All research presented here was approved by the Institutional Review Board and Institutional Animal Care and Use Committee at the University of Michigan. All children participated with the consent of their parents and their own verbal assent. The three dog ‘experimenters’ participated with the consent of their owners and their own eagerness to return to the study room.
Three dogs, Fiona, Henry and Seymour (Figure 1 ), participated as ‘experimenters’ in our study between 2015 and 2020. We initially chose to work with Fiona, who belonged to a close friend of one of the authors, because we observed Fiona to eagerly and gently interact with young children (iteration 1). To expand beyond a single dog, Henry, who belongs to one of the authors (who was never in the room with him during testing), joined the study (iterations 1–2). After Fiona moved away, we recruited Seymour, who belongs to a friend of one of the authors, as an “experimenter” (iteration 3). Fiona, Henry and Seymour spontaneously behaved differently in similar situations because of differences in personality, baseline activity and other factors. We did not attempt to control this variability but used it to assess children’s dog-directed behaviors in a generalizable way.
Children came from Ann Arbor, Michigan, US and surrounding towns where most families were middle class. By parental report, the majority of children (71%) identified as White, and 29% identified as multiracial, Black or African-American, Latino or Asian. A total of 36% of these children had a parent occupied in caring for them at home full-time, and all spoke English as their only language or fluently as one of their multiple languages.
We recruited 97 children of 1.7 to 3.1-year-old (51 female, 46 male; age range: 20–47 months) to participate in three iterations of our study at a university child laboratory, 44 of whom had pet dogs at home and 53 who did not. We chose this sample size to reveal comparable effects if children helped dogs at similar levels to their helping other humans. It was unclear what to expect with children helping dogs, but research with children helping adults in previous studies ( Warneken and Tomasello 2006 ; 2007) has consistently found significant helping with sample sizes ranging from 24 to 36 children. We chose our age group after initially piloting with 14-month-old to 3-year-old children (= 35). Younger children seemed overstimulated in the presence of the dog and had difficulty attending to the situation as was necessary for the study design used here. We also excluded children in our age range who were indicated by their parents (in a recruitment phone call) to be allergic to or highly fearful of dogs.
PROCEDURES
Children and their parents were taken to a testing room where one of three small dogs, either Fiona, Henry, or Seymour, sat in an enclosure created by a commercial baby-gate fence (Figures 1 and 2 , Supplemental Video).
Figure 2 . Child-dog interaction setup.
The enclosure had numerous holes and spaces to easily see the dog and to allow small objects (i.e., edible dog treats or toys) to be offered to the dog through the fence. We implemented this enclosure set up as a safety measure. Although all three dogs had no history of aggression and were gentle with children, the children were quite young and we were concerned that they or their parents might become startled or distressed if a dog initiated physical contact with them such as excited licking. Importantly, our initial IRB approval (iteration 1, n = 37 children) did not permit children to pet dogs. We were able to obtain IRB permission for petting for later participants (iterations 1 and 2, n = 60).
We used treats and toys as out-of-reach objects because both were of interest to dogs but might present different scenarios to children (e.g., feed the dog vs. play with the dog). Using both permitted us to distinguish whether children wanted to help the dog across different contexts rather than solely feeding or playing with the dog. There was also a ‘staging platform’ adjacent to the fence where we placed out-of-reach objects (the white box-like item in Figure 2 ). When children and parents entered the testing room, we briefly introduced them to the dog together. Then, we asked parents to sit in a chair in the back of the room and pretend to be busy reading a magazine while we engaged their child in further dog familiarization and then target out-of-reach object events.
Then the human experimenter again introduced the child to the dog, sat on the floor and read a book about a dog named Spot (‘kind of like Fiona/Henry/Seymour’) to the child in proximity to the enclosure so that the child was comfortable in the testing space and understood that they were permitted to interact with the dog. Importantly, this book included no mention or depiction of Spot’s goal-directed behavior. Based on our pilot testing, we had concerns that children might think that they were not permitted to interact with the dog in the testing room. During pilot sessions, when we presented children with our task and later interviewed them, several children verbally and emotionally expressed a desire to feed the dog. For instance, one child screamed to a ‘busy’ parent, ‘She’s hungry! She’s hungry! That dog is hungry!’ but did not provide the dog with treats themselves. When asked about their actions, children indicated that they thought they were not allowed to feed the dog because she was ‘in a cage.’ Consequently, we took several measures to ensure children understood that they were allowed to feed the dog but did not think that the experimenter desired for them to do so. In iteration 1, prior to receiving permission to allow petting and touching through the enclosure fence, we gave children a dog treat to hold but did not explicitly encourage or train them to feed the dog. In iterations 2 and 3 we received IRB permission for petting through the fence and permitted children to pet and feed the dog treats if they wished to. In the latter iterations, we then used toys not treats in out-of-reach object events rather than a combination of the two used in the first iteration (see below).
After the children were familiarized with the dog in the manner above, we initiated an out-of-reach ‘event,’ during which a treat or toy was placed on the staging platform within reach of the child and out-of-reach for the dog. Importantly, we did not want it to appear as though the human experimenter had purposely placed the treat or toy on the platform with the intention of giving it to or keeping it away from the dog. Thus, we made the placing appear accidental in the following ways: In iteration 1 (n = 37 children), we presented four events to each child in the same order. In the first event, the experimenter told the child, ‘I have to leave for one minute, but I’ll be right back, OK?’ then left the area while surreptitiously letting a single treat drop from her pocked onto the staging platform outside of the dog’s enclosure, within reach of the child. To initiate her departure in events 2–4, the experimenter used a remote to trigger a ringing sound on her laptop computer, which was kept on a cart at standing desk height at the back of the room. At the sound of the buzz, she would exclaim, ‘Oh! I’ll be right back!’ surreptitiously drop the treat or a toy on the platform, go to the computer, pretend to be engaged and type, and allow the dog to naturally beg for and attempt to access the out-of-reach food or toy for 30 seconds. Begging/attempting behavior included the dog scratching at the fence with her paws, making whining noises and/or looking at the item and looking at the child while cocking her head from side to side (see Supplemental Video). In events 1 and 4, the out-of-reach item was a treat, in event 2 it was two treats contained within a toy and in event 3 a toy alone (see Supplemental Video, Supplemental Protocol). In this study iteration, both dogs Fiona and Henry attempted to access the object in every test event (n = 145 events, 37 children).
After testing 37 children with the above method, we wanted to document child behavior in ‘control’ events where dogs would naturally ignore out-of-reach items compared to test events where the dogs attempted to access items as they had during iteration 1. We achieved this natural distinction by rubbing a rubber dog-bone-shaped toy with a dog treat or food in a manner that left no residue detectable to children but retained a strong smell that was attractive to dogs. Instead of creating four out-of-reach object events for each child as we had in iteration 1, we created two events, one where a scented bone was placed out of the dog’s reach and one where an unscented bone was. We counterbalanced the order of these ‘control’ and ‘test’ events between participants. Because Fiona moved from the area, Henry remained our only dog participant for iteration 2 (n = 37 child participants). Unexpectedly, particularly during events with the unscented toy, Henry not only ignored the out-of-reach item but became relatively disengaged, appearing uninterested in the child. In total, there were 70 events in which Henry attempted to access the out-of-reach toy but in only 37% of these, he was energetic and engaged (see coding details in ‘Analyses’ below). There were 77 events in which Henry ignored the out-of-reach toy and he was energetic and engaged in 8% of these events. This behavior was quite different from Fiona’s behavior and Henry’s own behavior during iteration 1 in which both dogs were engaged and attempted to access objects in 100% of events.
Our concern that Henry’s level of disengagement, more so than his actions towards the out-of-reach item, might influence children’s behavior, led us to work with a new dog, Seymour, who was consistently engaged. Trials with Seymour constituted iteration 3. These concerns about dog engagement level also contributed to our decision to aggregate data from out-of-reach events across all study iterations in a single analysis that permitted us to address dog engagement, among other potentially confounding factors (see ‘Analysis’ below).
As in iteration 2, in iteration 3 (n = 23 child participants) we presented each child with two out-of-reach toy target events, in this case, one where Seymour showed interest in the object and one where he mostly ignored it. Both objects were identical-looking rubber dog bone toys, differing only in colour. The different colours were intended to demonstrate to the child (not the dog, who, as in iteration 2 was responding to scent) that they were different items. Event order (dog showed interest vs. did not) and objects (which was scented and other was not) were counterbalanced between child participants. We also slightly altered the ‘accidental’ placing of objects that put them out-of-reach during these iteration 3 events. Instead of having an experimenter ‘accidentally’ drop an object, we had her ‘accidentally’ kick the staging platform with the target object on it away from the edge of the fence as she walked to attend to the computer buzzer. This change provided a more streamlined method and permitted us to achieve identical behavior between several experimenters who administered the tasks.
We were forced to end data collection for iteration 3 with the onset of the COVID-19 pandemic. At this time, we had collected data from only 23 children, the majority of whom (75%) did not have pet dogs at home. This imbalance, making it difficult to examine the effects of prior dog exposure when examining findings from iteration 3 alone, further contributed to our decision to aggregate events across all iterations for our primary analysis (see ‘Analysis’ below).