case of
Kuhn, G., Tatler, B., Findlay, J. M., & Cole, G. G. (2008). Misdirection in arithmetic word problems. Acta Psychologica, 129(3), 315–324.
magic: Implications for the relationship between eye gaze and Thomas, L. E., & Lleras, A. (2009). Swinging into thought: Directed
attention. Visual Cognition, 16(2), 391–405. movement guides insight in problem solving. Psychonomic Bulletin
Lamont, P., Henderson, J. M., & Smith, T. J. (2010). Where science and and Review, 16(4), 719–723.
magic meet: The illusion of a ‘‘science of magic’’. Review of General Weisberg, R. W., & Alba, J. W. (1981). An examination of the alleged role of
Psychology, 14(1), 16–21. ‘‘fixation’’ in the solution of several ‘‘insight’’ problems. Journal of
Lamont, P., & Wiseman, R. (1999). Magic in theory. Hertfordshire: Experimental Psychology: General, 110(2), 169–192.
University of Hertfordshire Press. Weisberg, R. W. (1995). Prolegomena to theories of insight in problem
MacGregor, J. N., & Cunningham, J. B. (2008). Rebus puzzles as insight solving: A taxonomy of problems. In R. J. Sternberg & J. E. Davidson
problems. Behavior Research Methods, 40(1), 263–268. (Eds.), The nature of insight (pp. 157–196). Cambridge, MA: MIT Press.