Animal Behaviour 80 (2010) 783e789 Contents lists available at ScienceDirect Animal Behaviour journal homepage: www.elsevier.com/locate/anbehav Kea, Nestor notabilis, produce dynamic relationships between objects in a second-order tool use task Alice M.I. Auersperg a,b,*, Gyula K. Gajdon a,b, Ludwig Huber a a Department of Neurobiology and Cognition Research, University of Vienna b Konrad Lorenz Institute for Ethology, Austria article info Studies on advanced forms of tool use in birds have mainly been concentrated on corvids. In this study, Article history: captive kea, a neophilic New Zealand parrot species, produced different orders of spatial object rela- Received 10 May 2010 tionships in a tube-lifting/object-inserting paradigm. Hence, we found that kea, which are neither Initial acceptance 6 July 2010 natural tool users nor nest builders, could readily solve a second-order tool use task. They also learned to Final acceptance 9 August 2010 produce highly complicated meansemeanseend sequences in a short period of time. Available online 19 September 2010 Ó 2010 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved. MS. number: 10-00321 Keywords: kea meanseend Nestor notabilis physical cognition tool use Creating sophisticated object relationships to achieve a goal has Secondary tool use has rarely been observed in nonhuman been considered to be one of the defining features of intelligent animals other than primates. There is, however, increasing evidence species (van Schaik et al. 1999; Reader & Laland 2002). The more of elaborate consideration of object relationships in two corvid complicated object relationships created by animals often involve species. The New Caledonian crow, Corvus moneduloides, which the use of tools as a means to reach a desired item. Tool use was routinely uses tools within its natural environment, has recently defined by Beck (1980, page 10) as ‘the external employment of an shown secondary tool manipulation in the laboratory in the form of unattached environmental object to alter more efficiently the form, meta-tool use (Taylor et al. 2007; Wimpenny et al. 2009). There is position, or condition of another object, another organism, or the also evidence that these birds have some causal understanding user itself when the user holds or carries the tool during or just about the effects that objects may have on one another (von Bayern prior to use and is responsible for the proper and effective orien- et al. 2009; Taylor et al. 2009). Recently, there have also even been tation of the tool’. Tool use competences can be categorized in impressive demonstrations of first- and second-order tool use in terms of levels of complexity: Fragaszy et al. (2004a) described the a nontool-using corvid, the rook, Corvus frugilegus (Bird & Emery relationships produced when a subject acts on an object and an 2009). To our knowledge, there has been no investigation of action on a second object occurs by default as zero-order relations, secondary tool use in any bird species other than corvids. as, for example, pulling a piece of cloth with a food reward resting There is currently debate on whether tool use stimulated the upon it (Auersperg et al. 2009). Acting with object A in relation to evolution of intelligence in tool-using species or whether it is more object B (for example pounding a nut with a stone) is classified as likely to be a by-product of general-purpose intelligence (Bird & a first-order relation. Finally, acting with object A in relation to Emery 2009). It has also been argued that most birds are object B following a placement of object B in relation to a third routinely establishing more complicated object relationships object C would be a second-order relationship. Wimpenny et al. during nest building than during tool use and that tool use might (2009, page 2) also described ‘secondary tool use’ as ‘using one merely have evolved to replace morphological adaptations (Hansell tool on another object to access it or modify it for the use as a tool’. & Ruxton 2008; Kacelnik 2009). It would therefore be interesting to test complex levels of tool use in species that lack sophisticated nest construction. * Correspondence: A. M. I. Auersperg, Department of Neurobiology and Cognition Research, University of Vienna, Althanstrasse 14, 1090 Vienna, Austria. The kea is a parrot, resident in the mountain region of New E-mail address: [email protected] (A.M.I. Auersperg). Zealand. This species is not known to build complex nest cups but 0003-3472/$38.00 Ó 2010 The Association for the Study of Animal Behaviour. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.anbehav.2010.08.007 784 A.M.I. Auersperg et al. / Animal Behaviour 80 (2010) 783e789 to breed in simple burrows (Jackson 1963). As a consequence of Table 1 a scarcity of predators and a seasonality of food availability, kea are The experimental conditions highly neophilic. Similar to other destructive foragers, such as the Condition Correct sequence Pc Trial no./ Session No. of Reward New World capuchin monkeys, Cebus libidinosus, which use tools in of actions session no. stoppers pierced the laboratory as well as in the wild (Fragaszy et al. 2004a, b), kea TL Choose correct tube 0.5 4 5 0 No display an extremely strong urge to manipulate and dismember TLA Choose correct tube; 0.5 2 15 2 No novel objects (Diamond & Bond 1999, 2004). This ‘haptic neophilia’ choose correct end TLB Choose correct end 0.5 2 15 2 No may expose the animals to functional object characteristics that are TLC Choose correct tube; 0.25 2 15 3 No unavailable to more neophobic species. It is likely that such choose correct end behaviour enhances innovativeness and facilitates learning. TLD Do not touch tubes d 2153No Observations of kea in the laboratory have shown remarkable SB Select correct object; 0.5 8 5 0 Yes insert into correct tube; learning capacities about their physical environment (Werdenich & lift at the same side the Huber 2006). They also show intense combinatorial activity during object was inserted object play, in particular when inserting objects into hollow spaces SBA Choose correct tube; 0.5 2 15 2 Yes (G. K. Gajdon, M. Lichtenegger & L. Huber, unpublished data). Since insert at correct end; combinatorial actions may allow access to otherwise inaccessible lift same end; lift opposite end resources and require an ability to coordinate objects relative to SBB Insert at correct end; 0.5 2 15 2 Yes each other, they can be considered as precursors of tool use in lift same end; lift nontool-using species (Fragaszy et al. 2004a). opposite end To examine the keas’ sensitivity to different orders of object SBC Choose correct tube; 0.25 2 15 3 Yes insert at correct end; relationships, we devised an experimental set-up in which subjects lift same end; lift had to establish increasingly complicated levels of spatial rela- opposite end tionships between objects by themselves. Our subjects had previ- Tube lifting (TL), side-restricted tube-lifting (TLAeD), spaghetti-breaking (SB) and ously participated in an ‘inserting as means paradigm’ (G. K. Gajdon side-restricted spaghetti-breaking tasks (SBAeC); P hypothesized mean; trial c ¼ et al., unpublished data), in which they had learned to drop the no./session number of trials per session; session no. number of sessions; no. of ¼ ¼ stoppers the total number of stoppers on both of the two tubes; reward pier- smaller of two differently sized cubes into an opaque tube. The ¼ ced the state of the reward (pierced by a piece of spaghetti or free inside the tube). opaque tube was firmly fixed in a vertically slanted position and ¼ contained a reward, which was pasted onto the middle of the tube floor using cream cheese. When an object was inserted into the and vegetables as well as a daily protein source (boiled eggs, upper end of the tube it would knock against the reward and cause cottage cheese with yogurt, corn or minced beef). The amount of it to fall out at the lower end of the tube. For the following exper- food was balanced weekly to keep the birds at their free feeding iment, we used a similar design but the tubes were lying on the weights. Fresh drinking water was available ad libitum. The ground without fixation. subjects were visually isolated from their groupmates, by an opa- In the first ‘tube-lifting’ paradigm, the reward was lying freely que sliding wall, within an experimental compartment (5 10 m) Â on the tube floor and the tube had to be lifted at one end to release during testing. All subjects that participated in the described it. The object relations produced here are zero-order relations experiments were housed in accordance with the Austrian Federal (Fragaszy et al. 2004a). The following ‘spaghetti-breaking’ para- Act on the Protection of Animals. Furthermore, as all kea studies digm investigated whether the animals could combine their tool were strictly noninvasive and based purely on behavioural tests, size expertise with their new experience from the tube-lifting they were classified as nonanimal experiments in accordance with paradigm to master a second-order tool use task. Here, the reward the Austrian Animal Experiments Act. was pinned inside the tube using uncooked spaghetti and could only be attained by inserting the appropriate object (a ball) and TUBE LIFTING thereafter lifting the tube at the side in which the object was inserted to break the spaghetti and release the reward. The object Task 1: Tube Lifting relationships produced in this case are second-order relationships (act with an object A (tube) in relation to an object B (ball) Methods following a placement of object B in relation to a third object A consolidation phase was designed to control whether the C (reward); see Fragaszy et al.