Guppies (Poecilia reticulata) are deceived by visual illusions during obstacle negotiation

Abstract

Animals travelling in their natural environment repeatedly encounter obstacles that they can either detour or go through. Gap negotiation requires an accurate estimation of the opening's size to avoid getting stuck or being injured. Research on visual illusions has revealed that in some circumstances, transformation rules used to generate a three-dimensional representation from bidimensional retinal images fail, leading to systematic errors in perception. In Ebbinghaus and Delboeuf illusions, the presence of task-irrelevant elements causes us to misjudge an object's size. Susceptibility to these illusions was observed in other animals, although with large intraspecific differences. In this study, we investigated whether fish can accurately estimate gap size and whether during this process they may be deceived by illusory patterns. Guppies were extremely accurate in gap negotiation, discriminating holes with a 10% diameter difference. When presented with two identical holes surrounded by inducers to produce Ebbinghaus and Delboeuf patterns, guppies misperceived gap size in the predicted direction. So far, researchers have principally considered illusions as useful tools for exploring the cognitive processing underlying vision. Our findings highlight the possibility that they have important ecological implications, affecting the everyday interactions of an animal with its physical world besides its intra- and interspecific relationships.

1. Introduction

Many species inhabit complex three-dimensional habitats that they need to navigate in order to forage, mate or avoid predators. The straight path is frequently intersected by obstacles that an animal could either circumvent or pass through. Researchers have so far examined opening negotiation in a handful of species [1–3]. Passing through obstacles requires an accurate estimation of the size of the opening. If the gap is too narrow, the animal can be injured, become stuck or be captured by predators. Many insects, birds and bats that fly through dense vegetation need to quickly select among different paths and openings of different size [4,5]. The same likely occurs in terrestrial species such as humans that live in cluttered environments or in aquatic species that inhabit highly vegetated habitats or rocky streams. In these species, selection is expected to favour the evolution of perceptual and cognitive mechanisms for rapid and accurate size discrimination, at the same time buffering the perceptual system against factors that may compromise a correct estimation. However, one century of studies of visual illusions has shown that humans can be induced to see objects that do not physically exist, to perceive motion in static images and to make errors in evaluating luminance, colour, distance or size of an object [6]. This occurs because the visual system uses a set of transformation rules to interpret retinal bidimensional images and translate them into a three-dimensional representation of the external word, and this process can generate systematic errors in perception. Animals are expected to face similar problems in three-dimensional scene representation, and in fact, other species appear susceptible to the same visual illusions; however, owing to differences in the visual system and in the characteristics of habitat, we do not expect that all species perceive visual illusions as humans do [6,7].

Visual illusion can naturally occur in the environment as a consequence of a particular spatial arrangement of the elements in the visual scene or the visual patterns generated by other organisms. Few studies have addressed the ecological implications of these systematic errors. In humans, illusions have been shown to alter the cinematics of grasping or lifting an object and to affect the judgement of the body mass of a conspecific or the size of food portions [8–10]. In other species, research has focused exclusively on the effects of visual patterns generated by an animal on the receiver's perception. For example, camouflage patterns create false marginal boundaries and internal high-contrast edges, thereby disrupting the predator's perception of the animal's true shape or movement [11].

In two well-studied illusions, a particular arrangement of the objects around a target alters the perception of its size. In the Ebbinghaus illusion (figure 1a), an object appears larger to humans when surrounded by smaller objects and smaller when surrounded by larger objects. Dogs, dolphins, pigeons and sharks perceive this illusion but in a reversed way [12–15]; conversely, two teleosts perceive the illusion as humans do [15,16]. In the Delboeuf illusion (figure 1b), a circle surrounded by a slightly larger circle appears larger than an equal-sized circle surrounded by a much larger circle. The Delboeuf illusion has gained popularity in dietary restriction plans, given that humans perceive a food portion as being larger when served on small plates [17]. Primates, cats and one reptile perceive this illusion as humans do [18–20]; conversely, guppies and five other teleost fishes perceive the illusion in a reversed way [21,22].

Figure 1.

In the Ebbinghaus illusion (a) the same object appears larger to humans when surrounded by smaller objects and smaller when surrounded by larger objects. In the Delboeuf illusion (b) humans underestimate the size of a circle enclosed in a much larger circle and overestimate the size of a circle enclosed in a slightly larger circle.

In this study, we investigated guppies in an ecologically relevant context, asking whether a species that inhabits cluttered environments is deceived by the presence of task-irrelevant cues during obstacle negotiation. After assessing in Experiment 1 whether guppies can accurately discriminate the larger of two holes, in Experiment 2 we investigated whether the perception of the relative size of two holes is affected by adding task-irrelevant objects so as to create illusory patterns. We predict that with the Ebbinghaus illusory pattern, guppies would prefer to pass through holes surrounded by smaller inducers and that with the Delboeuf illusory pattern, they would prefer the hole surrounded by a large circle.

2. Methods

(a) . Subjects

We tested 36 female guppies in each experiment. Guppies are a highly social species and may suffer from prolonged isolation. As experiments lasted one or two weeks, we tested guppies in small groups (six groups in each experiment). As in other studies [3,23], each cluster of six fish was considered as one datapoint with no distinction between subjects. A subsample of videos were analysed to check whether individuals contributed equally to the total number of passages. For each video, we identified the position of the six fish in the initial frame, and we analysed the same video six times, following a different subject each time. The analysis revealed that in all experimental groups, the six individuals did not differ in the number of passages (all p-values greater than 0.498; see electronic supplementary material).

(b) . Apparatus

A green plastic hourglass-shaped apparatus was housed in an 80 × 40 cm aquarium filled to a depth of 30 cm (figure 2a). A central corridor (28 × 13 cm) connected two lateral compartments containing natural plants. A removable divider—a black plastic panel (11 × 30 cm) with two holes—subdivided the apparatus into two identical halves.

Figure 2.

(a) Experimental apparatus: a, lateral compartments; b, aquatic plants; c, panel. (b) Panel with experimental stimuli used for size discrimination trials in Experiment 1, namely the depicted panel shows 0.9 diameter ratio (i). Panels used in Experiment 2 to study Ebbinghaus (ii) and Delboeuf (iii) illusions. (c) Results of Experiment 2: boxplots represent median, first quartile, third quartile, ranges and outliers. The x-axis refers to the different types of trials (control ratios and illusory tests). The y-axis refers to the proportion of passages through the bigger hole in control panels and through the hole that is perceived as larger in teleost fishes. Asterisks (*) denote a significant departure from chance level (p < 0.05).

(c) . Acclimatization phase

To accustom subjects to passing through the holes, they were left for 3 days with two large holes differing in diameter (2.7 and 2.1 cm, respectively). During the acclimatization phase of Experiment 2, in order to habituate subjects to the inducers used for the illusions, subjects were exposed to various replicas of the inducers which were glued in random position on four black panels.

(d) . Test phase

Subjects were observed each day for two periods of 3 h separated by a 1 h interval in which no panel was present in the tanks. In Experiment 1, the test phase lasted 6 consecutive days during which we randomly alternated panels presenting two holes with four different diameter ratios. The diameter of one hole was constant (0.9 cm), while the other varied: 1.3 cm (0.69 diameter ratio), 1.2 cm (0.75), 1.1 cm (0.82) and 1.0 cm (0.90) (figure 2b).

In Experiment 2, we presented two different types of panels: control panels and illusory panels. The control panels were the same as in Experiment 1. Intermixed with the control panels, we presented illusory panels with two same-sized holes (0.9 cm in diameter) surrounded by inducers to obtain Delboeuf or Ebbinghaus illusory patterns (figure 2b; see electronic supplementary material, additional methods). To prevent side biases, subjects faced the illusory patterns in opposite left–right positions on the two sides of the panel. The test phase lasted two consecutive weeks. Half the subjects were tested with the Delboeuf illusion in the first week and the Ebbinghaus illusion in the second. We tested the other groups in reverse order.

(e) . Data collection and analysis

All trials were videorecorded and analysed off-line. Analyses were performed in R. Further details on methods and statistical analyses are reported in electronic supplementary material, additional methods and results.

3. Results and discussion

In Experiment 1, guppies passed significantly more often through the bigger hole in all four ratios (table 1), with a significant increase in accuracy with increasing difference between holes (linear trend, F_1,17 = 7.617, p < 0.05). Guppies discriminated two holes that differed by 10% in diameter (corresponding to a 19% difference in area), confirming the good performance previously observed in this task by zebrafish [3,24]. In zebrafish, this capacity is probably innate as it is observed in the first days after hatching, and it is not influenced by experience. The ability to accurately estimate quantities likely confers to fish other important selective advantages, such as estimating the most profitable food item, or the size of an opponent or a prospective mate, and this might explain why fish are comparable to mammals in size discrimination acuity [3].

Table 1.

Group performances of the first experiment (number of passages through the bigger hole) in all the four ratios presented. Asterisks (*) denote a significant departure from chance level (p < 0.05).

group	ratio 0.69	ratio 0.75	ratio 0.82	ratio 0.90
group 1	146/229	164/216	118/188	83/139
	p < 0.001*	p < 0.001*	p < 0.001*	p = 0.027*
group 2	177/256	207/282	101/137	52/71
	p < 0.001*	p < 0.001*	p < 0.001*	p < 0.001*
group 3	159/213	93/132	55/74	87/155
	p < 0.001*	p < 0.001*	p < 0.001*	p = 0.148
group 4	426/657	559/802	435/661	242/430
	p < 0.001*	p < 0.001*	p < 0.001*	p = 0.011*
group 5	177/253	171/231	155/208	98/162
	p < 0.001*	p < 0.001*	p < 0.001*	p = 0.009*
group 6	190/246	175/233	105/141	80/115
	p < 0.001*	p < 0.001*	p < 0.001*	p < 0.001*
mean	0.702	0.737	0.717	0.631
95% CI	[0.650, 0.755]	[0.718, 0.755]	[0.669, 0.766]	[0.561, 0.701]
	p < 0.001	p < 0.001	p < 0.001	p < 0.001

In Experiment 2, guppies also significantly discriminated the four ratios (binomial tests: all p-values < 0.05; figure 2c). The performance did not differ between the two experiments (linear mixed-effects model: experiment, F_1,34 = 1.906, p = 0.175; ratio, F_1,34 = 15.097, p < 0.001; interaction, F_1,34 = 0.744, p = 0.394). In both illusory tests, guppies selected one hole more than chance, suggesting that they were susceptible to size illusions. In the Delboeuf illusory test, guppies significantly preferred the hole surrounded by the larger ring (mean 0.554, 95% CI [0.510, 0.593], p < 0.001; figure 2c). In the Ebbinghaus illusory test, guppies significantly preferred the hole surrounded by the smaller circles (mean 0.562, 95% CI [0.530, 0.594], p < 0.001; figure 2c). The degree of preference did not differ between the two illusions (linear mixed-effects model: F_1,44 = 0.233, p = 0.632; presentation order, F_1,44 = 1.288, p = 0.733; interaction, F_1,44 = 1.516, p = 0.225).

The results of Experiment 2 indicate that in guppies the presence of nearby objects, although irrelevant for the execution of the task, can influence optimal obstacle negotiation. As guppies inhabit a variety of habitats including rocky streams and rivers with aquatic vegetation, they have arguably experienced this situation in their evolutionary past. Similar problems may affect other fishes living in this type of habitat but also aerial and terrestrial organisms that navigate in cluttered environments. Historically, illusions have been considered an important tool for investigating the properties of the visual perceptual system. This study shows for the first time that visual illusions can lead to a suboptimal response to the challenges that animals commonly face in their environment.

Future research should investigate whether the presence of naturally occurring illusory patterns affects the optimal response in other functions that require accurate quantitative estimation, such as in landmark-based long-range navigation, in collective movements, or in assessment of mate quality or resource-holding potential [25]. Indeed, there is some evidence in humans that systematic inaccuracies similar to those observed here may occur in other circumstances. For example, humans perceive a target woman as thinner when surrounded by larger women, and the same food portion as larger if served on a small plate [10,17].

Although we used a novel paradigm that exploits the natural tendency of fish to select the largest available opening to go through an obstacle, our results confirm previous findings that guppies perceive the Ebbinghaus illusion in the same way as humans and, similar to other teleosts, the Delboeuf illusion in a reversed way. The measured response to the two illusions, although significant, is rather small even in relation to the 0.90 ratio. Likely, the illusory effect is very small and close to the perceptual threshold. There is no information on animal species, but in humans the Ebbinghaus illusory effect has been estimated at around 14% area difference in adults and 11% in 5-year-olds, considerably smaller than in the 0.90 ratio (23.5%) [21,26]. On the other hand, the magnitude of response to an illusion may vary with the method used. In another study, the response of guppy to the Delboeuf illusion varied from a 53% preference using a conditioning paradigm to 71% with a spontaneous food choice paradigm [21].

The reasons why guppies and other teleosts perceive the Delboeuf illusion oppositely to humans and other mammals have been thoroughly discussed elsewhere [21]. One popular explanation for the reversed perception of size illusions in some species is that this class of illusions is generally based on two opposite perceptual effects (called assimilation and contrast) and that different species may have different sensitivities to these two effects [14,21,25]. Guppies have shown a striking resemblance to humans in their response to visual illusions. This similarity extends to different categories of illusions, including motion illusions (rotating snakes [27]), contrast illusions (brightness illusion [28]), size illusions (Müller-Lyer, Ebbinghaus, horizontal–vertical illusions [29,30]) and numerosity illusions (solitaire illusion [31]). The exception represented by the Delboeuf illusion could be due to factors not specific to the task; for example, the Delboeuf illusory patterns could be reminiscent of biologically relevant objects, such as the mouth of a large fish, and hence elicit avoidance behaviour in guppies.

Given that estimation inaccuracy likely imposes fitness costs and that vulnerabilities of the perceptual system can be exploited by other organisms [32,33], in the course of evolution, natural selection would be expected to eliminate these perceptual failures. The explanation of this apparent paradox lies in the fact that rules of perception are essential for a correct reconstruction of the external world. For example, the phenomenon of colour constancy enables us to perceive an object as having the same colour under varying illumination conditions. Similarly, amodal completion allows the visual system to perceive objects as a whole, despite partial occlusion by other objects. In most cases, these rules allow a reliable 3D reconstruction of the outside world, but in certain circumstances they lead to perceptual errors. For example, in the brightness illusion, two identical objects appear at a different brightness if placed on different backgrounds, and in the Kanizsa triangle illusion, we perceive an object that does not physically exist [7,24].

Ethics

Experiments were conducted in compliance with the law of the country (Italy) in which they were performed (Decreto legislativo 4 marzo 2014, n. 26). The experimental procedures were approved by the Ethical Committee of Università di Padova (protocol no. 61/2018).

Data accessibility

The data of this study are available from the Dryad Digital Repository [34]: https://doi.org/10.5061/dryad.sbcc2fr7b. A detailed description of the data associated with this study is available in the electronic supplementary material.

Authors' contributions

M.S.: conceptualization, data curation, formal analysis, investigation, methodology, writing—original draft, writing—review and editing; A.B.: conceptualization, project administration, resources, supervision, writing—original draft, writing—review and editing; C.A.: conceptualization, project administration, resources, supervision, writing—original draft, writing—review and editing. All authors gave final approval for publication and agreed to be held accountable for the work performed herein.

Competing interests

We declare we have no competing interests.

Funding

The present work was carried out within the scope of the research program 'Dipartimenti di Eccellenza' (art.1, commi 314–337 legge 232/2016, MIUR grant to the Department of General Psychology, University of Padua) and was supported by the University of Padua under the STARS Grants programme (ANIM_ILLUS) to A.C.