Ravens, New Caledonian crows and jackdaws parallel great apes in motor self-regulation despite smaller brains

Overriding motor impulses instigated by salient perceptual stimuli represent a fundamental inhibitory skill. Such motor self-regulation facilitates more rational behaviour, as it brings economy into the bodily interaction with the physical and social world. It also underlies certain complex cognitive processes including decision making. Recently, MacLean et al. (MacLean et al. 2014 Proc. Natl Acad. Sci. USA 111, 2140–2148. (doi:10.1073/pnas.1323533111)) conducted a large-scale study involving 36 species, comparing motor self-regulation across taxa. They concluded that absolute brain size predicts level of performance. The great apes were most successful. Only a few of the species tested were birds. Given birds' small brain size—in absolute terms—yet flexible behaviour, their motor self-regulation calls for closer study. Corvids exhibit some of the largest relative avian brain sizes—although small in absolute measure—as well as the most flexible cognition in the animal kingdom. We therefore tested ravens, New Caledonian crows and jackdaws in the so-called cylinder task. We found performance indistinguishable from that of great apes despite the much smaller brains. We found both absolute and relative brain volume to be a reliable predictor of performance within Aves. The complex cognition of corvids is often likened to that of great apes; our results show further that they share similar fundamental cognitive mechanisms.

Overriding motor impulses instigated by salient perceptual stimuli represent a fundamental inhibitory skill. Such motor self-regulation facilitates more rational behaviour, as it brings economy into the bodily interaction with the physical and social world. It also underlies certain complex cognitive processes including decision making. Recently, MacLean et al. (doi:10.1073/pnas.1323533111)) conducted a large-scale study involving 36 species, comparing motor self-regulation across taxa. They concluded that absolute brain size predicts level of performance. The great apes were most successful. Only a few of the species tested were birds. Given birds' small brain size-in absolute terms-yet flexible behaviour, their motor self-regulation calls for closer study. Corvids exhibit some of the largest relative avian brain sizes-although small in absolute measure-as well as the most flexible cognition in the animal kingdom. We therefore tested ravens, New Caledonian crows and jackdaws in the so-called cylinder task. We found performance indistinguishable from that of great apes despite the much smaller brains. We found both absolute and relative brain volume to be a reliable predictor of performance within Aves. The complex cognition of corvids is often likened to that of great apes; our results show further that they share similar fundamental cognitive mechanisms. We did not use the A-not-B task for two main reasons: it is not clearly a motor self-regulatory task, and MacLean et al's study had methodological shortcomings regarding its use. Solving the task is an exceedingly dynamic process affected by such contextual features as the delay between hide and search, the visual distinctiveness-as well as number-of containers, and the level of attention directed towards the baited container [18,19]. Given this dynamic complexity, the A-not-B task has been widely questioned as an inhibitory task [18]. In the same vein, Jelbert and colleagues recently showed that New Caledonian crows who received training in tracking human hand actions were significantly better at solving the A-not-B task than those who were not trained [20].
MacLean et al. also implemented the task unsatisfactorily. Firstly, they conducted only a single trial per individual, compared to 10 trials per individual in the cylinder task. Inhibitory control is not a binary skill but a continuum-as shown in the cylinder task. Secondly, they used three containers, of which the middle container was never rewarded. As we understand it, every failure to retrieve the reward-including searching the middle container-was regarded as inhibitory failure. The middle location should not be regarded as an inhibition failure as no habit was formed in relation to it. Thirdly, as many as nine of the species examined could not be tested in the A-not-B task. Only four of the seven bird species received the A-not-B task. Overall the task does not provide much informative data.
We therefore replicated only the cylinder task conducted by MacLean and co-workers [9], comparing the performance on it in the same species, with particular focus on the birds. It must be noted, however, that MacLean and co-workers' findings regarding the correlation between absolute and relative brain size holds for this task alone. In our replication, we followed the experimental procedures described in MacLean et al.

Subjects and housing
Five adult ravens (three females), 10 adult New Caledonian crows (four females) and 10 adult jackdaws (five females) participated in the study. The ravens were tested at Lund University Corvid Cognition Station in Sweden, the other birds at the Avian Cognition Research Station associated with the Max Planck Institute for Ornithology, Seewiesen, Germany. All were housed in environmentally enriched outdoor aviaries. The New Caledonian crows had access to heated indoor compartments (6 m 2 ). The ravens and jackdaws were housed as a social group in 400 m 2 (ravens) and 120 m 2 (jackdaws) spaces, and the New Caledonian crows as pairs in 18-32 m 2 spaces. All birds had ad libitum access to food and water. Subjects were tested individually in familiar testing compartments.

Set-up and materials
The cylinder task employed two apparatuses: an opaque and a transparent hollow cylinder. The cylinders were open on both sides and attached to a wooden base. The openings allowed the subjects to insert their heads and retrieve a highly desirable food reward from the middle of the cylinder.

Procedure
First the birds were habituated to the opaque cylinder by having it placed inside the aviary. A bird was defined as habituated when it interacted with the cylinder. All subjects except two jackdaws and one raven habituated after 2 days. It took 3 days for two jackdaws and 6 days for one raven to habituate. Familiarization trials ensued 1 day after the habituation. In these trials, the birds were familiarized with the opaque cylinder and its particular shape. The experimenter hid the reward at the centre of the tube while the subject observed; then the subject was allowed to retrieve the reward. Correct responses were coded, for the first responses, as reaching through either opening to get the reward. Regardless of their first response, subjects were eventually allowed to obtain the reward. To proceed to the test trials, subjects had to succeed on four out of five consecutive trials. All birds completed the familiarization trials within one session in 1 day and they did not have any incorrect responses. Testing trials followed 1 day after the familiarization trials.
In the test trials, a transparent cylinder replaced the opaque cylinder. Ten trials were administered for every subject. The baiting procedure was identical to that used in the familiarization trials. Correct responses were coded, again for the first responses, when one of the openings was used to retrieve the reward (detour through the side), and an incorrect response was coded if physical contact was made   [9]. The right table shows new top 10 ranking when including the performance of the Corvus species in the current study (shown in bold).  Table 2. Summary of the two final models that investigated the effect of absolute brain volume, residual brain volume and trial number on the cylinder task performance for 10 bird species tested in this study and MacLean et al. [9]. with the front of the cylinder (probably in an attempt to directly reach for the reward). Subjects were eventually allowed to retrieve the reward regardless of the accuracy of their first responses.
To perform detailed comparison, we obtained unpublished trial-level results on the seven bird species from MacLean et al., along with endocranial brain volume and body mass data from the literature concerning the three Corvus species [21][22][23] analysed together with the mass and volume data for the other seven bird species reported in the replicated study (MacLean et al. supporting information table S1). Two generalized linear mixed-effect models (GLMM) were constructed to investigate the effect of absolute and residual brain volume on performance, with the latter calculated as the residual from a phylogenetic regression of absolute-brain-volume predicted body mass. For both models, trial number was added as a fixed effect. Both absolute and residual brain volume as well as trial number were found to be significant predictors of performance on the cylinder task (table 2). Comparison of the two models using the Akaike information criterion (AIC) provided stronger support for the absolute brain volume model (table 2). In addition, we investigated the effect of the trial number on performance separately for all bird species tested. We found a trial effect on performance in four of the bird species examined by MacLean et al., with no such effect in the Corvus species we examined. Performance improved significantly over successive trials for the zebra finches (p = 0.037), song sparrows (p = 0.014), swamp sparrows (p = 0.008) and orange amazons (p = 0.029; electronic supplementary material, table S1). The relationship between the absolute brain size and the cylinder task performance for 10 bird species tested in this study and MacLean et al. [9]. The trend line is based on a species mean percentage score on the cylinder task.

Discussion
The Corvus species performed on a similar level to the great apes, despite vastly smaller absolute brain sizes. A chimpanzee brain is roughly 26 times larger than a raven's; nevertheless, both species achieve 100% success. The jackdaws were more successful than either the bonobos or gorillas, despite a brain 70-94 times smaller. Clearly, absolute brain size is no overall predictor of motor self-regulation across a wider range of animal taxa. However, as among primates, absolute brain size does appear to be a significant predictor across bird species (figure 1), but on the other hand relative brain size is as well a significant predictor within birds.
Recent findings indicate the existence of different scaling relationships between total number of neurons and absolute brain size across mammalian taxa [24]: bigger brains do not necessarily contain more neurons than smaller ones. Neuronal numbers might perhaps better explain corvids' and apes' similar performance in the cylinder task than absolute or relative brain size, but such investigations have to await studies on neuronal counts in bird brains. Furthermore, the crude comparisons made both in our study and MacLean et al.'s may miss the importance of the relative or absolute sizes of certain brain regions related to inhibitory skills. The lack of data on different brain regions across bird species meant that we could not perform the requisite analysis.
Regardless of any comparisons of various brain measurements, one can conclude that the similarities between corvids and great apes in complex cognitive skills [11] also hold for at least one less taxing skill: motor self-regulation.
Some of the bird results from the cylinder task in MacLean et al.'s study are difficult to interpret. For example, we found a significant positive trial effect for four of the bird species; see the electronic supplementary material for details. This might indicate that the subjects representing these species lacked sufficient initial experience of the task but gained experience across trials. The cylinder task is not suitable if adequate experience is lacking: i.e. failures might result from other causes than limited inhibition. It has also been shown in human children that previous experience with transparent objects is crucial for this task [19]; therefore, one should not administer the cylinder task if it is not clearly established that the subject has sufficient experience about transparent surfaces. Recently, it was shown that Clark's nutcrackers (Nucifraga columbiana) perform better in the last five, out of the 10, trials in the cylinder task, even if they have passed the familiarization trials with an opaque cylinder; the authors concluded that this was an effect of learning about transparent surfaces [25]. We do not know whether similar trial effects hold for some of the mammalian species. Here it is worth mentioning that all our birds have had previous experience of transparent surfaces, both from various enrichment activities and experiments since they were young.
One might reasonably ask not only for reports on failures and successes, but also for information on the behavioural dynamics of how the tasks were solved by different species. Not all of the few failures reported from the Corvus species in our study were 'clean' failures such that the bird tried to reach the reward directly as if the barrier was air. Instead, some birds nudged the cylinder, while others behaved as if exploring the attachment of the cylinder. Without details of possible exploratory behaviour across species, it might be premature to make the far-reaching conclusions about evolution and brain size that MacLean et al. do [26]. What is without doubt is that great apes and Corvus corvids have pronounced motor self-regulatory behaviour in relation to the cylinder task, despite very different absolute different brain sizes.
Ethics. The study was conducted in accordance with national and EU legislation and guidelines for animal research. The Swedish regional ethical board approved the raven testing. According to German law, no such similar approval was necessary for the New Caledonian crows and the jackdaws, as the studies do not fall under the definitions of animal research.
Data accessibility. The datasets supporting this article have been uploaded as part of the electronic supplementary material.