Relative Quantity Judgments in the Beluga Whale (Delphinapterus Leucas) and the Bottlenose Dolphin (Tursiops Truncatus)

Behavioural Processes 96 (2013) 11–19

Contents lists available at SciVerse ScienceDirect Behavioural Processes

journal homepage: www.elsevier.com/locate/behavproc

Relative quantity judgments in the beluga whale (Delphinapterus leucas) and the bottlenose dolphin (Tursiops truncatus)

José Z. Abramson a,b,∗, Victoria Hernández-Lloreda a,c, Josep Call d, Fernando Colmenares a,b a Grupo UCM de Estudio del Comportamiento Animal y Humano b Departamento de Psicobiología, Facultad de Psicología, Universidad Complutense de Madrid, Campus de Somosaguas, 28223 Madrid, Spain c Departamento de Metodología de las Ciencias del Comportamiento, Facultad de Psicología, Universidad Complutense de Madrid, Campus de Somosaguas, 28223 Madrid, Spain d Department of Developmental and Comparative Psychology, Max Plank Institute for Evolutionary Anthropology, Leipzig, Germany article info abstract

Article history: Numerous studies have documented the ability of many species to make relative quantity judgments Received 16 July 2012 using an analogue magnitude system. We investigated whether one beluga whale, Delphinapterus leucas, Received in revised form and three bottlenose dolphins, Tursiops truncatus, were capable of selecting the larger of two sets of 18 December 2012 quantities, and analyzed if their performance matched predictions from the object file model versus Accepted 18 February 2013 the analog accumulator model. In Experiment 1, the two sets were presented simultaneously, under water, and they were visually (condition 1) or echoically (condition 2) available at the time of choice. In Keywords: experiment 2, the two sets were presented above the water, successively (condition 1) or sequentially, Numerical cognition Relative quantity judgment item-by-item (condition 2), so that they were not visually available at the time of choice (condition 1) Beluga whale or at any time throughout the experiment (condition 2). We analyzed the effect of the ratio between Bottlenose dolphin quantities, the difference between quantities, and the total number of items presented on the subjects’ Object file model choices. All subjects selected the larger of the two sets of quantities above chance levels in all conditions. Analog accumulator model However, unlike most previous studies, the subjects’ choices did not match the predictions from the accumulator model. Whether these findings reflect interspecies differences in the mechanisms which underpin relative quantity judgments remains to be determined. © 2013 Elsevier B.V. All rights reserved.

1. General introduction Davis et al., 1989; dogs: Bonanni et al., 2011; Ward and Smuts, 2007; sea lions: Abramson et al., 2011; elephants: Irie-Sugimoto Numerical cognition stands at the core of unique human cog- et al., 2009; horses: Uller and Lewis, 2009; crows: Smirnova et al., nitive achievements but its evolutionary origins and precursors 2000; parrots: Pepperberg, 1987, 2006; pigeons: Alsop and Honig, remain to be established (Uller, 2008). The ability to assess relative 1991; chicks: Rugani et al., 2007; amphibians: Uller et al., 2003; quantities is of adaptive value to deal with some of the ecological ﬁshes: Agrillo et al., 2007; Gòmez-Laplaza and Gerlai, 2011; insects: and social challenges that many animal species encounter in their Carazo et al., 2009). These ﬁndings suggest that human numer- lives (Shettleworth, 2010). Accordingly, there is growing evidence ical skills might build on an evolutionarily ancestral capacity, a of rudimentary numerical competence in numerous non-human language-independent representation of numbers, that supports species (great apes: Anderson et al., 2005; Beran et al., 2005; Boysen elementary arithmetic computations and that is somehow shared and Berntson, 1989; Call, 2000; Dooley and Gill, 1977; Hanus and by nonverbal animals (for overviews see Brannon and Terrace, Call, 2007; Rumbaugh et al., 1987; Silberberg and Fujita, 1996; 2002; Butterworth, 1999; Cantlon, 2012; Dehaene, 1997; Gallistel, monkeys: Anderson et al., 2000; Beran, 2007a,b; Hauser et al., 1990; Matsuzawa, 2009; Uller, 2008). 2000; lions: McComb et al., 1994; raccoons: Davis, 1984; rats: Several mechanisms have been proposed to account for this competence. Early proposals included perceptual mechanisms such as subitizing (Davis and Perusse, 1988) or prototype matching (Thomas, 1992). However, neither can account for some of the ∗ Corresponding author at: Departamento de Psicobiología, Facultad de Psicología, evidence currently available, for example, when individuals dis- Universidad Complutense de Madrid, Campus de Somosaguas, 28223 Madrid, Spain. criminate between quantities that fall outside the subitizing range Tel.: +34637003081. E-mail addresses: [email protected], [email protected] (Dooley and Gill, 1977), or when they discriminate between sets (J.Z. Abramson). that are presented sequentially or item-by-item (Beran, 2001; Call,

200; Hanus and Call, 2007; Irie-Sugimoto et al., 2009). In fact, the Cetacean odontocetes might be a particularly interesting group ability to solve the latter condition requires a mental comparison on which to investigate numerical abilities given the peculiarities of mechanism that goes beyond a merely perceptual one. This non- their socio-ecology and neurobiology compared with most terres- perceptual mechanism can take two forms, namely, the object file trial mammals. Bottlenose dolphins (Tursiops truncatus) can learn to model, which is a digital mechanism (Kahneman and Treisman, choose the object with the greatest fish value (Mitchell et al., 1985) 1984; Kahneman et al., 1992; Simon et al., 1995; Uller et al., and are capable of representing ordinal relations among numerosi- 1999), and the accumulator model, which is an analog mechanism ties (Jaakkola et al., 2005; Killian et al., 2003). It has been proposed (Dehaene, 1997; Gallistel and Gelman, 2000; Meck and Church, that dolphins’ quantity representation is supported by an analog 1983; Wynn, 1998). magnitude mechanism (Jaakkola et al., 2005). Although most would agree that both kinds of representations However, it is unknown whether bottlenose dolphins would could be used in numerical reasoning, there is no agreement about also be able to choose the largest of two quantities when they are its extent (Dehaene, 2001; Carey, 2001; Wiese, 2003). According to not perceptually available at the time of choice. Recall that quan- the object file model, numerical capacity works on mental (sym- tities in previous studies with dolphins were always perceptually bolic) representations of a set of visual objects, i.e., the object available at the time of choice. Moreover, as far as we know, no files. Each object file is a mental token that represents each ele- studies have investigated the quantitative abilities of other odon- ment of a set of elements, yielding exactly as many files (mental tocetes. Beluga whales (Delphinapterus leucas), in particular, may tokens) as objects that are filed in short-term memory (Uller et demonstrate cognitive abilities comparable to those of bottlenose al., 1999; Wiese, 2003). In making relative quantity judgments, dolphins given that they also possess large brains, live in sophisti- the subject would create one object file for each item presented cated social systems, and are highly trainable (Brodie, 1989; Defran in a set, and would thus produce a one-to-one correspondence and Pryor, 1980; Samuels and Tyack, 2000; Nowak, 1991). between “active” object files and the number of presented items In the present study, we investigated the ability of one beluga (Hauser and Carey, 1998). Quantitative differences between two and three bottlenose dolphins to compare and operate on quanti- arrays are detected by comparing the two representations. In ties of food items. We wanted to determine whether, in the absence contrast, according to the accumulator model, animals cannot dis- of training, subjects were able to (1) select the larger of two sets of criminate absolute numbers or label each separate object; instead quantities, (2) using both visual cues or echolocation, (3) when the they recognize quantities by means of an accumulated analog quantities were not perceptually available at the time of choice, (4) representation, the accumulation of continuous quantities in pro- for a range of quantities that allowed us to characterize the repre- portion to the number of quantified elements (Meck and Church, sentational system underpinning their responses. To address these 1983). That is, discrete quantities can be represented as mental questions, we tested four different conditions in two experiments. magnitudes that could be seen as an analog of the perceived dis- In experiment 1 the two sets of quantities were presented simulta- crete quantities. In making relative quantity judgments, the animal neously as whole sets, that is, the subjects could compare the two would have two noisy analog representations of the two sets. The sets directly (e.g., Call, 2000; Hanus and Call, 2007; Irie-Sugimoto extent to which these two fuzzy representations overlap would et al., 2009). In condition 1, the two sets were presented visually, determine the likelihood of confusion about their relative magni- whereas in condition 2 the two sets were presented in opaque tude. boxes, that is, to pass the test, the subject had to use echolocation. Although the object file model does provide an accurate repre- In experiment 2, the two sets of quantities were presented succes- sentation of quantities, it is, however, limited to small quantities, sively (condition 1) or sequentially, item-by-item (condition 2), so because a mechanism like this requires memory space, which that the totality of items were perceptually unavailable at the time implies a limited capacity of processing quantities greater than 3 of choice (condition 1) or at any time throughout the task (condition or 4 due to limitations for processing stimulus items that are per- 2) (Beran, 2001; Call, 2000; Hanus and Call, 2007; Irie-Sugimoto ceptually available simultaneously (Uller et al., 1999). By contrast, et al., 2009). Thus, to succeed in the two conditions of experi- although the analog magnitude representation of the accumulator ment 2, subjects were required to store and compare the quantities model is rather fuzzy (Gallistel and Gelman, 2000), as the accuracy mentally. Finally, to explore the nature of the mental mechanism of discrimination decreases with increasing quantities (when the underlying the quantitative cognition of the subjects, we analyzed, absolute differences are kept constant), it can, however, deal with for each type of presentation, the effect of the ratio between quan- larger sets, since it has no a priori limit (Dehaene, 2001; Gallistel tities, the difference between quantities, and the total number of and Gelman, 2000). items presented. A predominant influence of the ratio between The vast majority of studies on relative quantities support the quantities on the subject’s performance (less accurate discrimi- accumulator model rather than the object file model because they nation as the ratio between quantities increases) would support have reported performance that is ratio-dependent with no clear- an analog representational mechanism, as suggested by the accu- cut drop out beyond certain quantities. In fact, there are only mulator model. In contrast, the object-file model predicts a cut-off three studies (all in rhesus macaques) that have found some evi- point in discrimination ability when quantities exceed the number dence in favor of the object file model (Hauser and Carey, 2003; of four. Hauser et al., 2000; Wood et al., 2008). However, Beran (2007a,b) also tested the same species finding no support for the object file model. Interestingly, Irie-Sugimoto et al. (2009) found that 2. General methods Asian elephants consistently selected the larger of two quantities but found neither a ratio-dependent performance nor a clear-cut 2.1. Subjects and general procedure drop out beyond certain quantities. Irie-Sugimoto et al. (2009) suggested that Asian elephants by virtue of their memory capacity may We tested one female beluga whale (D. leucas) housed at possess a ‘large object-file model’. Note, however, that African ele- L’Oceanografic Aquarium in Valencia, Spain, and three female bot- phants, unlike their Asian counterparts, do show ratio-dependent tlenose dolphins (T. truncatus), two of which were housed at performance, which is consistent with the accumulator model (Irie, Marineland Aquarium in Antibes, France, and one was housed at 2012; Perdue et al., 2012). The putative difference between these Madrid Zoo Aquarium, in Spain (Table 1). two closely related species is intriguing and deserves further com- The beluga whale (Y) was mother-reared and captured when parative scrutiny. she was 1-year-old in the Okhotsk Sea. She had been housed at J.Z. Abramson et al. / Behavioural Processes 96 (2013) 11–19 13

L’Oceanografic since 2003 and had lived with a male beluga in the same pool ever since then. Y’s experimental sessions were con- ducted typically between 11:00 a.m. and 2:00 p.m. and between 4:00 p.m., and 6:00 p.m.; she participated in one to three sessions per day. The subject was fed approximately 18 kg of freshly thawed herring, hake, capelin, and pota each day, one half of which was typically consumed during experimental sessions. Two of the female bottlenose dolphins, S and MA, were captured in 1983 and 1990, respectively, and were mother reared. M was S’s daughter, born in captivity. All dolphins were socially housed. S was tested typically from 9:00 a.m. to 11:00; M from 11:00 to 12:00, and, finally, MA was tested typically from 12:00 a.m. to 14:00 p.m. The subjects were fed approximately 4–6 kg of freshly thawed herring and capelin each day, one half of which was typically consumed during experimental sessions. The three dolphins were trained for a variety of examination and exercise behaviors with standard operant conditioning procedures and fish reinforcement. Prior to this experiment, all of the subjects participated in biological studies and veterinary procedures, but none of them had ever participated in cognitive studies. The subjects were never deprived of food in any way, regardless of performance.

3. Experiment I: whole set presentation

3.1. Method

3.1.1. Subjects

We tested the beluga whale, Y, (conditions 1 and 2) and two Fig. 1. Whole set experimental set-ups: (a) condition 1 visual presentation; (b) con- bottlenose dolphins Sharky (S) and Mila (M) (condition 1) (Table 1). dition 2 echoic presentation. Right side of the ﬁgure b shows how the quantities are The two dolphin subjects were also tested in condition 2, however, hung on an inverted T shape stainless steel structure, like a skewer, and then placed they did not pass the pre-test. in the opaque box.

3.1.2. Apparatus the front panel. Like in condition 1, the boxes were hung from a 3.1.2.1. Condition 1: visual presentation. Two identical methacry- wooden platform situated outside the pool by two methacrylate late transparent boxes (35 cm × 25 cm × 7 cm) placed inside the braces attached to it. The entire boxes were below the surface of pool (see Fig. 1a) were used in this condition. The top of the the water in the pool. Pool water entered the box from the bottom boxes was open. We hanged the boxes by two methacrylate braces and the back, filling the interior to the water level outside the box. attached to a wooden platform outside the pool. This arrangement The quantities to be inspected echoically were hung underwater by held the boxes in a fixed orientation that was maintained through- means of an inverted T shape stainless steel structure. The pieces out all trials. The entire box was below the surface of the water in of fish hung on this structure, like a skewer, and were placed in the the pool. Pool water entered the box from three holes placed in the box across two slits located on the top of both side panels of each bottom panel, filling the interior to the water level outside the box. box. The quantities to be inspected visually were kept underwater by means of two methacrylate panels that contained six hooks where 3.1.3. Procedure the pieces of fish were hung. Once filled with the different quan- Subjects underwent a habituation period to get familiarized tities, these panels, which were attached to methacrylate handles, with the testing situation. All subjects were individually tested in were inserted into the underwater transparent boxes. their respective pool with the help of two trainers. Prior to each testing session, we required the subjects to undergo pretest tri- 3.1.2.2. Condition 2: echoic presentation. We exposed quantities to als that were designed to check that the subjects had acquired the subject’s echolocation sense alone by presenting them within two task’s basic prerequisites and were sufficiently motivated. If sub- identical opaque boxes (35 cm × 25 cm × 7 cm) placed inside the jects refused to participate in a given session, we rescheduled it pool (see Fig. 1b). The back and bottom of the boxes were open for a later time. In pretest trials, the subject was shown two boxes; and the front box consisted of a 0.3-cm-thick black polyethylene one box was baited and the other was empty. They were trained to panel that precluded any visual inspection of the contents of the approach and touch the baited box to receive the fish. When they box. In water, polyethylene of this type transmits sound well (e.g., reached 80% percent of correct choices we considered they were 75% transmissibility at 60 kHz; see Pack & Herman, 1995), allowing ready to be tested. As for the potential effect of a lack of motivation the subject to inspect the contents of the box echoically through on performance, we required the subject to undergo 1-0 and 6-1

Table 1 Subjects, settings, and testing conditions.

Subject Species Sex and age Setting Experiment and conditions

Yulka [Y] Beluga whale F, 12 years L’Oceanograﬁc Aquarium Experiment I, conditions 1 and Experiment II, condition 2 Sharky [S] Bottlenose dolphin F, 30 years Marineland Aquarium Experiment I, condition 1 Mila [M] Bottlenose dolphin F, 2 years Marineland Aquarium Experiment I, condition 1 Mancha [MA] Bottlenose dolphin F, 25 years Zoo Aquarium Experiment II, condition 1 14 J.Z. Abramson et al. / Behavioural Processes 96 (2013) 11–19

subjects could solve the tests visually (or visually and echoically, visual-echoic condition). In contrast, in condition 2 the quantities were not visually available at the time of choice, so the only way the subject could solve the tasks was by means of echolocation (echoic condition). Owing to differences in the facilities between the two aquariums, we used a slightly different protocol with the two species. Beluga whale: Trainer 1 and experimenter 1 filled each of the boxes with a certain amount of fish out of the subject’s view. In this protocol, we did not follow a strict double blind test, because we did not have enough trainers available. However, when trainer 2 gave the subject the command “go”, experimenter 1 and trainer 1, who helped to put the amounts and gave the selected set, and, therefore, knew the correct choice once the boxes were baited, faced away, outside the view of the subject, in order to avoid cuing her. When the subject made her choice by touching one box by the mouth, experimenter 2 gave the signal to trainer 1, who took away the Fig. 2. Experimental set-ups: (a) for the beluga whale; (b) for the bottlenose dolphins. unselected box and gave the content of the chosen box to the subject in case she selected the box with the larger amount of fish. In addition, when the subject went back to the starting position, pretest trials, and to choose the baited box for three consecutive experimenter 1 and trainer 1 set the bait into the boxes for the trials. next trial (Fig. 2a). Therefore, in each trial the subject went from its Once the subject met these two criteria, we presented the test initial position (with trainer 2) directly under water to the boxes, trials in which both boxes were baited with different amounts of without watching trainer 1 and making the decision before arriving fish. Each subject was presented with one to three daily sessions at the choice position. comprising between 10 and 12 two-alternative forced-choice tri- Bottlenose dolphins: We introduced an extra control and ran the als. Usually twelve comparisons were presented randomly to each experiment double blind due to the availability of more trainers. subject in each session: 1 versus 2; 1 versus 3; 1 versus 4; 1 versus The experimenter and another assistant filled each of the boxes 5; 2 versus 3; 2 versus 4; 2 versus 5; 2 versus 6; 3 versus 4; 3 versus with a certain amount of fish out of the subject’s and trainer 1’s 5; 3 versus 6 and 4 versus 6. The inter-trial interval was approx- view. Next, the experimenter signaled to trainer 1 to go to the test- imately 10 s. The subject’s response was considered correct when ing point to begin the trial. The subject waited at the waiting point she chose the box with the larger quantity. The quantities were (playing and doing exercises) until she was signaled by trainer 2 to randomly determined and counterbalanced on the right and left go to the testing area to make a choice. The subject responded by sides. Trials were repeated when the subject did not make a clear moving from the waiting area to the testing place where s/he would choice (she did not touch or approach any box or did not go from touch one of the two boxes with her nose (Fig. 2b). Once the subject the waiting place to the testing arena). If the subject lost inter- had made the choice, the experimenter gave the signal to trainer est during presentation, the trial was cancelled and restarted later. 1 to reward the subject in case of a correct choice. If the subject The subject received 6 presentations of each comparison. A session passed the pretest trials, she received 6 presentations of each com- was considered completed when the 12 comparisons were run or parison, There were three trials which were not recorded and had when trainers indicated that a subject was tired and/or unwilling to be excluded (one 5 vs. 1 trial for Mila and one 4 vs. 2 and one 5 vs. to participate. 3 trials for Sharky) because the camera was accidentally switched Trainer 1 gave the reward in case of a correct choice, and trainer 2 off. Finally, the numbers of any particular comparison varied across maintained the subject in a waiting point from which it was impos- conditions. sible for the subject to see the baiting procedure in the underwater boxes (Fig. 1a and b). During the entire session, trainer 2 remained 3.1.4. Data scoring and analysis at the starting point and gave only non-verbal commands “go” and All trials were live coded and videotaped with a digital full HD “stay”. Once the boxes were filled, trainer 2 was signaled to do the 1080 Sony HDR-XR520 camera. Our main dependent variable was command “go”. The subject typically approached the boxes and the percentage of trials in which subjects selected the larger of the touched one of them (Fig. 2). This was recorded as the subject’s two quantities (i.e., correct responses). A second observer scored choice of box. The trainer took away the content of the unselected 40% of the sessions to assess inter-observer reliability, which was box and gave the content of the chosen box to the subject. excellent (Cohen’s kappa = 0.97, p < 0.001). Subjects only received the reward if they chose the box with Given our small sample sizes, we analyzed the data per indi- the larger quantity. We adopted this protocol in order to avoid vidual. We used binomial tests for each subject to analyze whether potential interference from their conditioning-based training pro- subjects performed above chance levels (p = 0.5). We also calculated cedures, which were different for the two species. Although the the Pearson correlation coefficients between the percent of cor- beluga whale was trained for a variety of examination and exer- rect responses and the (1) difference (larger quantity minus smaller cise behaviors with standard operant conditioning procedures and quantity) (2) ratio (smaller quantity/larger quantity) and (3) total fish reinforcement, she did not participate in shows. In contrast, quantity (smaller quantity + larger quantity). dolphins participated in shows and were instrumentally trained to In order to test whether the data matched the prediction from display acrobatics and other behaviors. Subjects so trained, espe- the object file model, we divided the trials into two categories: cially dolphins, might face a problem when exposed to tasks in below and above the limit predicted by the object file model (i.e. 4 which they were expected to freely choose what they themselves items in a set; see, e.g. Beran and Beran, 2004; Hauser et al., 2000; wanted rather than what they might expect the trainer wanted Uller et al., 1999). In order to rule out the possibility that dolphins them to perform. That is why, contrary to other studies with and belugas have an object file limit different from 4 because of dif- untrained animals, we decided to use a reward-based approach. In ferences in memory skills, we explored the subjects’ performance Condition 1 we presented the quantities in transparent boxes, thus for each comparison by looking for some indication of a set size J.Z. Abramson et al. / Behavioural Processes 96 (2013) 11–19 15

Table 2 we tested did not fully grasp the task in this condition or they were Percentage of correct responses and correlation coefﬁcients between the percent- habituated not to use echolocation in general. age of correct responses and the predictors investigated in the visual condition of Experiment 1. 4. Experiment 2: non-simultaneous presentation Subject % Correct N Correlations with % correct

Ratio Diff Total Object 4.1. Condition 1: successive presentation Y72*** 72 −.25 .43 .27 −.13 M68** 71 .27 −.10 .18 .12 4.1.1. Method Sh 80*** 70 −.02 .07 .22 −.17 4.1.1.1. Subject. We tested one bottlenose dolphin, Mancha (MA). Note: Correlations coefficients (N = 12 in all cases) between percent correct and ratio, difference, and total are based on Pearson correlations whereas the correlations 4.1.1.2. Apparatus. We used the same methacrylate transparent between percent correct the object file is based on the point-biserial coefficient. boxes as in condition 1 of experiment 1, but in this case the boxes ** p < 0.01. were placed outside the pool. To hold them out of the water, we *** p < 0.001. attached the boxes to two horizontal methacrylate braces and fixed the whole set in the floor outside the pool in a manner that held limitation (such as an abrupt change in performance when tested the boxes in a fixed orientation that was maintained throughout all for comparisons over or below a particular quantity). The point- trials. In addition, an opaque lid mounted on hinges was placed on biserial correlation coefficient was calculated for each subject each of the boxes. between the proportion of correct trials and this categorical variable. 4.1.1.3. Procedure. The general testing procedure was similar to experiment 1 except that we used an operant conditioning pro- 3.2. Results cedure similar to what dolphins had experienced during training and public shows. In particular, the choice of the stimulus desig- Table 2 presents the percentage of correct trials per subject nated as correct (the larger quantity) was reinforced by delivering (and the correlation with the four predictor variables) in the visual 3 to 4 fishes, which was the standard (amount of) reward in any condition. All subjects chose the box with the larger quantity signif- correct behavioral performance during management practices and icantly above chance in the visual condition (Binomial tests: p < .005 during public shows. In case of a wrong choice, no fish reward was in all cases). However, there was no significant relation between given. performance and any of the four predictor variables (r < .44, p > .15 The reason for adopting this conditioning approach was in all cases, Table 2). motivated by the challenge of presenting successive amounts Dolphins failed the pre-test phase in the echoic condition. Con- underwater while keeping and monitoring the subject’s attention sequently, this condition was only tested in the beluga whale. on the stimuli presentation. Presenting the stimuli in the surface She selected the larger quantity significantly above chance (% was far more manageable but this location had the disadvantage correct = 65%; binomial test, n = 72, p = .0132). Pearson correlation that the trainers’ presence hindered the dolphins’ spontaneous analyses revealed that performance improved as the total amount responses (i.e., MA showed a tendency to wait to see what she was − of elements to be compared decreased (rtotal = .59, n = 12, p = .041). expected to do). Therefore, MA was trained to choose the larger In contrast, there was no significant relation between performance quantity in a sequential modal presentation using three pairs in − − and ratio (rratio =. 0.11, n = 12, .739) or difference (rdifference = .41, the training phase and was later tested to see if she had learned a n = 12, p = .183). Likewise, there was no significant relation between rule to respond to the largest of the two quantities by presenting performance and object file (robjectfile = .52, n = 12, p = .08). the subject with novel pairs in the testing phase. Unlike in experiment 1, the two sets of quantities were pre- 3.3. Discussion sented successively instead of simultaneously, and none of them was perceptually available at the time of choice (see Call, 2000). A beluga whale and two dolphins selected the larger of two Trainer 2 removed the lid from one of the boxes while leaving the quantities when they were simultaneously presented visually but, other box covered (Fig. 3; step one). After 5 s had elapsed, he cov- contrary to what has been found in most species studied so far, ered the first box with the lid, uncovered the second box (Fig. 3; including dolphins (Jaakkola et al., 2005), performance did not vary step two) and waited 5 s before covering it again. Then, Trainer 2 as a function of the ratio or difference between the two quan- waited an additional 3 s and gave a signal (hand clapping) to let the tities or the total number of elements present. Additionally, the subject make her choice by touching one of the two boxes with her beluga whale, unlike the dolphins, used echolocation to discrimi- rostrum (fish always out of Trainer 2’s view) (Figure 3; step three). nate between simultaneously presented quantities when they were 4.1.1.3.1. Training. We used three pairs: 0 vs. 1, 1 vs. 4, and 2 not visually available. Here her accuracy in detecting the larger of vs. 5. Initially, the discrimination was simplified as much as possi- two quantities decreased as the total amount of elements increased. ble using only a single pair: 0 vs. 1. Reinforcement followed when Our data in the visual condition supported neither the accumu- the subject chose the box containing the 1 piece of fish. After MA lator nor the object file model, thus joining a minority of studies that performed correctly on the 1 vs. 0 pair, she was trained with the have documented above chance performance without the signature other two quantities. Each of the daily sessions consisted of a total of a magnitude estimation or an object file system (Asian elephants: of 10 trials. The left-right position of the larger quantity changed in Irie-Sugimoto et al., 2009; dogs: Ward and Smuts, 2007; chim- a quasi-random sequence. Once the performance reached the 80% panzees: Beran, 2004). In contrast, the beluga data in the echoic correct responses we began the test phase. condition suggested the operation of a magnitude system. 4.1.1.3.2. Testing. The procedure was the same as in the train- We found that the beluga whale was capable of using echolo- ing phase except that we introduced four new pairs (8 vs. 2; 7 vs. cation (without the benefit of eyesight) to discriminate between 4; 10 vs. 4; 8 vs. 5) and the trials were run double blind. The novel quantities. Dolphins in the current study failed to distinguish pairs were chosen so that one of them had the same ratio as the between 1 versus 0 in the echoic condition, even though dolphins trained comparisons and the other had the same difference. More- can detect and discriminate objects by echolocation (Au, 1993). We over, two of the comparisons had as the small quantity the one do not have an explanation for this result. Perhaps the dolphins that trained to be chosen in the training trials in order to rule out that 16 J.Z. Abramson et al. / Behavioural Processes 96 (2013) 11–19

the 8 vs. 2 pair trials, 67% of the 7 vs. 4 and 8 vs. 5 pair trials, and 92% of the 10 vs. 4 pair trials.

4.2. Condition 2: sequential item-by-item presentation

4.2.1. Method 4.2.1.1. Subject. We tested the same beluga whale as in experiment 1.

4.2.1.2. Apparatus. We used a pair of identical square shaped opaque buckets (40 cm depth and 25 cm in diameter). These buckets were presented above the surface in a wooden platform, thus making them visually but not echoically perceptible, since belugas and dolphins cannot echolocate in air, because of the impedance mismatch between water and air is too great (Au, 1993).

4.2.1.3. Procedure. The trainer placed both buckets on the platform in an upright position, approximately 60 cm apart in front and 70 cm away from the subject. The trainer first took out pieces of fish from a cube and dropped them one at a time into each bucket in full view of the animal, starting with the left bucket and then the right one. In doing so, the subject could only see one item at a time falling into the buckets and/or listen to the sounds made when the food hit the bottom of the bucket. The buckets were high enough so that the subject could not see its content. Once the last fish was dropped into the right bucket, the trainer moved the platform in order to place the buckets within the subject’s reach so that she could make her choice. The trainer took away the unselected box and gave to the subject the amount of fish placed in the chosen box. As in the other conditions, we required the animal to undergo Fig. 3. Successive presentation experimental set-up. pretest trials prior to each testing sessions. In these pretest trials, we wanted both sets to have food, so that the rule “go where there is any food” could not be used. We chose the quantity set 1 versus subjects were responding to specific quantities rather than pairs of 6 fishes. After the presentation, the experimenter looked down to quantities. The order of pair presentation was: 8 vs. 2 and 7 vs. 4, avoid cueing the animal. As in experiment 1, the subject received 6 followed by the familiar 4 vs. 1 and 8 vs. 5 and 10 vs. 4, followed times each of the 12 pairings resulting in 72 trials in total. by the familiar 5 vs. 2 trials. MA received a total of 12 trials for each new set of quantities. 4.2.1.4. Data scoring and analysis. Data scoring and analysis were Each testing session consisted of 10 trials, eight trials containing the same as in experiment 1. training pairs and two test trials with novel pairs. When presenting the new quantities, every choice, whether correct or incorrect, was 4.2.2. Results rewarded to guard against associative learning. The order of the Y chose the box with the larger quantity significantly more sessions was as follow: six 8 vs. 2 sessions (training pairs: 1 vs. often than chance (proportion = .64, binomial test, n = 72, p = 0.02). 4); six 7 vs. 4 sessions (training pairs: 1 vs. 4); six 8 vs. 5 sessions No relation was found between performance and ratio, difference (training pairs: 5 vs. 2 trials) and six 10 vs. 4 sessions (training and total amount (n = 12; rdifference = .05, p = .89; rratio = .05, p = .88; − pairs: 5 vs. 2 trials). Thus while the training and testing pairs were rtotal = .17, p = .59; robjectfile = .27, p = .39). Table 3 summarizes the presented in blocks between sessions, their order within a session subject’s performance in each comparison. was randomized. The reason for presenting the various test pairs in blocks was to prevent a potential interfering effects between two 4.3. Discussion or more types of test pairs. Distance between the fishes inside the box was varied during the session, in order to control for the size of One bottlenose dolphin selected the larger quantity in all novel the area they covered in the box. To rule out potential motivational numerosity pairs even though she was prevented from seeing both effects, prior to each test trial, we required the animal to undergo quantities at the same time. Moreover, the beluga whale selected pretest trials: the subject had to succeed in three consecutive trials the larger quantity significantly above chance levels despite the fact with the trained pairs. that she was prevented from seeing both quantities as a whole, but had to rely on keeping track of the quantities added into each pair. These results demonstrate that both dolphins and beluga whales 4.1.1.4. Data scoring and analysis. Data scoring and analysis were are capable of making relative quantity judgments in the absence of the same as in experiment 1. the stimuli both when they are required to compare whole quantities mentally or they have to keep track of the addition of individual 4.1.2. Results items. Thus, dolphins and beluga whales are comparable to several Overall, MA chose the correct option on 75% of the novel pair tri- other species that have been shown to be capable of mentally com- als (binomial test, n = 48, p = .001). Although we do not have enough paring quantities (e.g., apes: Beran, 2001, 2004; Call, 2000; Hanus pairs of comparisons to analyze the patterns of errors related to and Call, 2007; elephants: Irie-Sugimoto et al., 2009; horses: Uller other variables, performance seemed to improve as the disparities and Lewis, 2009; dogs: Ward and Smuts, 2007; sea lions: Abramson of comparison increased. Thus, MA performed correctly in 75% of et al., 2011). J.Z. Abramson et al. / Behavioural Processes 96 (2013) 11–19 17

Table 3 Percentage of correct trials for each subject per comparison in Experiment 1, conditions 1 and 2 and experiment 2, condition 1.

Comparison % Correct

Experiment 1 Experiment 2

Condition 1 Condition 2 Condition 1

Beluga whale Botllenose dolphins Beluga whale Beluga whale YM SHYY

1 versus 2 67 67 67 83 50 1 versus 3 50 50 100 67 67 1 versus 4 100 83 6 67 67 1 versus 5 83 4 67 83 50 2 versus 3 67 83 67 83 83 2 versus 4 67 67 100 50 50 2 versus 5 67 83 100 67 67 2 versus 6 83 67 83 50 50 3 versus 4 100 67 67 100 67 3 versus 5 50 33 8 50 100 3 versus 6 100 67 67 67 100 4 versus 6 67 100 100 50 50

However, once again we found little support for either the accu- final pairs of sets were not perceptually accessible at the time of mulator or the object file model. Only the dolphin data seemed choice (successive condition) and rewards were no longer visible to indicate that performance improved as the disparity between after they were placed into the buckets (sequential condition). quantities increased. This suggests that the accumulator model is The findings from experiment 2 suggest that both species are the likely candidate mechanism responsible for the results in this capable of mentally comparing quantities, matching the results task. from previous studies on relative quantity discrimination in other non-human species (e.g., apes: Beran, 2001, 2004; Call, 2000; Hanus 5. General discussion and Call, 2007; elephants: Irie-Sugimoto et al., 2009; horses: Uller and Lewis, 2009; dogs: Ward and Smuts, 2007; sea lions: Abramson Both beluga whales and bottlenose dolphins were capable of et al., 2011). It is worth noting, however, that the percentage of selecting the larger of two quantities when they are presented correct responses by the beluga in the item-by-item presentation simultaneously as pairs of visually accessible whole sets (exper- (experiment 2, condition 2) contrasts with that obtained in the iment 1, condition 1). Additionally, the beluga succeeded when whole set presentation (experiment 1), 64% versus 72% of the tri- the stimuli were not visually available but could only be detected als, respectively. The higher attentional demands required by the by echolocation (experiment 1, condition 2). Our findings also item-by-item presentation could explain this discrepancy. Similar revealed that both species are capable of selecting the larger of two results have been found in primates where the performance in the quantities when these are presented successively, one set at a time item-by-item presentation is substantially lower than the perfor- (for bottlenose dolphins; experiment 2, condition 1) and sequen- mance in the whole set presentation (e.g., Evans et al., 2009; Hanus tially, item-by-item (for the beluga whale; experiment 2, condition and Call, 2007). 2). In both cases, this implies that they can select the larger quan- Remarkably, the beluga whale’s performance was quite similar tity even though the two final pairs of sets are not perceptually to that reported for sea lions (Abramson et al., 2011), i.e., 64% overall accessible at the time of choice. accuracy for the beluga versus 66% for the South American sea lion. Taken together, these results confirm those obtained in previous Thus, it seems that the capacity to mentally represent and compare studies on relative quantity judgment of whole sets in marine mam- quantities is not just present in terrestrial mammals or amphibians mals (sea lions: Abramson et al., 2011; dolphins: Jaakkola et al., marine mammals, such as sea lions and cetaceans also possess it. 2005; Killian et al., 2003) and extend the evidence in three main It is important to note that for both the echoic presentation and directions. First, subjects were capable of selecting the larger quan- the item-by-item presentation, the total quantity of items turned tity even though they were prevented from directly (perceptually) out to be the only variable that could explain part of the variability comparing the two sets of quantities during their presentation or in the beluga’s performance. One may speculate that the capacity at the time of choice. Second, it documents relative quantity judg- for echolocation could be responsible for a more precise “digital” ments using echolocation (in the beluga whale). Third, it reports mechanism, although further work is needed to investigate this relative quantity judgments in a species never tested before, the possibility. beluga whale. With regard to the variables that best explained performance, The results in experiment 1 parallel those reported for sea lions the results did not produce a completely consistent picture. Thus, (Abramson et al., 2011), some non-human primates (Call, 2000; while in both species the differences in the percentage of cor- Hanus and Call, 2007), and other terrestrial mammalian species rect choices shown in the simultaneous visual condition could be (dogs: Ward and Smuts, 2007; elephants: Irie-Sugimoto et al., accounted for neither by the ratio, the difference, nor by the total 2009). All of the subjects chose the larger quantity above chance quantity of elements, in the echoic condition the total quantity of in the simultaneous presentation, albeit we found important inter- elements emerged as the variable that better explained most of individual differences in the level of correct choices, with results the variation in the beluga’s performance. In the successive condi- ranging from 68% to 80%. However, success in choosing the larger tion, in contrast, the bottlenose dolphins’ performance improved quantity in experiment 1 could be based on directly comparing as the disparities between the two sets increased, whereas in the the two quantities, as both sets are simultaneously visible to the sequential condition, the beluga’s performance was related to the subjects at the time of choice. Experiment 2 addressed this issue total quantity of items, as was found to be the case in the echoic directly. It tested whether the subjects also discriminated between condition as well. Although it is true that in some sessions the quantities when those had to be mentally compared, as the two beluga exhibited a side bias effect that lowered her general score; 18 J.Z. Abramson et al. / Behavioural Processes 96 (2013) 11–19 overall, we did not find any bias effect that could account for the cognitive system engaged. Important advances may follow in the results. In the analysis of the beluga’s side effect, in Experiment 1 for field of numerical cognition if studies adopt a broadly comparative Condition 1 she chose 61 right vs. 39 left (Binomial test: p > 0.05, approach that takes into account different sensory modalities used although in two sessions she was clearly biased to the right); in by different species. Condition 2 (opaque boxes), she chose 54 right vs. 46 (Binomial test: p > 0.05); and in Experiment 2 (item by item), she chose 42 Acknowledgements left vs. 58 right (Binomial test: p > 0.05). Finally, in the analysis of the dolphins’ side effect, in Experiment 1, Condition 1: Mila chose We are grateful to the directors of the L’Oceanografic aquar- 60 left vs. 40 right and Sharky chose 54 left vs. 46 right (Binomial ium of Valencia, and Zoo Aquarium of Madrid for allowing us to test: p > 0.05). conduct our research. Special thanks go for Miguel Bueno, Manuel Previous studies in various species have shown that the accu- Castellote and Jose Antonio Esteban for his help and support. We mulator model is a good candidate general mechanism responsible appreciate the work of the animal keepers from the three aquari- for the capacity to discriminate the larger of two quantities, as ums: Jean Chaperon, Sebastian, Miguel and Nicky from Marineland, in most cases performance has been limited by the total quan- Antibes, Zoo Aquarium of Madrid head coach Carlos de las Parras tity (magnitude) and the difference (disparity) of the comparisons for dolphin training and management and the trainers team; Cacha (human infants and adults: Barth et al., 2003; Cantlon and Brannon, Martínez, Arancha Sánchez, Berta Romeral, Pablo Juárez and Anto- 2006; Feigenson et al., 2002; Kobayashi et al., 2005; Uller et al., nio Martínez and finally from Oceanogràphic Valencia head coach 1999; Xu and Spelke, 2000; chimpanzees: Beran, 2001; Boysen and Luis Rizzoto for beluga training and management and the trainers Berntson, 1995; Boysen et al., 2001; gorillas: Anderson et al., 2005; team; Iván Soto, Felipe Cabildo and Almudena Mena. In addition orangutans: Anderson et al., 2007; Call, 2000; sea lions: Abramson we thank Luis Franco for his enormous help in the design and man- et al., 2011; parrots: Aïn et al., 2009; fishes: Gòmez-Laplaza and ufacturing of the apparatus used in this study. Thanks to Ma José Gerlai, 2011). In the present study, however, we found that with the Hernández Lloreda for her comments and statistical advice. Spe- exception of the successive condition where differences between cial thanks go to Titaví for her artwork. Research reported in this quantities might have some predictive power, the ratio between study was partly supported by grant CCG08-UCM/SAL-4007 (Uni- quantities was not a good predictor of performance. This means that versidad Complutense de Madrid y Comunidad de Madrid) to M.V. most of our results fail to support the accumulator model -a find- Hernández Lloreda and by grants UCM-BSCH GR58/08 (Universi- ing that has also been reported in studies with Asian elephants, for dad Complutense de Madrid y Banco Santander Central Hispano) whom no disparity effect has been established in tasks with visual and D/018712/08 (Agencia Espanola˜ de Cooperación Internacional and non-visual conditions (Irie-Sugimoto et al., 2009). Neverthe- y Desarrollo) to F. Colmenares. less, African elephants, unlike their Asian counterparts, do show ratio-dependent performance, which is consistent with the accu- References mulator model (Perdue et al., 2012). Similar results (i.e. breakdown effects), where performance did not conform to predictions from Abramson, J.Z., Hernández-Lloreda, V., Call, J., Colmenares, F., 2011. Relative quantity Weber’s law under all testing conditions, have been found in dogs judgments in South American sea lions (Otaria flavescens). Anim. Cogn. 14 (5), (Ward and Smuts, 2007) and in chimpanzees (Beran, 2004). 695–706. Agrillo, C., Dadda, M., Bisazza, A., 2007. Quantity discrimination in female mosquito Although the aim of the present set of experiments was not to fish. Anim. Cogn. 10, 63–70. analyze numerical abilities but the subjects’ competence to assess Aïn, S.A., Giret, N., Grand, M., Kreutzer, M., Bovet, D., 2009. The discrimination of and compare quantities under different presentation conditions, discrete and continuous amounts in African grey parrots (Psittacus erithacus). Anim. Cogn. 12, 145–154. this study does not preclude that the relative quantity judgments Alsop, B., Honig, W.K., 1991. Sequential stimuli and relative numerosity discrimina- reported were based on quantitative cues other than numerous- tions in pigeons. J. Exp. Psychol. Anim. Behav. Proc. 17, 386–395. ness. For example, factors such as surface area, in experiment 1, or Anderson, J.R., Awazu, S., Fujita, K., 2000. Can squirrel monkeys (Saimiri sciureus) dropping duration, in experiment 2, are confounded with numer- learn self-control? A study using food array selection tests and reverse-reward contingency. J. Exp. Psychol. Anim. Behav. Proc. 26, 87–97. ousness. In a strict sense, then, we cannot claim that the subjects Anderson, U.S., Stoinski, T.S., Bloomsmith, M.A., Maple, T.L., 2007. Relative numer- have the ability of number appreciation, even though the use of ousness judgment and summation in Young, middle-aged, and older adult these quantitative cues might, on average, yield the same results Orangutans (Pongo pygmaeus ablii and Pongo pygmaeus pygmaeus). J. Comp. Psychol. 121, 1–11. as would numerosity (Anderson et al., 2007; Mix et al., 2002). Anderson, U.S., Stoinski, T.S., Bloomsmith, M.A., Marr, M.J., Smith, A.D., Maple, T.L., Although controlling for those factors would allow a more precise 2005. Relative numerousness judgment and summation in young and old West- assessment of this potential numerical competence (e.g. Jaakkola ern lowland gorillas. J. Comp. Psychol. 119, 285–295. Au, W.W.L., 1993. The Sonar of Dolphins. Springer-Verlag, New York. et al., 2005), this would however be achieved at the expense of a Barth, H., Kanwisher, N., Spelke, E.S., 2003. The construction of large number repre- diminished ecological validity and of preventing proper compar- sentations in adults. Cogn. 86, 201–221. isons from being done with other studies that have used a similar Beran, M.J., 2001. Summation and numerousness judgments of sequentially presented sets of items by chimpanzees (Pan troglodytes). J. Comp. Psychol. 115, protocol (Uller, 2008). 181–191. In summary, this study provides new data on the abilities to Beran, M.J., 2004. Chimpanzees (Pan troglodytes) respond to nonvisible sets after make relative quantity judgments in two cetacean species, the one-by-one addition and removal of items. J. Comp. Psychol. 118, 25–36. Beran, M.J., 2007a. Rhesus monkeys (Macaca mulatta) enumerate large and small beluga and the bottlenose dolphin. Although subjects managed to sequentially presented sets of items using analog numerical representations. J. choose the larger of the two sets of quantities, thus confirming the Exp. Psychol. Anim. Behav. Proc. 33, 42–54. results obtained in other species, the cognitive mechanism sup- Beran, M.J., Beran, M.M., 2004. Chimpanzees remember the results of one-by-one porting this performance is not totally clear. In fact the data does addition of food items to sets over extended time periods. Psychol. Sci. 15, 94–99. Beran, M.J., Beran, M.M., Harris, E.H., Washburn, D.A., 2005. Ordinal judgments and not squarely fit either the accumulator or the object file model. summation of nonvisible sets of food items by two chimpanzees (Pan troglodytes) Further studies including more individuals in each species and a and a rhesus macaque (Macaca mulatta). J. Exp. Psychol. Anim. Behav. Proc. 31, larger variety of quantities are needed to establish the cognitive 351–362. Beran, M.J., 2007b. Rhesus monkeys (Macaca mulatta) succeed on computerized test mechanisms that underpin numerical competence and to deter- designed to assess conservation of discrete quantity. Anim. Cogn. 10, 37–45. mine the extent to which such mechanisms are shared by, or unique Bonanni, R., Natoli, E., Cafazzo, S., Valsecchi, P., 2011. Free-ranging dogs assess the to, different species. It will also be of interest to ascertain if there is quantity of opponents in intergroup conflicts. Anim. Cogn. 14, 103–115. Boysen, S.T., Bernston, G.G., Mukobi, K.L., 2001. Size matters: impact of item size and any kind of link between the sensory modality involved in numer- quantity on array choice by chimpanzees (Pan troglodytes). J. Comp. Psychol. 115, ical discrimination, for example, sight versus echolocation, and the 106–110. J.Z. Abramson et al. / Behavioural Processes 96 (2013) 11–19 19

Boysen, S.T., Berntson, G.G., 1989. Numerical competence in a chimpanzee (Pan Jaakkola, K., Fellner, W., Erb, L., Rodriguez, M., Guarino, E., 2005. Understanding of troglodytes). J. Comp. Psychol. 103, 23–31. the concept of numerically less by bottlenose dolphins (Tursiops truncatus). J. Boysen, S.T., Berntson, G.G., 1995. Responses to quantity: perceptual versus cognitive Comp. Psychol. 119, 286–303. mechanisms in chimpanzees (Pan troglodytes). J. Exp. Psychol. Anim. Behav. Proc. Kahneman, D., Treisman, A., 1984. Changing views of attention and automaticity. In: 21, 82–86. Parasuraman, R., Davies, D. (Eds.), Varieties of Attention. Academic Press, New Brannon, E.M., Terrace, H.S., 2002. The evolution and ontogeny of ordinal numerical York, pp. 29–61. ability. In: Bekoff, M., Allen, C., Burghardt, G.M. (Eds.), The Cognitive Animal: Kahneman, D., Treisman, A., Gibbs, B.J., 1992. The reviewing of object files: object- Empirical and Theoretical Perspectives on Animal Cognition. The MIT Press, specific integration of information. Cogn. Psychol. 24, 175–219. Cambridge, pp. 197–204. Killian, A., Yaman, S., von Fersen, L., Güntürkün, O., 2003. A bottlenose dolphin Brodie, P.F., 1989. The white whale (Delphinapterus leucas). In: Ridgway, S.H., Har- discriminates visual stimuli differing in numerosity. Learn. Behav. 31, 133–142. rison, R. (Eds.), Handbook of Marine Mammals, vol. 4. River Dolphins and the Kobayashi, T., Hiraki, K., Hasegawa, T., 2005. Auditory-visual intermodal matching Larger Toothed Whales. Academic Press, San Diego, pp. 119–144. of small numerosities in 6-month-old infants. Develop. Sci. 8 (5), 409–419. Butterworth, B., 1999. The Mathematical Brain. Macmillan, London. Matsuzawa, T., 2009. Symbolic representation of number in chimpanzees. Curr. Call, J., 2000. Estimating and operating on discrete quantities in orangutans (Pongo Opin. Neurobiol. 19 (1), 92–98. pygmaeus). J. Comp. Psychol. 114, 136–147. McComb, K., Packer, C., Pusey, A., 1994. Roaring and numerical assessment in con- Cantlon, J.F., Brannon, E.M., 2006. Shared system for ordering small and large num- tests between groups of female lions (Panthera leo). Anim. Behav. 47, 379–387. bers in monkeys and humans. Psychol. Sci. 17, 401–406. Meck, W.H., Church, R.M., 1983. A mode control model of counting and timing Cantlon, J.F., 2012. Math, monkeys, and the developing brain. Proc. Natl. Acad. Sci. processes. J. Exp. Psychol. Anim. Behav. Proc. 9, 320–334. U.S.A. 109, 10725–10732. Mitchell, R.W., Yao, P., Sherman, P.T., O’Regan, M., 1985. Discriminative respon- Carazo, P., Sanchez, E., Font, E., Desfilis, E., 2009. Quantity discrimination in Tenebrio ding of a dolphin (Tursiops truncatus) to differentially rewarded stimuli. J. Comp. molitor: evidence of numerosity discrimination in an invertebrate? Anim. Cogn. Psychol. 99, 218–225. 12, 463–470. Mix, K., Huttenlocher, J., Levine, S., 2002. Multiple cues for quantification in infancy: Carey, S., 2001. Cognitive foundations of arithmetic: evolution and ontogenesis. is number one of them? Psychol. Bull. 128, 278–294. Mind. Lang. 16, 37–55. Nowak, R.M., 1991. Beluga, or white whale. In: Baltimore, M.D. (Ed.), Walker’s Mam- Davis, H., 1984. Discrimination of the number three by a raccoon (Procyon lotor). mals of the World, vol. 2, 5th ed. Johns Hopkins University Press, pp. 1008–1010. Anim. Learn. Behav. 12, 409–413. Pack, A.A., Herman, L.M., 1995. Sensory integration in the bottlenosed dolphin: Davis, H., Perusse, R., 1988. Numerical competence in animals: definitional Immediate recognition of complex shapes across the senses of echolocation and issues, current evidence, and a new research agenda. Behav. Brain. Sci. 11, vision. J. Acoust. Soc. Am. 98, 722–733. 561–615. Perdue, B.M., Talbot, C.F., Stone, A., Beran, M.J., 2012. Putting the elephant back in Davis, H., MacKenzie, K.A., Morrison, S., 1989. Numerical discrimination by rats the herd: elephant relative quantity judgments match those of other species. (Rattus norvegicus) using body and vibrissal touch. J. Comp. Psychol. 103, Anim. Cogn., http://dx.doi.org/10.1007/s10071-012-0521-y. 45–53. Pepperberg, I.M., 1987. Evidence for conceptual quantitative abilities in the grey Defran, R.H., Pryor, K., 1980. The behavior and training of cetaceans in captivity. In: parrot: labeling of cardinal sets. Ethol. 75, 37–61. Herman, L.M. (Ed.), Cetacean Behavior: Mechanisms and Functions. John Wiley Pepperberg, I.M., 2006. Grey parrot (Psittacus erithacus) numerical abilities: addition and Sons, New York, NY, pp. 319–362. and further experiments on a zero-like concept. J. Comp. Psychol. 120, 1–11. Dehaene, S., 1997. The Number Sense: How the Mind Creates Mathematics. Oxford Rugani, R., Regolin, L., Vallortigara, G., 2007. Rudimental numerical competence in University Press, New York. 5-day-old domestic chicks (Gallus gallus): identification of ordinal position. J. Dehaene, S., 2001. The Cognitive Neuroscience of Consciousness. MIT Press, Exp. Psychol. Anim. Behav. Proc. 33, 21–31. Cambridge. Rumbaugh, D.M., Savage-Rumbaugh, S., Hegel, M.T., 1987. Summation in the chim- Dooley, G.B., Gill, T.V., 1977. Acquisition and use of mathematical skills panzee (Pan troglodytes). J. Exp. Psychol. Anim. Behav. Proc. 13, 107–115. by a linguistic chimpanzee. In: Rumbaugh, D.M. (Ed.), Language Learn- Samuels, A., Tyack, P.L., 2000. Flukeprints: a history of studying cetacean societies. ing by a Chimpanzee: The Lana Project. Academic Press, New York, In: Mann, J., Connor, R.C., Tyack, P.L., Whitehead, H. (Eds.), Cetacean Societies: pp. 247–260. Field Studies of Dolphins and Whales. Academic Press, San Diego, pp. 9–44. Evans, T.A., Beran, M.J., Harris, E.H., Rice, D., 2009. Quantity judgments of sequen- Shettleworth, S., 2010. Cognition, Evolution, and Behaviour. Oxford University Press, tially presented food items by capuchin monkeys (Cebus apella). Anim. Cogn. 1, Oxford. 97–105. Silberberg, A., Fujita, K., 1996. Pointing at smaller food amounts in an analogue of Feigenson, L., Carey, S., Hauser, M., 2002. The representations underlying infants’ Boysen and Berntson’s (1995) procedure. J. Exp. Anal. Behav. 66, 143–147. choice of more: object files versus analog magnitudes. Psychol. Sci. 13, 150–156. Simon, T.J., Hespos, S.J., Rochat, P., 1995. Do infants understand simple arithmetic? Gallistel, C.R., 1990. The Organization of Learning. MIT Press, Cambridge. A replication of Wynn (1992). Cogn. Dev. 10, 253–269. Gallistel, C., Gelman, R., 2000. Non-verbal numerical cognition: from reals to inte- Smirnova, A.A., Lazareva, O.F., Zorina, Z.A., 2000. Use of number by crows: investi- gers. Trends Cogn. Sci. 4, 59–65. gation by matching and oddity learning. J. Exp. Anal. Behav. 73, 163–176. Gòmez-Laplaza, L.M., Gerlai, R., 2011. Spontaneous discrimination of small quan- Thomas, R.K., 1992. Primates’ conceptual use of number: Ecological perspectives tities: shoaling preferences in angelfish (Pterophyllum scalare). Anim. Cogn. 14, and psychological processes. In: Paper presented at the 13th Congress of the 565–574. International Primatological Society, Tokyo, Japan. Hanus, D., Call, J., 2007. Discrete quantity judgments in the great apes (Pan paniscus, Uller, C., 2008. Developmental and evolutionary considerations on numerical cog- Pan troglodytes, Gorilla gorilla, Pongo pygmaeus): the effect of presenting whole nition: a review. J. Evol. Psychol. 6 (4), 237–253. sets versus item-by-item. J. Comp. Psychol. 121, 241–249. Uller, C., Lewis, J., 2009. Horses (Equus caballus) select the greater of two quantities Hauser, M.D., Carey, S., 1998. Building a cognitive creature from a set of primitives: in small numerical contrasts. Anim. Cogn. 12 (5), 733–738. evolutionary and developmental insights. In: Cummins, D., Allen, C. (Eds.), The Uller, C., Carey, S., Huntley-Fenner, G., Klatt, L., 1999. What representations might Evolution of Mind. Oxford University Press, Oxford, England, pp. 51–106. underlie infant numerical knowledge? Cogn. Dev. 14, 1–36. Hauser, M.D., Carey, S., 2003. Spontaneous representations of small numbers of Uller, C., Jaeger, R., Guidry, G., Martin, C., 2003. Salamanders (Plethodon cinereus)go objects by rhesus macaques: examinations of content and format. Cogn. Psychol. for more: rudiments of number in an amphibian. Anim. Cogn. 6, 105–112. 47, 367–401. Ward, C., Smuts, B., 2007. Quantity-based judgments in the domestic dog (Canis Hauser, M.D., Carey, S., Hauser, L., 2000. Spontaneous number representation in lupus familiaris). Anim. Cogn. 10, 71–80. semi-free-ranging rhesus monkeys. Proc. Royal. Soc. B 267, 829–833. Wiese, H., 2003. Iconic and non-iconic stages in number development: the role of Irie, N., 2012. Numerical cognition of elephants: relative quantity judgment by language. Trends Cogn. Sci. 7, 385–390. an African Forest Elephant (Loxodonta cyclotis) and an African Savannah Ele- Wood, J., Hauser, M., Glynn, D., Barner, D., 2008. Free-ranging rhesus monkeys phant (Loxodonta Africana). In: Aranovich, M., Dufresne, O. (Eds.), Elephants’s spontaneously individuate and enumerate small numbers of non-solid portions. Ecology, Behavior and Conservation. Nova Science Publisher, Inc., New York, Cogn. 106, 207–221. pp. 145–151. Wynn, K., 1998. Psychological foundations of number: numerical in human infants. Irie-Sugimoto, N., Kobayashi, T., Sato, T., Hasegawa, T., 2009. Relative quan- Trends Cogn. Sci. 2, 296–303. tity judgment by Asian Elephants (Elephas maximus). Anim. Cogn. 12 (1), Xu, F., Spelke, E.S., 2000. Large number discrimination in 6-month-old infants. Cogn. 193–199. 74, 1–11.