Behavioral persistence in captive bears: a response to Criswell and Galbreath Sophie S. Vickery' and Georgia J. Mason2 behavior [Odberg 1978, Mason 1991]) of captive animals can have detrimental correlates. They have AnimalBehaviour Research Group, been associated with reduced behavioral diversity Departmentof Zoology, OxfordUniversity, (Stolba et al. 1983, Shepherdsonet al. 1993, Swaisgood South Parks Road, OxfordOX1 3PS, UK et al. 2001), decreasedresponsiveness to environmental stimuli (Fentress 1977, Ormrod 1992, Langenhorst Ursus 16(2):274-279 (2005) 1998, Clubb 2001), reducedexploratory behavior (Lien and Klopfer 1978), poor maternalcompetence (Mason et al. 1995), injuryand reduced body condition(Sambraus 1985, Fraser and Broom 1990, Moon-Fanelli and This paper responds to criticisms raised by Criswell Dodman 1998), and reduced reproductive success and Galbreath(2005) regardinganalyses we used in an (Shepherdson 1994). Comparedto their non-stereotyp- earlierpaper (Vickery and Mason 2003), and shows that ic counterparts, stereotypic animals may also have their criticisms are unfounded. However, first, we altered neurochemistry, for example showing alter- present some backgroundto the original work. Vickery ations in dopamine (Dantzer 1986), serotonin (Vande- and Mason (2003), written toward the end of a long- broek et al. 1995), and opioid (Zanella et al. 1996, term study of captive bears, was timed to allow for Zanella and Mason 1997) functioning. Furthermore, presentationin 2002. One of us (SSV) was based at recent work on bank voles (Clethrionomysglareolus), Banglamung Wildlife Breeding Station in Chonburi mice (Mus musculus), parrots (Amazona amazonica), province, Thailand, a sanctuary which individually and 2 species of songbird (Parus caeruleus and P. housed Asiatic black bears (Ursus thibetanus) and palustris) and rhesus monkeys (Macaca mulatta) has Malayan sun bears (Helarctos malayanus) in virtually shown that stereotypic animals are more 'persevera- identical conditions, providing a unique opportunityto tive', i.e. more behaviorally persistent than are their investigatebears' behavioralresponses to captivity. The less stereotypic conspecifics (Gamer 1999, Garnerand behaviorof 29 bears was observed over 2.5 years. This Mason 2002, Gamer et al. 2003a,b, Lutz et al. 2004, was the total numberof individualsavailable for study, Latham 2005). Our aim was to test for a similar barring those that were sick or whose records were correlation in bears, to investigate the generality of incomplete. Of these, 21 were put through learning these findings. For species such as these, such an tasks to test the hypothesis that their stereotypy effect could also have conservation implications, stemmed from a general inability to switch behaviors potentially helping to explain the low reintroduction appropriately.These learning tasks were time-consum- success currentlydisplayed by captive-bredanimals. In ing to run, and thus it was not possible to test all 29 reintroductionprograms, captive-bred animals general- bears. Twelve bears were tested in Year 1 (the data ly fare far worse than translocatedwild conspecifics presentedin Vickery and Mason [2003]), and a further9 (Griffith et al. 1989, Ginsberg 1994, Frantzen et al. bears in Year 2. Both years' data are presented in 2001), and this certainly seems true for bears (Alt and Vickery and Mason (2005). In addition, the properties Beecham 1984, Stiver et al. 1997, Eastridgeand Clark of all 29 bears' stereotypieswere analyzed for effects of 2001, Clark et al. 2002). For example, Stiver et al. time in captivity, sex, and species (presentedbriefly in (1997) reportedthat of 23 pen-rearedAmerican black Vickery and Mason [2003], and more fully in Vickery bears (Ursus americanus) released in east Tennessee and Mason [2004]). between 1982 and 1995, 6 were known to die, 5 It is well recognized that the stereotypies(repetitive, caused nuisance problems, and 2 were killed illegally invariantand seemingly functionlessforms of abnormal by humans. The 2 main findingspresented in Vickery and Mason (2003) were: (1) a positive correlation between an 1Presentaddress: Animal Welfare Division, Department for individual's stereotypy frequency and the number of Environment,Food and Rural, Affairs, 1A Page Street, trials for which it continuedto respond without reward LondonSW1P 4PQ,UK; [email protected] (a measureof behavioralpersistence), and (2) a positive 2Presentaddress: Centre for the Studyof AnimalWelfare, correlationbetween stereotypyfrequency and time spent Departmentof Animal and PoultryScience, Universityof we divide Criswell and Galbreath's Guelph,Ontario N1G 2W1, Canada;[email protected] in captivity. Here, 274 BEHAVIORALPERSISTENCE: RESPONSE 275 criticisms of our work into 2 categories:those aimed at Statistical analyses our general scientific approach,and those aimed at our We categorize Criswell and Galbreath's statistical specific statisticalanalysis. criticisms into those that are wrong and those that are right but unimportant.First, we deal with issues about which they are simply wrong. Criswell and Galbreath Our aims and suggest that we would have obtaineda betterfit had we general chosen In S' (mean stereotypylevels across up to 6 ob- scientific approach servationperiods between June 2000 and May 2002) in we and did we use Why did do this work, why the place of In S (stereotypy levels measured immediately we did? One of Criswell and Galbreath'swor- subjects prior to task presentation).Although we appreciatethe ries is that "It was never made clear ... 29 bears why" suggestion, this is irrelevantbecause these 2 stereotypy were used in one and 12 in another.The analysis impli- measures are far from interchangeable.Levels of ste- cation is that we selected a non-random subset of reotypy can vary with season, time in captivity,environ- animals from which it would be to clearly illegitimate mentalchange andother factors, and an individual'slevel further. However, we did no such extrapolate thing. of extinction responding might also vary over time This was an in which we opportunisticstudy simply (Garner1999, Garneret al. 2003a). Thus, when inves- used all the animals that were available to us and tigatinga relationshipbetween stereotypyand extinction, collected as much data as possible within logistic it is sensible to measure these 2 variables as closely constraints. For example, we ran the learning experi- togetheras possible. The measureof stereotypydenoted ments over 2 years because so few bears could be S was used in our extinction analysis (Criswell and investigated simultaneously. Galbreath[2005] Eq. 1) because this was taken imme- Second, Criswell and Galbreath claim that we diately priorto task presentation.However, for the same "lumped all the bears available" into a single treatment reason - because an individual's stereotypyfrequency group, thereby foregoing the opportunity to study a might differover the courseof a year- when looking for control group of bears. However, in fact, there was a relationshipbetween averagestereotypy frequency and no appropriatecontrol group, because all the bears that the number of years spent in captivity, it is most could be studied (i.e. those that were single-housed) appropriateto assess on several were housed in near-identicalcages and all except 2 stereotypy frequency occasions the and calculate an overall exhibited stereotypy. It was for this reason that our throughout year mean. Thus the measure used in our 'time in study was correlational. Third, they claim that our captivity' and Galbreath S', "... inference would have been strongerhad [we] used analysis (Criswell [2005] Eq. 2), a mean of measured an experimentaldesign in which those bears that had represents stereotypy frequencies over to 6 observation previously been removed from enriched enclosures to up periods. claim that it would be small cages were compared to bears lucky enough to Second, they impossible to datasets remain behind in the enclosures." Here, we emphasize properlylog-transform containingzeros. In fact, this is that we did not move the animals; ratherour subjects problem commonly solved in practiceby adding 1 to were bears that had been housed in cages long before each value prior to transformation(Martin and we arrived. Bateson 1993:151). Finally, they incorrectly imply that we condemn We now respond to statistical critiques that are reintroductionas a conservationtool. For example, they technically correct but are ultimately unimportantfor assert, "These findingslead Vickery and Mason, in turn, inference in this study. Criswell and Galbreath(2005) to question the potential success of reintroduction first claim that we reversedthe "directionof causality". programs with bears held captive for substantialtime Our use of stereotypy as the dependent variable was periods". In fact, there was alreadyample evidence that appropriatebecause we were suggesting stereotypy to reintroductionprograms involving captive-bred bears, be a productof underlyingchanges that increasegeneral and indeed captive-bredanimals of a range of species, behavioralpersistence (see also Garnerand Mason 2002). are less successful than those involving their wild However, as Criswell and Galbreath(2005) point out conspecifics. Our paper simply sought to raise and test themselves, "in a correlationalstudy, that switch is not so ideas that might explain why this is so, with the aim of serious." Our study was indeed exactly that-a correla- helping biologists improve reintroduction success