top of page

What do dogs understand about humans?

Updated: Feb 3

Delve into the intriguing world of canine understanding in our upcoming Zoom event on Saturday, February 3rd. Inspired by Chapter 10 "What Dogs Understand About Humans" by Juliana Braüer of the book "The Social Dog Behaviour and Cognition" by Juliane Kaminski, and Sarah Marshall-Pescini. We will explore the profound questions posed by Juliana Braüer in 2014. Join the discussion as we navigate topics such as what dogs comprehend about their human companions, their awareness of human perceptions, and their ability to decipher human goals and intentions. As we bridge the gap between this enlightening chapter and the present day information, we invite you to share your thoughts, questions, and discoveries.

Summary of Chapter

In this chapter the author sets out to answer the following questions;

  • "What do [dogs] understand about the human companions?"

  • "Do dogs know what humans can perceive?"

  • "And if so, how do they assess what humans can perceive?"

  • "Do they understand humans' goals, and intentions?"

  • "Are they able to attribute mental states to others such as beliefs and knowledge?"

The chapter is divided into sections, each detailing a precondition to answering these questions.

10.1 Monitoring Monitoring is defined as being attentive to another, often by looking at them. This is considered a precondition for understanding another as a means of gathering information about and from them. Dogs show an exceptional tendency to monitor humans, particularly when compared to wolf pups who were raised under identical conditions.

The author notes that dogs not only constantly monitor humans, but have the ability to understand human communicative cues, such as a pointing gesture.

It was noted that dogs use social-referencing similarly to human children, in which their understanding and behavior is based on the information they seek from a human. When humans displayed apprehension, dogs would hesitate when moving toward a novel object, vs. moving closer to an object and interacting with it sooner when the human displayed a positive attitude. (Merola et al., 2011; Merola et al., 2012) A look at numerous studies concluded that dogs are adept at following the gaze of a human, in order to find hidden food, labeled an object-choice task. However, these results were less successful in setups in which the gaze directed the dog to an open or ambiguous area. Two potential explanations are proposed for this discrepancy, the first may be due to differences in how the dog's attention was obtained prior to shifting the gaze, whether they were called to, or spoken to in a certain tone of voice. The second could be a result of different measures used, one looked at duration of gaze, while another measured which object the dog first attended to.

10.2 Perspective Taking This chapter sets out to first address whether dogs understand what humans can see.

In a study in which dogs were given the choice to beg for food between two humans that were eating, dependent on whether they could see the human's eyes or the direction of their face. Dogs showed a preference for the attentive human over the inattentive human. (Gacsi et al., 2004; Virányi et al., 2004; Udell et al., 2011; see also Udell & Wynne, 2011; and Virányi & Range, 2011).

Another set of studies (Virányi et al., 2004; Call et al., 2003; and Huber, 2006) Demonstrated that dogs respond better when the owner looks at the dog while delivering a cue, vs. looking at another person, turning their back, etc.

This confirmed that dogs are sensitive to human attention, but researchers wanted to dive deeper into examining whether dogs could understand what humans can see.

In one experiment, researches set forbidden food behind a small barrier, a big barrier, and a big barrier with a plexiglass window. The dogs ate more forbidden food when the barrier was large. However, this did not rule out that the dog was responding to discrimination between the type of barrier, rather than truly understanding what humans can see. (Braüer et al., 2004) To address this concern, Kaminski et al. (2009) sought to determine whether dogs could take the visual perspective of humans when that perspective differed from their own. Two toys were placed behind two small barriers so the dog could see both, one barrier was transparent, the other opaque. There were three experimental conditions, in one version the human sat opposite the barriers and could see only one toy. In the second, the human sat opposite to the barriers with their back turned so they could see no toys, and in the third the human sat behind the barriers with their back turned, so the barriers were no longer between the human and the toy, but they were nonetheless unable to see the toy. The study found that if the human could see only one of the toys, the dogs brought back the toy the humans could see, despite the fact they they themselves were able to see both toys. Interestingly, dogs seem to display a tendency to act in accordance with this perspective taking only when they can also see the humans, indicating that this flexible perspective taking may be limited by the belief that "If I don't see the human, they don't see me". In similar experimental setups (Kundey et al., 2010) looked at whether dogs can understand what humans can hear, with similar results, dogs preferred silent pathways to obtaining forbidden food when the human was present, whether or not they were attending to them.

10.3 Seeing Leads to Knowing

In this section, the author sets out to examine whether dogs understand not only what humans can see, but what they know as a result of what they have seen.

To test this, Virányi et al., (2006) looked to the "Ignorant Helper" paradigm invented by Gomez (1996; see also Witten, 2000). Two objects were hidden in a room, the dog's favorite toy, and a stick needed to retrieve the toy which was inaccessible. The human was either absent or present while the toy was hidden, dogs observed the whole hiding process, but could not reach the toy on their own. Previous studies have shown that dogs will indicate the location of a hidden object without prior training (Miklósi et al., 2000; Kaminski et al., 2011a). The dogs did not show the location of the stick, which seems to point to the fact that they did not understand that the human needed the stick to retrieve the toy. They did however show the location of the toy more often when the human was not present during the hiding process. Cooper et al. (2003) used the Guesser-Knower paradigm invented by Povinelli et al. (1990) in which the dogs are presented with 3 locations, one of which contains food. There are two human informants, one that was present as the food was placed, the "knower", and one that was absent, the "guesser" who both point to different locations. In the first trial 93% of the dogs chose the location pointed to by the knower. However this effect was only noted in the first trial, and did not persist through the overall performance. It is noted that this experiment was never properly documented, and therefore it is impossible to know if other variables can account for this result. Kaminski et al. (2009a) did not find evidence that dogs understand what a human has seen in the past. They placed two toys on the dog's side of two small barriers, while the dog observed the toys being placed. The human was out of the room during the placement of one toy, and sat opposite the barriers observing the placement of the other toy. After that, the experimenter asked the dog to fetch the toy, with the intent of seeing whether the dogs would differentiate between the toys based on the one that the human had the most information about, however that was not the case. 10.4 Intentions

In this section the author examines whether dogs can understand the goals and intentions of humans.

Range et al. (2007) found that dogs copy the actions of others more often when those actions provide an effective and efficient solution to a problem. For example, they used a problem that involved opening a rod. At baseline they found that dogs preferred to use their mouth, but would copy a demonstrator who used their paw because their mouth was obstructed by a ball. The test dogs did not have a ball, so their mouth was usable, however they still opted to use the paw. Kaminski et al. (2011b) found that dogs do not imitate rationally when they replicated the study with another control condition. They suggest that dogs were simply distracted by the ball, and did not attend to the rationality of the action.

In a second experiment, Kaminski et al. compared an unusual cue toward the location of hidden food given in a rational and irrational context. In one presentation the human would be holding an object, making their hands full, and would use their leg to signal toward the location, as a rational alternative to communication. In the second, their hands were empty and they would use their leg to communicate in an irrational setup. The dogs did not differentiate between the two, indicating that they did not take into account the situational constraints.

Kaminski et al. (2012) used an object choice task experiment to compare the effect of intentional communicative movements to non-intentional communicative movements. The human either pointed at the correct cup of hidden food, alternating gaze between the dog and the cup, or pointed while pretending to check the time on a watch, or clock on the wall. Dogs followed intentional pointing more often. Csibra, (2003) and Behne et al., 2005 demonstrate that dogs are very adept at determining when communication is intended for them, and utilize the same communicative cues that human infants use to identify their intent. A few preceding studies found arguably insufficient results to determine whether a dog could use their ability to ascertain intent, to behave cooperatively, assisting in a problem scenario. To address this, Braüer et al. (2013) aimed to address whether the dog's understanding of the human's goal was unclear in these previous setups. They set out to test whether dogs would help a human if the human's goal was made as obvious as possible. The human would attempt to enter a target room to get a key. The tested dog had demonstrated an ability to press a button to open the door. The help conditions in which the human attempted to enter the room, were compared to humans that did not attempt to enter the room. The experimenters found that the dogs helped the human when the human explicitly communicated their goal to the dog. 10.5 Theory of Mind In Domestic Dogs?

Theory of mind is defined by the author as the ability to attribute mental states such as beliefs, intents, desires, pretending, and knowledge that are different from their own to another. The author states that there is no evidence of this to date, and acknowledges that this is difficult to test and there is no clear-cut test or no answer to this question.

The final section,"mind reading or behavior reading" provides a recap of what has been learned within the chapter, and closes with this:

"In conclusion, although there is no evidence that domestic dogs posses a humanlike theory of mind, they 'understand' a lot about humans in their own way. They are very successful in solving social problems in their human environment, as they constantly monitor humans, learn valid associations, make adequate generalizations, and use egocentric strategies."

It is also important to note that this book was written in 2014, and continuous attempts to capture these complex subjects in experimental design are yielding new results with time.


Join us Saturday, February 3rd for a group zoom event to discuss this topic. As we bridge the gap between this enlightening chapter and the present day, we invite you to share your thoughts, questions, and discoveries. Become a member of R+evolution to join a welcoming community of behavior and training enthusiasts, where friendships are made, knowledge is shared, and learning is a collective adventure. We hope to see you there!

9 AM Pacific

10 AM Mountain

11 AM Central

12 PM Eastern

Summary of The Social Dog, Ch 10
Download PDF • 92KB


bottom of page