1 Picture Recognition of Food by Sloth Bears (Melursus ursinus) 
2 Tabellario, S., Babitz, M.A., Bauer, E.B., Brown-Palsgrove, M. 
3  
4 Abstract 
5 Pictures are often used in cognitive research to represent objects and many species have demonstrated the ability to 
6 recognize two-dimensional pictures as representations of their three-dimensional counterparts. However, for ursids 
7 picture recognition has been reported in only one study of a single 11-year-old female American black bear 
8 (Johnson-Ulrich et al. 2016).  We tested the picture recognition abilities of an additional species, the sloth bear. 
9 After a food preference test by which the bears’ food options were ranked and categorized as high-, mid-, and low-
10 preference items, we tested a sub-adult male and an adult female sloth bear by presenting two pictures of food in 
11 each testing trial – a high-preference food and a low-preference food. Both bears met criterion by choosing the 
12 pictures of their preferred foods in at least 80% of the trials in three consecutive testing sessions. We then presented 
13 never before used pictures of high-preference versus low-preference food items and they again met our criterion. 
14  
15 Keywords  
16 Picture recognition, sloth bear, Melursus ursinus, choice, control 
17  
18 Introduction 
19 In 2017, Amici pointed out that “studies of animal cognition are not equally distributed across taxa”. Scientists focused 
20 mainly on non-human primates or birds. However in investigations of how animals see the world and process 
21 information, research has demonstrated that several other species have the ability to recognize two-dimensional (2D) 
22 pictures as representations of their three-dimensional (3D) counterparts, including a number of sheep, fish, and lizards, 
23 as reviewed by Bovet and Vauclair (2000), horses (Hanggi 2001), and tortoises (Wilkinson et al. 2013). However only 
24 one bear, an American black bear (Johnson-Ulrich et al. 2016) is noted in the literature to have displayed this ability.  
25 Recognizing 2D pictures can be a starting point for many cognitive tests – discrimination tasks, categorization, and 
26 memory tests.  It is not a given that all animals understand the correspondence between 2D pictures and the real life 
27 objects that they represent. Even some adult humans who have not previously been exposed to photographs are not 
28 able to identify objects in photographs (Miller 1973) and on some occasions chimpanzees have been unable to 
29 recognize 2D photographs (Winner and Ettlinger 1979).  
30 People generally perceive bears as intelligent (Nakajima et al. 2002). In fact, animal trainers and zookeepers regard 
31 bears as the most intelligent carnivore, with intelligence levels comparable to primates (Breiter 2008). However, the 
32 cognitive abilities of bears are understudied (Perdue 2016; Vonk 2016; Vonk and Jett 2018) and there is little scientific 
33 research to back up these claims. Even less is known about sloth bears in particular. Finding evidence that sloth bears 
34 can recognize 2D pictures as representations of their real life counterparts will add to our knowledge base about sloth 
35 bear cognition (Amici et al. 2017; Hartmann et al. 2017). Once we know this cognitive task is possible, other avenues 
36 of investigation open up, allowing the development of a deeper understanding of how the cognitive ability of bear 
37 species compare to one another, and to other carnivores.  
38  If we can learn more about the cognitive abilities of bears, then we will be able to take better care of them in captivity, 
39 managing their mental and emotional well-being in addition to their physical health (McGuire et al. 2017; Perdue 
40 2016; Vonk 2016).  
41 One main source of abnormal stereotypic behavior in captive animals is an animal’s lack of control over their 
42 environment (Leotti et al. 2010). Bears in particular are a species that are prone to stereotypies in captivity (Shih et al. 
43 2016; Vickery and Mason 2004). Giving an animal an opportunity to make choices gives them the perception of 
44 control (Leotti et al. 2010). Control and the choices that are so closely associated with feeling in control in captivity 
45 are critical elements of good animal welfare (Mellor 2014; Reiss 2006). Giving captive animals a way to communicate 
46 their preferences and increasing their opportunities to make choices could help reduce stereotypic behaviors and 
47 improve their well-being, consequently increasing their overall welfare (Buchanan-Smith 2011; Ross 2006). 
48 This project emerged as a novel approach to give Smithsonian National Zoological Park’s (NZP) sloth bears an 
49 opportunity to make choices and communicate their preferences to their keepers about their daily management.  
50  
51 Methods 
52 Subjects 
53 Two captive born sloth bears participated in the study – one 3-year-old, mother-reared male (Niko) and one 5-year-
54 old, hand-reared female (Remi). They both reside at NZP. Although they regularly participate in operant 
55 conditioning tasks for husbandry behaviors, they were both experimentally naive. Neither bear had any history with 
56 pictures of food prior to these experiments. Each bear was tested in his or her regular night house while temporarily 
57 separated from his or her conspecific. Each testing session was voluntary, and the subject could choose to participate 
58 or not.  
59  
60 Materials 
61 The testing board was approximately 25 ½” x 9” x 3” (65cm x 23cm x 8cm). During the food preference test, the 
62 surface was flat, and items were placed approximately 16” (40cm) apart in plastic plates on opposite ends of the flat 
63 board (Figure 1). During the picture recognition tests, a card stand was added to the board so that the picture cards 
64 would stand at a 45 angle facing the bear. Dividers kept the centers of the cards 12” (30cm) apart (Figure 2).  
65 The picture cards were color photos of familiar foods displayed in the same form as the foods were presented to the 
66 bears throughout the food preference test (for example, a 1” (2.5cm) piece of apple, 3 almonds, 1 prune, etc.). The 
67 pictures were printed life sized, in color with a matte finish and white background on 5”x7” (13cm x 18cm) 
68 cardstock. There was a single image of each food, but the picture cards were randomly rotated 180 degrees 
69 throughout the tests.  
70  
71 Procedure 
72 No pre-training took place before the food preference test or before either phase of the picture recognition tests. No 
73 novel food items were included in any of the tests.  The testing took place in the afternoons when the bears would 
74 normally have a training session – after a snack, but before their large PM diet. The coder was blind to the “right” 
75 answer.  
76  
77 Food Preference Test 
78 Twenty-five foods were presented in all possible combinations, two at a time using a paired stimulus preference 
79 assessment (Hopper et al. 2019). Foods were counter balanced for left/right placement and trial order was 
80 randomized as determined by an online generator. 
81 Foods were presented simultaneously side by side on the testing board approximately 16” (40cm) from the mesh. 
82 Then the board was immediately pushed up to the mesh bear enclosure (see Figure 1) so the bear could make a 
83 choice.  Each bear was tested in a total of 42 sessions. Each of the first 41 sessions contained 14 trials. The last 
84 session contained 26 trials to reach the 600 trials needed to make all possible combinations of food counterbalanced 
85 for left/right placement. During the first trial of the first session of the food preference test the bear was given free 
86 choice to choose how to indicate their choice (by blowing, pawing, or nosing towards the food they wanted). Both 
87 bears chose to indicate by blowing in the direction of one item or the other. The bear was given the food they chose, 
88 which they were free to eat or discard. The food the bear did not choose was taken away. If the bear did not select 
89 either food after five seconds, the testing board was momentarily withdrawn and the trial was immediately repeated. 
90 Foods were scored based on how many times the bear selected each food and whether or not a chosen food was 
91 eaten or discarded (1 point for choosing and eating, 0.5 points for choosing and discarding, 0 points if not chosen). 
92 Out of all of the foods in the preference test, 12 were selected by the experimenters to use in the picture recognition 
93 phase– these foods included six of the most preferred foods (mealworms, almonds, leafeater biscuits, raisins, prunes, 
94 and grapes for Niko and raisins, prunes, mealworms, pecans, brazil nuts, and cheerios for Remi) and six of the least 
95 preferred foods (butternut squash, turnip, banana, omnivore kibble, orange, and mango for Niko and honeydew, 
96 orange, omnivore kibble, broccoli, celery, and zucchini for Remi), omitting foods in the mid-value preference range 
97 to clearly delineate high-value versus low-value preferences. Niko’s preferences remained consistant over time. 
98 Remi’s preferences changed during hyperphagia, when bears typically eat more. At this time, Remi would eat even 
99 her least preferred foods, causing the experimenters to take a break from testing until her appetite returned to 
100 normal. Remi was tested on 42 days over an 8 month period. Niko was tested on 42 days over a 2 month period. 
101 Roughly the same volume of each type of food was used so that quantity would not likely influence the bears’ 
102 choices.  
103  
104 Picture Recognition - Phase 1 
105 Using the testing board from the food preference test with the addition of a card stand (see Figure 2), the bear was 
106 presented with two picture cards. Again we presented the stimuli – this time the pictures of food – simultaneously 
107 while the testing board was approximately 16” (40cm) away from the mesh. Then the board was immediately 
108 pushed up to the mesh for the bear to make a choice. We used three of each bear's most preferred foods and three of 
109 each bear's least preferred foods from the food preference test (the other three most preferred and three least 
110 preferred food items were reserved for use in Phase 2). When each bear indicated one of the pictures by blowing at 
111 it, the keeper gave them the food they chose. The picture of the unselected food was taken off the testing board and 
112 the picture of selected food remained on the board while the bear ate or discarded the food. Across all phases of the 
113 experiment, roughly the same volume was used for each type of food so that the bear would not likely learn to 
114 associate larger or smaller quantities to the pictures.  
115  
116 Photos were paired by matching every most preferred food with every least preferred food and counterbalancing 
117 those pairings for left/right placement. Each session consisted of 15 randomized individual trials. When each bear 
118 selected their preferred foods in 80% of the trials over three consecutive testing sessions, we concluded that that bear 
119 recognized the pictures and moved to Phase 2. Note that this is a more stringent criteria than that used in other 
120 similar studies, which required reaching 80% in 2 consecutive trials (Spetch and Friedman 2006; Truppa et al. 2009; 
121 Wein et al. 2015). Remi was tested on 7 days over a 9 day period. Niko was tested on 3 days over a 7 day period.  
122  
123 Picture Recognition Transfer - Phase 2   
124 In Phase 2, we tested whether the picture recognition transferred to novel stimuli (i.e., pictures that the bear had 
125 never seen before). For the transfer, we used pictures of the three most preferred foods and three least preferred 
126 foods that were not used in Phase 1 of the experiment. The experiment followed the same procedure as in Phase 1.  
127 Each session consisted of 15 individual trials. When the bear selected their preferred food in 80% of the trials on 
128 three consecutive testing sessions, we concluded that the bear recognized the pictures. For Phase 2, Remi was tested 
129 on 5 days over a 5 day period and Niko was tested on 3 days over an 8 day period. 
130  
131 Results 
132 Phase 1: With no prior exposure to pictures of food, Niko showed evidence of spontaneous picture recognition, 
133 scoring at or above the 80% correct criteria in the first three sessions. Remi also showed evidence of picture 
134 recognition; however, it took her seven sessions to reach the 80% correct criteria in three consecutive sessions 
135 (Table 1). Although, Remi scored 93% and 100% on her first and second sessions, her performance dropped off 
136 before she was successful.  
137 Phase 2: In the picture recognition transfer phase, Niko met the criteria immediately, scoring above 80% correct on 
138 the first three trials. It took Remi five trials to meet the criteria in this phase (Table 2).      
139  
140 Discussion 
141 This task showed that sloth bears have the ability to recognize 2D pictures as equivalents of their real, 3D 
142 counterparts. Both bears responded to the pictures of food just as they did the pieces of food during the preference 
143 test phase, by using a characteristic sloth bear behavior of blowing (Laurie and Seidensticker 2009) to select the one 
144 they wanted. While both bears recognized the pictures relatively quickly, Niko recognized the pictures 
145 spontaneously, without any training, in both phases of the experiment. Remi acquired the skill through learning 
146 (Bovet and Vauclair 2000) after a few sessions. We suspect it took Remi longer to acquire this skill because her 
147 testing was impacted by factors of natural bear biology – hyperphagia, breeding season, and a pseudo-pregnancy – 
148 which affected her food preferences, hunger, and motivation. Hyperphagia and breeding season did not affect Niko’s 
149 performance since he was immediately successful and therefore went through the tests so quickly that he did not 
150 encounter hyperphagia and breeding season during his tests. These physiological changes should be considered 
151 when doing cognitive tests with any bear species. In order to properly assess food preferences in a species with such 
152 seasonal variation, a wide variety of choices should be tested to create the clearest possible delineation between high 
153 and low value foods. Alternatively, preferences could be tested at various times throughout the year, addressing each 
154 physiological change.  
155 Notably, Niko was able to demonstrate a spontaneous preference for photos that represented his preferred foods. 
156 However, it is unclear what this truly means about a sloth bear’s cognition. Some hypothesize that animals that 
157 recognize pictures spontaneously are simply confusing the pictures with the real objects (Bovet and Vauclair 2000). 
158 It is unlikely that an animal with such a strong olfactory sense would make that mistake. Interestingly, when humans 
159 are capable of  picture recognition from a very early age (2-3 months old), those same researchers hypothesize that 
160 the ability is simply innate (Bovet and Vauclair 2000). More research is needed to reveal the cognitive implications 
161 of Niko recognizing all of the pictures spontaneously.  
162 When animal caretakers are more familiar with the cognitive abilities of the animals in their care, they are better 
163 able to provide the best possible care to their animals, addressing both their physical and mental needs. Utilizing 
164 methods similar to those used in this study (choice between picture cards) could provide an opportunity for sloth 
165 bears to communicate other preferences to their keepers (for example, preferences for specific enrichment items, 
166 yard or enclosure access, or social partners). Adding this type of choice and control into their daily management can 
167 positively enhance the welfare of bears under human care (Mellor 2014; Reiss 2006). 
168 In conclusion, this study provides evidence that sloth bears can recognize 2D pictures of food as representations of 
169 their 3D counterparts. Their ability to recognize 2D pictures of other items is likely. However, further research on 
170 the cognitive abilities of sloth bears is recommended. Since little is known about the cognitive abilities of bears in 
171 general, comparative studies with other bear species are also warranted. Future studies could include increasing the 
172 sample size and comparing inter-individual differences for this methodology, comparing 2D picture recognition 
173 ability in sloth bears to other bear species, as well as exploring whether this ability transfers to more abstract picture 
174 representations such as black and white pictures, illustrations, or line drawings. 
175  
176 Acknowledgments 
177 Thank you to Dr. Betsy Herrelko for her guidance and support as well as the entire NZP Asia Trail team who made 
178 this work possible.  
179  
180 Compliance with ethical standards 
181 All applicable international, national and/or institutional guidelines for the care and use of animals were followed. 
182 All procedures performed in studies involving animals were in accordance with the ethical standards of the 
183 institution or practice at which the studies were conducted. (Smithsonian National Zoological Park Animal Care and 
184 Use Committee #18-15)  
185  
186 References 
187 Amici F, Cacchione T, Bueno-Guerra N (2017) Understanding of object properties by sloth bears, Melursus ursinus 
188 ursinus. Animal Behaviour 134:217–222 
189 Bovet D, Vauclair J (2000) Picture recognition in animals and humans. Behavioural Brain Research 109:143–165. 
190 doi: 10.1016/S0166-4328(00)00146-7 
191 Breiter M (2008) Bears: A Year in the Life. A & C Black 
192 Buchanan-Smith HM (2011) Environmental enrichment for primates in laboratories. Advances in Science and 
193 Research 5:41–56. doi: 10.5194/asr-5-41-2010 
194 Hanggi EB (2001) Can horses recognize pictures. In: Proceedings of the third international conference of cognitive 
195 science, Beijing, China. pp 52–56 
196 Hartmann D, Davila-Ross M, Wong ST, et al (2017) Spatial transposition tasks in Indian sloth bears (Melursus 
197 ursinus) and Bornean sun bears (Helarctos malayanus euryspilus). Journal of Comparative Psychology 
198 131:290–303. doi: 10.1037/com0000077 
199 Hopper LM, Egelkamp CL, Fidino M, Ross SR (2019) An assessment of touchscreens for testing primate food 
200 preferences and valuations. Behav Res 51:639–650. doi: 10.3758/s13428-018-1065-0 
201 Johnson-Ulrich Z, Vonk J, Humbyrd M, et al (2016) Picture object recognition in an American black bear (Ursus 
202 americanus). Animal Cognition 19:1237–1242. doi: 10.1007/s10071-016-1011-4 
203 Laurie A, Seidensticker J (2009) Behavioural ecology of the Sloth bear (Melursus ursinus). Journal of Zoology 
204 182:187–204. doi: 10.1111/j.1469-7998.1977.tb04155.x 
205 Leotti LA, Iyengar SS, Ochsner KN (2010) Born to choose: the origins and value of the need for control. Trends in 
206 Cognitive Sciences 14:457–463. doi: 10.1016/j.tics.2010.08.001 
207 McGuire MC, Vonk J, Johnson-Ulrich Z (2017) Ambiguous Results When Using the Ambiguous-Cue Paradigm to 
208 Assess Learning and Cognitive Bias in Gorillas and a Black Bear. Behav Sci (Basel) 7:. doi: 
209 10.3390/bs7030051 
210 Mellor E (2014) Choice tests; application and relevance in terms of improving husbandry methods and welfare of 
211 captive animals. The Plymouth Student Scientist 7:191–200 
212 Miller RJ (1973) Cross-cultural research in the perception of pictorial materials. Psychological Bulletin 80:135 
213 Nakajima S, Arimitsu K, Lattal KM (2002) Estimation of animal intelligence by university students in Japan and the 
214 United States. Anthrozoös 15:194–205 
215 Perdue B (2016) The Effect of Computerized Testing on Sun Bear Behavior and Enrichment Preferences. 
216 Behavioral Sciences 6:19. doi: 10.3390/bs6040019 
217 Reiss D (2006) Enriching animals while enriching science: Providing choice and control to dolphins. In: Seventh 
218 International Conference On Environmental Enrichment. p 26 
219 Ross SR (2006) Issues of choice and control in the behaviour of a pair of captive polar bears (Ursus maritimus). 
220 Behavioural processes 73:117–120 
221 Shih H-Y, Yu J-F, Wang L-C (2016) STEREOTYPIC BEHAVIORS IN BEARS. Taiwan Veterinary Journal 42:11–
222 17 
223 Spetch ML, Friedman A (2006) Pigeons see correspondence between objects and their pictures. Psychological 
224 Science 17:966–972 
225 Truppa V, Spinozzi G, Stegagno T, Fagot J (2009) Picture processing in tufted capuchin monkeys (Cebus apella). 
226 Behavioural processes 82:140–152 
227 Vickery S, Mason G (2004) Stereotypic behavior in Asiatic black and Malayan sun bears. Zoo Biology: Published in 
228 affiliation with the American Zoo and Aquarium Association 23:409–430 
229 Vonk J (2016) Advances in Animal Cognition. Multidisciplinary Digital Publishing Institute 
230 Vonk J, Jett SE (2018) “Bear-ly” learning: Limits of abstraction in black bear cognition. Animal Behavior and 
231 Cognition 5:68–78. doi: 10.26451/abc.05.01.06.2018 
232 Wein A, Gajdon GK, Schwing R (2015) Picture - Object Recognition in Kea ( Nestor notabilis ). Ethology 
233 121:1059–1070. doi: 10.1111/eth.12423 
234 Wilkinson A, Mueller-Paul J, Huber L (2013) Picture–object recognition in the tortoise Chelonoidis carbonaria. 
235 Animal Cognition 16:99–107. doi: 10.1007/s10071-012-0555-1 
236 Winner E, Ettlinger G (1979) Do chimpanzees recognize photographs as representations of objects? 
237 Neuropsychologia 17:413–420 
238  
239  
240  
Table 1: Phase 1 Results 
Session 1 2 3 4 5 6 7 
14/15 14/15 12/15 
NIKO     
(93%) (93%) (80%) 
14/15 15/15 10/15 8/15 12/15 12/15 13/15 
REMI 
(93%) (100%) (67%) (53%) (80%) (80%) (87%) 
241  
242 Caption for Table 1: Number of correct trials out of 15 trials per session for Picture Recognition – Phase 1.  
Table 2: Phase 2 Results 
Session 1 2 3 4 5 
13/15 13/15 14/15 
NIKO   
(87%) (87%) (93%) 
9/15 10/15 12/15 12/15 12/15 
REMI 
(60%) (67%) (80%) (80%) (80%) 
243  
244 Caption for Table 2: Number of correct trials out of 15 trials per session for Picture Recognition Transfer – Phase 
245 2.  
246