Prevalence of CE

Thirty published studies met our search criteria (Table S1; Supplementary material). The majority of CE studies (18 articles, 60.0%) were performed on primates or marine mammals. We found that 22/30 studies (73.3%) cited the original CE definition (Clark Reference Clark2011). Three of these empirical studies were authored by Clark and colleagues. Studies that did not cite the original 2011 definition (26.6%) provided similar definitions, including references to problem-solving and/or having additional control over the environment. Initially, we also hoped to assess whether the CE studies in our literature search met the conditions of CE according to scenario 3 in Table 1. However, this assessment was not possible because studies seldom reported whether the putative CE presented a high enough level of challenge to animals, or the net welfare outcome. Due to positive publication biases in animal welfare science and allied fields (Van der Schot & Phillips Reference Van der Schot and Phillips2013), enrichment practitioners might be deterred from using CE if it is too difficult to evidence a significant improvement in animal welfare. In a survey of 177 zoo animal care staff (Hall et al. Reference Hall, McGill, Sherwen and Doyle2021), respondents generally saw the value of CE but felt discouraged from using CE because they felt that it required specialist knowledge and resources.

CE subtypes and rationale

We identified five broad subtypes of CE: maze navigation (n = 9; 30.0%); object manipulation (n = 9; 30.0%); associative learning opportunities (n = 5; 16.7%); cognitive task batteries, such as match-to-sample or object discrimination tasks traditionally found in cognitive research (n = 5; 16.7%); and social group manipulation (n = 2; 6.6%). For our classification purposes, we defined a maze as a system of intersecting paths and blind alleys that must be navigated (VandenBos Reference VandenBos2015). One hundred percent of studies in our literature search met at least one of their stated goals of CE. Our literature search showed that, despite a high level of agreement on the definition of CE, the rationale for using CE varied across studies. In some cases, CE aimed to stimulate the expression of cognitive skills that were absent, assumed to be weak, or deemed to be important for optimal biological functioning (e.g. rainbow trout [Oncorhynchus mykiss] were trained to associatively learn the availability of food to forage more effectively; Kleiber et al. Reference Kleiber, Le-Calvez, Kerneis, Batard, Goardon, Labbé, Brunet, Ferreira, Guesdon, Calandreau and Colson2022). While in other cases, existing or traditional forms of husbandry were evaluated as putative forms of CE (e.g. mixed-species housing of tufted capuchin monkeys [Cebus apella] and squirrel monkey [Saimiri sciureus]; Daoudi et al. Reference Daoudi, Badihi and Buchanan-Smith2017).

Summary

In summary, Clark’s Reference Clark2011 CE definition has been frequently cited within the past 15 years and appears to be the most common definition of CE used in peer-reviewed research on captive animal welfare. However, it is difficult to accurately report whether CE has been sufficiently challenging, coped with, or has led to sufficiently positive welfare outcomes, and subtypes of CE are difficult to compare against each other. So, after 15 years of advancement in the field of animal welfare science, CE deserves a critical appraisal, including the definition of CE itself. The remainder of this opinion paper will discuss current scholarly consensus on positive animal welfare and challenge (see Positive animal welfare, skill development and expression), leading to an updated definition of CE (see An updated definition of cognitive enrichment) and our proposal for game-based CE (see Gameplay: The next frontier in cognitive enrichment? and Flow game case study).

Issue

Although the original definition was inspired by Meehan and Mench’s (Reference Meehan and Mench2007) challenge framework, it does not explicitly mention challenge.

Solution

A more conditional definition (i.e. ‘CE arises when X, Y and Z happen’) avoids the ‘near-miss’ scenarios illustrated in Table 1.

Issue

CE was originally defined quite narrowly as a ‘task’. A task is normally associated with high levels of experimental control, which could be off-putting to many captive settings.

Solution

A challenge does not have to be a task. It can be spontaneously induced by changes of a biotic (e.g. social) and abiotic (e.g. climatic) origin or artificially induced by a ‘game’ (see Gameplay: The next frontier in cognitive enrichment?).

The natural or artificial origin of CE is not important, as long as an animal’s response can be evaluated against the operational definition.

Issue

We agree with a criticism of the original 2011 definition (Jaakkola Reference Jaakkola2024) that ‘engaging cognitive skills’ does not provide a clear enough distinction between CE and cognition in everyday life. Jaakkola used the simple example of eating food from a bowl, which engages several cognitive skills in an attempt to solve a problem (hunger) and control some aspect of the environment (how much food to eat and when to leave).

Solution

Our updated definition asserts that CE requires an animal to ‘develop and express its evolved cognitive skills’. In line with Rault et al. (Reference Rault, Bateson, Boissy, Forkman, Grinde, Gygax and Jensen2025), this acknowledges how skill development (i.e. becoming more competent) and skill expression (i.e. expressing competence) are aspects of cognitive operation that may go beyond an animal’s everyday baseline.

Issue

The original definition mentions opportunities to solve problems and have control over the environment. With hindsight, the word ‘opportunity’ is ambiguous, and it is unclear whether an overall solution is necessary in CE. As discussed in Positive animal welfare, skill development and expression, finding a final solution to a challenge may be associated with an acute positive affective state in animals, but high engagement and making progress towards finding a solution may also be beneficial. Animals may progress or find the solution to a problem through hard work or by accident (i.e. innovation; Reader & Laland Reference Reader and Laland2003).

Solution

Therefore, we have replaced ‘solve’ and ‘control’ in the original definition with ‘cope’. We also refer to coping as a ‘process’ rather than a discrete event, to align with the concepts of challenge as a process (Meehan & Mench Reference Meehan and Mench2007) and cognitive processes (Mendl et al. Reference Mendl, Burman and Paul2010).

Issue

The original CE definition did not actually specify that well-being should be in the positive direction. For completeness, we have added ‘positive’ to the definition. Furthermore, ‘validated indicators of welfare’ implied that CE is only suitable for species with a rich empirical literature on validated behavioural (and perhaps physiological) welfare indicators, which could be off-putting to many animal settings (Jaakkola Reference Jaakkola2024).

Solution

CE should aim for ‘a net positive affective state’ compared to the animal’s baseline, in a range of species. This aligns CE with the concept of affective balance (Webb et al. Reference Webb, Veenhoven, Harfeld and Jensen2019) and predominantly positive affective states (Rault et al. Reference Rault, Bateson, Boissy, Forkman, Grinde, Gygax and Jensen2025).

A game fosters both skill development and expression (competence)

Humans play games for general stimulation, leisure/occupation/distraction, social connection and external rewards (see Lazzaro Reference Lazzaro2009). But from a psychological research perspective, games have typically been used to develop a participant’s skill over a particular period of time (e.g. to improve working memory over a period of several hours or days). Existing and newly developed skills are also expressed during gameplay. Psychologists have also recognised the value of games in extending participant engagement and performance (Allen et al. Reference Allen, Brändle, Botvinick, Fan, Gershman, Gopnik and Schulz2024). Traditional cognitive tasks that test a participant’s skill at a particular point in time (e.g. to find if working memory exists at a particular threshold) can be ‘gamified’ to extend their duration and therefore shift from skill testing to skill development (Lumsden et al. Reference Lumsden, Edwards, Lawrence, Coyle and Munafò2016; Suh et al. Reference Suh, Wagner and Liu2018). In contrast to a traditional task, where a skill is often scored as present or absent during a single trial, game performance can be assessed in more flexible, non-binary ways.

A game imposes measurable parameters

A game’s parameters are substantially known and measurable, thus making it a powerful tool for research (Allen et al. Reference Allen, Brändle, Botvinick, Fan, Gershman, Gopnik and Schulz2024; Clark Reference Clark2025). And unlike a task, a game tends to have layers or combinations of rules that change as gameplay unfolds, dependent or independent of the player’s performance. Game parameters are consistent with the zone of proximal development (ZPD) theory from human developmental psychology (Vygotsky Reference Vygotsky1978). ZPD theory posits that skill development (in children) is facilitated by creating a ‘scaffold’ environment that provides just the right level of support to help them reach a high level of potential and agency. The supportive element of the ZPD is crucial: a person would not develop to such a high level completely independently. Examples of scaffolding are rules, training and rewards, and demonstrating solutions or ‘teaching’. In adults, the ZPD is similar to the human skill-challenge balance (Positive animal welfare, skill development and expression) and avoiding boredom and cognitive overload facilitates a psychological state called ‘flow’ (Nakamura & Csikszentmihalyi Reference Nakamura, Csikszentmihalyi, Snyder and Lopez2002; Csikszentmihalyi et al. Reference Csikszentmihalyi, Csikszentmihalyi, Abuhamdeh, Nakamura and Csikszentmihalhi2014). While the ZPD supports reaching a high level of potential, it is important to note ZPD does not support ‘life or death’ activities and does not lead to maximum possible skill development and expression. A state of maximum performance known as ‘clutch state’ in humans occurs during high-endurance sporting activities; it cannot be sustained long-term due to high energetic demands and is associated with maximum physiological arousal and negative affective state (Swann et al. Reference Swann, Jackman, Schweickle and Vella2019).

To our knowledge, ZPD theory is yet to be applied to animal enrichment. This is surprising because, as previously discussed, a skill-challenge balance has already been identified as beneficial for animal welfare because it minimises the negative affective states of boredom and frustration (for a review, see Meehan & Mench Reference Meehan and Mench2007). If we apply ZPD theory to animal CE, we can think of a game as having an inbuilt scaffold designed and controlled by the practitioner, supporting the animal with rules and feedback and matching the level of challenge to their current skill level.

Games involve two forms of play

Using simple and accessible language, games are ‘played’, but it is important to distinguish between two possible types of play during game-based CE. First, ‘gameplay’ is how an animal directly interacts with a game, through a tangible or digital interface. This type of play is linked to functional skill development or ‘training for the unexpected’ (Held & Špinka Reference Held and Špinka2011; Burghardt Reference Burghardt2005). Second, ‘joyful’ (‘luxury’) play, often involving one’s own body, toys, or social partners, is an established positive animal welfare indicator (Burghardt Reference Burghardt2005; Yeates & Main Reference Yeates and Main2008; Nelson et al. Reference Nelson, Taylor, Cartmill, Lyn, Robinson, Janik and Allen2023, although see Ahloy-Dallaire et al. Reference Ahloy-Dallaire, Espinosa and Mason2018 for a counter-argument).

Given that several lines of evidence already link skill development and intrinsic reward or affective state in animals (see Positive animal welfare, skill development and expression), it is probable that gameplay in some animal taxa will be associated with emotional responses indicating how well the animal perceives itself coping with challenge. The evolution of high technical skills in humans has been linked to the evolution of positive feelings as a form of intrinsic reward (Broaden and Build theory; Fredrickson Reference Fredrickson1998, Reference Fredrickson2004), so if we assume some evolutionary continuity, skill development in some taxa may also be moderated by their affective state. In humans, there is still debate over two prevailing theories of the evolution of ‘happiness’ or positive affect: hedonism (a more immediate drive towards behaviours that cause pleasure and/or avoid pain) and eudaimonism (a longer-term drive towards the satisfaction of self-development and full functionality: Ryan & Deci Reference Ryan and Deci2001; Delle Fave et al. Reference Delle Fave, Brdar, Freire, Vella-Brodrick and Wissing2011; Mauss & Ford Reference Mauss and Ford2026). Rather than one happiness theory being correct, hedonism and eudaimonism may be expressed across different time-frames. An interesting future research question is whether an animal’s current (in-the-moment) and retrospective experiences of CE differ. Welfare indicators during CE might be more hedonistic (pleasure-like), whereas welfare indicators after CE might be more eudaimonic (satisfaction-like).

Innovation games

Innovation refers to inventing a novel behaviour to reach a win state or using an existing behaviour on a novel problem to win suddenly (innovation as conceptualised by Reader & Laland Reference Reader and Laland2003; Tebbich et al. Reference Tebbich, Griffin, Peschl and Sterelny2016). The justification for an innovation game is that finding sudden solutions to abstract problems (Figure 1, top) can cause acute bursts of positive emotion and/or positive physiological arousal in animals, the so-called ‘Eureka effect’ (Positive animal welfare, skill development and expression). While we acknowledge evidence of this effect is currently restricted to cattle, dogs and possibly dolphins, it is a promising CE avenue to explore. The Eureka effect in humans has been studied by comparing participant responses on tasks that do and do not naturally converge towards a logical solution; in other words, comparing the emotional effect of an expected and unexpected solution (Lu & Singer Reference Lu and Singer2023). The same experimental design could be applied to animals and is supported by evidence that Japanese macaques (Macaca fuscata) showed increased behavioural signs of arousal or ‘surprisal’ when they encountered an unexpected reward during cognitive testing (Iki & Adachi Reference Iki and Adachi2025). Interestingly, the behavioural sign of arousal in the macaque study was self-scratching, which is typically used to indicate negative arousal (anxiety) in primates (Maestripieri et al. Reference Maestripieri, Schino, Aureli and Troisi1992). Two further research questions are whether the Eureka effect is more intense in animals exposed to a higher level of challenge relative to their everyday baseline, and whether animals that can acquire a game solution via social learning (direct observation, then emulation or imitation) have a weaker Eureka moment than asocial innovators.

An Innovation game has a low-intensity ZPD because it encourages the animal to try out many possible behavioural solutions and therefore express many pre-existing skills to overcome a challenge. Imagine the following example for a pig: a novel wheel apparatus contains food occluded behind a rotating plate (resembling an old rotary telephone; Clark et al. Reference Clark, Burdass, Kavanagh and King2025). The intended solution is to rotate the plate until food is exposed through a single hole. The pig may attempt many different behaviours in a trial-and-error process to find the win state to this very novel challenge, such as general nosing, pressing and pulling. It should be understood that spontaneous innovation to any challenge is never guaranteed. The ability to innovate is not universal across the animal kingdom, and even within innovating species, some individuals in a group do not innovate (Reader & Laland Reference Reader and Laland2003; Ramsey et al. Reference Ramsey, Bastian and Van Schaik2007; Brosnan & Hopper Reference Brosnan and Hopper2014). This being said, the capacity to innovate continues to be reported in unsuspected and previously understudied species (e.g. bat-eared foxes [Otocyon megalotis]; Petelle et al. Reference Petelle, Jacobs and le Roux2023; orchard bees [Osmia cornuta]; Collado et al. Reference Collado, Menzel, Sol and Bartomeus2021). Linked to Pre-existing CE games, we highly recommend evaluating existing innovation tasks used in comparative cognition research as putative CE.

Flow games

As discussed in Gameplay: The next frontier in cognitive enrichment?, Flow is a psychological state in humans that integrates high focus and skill development with positive affect (Nakamura & Csikszentmihalyi Reference Nakamura, Csikszentmihalyi, Snyder and Lopez2002; Csikszentmihalyi et al. Reference Csikszentmihalyi, Csikszentmihalyi, Abuhamdeh, Nakamura and Csikszentmihalhi2014). Flow state has not been definitively evidenced in animals to date, but evidence of high task engagement and a motivation to forego extrinsic rewards has begun to shape an exciting new direction of research (Clark Reference Clark2023; Hintze & Yee Reference Hintze and Yee2023). For now, we suggest working towards the measurement of one or more indicators of Flow evidenced in humans: high focus, disinterest in the wider environment and extrinsic reward, easeful movements, and positive emotion. As in humans, Flow games for animals should create a skill-challenge balance (see Positive animal welfare, skill development and expression), thereby avoiding frustration/anxiety and boredom/apathy (Table 1; Section 4.1, bottom box). A Flow game has relatively high-intensity ZPD scaffolding because skills are developed gradually within quite narrow parameters.

A Flow game should adaptively respond so that the skill-challenge balance is maintained during play. The analogy of a tightrope is useful here. Maintaining a skill-challenge balance (staying on the tightrope) and avoiding frustration/anxiety and boredom/apathy (falling off the tightrope) requires constant, active challenge adjustments. If the game design is too simplistic, surplus skill development will naturally pull the player off the tightrope and towards boredom. In contrast, a game that has elements that are perceptually or cognitively inappropriate will pull the player towards frustration.

Computerised systems that can perform frequent ‘tightrope’ calibrations will be very beneficial for Flow games. Fortunately, computer technology can recognise individual animals housed in groups and offer customised tasks using dynamic difficulty adjustment (Calapai et al. Reference Calapai, Cabrera-Moreno, Moser and Jeschke2022, Reference Calapai, Pfefferle, Cassidy, Nazari, Yurt, Brockhausen and Treue2023; see Flow game case study). However, we fully acknowledge that computerised systems are a substantial investment and are not feasible for most captive settings. A low-tech alternative is to present different levels of a modular analogue game over time, based on an animal’s current performance (for example, providing a harder module if the animal is performing at a high level). The ability to recognise and rapidly respond to an animal’s gameplay performance is crucial for this to be effective, and the rules of the game need to be explicit to both the player and the practitioner. Referring to the tightrope analogy once again, this may also involve rapid recognition of behavioural indicators of frustration and boredom/disengagement if relevant, in addition to the rate at which the animal is progressing through levels of challenge. While the low-tech approach to flow games might sound difficult in practice, animal cognition researchers are experienced in rapidly presenting tasks to animals during blocks of trials and responding to their immediate performance, so this experience could be leveraged by incorporating flow games into studies of animal cognition.

Choosing Innovation versus Flow

We encourage CE practitioners to explore options for Innovation and Flow games – either through the extension and modification of existing CE and cognitive tasks, or through novel designs. CE practitioners should consider whether their target species/individual is likely to innovate and how sensitive it may be to reward loss or failure in a game. For example, even though great apes can innovate, not all individuals in a group usually do so (Manrique et al. 2013). Great apes can also develop high emotional sensitivity to response errors, reward loss, or inconsistent task feedback but this can vary per individual (Elder & Menzel Reference Elder and Menzel2001; Yamanashi & Matsuzawa Reference Yamanashi and Matsuzawa2010), which could make a Flow game more suitable than an Innovation game, as long as the game is responsive to individual performance. Conversely, an Innovation game may be more suitable for settings with low experimental control and low technology.

To end this section, we cover some more general CE game design principles. As a first step, it is wise to work with species that have already undergone extensive cognitive research, such as rats and mice, chickens, corvids, dogs, pigs and cows, primates, dolphins, and sea lions (Farrar et al. Reference Farrar, Altschul, Fischer, Van Der Mescht, Placì, Troisi, Vernouillet, Clayton and Ostojić2020), so that game design is based on cognitive evidence. CE game design should be compatible with a target species’ sensory system, anatomy, ultimate (evolved) motivations, and proximate motivations (Meehan & Mench Reference Meehan and Mench2007; Clark Reference Clark2011). If existing cognitive tasks will be gamified (i.e. adding more measurable parameters, rules and opportunities to develop skill), we caution that humans can develop disordered, addictive gaming behaviour towards very high levels of gamification. This includes commercial electronic gambling machines (i.e. slot machines) that encourage rapid and repetitive actions and have ambiguous performance indicators such as losses disguised as wins (Murch & Clark Reference Murch and Clark2016). Even though making risky decisions is a naturalistic strategy for animals to encounter and exploit new resources (De Petrillo & Rosati Reference De Petrillo, Rosati, Kaufman, Call and Kaufman2021), ‘unskilled gambling’ (i.e. risking resources for little or no beneficial reward pay-off) can readily occur in rodents, primates and pigeons in a laboratory setting (Laude et al. Reference Laude, Stagner and Zentall2014; Hall et al. Reference Hall, Smith, Russell, Lambeth, Schapiro and Brosnan2019; Zentall Reference Zentall2023). This being said, there is a lack of research on precisely how gambling could become addictive to the point of harming welfare in animals (Clark Reference Clark2025), and our approach to CE throughout this paper has emphasised close animal monitoring and being prepared to modify challenge if it has negative welfare outcomes.

Fine temporal resolution

CE evaluation has previously focused on CE as an environmental ‘event’ rather than a process (Table S1; Supplementary material) and has also focused on fairly broad behavioural effects. As previously discussed, this does not align with how animals naturally encounter and deal with challenges (Meehan & Mench Reference Meehan and Mench2007) and does not capture all of an animal’s meaningful cognitive and emotional responses happening in sequence. For example, between-group CE experiments (enriched group vs control group) and within-group CE experiments (baseline vs enrichment vs post-enrichment) have evaluated how total durations, frequencies or averages of behaviours such as exploration, aggression and play are affected by CE presence/absence (between-group example: zebra fish [Danio rerio]; Gatto et al. Reference Gatto, Varracchio, Bertolucci and Lucon-Xiccato2024; domestic cats [Felis catus]; Gourkow & Phillips Reference Gourkow and Phillips2016; within-group example: black rhinoceros [Diceros bicornis]; Krebs & Watters Reference Krebs and Watters2017). Although it is important to assess the lingering effects of enrichment that extend beyond its physical presence (e.g. Fairhurst et al. Reference Fairhurst, Frey, Reichert, Szelest, Kelly and Bortolotti2011; Woodward et al. Reference Woodward, Winder and Watt2019), it is also important to pay closer attention to what happens during CE presence to reflect the increasing integration of cognition, emotion and mood in animal welfare. In particular, the ‘Affective Balance’ model portrays welfare as a dynamic sum or accumulation of positive and negative experiences over time (Webb et al. Reference Webb, Veenhoven, Harfeld and Jensen2019; Kremer et al. Reference Kremer, Holkenborg, Reimert, Bolhuis and Webb2020; Paul & Mendl Reference Paul and Mendl2025) and requires a fine scale of analysis. For example, instantaneous sampling at 10-min intervals could easily miss a sudden ‘Eureka moment’ in an Innovation game.

To illustrate how future CE game evaluation could work in practice, imagine the following game scenario (Innovation or Flow). A chimpanzee (Pan troglodytes) plays a maze game, and all responses (gameplay and off-game behaviour) are plotted across time (Figure 2). The Bayesian brain framework of animal welfare (Lecorps & Weary Reference Lecorps and Weary2024) predicts that prior game performance should impact future emotional response. Visualising responses to a CE game on a timeline thus allows us to use Bayesian framework to assess temporal relationships, such as whether negative emotions follow a losing streak or sudden loss, and positive emotions follow a winning streak or sudden win.

Figure 2.

Game engagement, performance and emotion synchronised across time during a cognitive enrichment session for a chimpanzee (Pan troglodytes). The Engagement domain refers to whether the chimpanzee is using the game or not. The Performance domain is a deeper measurement of how well the chimpanzee is coping with the challenge and progressing towards the win state. The Emotion domain refers to any behavioural response that may reflect emotion, such as game-directed aggression or self-scratching. And finally, the Affective Balance domain represents the sum of positive and negative emotions across time. Critically, all responses are accurately plotted on the timeline to examine cause-effect relationships across multiple domains.

Figure 2. Long description

There is a photograph of a chimpanzee on the left panel. The chimpanzee has their back to the camera, and is interacting with a maze-like game with multiple interconnected cells like a honeycomb. The right panel, illustrates a theoretical scenario where the chimpanzee experiences fluctuations in engagement, performance and emotion during gameplay. The chimpanzee is initially disengaged yet expresses some positive emotion. Then they have a period of high game engagement with rising performance. They disengage briefly and express a negative emotion after their performance drops. The same sort of pattern occurs one again where positive emotions follow a rise in performance, and a disengagement and negative emotions follow a fall in performance.

The functional role of frustration in cognitive enrichment

Although game-based CE broadly aims to avoid frustration by scaffolding (Innovation and Flow games) and maintaining a skill-challenge balance (Flow games), we expect that all CE will induce at least a small amount of initial frustration. We previously described how environmental change can either challenge or comfort an animal (or lead to boredom, see A critical review of CE to date). To expand on this, challenge activates an animal’s physiological stress response (Wiepkema & Koolhaas Reference Wiepkema and Koolhaas1993; Moburg & Mench 2000; Broom & Johnson Reference Broom and Johnson2019; Wu et al. Reference Wu, Tilbrook, Maloney and Blache2025). Even though ‘stress’ often has negative connotations and is conflated with poor animal welfare (Morgan & Tromborg Reference Morgan and Tromborg2007), stress in response to challenge is normal and functional. And at a behavioural level, Meehan and Mench (Reference Meehan and Mench2007) discussed how frustration is normal and functional when an animal is challenged because it represents motivation to alter its conditions (see also Toates Reference Toates1986; Wiepkema & Koolhaas Reference Wiepkema and Koolhaas1993). Meehan and Mench (Reference Meehan and Mench2007) proceeded to argue that while initial frustration is to be expected when an animal perceives a challenge, the challenge must either be solvable or escapable so that this frustration can ultimately be removed.

It is ethically important to acknowledge whether the animal and/or the practitioner has control over an animal’s experience of frustration. An animal may or may not have control over removing its experience of frustration; it may happen actively by solving or escaping the challenge, or passively by the environment changing. Given the physiological and behavioural ‘normalcy’ of stress and frustration, we imagine that most CE practitioners will be ethically satisfied if CE causes some initial frustration and it can easily be reduced or removed. This also aligns with Rault et al.’s (Reference Rault, Bateson, Boissy, Forkman, Grinde, Gygax and Jensen2025) definition of positive animal welfare which is ‘predominantly’ but not exclusively a positive affective state, allowing for some negative emotions to arise over time. And in human flow, the Affective Change hypothesis (Baumann & Scheffer Reference Baumann and Scheffer2010) emphasises that experiencing positive and negative affective states in close temporal proximity (the alternation between ‘seeing difficulty’ and ‘mastering difficulty’) helps a person reach an overall positive experience.

The fine-scale temporal analysis we proposed above (Figure 2) is useful because it allows us to observe not only the frequency of frustration events, but their timing and context. But how much frustration is too much? Currently, the field of animal affect (also known as affective neuroscience; Panksepp Reference Panksepp2004) has not been able to produce a full cost-benefit analysis of negative versus positive emotions for any species. For example, we cannot say for a given species that one frustrating event is outweighed by three joyful play events. The Affective Balance model (Figure 2, bottom) is currently a rough binary method based on emotional valence (i.e. score 1 for a positive emotion, score minus 1 for a negative emotion), which clearly has limitations (Webb et al. Reference Webb, Veenhoven, Harfeld and Jensen2019). But it is worth remembering that humans also differ in the extent to which positive emotions are valued and weighed against negative emotions (Affect Valuation theory: Tsai et al. Reference Tsai, Knutson and Fung2006; Willroth et al. Reference Willroth, Young, Tamir and Mauss2023).

If an animal cannot solve CE, this could be assessed as ‘distress’ or a failure to cope (Broom & Johnson Reference Broom and Johnson2019). Distress is associated with a lack of skill or control (Wu et al. Reference Wu, Tilbrook, Maloney and Blache2025), whereas eustress is associated with challenges that can be overcome because skills are developed and expressed. We hope that by using fine-scale assessment, CE would never reach a point of animal distress because the practitioner would already be prepared to modify the challenge if necessary. This could be through rule simplification, rule reduction, enhanced positive feedback, positive reinforcement training, or demonstrating solutions.

Although one CE exposure is associated with acceptable brief initial frustration, repeated CE exposures over time could add up to repeated instances of frustration. Our counter-argument is that animals should regularly encounter challenges (Toates Reference Toates1986; Moberg & Mench Reference Moberg and Mench2000) and making challenge a life-long process may actually decrease frustration across the lifespan, as the animal gains competence and resilience (Rault et al. Reference Rault, Bateson, Boissy, Forkman, Grinde, Gygax and Jensen2025). Clark (Reference Clark2011) used the analogy of a British television game show ‘The Crystal Maze’ where a team of human contestants voluntarily took part in challenges in different game zones based on their respective strengths (e.g. motor skills, memory, strategy) and motivations. A vision of animal enclosures containing voluntary gameplay zones is ambitious, even for the most major zoos and research centres. But if ambition eventually becomes reality, the concept of CE could evolve into cognitive husbandry (i.e. CE has evolved into CH), a lifelong commitment to avoid loss of brain and cognitive function in captive environments (Clark et al. Reference Clark, Greggor, Montgomery and Plotnik2023).