Monday, March 9, 2015

Contexts and cues: the reactive dog brain

A dog on leash, seeing another dog, explodes into a fury of barking and lunging. Reactive dogs, dogs who respond with arousal or aggression to what should be innocuous stimuli, can be very difficult for their owners to manage safely. I've written previously about hormonal changes in individuals experiencing this kind of arousal. But why do their brains trigger the stress response in such inappropriate situations in the first place?

Learning and memory

Past learning, stored as memories, has a lot to do with current behavior. If a dog has made bad associations with something in the past, he has a good chance of expecting a similarly unpleasant experience the next time he encounters it or something that reminds him of it. How he chooses to deal with this situation — aggression or withdrawal — is one interesting question, but right now I’m writing about how he makes associations in the first place and how he retrieves them later.

Learning and memory can mean a lot of different things depending on their context. I’ll be using them in a very narrow sense.

Learning: making an association between a stimulus and a consequence
Memory: the ability to retrieve a previously-formed association
So if a puppy is attacked by another dog, he may learn to associate other dogs with pain and fear. When he later encounters another dog, he uses his memory to retrieve that association. Two parts of the brain which are deeply associated with this type of learning and memory are the amygdala and the hippocampus.

The amygdala is associated with threat evaluation: is that twisty shape I see out of the corner of my eye a stick, or a poisonous snake? Is the dog I am greeting friendly, or about to attack me? People with damage to their amygdalas may have difficulty evaluating threats, to the extent that they may not be able to feel fear. As a result, the amygdala functions in emotional learning: people told scary stories remember them better than less exciting stories partly because of the emotional contributions of their amygdala, which tells them that an experience has some level of threat and should be recorded in memory with particular care.

The hippocampus, on the other hand, is famous for its contributions to learning different locations. London cab drivers must spend years memorizing the twisty street map of their city, and when they are done, their hippocampuses are actually larger in size compared to people who haven’t gone through the training.

When they work well, these two brain structures are an important part of the process of identifying appropriate threats and discarding stimuli that aren’t threatening, based on previous experience. So what exactly is going on when they operate as they should?

Fear conditioning: contexts and cues

The most effective studies that have been done to determine exactly how the hippocampus and amygdala function in learning and memory have used fear conditioning, often in laboratory rodents. Dog trainers use classical conditioning to associate stimuli that a dog considers threatening with something positive, to change the dog’s emotional response to that stimulus — for example, to teach a dog who fears other dogs that they will reliably get food when other dogs approach, so that the dog comes to look forward to the approach of another dog as a chance to get a treat. Fear conditioning researchers do the opposite, teaching a laboratory rat that something previously benign (like the sound of a bell) predicts something aversive (like an electric shock).

It’s unfortunate that so much research has been done on how to teach fear, something we don’t actually want to do in real life. However, what we learned from these studies should translate to the types of classical conditioning we do with dogs, and be even more relevant to helping us understand how fear-based behavior issues come about in the first place.

These studies have shown that that contexts and cues are important in classical conditioning. If you put a rat into a blue cage and then repeatedly play a bell right before shocking him, he will learn to fear the sound of the bell. The blue cage is the context; the bell is the cue. If you move the rat into a purple cage and play the tone without a subsequent shock, the rat will learn that the purple cage represents a different context, and that he does not need to fear the cue in that context. So the cue and the context contribute differently to classical conditioning.

Source:  Nature Reviews Neuroscience 14, 417–428 (2013)

In the case of a reactive dog, we might imagine that this dog spent time in a rough playgroup as a puppy, and learned to associate other dogs with being bullied. Here, the cue is another dog, and the context is the room the playgroup was in.

The hippocampus: learning in context

One of the jobs of the hippocampus is to encode contexts. Those London cab drivers with oversized hippocampuses have countless contexts encoded to represent many different locations around London. The hippocampus of the puppy who had a tough time at playgroup encoded the room where playgroup happened as a context.

In the case of our laboratory rats, the hippocampus encodes the blue cage as one context and the purple cage as another. With a healthy hippocampus, the rat can differentiate between the two contexts, and is fearful of the cue only in the appropriate context. But with a damaged hippocampus, the rat can’t differentiate between the blue and the purple cage. Although he was trained that the bell only predicts a shock in the blue cage, he fears both cages, because his hippocampus is unable to properly represent the context of the blue cage.

The associative amygdala

One of the jobs of the amygdala, on the other hand, is to encode associations. It encodes the association between cue and stimulus (bell predicts shock) and between context and stimulus (the shock only happens in the context of the blue cage). When humans were tested with functional MRI to see which regions of their brain became more active during a fear conditioning trial, the amygdala and hippocampus responded in different situations. When humans were trained to associate a cue with a shock, their amygdala activated in response to the cue. When they were trained to associate only a context with a shock, both their amygdala and their hippocampus activated when they were exposed to that context. The amygdala activated in both cases because the association was being recalled in both cases, but the hippocampus was only activated when the particular context was recalled. Fascinatingly, this study also found that humans with larger hippocampus volume had greater fear responses in fear conditioning trials. There was no association between amygdala size and fear response.

Prefrontal cortex as mediator

We are not, thankfully, completely at the mercy of the whims of our hippocampus and amygdala, subject to uncontrollable fears based on past bad experiences. We have some ability to take a step back and calm ourselves down. One of the parts of the brain involved in this higher-order cognition is the prefrontal cortex (PFC). This region of the brain has direct connections to both the hippocampus and the amygdala and appears able to mediate some of the signals coming from those two regions. Functional MRI studies tell us that while fear acquisition involves the amygdala, fear extinction (learning to let go of a fear) involves the PFC as well. We also know that people who have thicker PFCs are better at extinguishing fear associations. This mediation by the PFC is what lets us take a deep breath and choose not to give in to our fears.

Do dogs have this ability to take a step back and try consciously to decrease their fears? Certainly they are not as good at this skill as humans are, but I wonder if they do have some ability to do this. In a recent post at Reactive Champion, a reactive dog owner describes a situation in which she believes her reactive dog did just that.

PTSD: failure to contextualize?

When this system goes wrong, how does it go wrong? One hypothesis suggests that post-traumatic stress disorder (PTSD) is a disease of failure to contextualize. Humans with PTSD report having flashbacks to previous trauma unexpectedly and uncontrollably, and in inappropriate contexts. If you were in a drugstore during a robbery, it would be appropriate for you to remember that traumatic event when you returned to that location, and even to feel trepidation about entering that store again. You’d probably think about the event a lot for the first days, weeks, perhaps months afterwards, in many other contexts, as well. But your brain should recover, and you should eventually come to not think of it constantly, and only be reminded of it in similar contexts, such as the same or similar locations.

People with PTSD, however, may have trouble limiting their recall of traumatic events to similar contexts, so that they may be retrieving these memories (often vividly) in any and all contexts, years after the trauma has passed. The problem may lie with their hippocampus, which may have difficulty limiting recall by context. And indeed, studies have shown that people with PTSD often have smaller sized hippocampuses compared to the healthy population.

The perspective of the reactive dog

On to the realm of pure speculation, then, because studies haven’t been done in hippocampus function in reactive dogs. But I think the story of the person involved in a trauma who can’t appropriately contextualize her memories is similar to the story of the dog who was involved in a trauma (dog attack, overwhelming experience in a crowded area as a puppy) and can’t contextualize the experience. A dog who is attacked by other dogs at a dog park may learn to fear the dog park, but if never attacked outside of the dog park, should he learn to fear all dogs, everywhere? I’d argue that that’s an inappropriate association for his brain to make, and that the mechanism of failure might have to do with a failure of the hippocampus to appropriately contextualize, just as in someone with PTSD.

I’m certainly not saying that all reactive dogs have PTSD, but I am speculating that the mechanisms might be similar. Does hippocampal function vary across a spectrum, with some individuals having high-functioning hippocampuses and others not so effective ones? Do dogs with hippocampuses on one end of that spectrum have difficulty limiting their negative associations, such that they are more likely to suffer from fearfulness and possibly fear aggression? I don’t know, and I don’t know if the research will ever be done, but it’s an intriguing story to consider.


  • Maren, Stephen, K. Luan Phan, and Israel Liberzon. "The contextual brain: implications for fear conditioning, extinction and psychopathology." Nature Reviews Neuroscience 14.6 (2013): 417-428. [PDF]
  • Feder, Adriana, Eric J. Nestler, and Dennis S. Charney. "Psychobiology and molecular genetics of resilience." Nature Reviews Neuroscience 10.6 (2009): 446-457. [HTML]


  1. "In the case of a reactive dog, we might imagine that this dog spent time in a rough playgroup as a puppy, and learned to associate other dogs with being bullied."

    This is why I don't like, and no longer recommend "puppy play school" :-( My only two dogs who attended on of these ended up reactive to other dogs :-(

  2. Yes, it's super important to have the puppy playgroup experience be a good one. I do think they're very valuable if well-run, but there are plenty of them out there which are not well-run.