The authors hereby declare that funding received by a commercial source (L′Oreal) does not alter their adherence to all the PLOS ONE policies on sharing data and materials.
Conceived and designed the experiments: CAFW TB. Performed the experiments: CAFW. Analyzed the data: CAFW. Wrote the paper: CAFW TB.
Sensitivity to inequity is considered to be a crucial cognitive tool in the evolution of human cooperation. The ability has recently been shown also in primates and dogs, raising the question of an evolutionary basis of inequity aversion. We present first evidence that two bird species are sensitive to other individuals' efforts and payoffs. In a token exchange task we tested both behavioral responses to inequity in the quality of reward (preferred versus non-preferred food) and to the absence of reward in the presence of a rewarded partner, in 5 pairs of corvids (6 crows, 4 ravens). Birds decreased their exchange performance when the experimental partner received the reward as a gift, which indicates that they are sensitive to other individuals' working effort. They also decreased their exchange performance in the inequity compared with the equity condition. Notably, corvids refused to take the reward after a successful exchange more often in the inequity compared with the other conditions. Our findings indicate that awareness to other individuals' efforts and payoffs may evolve independently of phylogeny in systems with a given degree of social complexity.
The sensitivity regarding ones benefits of actions as well as efforts and payoffs for both the acting individual and its cooperation partner is proposed to be a key mechanism in the evolution of cooperation
We here tested the latter assumption by expanding research on IA to large-brained birds. Specifically, we investigated sensitivity to unequal reward structure and working effort in two corvids, carrion crows (
We tested the birds' performance in a token exchange task, comparable to that used by Brosnan and colleagues
This study complied with Austrian and local government guidelines, permission to conduct non-invasive cognition experiments was received by CW from the head of the Konrad Lorenz research station (KLF), Prof. Kurt Kotrschal. As experiments were entirely non-invasive, no further animal experimental license was required (
Corvids stayed in adjacent experimental compartments, separated from each other by a wire mesh. The experimenter kneeled in front of the compartments, having both reward typed visible for the birds in front of her.
Tests were conducted from January to July 2010 on six captive carrion crows (three males, three females) and four captive common ravens (three males, one female) at the KLF and the Cumberland game park, Austria. Individuals are housed in pairs in large outdoor aviaries and were tested in fixed dyads; four dyads consisted of affiliated pair partners/mates and one (raven) dyad consisted of non-affiliated but adjacently housed males. Crow dyads were tested individually in adjacent experimental compartments, separated from each other and the experimenter by a wire mesh (
In each task, we performed two sessions per subject in each condition, with 24 trials per session. Grapes (1/4) were used as low quality food rewards and cheese (0.5 cm×0.5 cm×0.1 cm) as high quality reward in both tasks. Grapes are a regular part of the birds' daily diet, whereas cheese is exclusively used as rewards in experiments. In all conditions, both reward types were present and visible to the focal subjects, in two separate transparent containers. In the raven dyads, containers where placed one and two meters away from the birds, in the middle of the experimental compartment. In the crow dyads approximately 50 cm away from the birds placed directly in front of the experimenter (
success | exchange | individual took the initial item with the beak and upon request gave it back into the hand of the experimenter |
refuses reward | individual successfully exchanged but refuses to take the reward | |
failure | refuses initial item | individual refuses to take the initial item |
drop | individual drops initial item but not into the hand of the experimenter | |
does not give back | individual does not give back the initial item upon request |
condition | task | action focal | action model | reward focal | reward model | |
equity | task A | exchange | exchange | cheese and grape | cheese and grape | |
equity low quality | task B | exchange | exchange | grape | grape | |
effort control | task A | exchange | cheese and grape | cheese and grape | ||
task B | grape | grape | ||||
quality control | task A | exchange | not present | grape | ||
no reward no partner | task B | exchange | not present | not rewarded | ||
both no reward | task B | exchange | exchange | not rewarded | not rewarded | |
inequity | task A | exchange | exchange | grape | cheese | |
task B | exchange | exchange | not rewarded | grape |
We used general linear mixed models (GLMMs) with binomial error distribution and a logit function. Response variables were the individual's behavioral response in each trial: exchange yes/no or refusal to take the reward. Condition, species, sex, dyad, session, start as focal or model, reward in the previous trial and type of reward (grape as low quality food, cheese as high quality food) served as fixed factors. In order to account for repeated measures for each individual, the individual identity was included as a random factor. All interactions between fixed factors as well as fixed factors and individual identity, remaining in the final model were included in the model. We selected the final model using a combination of a backward stepwise selection eliminating least significant factors from the model in order to reach the best model. This was determined using second order Akaike's information criteria (AICc), which compares the adequacy of several models and identifies the model which best explains the variance of the dependent variable as that with the lowest AICc value
Our final model included condition, reward in the previous trial, the quality of reward and the interactions between condition*individual and session*individual as significant predictors of success rate in exchanging with the human experimenter in task A (
Graph shows mean percentage of successful exchanges ± SE. *p<0.05; **p<0.01; ***p<0.001; Alpha after Bonferroni correction 0.0125.
task A: inequity in reward quality | |||||
fixed factors | numerator df | denominator df | F | p | |
condition | 3 | 1.787 | 3.278 | 0.02 | |
type of reward | 1 | 1.787 | 11.769 | 0.001 | |
reward in previous trial | 1 | 1.787 | 7.206 | 0.007 | |
session | 1 | 1.787 | 2.557 | 0.11 | |
interactions | |||||
condition * individual | 27 | 1.787 | 2.568 | <0.001 | |
session * individual | 9 | 1.787 | 4.042 | <0.001 |
task B: inequity in the absence of food reward | ||||
session | 1 | 2.335 | 16.617 | <0.001 |
condition | 4 | 2.335 | 5.369 | <0.001 |
type of reward | 1 | 2.335 | 0.011 | 0.74 |
interactions | ||||
session*condition | 4 | 2.335 | 7.073 | <0.001 |
session*individual | 9 | 2.335 | 9.969 | <0.001 |
individual*condition | 36 | 2.335 | 5.12 | <0.001 |
Factors and interactions (in bold) significantly influence exchange performance.
equity | effort control | quality control | inequity | ||
grape | cheese | ||||
task A: inequity in reward quality | |||||
Crows | |||||
3109301m | 23 (1) | 23 (0) | 48 (0) | 34 (0) | 38 (2) |
HF54707m | 19 (1) | 23 (0) | 39 (0) | 34 (0) | 35 (4) |
HF54710f | 19 (0) | 22 (1) | 34 (2) | 44 (8) | 28 (13) |
HF54709m | 16 (2) | 19 (0) | 30 (0) | 19 (0) | 19 (4) |
HF54706f | 20 (0) | 22 (0) | 48 (0) | 25 (2) | 43 (4) |
HF54708f | 15 (5) | 21 (0) | 19 (0) | 22 (10) | 38 (28) |
Ravens | |||||
JC46070f | 23 (0) | 24 (0) | 41 (6) | 34 (0) | 37 (15) |
JC38974m | 24 (0) | 24 (0) | 44 (1) | 46 (0) | 46 (4) |
JC38981m | 24 (0) | 24 (0) | 35 (0) | 32 (0) | 27 (0) |
JC46083m | 24 (0) | 24 (0) | 48 (0) | 48 (0) | 46 (0) |
task B: inequity in the absence of food reward | |||||
equity low quality | effort control | no reward, no partner | both no reward | inequity | |
Crows | |||||
3109301m | 43 (0) | 35 (0) | 18 | 10 | 24 |
HF54707m | 43 (20) | 34 (0) | 35 | 41 | 44 |
HF54710f | 11 (0) | 3 (0) | 8 | 24 | 20 |
HF54709m | 18 (1) | 22 (0) | 5 | 24 | 9 |
HF54706f | 30 (0) | 13 (0) | 18 | 17 | 35 |
HF54708f | 34 (0) | 32 (0) | 4 | 14 | 19 |
Ravens | |||||
JC46070f | 25 (0) | 22 (0) | 21 | 13 | 6 |
JC38974m | 26 (0) | 1 (0) | 14 | 10 | 8 |
JC38981m | 20 (0) | 33 (0) | 33 | 24 | 16 |
JC46083m | 44 (0) | 46 (0) | 29 | 19 | 12 |
Please note that in task B (inequity in the presence/absence of reward), reward refusals where not possible in the no reward, no partner, both no reward and inequity conditions, as focal individuals did not receive any reward for exchanging. In each condition each individual received a total of 48 trials (2 sessions á 24 trials). f female individuals, m male individuals.
T | p-value | |
|
||
equity high-equity low | 0.171 | 0.87 |
|
|
|
|
|
|
|
|
|
equity low-effort control high | 1.72 | 0.14 |
|
|
|
equity low quality reward-quality control | ||
|
|
|
|
|
|
effort control high-effort control low | 2.38 | 0.06 |
effort control low-quality control | 0.32 | 0.76 |
effort control low-inequity | 1.03 | 0.35 |
effort control high-inequity | 1.6 | 0.17 |
inequity-quality control | 1.826 | 0.12 |
|
||
|
|
|
|
|
|
|
|
|
|
|
|
effort control-no reward no partner | 2.37 | 0.055 |
effort control-both no reward | 1.56 | 0.169 |
effort control-inequity | 1.65 | 0.15 |
no reward no partner-both no reward no partner | 0.93 | 0.388 |
no reward no partner-inequity | 0.83 | 0.438 |
both no reward- inequity | 0.87 | 0.417 |
Differences between the parameter estimates have been divided by the standard error (SE) differences between pairs and interpreted the output as a t-test.
According to our expectation concerning reward distribution, exchange rate was higher in the equity condition than in the inequity condition, whereby the effect was stronger with high quality food (T = 5.35, p<0.001) than with low quality food (T = 2.61, p = 0.04; n.s. after Bonferroni). According to our expectation concerning sensitivity to effort, exchange rate was higher in the equity condition than in the effort control, whereby the effect was smaller with high quality food (T = 2.88, p = 0.03; n.s. after Bonferroni) than with low quality food (T = 4.14, p<0.001; see
Equity- effort control (T = 0, p = 1); equity- quality control (T = 7, p = 0.001); equity-inequity (T = 10.5, p = 0.001); effort control- quality control (T = 7.5, p = 0.001); effort control- inequity (T = 10.6, p = 0.001); quality control-inequity (T = 12.25, p = 0.01).
In task B, condition significantly influenced exchange performance (
Graph shows mean percentage of successful exchanges ± SE. Alpha after Bonferroni correction 0.01.
In conditions where the focal individual did not receive any reward for exchanging, performance dropped dramatically compared to the equity condition (equity- no reward no partner T = 4.44, p = 0.004; equity-both no reward T = 3.65, p = 0.01; equity- inequity T = 3.74, p = 0.009; see
To our knowledge, this is the first evidence of behavioral responses to inequity in birds. As expected, crows and ravens responded to inequity in outcome, i.e. to difference in quality (task A) and to presence/absence of the reward (task B); moreover, they were highly sensitive to inequity in working effort in both tasks. In addition, the quality of the offered food (grapes and cheese) influenced the birds' exchange performance, whereby the differences to inequity and quality were most pronounced with highly valued food (cheese).
Interestingly, the strongest effect of the current study was found in respect to working effort: when the model individual received the reward as a ‘gift’ without exchanging a token first, the focal bird, required to exchange for the reward, decreased its performance. These results are of special interest, as a main criticism regarding inequity aversion (IA) in non-human animals is that their behavioral responses might reflect individual frustration and/or response to negative contrast (i.e. receiving less preferred food after receiving more preferred food) rather than sensitivity to inequity
Regarding inequity in the outcome, corvids decreased their exchange performance in the inequity, compared to the equity condition in both tasks. However, in task A, it is difficult to disentangle the effects of food quality and inequity. Exchange rates in the inequity condition are not significantly different to rates in the quality control, where individuals were tested alone and received a low quality food reward for exchanging. Refusals in the inequity condition could thus simply be due to the low quality of the offered reward.
The fact that crows and ravens did respond to the non-social controls of tasks A and B with a drop in performance underlines that they are highly sensitive to the social context. Indeed, the mere presence of a social partner seemed to have a facilitating effect on the birds' motivation to exchange. In parallel studies using the same paradigm, we could perform eight trials per sessions when individuals were tested alone
To control for a possible frustration effect in our tasks, the prospected reward was always presented to both, the model and the focal individual in advance of the exchange. Therefore, individuals were expected to refuse to exchange when treated unequal, but not to complete the task and then refuse the reward. Refusing the reward after exchanging the initial item presents a clear cost, as individuals are outputting the working effort but not accepting the reward. That the exchange task reflects some working effort for corvids is illustrated by the birds' behavior in the effort control condition of the current study. Moreover, in another study, crows preferred larger quantities of food over smaller ones in a choice task, but were not willing to exchange one piece of food for a larger number of the same reward
A limitation of the presented study is the low sample size. As only six crows and fours ravens were available for testing, we must be cautious with generalizing our findings. For instance, it might be possible that species differences between crows and ravens exist, but could not be detected with our sample. Also, the interactions between fixed factors (condition, session) and individual identity indicates that individuals do respond differently to inequity. For example regarding the significant interaction between session*individual, most birds increased performance in the course of the experiment, except two which decreased exchange rates in later sessions. This could be due to many factors such age, sex, affiliation status and ‘personality’
Despite those limitations, we would like to stress that the birds in our study performed qualitatively similarly to some primates
We are grateful to Kurt Kotrschal and Rachael Miller for support and general discussion.