The aim of this study was to test instrumental mirror use in Eurasian jays by presenting a modified version of the string-pulling task. Jays had to understand the direct link between the reward’s position in space and the reflection provided by the mirror to pull the correct string. This is the first investigation approaching a mirror-guided reaching task in birds. After an initial familiarisation with the string-pulling task, four out of seven birds learned to obtain a reward by pulling the correct string. However, when the baited and un-baited strings could only be seen when reflected by a mirror, and not directly, none of the jays tested pulled the baited string at significant levels.
The results of Experiment 1 are comparable to those available from several species of corvids tested on a horizontal patterned string-pulling task (Obozova et al. 2014; Jacobs and Osvath 2015). Although the criteria for passing and the number of trials conducted across studies are not identical, what emerges is that the number of subjects that pull the string in a goal-directed manner (in a perpendicular string pattern) is quite small. In a seminal study, all three ravens presented with horizontal strings first touched baited and un-baited strings at chance, though they preferentially then switched to obtain baited strings (Heinrich 1995). Three out of four hooded crows and ravens (Bagotskaya et al. 2012) solved the parallel string pattern within 32 trials, whereas only one California scrub-jay out of five (Hofmann et al. 2016), and one oriental magpie out of eight (Wang et al. 2020a) were able to successfully pass this test (within 50 and 30 trials, respectively). A modified version of the parallel configuration was necessary to increase the number of California scrub jays able to pull the correct string (Hofmann et al. 2016): the location of the reward was made more visible and with higher contrast compared with the unrewarded string; four out of five birds managed to pass this re-designed task within 50 trials. Parrots were more successful in string-pulling tasks in various patterns, however, they were mainly tested using a vertical configuration (Werdenich and Huber 2006; Schuck-Paim et al. 2009; Krasheninnikova and Wanker 2010; Krasheninnikova 2013; Krasheninnikova et al. 2013; Krasheninnikova and Schneider 2014; Molina et al. 2019; Ortiz et al. 2019, see van Horik and Emery 2016 for use of an horizontal variation). Although the vertical version appears to be more challenging than the horizontal configuration, there are some exceptions. In one study, none of the grey parrots passed the horizontal parallel and more complex patterned configurations, though they passed a vertical string-pulling task (Jacobs and Osvath 2015; Molina et al. 2019). In our study, Hoy and Lima were able to pass the task within 30 and 40 trials, whereas Washington and Romero passed in 110 and 120 trials, respectively. Romero also had an overall left side bias, especially during the initial trials. Due to the stringent criteria we applied, the range of trials required for each bird to pass was quite wide, although two birds did pass the test within a similar number of trials to those in other studies. The three other birds stopped pulling the string over four consecutive sessions, performing a range of 52 to 97 trials, with two of them preferentially pulling the string on the right side (Caracas and Wellington).
Limited conclusions can be drawn about causal reasoning in the jays that passed, as it was beyond the purpose of this study to disentangle the different cognitive mechanisms involved in the string-pulling task, which would have required additional configurations (Jacobs and Osvath 2015). Establishing whether jays used goal-directed behaviour in this study is also very difficult in the absence of control experiments testing whether the birds can solve the string-pulling task without relying on proximity with the reward, whether they can generalise the rule to modified versions of the task, and whether their performance on the task is independent of perceptual cues (Schuck-Paim et al. 2009; Jacobs and Osvath 2015). However, the performance of at least four of the jays tested here, those that passed, may be consistent with some understanding of causal reasoning and connectivity being present in Eurasian jays, which may warrant direct investigation in the future studies.
Experiment 2 was performed with only the four jays that successfully completed the horizontal parallel string-pulling task, to investigate whether they could solve a complex visual spatial task using mirrored information. The task presented here, inspired by mirror-guided reaching tasks, is technically and qualitatively different from the tasks developed for primates, to account for the morphological differences between species, but is conceptually similar (Pepperberg et al. 1995). When tested in a mirror-guided reaching task, chimpanzees were able to track the movements of their hands towards the food or object by looking at the mirror, as well as a live video in some conditions, even when the object was reversed laterally, inverted by 180°, or both reversed and inverted (Menzel et al. 1985). Control experiments indicated that two chimpanzees distinguished between pre-recorded and live videos. This level of proficiency was also evident in a subsequent study, in which four macaques were tested in a mirror-guided reaching task (Anderson 1986). Only two macaques out of four were able to use the mirror to locate food and guide their hands to it. Well-designed controls were performed to make sure that the animals realised the association between their hands’ movements and the reflection, and subjects that showed understanding of this relationship behaved accordingly. In trials in which no food was hidden, when the usual movements were performed by the experimenter and the mirror was available, subjects reduced their searching behaviour, whereas they tried to search for the food more often when the mirror was removed. In contrast, the two subjects that were not able to use the mirror instrumentally searched for the food equally in both mirror and mirror-less, un-baited conditions.
A few bird species tested in a similar, mirror-mediated spatial locating task had to use a mirror to locate the bait in one of three or four adjacent locations, normally placed below a countertop. Parrots performed quite well (using three compartments, above 75% correct), even on transfer tests (four compartments, above 60%; and with an overhead mirror, 75%; Pepperberg et al. 1995). New Caledonian crows were tested on a similar task after receiving initial training on a two-box apparatus (three birds performed below 50% correct, but one bird made 90% correct choices over 30 trials): in a four-box apparatus, all the birds (10/10 correct for one crow, with 6/10 trials in the three-choice task) were able to pick the bait correctly within 20–30 trials. The crows’ results are interesting because they were better able to locate the bait in a four-box than a two-box apparatus (Medina et al. 2011). One plausible explanation for this is that they learned over time to associate the location of the food and the image reflected in the mirror. According to the definition of mirror-mediated spatial locating, the subjects must form a mental representation that the stimulus (food or object) reflected in the mirror is the same stimulus located in real space (Povinelli 1989; Medina et al. 2011). Accordingly, it is very likely that the crows (at least three out of four subjects) did not comprehend the use of the mirror in such a fashion. However, they did not simply use the mirror as a trigger to initiate searching behaviour, either (i.e. they selected the correct compartment in the four-box apparatus).
The task conducted in our study cannot be classified as a mirror-guided reaching task because the birds did not need to constantly monitor the movement of the strings to obtain the reward successfully. However, it is still more complex than the mirror-mediated spatial locating task performed with grey parrots and New Caledonian crows (Pepperberg et al. 1995; Medina et al. 2011). In fact, jays could not associate the food with their own image (which they could not see) moving towards the mirror and the correct location, because they needed to use a string to pull the food towards them instead (although, had they been successful, they may have learned to associate the side in which the food was visible with the correct string instead—had this been the case, further configurations, such as crossed-string paradigms, could have been further used to disentangle the mechanisms behind their success). Furthermore, to get the reward, they had to understand that this was possible only using the string, and that the image reflected in the mirror represented both the same string whose end they could see in front of them, as well as the real reward. Moreover, the visual-perceptual motor feedback intrinsically present in the classical string-pulling task is limited in our study. The food is completely hidden apart from in the mirror, and the movement of the string attached to the food cannot be monitored because as it is being pulled it moves out of view of the mirror (Taylor et al. 2010). In our study, birds would need to look to a different, possibly non-intuitive initial location of the food in Experiment 2 (slightly above them, as reflected by the suspended mirror), assume connectivity of a string to the food, a connectivity that cannot be seen directly in front of them, and disregard the fact the food disappears while the correct string is pulled, to learn to solve Experiment 2 correctly. These disruptions of the visual presentation and subsequent visual feedback afforded by Experiment 1 may have not given the jays enough to go on to learn to use the mirror to solve Experiment 2 during the duration of the study. The performance of the jays, overall, does not indicate that they were able to solve this complex mirror-mediated spatial locating task. Based on their performance, it is, however, possible to suggest that they used the mirror as a trigger to initiate searching behaviour. For example, a mirror-mediated spatial locating task recently performed in azure-winged magpies showed that four out of five birds did not search for the food in the location where it was hidden (i.e. in a compartment above their heads; instead the birds often went behind the mirror), although they looked at the mirror more often when the food was present than when it was not and likely used the mirror as a cue to trigger a search (Wang et al. 2020b). Therefore, the performance of the Eurasian jays could have been due to the limited direct visual-perceptual motor feedback provided, supporting the argument that this feedback is crucial for the solution of string-pulling tasks (Taylor et al. 2010, 2012; Jacobs and Osvath 2015), rather than due to an inability to use mirrored information, although other explanations, discussed as follows, are also possible.
It is possible that other factors implicit in the setup of this novel task, aside from the requirement to use the mirror to find the food, may have accounted for the jays’ behaviour. For example, the limited visual information available in general, aside from the lack of visual-perceptual motor feedback, may account for the birds’ performance. Jays could only see the end of the string protruding from the apparatus, and could see the rest of the string and the reward only reflected in the mirror. Thus, jays may have been unable to understand that connectivity between the string available to them and the string reflected in the mirror was preserved, rather than interrupted. In studies in which “broken strings” were used, birds with previous string-pulling experience tended to select the unbroken strings, whereas naïve birds tended to choose at chance (see a discussion of this in Bastos et al. 2021). It is possible that these experienced Eurasian jays may have expected connectivity as a requirement of the task, and that this lack of visible connectivity may have hindered their performance. After all, many corvid species tested, including Eurasian jays, have shown evidence of fairly sophisticated physical cognition abilities (Emery and Clayton 2009; Cheke et al. 2011), and Eurasian jays have even shown evidence of expecting sufficient physical support for objects in a violation of expectations paradigm (Davidson et al. 2017). Consequently, the jays may have an expectation about the necessity for strings to be connected. In addition, it is possible that other executive functions, such as an inability to inhibit selecting a string directly, without pausing to observe and inspect the mirror, may also account for this failure: the performance of corvid species on delayed gratification tasks is variable (Miller et al. 2019), and inhibiting the choice of a string to obtain more information from the apparatus may have been difficult for the jays.
To our knowledge, this is the first study to investigate a complex form of instrumental mirror use in Eurasian jays, a task that, like previously designed mirror-mediated spatial locating tasks, can be tailored such that solving it would require understanding of mirrored information as depicting a real object in real space. However, detangling the specific mechanisms subjects may use to solve it would require additional controls and configurations than those used here, in which the birds never solved the task. Although this task cannot be categorised as mirror-guided reaching, since the birds did not need to constantly monitor, and re-adjust, the movement of the strings using the mirrored information, it likely adds an additional level of cognitive demand to the classic mirror-mediated spatial locating task, through the requirement to use mediating strings. Although the Eurasian jays’ performance in this task did not reveal an ability to use mirrored information, this study does provide a novel methodology to study how birds may process mirrored information, in a manner more similar to the primate mirror-guided reaching task. Given the additional conceptual difficulties the task may require compared to simpler mirror-guided spatial locating tasks, it would be worth investigating whether other avian species, especially those successful at other mirror use tasks, succeed on this novel task, or whether it poses sufficient additional cognitive demands to hinder their performance as well.