- Email: [email protected]
Perception of the Müller–Lyer illusion in capuchin monkeys (Cebus apella)
Behavioural Brain Research 182 (2007) 67–72
Research report
Perception of the M¨uller–Lyer illusion in capuchin monkeys (Cebus apella) Elisa Suganuma, Valdir Filgueiras Pessoa, Victoria Monge-Fuentes, Br´aulio Magalh˜aes Castro, Maria Clotilde Henriques Tavares ∗ Department of Physiological Sciences, Institute of Biology, University of Bras´ılia, CEP 70.910-900 Brasilia, DF, Brazil Received 28 November 2006; received in revised form 30 April 2007; accepted 8 May 2007 Available online 13 May 2007
Abstract Visual illusions are formed by differences between the perception of one figure and its real physical characteristics. The M¨uller–Lyer illusion is the best known and most studied geometric illusion, consisting in the subject’s judgment between two parallel lines that have the same size, one flanked with outward-pointing arrowheads, and the other with inward-pointing arrowheads. These arrowheads act as inductors that make the lines to be perceived as having different sizes, inward-pointing stimuli being estimated as longer. This study aimed to investigate the M¨uller–Lyer illusion in capuchin monkeys (Cebus apella), a New World primate not yet investigated for this illusion. For this purpose, stimuli were presented on a touch screen monitor. Ten adult subjects (five females and five males) were used. Before the tests, they were trained to discriminate between two physically different lines with and without arrowheads. The longer lines were always the positive (rewarded) stimuli. Regarding the M¨uller–Lyer Illusion test, all monkeys, unrespective of gender, demonstrated susceptibility to the illusion, by choosing preferentially the line with inward-pointing arrowheads. In order to determine the degree of the illusion, a point of subjective equality test (PSE) was performed. The PSE without arrowheads values were lower than the PSE with arrowheads. Thus, it was demonstrated that capuchin monkeys were susceptible to the M¨uller–Lyer illusion, once the perception of the lines’ size was influenced by the presence of the arrowheads and by their orientation. © 2007 Elsevier B.V. All rights reserved. Keywords: M¨uller–Lyer; Visual illusion; Capuchin monkeys; Cebus apella; Perception
1. Introduction Visual illusions occur when the individual perceives a figure as having characteristics different to its own real physical properties [20]. Geometrical optical illusions belong to one class of visual illusions where figures present context-induced subjective distortions of their visual features, such as length, orientation, or curvature of lines. One of the best-known and most extensively investigated geometrical illusion is the M¨uller–Lyer configuration [6], in which two identical straight lines appear perceptually different in length. In this case, one of the lines is flanked by inward-pointing arrowheads, always appearing as being longer, while the other is flanked by outward-pointing arrowheads, always appearing shorter [12].
∗
Corresponding author. Tel.: +55 61 3307 2175; fax: +55 61 3274 1251. E-mail addresses: [email protected], [email protected] (M.C.H. Tavares). 0166-4328/$ – see front matter © 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.bbr.2007.05.014
Many different physiological and cognitive theories have attempted to explain the M¨uller–Lyer illusion, but none of them has been fully satisfactory. It is possible that a complete explanation of this illusion involves a combination of several theories, which explain multiple mechanisms that are involved in many of the illusions. Also, small modifications to the figure enhance some of the mechanisms while inhibit others. Despite of this, the most cited cognitive theory that explains the M¨uller–Lyer illusion is Gregory’s classical perspective theory, in which the linear perspective perceived by the observer, either consciously or unconsciously, elicits “inappropriate” compensation based on the size constancy leading to a perceptual magnification of the figures that appear farther. This occurs because the visual system tries to adjust the small retinal image to maintain size constancy [10]. The similarity between humans and non-human primates’ visual system allows the use of primates as models for human illusory perception. Therefore, studies performed with non-human primates have permitted a better comprehension
68
E. Suganuma et al. / Behavioural Brain Research 182 (2007) 67–72
of the neural mechanisms involved in illusory perception [11,14]. In line with that, some non-human primates have shown to be susceptible to geometric illusions. For instance, rhesus monkeys (Macaca mulatta) [2,8,9] and chimpanzee (Pan troglodytes) [9] perceived the Ponzo illusion; chimpanzees were also susceptible to the Kanizsa illusion [7]; olive baboon (Papio anubis) experienced the Z¨oellner illusion [3]; baboons (Papio papio) were susceptible to the corridor illusion [1]; and rhesus monkeys (Macaca mulatta), sooty mangabey (Cercocebus fuliginosus), and capuchin monkeys (Cebus capucinus) perceived the vertical–horizontal illusion [5]. However, as far as we know, the M¨uller–Lyer illusion has not yet been studied in non-human primates, pigeons standing for the only non-human animal in which this illusion has been tested [15,20]. Capuchin monkeys were chosen as subjects for this study for several reasons: (1) visual acuity and scotopic sensitivity are similar to humans’ [4]; (2) their ability for solving problems using an abstract rule [19]; (3) high encephalic coefficient, indicating their developed cognitive capacity [13]. Thus, this study aimed to investigate the visual perception of the M¨uller–Lyer illusion in capuchin monkeys (Cebus apella), a primate for which there is no literature reference on this type of research.
Fig. 1. Sequence of experimental stages. Stimuli consisted of black graphic patterns on a white background with a 180◦ orientations and 1.0 mm thickness. The viewing distance was about 200 mm.
2. General method
2.4. Data analysis
2.1. Subjects
Statistical analysis was run using Statistical Package for Social Sciences (SPSS) 13.0 Software. The Kolmogorov–Smirnov test was applied to determine the kind of test (parametric or non-parametric) to use. In Section 3.5, the Wilcoxon test was used to compare the number of trials done to acquire the learning criterion among each phase. In addition, to compare the performance between genders in all training phases, the Student t-test (independent simple t-test) was applied. The same test was carried on to verify the animals’ performance due to the previous apparatus experience in phase 1. For phases 2–4, the Mann–Whitney test was used. In Section 4.1, the Wilcoxon test was applied to analyze the number of times that the line with inward and outward-pointing arrowheads was chosen. The same test was run to compare the animals’ performance in the 10 initial warm-up trials and the 10 constancy verification trials. To compare the performance between genders, the Mann–Whitney test was conducted. Furthermore, the Student t-test (paired simple t-test) was applied to compare the mean number of correct responses in the first 10 test trials with the last 10 test trials and in the first 20 test trials with the last 20 test trials. In Section 5.3, the Student t-test (paired simple t-test) was conducted when the group was analyzed as a whole in the PSE test with and without arrowheads. Finally, the Student t-test (independent simple t-test) compared the performance between females and males.
Ten adult (6–10 years old) capuchin monkeys (C. apella), five females and five males, were used as subjects in this study. All subjects had no experience in illusory tests, although four females and three males had experience in tests involving the use of computer apparatus. They were kept in cages (4 m long, 2.9 m wide, and 2 m high) surrounded by natural vegetation and were maintained in couples or in groups of three at the Primate Center of the University of Brasilia, Brazil. The animals were tested in their own cages and were only separated from the rest of the group during the experimental session. They were not food deprived and water was available ad libitum except during the sessions. All procedures were approved by the University of Brasilia Animal Care and Use Committee.
2.2. Apparatus and computer program A laptop (Acer, TravelMate 521TE) connected to a 15 in. touch-screen monitor (LG Studio Works 440, Microtouch) was used to collect all data. A food pellet dispenser (Med Associates ENV-203) was not controlled by the computer software, but operated manually by one of the experimenters. The experimental apparatus was set up inside a portable wooden cart set in a way that the monitor faced the cage. A computer program ILU (using Delphi language and Windows XP compatible), developed to conduct the experiments, permitted the display of illusory figures, manipulation of the stimuli’s exposure time, intervals between stimuli, number of trials, and screen’s background color. Data such as subject, session, trial, phase of the experiment, and animal response accuracy were recorded by the software as well as manually by the experimenter.
2.3. General procedure and stimuli This study was divided into three stages: training (phases 1–4), the M¨uller–Lyer illusion test (MLT), and determination of the point of subjective equality (PSE) with and without arrowheads (Fig. 1).
3. Training Training was composed of four phases, each one lasting until the subjects reached the criterion of 90% correct answers in two consecutive sessions, composed of 45 trials each. The intertrial intervals were 2000 ms. Monkeys were always rewarded with food pellet when they accomplished the phase’s objective. 3.1. Phase 1: touching on the stimulus The stimuli were single horizontal lines (60, 70, and 80 mm) not flanked by arrowheads. The line was located either on the
E. Suganuma et al. / Behavioural Brain Research 182 (2007) 67–72
69
Fig. 2. Stimuli used throughout the experiment: (a) training phase 1; (b) training phase 2 and PSE without arrowheads; (c) and (d) training phase 3, with outwardpointing arrowheads and inward-pointing arrowheads, respectively; (e) and (f) training phase 4, inward-pointing arrowheads flanking the longest line and outwardpointing arrowheads flanking the longest line, respectively; (g) MLT; (h) and (i) PSE with arrowheads, outward-pointing arrowheads flanking the longest line and inward-pointing arrowheads flanking the longest line, respectively.
upper or lower half of the screen (Fig. 2a). Due to the fact that three of the subjects were na¨ıve in the use of computer screens, all animals were initially rewarded when they simply touched the screen. After this was accomplished, the animal was only rewarded when it touched properly on the stimulus. In the following three phases stimuli consisted of two lines, with or without arrowheads, physically different in size. The animals’ task was to identify and touch only on the line perceived to be the longest, in which case, monkeys were reinforced. To prevent association of the positive stimulus with position, the location of the longest line was presented half of the time in the upper position and half in the lower position. The distance (50 mm) between the stimuli was kept the same during all phases of the experiment. Throughout training phase 1, the animals learned to touch the screen using one of its fingers. 3.2. Phase 2: discrimination between two physically different lines without arrowheads Four pairs of lines (70/20, 70/30, 80/20 and 80/30 mm) without arrowheads in its extremities were randomly presented (Fig. 2b). Learning discrimination was laborious and took several months. 3.3. Phase 3: discrimination between two physically different lines, both with arrowheads pointing inward or outward Eight pairs of horizontal lines (75/15, 75/20, 80/20, 80/25, 85/45, 85/48, 90/48, and 90/51 mm) with arrowheads were
randomly shown. Each pair of lines was flanked with both arrowheads pointing inward (> <) or outward (< >). The arrowheads consisted of two fins, each 25 mm long and formed a 135◦ angle with the horizontal line in the inward-pointing arrowhead condition and a 45◦ angle in the outward-pointing arrowhead condition (Fig. 2c and d). 3.4. Phase 4: discrimination between two physically different lines, one with arrowheads pointing inward and the other pointing outward Eight pairs of horizontal lines (85/45, 85/48, 90/45, 90/48, 100/15, 100/18, 105/15, and 105/18 mm) with arrowheads were randomly presented. The characteristics for the arrowheads were the same as for phase 3 (Fig. 2e and f). The objective for both phases 3 and 4 was to discriminate the longest line independent of the arrowheads’ orientation. 3.5. Results Fig. 3 shows the general results obtained in C. apella for the number of trials needed in order to reach the learning criterion in the four training phases. In phase 2, animals needed a greater number of trials due to the learning of a new rule (choosing the longest line), but there was no significant difference between phases 3 and 4 (Z = −2.103, p = 0.210) demonstrating the generalization of the rule learned in phase 2. Similarly, in all training phases no significant difference was found between genders (phase 1: U = 8, p = 0.347; phase 2: t = 1.200, d.f. = 9, p = 0.264;
70
E. Suganuma et al. / Behavioural Brain Research 182 (2007) 67–72
Fig. 4. Percentage mean of C. apella choice (n = 10) for the stimulus with outward and inward-pointing arrowheads in the M¨uller–Lyer illusion test. Fig. 3. Number of trials to criterium for the Cebus apella group (n = 10) in the four training phases. The criterium was a minimum of 90% of correct responses. Rectangles indicate the interval between quartiles containing 50% of the number of trials. Bars represent the highest and lowest values, and the thick horizontal lines represent the median values. Extreme values were not represented in figure. *p ≤ 0.03 in the comparisons for phases 1 and 2 with all the phases.
phase 3: t = −0.159, d.f. = 9, p = 0.877; phase 4: t = −1.857, d.f. = 9, p = 0.100). The experience of animals with computer apparatus was also compared. Significant difference was found only for phase 1 (t = 2.403, d.f. = 9, p = 0.043), in which the number of trials for the non-experient animals (1184.85 ± 217.89) was greater compared to the experient ones (559.35 ± 142.605), indicating that previous experience did not influence the animals’ performance after phase 1 rule was learned.
¨ 4. Muller–Lyer illusion test (MLT) Consisted on the discrimination between two physically equal lines, one with arrowheads pointing inward and the other with arrowheads pointing outward. Each session (one per animal) was composed by 10 initial warm-up trials, followed by 40 test trials and then 10 constancy verification trials (probe). The warm-up and constancy verification stimuli were the same as the ones used in training phase 4 and had the objective to verify if the subjects were discriminating the longest line by itself or if the choice was influenced by the arrowheads’ orientation. Stimuli for the 40 test trials were pairs of lines, both with the same lengths (72, 76, 80, 84 or 88 mm), with arrowheads and randomly presented. The size of each fin was equivalent to 30% of the horizontal line size. The fin’s angle and the distance between the stimuli was the same as for training phases 3 and 4. The rewarded stimulus was the illusory longest one (inward-pointing arrowhead) because both lines had the same physical size, but different arrowheads’ orientation (Fig. 2g).
4.1. Results Fig. 4 shows the general result for the C. apella for the MLT. The number of times that stimuli consisting of lines with inward-pointing arrowheads was chosen (31.8 ± 0.952), was significantly greater (Z = −2.820, p = 0.005) when compared to the lines with outward-pointing arrowheads (8.2 ± 0.952). This indicates that the animals judged longer the line with inwardpointing arrowheads than the line with the outward-pointing arrowheads, demonstrating that they were susceptible to the illusion. No significant difference was found for females and males’ performance (U = 3.500, p = 0.053). Furthermore, there was no significant difference (Z = −0.368, p = 0.713) between the 10 initial warm-up and the 10 final constancy verification trials. Additionally, no significant difference was observed between the mean number of correct response in the first 10 test trials with the last 10 test trials (t = 0.264, d.f. = 9, p = 0.798) neither in the first 20 test trials with the last 20 test trials (t = −0.952, d.f. = 9, p = 0.366). This indicates that throughout the test the animals maintained the same rule, suggesting that there was no learning effect. 5. Point of subjective equality (PSE) Refers to the estimation of the point where the stimuli were considered equally long for stimuli with and without arrowheads in order to determine the degree of the illusion. 5.1. PSE without arrowheads Done after completion of training phase 2. Each session was composed of 45 trials. With the purpose of reminding the subject about the rule of choosing the longest line, the first 5 trials contained the same stimuli used in training phase 2. Test stimuli had the same characteristics as training phase 2 stimuli except that one of the lines always had a fixed size length (80 mm) and the other varied in 8–50% difference along 2% intervals in
E. Suganuma et al. / Behavioural Brain Research 182 (2007) 67–72
71
Table 1 Percentage and centimeter values for PSE without arrowheads and PSE with arrowheads, and difference for these values for Cebus apella (n = 10) PSE without arrowheads
PSE with arrowheads
PSE with arrowheads minus PSE without arrowheads
Difference (%)
Difference (cm)
Difference (%)
Difference (cm)
Difference (%)
Difference (cm)
Female 1 Female 2 Female 3 Female 4 Female 5 Male 1 Male 2 Male 3 Male 4
8 16 26 6 18 6 18 18 12
0.64 1.28 2.08 0.48 1.44 0.48 1.44 1.44 0.96
40 38 38 36 30 46 38 38 40
3.20 3.04 3.04 3.04 2.40 3.68 3.04 3.04 3.20
32 22 12 30 12 40 20 20 28
2.56 1.76 0.96 2.40 0.96 3.20 1.60 1.60 2.24
Average
14
1.12
38
3.08
24
1.92
Results are presented only for four males once one died before the PSE with arrowheads.
relation to the fixed size stimulus (Fig. 2b). The subject always started with the stimulus with the 50% difference, which was gradually reduced in 10% intervals according to the subject’s performance. The animal had a maximum of three consecutive sessions to reach the criterion in each interval. Final PSE was determined by decreasing stimuli’s size to the last 10% interval from the point of discrimination limit and afterwards reducing it in 2% intervals. The percentage of correct responses in each session was calculated and the criterion was determined by the binomial test constructed for 95% confidence limits around chance performance based on the number of test trials. Thus, for 40 trials the upper limit was calculated as 67%. The performance of all subjects was compared to these confidence limits and any performance above the upper limit was considered significant (p < 0.05). 5.2. PSE with arrowheads Done after MLT. The procedure was the same used for the phase with stimuli without arrowheads, but with three differences. First, the stimuli consisted of two lines, one with arrowheads pointing outward and the other with arrowheads pointing inward, the longest line being the former (2h) or the later (2i). Second, the first five trials contained the same stimuli used in training phase 4 to assure that the same rule was being followed. Third, the animal had a maximum of six consecutive sessions to reach the criterion in each interval due to the fact that in each session 50% of the stimuli were illusory (the rewarded stimulus was the line physically longest with outward-pointing arrowheads) and the other 50% were control (the rewarded stimulus was the line physically longest with inward-pointing arrowheads). However, only the illusory stimuli were considered in the analysis. 5.3. Results Individual results for PSE are shown in Table 1. Despite individual differences, when the group was analyzed as a whole, the animals’ performance was significantly better for the PSE without arrowheads than for the PSE with arrowheads (t = −8.661, d.f. = 8, p < 0.001). No significant differences
were found between females and males for the PSE without arrowheads (t = 0.270, d.f. = 8, p = 0.786) nor for the PSE with arrowheads (t = −1.600, d.f. = 8, p = 0.155). These results demonstrate that the arrowheads’ orientation influenced the perception of the lines’ length. 6. Discussion All animals chose mainly the stimuli that presented inwardpointing arrowheads, showing their susceptibility to the illusion. No gender differences were verified. An important aspect in this test was the similarity between the results in the initial 10 warmup trials and the final 10 constancy verification trials, showing that the rule learned in the training stage remained unaltered. Additionally, to be sure that there was no learning effect in the M¨uller–Lyer test, we compared the response accuracy considering the first 10 and last 10 test trials, and also the first 20 and last 20 test trials. In both cases, there was no significative statistical difference. These results strongly suggest that capuchin monkeys were susceptible to the M¨uller–Lyer illusion. A cleaner and alternative procedure to avoid the learning effect would have been to include in each session a few number of randomly reinforced test trials intermixed by chance with a higher number of training rewarded trials. This could exclude the possibility of inducing a response bias in the test trials due to a learning effect. Future studies may take this into account. Stuart et al. [18] using the same forced choice procedure for same size stimuli, but with arrowheads pointing in opposite orientations, also verified the illusory effect in humans. However, in their experiments, human subjects were verbally instructed to look at the figure as a whole and not to concentrate their attention in the arrowheads’ orientation. Our protocol included training phases 3 and 4 to counterbalance the impossibility of verbal instructions to the monkeys. Another study that adopted the same forced choice procedure was done by Warden and Baar [20]. They demonstrated pigeons’ susceptibility to the illusion. An important difference in this procedure was that both stimuli were elegantly presented side by side, avoiding direct comparison between the lines’ length. It is a difficult task to make any kind of comparisons for humans, pigeons, and non-human primates due to the fact that
72
E. Suganuma et al. / Behavioural Brain Research 182 (2007) 67–72
each study used different procedures and methodologies. However, any possible interspecific difference in the visual and perceptual systems could not be excluded. The values for the PSE with and without arrowheads varied depending on the animal tested. Nevertheless, when the group was analyzed for each condition, no significant differences appeared. When comparing the data for PSE with and without arrowheads, the results demonstrated significant difference indicating that the arrowheads influenced the judgment for the lines’ length. Previously mentioned methodological differences preclude comparisons between the results obtained in this study with results from other investigations. Cognitive and physiological theories formulated to explain illusory effects in humans are not entirely satisfactory for the majority of the illusions. A possible explanation involves a combination of multiple mechanisms. In spite of that, non-human primate studies have allowed a great understanding of the putative neural mechanisms related to the illusions. One positive example is the investigation of contour perception in rhesus monkeys, which provided fundamentals for physiological explanation of the Kanizsa illusion [11,14]. Although the physiological mechanisms for illusion perception are not still understandable [7], the demonstration of illusory effect in non-human primates rises ideas about the origin of this phenomena from an evolutionary perspective. 7. Conclusions The results obtained in this study apparently indicated the presence of the illusory effect in capuchin monkeys (C. apella) for the M¨uller–Lyer illusion. Strictly speaking, it is impossible to assume that the monkeys are actually illuded with the size of the lines when the M¨uller–Lyer illusion test was performed. However, as a whole, the animals were able to present a homogenous accurated performance in the sense of illusion. Our study is the first to investigate the M¨uller–Lyer illusion in C. apella. As observed by Dominguez [5], capuchin monkeys seem to be susceptible to the illusory effect. The indication that Cebus can share with humans similar perceptual mechanisms entails to further comparative studies, in which other parameters that can influence the illusion can be investigated. It would be important, for example, to know if characteristics such as the angle variation of arrowheads [17], orientation of the lines [16], and contrast between lines and arrowheads [6] can also affect the monkeys’ perception as it occur in humans. Although it cannot be possible to comprehend the extension for which the illusions can be important to the monkeys’ interaction with their environment, our results support the existence of similarities between the visual perception in humans and non-human primates, taking C. apella as an example.
Acknowledgments This study was funded by CAPES (graduate fellowship to E. S.) and FINATEC. Our thanks go to Ricardo Sussumu Chiba for the development of the computer software and to Claudia Yrlanda Simon for her contribution in the initial steps of this study. We also thanks to Raimundo Oliveira da Silva and Geinaldo Vieira da Silva for their technical assistance and animal care. Our special thanks goes to Carlos Enrique Uribe for the statistical processing of the results obtained. We are also grateful to the two anonymous referees for their useful comments that contributed to improve the quality of the manuscript. References [1] Barbet I, Fagot J. Perception of the corridor illusion by baboons (Papio papio). Behav Brain Res 2002;132:111–5. [2] Bayne KAL, Davis RT. Susceptibility of rhesus monkey (Macaca mulatta) to the Ponzo illusion. Bull Psychon Soc 1983;21:476–8. [3] Benhar E, Samuel D. Visual illusions in the baboons (Papio anubis). Anim Learn Behav 1982;10:115–8. [4] De Valois RL. Vision. In: Schrier AM, Stollnitz F, editors. Behavior of nonhuman primates. Modern research trends, vol. 3. New York: Academic Press; 1971. p. 107–57. [5] Dominguez KE. A study of visual illusion in the monkey. J Genet Psychol 1954;85:105–27. [6] Dragoi V, Lockhead G. Context-dependent changes in visual sensitivity induced by M¨uller–Lyer stimuli. Vision Res 1999;39:1657–70. [7] Fagot J, Tomonaga M. Effects of element separation on perceptual gruping by humans (Homo sapiens) and chipanzees (Pan troglodytes): perception of Kanizsa illusory figures. Anim Cogn 2001;4:171–7. [8] Fujita K. Linear perspective and the Ponzo illusion: a comparison between rhesus monkeys and humans. Jpn Psychol Res 1996;38:136–45. [9] Fujita K. Perception of Ponzo illusion by rhesus monkey, chimpanzees, and humans: similarity and difference in the three primate species. Percept Psychophys 1997;59:284–92. [10] Gregory RL. Eye and brain: the psychology of seeing. New York: McGrawHill; 1974. p. 255. [11] Heydt R, Petergans E. Mechanisms of contours perception in monkey visual cortex. I. Lines of pattern discontinuity. J Neurosci 1989;9:1731–48. [12] Howe CQ, Purves D. The M¨uller–Lyer illusion explained by the statistics of image–source relationships. PNAS 2005;102:1234–9. [13] Jerison HJ. Evolution of the brain and intelligence. New York: Academic Press; 1973. p. 482. [14] Lam DMK, Gilbert CD. Neural mechanisms of visual perception. Portfolia: Woodlands; 1989. p. 304. [15] Malott RW, Malott MK, Pokrzywinski J. The effects of outwardpointing arrowheads on the M¨uller–Lyer illusion in pigeons. Psychon Sci 1967;9:55–6. [16] Miller D. The wisdom of the eye. Academic press; 2000. p. 231. [17] Pressey A, Martin NS. The effects of varying fins in M¨uller–Lyer illusion and Holding illusion. Psychol Res 1990;52:46–53. [18] Stuart GW, Day RH, Dickinson RG. M¨uller–Lyer: illusion of size or position? Quart J Exp Psychol 1984;36:663–72. [19] Tavares MCH, Tomaz C. Working memory in capuchin monkeys (Cebus apella). Behav Brain Res 2002;131:131–7. [20] Warden CJ, Baar J. The M¨uller–Lyer illusion in the ring dove, Turtur risorius. J Comp Psychol 1929;9:275–92.
Research report
Perception of the M¨uller–Lyer illusion in capuchin monkeys (Cebus apella) Elisa Suganuma, Valdir Filgueiras Pessoa, Victoria Monge-Fuentes, Br´aulio Magalh˜aes Castro, Maria Clotilde Henriques Tavares ∗ Department of Physiological Sciences, Institute of Biology, University of Bras´ılia, CEP 70.910-900 Brasilia, DF, Brazil Received 28 November 2006; received in revised form 30 April 2007; accepted 8 May 2007 Available online 13 May 2007
Abstract Visual illusions are formed by differences between the perception of one figure and its real physical characteristics. The M¨uller–Lyer illusion is the best known and most studied geometric illusion, consisting in the subject’s judgment between two parallel lines that have the same size, one flanked with outward-pointing arrowheads, and the other with inward-pointing arrowheads. These arrowheads act as inductors that make the lines to be perceived as having different sizes, inward-pointing stimuli being estimated as longer. This study aimed to investigate the M¨uller–Lyer illusion in capuchin monkeys (Cebus apella), a New World primate not yet investigated for this illusion. For this purpose, stimuli were presented on a touch screen monitor. Ten adult subjects (five females and five males) were used. Before the tests, they were trained to discriminate between two physically different lines with and without arrowheads. The longer lines were always the positive (rewarded) stimuli. Regarding the M¨uller–Lyer Illusion test, all monkeys, unrespective of gender, demonstrated susceptibility to the illusion, by choosing preferentially the line with inward-pointing arrowheads. In order to determine the degree of the illusion, a point of subjective equality test (PSE) was performed. The PSE without arrowheads values were lower than the PSE with arrowheads. Thus, it was demonstrated that capuchin monkeys were susceptible to the M¨uller–Lyer illusion, once the perception of the lines’ size was influenced by the presence of the arrowheads and by their orientation. © 2007 Elsevier B.V. All rights reserved. Keywords: M¨uller–Lyer; Visual illusion; Capuchin monkeys; Cebus apella; Perception
1. Introduction Visual illusions occur when the individual perceives a figure as having characteristics different to its own real physical properties [20]. Geometrical optical illusions belong to one class of visual illusions where figures present context-induced subjective distortions of their visual features, such as length, orientation, or curvature of lines. One of the best-known and most extensively investigated geometrical illusion is the M¨uller–Lyer configuration [6], in which two identical straight lines appear perceptually different in length. In this case, one of the lines is flanked by inward-pointing arrowheads, always appearing as being longer, while the other is flanked by outward-pointing arrowheads, always appearing shorter [12].
∗
Corresponding author. Tel.: +55 61 3307 2175; fax: +55 61 3274 1251. E-mail addresses: [email protected], [email protected] (M.C.H. Tavares). 0166-4328/$ – see front matter © 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.bbr.2007.05.014
Many different physiological and cognitive theories have attempted to explain the M¨uller–Lyer illusion, but none of them has been fully satisfactory. It is possible that a complete explanation of this illusion involves a combination of several theories, which explain multiple mechanisms that are involved in many of the illusions. Also, small modifications to the figure enhance some of the mechanisms while inhibit others. Despite of this, the most cited cognitive theory that explains the M¨uller–Lyer illusion is Gregory’s classical perspective theory, in which the linear perspective perceived by the observer, either consciously or unconsciously, elicits “inappropriate” compensation based on the size constancy leading to a perceptual magnification of the figures that appear farther. This occurs because the visual system tries to adjust the small retinal image to maintain size constancy [10]. The similarity between humans and non-human primates’ visual system allows the use of primates as models for human illusory perception. Therefore, studies performed with non-human primates have permitted a better comprehension
68
E. Suganuma et al. / Behavioural Brain Research 182 (2007) 67–72
of the neural mechanisms involved in illusory perception [11,14]. In line with that, some non-human primates have shown to be susceptible to geometric illusions. For instance, rhesus monkeys (Macaca mulatta) [2,8,9] and chimpanzee (Pan troglodytes) [9] perceived the Ponzo illusion; chimpanzees were also susceptible to the Kanizsa illusion [7]; olive baboon (Papio anubis) experienced the Z¨oellner illusion [3]; baboons (Papio papio) were susceptible to the corridor illusion [1]; and rhesus monkeys (Macaca mulatta), sooty mangabey (Cercocebus fuliginosus), and capuchin monkeys (Cebus capucinus) perceived the vertical–horizontal illusion [5]. However, as far as we know, the M¨uller–Lyer illusion has not yet been studied in non-human primates, pigeons standing for the only non-human animal in which this illusion has been tested [15,20]. Capuchin monkeys were chosen as subjects for this study for several reasons: (1) visual acuity and scotopic sensitivity are similar to humans’ [4]; (2) their ability for solving problems using an abstract rule [19]; (3) high encephalic coefficient, indicating their developed cognitive capacity [13]. Thus, this study aimed to investigate the visual perception of the M¨uller–Lyer illusion in capuchin monkeys (Cebus apella), a primate for which there is no literature reference on this type of research.
Fig. 1. Sequence of experimental stages. Stimuli consisted of black graphic patterns on a white background with a 180◦ orientations and 1.0 mm thickness. The viewing distance was about 200 mm.
2. General method
2.4. Data analysis
2.1. Subjects
Statistical analysis was run using Statistical Package for Social Sciences (SPSS) 13.0 Software. The Kolmogorov–Smirnov test was applied to determine the kind of test (parametric or non-parametric) to use. In Section 3.5, the Wilcoxon test was used to compare the number of trials done to acquire the learning criterion among each phase. In addition, to compare the performance between genders in all training phases, the Student t-test (independent simple t-test) was applied. The same test was carried on to verify the animals’ performance due to the previous apparatus experience in phase 1. For phases 2–4, the Mann–Whitney test was used. In Section 4.1, the Wilcoxon test was applied to analyze the number of times that the line with inward and outward-pointing arrowheads was chosen. The same test was run to compare the animals’ performance in the 10 initial warm-up trials and the 10 constancy verification trials. To compare the performance between genders, the Mann–Whitney test was conducted. Furthermore, the Student t-test (paired simple t-test) was applied to compare the mean number of correct responses in the first 10 test trials with the last 10 test trials and in the first 20 test trials with the last 20 test trials. In Section 5.3, the Student t-test (paired simple t-test) was conducted when the group was analyzed as a whole in the PSE test with and without arrowheads. Finally, the Student t-test (independent simple t-test) compared the performance between females and males.
Ten adult (6–10 years old) capuchin monkeys (C. apella), five females and five males, were used as subjects in this study. All subjects had no experience in illusory tests, although four females and three males had experience in tests involving the use of computer apparatus. They were kept in cages (4 m long, 2.9 m wide, and 2 m high) surrounded by natural vegetation and were maintained in couples or in groups of three at the Primate Center of the University of Brasilia, Brazil. The animals were tested in their own cages and were only separated from the rest of the group during the experimental session. They were not food deprived and water was available ad libitum except during the sessions. All procedures were approved by the University of Brasilia Animal Care and Use Committee.
2.2. Apparatus and computer program A laptop (Acer, TravelMate 521TE) connected to a 15 in. touch-screen monitor (LG Studio Works 440, Microtouch) was used to collect all data. A food pellet dispenser (Med Associates ENV-203) was not controlled by the computer software, but operated manually by one of the experimenters. The experimental apparatus was set up inside a portable wooden cart set in a way that the monitor faced the cage. A computer program ILU (using Delphi language and Windows XP compatible), developed to conduct the experiments, permitted the display of illusory figures, manipulation of the stimuli’s exposure time, intervals between stimuli, number of trials, and screen’s background color. Data such as subject, session, trial, phase of the experiment, and animal response accuracy were recorded by the software as well as manually by the experimenter.
2.3. General procedure and stimuli This study was divided into three stages: training (phases 1–4), the M¨uller–Lyer illusion test (MLT), and determination of the point of subjective equality (PSE) with and without arrowheads (Fig. 1).
3. Training Training was composed of four phases, each one lasting until the subjects reached the criterion of 90% correct answers in two consecutive sessions, composed of 45 trials each. The intertrial intervals were 2000 ms. Monkeys were always rewarded with food pellet when they accomplished the phase’s objective. 3.1. Phase 1: touching on the stimulus The stimuli were single horizontal lines (60, 70, and 80 mm) not flanked by arrowheads. The line was located either on the
E. Suganuma et al. / Behavioural Brain Research 182 (2007) 67–72
69
Fig. 2. Stimuli used throughout the experiment: (a) training phase 1; (b) training phase 2 and PSE without arrowheads; (c) and (d) training phase 3, with outwardpointing arrowheads and inward-pointing arrowheads, respectively; (e) and (f) training phase 4, inward-pointing arrowheads flanking the longest line and outwardpointing arrowheads flanking the longest line, respectively; (g) MLT; (h) and (i) PSE with arrowheads, outward-pointing arrowheads flanking the longest line and inward-pointing arrowheads flanking the longest line, respectively.
upper or lower half of the screen (Fig. 2a). Due to the fact that three of the subjects were na¨ıve in the use of computer screens, all animals were initially rewarded when they simply touched the screen. After this was accomplished, the animal was only rewarded when it touched properly on the stimulus. In the following three phases stimuli consisted of two lines, with or without arrowheads, physically different in size. The animals’ task was to identify and touch only on the line perceived to be the longest, in which case, monkeys were reinforced. To prevent association of the positive stimulus with position, the location of the longest line was presented half of the time in the upper position and half in the lower position. The distance (50 mm) between the stimuli was kept the same during all phases of the experiment. Throughout training phase 1, the animals learned to touch the screen using one of its fingers. 3.2. Phase 2: discrimination between two physically different lines without arrowheads Four pairs of lines (70/20, 70/30, 80/20 and 80/30 mm) without arrowheads in its extremities were randomly presented (Fig. 2b). Learning discrimination was laborious and took several months. 3.3. Phase 3: discrimination between two physically different lines, both with arrowheads pointing inward or outward Eight pairs of horizontal lines (75/15, 75/20, 80/20, 80/25, 85/45, 85/48, 90/48, and 90/51 mm) with arrowheads were
randomly shown. Each pair of lines was flanked with both arrowheads pointing inward (> <) or outward (< >). The arrowheads consisted of two fins, each 25 mm long and formed a 135◦ angle with the horizontal line in the inward-pointing arrowhead condition and a 45◦ angle in the outward-pointing arrowhead condition (Fig. 2c and d). 3.4. Phase 4: discrimination between two physically different lines, one with arrowheads pointing inward and the other pointing outward Eight pairs of horizontal lines (85/45, 85/48, 90/45, 90/48, 100/15, 100/18, 105/15, and 105/18 mm) with arrowheads were randomly presented. The characteristics for the arrowheads were the same as for phase 3 (Fig. 2e and f). The objective for both phases 3 and 4 was to discriminate the longest line independent of the arrowheads’ orientation. 3.5. Results Fig. 3 shows the general results obtained in C. apella for the number of trials needed in order to reach the learning criterion in the four training phases. In phase 2, animals needed a greater number of trials due to the learning of a new rule (choosing the longest line), but there was no significant difference between phases 3 and 4 (Z = −2.103, p = 0.210) demonstrating the generalization of the rule learned in phase 2. Similarly, in all training phases no significant difference was found between genders (phase 1: U = 8, p = 0.347; phase 2: t = 1.200, d.f. = 9, p = 0.264;
70
E. Suganuma et al. / Behavioural Brain Research 182 (2007) 67–72
Fig. 4. Percentage mean of C. apella choice (n = 10) for the stimulus with outward and inward-pointing arrowheads in the M¨uller–Lyer illusion test. Fig. 3. Number of trials to criterium for the Cebus apella group (n = 10) in the four training phases. The criterium was a minimum of 90% of correct responses. Rectangles indicate the interval between quartiles containing 50% of the number of trials. Bars represent the highest and lowest values, and the thick horizontal lines represent the median values. Extreme values were not represented in figure. *p ≤ 0.03 in the comparisons for phases 1 and 2 with all the phases.
phase 3: t = −0.159, d.f. = 9, p = 0.877; phase 4: t = −1.857, d.f. = 9, p = 0.100). The experience of animals with computer apparatus was also compared. Significant difference was found only for phase 1 (t = 2.403, d.f. = 9, p = 0.043), in which the number of trials for the non-experient animals (1184.85 ± 217.89) was greater compared to the experient ones (559.35 ± 142.605), indicating that previous experience did not influence the animals’ performance after phase 1 rule was learned.
¨ 4. Muller–Lyer illusion test (MLT) Consisted on the discrimination between two physically equal lines, one with arrowheads pointing inward and the other with arrowheads pointing outward. Each session (one per animal) was composed by 10 initial warm-up trials, followed by 40 test trials and then 10 constancy verification trials (probe). The warm-up and constancy verification stimuli were the same as the ones used in training phase 4 and had the objective to verify if the subjects were discriminating the longest line by itself or if the choice was influenced by the arrowheads’ orientation. Stimuli for the 40 test trials were pairs of lines, both with the same lengths (72, 76, 80, 84 or 88 mm), with arrowheads and randomly presented. The size of each fin was equivalent to 30% of the horizontal line size. The fin’s angle and the distance between the stimuli was the same as for training phases 3 and 4. The rewarded stimulus was the illusory longest one (inward-pointing arrowhead) because both lines had the same physical size, but different arrowheads’ orientation (Fig. 2g).
4.1. Results Fig. 4 shows the general result for the C. apella for the MLT. The number of times that stimuli consisting of lines with inward-pointing arrowheads was chosen (31.8 ± 0.952), was significantly greater (Z = −2.820, p = 0.005) when compared to the lines with outward-pointing arrowheads (8.2 ± 0.952). This indicates that the animals judged longer the line with inwardpointing arrowheads than the line with the outward-pointing arrowheads, demonstrating that they were susceptible to the illusion. No significant difference was found for females and males’ performance (U = 3.500, p = 0.053). Furthermore, there was no significant difference (Z = −0.368, p = 0.713) between the 10 initial warm-up and the 10 final constancy verification trials. Additionally, no significant difference was observed between the mean number of correct response in the first 10 test trials with the last 10 test trials (t = 0.264, d.f. = 9, p = 0.798) neither in the first 20 test trials with the last 20 test trials (t = −0.952, d.f. = 9, p = 0.366). This indicates that throughout the test the animals maintained the same rule, suggesting that there was no learning effect. 5. Point of subjective equality (PSE) Refers to the estimation of the point where the stimuli were considered equally long for stimuli with and without arrowheads in order to determine the degree of the illusion. 5.1. PSE without arrowheads Done after completion of training phase 2. Each session was composed of 45 trials. With the purpose of reminding the subject about the rule of choosing the longest line, the first 5 trials contained the same stimuli used in training phase 2. Test stimuli had the same characteristics as training phase 2 stimuli except that one of the lines always had a fixed size length (80 mm) and the other varied in 8–50% difference along 2% intervals in
E. Suganuma et al. / Behavioural Brain Research 182 (2007) 67–72
71
Table 1 Percentage and centimeter values for PSE without arrowheads and PSE with arrowheads, and difference for these values for Cebus apella (n = 10) PSE without arrowheads
PSE with arrowheads
PSE with arrowheads minus PSE without arrowheads
Difference (%)
Difference (cm)
Difference (%)
Difference (cm)
Difference (%)
Difference (cm)
Female 1 Female 2 Female 3 Female 4 Female 5 Male 1 Male 2 Male 3 Male 4
8 16 26 6 18 6 18 18 12
0.64 1.28 2.08 0.48 1.44 0.48 1.44 1.44 0.96
40 38 38 36 30 46 38 38 40
3.20 3.04 3.04 3.04 2.40 3.68 3.04 3.04 3.20
32 22 12 30 12 40 20 20 28
2.56 1.76 0.96 2.40 0.96 3.20 1.60 1.60 2.24
Average
14
1.12
38
3.08
24
1.92
Results are presented only for four males once one died before the PSE with arrowheads.
relation to the fixed size stimulus (Fig. 2b). The subject always started with the stimulus with the 50% difference, which was gradually reduced in 10% intervals according to the subject’s performance. The animal had a maximum of three consecutive sessions to reach the criterion in each interval. Final PSE was determined by decreasing stimuli’s size to the last 10% interval from the point of discrimination limit and afterwards reducing it in 2% intervals. The percentage of correct responses in each session was calculated and the criterion was determined by the binomial test constructed for 95% confidence limits around chance performance based on the number of test trials. Thus, for 40 trials the upper limit was calculated as 67%. The performance of all subjects was compared to these confidence limits and any performance above the upper limit was considered significant (p < 0.05). 5.2. PSE with arrowheads Done after MLT. The procedure was the same used for the phase with stimuli without arrowheads, but with three differences. First, the stimuli consisted of two lines, one with arrowheads pointing outward and the other with arrowheads pointing inward, the longest line being the former (2h) or the later (2i). Second, the first five trials contained the same stimuli used in training phase 4 to assure that the same rule was being followed. Third, the animal had a maximum of six consecutive sessions to reach the criterion in each interval due to the fact that in each session 50% of the stimuli were illusory (the rewarded stimulus was the line physically longest with outward-pointing arrowheads) and the other 50% were control (the rewarded stimulus was the line physically longest with inward-pointing arrowheads). However, only the illusory stimuli were considered in the analysis. 5.3. Results Individual results for PSE are shown in Table 1. Despite individual differences, when the group was analyzed as a whole, the animals’ performance was significantly better for the PSE without arrowheads than for the PSE with arrowheads (t = −8.661, d.f. = 8, p < 0.001). No significant differences
were found between females and males for the PSE without arrowheads (t = 0.270, d.f. = 8, p = 0.786) nor for the PSE with arrowheads (t = −1.600, d.f. = 8, p = 0.155). These results demonstrate that the arrowheads’ orientation influenced the perception of the lines’ length. 6. Discussion All animals chose mainly the stimuli that presented inwardpointing arrowheads, showing their susceptibility to the illusion. No gender differences were verified. An important aspect in this test was the similarity between the results in the initial 10 warmup trials and the final 10 constancy verification trials, showing that the rule learned in the training stage remained unaltered. Additionally, to be sure that there was no learning effect in the M¨uller–Lyer test, we compared the response accuracy considering the first 10 and last 10 test trials, and also the first 20 and last 20 test trials. In both cases, there was no significative statistical difference. These results strongly suggest that capuchin monkeys were susceptible to the M¨uller–Lyer illusion. A cleaner and alternative procedure to avoid the learning effect would have been to include in each session a few number of randomly reinforced test trials intermixed by chance with a higher number of training rewarded trials. This could exclude the possibility of inducing a response bias in the test trials due to a learning effect. Future studies may take this into account. Stuart et al. [18] using the same forced choice procedure for same size stimuli, but with arrowheads pointing in opposite orientations, also verified the illusory effect in humans. However, in their experiments, human subjects were verbally instructed to look at the figure as a whole and not to concentrate their attention in the arrowheads’ orientation. Our protocol included training phases 3 and 4 to counterbalance the impossibility of verbal instructions to the monkeys. Another study that adopted the same forced choice procedure was done by Warden and Baar [20]. They demonstrated pigeons’ susceptibility to the illusion. An important difference in this procedure was that both stimuli were elegantly presented side by side, avoiding direct comparison between the lines’ length. It is a difficult task to make any kind of comparisons for humans, pigeons, and non-human primates due to the fact that
72
E. Suganuma et al. / Behavioural Brain Research 182 (2007) 67–72
each study used different procedures and methodologies. However, any possible interspecific difference in the visual and perceptual systems could not be excluded. The values for the PSE with and without arrowheads varied depending on the animal tested. Nevertheless, when the group was analyzed for each condition, no significant differences appeared. When comparing the data for PSE with and without arrowheads, the results demonstrated significant difference indicating that the arrowheads influenced the judgment for the lines’ length. Previously mentioned methodological differences preclude comparisons between the results obtained in this study with results from other investigations. Cognitive and physiological theories formulated to explain illusory effects in humans are not entirely satisfactory for the majority of the illusions. A possible explanation involves a combination of multiple mechanisms. In spite of that, non-human primate studies have allowed a great understanding of the putative neural mechanisms related to the illusions. One positive example is the investigation of contour perception in rhesus monkeys, which provided fundamentals for physiological explanation of the Kanizsa illusion [11,14]. Although the physiological mechanisms for illusion perception are not still understandable [7], the demonstration of illusory effect in non-human primates rises ideas about the origin of this phenomena from an evolutionary perspective. 7. Conclusions The results obtained in this study apparently indicated the presence of the illusory effect in capuchin monkeys (C. apella) for the M¨uller–Lyer illusion. Strictly speaking, it is impossible to assume that the monkeys are actually illuded with the size of the lines when the M¨uller–Lyer illusion test was performed. However, as a whole, the animals were able to present a homogenous accurated performance in the sense of illusion. Our study is the first to investigate the M¨uller–Lyer illusion in C. apella. As observed by Dominguez [5], capuchin monkeys seem to be susceptible to the illusory effect. The indication that Cebus can share with humans similar perceptual mechanisms entails to further comparative studies, in which other parameters that can influence the illusion can be investigated. It would be important, for example, to know if characteristics such as the angle variation of arrowheads [17], orientation of the lines [16], and contrast between lines and arrowheads [6] can also affect the monkeys’ perception as it occur in humans. Although it cannot be possible to comprehend the extension for which the illusions can be important to the monkeys’ interaction with their environment, our results support the existence of similarities between the visual perception in humans and non-human primates, taking C. apella as an example.
Acknowledgments This study was funded by CAPES (graduate fellowship to E. S.) and FINATEC. Our thanks go to Ricardo Sussumu Chiba for the development of the computer software and to Claudia Yrlanda Simon for her contribution in the initial steps of this study. We also thanks to Raimundo Oliveira da Silva and Geinaldo Vieira da Silva for their technical assistance and animal care. Our special thanks goes to Carlos Enrique Uribe for the statistical processing of the results obtained. We are also grateful to the two anonymous referees for their useful comments that contributed to improve the quality of the manuscript. References [1] Barbet I, Fagot J. Perception of the corridor illusion by baboons (Papio papio). Behav Brain Res 2002;132:111–5. [2] Bayne KAL, Davis RT. Susceptibility of rhesus monkey (Macaca mulatta) to the Ponzo illusion. Bull Psychon Soc 1983;21:476–8. [3] Benhar E, Samuel D. Visual illusions in the baboons (Papio anubis). Anim Learn Behav 1982;10:115–8. [4] De Valois RL. Vision. In: Schrier AM, Stollnitz F, editors. Behavior of nonhuman primates. Modern research trends, vol. 3. New York: Academic Press; 1971. p. 107–57. [5] Dominguez KE. A study of visual illusion in the monkey. J Genet Psychol 1954;85:105–27. [6] Dragoi V, Lockhead G. Context-dependent changes in visual sensitivity induced by M¨uller–Lyer stimuli. Vision Res 1999;39:1657–70. [7] Fagot J, Tomonaga M. Effects of element separation on perceptual gruping by humans (Homo sapiens) and chipanzees (Pan troglodytes): perception of Kanizsa illusory figures. Anim Cogn 2001;4:171–7. [8] Fujita K. Linear perspective and the Ponzo illusion: a comparison between rhesus monkeys and humans. Jpn Psychol Res 1996;38:136–45. [9] Fujita K. Perception of Ponzo illusion by rhesus monkey, chimpanzees, and humans: similarity and difference in the three primate species. Percept Psychophys 1997;59:284–92. [10] Gregory RL. Eye and brain: the psychology of seeing. New York: McGrawHill; 1974. p. 255. [11] Heydt R, Petergans E. Mechanisms of contours perception in monkey visual cortex. I. Lines of pattern discontinuity. J Neurosci 1989;9:1731–48. [12] Howe CQ, Purves D. The M¨uller–Lyer illusion explained by the statistics of image–source relationships. PNAS 2005;102:1234–9. [13] Jerison HJ. Evolution of the brain and intelligence. New York: Academic Press; 1973. p. 482. [14] Lam DMK, Gilbert CD. Neural mechanisms of visual perception. Portfolia: Woodlands; 1989. p. 304. [15] Malott RW, Malott MK, Pokrzywinski J. The effects of outwardpointing arrowheads on the M¨uller–Lyer illusion in pigeons. Psychon Sci 1967;9:55–6. [16] Miller D. The wisdom of the eye. Academic press; 2000. p. 231. [17] Pressey A, Martin NS. The effects of varying fins in M¨uller–Lyer illusion and Holding illusion. Psychol Res 1990;52:46–53. [18] Stuart GW, Day RH, Dickinson RG. M¨uller–Lyer: illusion of size or position? Quart J Exp Psychol 1984;36:663–72. [19] Tavares MCH, Tomaz C. Working memory in capuchin monkeys (Cebus apella). Behav Brain Res 2002;131:131–7. [20] Warden CJ, Baar J. The M¨uller–Lyer illusion in the ring dove, Turtur risorius. J Comp Psychol 1929;9:275–92.