At least one individual of each species discovered all four available methods. This proves that in principle, the affordances of the tasks lay within the cognitive and physical capacity of both species.
There were, however, interesting differences in performance with the present set of tasks. The kea were faster in discovering multiple solutions and showed more individual variation than the naturally tool-using NCCs. Within the first session, all kea successfully employed at least two or three solutions while none of the crows used more than one. The kea also switched to other solutions quicker once previously mastered solutions were blocked. Although both species had experience with compact objects 
, using the ball was acquired faster by the kea, while in the stick option the naturally stick-tool using NCCs were faster. Only one kea succeeded in inserting the stick tool into the correct opening, although all attempted to do so.
To a large extent, differences in exploration patterns and affordance learning as well the balance between neophilia/neophobia seem to be responsible for the differential performance. The kea showed more haptic exploration while the crows, probably due to their higher level of neophobia, seemed to explore more in a visually guided manner. Similar differences exist between kea and common ravens 
. The kea’s higher readiness to manipulate i.e. act on novel objects, may help them to detect functional affordances. In their naturally low-risk, variable environment, neophilia may reflect low predation risk 
. The crows, in contrast, approached the apparatus hesitantly, and touched it less often than the kea. Two crows never explored the box thoroughly with their beaks. One of them had to be excluded from the study because it never approached the experimental setup. Neophobia hampered the crows’ performance in other respects. For instance, despite their predisposition, experience and competence for stick tool use, the crows did not use the available tools as first option and also took some time to use tools after the string pulling option had been removed. (The string option was the first to be blocked for both species. It may have represented one of the more conspicuous affordances of the apparatus or evoked little neophobia since both species were familiar with strings; see methods
). It is possible that because the available sticks were smoother, painted, and considerably thicker than those the crows naturally choose in the wild 
ergonomic difficulties synergised with neophobia, and may have affected their performance. Once the crows had established stick tool use, they continued to use the sticks for exploring the box instead of touching it directly. This is reflected in the number of ITA with sticks when confronted with the ball and window option, i.e. after the stick option was blocked. This may indicate a preference to explore unfamiliar objects with tools rather than touching them directly, as shown by Wimpenny et al. (2010) 
. Such behaviour is, however, disadvantageous in situations where the functionalities of the objects are not detectable by simply applying pressure, such as in the case of the window option, illustrating one cost of the otherwise useful capacity to use tools for exploratory goals.
Another strong difference was that the kea showed greater destructiveness than the crows, as a consequence of their forceful and frequent use of pulling and tearing actions. One kea, Luke, even broke the Plexiglas® on the top of the box while trying to force it open, and most kea attempted to turn over the MAB, which had to be fixed to the aviary floor. Wild kea are well-known for what human observers usually describe as curiosity, playfulness and urge to tear apart objects such as cars’ windshield wipers or picnic baskets 
Wild (as well as naïve captive) NCCs also use tearing actions when manufacturing tools from pandanus leaves 
. They however did not use such behaviours in this setup. We can at this point speculate that tearing behaviours may mainly be orientated towards tool making and nest building rather than to exploration of novel objects or food extraction. The slight curvature at the beak tip of most corvids has been interpreted as an adaptation to feeding on carcasses 
. NCCs might have secondarily lost their beak curvature making their beaks less suited for violent tearing actions. Very recently Rutz et al. (2010) 
were able to show that a substantial amount of the crows’ protein and lipid intake came from wood boring beetle larvae obtained with stick tools, also indicating that explorative foraging is more concentrated on probing actions. It seems therefore possible that the range of exploration techniques during foraging in NCC may be constrained by the adaptive specialization for tool use and that this will affect the “zone of latent solutions” 
within the species' cognitive repertoire.
Our results also provide the first experimental evidence of stick tool use in a parrot. Kea are neither natural tool users like New Caledonian crows, nor do they construct nests with twigs as all corvids do, thus lacking the predisposition of nest builders for handling twigs and other elongated stick-like objects 
. Instead, kea use or dig burrows for laying their eggs 
. Also importantly, ergonomically, the use of sticks is clearly difficult for kea. The curvature of their beaks and pronounced size difference between upper and lower beak, precludes a good grip and control of long, straight tools. NCCs, maybe as an adaptation to tool use 
, have short straight beaks with the mandible almost as long as the maxilla, allowing them to effectively hold sticks directly forwards, functionally elongating their beaks and increasing their reach.
To overcome these difficulties, the single successful individual kea developed a complicated stepwise technique, involving carefully concerted foot and bill actions (see Movie S1
). This permitted the subject to insert and direct a stick tool despite the species’ morphological constraints. Kermit’s performance indicates a high degree of deliberate control over his movements, suggestive of anticipation of their effect and perhaps a representation of the goal action, i.e. of inserting the stick into the opening. Proof of goal directedness or goal representation, however, requires specific tests that were not implemented here 
Our study illustrates the difficulties of comparative cognition research and points to some partial solutions. Clearly, no single-task exploration can be used to assess problem-solving ability or make claims for advanced general intelligence or innovativeness. This caveat applies to within as well as between species comparisons. Problem solving is intrinsically multi-dimensional and it is to be expected that individuals or species will outperform each other in different dimensions. Batteries of tasks designed with this in mind may however be highly informative about the different predispositions and cognitive competences across individuals and species.