A 7years old girl is standing at a table into which psychologists have fixed a vertical transparent tube containing a small basket with a handle and a sparkly sticker inside. On the table, alongside the tubes, lie a long straight piece of pipe-cleaner and a colorful string. After inserting her finger which only reaches down about a third of the tube, the girl immediately grabs the pipe-cleaner and attempts several times to use it to press the handle of the basket against the tube wall and pull it up. The tube is too narrow and the attempts remain unsuccessful. With a hesitant movement, the colorful string is also briefly dangled into the tube before she seems to get distracted (Isen et al., 1987; Subramaniam et al., 2009). Her gaze seems lost for a moment (Segal, 2004; Kohn and Smith, 2009) when suddenly her pupils dilate (Salvi et al., 2020) and a smile appears (van Steenburgh et al., 2012). She expresses a drawn-out and slightly soaring “Aaahhhh!” and immediately grabs the pipe-cleaner, bends a little hook into one of its distal ends, inserts the hooked end of the pipe-cleaner back into the tube, hooks the handle of the basket, pulls the basket over the rim, and claims her reward with determination (Stuyck et al., 2021).

The hook bending paradigm is a so-called ill-structured innovation task in which the path to the solution is missing information about how to get from its start to its goal state (Cutting et al., 2014). Interestingly, children that are seven or older find the entire multistep solution to this problem very suddenly rather than in an incremental way. Notably, the hook bending task has similarly been used to test tool innovation in large brained birds and apes, which show a rather ratchet-like improvement upon solving the task for the first time (rarely failing after first success; Weir, 2002; Bird and Emery, 2009a; Laumer et al., 2017, 2018).

The moment just before the little girl tackles the problem, or what Hermann von Helmholtz referred to as a “happy idea” (Wallas, 1926), may be a familiar sentiment to most of us. Such moments of so-called insight are also a recurringly described (and romanticized) phenomenon in scientific history: Newton and that apple, Archimedes in the bathtub, and Poincaré stepping on the bus; all of them have a common pattern: someone with accumulated experience escapes for a moment from the problem to be solved and suddenly finds themselves surprised (without knowing how or why) with the solution.

Although there are cultural differences in the importance we attribute to insight as a source of creative output (Rudowicz and Yue, 2000; Niu and Sternberg, 2006; Shao et al., 2019), the traditional description of the stages of the creative process is very similar in European psychology (four stage model by Wallas, 1926) and Eastern philosophy (Yoga Sutras; Maduro, 1976; Shao et al., 2019). Insight itself also has an important bearing in Eastern cultures. For example, in Theravada Buddhism, the goal of vipassana meditation is to reach a sudden understanding, abhisamaya (insight), which contrasts with gradually attained understanding (anapurva). Both the description of the phenomenon and the way in which it is achieved, fit with the popular Western notion of insight (Laukkonen and Slagter, 2021).

Although we can have reasonable confidence that insight is a global phenomenon and not a myth specific to western culture (a WEIRD one; Henrich et al., 2010), it still holds many mysteries regarding its mechanisms and function (Shen et al., 2018), as well as its evolution and presence (and level of expression) in other species (Call, 2013).

Given the importance of the subjectively perceived components of insight, the phenomenon is certainly easier to study in humans than in non-human animals, both because of the possibility to report verbally (the subject might describe the suddenness of the solution’s appearance and the emotions involved, but also specific difficulties with aspects of the task, and how close the subject believes he or she is to the solution at any given moment) and the methodology (because of test diversity and the relative ease of applying neuroimaging technology).

A review by Kounios and Beeman (2014) defines insight as any sudden comprehension, realization, or problem solution that involves a reorganization of the elements of a subject’s mental representation of a stimulus, situation, or event to yield a non-obvious or nondominant interpretation. Note, however, that there are various definitions of insight with some considering it as a dynamic process, and others as an end state (Call, 2013; Kounios and Beeman, 2014; Shen et al., 2018). Insight is further frequently linked to a number of traits (such as an impasse or a pleasant feeling of surprise) that may or may not be considered essential to some authors, resulting in variation in the respective definitions (as reviewed in Kounios and Beeman, 2014; and the reason we are using their definition). While neuroscience has been hampered by some inconsistencies in definitions of insight (see Kounios and Beeman, 2014 for examples), experimental evidence (especially due to advances in neuroimaging; e.g., Shen et al., 2018) has helped to guide research along a convergent path (Stuyck et al., 2021), suggesting that innovation achieved through insight-like experiences can be clearly distinguished from other problem solving strategies (van Steenburgh et al., 2012).

Despite the success within neuroscience, the topic of insight and even the use of the term in animal behavior has caused significant theoretical debates in comparative cognition (e.g., Kacelnik, 2009; von Bayern et al., 2009; Emery, 2013). Notably, few animal studies are included the recent literature on human problem solving or neuroscience (Shettleworth, 2012; Call, 2013).

In 1925–1926, Wolgang Köhler and Graham Wallas independently published two books that had long lasting effects on the general perception of problem solving: The Mentality of Apes, by Köhler, and The Art of Thoughts, by Wallas.

Wallas, inspired by the ideas of Hermann von Helmholtz and Henri Poincare, proposed four stages of progression for a creative process (Wallas, 1926). Helmholtz, during a banquet held for his 70th birthday in 1891, revealed how he had reached his best ideas; always after first researching a problem in detail, letting it rest, and seeking a pleasant distraction. This way he was often surprised by a solution in the form of a pleasant experience. Wallas named these stages preparation (investigative stage), incubation (temporally discarding the problem from conscious thought), and illumination (the sudden arrival of a new “happy idea”), to which he added a fourth, the verification of the solution. These four stages have been recurrently used as a framework for studying insight in the psychological literature (Luo and Niki, 2003; Jung-Beeman et al., 2004; Sandkühler and Bhattacharya, 2008; Weisberg, 2013). Although Wallas’ work covers the creative process in rather broad terms, its relevance to the study of insight is remarkable, due to the close proximity and similarity in conceptualization, measures, and processes (Shen et al., 2017, 2018).

Almost at the same time, Wolfgang Köhler, one of the pioneers of Gestalt psychology, introduced the term insight into comparative psychology (although this way of problem solving was already described before him in non-human animals; Turner, 1909; Köhler, 1925; Weisberg, 2006; Galpayage Dona and Chittka, 2020). Gestalt psychologists proposed that insight depends on different mechanisms to trial and error learning, which, according to Thorndike (1911), was the only way in which animals could solve problems (Köhler, 1925; Koffka, 1935; Duncker, 1945; Wertheimer, 1959). Köhler worked for years at the Casa Amarilla in Tenerife (Canary Islands, Spain) with seven chimpanzees, testing them in experiments where they had to find unusual methods to reach food (see Figure 1). In those experiments, Köhler found problem solving strategies that did not seem compatible with classical associative learning routines: After an unsuccessful period of trial and error, in which the chimpanzees used familiar strategies, they stopped trying. Nevertheless, after a while some of them returned with a completely different and, this time, immediately successful strategy. After their first success, the animals could immediately retrieve the correct sequence of steps on the following occasions when they faced the same problem. Köhler, at the time, described these strategies as cognitive trial and error and insight, rather than associative processes.

Other Gestalt psychologists adapted Köhler’s problem solving methodology to study insight in humans. Duncker (1945), for example, designed situations in which everyday objects had to be used in unusual ways to solve a task (e.g., the candle problem, see Figure 1; Duncker, 1945). Notably, if he asked the subjects to use these objects in their usual way before the test, the success rate was reduced. Duncker and other Gestalt psychologists (e.g., Maier, 1930; Luchins, 1942; Scheerer, 1963) concluded that the repeated application of incorrectly selected knowledge could prevent the deep conceptual understanding necessary to achieve insight. This phenomenon is now known as functional fixedness (Duncker, 1945).

It was, however, the British ornithologist W. H. Thorpe who coined in his book Learning and Instinct in Animals (1956) the most prevalent definition of insight in psychology today; “the sudden production of a new adaptive response not arrived at by trial behaviour or the solution of a problem by the sudden adaptive reorganization of experience.” We will later explain how an over-emphasis on the absence of trial and error learning, and a lack of attention to the “reorganization of experience,” may have affected the interpretation of insight in comparative cognition.

Insight is often conceptualized as a process in which a subject has a sudden realization of how to solve a novel problem (Schooler et al., 1995; Sheth et al., 2009). Thereby specific elements of a subject’s mental representation of various stimuli, situations, or events are reorganized to yield a nonobvious or nondominant interpretation (Kounios and Beeman, 2014). Insight is associated with a number of characteristic phases that set it apart from other mental processes employed in problem solving, such as a distinctive subjective momentary experience of surprise and delight, the “aha” or “eureka” moment (Bowden et al., 2005).

Neuroscience typically contrasts insight with analytical reasoning within problem solving. A directly perceivable difference between the two seems to be a more or less gradual progress toward a solution in analytical thinking (Smith and Kounios, 1996), while individuals are abruptly surprised by the latter during an insightful solution (Metcalfe and Wiebe, 1987). Thus, insight is believed to depend by a large degree (but not completely) on unconscious mental processing, as we will see in the next sections (Sandkühler and Bhattacharya, 2008; Shen et al., 2013, 2018; Weisberg, 2013).

Comparative cognition has attempted to tackle the presence of insight in animals by rating the speed of their performance on technical problem or their ability to transfer information from one task to another (Seed and Boogert, 2013).

One issue with this may be that, as mentioned earlier, there are no insight problems, only insight solutions; a problem designed to be solved by insight can also be solved by other processes (van Steenburgh et al., 2012). Epstein et al. (1984) tried to highlight this issue in a popular paper which showed that pigeons solved seemingly complex problems spontaneously by “chaining” blocks of previously learned information.

Neuroscience’s results and advances have been able to compensate a lack of theoretical consistency regarding insight. Cognitive research on animal insight, on the other hand, has been limited to the creativity of experimental designs, with no apparent chance of ending long-running debates stemming from two opposing schools of thought, cognitive psychology and behaviorism, “romantics” against “killjoys” (Shettleworth, 2010, 2012; Call, 2013; Starzak and Gray, 2021). While we believe that the progress of comparative cognition feeds (as a dissipative structure) on the continued conflict between the two positions, the lack of experimental progress has kept these discussions in an impasse (e.g., Heinrich, 1995; Kacelnik, 2009; Chittka et al., 2012; Taylor et al., 2012; Emery, 2013; Starzak and Gray, 2021).

Today we know that insight is a measurable phenomenon with a physiological basis that is beginning to be revealed (Shen et al., 2018). Moreover, it makes little sense to set the phenomenon apart from associative learning and experience (Shettleworth, 2010, 2012; Hanus et al., 2011; Call, 2013; Shen et al., 2018; Ebel et al., 2020). Insight does not mean developing de novo behaviors to solve a problem, but to find a solution by restructuring the problem, even if the agent reorganizes old experiences to apply them to a novel context.

Although insight involves making the nonobvious seem obvious, and even tends to correlate with a higher success rate at problem solving (higher successful rate, Salvi et al., 2016; Webb et al., 2016; but see, Stuyck et al., 2021), a successful restructuring does not necessarily imply a correct conceptualization of the full nature of the problem, and an answer obtained by insight need not necessarily be correct (Kounios and Beeman, 2014). Just as a feeling of understanding does not equate to a true understanding of the problem, we must thus be careful in equating insight with understanding or suggesting that one predicts the other.

Insight may exist in animals outside humans and could even be relatively widespread in nature (e.g., Shettleworth, 2012; Pezzulo et al., 2014). Yet to proficiently tackle the phenomenon in non-verbal species is an unsolved problem in comparative cognition.

While rodent studies suggest that insight does not require sophisticated cognition, the role of the prefrontal cortex in important insight stages may suggest insightful solutions are more likely to emerge in species that have highly developed and functionally equivalent brain regions (Shettleworth, 2010, 2012; Call, 2013; Olkowicz et al., 2016; Shen et al., 2018).

Methodologies, such as the priming of different brain hemispheres, related to insight (which function similarly in non-human primates as in humans) as well as new technologies in animal eye tracking open the door to technically challenging targeted studies in species other than our own (Krupenye et al., 2016; Shen et al., 2018; Völter et al., 2020; Ben-Haim et al., 2021).

The crucial role of subjective experience in insight, as well as the traditional reliance on verbal reports in a large number of studies, makes it tempting to conclude that the study of insight is inaccessible in non-human animals. Nonetheless, other signatures of insight do exist (e.g., Kounios and Beeman, 2014). Apart from EEG and fMRI studies, evidence of human insight stems also from eye tracking studies (e.g., Salvi, 2013; Salvi et al., 2016; Huang, 2017), grip strength (Laukkonen et al., 2021), heart rate (Hill and Kemp, 2018), pupil dilation, and eye movement (with pupil dilation happening only just prior to an insightful event, and an increase in microsaccade rate coinciding with analytic responses; Salvi et al., 2020). Moreover, it has been shown repeatably that agents do not even necessarily need to solve the problem. A promising approach could be to confront an animal with a problem and then, after a period unsuccessful interaction, to suddenly show the solution and record the response (e.g., Kizilirmak et al., 2016; Webb et al., 2019).

Even the “aha” moment itself might be accessible to study in non-verbal subjects, given the expected physiological emotional response that follows it. We know that many animals show an emotional response while learning how to solve tasks (independent from the presence of a reward; e.g., cows, Hagen and Broom, 2004; goats, Langbein et al., 2004; horses, Mengoli et al., 2014; dogs, McGowan et al., 2014; dolphins, Clark et al., 2013). Studying insight through the presentation of a solution would thus require both a behavioral analysis (as in traditional contrafreeloading tests or yoked experimental designs; e.g., Hagen and Broom, 2004; Rosenberger et al., 2020) as well as a physiological one. Artificially altering the transparency of the path toward the solution, and altering the time spent at an apparent impasse, may allow us to predict and modify the intensity of the respective physiological (as it would be an increased heart rate; Hill and Kemp, 2018) and behavioral responses (e.g., in dogs, we would predict pupil dilation, tail wagging, and increased general activity; McGowan et al., 2014; Webb et al., 2019; Salvi et al., 2020).

Insight is a measurable phenomenon in humans, and the mechanisms by which it occurs may well be accessible to species other than our own. Thanks to recent progress in neuroscience and human psychology, we are beginning to clarify the (in some cases subtle) differences that distinguish insight problem solving from other processes. Comparative cognition, however, has so far been limited in its approach. Performance-based setups using technical problems in both birds and mammals have produced highly interesting and suggestive, yet, ambivalent evidence on animal insight (e.g., Heinrich, 1995; Mendes et al., 2007; Bird and Emery, 2009a,b; Laumer et al., 2017, 2018; von Bayern et al., 2018). We are optimistic that accomplishments in neuroscience and human psychology over the past decade can be incorporated into and inspire future comparative cognition studies in their ongoing quest to learn about the capacity for insight in species other than our own.

AO-M wrote the first draft. AO-M and AA finished the manuscript. All authors contributed to the article and approved the submitted version.

The authors are funded by the WWTF Project CS18-023 and START project Y 01309 by the Austrian Science Fund (FWF) to AA.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.