ALEXANDRA HOROWITZ

The mind of a dog
 
   
 
I study dog behavior and cognition. To date, my research has taken four forms:
naturalistic observations of social play;
observations of dog-human play;
tests of anthropomorphisms used with dogs; and, most recently,
field recordings of vocalizations during play, and the design of experiments to test the communication and attention abilities of dogs.
 

play signalling
Dog social play research
The behaviors of theories of mind, and a case study of dogs at play

In my dissertation, I considered the theoretical construct "theory of mind" and tested how behaviors of animals observed in a natural setting might indicate incipient mindful understanding.

In particular, in this research I observed domestic dogs at play, making note of the order in which dogs do certain actions. Some behaviors serve to communicate something; others are expressions of excitement, fear, or anger; others are non-reducibly playful. The amazing success that unfamiliar dogs have in negotiating an all-out play-fight indicates that they are communicating their intent to each other very well. So well, in fact, that my research led me to suggest that they need to have some implicit understanding of the other animal's perspective on things -- his mind -- in order for play to work.

An analysis of the sequences of actions in play showed that dogs seem to differentially signal play intentions to their partners based on the perceived attention of that partner. In most cases, play signals were only sent to attentive, present audiences. When a play signal is not seen, the signaler moves to get attention and then re-indicates his intentions in the view of the partner. The dogs appear to appreciate the importance of attention in communication. Furthermore, dogs use attention-getters appropriate to the level of inattention of their playmates -- using the most forceful attention-getters only with distracted playmates.

These behaviors are interesting, as capable use of attention -- and of attention in communication -- is often implicated, in humans, in developing a theory of mind. For instance, as a well-functioning adult human with a fully developed theory of mind, I know that you won't be able to hear or understand what I'm saying if you're across the room, talking on the phone. I know that I need to get your attention, and then use a language and words you can comprehend.

The data of this research show that to a limited extent, dogs show an ability that humans have: perspective-taking. The playing dog appreciates the importance of others' attention in communication. I suggest that this is a kind of "rudimentary" theory of mind.

The dissertation as pdf: text (2.5 mb) (please cite as per the title and frontmatter).

Related: a paper about talking about dogs and another about dog cognition.


 
dog-human dyad

Dog-human play
Behavioral design of life-like human-robot interaction

In association with Javier Movellan and the Machine Perception Lab at the University of California, San Diego, I conducted ethological research of the interspecies play of people and dogs. This work was funded by Sony, whose "Aibo" robot dog is designed to interact smoothly and realistically with its human owners.

This ethological research took the form of an observational study of a large sample of human-dog dyads in specified social situations. The primary means of information-gathering was videotape of naturally-occurring interactions. Analysis emerged from in-depth examination of the videotape, characterization of the activities of the participants, and consideration of the tallied data from the group.

Observational data-gathering consisted of videotaping of play routines between dogs and owners. People who were playing with their dogs were approached, and, upon consent to videotape, were recorded in play with their dogs. Permissions were secured for each taping. These tapes were reviewed and converted to computer video form. This data was and is still being assessed for patterns.

Analysis of the video data allows me to deconstruct each play bout into constituent behaviors. From this, a characterization of the specific components of a dog-human play interaction is possible. The data will allow development of a listing of dog-typical behaviors which indicate emotional state, receptivity, or which accompany communication or provoke human response. In addition, I will be able to specify the kinds of interaction routines that are canonical between humans and their dogs.

Click here for the original research proposal.


 


Testing anthropomorphisms

Anthropomorphisms are claims which are generally unsupported by scientific research. Commonly, animal behavior is compared to human behavior, and where there is superficial matching, the attribution (of understanding, emotion, or knowledge) that is made to the human is extended to the animal.

While vilified by science, anthropomorphisms can be starting points to considering animal behavior. By attending to the causes of the seen behavior of the animal, one can in effect empirically test the anthropomorphic claim. In the latest study, the anthropomorphism investigated was the claim that the so-called "guilty look" shows that dogs feel guilt or understand that they have disobeyed.

The tested dogs did not show more behaviors associated with the guilty look when they performed a forbidden act--eating an available treat--than when they did not. Instead, more "guilty look" behaviors were seen in trials when the owner scolded the dog, whether the dog had disobeyed or not. Furthermore, the effect of scolding was more pronounced when the dog was obedient, and did not eat the treat. This is consonant with the hypothesis that it is not disobedience but scolding behavior, or perhaps fearful anticipation of scolding, that causes an increase in the magnitude of the guilty look. Finally, dogs who had been in obedience training also showed more pronounced guilty looks than those who had not.

 


Measures of communication and attention

1."Vocalizations between dogs (Canis familiaris) in social interaction: Barks, growls, and whines as communicative utterances."
Our Dog Cognition lab is examining the vocal behavior of domestic dogs in the context of social play with other dogs in natural settings. Though there are numerous uttered sounds generated during play, no research has investigated what the communicative function of the vocalizations may be. Through video and audio recordings of ordinary social interactions in natural settings, we are attempting to isolate vocal signals and identify the specific context in which the signal is used. Our aim is to characterize and fully describe a repertoire of vocal signals used by dogs in interaction not with humans, but with other dogs. We distinguish the signals by acoustic structure(s), and identify contexts associated with specific vocalizations.

2. I am also designing experiments to further test dogs' communication methods and use of others' attention. Previous research has shown dogs to be very capable at following human-centered communicative cues: pointing, head nods, eye gaze; their use of vocalizations and their responses to vocalizations of others is well known. I will be looking at communicative methods in modalities other than vision, and at use of species-specific gestures. In addition, I will test dogs' use of attention in interaction with humans.

Return to my home