In its purest form, curiosity refers to the drive to gain information in the absence of clear instrumental goals such as food or shelter (1). From a behavioral biology perspective, curiosity is understood to have evolutionary value in that gaining information may ultimately enable an individual to better exploit resources and manage threats (2). In the immediate-term, however, information-seeking itself can become its own reward and is sometimes prioritized over clear, material gains (1). As such, curiosity has fascinated scholars throughout history, especially in the middle of the 20th century when psychologists and ethologists intensively studied curiosity in species across the animal kingdom (3–5). The past decade has seen a resurgence of interest in the study of curiosity across multiple disciplines, with investigations of curiosity in a range of species including raptors (6), tortoises (7), orangutans (8), honeybees (9), and in one multispecies study, sheep, penguins, lemurs, cockatoos, tortoises, and kangaroos (10).

To understand what features may drive curiosity, Loewenstein (11) proposed that information seeking behavior is elicited when there is an information gap between what is known and what is unknown, without going beyond the cognitive capacity of the individual. Two ways for perceptual objects to have enhanced information potential are by (i) being complex or (ii) being unusual. From a visual perspective, complex objects are those with many parts, shapes, patterns, and/or colors. Previous work has confirmed that visual complexity attracts greater attention from, for example, raptors (6), rats (12), and humans (13). Unusual objects are either ones that are very different from those that the individual has experienced in its own lifetime or, from an evolutionary perspective, those which would not normally be encountered in the ecological niche of that species—in other words, objects that are ecologically implausible. Recent work in orangutans, for example, has shown that they are particularly interested in unusual, ecologically implausible objects (8). Interestingly, some prey animals are also known to be highly motivated to gather information about predators, readily inspecting potential predators despite the evident risks (14, 15). Thus, previous theoretical and empirical work suggests that complexity, unusualness, and predatory potential are all relevant dimensions in eliciting perceptual curiosity.

As these proposed curiosity-eliciting dimensions indicate, experiencing curiosity and curiosity-producing situations are ambivalently charged—they have the possibility of taking on a positive or negative valence, thus complicating the relationship between curiosity and welfare. Human studies have found that heightened curiosity can be associated with positive affect and wellbeing [e.g., (16)], but also frustration (17, 18), sensation-seeking, boredom-avoidance, and substance abuse (19). When measured as exploratory behavior, curiosity's bivalent nature is evident in nonhuman animals as well. On the one hand, opportunities for exploration and exploratory behavior are associated with improved welfare and cognition (20–23). For example, rats maintained in complex social housing were found to have enhanced exploratory tendencies (giving up known rewards and risking aversive conditions in order to explore new spaces) compared to isolated rats in barren cages (24). Conversely, other forms of exploration have been associated with poor animal welfare. For example, undifferentiated exploratory behavior of novel stimuli (i.e., neophilia) in mink was more apparent in individuals housed in non-enriched cages compared to those housed in enriched cages (25, 26).

These and other studies suggests that exploratory behavior in the form of indiscriminate neophilia (quick, fleeting exploratory behavior) may be reflective of an avoidance motivation (escaping aversive or fear-inducing conditions), especially in situations likely to induce boredom. Boredom, a negative state caused by a lack of behavioral opportunities, is identified as a serious welfare concern (27) and an important area of study in animal behavior (28). In contrast, exploratory behavior in the form of free-choice, targeted, and sustained information seeking has been associated with positive wellbeing in both humans and other animals (16, 22, 23). Distinguishing between these two types of exploratory behavior—indiscriminate neophilia vs. targeted and sustained information-seeking—is necessary when attempting to understand the relationship between curiosity and current welfare state.

By investigating curiosity in zebrafish (Danio rerio), the present study aimed to provide crucial data for interdisciplinary, comparative work looking at curiosity and welfare across the animal kingdom. Knowledge about curiosity in zebrafish represents a particularly useful starting place for studying curiosity in fish because it can indicate whether similar issues may be at stake for other species of fish, including the welfare of fish in aquaculture (29). As one of the most studied species of fish (30, 31), information about zebrafish capacities and interests has the potential to set expectations for fishes in general and can influence fish welfare standards. Moreover, as a member of the Cyprinidae family, zebrafish can provide insights into patterns of behavior and motivation for Cyprinidae, one of the most farmed families of fishes (32).

By characterizing the exploratory behavior of groups of zebrafish toward novel objects, we hypothesized that initial interest could range from neophobic (manifesting as avoidance and fear responses), to indifference (little to no behavioral changes in response to the objects), to curiosity (exploration of and attraction to the objects). We further distinguished between indiscriminate neophilia and targeted information-seeking. We operationalized indiscriminate neophilia as fleeting and uniform or haphazard interest across objects, i.e., short, irregular, nondifferentiated interest durations. In contrast, we operationalized targeted information-seeking behavior as sustained, differential attention, i.e., the presence of distinctive and consistently high interest for protracted durations of time for only some object presentations.

Pending evidence of targeted information-seeking behavior, we also sought to determine whether there was consistency across tanks as to which objects attracted the most sustained attention. General consistency in object-level attraction across tanks would indicate that some perceptual property of the objects themselves were driving the differential interest. Accordingly, we aimed to select objects that varied in complexity, ecological plausibility, and predator resemblance, allowing us to assess (i) which of these dimensions might produce the most sustained information seeing, (ii) how the introduction of objects may affect welfare, and (iii) what object properties might provide appropriate cognitive enrichment for this species (31).

To achieve these aims, we observed the behavior of groups of zebrafish toward a range of 30 different novel objects presented one at a time in their home tanks. The zebrafish were housed in relatively large (110 L), semi-naturalistic aquaria that could accommodate multiple behaviors, thus providing the fish with a low-stress, free-choice (explore or not) set up that allowed them to engage with a particular novel object or not. In addition to object-directed behavior, we also recorded aggression, shoal cohesion, swimming coordination, and diving behavior during a baseline period without a novel object and while the object was in the tank. Zebrafish social dynamics are sensitive to environmental manipulations (33), with recent empirical work suggesting that free-choice exploration can increase prosocial behavior and decrease diving (34), whereas potential stressors have the reverse effects (35). Together, this suite of behaviors can provide indications of information-seeking motivation in zebrafish as well as preliminary indications as to the valence (positivity/negativity) of the zebrafishes' response to the novel object exploration opportunities.

Zebrafish began exploring objects almost immediately with a median approach latency of <1 s, a maximum approach latency of 48 s, and 95% of all first approaches occurring within 6 s.

At the beginning of the study (i.e., object presentations 1–10), fish were in the object area (within two body-lengths of the object) more often than expected above baseline, during both the first 100 s and the final 100 s of the 10-min novel object presentation (first-100 s: 0.77 above baseline, 95% Confidence Interval [0.51, 1.03] (henceforth indicated by square brackets only), t(154.66) = 5.77, p < 0.0001; final-100 s: 0.36 [0.10, 0.62] above baseline, t(154.66) = 2.67, p < 0.01; Figure 2). By the end of the study (i.e., object presentations 21–30), overall fish presence in the object area was indistinguishable from baseline (first-100s: 0.06 [−0.10, 0.22] above baseline, t(128.97) = 0.69, p < 0.4; final-100s: 0.16 [0.00, 0.32] above baseline, t(128.97) = 1.93, p < 0.06; Figure 2).

High interest during the first 100 s of a novel object presentation generally corresponded to higher interest during the final 100 s of that presentation, a pattern that remained significant throughout the study (object presentations 1–10: 0.40 [0.28, 0.52], t(58.00) = 6.36, p < 0.0001; object presentations 11–20: 0.51 [0.27, 0.75], t(55.62) = 4.19, p < 0.001; object presentations 21–30: 0.26 [0.06, 0.46], t(49.31) = 2.62, p < 0.02; Figure 3).

In addition to the greater overall interest at the beginning of the study compared to the end of the study, there was also greater variability in interest at the beginning compared to the end: The range in interest for presentations 1–10 was −2.2 to 5.1 vs. the range in interest for presentations 21–30 was −1.3 to 1.9 (Figure 3). At the beginning of the study (object presentations 1–10), object ID accounted for 11% of the variability in interest, with some objects consistently attracting more attention and some objects consistently attracting less attention than other objects [Likelihood Ratio Test: χ2(1) = 7.08, p < 0.01; Figure 4]. Tank ID accounted for 21% of the variability in interest [Likelihood Ratio Test: χ2(1) = 15.30, p < 0.0001], leaving 68% of the variability in object interest unexplained. None of the human-scored dimensions of object characteristics—complexity, ecological plausibility, predator-resemblance—explained variability in zebrafish interest during this period (p's > 0.4). At the end of the study (object presentations 21–30), neither object ID nor tank ID accounted for the (relatively small) variability in interest (p's > 0.2).

The experimental procedures of introducing of novel objects altered zebrafish social behavior and diving behavior, but unlike interest, these effects remained consistent throughout the study. Compared to baseline: agonistic behavior decreased during the first 100 s (-0.26 [-0.32,−0.20], t(13.73) = 7.86, p < 0.0001) and returned to baseline levels by the final 100 s (−0.01 [−0.07, 0.05], t(13.73) = 0.21, p > 0.8); cohesion increased during the first 100 s (0.47 [0.33, 0.61], t(5.91) = 7.04, p < 0.001) and returned to baseline by the final 100 s (0.05 [−0.09, 0.19], t(5.91) = 0.80, p > 0.4); and coordination also increased during the first 100 s (0.68 [0.54, 0.82], t(6.31) = 10.31, p < 0.0001) and returned to baseline by the final 100 s (0.13 [−0.01, 0.27], t(6.31) = 2.05, p < 0.09).

Throughout the study, diving behavior decreased during the first 100 s of object presentations (−0.26 [−0.30, −0.22], t(15.22) = 10.46, p < 0.0001) and returned to baseline by the final 100 s (-0.03 [−0.07, 0.01], t(11.90) = 1.37, p > 0.1).

Importantly, at the beginning of the study (object presentations 1–10), changes in social behavior and diving behavior corresponded to zebrafish interest: novel object presentations that elicited greater interest (vs. less interest) corresponded to decreased aggression (−0.06 [−0.10, −0.02], t(107.40) = 2.42, p < 0.02), increased cohesion (0.10, [0.02, 0.18], t(92.22) = 2.64, p < 0.01), increased coordination (0.08 [0.00, 0.16], t(67.24) = 2.02, p < 0.05), and decreased diving (−0.07 [−0.13, −0.01], t(30.99), = 2.10, p < 0.05). At the end of the study (object presentations 21–30), the linkage between interest scores and these other behaviors was no longer significant for any behavior (all p's > 0.1).

Across the study, zebrafish readily approached novel objects introduced into their home-tanks with a median approach latency of 1 s. For comparison, a zoo study investigating curiosity across several taxa (sheep, tortoises, penguins, kangaroos, cockatoos, and lemurs) reported that latency to orient to novel objects ranged from 25 s to over 15 min, with many taxa never coming within two body-lengths of some objects (10).

At the beginning of the present study, i.e., object presentations 1–10, we found evidence for sustained information-seeking: differential, prolonged (up to 10 min), interest consistently displayed toward some objects and not toward others. During this period, we found that interest varied systematically within tank such that objects that attracted high attention during the first 100 s, typically continued to attract above-average attention nearly 10 min later. Further, object ID accounted for 11% of the variability in interest across tanks, indicating that there was some object property driving the zebrafish interest behavior. This type of differential, object-driven sustained attention is more consistent with information seeking than it is with sensation-seeking or indiscriminate neophilia.

None of the dimensions coded by human observers— complexity, ecological plausibility, predator-resemblance—predicted differential zebrafish behavioral responses to the objects. One possible interpretation of this null result is that while we classified objects based on their visual characteristics from a human perspective, the zebrafishes' interest may have been determined by multiple sensory features as perceived by them, including, for example, chemosensory properties and fluid dynamics [zebrafish are sensitive to water flow; (45)]. Thus, it is possible that the theories about what predicts greater curiosity may have been borne out from the fishes' perspective: the objects with the most complex and most unusual features in multi-sensory space in zebrafish perception could have attracted the most attention from the zebrafish. For instance, across tanks, a small white shell and a large neon pink artificial ball both attracted a great deal of attention at the beginning of the study, despite being markedly different in human-coded complexity, ecological-plausibility, and predator-resemblance, and different in terms of shape, size, color, and material. In a multisensory space from a zebrafish's perspective, however, perhaps the white shell and pink sponge ball were equivalently unusual or complex.

Determining what parameters drive intrinsically motivated exploratory behavior in zebrafish is an area for future research, but as indicated in the present results, serial presentations of multiple static objects (the design employed for this study) is unlikely to yield the best data. At the end of the study, approaches remained fast, but interest was relatively uniform (variability in interest between object presentations was low), fleeting (interest did not extend throughout even the first 100 s of object presentation), and any variability that did exist was not driven by object characteristics. In other words, by the end of a serial presentation of 30 novel static objects (by object presentations 21–30), there was little evidence for information-seeking, instead the pattern of behavior was more consistent with sensation-seeking.

Importantly, objects were presented in a pseudo-randomized order, ensuring that across tanks, the objects during the early period were the same as the objects at the end of the study. As such, the drop-off in information seeking cannot be due to some object-level feature. Instead, the loss of interest could be due to some form of habituation to the task. At the beginning of the study, the task itself contained new information as the zebrafish learned what to expect from the procedures. By the end of the study, there was less information to be gained about the situation itself, with the only unknown aspects being those associated with the stationary, nonmanipulable, static objects themselves. It is possible, therefore, that at the beginning of the study, one source of the zebrafishes' curiosity was figuring out what to expect with the task, including what the object itself might do or not do and what could or could not be done with the object. If so, introducing more dimensions of variability into the study procedures—e.g., using objects or images that move, change, or are interactive—might facilitate continued information-seeking behavior throughout repeated presentations.

In sum, these results show that in addition to neophilia, zebrafish engage in behaviors that are consistent with the motivation for information gain. That such a motivation exists in zebrafish falls into a long tradition of animal behavior research pointing toward the deep evolutionary roots of an intrinsic motivation to explore, learn, and make sense of the world and suggests the possibility that zebrafish, like other species, may find information acquisition rewarding (12, 21–23, 46).

We found evidence that, in addition to being neophilic, zebrafish engage in sustained information-seeking behavior. As such, this study provides justification for future work exploring the extent to which fish might find cognitive stimulation rewarding and the role cognitive engagement cognitive engagement plays in their overall welfare. Further investigation into cognitive enrichment for fish will contribute to the growing literature on positive wellbeing, which has become an important topic of research across disciplines. Within the human literature, positive psychology theories have proved to be generative (62) and in the animal literature, a similar pattern is beginning to emerge (63–66). Extending the existing research exploring whether and how fish experience pain (67–69), the present research contributes evidence that fish also have the capacity for positively valenced experiences and are thus good candidates for future research on positive welfare.

Fish welfare is threatened by multiple forms of human activity—from industrial fishing to farming to scientific research to entertainment. There is an urgent need to understand and protect their wellbeing, yet comparatively little species-specific research investigates fish welfare (32) and fish are generally perceived as less deserving of enrichment than other vertebrates (70). One source of this discrepancy may be the general perception that fish are “primitive” or “lower” vertebrates, a notion that is demonstrably false—as a group, fish are just as evolved as any other extant animal taxa and are cognitively sophisticated (68, 71), with some species of fish even outperforming primates on cognitive tests (72) and passing the mirror-self-recognition test (73, 74). The misclassification of fish as somehow inferior to other vertebrates suggests that research on positive welfare in fishes may also have a special role to play in changing the narrative around what fishes can experience and, therefore, what sort of care and protections they require.

Fishes are also some of the most studied organisms in modern science, but they are rarely studied for their own sake. Much of the research involving fish instead focuses on modeling human biological processes [e.g., gene expression in the brain (75)], testing general theories in behavioral biology [e.g., Optimal Foraging Theory (76)], or determining efficient ways to increase farming and fisheries production. In compliment and in contrast to these anthropocentric forms of fish research, adopting a fish-centric research agenda can facilitate unique basic science contributions (e.g., evolutionary patterns of curiosity across the animal kingdom), ethical insights (fish are capable of positive experiences and may suffer from boredom and other forms of poor welfare in their absence), and practical solutions [e.g., cognitive stimulation may be a valuable source of enrichment for fish; (56)]. Moreover, studying fish in natural or semi-natural conditions can help elucidate the degree to which barren laboratory housing induces abnormal biological states that may reduce the generalizability of data collected.

In conclusion, while we did not find that zebrafish follow the same patterns of curiosity as those found in previously studied (terrestrial) animals, we did find evidence that they find some objects to be more interesting than others. As these exploration opportunities decreased agonistic and diving behavior while also increasing affiliative behavior, it is possible that zebrafish, like many other species, benefit from activities that engage their cognitive abilities and preferences. This research builds on our understanding of the determinants and consequences of curiosity across species and opens new avenues of investigation regarding the role that exploration and learning play in the lives and welfare of fishes.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

All husbandry and experimental procedures were approved by the University of British Columbia's Animal Care Committee, protocol number A14-0119.

All authors listed have made a substantial, direct, and intellectual contribution to the work and approved it for publication.