The authors have declared that no competing interests exist.
Domestic dogs have been shown to reciprocate help received from conspecifics in food-giving tasks. However, it is not yet known whether dogs also reciprocate help received from humans. Here, we investigated whether dogs reciprocate the receipt of food from humans. In an experience phase, subjects encountered a helpful human who provided them with food by activating a food dispenser, and an unhelpful human who did not provide them with food. Subjects later had the opportunity to return food to each human type, in a test phase, via the same mechanism. In addition, a free interaction session was conducted in which the subject was free to interact with its owner and with whichever human partner it had encountered on that day. Two studies were carried out, which differed in the complexity of the experience phase and the time lag between the experience phase and test phase. Subjects did not reciprocate the receipt of food in either study. Furthermore, no difference was observed in the duration subjects spent in proximity to, or the latency to approach, the two human partners. Although our results suggest that dogs do not reciprocate help received from humans, they also suggest that the dogs did not recognize the cooperative or uncooperative act of the humans during the experience phase. It is plausible that aspects of the experimental design hindered the emergence of any potential reciprocity. However, it is also possible that dogs are simply not prosocial towards humans in food-giving contexts.
Cooperation is observed in nature across a wide range of species and contexts [
Domestic dogs are a promising model species for the study of reciprocity. Dogs have been shown to express helpful or prosocial behaviours (i.e. behaviours that benefit others [
Dogs are also particularly interesting as they have been shown to express prosocial behaviours in interspecific interactions. Two recent experimental studies reported dogs directing prosocial behaviours towards humans. First, Carballo et al. [
Further to the expression of prosocial behaviours towards conspecifics and humans, dogs have recently been shown to reciprocate help received from conspecifics. In a recent study conducted by Gfrerer and Taborsky [
In a subsequent study, Gfrerer and Taborsky [
Dogs possess additional characteristics which might suggest the propensity to reciprocate help received from humans. Apart from a long history of dog-human cooperation and communication [
Carballo et al. [
Several additional studies present similar findings, with dogs discriminating between a cooperative and uncooperative human [
Given dogs’ long history of cooperating with humans, their specialized skills for such interaction with humans, their capacity to develop bonds with humans, their ability to discriminate between a cooperative and an uncooperative human, and their propensity to reciprocate help received from conspecifics, we investigated whether pet dogs reciprocate help received from humans. To investigate potential reciprocation of help received, we applied a design generally matching that used to demonstrate reciprocity in dogs [
All procedures were approved by the ethics committee of the University of Veterinary Medicine, Vienna (ethical protocol no.’s: ETK-26/02/2019; ETK-134/07/2019). Additionally, dog owners were required to sign a consent form prior to participation in the study.
All dogs included in the study were pet dogs and were recruited via advertisements posted on social media and via word of mouth. The final sample tested in this study comprised 21 dogs (12 females; 9 males) of varying age (mean ± SD: 5.8 ± 2.8 years) and from a range of breeds, including mixed breeds (see
Subject ID | Age (years) | Sex | Breed | No. of sessions to reach criterion |
---|---|---|---|---|
A1 | 5 | F | Australian shepherd | 1 |
B1 | 10 | M | Border collie | 1 |
C1 | 6 | M | Border collie | 1 |
D1 | 11 | F | Border collie | 1 |
E1 | 8 | F | Border collie | 1 |
F1 | 7 | M | Greyhound | 3 |
G1 | 5 | F | Terrier mix | 1 |
H1 | 11 | M | Airedale terrier | 3 |
I1 | 4 | M | Hungarian vizsla | 3 |
J1 | 5 | M | Mix | 1 |
K1 | 5 | F | Yorkshire terrier | 3 |
L1 | 6 | M | Dachshund | 1 |
M1 | 4 | M | Spanish galgo | 3 |
N1 | 4 | M | Australian cattle dog mix | 1 |
O1 | 2 | M | Labrador retriever mix | 1 |
P1 | 1 | F | Border collie | 1 |
Q1 | 7 | F | Podenco | 2 |
R1 | 3 | F | Beagle | 2 |
S1 | 3 | M | German hunting terrier mix | 2 |
T1 | 5 | M | Australian shepherd mix | 1 |
U1 | 9 | F | Australian shepherd | 2 |
F, female; M, male.
Human participants, who acted as helpful or unhelpful partners in the setup (see below), were recruited through personal contact. They were briefly instructed on their task prior to participation. The final sample size of human participants was 12 (8 females; 4 males). The helpful and unhelpful partners were humans who were unfamiliar to the particular subject with which they were tested.
The study took place in a rectangular room (approx. 7 m x 6 m) and primarily within two adjacent, square enclosures (1.5 m x 1.5 m; see
Two circles (each approx. 1 m in radius) were marked on the floor of the room, for the free interaction session. A red water bowl was also present in the room. A food dispenser and button are depicted in the middle of the room (note: the dispenser and button were only in this position for the first stage of training). Black curtains surrounding some of the fences are represented by thick black lines.
The cooperative act performed by the helpful human comprised the provision of food. More specifically, the helpful human pressed a button to release dry dog food from a commercially available food dispenser (Trixie, Dog Activity Memory Trainer; cat no. 32040; TRIXIE Heimtierbedarf GmbH & Co. KG, Tarp, Germany). The unhelpful human pressed a non-functional button (i.e. a button that was identical to the functional button but that was switched off) and, therefore, failed to provide food. To provide food in return, dogs could press the same type of button, which released either small, round chocolates, chocolate covered almonds, or plain hazelnuts, depending on the human’s preference (this was determined prior to the experiment by the experimenter).
The button, when switched on, emitted an audible sound when pressed and remotely controlled the food dispenser such that food was released. Regardless of whether a particular button was switched on or off, it was always presented in a green rubber holder which came with the product. The food dispenser could be adjusted so as to control the amount of food released each time the functional button was pressed. However, the amount of food could not be controlled precisely; thus, for dogs, the food dispenser released approximately five pieces of dry food each time the button was pressed and for humans the dispenser released approximately two pieces of food.
When the button was situated in the dog’s enclosure, it was covered by a wooden box. The box was attached to a wooden base by hinges and springs. A rope was attached to the box so that the box could be opened from the experimenter’s position by pulling. Only when the box was open, could the dog press the button. When the experimenter released pressure on the rope, the box closed again due to the force of the springs. A trial for the subject was defined as a single opening and closing of the box (i.e. a single presentation of the button).
Four cameras (one in each corner of the room) recorded general events that occurred in the room, while two webcams were used to record the activities within the two enclosures. Three outer fences of the enclosures were covered with black curtains so that the experimenter and owner could remain hidden from view during most phases of the experiment. The experimenter sat next to the human’s enclosure and watched the proceedings via the two webcams. Dog owners could sit next to or behind the experimenter and could also watch the session on the experimenter’s laptop, via the live feed from the webcams. Seven of the subjects were uncomfortable in the absence of their owner; therefore, the owner was permitted to sit behind the subject’s enclosure, in view of the subject.
All testing took place at the Clever Dog Lab of the Messerli Research Institute, located at the University of Veterinary Medicine, Vienna.
The training phase was conducted in five stages, as follows (see
Stage 1:
In the first stage of training, the button and dispenser were placed in the middle of the room, approximately half a metre apart. The subject was trained to press the button; the owner or experimenter attracted the subject’s attention to the button verbally, by pointing, or by pressing the button. Each time the button was pressed, food rewards were released from the dispenser. The subject was permitted to retrieve these rewards. Once the subject had successfully pressed the button five consecutive times independently (i.e. without the assistance of the owner or the experimenter), it proceeded to the next stage of training.
Stage 2:
In stage two of the training, the button and dispenser were situated in the subject’s test enclosure. The button remained in the box which was permanently open, and the food dispenser was positioned approximately half a metre away. Again, the dog was required to press the button and retrieve the food rewards from the dispenser five consecutive times to proceed to the next stage of training. The curtain covering one fence of the dog’s enclosure was typically not lowered during this stage of training so that the dog could still see out into the room and see its owner.
Stage 3:
In stage three, movement of the box was introduced to separate trials. After each time the dog pressed the button, the box was closed for approximately four seconds and then reopened. The subject was required to press the button on five consecutive trials to progress to the next stage. Box opening and closing was continued for all subsequent stages. For dogs that were uncomfortable with the opening or closing of the box while eating, the dispenser was placed against the opposite fence. For one very tall dog, the dispenser was placed on an upturned box against the opposite fence so that it could eat from the food dispenser comfortably.
Stage 4:
In stage four, the food dispenser was placed in the human enclosure, directly adjacent to the box containing the button. A sliding door on the fence separating the two enclosures remained open. The subject was required to press the button in its own enclosure and then enter the neighbouring enclosure through the open sliding door, retrieve the food, and return to press the button again. Subjects were permitted to proceed to the next stage of training once they had pressed the button and retrieved the food independently on ten consecutive trials.
Stage 5:
Stage five of the training was similar to stage four; however, the sliding door between the two enclosures was closed. Thus, if a subject pressed the button, food was released by the dispenser in the human’s enclosure, but the subject could not enter the human’s enclosure to retrieve it. Ten trials were conducted with this setup; however, the subjects were not required to press the button to proceed to the next stage. The purpose of this stage of training was to give subjects the opportunity to learn that, when the dispenser was on the other side of the fence and there was no opening, they could not retrieve the food. This stage of training matched the asocial control condition (see below), though the fence separating the enclosures in the asocial control condition did not have a sliding door. The curtain was typically lowered on one fence of the dog’s enclosure in this stage of training.
Five motivational sessions were conducted after this stage of training. These motivational sessions were identical to stage three of training and were included to ensure that the button pressing behaviour was not forgotten or extinguished by the previous experience of not receiving food.
Stage 6:
The final stage of training matched the previous stage; however, the human enclosure was occupied by a human who ate the food (human food) released from the food dispenser each time the subject pressed the button. The human eating the food in this stage was either the experimenter or a human unfamiliar to the dog; they were not one of the humans who would later act as the helpful or unhelpful human. Again, ten trials were conducted but the subject was not required to press the button in order to meet the criteria for inclusion in the study. The purpose of this stage of training was to give the subjects the opportunity to learn that, if the dispenser was in the human’s enclosure, and a human was present, pressing the button resulted in the human being rewarded. Five motivational sessions were also conducted after this stage of training.
In stage 1, the owner and experimenter trained the subject to press the button to release food from the dispenser in the middle of the room. In stage 2, the subject was trained to press the button inside the permanently opened box in the enclosure to release food from the dispenser. In stage 3, opening and closing of the box was introduced (this was controlled by the experimenter pulling or releasing the rope, respectively). The subject could press the button each time the box was open to activate the dispenser. In stage 4, the dispenser was placed in the adjacent enclosure and a sliding door between enclosures was open. The subject could press the button each time the box was opened and then retrieve food from the dispenser in the adjacent enclosure. In stage 5, the sliding door between the enclosures was closed. The subject could press the button each time the box opened and food would be released from the dispenser; however, the subject could not access this food. In stage 6, the dispenser and a human were present in the adjacent enclosure. If the subject pressed the button when the box was opened, the human ate the food that was released from the dispenser. Note: in stage 6, typically a second person was not available; therefore, the experimenter played the role of the human and controlled the opening of the box while kneeling in the human enclosure.
Between the stages of training, subjects were released from the enclosures and allowed to roam freely in the test room where they had access to their owner and water. If a subject showed signs of stress or lack of motivation, it was also given a break. However, if more than two breaks were required within a single stage of training, the session was terminated and continued on another day at that particular stage of training, preceded by a motivational session (provided the subject had already succeeded in stage three of training). Twelve subjects completed the training in a single session of one hour or less, four subjects required two separate sessions (i.e. two separate days), and five subjects required three sessions (i.e. three separate days).
Subjects received two experience phase sessions, one with the helpful human and one with the unhelpful human, each on a separate day (see
Immediately prior to each experience phase, the human entered his/her enclosure before the dog entered the room so that the first encounter with the human was in this experimental setting. During the experience phase, the human knelt in the enclosure, facing the dog’s enclosure (see
With the exception of the first two trials of the experience phase, the helpful human knelt in front of the functional button, which was adjacent to the food dispenser, and pressed this button on each trial, thereby releasing food for the subject. With the exception of the first two trials, the unhelpful human knelt in front of the non-functional button and pressed this on each trial resulting in no food being released for the subject.
The humans were instructed to press a button once every ten seconds, which was signaled by a timer hanging from the top of the fence separating the two enclosures, just above eye level. Humans were instructed not to talk to, and not to interact physically with, the subject. They were permitted to look at the subject but were asked not to stare at it.
For his/her first two trials, the helpful human knelt in front of the non-functional button and pressed this, before moving to the other, functional button, to press this on all remaining trials. The unhelpful human did the opposite, beginning with the functional button for the first two trials before switching to the non-functional button for all remaining trials. Each human began with the opposite button to which they were supposed to press, to facilitate the subject’s understanding that the human intentionally provided or intentionally did not provide food. The unhelpful human was required to press a non-functional button to ensure that any difference in the subjects’ propensity to press the button for the two humans in the test phase could not be due to differences in the activity level of the humans during the experience phase and could not be due to mimicking the behaviour of either one of the humans.
In between sessions, the subject was released from the enclosure for an approximately two-minute break in which they were free to roam around the room, interact with their owner or the experimenter, or drink from a bowl of water. The human remained in the enclosure during these breaks. After being released from the enclosure after each session, the subjects were provided with additional pieces of food in a food bowl. After sessions with the helpful human, they were provided with three to five pieces of food. However, after sessions with the unhelpful human, they received approximately 50 pieces of food to ensure that they received roughly the same amount of food independent of their experience with the human partner.
The test phase consisted of two test days, one with the helpful human and one with the unhelpful human. The order in which subjects were tested with the helpful and unhelpful humans was counterbalanced across subjects and was independent of the order in which they experienced these two humans in the experience phase.
Each test day consisted of three conditions, four motivational sessions, and a free interaction session (described below). The three conditions included the test condition, the social facilitation control, and the asocial control (see
A, test condition; B, social facilitation control; C, asocial control. In the test condition, each time the subject pressed the button, food was released from the dispenser and the human ate this. In the social facilitation control, each time the subject pressed the button, food was released from the dispenser but the human could not access it. In the asocial control condition the human was not present. Each time the button was pressed in this condition, food was released but no human ate it.
In the test condition, the human knelt in the middle of the human enclosure, facing the dog’s enclosure (see
The social facilitation control was similar to the test condition except that the food dispenser was positioned outside the subject’s enclosure (see
The asocial control matched the test condition except that the human was not present in the enclosure or in the room (see
The order in which subjects experienced the three conditions was randomized across subjects. However, the order in which a particular subject experienced these three conditions was the same on both test days.
Each condition on the test day was preceded and succeeded by a motivational session in which the human partner left the room and the food dispenser was placed in the subject’s enclosure for five trials such that the subject could reward itself five times. In total, there were four motivational sessions on each test day. There was an approximately one-minute break in between motivational sessions and experimental conditions. During these breaks, the experimenter changed the setup as necessary and the dog was free to explore the room and drink water.
Each of the test days ended with a five-minute free interaction session in which the subject was free to roam around the room and interact with either the owner or the human with which it was tested on that particular day.
Two circles were marked with tape on the floor of the room, each with an approximately two-metre diameter, and with approximately one metre separating them (see Figs
The owner and the human partner with which the subject was tested on that day (helpful or unhelpful) knelt opposite each other in the middle of circles that were marked on the floor with tape (diameter = 2 m). The subject was released by its owner to roam free in the room, and to interact freely with the human partner or its owner, for 5 minutes.
Behaviour coding was performed using Solomon coder (version beta 19.08.02 [
The latency to reach proximity to the test human and the duration spent in proximity to the test human in the free interaction session were also coded. A subject was considered in proximity to the human when at least one of its paws had crossed the line marking the circle.
A single video recording of one free interaction session for one subject was unavailable due to a technical malfunction. Therefore, this session could not be included in the final analysis. One person coded all the videos and a second experimenter coded approximately 20% of the test condition videos and free interaction session videos to assess interobserver reliability.
All models were fitted in R (version 3.6.2–4.0.2 [
We included random intercept effects of subject (i.e. the identity of the dog), human (the identity of the helpful or unhelpful human), the dog-human dyad, and observation (i.e. the identity of each individual observation; hereafter “observation level random effect”). The observation level random effect was included to account for session to session variation in the propensity to press the button. In addition to random intercepts, to avoid overconfidence regarding the precision of estimates for the fixed effects, and to ensure that the type
We fitted the model using the function “glmer” from the package “lme4” (version 1.1–23 [
We confirmed that the best linear unbiased predictors (BLUPs) did not deviate from a normal distribution by visual inspection of histograms [
As an overall test of the effects of condition, human type, and their interaction, with the aim of avoiding cryptic multiple testing, we conducted a full-null model comparison [
To assess interobserver reliability, we calculated the intraclass correlation coefficient using the function “icc” in the package “irr” (version 0.84.1 [
As only one frailty term could be included in the model, we fitted two models, one with subject as the frailty term and one with human as the frailty term. We compared the log-likelihood and model complexity of the two models and selected the model with human as the frailty term for further inference (log-likelihoods: model with subject: -106.1143 [
We fitted the model using the function “coxph” in the package “survival” (version 3.1–8 [
We tested the assumption of proportional hazards using the “cox.zph” function from the package “survival” (version 3.1–8 [
As an overall test of the effect of human type, we conducted a full-null model comparison [
We included the fixed effect of human type as the primary term of interest. To control for their possible effects, we also included test day order and first human experience as fixed effects. We included random intercept effects of subject and human. No random slopes were identifiable for this model; therefore, no random slopes were included.
We fitted the model using the function “glmmTMB” from the package “glmmTMB” (version 1.0.0 [
The model was not overdispersed (dispersion parameter: 1.034). We determined VIFs [
A full-null model comparison was carried out, as above, as an overall test of the effect of human type. The sample for this model included a total of 41 observations across 21 subjects and 12 humans. Interobserver reliability for duration of proximity to the test human was excellent (ICC = 0.979, nobservations = 8, nraters = 2,
Overall, neither condition, human type, nor the interaction between condition and human type had a significant effect on the proportion of times the subjects pressed the button (full-null model comparison: χ2 = 4.388,
Pressing for the helpful human is represented in white and pressing for the unhelpful human is represented in grey. Boxes display the interquartile range, black horizontal bars represent the median, whiskers represent the range of data points within 1.5 times the interquartile range from the upper and lower hinge, and circles represent individual data points. T, test; SFC, social facilitation control; AC, asocial control.
There was no significant difference in the latency to approach the human in the interaction session based on whether they had been helpful or unhelpful (full-null model comparison: χ2 = 0.324,
Human type did not have a significant effect on the proportion of time subjects spent in proximity to the human (full-null model comparison: χ2 = 0.035,
Boxes display the interquartile range, black horizontal bars represent the median, whiskers represent the range of data points within 1.5 times the interquartile range from the upper and lower hinge, and circles represent individual data points.
Dogs did not reciprocate the help received from the humans in this study. There are two aspects of the setup which might have hindered the emergence of any reciprocity. First, there was a considerable time lag between the experience phase and the test phase for most subjects. For example, the number of days from the last experience phase session to the first test phase session ranged from 1 to 22 with a mean (± SD) of 8.5 (± 4.5) days and a median of 7 days. This could have placed large demands on the subjects’ memory, a cognitive capacity which has been identified by other authors as a major factor constraining the emergence of reciprocity [
Second, the procedure during the experience phase, with both humans pressing a button, may have introduced too much complexity such that it confused the subjects or at least reduced the likelihood of them interpreting the actions of both humans accurately. We, therefore, conducted a second study to address these potential issues.
Study 2 largely followed the design and approach of study 1. There were, however, two major differences introduced in order to simplify the setup.
The first difference relates to the experience phase. During the experience phase in study 2, the human was presented with only one button, in contrast to study 1 in which the human was presented with two buttons. The human knelt in the middle of the human enclosure, facing the dogs’ enclosure. The button was presented to the human in a box, as it was presented to subjects in study 1. The experience phase consisted of two sessions each of ten trials (a trial being defined by the opening and closing of the box). On each trial, the experimenter opened the box for five seconds. In the experience phase with the helpful human, the human pressed the button after three seconds of the box opening. This resulted in food being released from the dispenser for the subject. In experience phase sessions with the unhelpful human, the button was not pressed upon opening of the box. Therefore, the subject did not receive any food from the food dispenser. As in study 1, the experience phase with each human type was conducted on a separate day and the order in which the two human types were experienced was counterbalanced across subjects.
In between the two sessions of the experience phase with each human, there was an approximately two-minute break in which the subject was released from the enclosure and was free to interact with its owner, the experimenter, or to drink water. Subjects were not provided with food in between experience phase sessions; thus, the amount of food received in each experience phase session differed.
The second major difference in the design of study 2 is that subjects were given the opportunity to reciprocate on the same day that they had the experience phase with a particular human type. The test conditions proceeded exactly as described for the test day in study 1, with a motivational session preceding and succeeding each condition. Other differences between the studies are identified below. In between the experience phase and the test phase, subjects were given an approximately two-minute break while the setup was changed. Subjects were free to roam around the room and drink water in this time.
This study also tested 21 subjects (11 females; 10 males) from a range of breeds, including mixed breeds (see
Subject ID | Age (years) | Sex | Breed | No. of sessions to reach criterion |
---|---|---|---|---|
A1 | 6 | F | Australian shepherd | 1 |
B1 | 10 | F | Border collie | 1 |
C1 | 7 | F | Border collie | 1 |
D1 | 12 | F | Border collie | 1 |
E1 | 8 | F | Border collie | 1 |
F2 | 12 | M | Border collie | 1 |
G2 | 5 | M | Border collie | 1 |
H2 | 8 | F | Australian shepherd | 2 |
I2 | 7 | F | Mix | 2 |
J2 | 5 | M | Border collie mix | 1 |
K2 | 7 | F | Golden retriever | 1 |
L2 | 6 | M | Bernese mountain dog | 1 |
M2 | 10 | M | Labradoodle | 2 |
N2 | 4 | F | Mix | 2 |
O2 | 5 | M | Border collie | 1 |
P2 | 2 | M | Border collie | 1 |
Q2 | 10 | M | German shepherd | 2 |
R2 | 6 | F | Bernese mountain dog | 2 |
S2 | 7 | F | Border collie mix | 1 |
T2 | 3 | M | Petit Brabançon | 1 |
U2 | 5 | M | Shetland sheepdog | 1 |
F, female; M, male.
In the social facilitation control of this study, the human remained in the enclosure; however, rather than placing the food dispenser outside the dog’s enclosure, as in study 1, the food dispenser was removed from the setup completely and turned off. If the subject pressed the button in this condition, the human would not receive any food. This control condition differed from that in study 1, as it was deemed plausible that the novel position of the food dispenser, which was closer to the subject than to the human in study 1, might have altered subject’s motivation by giving them the impression they had a better chance of obtaining the food compared with the test condition. The curtain also remained in place on the side of the subject’s enclosure in the social facilitation control of study 2.
Two additional training stages were included at the end of the training phase of this study. The first new training stage matched the social facilitation control in terms of the setup (i.e. the subject had access to the button in its enclosure, and a human was present in the human enclosure but the food dispenser was not present). Ten trials were conducted with this setup. The subject was not required to press the button on any trial to proceed to the next stage. The next stage of training mimicked the asocial control condition in that the food dispenser was in the human’s enclosure but no human was present. This also matched stage 5 of training. Again, ten trials were conducted and the subject was not required to press on any trial to proceed to the experiment. These training stages were both preceded and succeeded by a motivational session. The reason for inclusion of these new training stages was to ensure subjects had experienced each possible condition they would encounter in the test setting.
Three experimenters coded the videos. All three experimenters also coded the same 23% of videos (approximately) for interobserver reliability coding.
Analysis of the effect of human type, condition, and their interaction, on the proportion of times subjects pressed the button matched that for study 1 with a few differences. For instance, the fixed effect of “first human experience” (i.e. whether the subject experienced the helpful or unhelpful human in their first experience phase) was excluded from this model and all other models in study 2. The reason for this exclusion was that the experience phase and test phase with a particular human took place on the same day, making first human experience redundant with the fixed effects of human type (helpful or unhelpful) and test day order (i.e. whether it was the first or second test day). In addition, a random slope of first human experience was not included within the random effect of human, and a random slope of condition was not included within the random effect of the dog-human dyad.
As with study 1, the full model here could only be fitted after exclusion of the correlations among random intercepts and random slopes. Also, for our full-null model comparison we chose the model which lacked the random slope of the interaction between condition and human type within the random effect of subject (log-likelihoods: model with complete set of random effects: -267.8015 [
The BLUPs did not deviate from a normal distribution, as confirmed by visual inspection of histograms [
The sample for this model included a total of 126 observations across 21 subjects, 24 humans, and 42 dyads. Interobserver reliability for the number of times the button was pressed was excellent (ICC = 0.994, nobservations = 29, nraters = 3,
The effect of human type on latency to approach the human was analysed in the same manner as study 1. We also selected the model with human as the frailty term, though the exact log-likelihood values differed (log-likelihoods: model with subject: -111.1651 [
The effect of human type on the proportion of time subjects spent in proximity to the human in the free interaction session was analysed in the same manner as study 1 (see
Overall, neither condition, nor human type, nor the interaction between condition and human type had a significant effect on the proportion of times the subjects pressed the button (full-null model comparison: χ2 = 8.856,
Pressing for the helpful human is represented in white and pressing for the unhelpful human is represented in grey. Boxes display the interquartile range, black horizontal bars represent the median, whiskers represent the range of data points within 1.5 times the interquartile range from the upper and lower hinge, and circles represent individual data points. T, test; SFC, social facilitation control; AC, asocial control.
There was no difference in the latency to approach the human in the interaction session based on whether they had been helpful or unhelpful (full-null model comparison: χ2 = 1.493,
Human type did not have a significant effect on the proportion of time subjects spent in proximity to the human in the interaction session (full-null model comparison: χ2 = 0.130,
Boxes display the interquartile range, black horizontal bars represent the median, whiskers represent the range of data points within 1.5 times the interquartile range from the upper and lower hinge, and circles represent individual data points.
Despite amending the procedure, dogs still did not reciprocate help received from humans. Although we reduced the complexity of the experience phase with regards to the activity of the humans and the number of buttons available, the new design may have introduced a confound: as the button was presented to the human in the same box as that used for subjects, it would have been difficult for subjects to see the button during the experience phase. Thus, it may not have been clear to the subjects that the human was pressing the button. Although this is plausible, the subjects could also see the movements of the human’s arm, they were familiar with the sounds of the box opening and the sound emitted by the button when pressed, and dogs are capable of geometric perspective taking [
It is, nonetheless, conceivable that subjects did not pay enough attention to the humans in the experience phase of either study to permit recognition of the action of the human. We investigated this possibility further by determining whether gazing at the human occurred at all. We focused on the experience phase of study 1 due to better video quality and, more specifically, the experience phase with the helpful human, as it is arguably the more important human to be attentive to in order to reciprocate.
Visual inspection of plotted data representing the proportion of time spent gazing at the helpful human immediately before button pressing (see
We also fitted a model to investigate whether the proportion of time spent gazing at the human changed across trials (see
Regarding overall performance in the test phase, although we did not compare the results of the two studies statistically, median performance in terms of the number of times subjects pressed the button appeared lower in study 2 than in study 1 (see Figs
Across the two studies conducted here, dogs failed to reciprocate help received from the human. In addition, dogs did not appear to develop a preference for a particular human type (i.e. helpful or unhelpful) in the free interaction session, suggesting that either they did not distinguish between the two humans based on helpfulness or they did distinguish between them but developed no preference.
Our assessment of subjects’ preference using the free interaction sessions was admittedly opportunistic and was also indirect, as subjects were never faced with a choice between the helpful and unhelpful human. As a result, our approach is arguably less sensitive than a direct preference test. However, the free interaction session offered the advantage, at least in study 2, of observing subjects interacting with the respective human at a standardized time point after the experience phase and test phase. Had a direct preference test been conducted, the experience with one human would have been more recent than the experience with the other, potentially biasing subjects’ choice. Moreover, free interaction sessions like ours have been used to observe differential approach behaviour towards an experimenter based on experimental treatments immediately prior in inequity aversion studies [
There are numerous potential explanations for the lack of reciprocity in the current studies, primarily related to the experimental approach. For example, the use of food as the currency here may have been inappropriate, as the provision of food in the dog-human relationship is typically asymmetrical, that is, humans usually provide pet dogs with food but dogs rarely provide humans with food. A similar explanation has been put forward (see Carballo et al. [
The lack of a preference for either human in the free interaction session, in both studies, suggests that the lack of reciprocity may have little to do with dogs’ propensity to provide humans with food and more to do with a failure to recognize the helpful act of the human. A variety of interrelated explanations might account for dogs’ failure to recognize the cooperative act of the human. First, the method of food provision in our studies involved an effectively abstract connection between a button and a food dispenser. Most studies demonstrating successful food provisioning in dogs so far have used methods involving clear physical relationships (e.g. a rope connected to a tray [
Two experimental studies have demonstrated prosociality in dogs using abstract connections similar to ours. First, Bräuer et al. [
Two recent prosociality studies characterized by similarly abstract connections between the required behaviour and the food delivery (one based on a touch screen task [
An additional possible explanation for subjects’ potential lack of registering the cooperative act of the human, is that the behaviour performed by the human was not a natural cooperative behaviour. There are two clear aspects common to natural cooperative interactions which were missing here. First, typical behaviours that are performed prosocially and reciprocated in nature include those such as grooming and active food sharing [
Second, natural interactions typically involve communicative signals between interacting individuals. In our studies, the human refrained from active communication with the subject throughout the experience phase. The absence of natural communication was also highlighted by Quervel-Chaumette et al. [
Although both of the above arguments (i.e. the absence of a natural cooperative behaviour and physical contact, and the absence of natural communication) seem quite plausible, particularly in combination, it would be surprising to some extent if these accounted for our findings in whole. In both studies of Gfrerer & Taborsky [
Finally, it is possible that, as a result of the training, the button and the dispenser were more salient than the human and occupied most of the dogs’ attention throughout both studies. Consequently, the dogs may not have paid enough attention to the human to register fully the relevance of the human’s actions, even though most subjects did at least look at or see the human to some degree. Increasing the salience of the human and the human’s actions could overcome this issue and facilitate the subjects’ registering of the cooperative act.
In this context, it is conceivable that the limited training and exposure to the action of the humans also diminished the dogs’ likelihood of perceiving or understanding the human’s action. Gfrerer and Taborsky’s [
In the context of learning about crucial aspects of the setup, it is worth noting that to reciprocate directly in our setup, dogs would have needed to discriminate between the helpful and unhelpful human. The discrimination learning required may have been too demanding for such a small number of sessions. Visual and olfactory discrimination learning in dogs can take a considerable number of training sessions [
Despite the apparent difficulty of discrimination learning, however, it is worth keeping in mind that dogs were previously shown by Gfrerer and Taborsky [
The explanations presented here for the failure of subjects to reciprocate tend to imply that dogs would have reciprocated were it not for methodological issues. However, it is possible that dogs do not have a predisposition to engage in reciprocal cooperation with humans naturally. In fact, a recent study comparing inequity aversion in dog breed groups that vary in the extent to which they were selected for cooperation with humans, suggested that dog-human cooperation may not involve this form of reciprocity [
In future studies, the familiarity of the partners may also need to be considered. Our use of unfamiliar partners prevented results from being confounded by previous experience with those specific individuals. However, familiarity appears to influence prosociality in dogs, at least in experimental tasks with conspecifics [
In conclusion, dogs in the current studies failed to reciprocate help received from humans and a preference for either human type (helpful or unhelpful) was not evident based on a free interaction session. Given that dogs have already been shown to reciprocate help received from conspecifics in experimental studies, the absence of reciprocity here may be explained by methodological inadequacies, though it is also possible that dogs are not predisposed to engage in such cooperative interactions with humans naturally. Future studies investigating direct reciprocity between dogs and humans may benefit from the application of rescuing as the cooperative behaviour required of the dog, and a natural behaviour such as food-giving or grooming, involving physical contact and active communication, as the cooperative behaviour provided by the human. In addition, a more extensive training procedure than that applied here, including the repeated exchange of roles between the dog and the human may facilitate the emergence of reciprocity. Moreover, increasing the difference between the helpful and unhelpful partners could aid the subjects in discriminating between them. Finally, repetition of the current setup but with intraspecific rather than interspecific interactions would help determine whether the current results reflect dogs’ disinclination to reciprocate help received from humans, or whether they are best explained by methodological issues.
Model stability plot for the proportion of times subjects pressed the button, representing the range of model estimates for each term in the model when the levels of the random effects are excluded one at a time. Names including an ampersat symbol (@) refer to random effects: the first term in the name is the grouping variable; the term after the first ampersat is either a random intercept (indicated by “(Intercept)”) or a random slope.
(SVG)
Scaled Schoenfeld residuals [
(SVG)
Dfbeta values [
(SVG)
Model stability plot for duration of proximity to the human, representing the range of model estimates for each term in the model when the levels of the random effects are excluded one at a time. Names beginning with “cond” refer to the mean part of the model. The name beginning with “disp” refers to the dispersion part of the model. Names including more than one ampersat symbol (@) refer to random intercepts.
(SVG)
Model stability plot for the proportion of times subjects pressed the button, representing the range of model estimates for each term in the model when the levels of the random effects are excluded one at a time. Names including an ampersat symbol (@) refer to random effects: the first term in the name is the grouping variable; the term after the first ampersat is either a random intercept (indicated by “(Intercept)”) or a random slope.
(SVG)
Scaled Schoenfeld residuals [
(SVG)
Dfbeta values [
(SVG)
Model stability plot for duration of proximity to the human, representing the range of model estimates for each term in the model when the levels of the random effects are excluded one at a time. Names beginning with “cond” refer to the mean part of the model. The name beginning with “disp” refers to the dispersion part of the model. Names including more than one ampersat symbol (@) refer to random intercepts.
(SVG)
Transparent dots represent data points. The solid line shows the fitted model and the dotted lines show its 95% confidence interval.
(SVG)
Transparent dots represent data points. The solid line shows the fitted model and the dotted lines show its 95% confidence interval.
(SVG)
(DOC)
(XLSX)
(MP4)
We would like to thank the dogs and their owners for participation in this study. We are also very grateful to all those who participated in the study as helpful and unhelpful human partners. We thank Roger Mundry for statistical support. We also thank Karin Bayer, Jennifer Bentlage, Karoline Bürger, Aleksandar Orlic, Brigitte Pavlik, and Christina Szaga-Doktor for administrative support, as well as Peter Füreder and Wolfgang Berger for technical support.
PONE-D-20-35864
Dogs fail to reciprocate the receipt of food from a human in a food-giving task
PLOS ONE
Dear Dr. McGetrick,
Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.
It was reviewed by two experts in the field and they have recommended a few modifications be made prior to acceptance.
I therefore invite you to consider the suggested changes and resubmit your manuscript, along with a response to reviewers to aid re-revision.
Please submit your revised manuscript by Jan 25 2021 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at
Please include the following items when submitting your revised manuscript:
A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'. A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'. An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.
If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.
If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see:
We look forward to receiving your revised manuscript.
I wish you the best of luck with your revisions.
Hope you are keeping safe and well in these difficult times.
Kind regards,
Simon Clegg, PhD
Academic Editor
PLOS ONE
Journal requirements:
When submitting your revision, we need you to address these additional requirements.
1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at
Reviewers' comments:
Reviewer's Responses to Questions
1. Is the manuscript technically sound, and do the data support the conclusions?
The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.
Reviewer #1: Yes
Reviewer #2: Yes
**********
2. Has the statistical analysis been performed appropriately and rigorously?
Reviewer #1: Yes
Reviewer #2: Yes
**********
3. Have the authors made all data underlying the findings in their manuscript fully available?
The
Reviewer #1: Yes
Reviewer #2: Yes
**********
4. Is the manuscript presented in an intelligible fashion and written in standard English?
PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.
Reviewer #1: Yes
Reviewer #2: Yes
**********
5. Review Comments to the Author
Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)
Reviewer #1: This is an interesting, relevant and original work. The manuscript is clear and well-written and the authors have tried to do a thorough analysis of the results. They found that dogs were not able to reciprocate the help received from humans. Even when the lack of this ability in a food-giving task is a possibility that this work as well as other previous studies have suggested, the limitations in the experimental designs force us to be cautious.
The authors clearly described and analyzed some of these limitations. However, another plausible and parsimonious explanation of the results is that dogs were repeatedly reinforced throughout the procedure for pressing the button. So, they could continue pressing the button, just because this behavior had a strong association with the food, regardless of the rest of the conditions. In addition, considering that reinforced and not reinforced trials were mixed (because of the motivational sessions) in the design, dogs were exposed to a partial reinforcement schedule, which increases the resistance to extinction. In this way, the association between pressing the button and food could be higher than anything else. Moreover, in order to successfully solve the task, the dogs must perform a discrimination learning in which they have to inhibit their learned response of pressing the button in some contexts. This kind of learning usually requires long trainings. You have to state this limitation in the discussion.
Anyway, I think this problem always appears when you have to train the behavior required to help the other individual. For this reason, it is better to choose a response that is already part of the repertoire of the animals.
On the other hand, one crucial aspect in this kind of experiments is whether dogs are able to understand the situation. In this sense, an overshadowing phenomenon may have occurred. As a consequence of the training, the device (button and dispenser) was very salient. So, when dogs pressed the button, it is likely that they were paying more attention to the device than to the humans. If they paid little or no attention to them, the likelihood of learning about their behavior is low. Do you have any assessment of the behavior of the dogs during the conditions in which humans were present? Can you assess gaze duration towards humans?
As far as I understand, you did not perform a preference test between the helpful and the unhelpful human, so you cannot conclude that dogs were not able to discriminate them. Your comparison is indirect and therefore less sensitive. The length of time in which dogs were near each person compared to the owner is not a good indicator of the helper-unhelper discrimination. For example, the presence of the owner during the free interaction session could interfere, decreasing the negative reaction towards the unhelpful human, decreasing the interest in the unfamiliar person, etc. You have to relativize your conclusions in this regard.
It is necessary to include some important information about the characteristics of the sample: the description of the breeds and the level of previous training.
Which was the reward used? I imagine that the helpful human gave the dog a better reward than the one the dog obtained by itself during training and motivational trials.
I can see by comparing Fig 6 and 9 that the performance of the dogs decreased between the two studies. Can you explain this difference?
I think that the legend of Fig 1 is too long and includes many details that are already included in the text. I suggest reducing it.
Please describe the interval between phases.
L 324-325 I do not understand the objective of modifying the first trial procedure.
L416 “The video for one of the free interaction sessions was unavailable” Do you mean the video from one dog or the videos of all the dogs during one session? In this case, which was the session?
426 and 551 vs 427. Did you use the number of times or the proportion of times?
Reviewer #2: General comments
In the present study authors tested if dogs would reciprocate a pro-social act from an unknown human in a context involving food, using a novel experimental design adapted from previous studies in dogs and other species. In the first study the authors did not find any evidence of reciprocity in dogs and speculate that this result could have been due to the complexity of the task or the mnemonic demands imposed to the dogs by the time gap between the training and testing phases. Thus, in a second study the authors simplify the task and run the training and testing phases in the same day. Again, dogs failed to reciprocate the human pro social behavior in this simplified version of the task. These findings led the authors to conclude that dogs do not spontaneously reciprocate the human pro-social behavior in contexts involving food.
I want to congratulate the authors for their work; I found the paper very interesting and carefully design. I will suggest some minor revisions that could probably improve the paper.
Specific comments
Introduction
The review of the previous literature is precise updated and well written. Nevertheless:
L. 43-44. I would remove the phrase. “Reliable models of reciprocity that can be
manipulated experimentally are scarce”. There is a vast literature on reciprocity on rats, mice, and insects. And I believe that there are all valuable models for the study of reciprocity (e.g. Taborsky’s works).
L. 89 -102. I found this paragraph a bit unnecessary. Although, as mention, all the listed characteristics made the dog a good model for the study of pro-sociality in general these are not specific characteristics needed for reciprocity. I would focus in reciprocity (See. Trivers, 1971).
L. 160-161. In line with the previous comment. Some of the aspects that would enhance direct reciprocity are: the possibility of future encounters (Trivers, 1971), and the bond between the subjects (i.e. De Wal, 1997; Freidin et al., 2013). So I think that one big difference with some previous studies (Quervel-Chaumette et al., 2015; 2016, Sanford et al., 2018; Carballo et al., 2020) is the identity of the receiver. Most evidences showing spontaneous pro social behavior in dogs (or the lack of it, Quervel-Chaumette 2016) came from studies with the owners, or some familiar human, as partner. I understand the methodological choice of the authors to use unknown humans as partners which eliminate a lot of uncontrolled variations that could mixed the results. But I believe that the best solution would be to run a third study whit the owners as partners and counterbalance their role (Helper vs. Unhelper) between subjects. The inclusion or non-inclusion of this third study would not influence in my personal decision to suggest the publication of the paper. I simply mention that this would be an interesting aspect to take into account and probably the authors would like to include a further study or mention this aspect in the conclusion.
Methods and analysis
I found the methodology very well design and implemented. I think the data presented by the authors in the data set of supplementary material and the variables analyzed are clear and with a straightforward interpretation. So I found the statistical analyses unnecessarily a bit too complex for the data and the number of factors taken in consideration. But authors may well prefer to keep the presented analyses.
L. 261. Who was the human who eat the food?
Discussion
Again, I found the discussion well written and pertinent. And I mostly agree with the authors’ interpretation of the data. Nevertheless, I would include some references to the role of learning mechanisms in the development of pro social behavior in dogs and in reciprocity in particular. The authors mentions this when they speculate that with further training and switching roles between dogs and humans dogs probably could reciprocate the human but they do not provide and explicit theoretical link. For example, dogs could be focusing in some other aspect of the experimental situation which could overshadow the role of the helper, thus making the actions of the helper more salient could improve dogs performance. Furthermore, the discriminations between unknown humans could be demanding for dogs, thus increasing the difference between helper and un-helper could also improve dogs’ performance or the testing could be done with familiar humans. These aspects are related to the factors that could facilitate reciprocity that I mention earlier (familiarity, bonding, repetition of the encounters etc.) and are usually taken into account when discussing pro-social behavior in dogs.
L. 885: include “direct” before reciprocity.
**********
6. PLOS authors have the option to publish the peer review history of their article (
If you choose “no”, your identity will remain anonymous but your review may still be made public.
Reviewer #1: No
[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]
While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool,
Reviewer #1: This is an interesting, relevant and original work. The manuscript is clear and well-written and the authors have tried to do a thorough analysis of the results. They found that dogs were not able to reciprocate the help received from humans. Even when the lack of this ability in a food-giving task is a possibility that this work as well as other previous studies have suggested, the limitations in the experimental designs force us to be cautious.
The authors clearly described and analyzed some of these limitations. However, another plausible and parsimonious explanation of the results is that dogs were repeatedly reinforced throughout the procedure for pressing the button. So, they could continue pressing the button, just because this behavior had a strong association with the food, regardless of the rest of the conditions. In addition, considering that reinforced and not reinforced trials were mixed (because of the motivational sessions) in the design, dogs were exposed to a partial reinforcement schedule, which increases the resistance to extinction. In this way, the association between pressing the button and food could be higher than anything else. Moreover, in order to successfully solve the task, the dogs must perform a discrimination learning in which they have to inhibit their learned response of pressing the button in some contexts. This kind of learning usually requires long trainings. You have to state this limitation in the discussion.
We have now included two paragraphs in the discussion on discrimination learning. The first points out that discrimination learning would have been required for successful reciprocation here and that discrimination learning requires many sessions in studies with dogs. The second paragraph points out that although discrimination learning clearly takes a long time in dogs based on discrimination learning studies, dogs were able to discriminate between a cooperative and uncooperative partner after a very small number of trials in some studies:
L899 – 922: In the context of learning about crucial aspects of the setup, it is worth noting that to reciprocate directly in our setup, dogs would have needed to discriminate between the helpful and unhelpful human. The discrimination learning required may have been too demanding for such a small number of sessions. Visual and olfactory discrimination learning in dogs can take a considerable number of training sessions (97,105,106). In fact, in a discrimination learning task in which dogs had to discriminate between their owner and a familiar person based on visual information from the faces and heads, an average of 6.5 sessions, each lasting ten trials, were required for dogs to successfully discriminate (107). Similarly, dogs were found to require between three and eight sessions of 20 trials each to successfully discriminate between their owner’s and a stranger’s face (108). Moreover, discrimination between unfamiliar humans in our studies may have been particularly demanding. Increasing the perceptual differences between the humans may help with learning the discrimination in future studies.
Despite the apparent difficulty of discrimination learning, however, it is worth keeping in mind that dogs were previously shown by Gfrerer and Taborsky (35) to reciprocate after just 14 trials of experience with each partner type. It is plausible that the more extensive training procedure, outlined above, facilitated discrimination by priming subjects to attend to the partner. Furthermore, partner discrimination may be easier if the partners are conspecifics. However, as subjects received an experience phase and test phase with a single partner in a two-day sequence before encountering a new partner (35), discrimination learning may not have been necessary to facilitate reciprocity. Nonetheless, Heberlein et al. (56) and Carballo et al. (57) found that dogs could discriminate between humans that differed in cooperativeness after just six and twelve trials respectively. As subjects in our studies were presented with each human across at least 20 trials, it is not clear that the challenge of discrimination learning can account for the lack of reciprocity.
It is plausible that the training procedure caused the button and dispenser to become very salient for the dogs, thereby reducing their attentiveness to the human. We have included discussion points on this (see below). However, we did not get ceiling effects with regards to the number of times the dogs pressed the button. In fact, in study 1 and study 2 performance was relatively low e.g. median number of presses of ~7 – 10 in study 1 and ~2 – 4 in study 2, indicating that subjects were well able to inhibit pressing. Furthermore, at least 2 other studies demonstrating prosociality in dogs in food-giving tasks used similar training procedures to ours (see Quervel-Chaumette et al. 2015 and Dale et al. 2016) and the training did not appear to inhibit the subjects’ ability to make the decision to provide the partner with food. Thus, we do not believe the reinforcement schedule had a major impact on the outcome of either study.
Anyway, I think this problem always appears when you have to train the behavior required to help the other individual. For this reason, it is better to choose a response that is already part of the repertoire of the animals. On the other hand, one crucial aspect in this kind of experiments is whether dogs are able to understand the situation. In this sense, an overshadowing phenomenon may have occurred. As a consequence of the training, the device (button and dispenser) was very salient. So, when dogs pressed the button, it is likely that they were paying more attention to the device than to the humans. If they paid little or no attention to them, the likelihood of learning about their behavior is low. Do you have any assessment of the behavior of the dogs during the conditions in which humans were present? Can you assess gaze duration towards humans?
We have now coded the duration of gaze at the helpful human in the experience phase of study 1 for the periods immediately before button pressing and immediately after button pressing. We fitted a model to analyse the effect of trial number on gazing. The following details covering the video coding and analysis are included in S1 Appendix:
Gazing at the helpful human in the experience phase
Methods
Coding
The duration of gaze at the helpful human in experience phase sessions was coded. Instances of looking at or seeing any part of the human’s body (e.g. seeing the human’s hand on the button) were coded. Coding for gaze duration was divided among five experimenters and for interobserver reliability, all five experimenters coded the same 20% of the videos. Six trials from one session with one subject were missing due to the absence of a video recording.
Statistical analysis
For each trial we determined the proportion of time spent gazing at the helpful human in the 3 second period immediately before button pressing and in the 3 second period immediately after button pressing. For one single trial with one subject, there were less than 3 seconds on video prior to button pressing; therefore, the total duration used to determine the proportion was adjusted.
To determine whether trial number had an influence on the proportion of time subjects spent looking at the helpful human before button pressing, we fitted a Generalized Linear Mixed Model (GLMM) with a beta error distribution and a logit link function (1–3). As such beta models cannot handle zeroes and ones, and our dataset comprised both, we transformed the response using the following formula (where “x” is the variable to be transformed; see Smithson and Verkuilen, 3):
x’ = x.(length(x)-1) + 0.5
length(x)
We included the fixed effects of trial and trial squared. We included random intercept effects of subject and human, and random slopes of trial and trial squared were included within both.
We fitted the model using the function “glmmTMB” from the package “glmmTMB” (version 1.0.2.1; 4). Prior to fitting the model, and prior to squaring trial number, we z-transformed trial number to a mean of zero and a standard deviation of one to allow for an easier interpretation of results and to ease model convergence. All correlations between random slopes and random intercepts were removed due to convergence issues. The model was not overdispersed (dispersion parameter: 0.6989). A full-null model comparison was carried out, as an overall test of the effect of trial number and trial number squared on the proportion of time spent looking at the helpful human before button pressing.
Confidence intervals of model coefficients were derived using 1,000 parametric bootstraps using the function “simulate” of the package “glmmTMB” (version 1.0.2.1; 4) and a wrapper kindly provided by Roger Mundry. Tests of the individual fixed effects were derived with likelihood ratio tests (5) by using the R function “drop1” and setting the argument “test” to “Chisq”.
The effect of trial on the duration of gaze at the human after button pressing was analysed using an identical model structure; however, no random slopes were included due to convergence issues. This model was not overdispersed (dispersion parameter: 0.8290). The sample for both models included a total of 1,044 observations across 21 subjects and 12 humans. Interobserver reliability for duration of gaze at the helpful human before and after button pressing was moderate (before: ICC = 0.725, nobservations = 193, nraters = 5, P < 0.001; after: ICC = 0.61, nobservations = 193, nraters = 5, P < 0.001). Models were plotted using R (version 4.0.2; 6).
Results
Proportion of time spent gazing at the human before button pressing
Overall, there was a significant effect of the fixed effects on the proportion of time spent gazing at the helpful human before button pressing (full-null model comparison: χ2 = 11.437, df = 2, P = 0.003). More specifically, with a greater number of trials, the proportion of time spent gazing at the helpful human increased (see S1 Table and S1 Fig).
S1 Table Results of the model analysing the effects of trial and trial squared on the proportion of time subjects spent gazing at the helpful human before button pressing in the experience phase. Estimates are presented along with standard errors, confidence intervals, and likelihood ratio test results.
Term Estimate SE Lower CI Upper CI χ2 df P
Intercept -0.794 0.214 -1.209 -0.391 - - -
Triala 0.290 0.084 0.117 0.460 8.792 1 0.003
Trial squareda -0.084 0.046 -0.174 0.010 2.668 1 0.102
aTrial was z-transformed to a mean of 0 and a standard deviation of 1 prior to inclusion in the model and prior to squaring.
Proportion of time spent gazing at the human after button pressing
Overall, a trend was revealed for the influence of the fixed effects on the proportion of time spent gazing at the helpful human after button pressing (full-null model comparison: χ2 = 5.569, df = 2, P = 0.062). There was a weak effect whereby the proportion of time spent gazing at the human increase with more trials (see S2 Table and S2 Fig).
S2 Table Results of the model analysing the effects of trial and trial squared on the proportion of time subjects spent gazing at the helpful human before button pressing in the experience phase. Estimates are presented along with standard errors, confidence intervals, and likelihood ratio test results.
Term Estimate SE Lower CI Upper CI χ2 df P
Intercept -2.310 0.126 -2.572 -2.083 - - -
Triala 0.065 0.032 0.002 0.130 4.019 1 0.045
Trial squareda 0.043 0.036 -0.026 0.111 1.408 1 0.235
aTrial was z-transformed to a mean of 0 and a standard deviation of 1 prior to inclusion in the model and prior to squaring.
We have discussed this in the discussion for study 2:
L744 – 761: It is, nonetheless, conceivable that subjects did not pay enough attention to the humans in the experience phase of either study to permit recognition of the action of the human. We investigated this possibility further by determining whether gazing at the human occurred at all. We focused on the experience phase of study 1 due to better video quality and, more specifically, the experience phase with the helpful human, as it is arguably the more important human to be attentive to in order to reciprocate.
Visual inspection of plotted data representing the proportion of time spent gazing at the helpful human immediately before button pressing (see S9 Fig) and immediately after button pressing (see S10 Fig) reveals that most subjects generally did look at or see the human in the experience phase for some duration, particularly before button pressing occurred. Much of this appeared to be due to the subjects already focusing on the button before the human pressed it, or due the subject tracking the human’s hand when he/she began to move it towards the button. Gazing at the human’s face did not seem common.
We also fitted a model to investigate whether the proportion of time spent gazing at the human changed across trials (see S1 Appendix for details). With increasing trial number, the proportion of time subjects spent looking at the helpful human before and after button pressing increased significantly. This finding may indicate that subjects were learning the importance of the human’s action.
We have now made reference in the general discussion to the possibility that the button and dispenser were more salient than the human and that the dogs did not pay enough attention to the human:
L880 – 885 (General discussion): Finally, it is possible that, as a result of the training, the button and the dispenser were more salient than the human and occupied most of the dogs’ attention throughout both studies. Consequently, the dogs may not have paid enough attention to the human to register fully the relevance of the human’s actions, even though most subjects did at least look at or see the human to some degree. Increasing the salience of the human and the human’s actions could overcome this issue and facilitate the subjects’ registering of the cooperative act.
We have also maintained our discussion from the original version of the manuscript which suggests that natural behaviours may be better to use (L843 – 879).
As far as I understand, you did not perform a preference test between the helpful and the unhelpful human, so you cannot conclude that dogs were not able to discriminate them. Your comparison is indirect and therefore less sensitive. The length of time in which dogs were near each person compared to the owner is not a good indicator of the helper-unhelper discrimination. For example, the presence of the owner during the free interaction session could interfere, decreasing the negative reaction towards the unhelpful human, decreasing the interest in the unfamiliar person, etc. You have to relativize your conclusions in this regard.
We do not agree fully with the argument here. There is no clear reason why the presence of the owner and their potential influence on the dogs’ approach behaviour in the free interaction session would be any different in a direct preference test. Furthermore, a free interaction session like ours has been used successfully in previous studies to observe differences in approach behaviour towards an experimenter at different time points after different experimental treatments. We have nonetheless included a discussion point to address the issue of a free interaction session vs a direct preference test:
L774 – 785: Our assessment of subjects’ preference using the free interaction sessions was admittedly opportunistic and was also indirect, as subjects were never faced with a choice between the helpful and unhelpful human. As a result, our approach is arguably less sensitive than a direct preference test. However, the free interaction session offered the advantage, at least in study 2, of observing subjects interacting with the respective human at a standardized time point after the experience phase and test phase. Had a direct preference test been conducted, the experience with one human would have been more recent than the experience with the other, potentially biasing subjects’ choice. Moreover, free interaction sessions like ours have been used to observe differential approach behaviour towards an experimenter based on experimental treatments immediately prior in inequity aversion studies (66,99). Thus, it seems likely that our assessment was sensitive enough to detect the subjects’ preference (or lack thereof) for either human.
We have also reworded parts of the discussion slightly in relation to the conclusions based on the free interaction session:
L770: …dogs did not appear to develop a preference for a particular human type….
L771 – 773: …suggesting that either they did not distinguish between the two humans based on helpfulness or they did distinguish between them but developed no preference.
L947 – 949: …and a preference for either human type (helpful or unhelpful) was not evident based on a free interaction session.
It is necessary to include some important information about the characteristics of the sample: the description of the breeds and the level of previous training.
We have now included tables with subject information for studies 1 and 2 (see Table 1 and Table 2):
Table 1. Subject information including age, sex, breed, and the number of training sessions required to reach the criterion.
Subject ID Age (years) Sex Breed No. of sessions to reach criterion
A1 5 F Australian shepherd 1
B1 10 M Border collie 1
C1 6 M Border collie 1
D1 11 F Border collie 1
E1 8 F Border collie 1
F1 7 M Greyhound 3
G1 5 F Terrier mix 1
H1 11 M Airedale terrier 3
I1 4 M Hungarian vizsla 3
J1 5 M Mix 1
K1 5 F Yorkshire terrier 3
L1 6 M Dachshund 1
M1 4 M Spanish galgo 3
N1 4 M Australian cattle dog mix 1
O1 2 M Labrador retriever mix 1
P1 1 F Border collie 1
Q1 7 F Podenco 2
R1 3 F Beagle 2
S1 3 M German hunting terrier mix 2
T1 5 M Australian shepherd mix 1
U1 9 F Australian shepherd 2
F, female; M, male.
Table 2. Subject information including age, sex, breed, and the number of training sessions required to reach the criterion.
Subject ID Age (years) Sex Breed No. of sessions to reach criterion
A1 6 F Australian shepherd 1
B1 10 F Border collie 1
C1 7 F Border collie 1
D1 12 F Border collie 1
E1 8 F Border collie 1
F2 12 M Border collie 1
G2 5 M Border collie 1
H2 8 F Australian shepherd 2
I2 7 F Mix 2
J2 5 M Border collie mix 1
K2 7 F Golden retriever 1
L2 6 M Bernese mountain dog 1
M2 10 M Labradoodle 2
N2 4 F Mix 2
O2 5 M Border collie 1
P2 2 M Border collie 1
Q2 10 M German shepherd 2
R2 6 F Bernese mountain dog 2
S2 7 F Border collie mix 1
T2 3 M Petit Brabançon 1
U2 5 M Shetland sheepdog 1
F, female; M, male.
Which was the reward used? I imagine that the helpful human gave the dog a better reward than the one the dog obtained by itself during training and motivational trials.
This has now been clarified in the manuscript:
L298 – 300: The food rewards used in the experience phase were dry food pieces. The type of dry food used in the experience phase for a subject was the same as that used in the training and motivational sessions for that subject.
I can see by comparing Fig 6 and 9 that the performance of the dogs decreased between the two studies. Can you explain this difference?
It is not clear why this difference exists. We have now included a discussion point on this in the discussion for study 2:
L762 – 767: Regarding overall performance in the test phase, although we did not compare the results of the two studies statistically, median performance in terms of the number of times subjects pressed the button appeared lower in study 2 than in study 1 (see Fig 6 and Fig 9). The reason for this difference is not clear, though it is possible that either the extra training steps in study 2, or having an experience phase and a test phase on the same day, as was the case in study 2, reduced motivation to press in general in the test phase.
I think that the legend of Fig 1 is too long and includes many details that are already included in the text. I suggest reducing it.
This has now been shortened:
L162 – 165: Fig 1. Layout of test room. Two circles (each approx. 1 m in radius) were marked on the floor of the room, for the free interaction session. A red water bowl was also present in the room. A food dispenser and button are depicted in the middle of the room (note: the dispenser and button were only in this position for the first stage of training). Black curtains surrounding some of the fences are represented by thick black lines.
Please describe the interval between phases.
It is not clear to which part of the manuscript this refers specifically; however, we have added a description of the break between motivational sessions and test conditions in study 1 and of the break between the experience phase and test phase of study 2:
Description of break between motivational sessions and test conditions (study 1):
L381 – 383: During these breaks, the experimenter changed the setup as necessary and the dog was free to explore the room and drink water.
Description of break between the experience phase and test phase (study 2):
L615 – 617: In between the experience phase and the test phase, subjects were given an approximately two-minute break while the setup was changed. Subjects were free to roam around the room and drink water in this time.
L 324-325 I do not understand the objective of modifying the first trial procedure.
This has now been clarified in the manuscript:
L320 – 322: Each human began with the opposite button to which they were supposed to press, to facilitate the subject’s understanding that the human intentionally provided or intentionally did not provide food.
L416 “The video for one of the free interaction sessions was unavailable” Do you mean the video from one dog or the videos of all the dogs during one session? In this case, which was the session?
This has now been clarified in the manuscript:
L413 – 414: A single video recording of one free interaction session for one subject was unavailable due to a technical malfunction.
426 and 551 vs 427. Did you use the number of times or the proportion of times?
We used the proportion and plotted the number of times. This has been amended in the manuscript and supporting information.
Reviewer #2:
General comments
In the present study authors tested if dogs would reciprocate a pro-social act from an unknown human in a context involving food, using a novel experimental design adapted from previous studies in dogs and other species. In the first study the authors did not find any evidence of reciprocity in dogs and speculate that this result could have been due to the complexity of the task or the mnemonic demands imposed to the dogs by the time gap between the training and testing phases. Thus, in a second study the authors simplify the task and run the training and testing phases in the same day. Again, dogs failed to reciprocate the human pro social behavior in this simplified version of the task. These findings led the authors to conclude that dogs do not spontaneously reciprocate the human pro-social behavior in contexts involving food. I want to congratulate the authors for their work; I found the paper very interesting and carefully design. I will suggest some minor revisions that could probably improve the paper.
Specific comments
Introduction
The review of the previous literature is precise updated and well written. Nevertheless:
L. 43-44. I would remove the phrase. “Reliable models of reciprocity that can be manipulated experimentally are scarce”. There is a vast literature on reciprocity on rats, mice, and insects. And I believe that there are all valuable models for the study of reciprocity (e.g. Taborsky’s works).
This has now been removed:
L40 – 42: Many aspects of reciprocity remain poorly understood, however, particularly the proximate mechanisms.
L. 89 -102. I found this paragraph a bit unnecessary. Although, as mention, all the listed characteristics made the dog a good model for the study of pro-sociality in general these are not specific characteristics needed for reciprocity. I would focus in reciprocity (See. Trivers, 1971).
This paragraph has now been removed and more focus has been placed on characteristics needed for reciprocity:
L88 – 93: Dogs possess additional characteristics which might suggest the propensity to reciprocate help received from humans. Apart from a long history of dog-human cooperation and communication (36–44) for which dogs appear to have evolved complex social cognitive traits (45,46; but see 47), and the development of strong bonds with humans (48–52) which could facilitate reciprocity (53–55), dogs seem to distinguish between cooperative and uncooperative humans.
L118 – 122: Given dogs’ long history of cooperating with humans, their specialized skills for such interaction with humans, their capacity to develop bonds with humans, their ability to discriminate between a cooperative and an uncooperative human, and their propensity to reciprocate help received from conspecifics, we investigated whether pet dogs reciprocate help received from humans.
L. 160-161. In line with the previous comment. Some of the aspects that would enhance direct reciprocity are: the possibility of future encounters (Trivers, 1971), and the bond between the subjects (i.e. De Wal, 1997; Freidin et al., 2013). So I think that one big difference with some previous studies (Quervel-Chaumette et al., 2015; 2016, Sanford et al., 2018; Carballo et al., 2020) is the identity of the receiver. Most evidences showing spontaneous pro social behavior in dogs (or the lack of it, Quervel-Chaumette 2016) came from studies with the owners, or some familiar human, as partner. I understand the methodological choice of the authors to use unknown humans as partners which eliminate a lot of uncontrolled variations that could mixed the results. But I believe that the best solution would be to run a third study whit the owners as partners and counterbalance their role (Helper vs. Unhelper) between subjects. The inclusion or non-inclusion of this third study would not influence in my personal decision to suggest the publication of the paper. I simply mention that this would be an interesting aspect to take into account and probably the authors would like to include a further study or mention this aspect in the conclusion.
Thank you for the suggestion. We have now included a paragraph on the potential use of familiar partners in the general discussion and we will keep this in mind for future studies:
L939 – 946: In future studies, the familiarity of the partners may also need to be considered. Our use of unfamiliar partners prevented results from being confounded by previous experience with those specific individuals. However, familiarity appears to influence prosociality in dogs, at least in experimental tasks with conspecifics (21,22). Thus, the use of unfamiliar partners may have precluded any prosocial food-giving by the dogs. In addition, much reciprocity in non-human animals is thought to occur within long-term affiliative relationships (53). It may, therefore, be advantageous to study reciprocity between dogs and humans by using humans with whom the dogs are already familiar.
We have also made reference to the unfamiliarity of the partner in relation to discrimination between two partners:
L908 – 909: Moreover, discrimination between unfamiliar humans in our studies may have been particularly demanding.
Methods and analysis
I found the methodology very well design and implemented. I think the data presented by the authors in the data set of supplementary material and the variables analyzed are clear and with a straightforward interpretation. So I found the statistical analyses unnecessarily a bit too complex for the data and the number of factors taken in consideration. But authors may well prefer to keep the presented analyses.
We have kept the analysis as is but will take this comment on board for future publications.
L. 261. Who was the human who eat the food?
This has now been clarified in the manuscript:
L252 – 254: The human eating the food in this stage was either the experimenter or a human unfamiliar to the dog; they were not one of the humans who would later act as the helpful or unhelpful human.
Discussion
Again, I found the discussion well written and pertinent. And I mostly agree with the authors’ interpretation of the data. Nevertheless, I would include some references to the role of learning mechanisms in the development of pro social behavior in dogs and in reciprocity in particular. The authors mentions this when they speculate that with further training and switching roles between dogs and humans dogs probably could reciprocate the human but they do not provide and explicit theoretical link. For example, dogs could be focusing in some other aspect of the experimental situation which could overshadow the role of the helper, thus making the actions of the helper more salient could improve dogs performance.
We have now included discussion points on the potential greater salience of the button and dispenser and the potential benefit of increasing the salience of the human partners and their actions:
L880 – 898 (General discussion): Finally, it is possible that, as a result of the training, the button and the dispenser were more salient than the human and occupied most of the dogs’ attention throughout both studies. Consequently, the dogs may not have paid enough attention to the human to register fully the relevance of the human’s actions, even though most subjects did at least look at or see the human to some degree. Increasing the salience of the human and the human’s actions could overcome this issue and facilitate the subjects’ registering of the cooperative act.
In this context, it is conceivable that the limited training and exposure to the action of the humans also diminished the dogs’ likelihood of perceiving or understanding the human’s action. Gfrerer and Taborsky’s (30,35) training procedure, which included providing a partner with food and receiving food from a partner, required approximately 14 to 19 days with two sessions per day for each subject. Moreover, they incorporated the exchange of roles with a partner. The training protocol in experimental studies in which rats were observed to reciprocate was similar, with repeated experience of exchanging roles prior to the experimental procedure (103,104). The drawn out experience of the receipt of food from a partner, and providing food to a partner, over a number of days, combined with the rapid exchange of roles, may have facilitated an understanding of the significance of the partner’s cooperative or non-cooperative behaviour. In contrast, in our design, the subjects had a single experience day with each human, albeit consisting of numerous trials. Extensive training may be required to facilitate subjects’ attentiveness to the crucial aspects of the setup.
L744 – 761 (study 2 - discussion): It is, nonetheless, conceivable that subjects did not pay enough attention to the humans in the experience phase of either study to permit recognition of the action of the human. We investigated this possibility further by determining whether gazing at the human occurred at all. We focused on the experience phase of study 1 due to better video quality and, more specifically, the experience phase with the helpful human, as it is arguably the more important human to be attentive to in order to reciprocate.
Visual inspection of plotted data representing the proportion of time spent gazing at the helpful human immediately before button pressing (see S9 Fig) and immediately after button pressing (see S10 Fig) reveals that most subjects generally did look at or see the human in the experience phase for some duration, particularly before button pressing occurred. Much of this appeared to be due to the subjects already focusing on the button before the human pressed it, or due the subject tracking the human’s hand when he/she began to move it towards the button. Gazing at the human’s face did not seem common.
We also fitted a model to investigate whether the proportion of time spent gazing at the human changed across trials (see S1 Appendix for details). With increasing trial number, the proportion of time subjects spent looking at the helpful human before and after button pressing increased significantly. This finding may indicate that subjects were learning the importance of the human’s action.
Furthermore, the discriminations between unknown humans could be demanding for dogs, thus increasing the difference between helper and un-helper could also improve dogs’ performance or the testing could be done with familiar humans. These aspects are related to the factors that could facilitate reciprocity that I mention earlier (familiarity, bonding, repetition of the encounters etc.) and are usually taken into account when discussing pro-social behavior in dogs.
We have now included discussion points on the demands of discriminating between the two partners and the potential benefit of increasing the difference between them or testing with familiar humans:
L899 – 910: In the context of learning about crucial aspects of the setup, it is worth noting that to reciprocate directly in our setup, dogs would have needed to discriminate between the helpful and unhelpful human. The discrimination learning required may have been too demanding for such a small number of sessions. Visual and olfactory discrimination learning in dogs can take a considerable number of training sessions (97,105,106). In fact, in a discrimination learning task in which dogs had to discriminate between their owner and a familiar person based on visual information from the faces and heads, an average of 6.5 sessions, each lasting ten trials, were required for dogs to successfully discriminate (107). Similarly, dogs were found to require between three and eight sessions of 20 trials each to successfully discriminate between their owner’s and a stranger’s face (108). Moreover, discrimination between unfamiliar humans in our studies may have been particularly demanding. Increasing the perceptual differences between the humans may help with learning the discrimination in future studies.
L939 – 946: In future studies, the familiarity of the partners may also need to be considered. Our use of unfamiliar partners prevented results from being confounded by previous experience with those specific individuals. However, familiarity appears to influence prosociality in dogs, at least in experimental tasks with conspecifics (21,22). Thus, the use of unfamiliar partners may have precluded any prosocial food-giving by the dogs. In addition, much reciprocity in non-human animals is thought to occur within long-term affiliative relationships (53). It may, therefore, be advantageous to study reciprocity between dogs and humans by using humans with whom the dogs are already familiar.
L909 – 910: Increasing the perceptual differences between the humans may help with learning the discrimination in future studies.
We have also pointed out that some studies observed successful discrimination between a cooperative and an uncooperative partner after quite a small number of trials:
L911 – 922: Despite the apparent difficulty of discrimination learning, however, it is worth keeping in mind that dogs were previously shown by Gfrerer and Taborsky (35) to reciprocate after just 14 trials of experience with each partner type. It is plausible that the more extensive training procedure, outlined above, facilitated discrimination by priming subjects to attend to the partner. Furthermore, partner discrimination may be easier if the partners are conspecifics. However, as subjects received an experience phase and test phase with a single partner in a two-day sequence before encountering a new partner (35), discrimination learning may not have been necessary to facilitate reciprocity. Nonetheless, Heberlein et al. (56) and Carballo et al. (57) found that dogs could discriminate between humans that differed in cooperativeness after just six and twelve trials respectively. As subjects in our studies were presented with each human across at least 20 trials, it is not clear that the challenge of discrimination learning can account for the lack of reciprocity.
L. 885: include “direct” before reciprocity.
This has now been included (see L952).
Submitted filename:
Dogs fail to reciprocate the receipt of food from a human in a food-giving task
PONE-D-20-35864R1
Dear Dr. McGetrick,
We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.
Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.
An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at
If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact
Kind regards,
Simon Clegg, PhD
Academic Editor
PLOS ONE
Additional Editor Comments:
Many thanks for resubmitting your manuscript to PLOS One
The reviewers are happy that you have addressed all the comments and the manuscript reads well, therefore I have recommended it for publication
You should hear from the Editorial Office shortly.
It was a pleasure working with you and I wish you the best of luck for your future research
Hope you are keeping safe and well in these difficult times
Thanks
Simon
Reviewers' comments:
Reviewer's Responses to Questions
1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.
Reviewer #1: All comments have been addressed
Reviewer #2: All comments have been addressed
**********
2. Is the manuscript technically sound, and do the data support the conclusions?
The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.
Reviewer #1: (No Response)
Reviewer #2: Yes
**********
3. Has the statistical analysis been performed appropriately and rigorously?
Reviewer #1: (No Response)
Reviewer #2: Yes
**********
4. Have the authors made all data underlying the findings in their manuscript fully available?
The
Reviewer #1: (No Response)
Reviewer #2: Yes
**********
5. Is the manuscript presented in an intelligible fashion and written in standard English?
PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.
Reviewer #1: (No Response)
Reviewer #2: Yes
**********
6. Review Comments to the Author
Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)
Reviewer #1: (No Response)
Reviewer #2: I found all my comments responded and congratulate you for your work.
One final comment. In the figures 6 and 9 you represent the number of times the subject press the button while in the analysis and results sections you refered to the proportion fo times dogs press the button. For consistency I would change the figure's y axis
**********
7. PLOS authors have the option to publish the peer review history of their article (
If you choose “no”, your identity will remain anonymous but your review may still be made public.
Reviewer #1: No
PONE-D-20-35864R1
Dogs fail to reciprocate the receipt of food from a human in a food-giving task
Dear Dr. McGetrick:
I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.
If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact
If we can help with anything else, please email us at
Thank you for submitting your work to PLOS ONE and supporting open access.
Kind regards,
PLOS ONE Editorial Office Staff
on behalf of
Dr. Simon Clegg
Academic Editor
PLOS ONE