Solving problems by insight has been proposed to be beneficial to long-term memory (LTM) encoding (Auble, Franks, & Soraci, 1979; Danek, Fraps, von Müller, Grothe, & Ollinger, 2013; Ludmer, Dudai, & Rubin, 2011), a proposal that could have considerable educational as well as neuroscientific relevance. It would suggest that educators can optimize learning by actively creating situations during which their students experience insight. Understanding the relevant cognitive processes could help devising such educational strategies, in addition to addressing the fundamental brain basis of learning and memory. However, insight has been operationalized very differently across studies (e.g., Auble et al., 1979; Bowden, Jung-Beeman, Fleck, & Kounios, 2005; Kounios & Beeman, 2009; Ludmer et al., 2011). For example, Mednick operationalized insight as solving a problem that requires creativity (Mednick, 1962), whereby creativity was understood as the process of searching successfully for a unique common associate between a provided set of three words that are only remotely associated by that word (e.g., cottage, blue, goat–cheese). If the words were already closely associated, it would not require creativity to find the missing link. Others have asserted that the presence of a subjective “aha!” experience is necessary to interpret a solution to a problem as an insight (Bowden & Jung-Beeman, 2003a; Bowden et al., 2005; Danek et al., 2013). In contrast to this definition, Mednick’s operationalization is relatively more objective as the correctness of a solution to a problem that has only one possible solution can be judged by a third person. Again others required only a state of previous incomprehension of the solution to a problem, followed by a sudden change to a state of understanding the solution, which could even be induced by presenting the solution (Auble et al., 1979; Ludmer et al., 2011). Finally, some researchers have proposed that insight does not necessarily include a state of incomprehension, but needs mental restructuring (Wills, Estow, Soraci, & Garcia, 2006), for which the common term in the insight literature is a representational change of the problem/solution space (Ohlsson, 1984; Öllinger, Jones, & Knoblich, 2013). Therefore, it is difficult to say by what means insight problem-solving may facilitate encoding.
In the current study, we are focusing on two aspects of the different operationalizations of insight that may be especially relevant to learning, but which are not used consistently: generation, that is, whether the solution to an insight problem is found by the participant or presented by the experimenter, and the subjective “aha!” experience,Footnote 1 which is a subjective feeling of comprehending the solution suddenly while being convinced of its truth. We now address these components in turn and explain their relevance to learning.
The generation effect
The generation effect is the superiority of items that are generated by participants over items that are presented to them in regard to later LTM performance (Burns, 1992; Slamecka & Fevreiski, 1983; Slamecka & Graf, 1978). Generation may involve, for example, coming up with a word associated to a cue word, completing a word stem, solving an anagram, finding an antonym, or—in terms of the question at hand—finding a solution to a problem in contrast to being presented with the solution. It has been consistently found that requiring generation, in comparison to presenting the solution, leads to enhanced later long-term memory as tested by recognition memory, cued recall, and free recall (e.g., Burns, 1992; McNamara & Healy, 2000; Slamecka & Graf, 1978). In addition, this effect is larger the longer the retention interval (for a summary and overview, see the meta-analysis by Bertsch, Pesta, Wiscott, & McDaniel, 2007). There have been several attempts to explain the effect, the main hypotheses being (1) cognitive effort, (2) transfer-appropriate processing, and (3) level of processing (LOP).
The first explanation suggests higher cognitive effort used to engage with the items in contrast to mere reading/perceptual processing. However, in an extensive meta-analysis, Bertsch et al. found no dependency on cognitive effort, that is, task difficulty (Bertsch et al., 2007). The second explanation, transfer-appropriate processing, refers to the finding that memory performance is better, the closer the processing required by the memory test corresponds to the processing that occurred in the encoding phase. However, the evidence shows only partial support for this idea (Bertsch et al., 2007). The third explanation, level of processing, is probably the strongest so far, that is, that generating an item leads to a more elaborate semantic processing of the item than reading the solution, thereby enhancing memory (McElroy & Slamecka, 1982; Slamecka & Graf, 1978). There is strong evidence that better semantic integration for the generate condition is at the core of the generation effect. When comparing words with non-words in a generate and read condition, there is no generation effect for non-words—not even when the generation rule is rhyming, that is, lexical-phonologic, or when the target word to a non-word can be a word (McElroy & Slamecka, 1982). This result suggests that the semantic spreading activation of related concepts during generation and a better integration of the generated items into existing semantic networks is mostly responsible for the later memory advantage (Anderson, 1983; Collins & Loftus, 1975).
The subjective “aha!” experience
The “aha!” experience is defined as the sudden, unexpected comprehension of a solution to a problem that comes with an ease of understanding the solution. It is usually accompanied by a positive feeling that is not the feeling of pride to have solved the item, because it comes before the assessment of the solution. Moreover, there is complete confidence in the truth of the solution (Topolinski & Reber, 2010). The generation of a solution per se does not necessarily provide any information about how participants arrived at the solution, e.g., whether they arrived at it via sudden insight, by means of a analytical problem-solving approach, or otherwise—only if one assumes that a problem is not solvable without insight (such as Mednick, 1962). We therefore investigated not only the relationship between generating solutions to insight problems, but also how generation relates to the subjective “aha!” experience, and their possible interaction in regard to learning. Surprisingly, none of the authors who have previously used the subjectively reported “aha!” as a criterion for insight has reported whether such “aha!” experiences may also be reported when the solution is presented after a failed attempt at problem-solving.
The “aha!” experience was coined by Bühler (1907) who referred to it as a sudden verbal-cognitive understanding of a foreign idea/sentence. Something novel is learned by means of a complex interaction between one’s declarative memory content and problem-solving processes tackling the sentence. A more recent operationalization of the subjective “aha!” experience was introduced by Bowden and Jung-Beeman (2003b). Before the experiment, participants were provided with the following definition of an “aha!” experience: “A feeling of insight is a kind of ‘Aha!’ characterized by suddenness and obviousness. You may not be sure how you came up with the answer, but are relatively confident that it is correct without having to mentally check it. It is as though the answer came into mind all at once—when you first thought of the word, you simply knew it was the answer. This feeling does not have to be overwhelming, but should resemble what was just described”. (p. 507, Jung-Beeman et al., 2004). During the experiment, whenever the participants solved a problem, they were asked whether they solved it with an “aha!” experience or not. The authors interpreted items being solved without an “aha!” experience as being solved analytically, that is, by means of applying known problem-solving strategies and continuously getting closer to the solution until it has been found.
However, whether such a clear-cut binary categorization can be made is questionable. For example, Klein and Jarosz (2011) investigated naturally occurring insights, by collecting reports of such incidents in everyday life. They observed that reaching an impasse occurred only in some cases, although a change in understanding happened most of the time and helped coming up with new ways to approach solving the problem (representational change of the problem/solution space). Moreover, insights were often reached gradually: an intuitive feeling of coming closer to the solution often preceded the insight for a long time. During this time, inconsistencies with one’s own assumptions and with information that at first seemed coincidental was gradually perceived to form a pattern that finally led to the solution (Klein & Jarosz, 2011). They therefore proposed that the subjective “aha!” experience might be an epiphenomenon that accompanies some of the insight solutions, probably those that are found after an impasse has been suddenly overcome, but not all. On the other hand, Bowers suggests that the intuitive feeling of getting closer to the solution is due to automatic spreading activation from all known pieces of information to associated semantic information that gradually form coherence which at some point emerges into consciousness (Bowers, Regehr, Balthazard, & Parker, 1990). This theory of insight problem-solving does not necessitate impasse for the “aha!” experience which merely indicates the sudden emergence into consciousness of a hypothesis about the solution.
In the current study, we used a definition of the subjective feeling of “aha!” that was based on the Bowden and Jung-Beeman (2003a, b) definition, but changed it so that the generation of a solution was not declared a necessary precondition for “aha!”. This way, subjects were unbiased, therefore enabling us to test whether “aha!” experiences do indeed depend on generation. It should be noted that we did not specify the occurrence of an impasse or representational change as a defining criterion for the subjective “aha!” experience, because it is not clear whether it is necessary precondition as discussed above. Moreover, we do not see no “aha!” as analytical problem-solving but leave it undefined, as it is only clear that subjects did not comprehend the solution in the specified way for “aha!”, but not by what means they reached an understanding.
Insight and long-term memory encoding
We assume that insight can enhance LTM encoding (Auble et al., 1979; Danek et al., 2013). There are several factors that may lead to such a positive influence on memory formation. First, generation can enhance LTM encoding for an item although, as previously discussed, the explanation for this effect is not completely understood. Thus, learning from insight may occur mainly due to the generation effect. Second, Auble et al., found that learning only occurs when an initial state of incomprehension is followed by comprehension (Auble et al., 1979). In that sense, we hypothesize that the subjective feeling of “aha!”, indicating the sudden change from incomprehension to comprehension, may also occur when the solution is presented after an unsuccessful attempt at problem-solving. However, an “aha!” experience during generation has been found to enhance encoding even more than generation without “aha!” (Danek et al., 2013). Third, the novelty of the sudden coherence of the solution may be another factor facilitating learning. Novelty responses in the brain are similar to reward responses; novelty itself can be seen as rewarding as summarized in the novelty exploration bonus hypothesis (Kakade & Dayan, 2002). Associative novelty and reward are both thought to facilitate memory encoding by means of increased dopaminergic input from the midbrain to the hippocampus (amongst other regions), which is a key region of memory consolidation (e.g., Bunzeck, Doeller, Dolan, & Düzel, 2012; Kumaran & Maguire, 2009; Lisman & Grace, 2005; Otmakhova, Duzel, Deutch, & Lisman, 2013). Thus, learning from insight may further be based on the rewarding feeling that is based on the suddenness of understanding a novel relationship that previously seemed incomprehensible and is accompanied by a positive emotional response. Lastly, evidence suggests that detecting novel information that fits into an existing schema can accelerate learning (Tse et al., 2007; Van Kesteren, Ruiter, Fernández, & Henson, 2012). This was in fact one of the original Gestalt theories on learning from insight, i.e., that learning from insight occurs as novel information builds a novel schema with information already present in semantic memory. The previously perceived ambiguity is suddenly resolved as the association becomes clear (Bühler, 1907; Vollmers, 2014). To use the example of Archimedes: he suddenly realized the connection between the water level in his bath tub rising and the mass submerged into the water, although all necessary pieces of information were already known to him.
The subjective “aha!” experience may reflect a specific quality of comprehension. It is possible that the subjective “aha!” experience is evoked by a particularly strong novelty response due to the suddenness of comprehension. Moreover, while successfully generating a solution may already evoke a positive emotional response, the “aha!” experience is additionally associated with positive affect (Danek et al., 2013; Topolinski & Reber, 2010).Footnote 2 This positive emotional response may reflect the joy of comprehension or relief from the state of tension due incomprehension (Danek, Fraps, von Müller, Grothe, & Öllinger, 2014). We would therefore expect that the “aha!” experience per se can enhance memory formation, even when the solution is presented after failing to generate it, and when the additional positive emotional response related to solving the problem is therefore absent.
Aims of the current study
In the current study, we investigated the extent to which the generation of a solution and the subjective “aha!” experience are associated with LTM encoding in regard to two different measures of memory. We further wanted to evaluate the qualitative difference of the sudden comprehension for generated and non-generated solutions, as well as solutions understood with an “aha!” experience and without. The latter was especially interesting, because most previous studies reflect the indirect assumption that the subjective “aha!” experience can only occur when the solution is generated instead of being presented, without looking at generation and “aha!” in a fully crossed design.
For this purpose, we chose a pictorial problem-solving task in which we presented so-called Mooney stimuli, that is, black and white photos without any shades of gray (Imamoglu, Kahnt, Koch, & Haynes, 2012; Mooney, 1957). The participants’ task during the encoding phase was to try to figure out what was presented in the Mooney image (see Fig. 1a). The encoding phase was incidental, that is, participants were not told that their memory of the Mooney images and their motifs would be tested later on. After 1 week, memory was tested with an indirect (incidental) and a direct (intentional) memory test via solving old and new items and an old/new recognitionFootnote 3 memory question, respectively (Richardson-Klavehn & Bjork, 1988). Direct memory tests are more likely to tap into voluntary, controlled retrieval, whereas indirect memory tests are more likely to tap into more automatic, involuntary retrieval (Richardson-Klavehn, 2010). This paradigm enabled us to investigate the insight components generation and “aha!” and their relationships to voluntary (direct test: old/new decision) versus more automatic retrieval processes (indirect test: generating solutions to old items) (see also Schott et al., 2005). We would like to emphasize that we are investigating the relationships of generating a solution and the feeling of “aha!” to long-term memory independently. Thus, we are also testing whether a subjective feeling of “aha!” may also occur when the solution is presented after a failed attempt at generating a solution, and whether this may still be related to learning.
We further incorporated an emotional valence rating of how participants felt when comprehending the solution, in order to compare the quality of the sudden comprehension for generated and non-generated solutions, as well as solutions understood with an “aha!” experience and without.
Hypotheses
First of all, we hypothesized that both generation and “aha!” would be related to a relatively more positive emotional valence rating than no generation and no “aha!”. In the case of generation accompanied by a feeling of “aha!”, Danek et al. report feelings of joy, tension release, as well as performance-related emotions, such as pride, satisfaction, competitiveness (Danek et al., 2014). While most of these aspects of a positive emotional response can be expected for both generating a solution (with or without “aha!”) or for the “aha!” experience (with or without generation), the performance-related emotions depend on generation and should therefore not be related to the “aha!” experience. Thus, the emotional response is probably slightly more positive for generation than for aha.
In regard to the above discussed characteristics of insight that may be related to learning (Sect. “Insight and long-term memory formation”), we hypothesized that later memory performance should be better for generated solutions during encoding, both for identification and recognition memory performance at test, due to the generation effect. Second, the subjective feeling of “aha!” should lead to enhanced encoding, probably due to the positive emotional response (Danek et al., 2013; Ludmer et al., 2011). Third, as for the relationship between generation and “aha!”, it is unclear what to expect of non-generated solutions, because the relationship between the subjective “aha!” experience and later memory performance has thus far only been tested for generated solutions (Bowden & Jung-Beeman, 2003a; Danek et al., 2013). The implicit assumption of previous studies behind the analysis of aha/no aha responses only for generated solutions was probably that an “aha!” experience can only occur for self-solved items. If so, one might expect an interaction between the variables generation and aha in relation to later memory performance, such that an “aha!” experience is particularly important for later memory of items successfully generated at encoding. However, it could also be that the “aha!” experience can also be invoked by a presented solution after fruitless efforts of trying to solve the problem (Bowden & Jung-Beeman, 2003b), possibly indicating an induced representational change (Ohlsson, 1984, 1992). In this case, the relationships between generation and “aha!” with later memory might be independent.
As for the different measures of memory, we asked (1) whether generation during encoding facilitates identification during testing more than it facilitates recognition memory, which would represent transfer-appropriate processing, and (2) whether the “aha!” experience is more strongly correlated with an increased probability of recognizing the item as old during testing compared to later identification during testing, which might be expected if “aha!” experiences leave behind a distinctive episodic memory trace. The results thus gave an additional opportunity to examine whether successful generation and the “aha!” experience during encoding involve similar or partly distinct processes.
Our data revealed that both the generation of a solution and the “aha!” experience are independently accompanied by a positive emotional response and are also independently related to later memory. As for the different measures of memory, we found that generation at encoding had a stronger relationship to later identification than to later recognition memory at test, whereas the “aha!” experience at encoding had a slightly larger relationship to later recognition memory than to later identification at test.