Pangram verdict · v3.3
We believe that this text is a mix of AI, AI-assisted, and human-written content.
AI likelihood · overall
MixedArticle text · 1,399 words · 5 segments analyzed
IntroductionSince Gallup’s pioneering study of chimpanzees1 and Amsterdam’s later adaptation for children2, the mirror self-recognition (MSR) test has been widely regarded as the primary method for assessing self-awareness across species. Although the MSR task has been highly influential, its comparative value has been widely disputed because it relies heavily on vision and may therefore be less ecologically valid for species with different dominant sensory modalities. Moreover, reports of impressive MSR-like behavior in a relatively small-brained animal such as the cleaner wrasse (Labroides dimidiatus)3 have sparked debate over what the test actually captures. This prompted a shift towards a gradualist, modular framework that accounts for species-specific sensory modalities and ecological contexts4,5,6. As de Waal argued, animals must represent the affordances of their own bodies in relation to the physical and social environment, suggesting that mirror responses alone should not remain the sole benchmark for studying self-awareness6. A highly promising complementary approach is the “body-as-an-obstacle” (BAO) task. Unlike MSR, which focuses on visual self-recognition, BAO assesses whether subjects treat their own body as a causal obstacle during goal-directed action.The BAO task measures an individual’s ability to recognize parts of their own body as an impediment to achieving a goal, thereby assessing body-related self-representation. Developed for human infants7,8, the task requires subjects to hand over an object while standing on a mat directly attached to the object; success requires stepping off the mat to eliminate their own body weight as the source of obstruction. In a control condition, the obstruction is external (e.g., a heavy object), so stepping off is unnecessary. In humans, BAO performance emerges at roughly the same age as MSR (18–24 months) and correlates moderately with it7,9. Cross-cultural work shows that infants from autonomy-oriented cultures tend to pass MSR earlier, whereas those from relatedness-oriented cultures succeed earlier on BAO9. These findings indicate that MSR and BAO follow distinct developmental trajectories and rely on distinct processes. At the same time, they represent different modes of self-representation. Because BAO targets the embodied self, it offers a valuable alternative for species in which vision is not the dominant sensory modality.From a cognitive standpoint, both MSR and BAO require causal reasoning, but they differ in sensory demands. MSR depends on recognizing visual contingencies and features, whereas BAO relies on awareness of direct body-environment interaction without requiring an external visual representation7,9. Because BAO lacks visual prerequisites, it does not require prior mirror experience or visual inspection of the body, which may render it more ecologically valid for animals that rely primarily on non-visual modalities10. These advantages have motivated recent applications of BAO to non-human species.Researchers have increasingly used the BAO paradigm to assess body awareness in nonhuman animals, with positive results in elephants, dogs, chimpanzees, and lowland gorillas10,11,12. Because elephants’ size and strength precluded an “impossible” control, Dale and Plotnik included an unattached control and a foot-discomfort control; elephants stepped off the mat significantly more often in the test than in either control, whereas the two controls did not differ significantly from one another11. In dogs, Lenkei et al. implemented both controls and an attached-to-ground condition, and dogs again left the mat more often and sooner in the test than in the critical attached control, including in the first trial10. Vanhooland et al.
later adapted BAO for great apes using a non-social design and a single control analogous to the unattached condition used in earlier studies12.Together, these three studies showed that success on BAO, although correlated with MSR in humans, can be passed independent of having passed MSR in these animals. Adult Asian elephants and great apes have demonstrated MSR under certain conditions (e.g., after appropriate prior experience;13,14,15), whereas dogs have not16, suggesting that body awareness may be evolutionarily widespread and may represent a fundamental, ecologically relevant component of self-representation.4Although BAO has not yet been adapted for birds, there are strong reasons to do so. Birds, particularly parrots and corvids, have demonstrated a cognitive toolkit comparable to that of great apes17,18, yet they have often yielded inconsistent evidence under MSR paradigms (e.g., replication attempts have yielded different results19,20). This evidence, together with behavioral and neuroanatomical findings often linked to MSR in mammals21,22,23, has led several authors to advocate avian species as compelling non-primate candidates for testing self-awareness using better-adapted methodology5.Goffin’s cockatoos (Cacatua goffiniana) are highly innovative problem-solvers: they flexibly manufacture, modify, and use tools, and they engage extensively in haptic exploration with their beak and feet18,24,25,26. This tactile interaction style aligns well with the embodied, action-based demands of BAO. Like other parrots tested thus far, Goffin’s cockatoos have failed in attempts to test MSR27.However, sulfur-crested cockatoos have been observed opening rubbish bins in a manner reminiscent of BAO-type problem-solving28, although the learning mechanism remains unknown: they have learned to step off the lid of a rubbish bin onto the rim of the bin, avoiding the impediment of their own body weight for opening the lid.These observations suggest that other, highly flexible white Cacatua species, such as the Goffin’s cockatoos, are strong candidates for BAO testing.Here, we present the first avian adaptation of the BAO paradigm, using Goffin’s cockatoos as a model species (Fig. 1; Supplementary Video S1). We employ a “give-me” command to preserve the social element of the original design and include the key control comparisons used in mammalian BAO studies (unattached and attached-to-ground), plus a pre-test controlling for foot discomfort.
We conceal visual differences between test and control conditions to prevent cue-based anticipation, and we focus the analysis on the first test session to reduce the likelihood that the main condition effect reflects repeated exposure across sessions.
We hypothesize that Goffin’s cockatoos would be able to treat their own body as a causal obstacle to task completion in a body-as-obstacle task.Fig. 1Setup and testing procedure during the test condition: (A) picking up the ball after the “give-me” command, (B) pulling and stepping out of the mat, (C) transferring the ball while out of the mat. (D) schematic representation of the setup during the transfer sequence (including mat, corridor and left camera).ResultsPre-test performanceAll fifteen Goffin’s cockatoos successfully passed the pre-test. Out of 180 pre-test trials (15 subjects × 12 trials), 179 were completed successfully, corresponding to a success rate of 99.4%. Only one bird stepped off the mat during a pulling attempt; however, because this individual completed all remaining pre-test trials successfully, all subjects were retained for the main experiment.Overall performance across conditionsAcross all experimental sessions, cockatoos stepped off the mat in the test condition in 97.8 ± 0.9% of trials (mean ± SE across subjects; 528/540 trials overall). In contrast, they stepped off in only 0.7 ± 0.3% of trials in the unattached control (4/540 overall) and 39.1 ± 3.9% of trials in the attached-to-ground control (211/540 overall).Birds were substantially more likely to step off the mat in the test condition than in either control condition (Fig. 2). A binomial GLMM showed a significant effect of condition on step-off probability (likelihood-ratio test: χ2 (2) = 88.85, p < 0.001).
Post hoc pairwise comparisons confirmed that the test condition differed significantly from both the attached-to-ground and unattached controls, and that the two control conditions also differed (all adjusted p < 0.001). The grouping factor (subject group identity) did not significantly affect step-off probability (p = 0.92), suggesting that group assignment (see methods section) was unlikely to account for the observed pattern.Fig. 2Probability of stepping off the mat across test and control conditions: Model-estimated probability of stepping off the mat (95% confidence intervals) in Goffin’s cockatoos under three conditions: Test (object attached to the mat on which the bird stood), Attached (attached-to-ground), and Unattached (object not attached). The left panel shows estimates for all sessions, and the right panel shows estimates for Session 1 only.Behavioral responses in the attached-to-ground control conditionIn the attached-to-ground control condition, birds could step off the mat while still holding the ball, but stepping off could not solve the task. Because this condition was unsolvable by design, trials were not classified according to successful object transfer. Instead, we recorded whether the bird disengaged from the unsuccessful attempt within the trial window. Behaviors counted as disengagement included walking away, redirecting activity, or showing signs of frustration. Across all sessions, 329 of 540 attached-to-ground trials (60.9%) met this disengagement criterion. Among trials that ended in disengagement, the most common outcome was walking away (158/329, 48.0%), whereas the remaining disengagement trials mainly involved redirected contact behaviors during unsuccessful attempts to obtain the object.