Patients with hippocampal amnesia lack imagination

We have been talking about this paper in PNAS around the lab, so I thought I would share.

Hassabis et al, publishing in PNAS, have shown that patients with hippocampal damage lack the ability to imagine novel situations. This is a truly interesting finding, but it isn't why I want to talk about this paper. Actually, I want to talk about this paper because of how they explain this finding.

To tip my hand a little, Hassabis et al attribute this lack of novel mental imagery to a deficit in spatial context provided by the hippocampus. This would fit squarely in the body of thought that characterizes the function of the hippocampus as associating episodic information with a spatial framework. However, I will argue that this isn't the only way to view the data. But I am getting ahead of myself. Let's start with some background.

Background to the Hippocampus

The hippocampus is a structure in medial temporal lobe that is involved in learning and memory.

We know this because it has been observed in patients such as the canonical HM that if you lose both your hippocampi, you lose the ability to form new memories but have an intact long-term memory. Since this recognition, there has been a constant stream of research into determining what precisely the hippocampus does and how it does it.

Part of the debate about this research has focused on whether the hippocampus encodes new memories in a manner that is fundamentally spatial or whether the hippocampus is more generally associative.

On the side of a spatial function of the hippocampus, hippocampal neurons fire selectively in what are called place fields. Place fields are different portions of space during a task. For example, if you put an electrode into hippocampal neurons of a rat during a task that requires the use of spatial memory, each neuron will fire when the animal is sitting in a particular region in space. It was hypothesized that the set of neurons with different place fields sum together to form a map of the spatial world. The function of the hippocampus under this system is to associate particular contextual cues with points in the spatial map. (This view is exemplified in The Hippocampus as a Cognitive Map by O'Keefe and Nadel, 1978.)

There is an alternative view, however. This view holds that the hippocampus does indeed encode spatial information, but it also encodes a wide variety of other information. In this sense, it is a more general associative machine.

Evidence for this view can be found in the funny things that hippocampal neurons do when you modify the context under which it is acting. For example, you can train a rat to seek out a different goal box depending on whether it is hungry or thirsty -- whether it wants food or water. In this sense the internal state of the animal (hunger or thirst) is part of the context that rat uses to know what task to perform. What you find is the particular goal for an activity changes the location of the firing patterns for hippocampal neurons. The internal state modifies place fields. Similarly, hippocampal neurons have been found to fire specifically at the goal arm or the start arm of mazes. They can also show modified patterns of firing depending on the course the rats is taking, suggesting that prospective intention affects their firing rate.

All of these are funky things that happen to place fields that aren't really spatial in the strictest sense. Surely they happen in space, but they seem to represent other axes of information encoded by the hippocampus.

In light of this evidence, we can posit an alternative view of hippocampal function. The hippocampus does indeed associate place with context, but it can also associate context with other things. (This view is reviewed here. Full Disclosure: The review was written by my current boss Matthew Shapiro and one of our graduate students Pam Kennedy.)

I summarize the views above so that you can get a flavor for the debate in the hippocampus community. There are essentially two camps. One side views what the hippocampus does as spatial. The other side views it as partly spatial, but partly something else.

The Paper at Hand

In this context, along comes Hassabis et al who found that people with hippocampal damage (like HM above) have trouble imagining new experiences.

To give you a flavor of the difference between the vividness of the mental imagery in a normal patient and someone with hippocampal damage, here is an excerpt from their interviews with patients (click to enlarge). (The patients are P03 and P05; Control subjects are CON.)

When you read the patient cases, they can see fragments of context like the color blue, but on the whole it is just not happening. They lack coherence. They are not being associated into one experience.

This finding is interesting for a couple of reasons.

The first is that you can extend it to make a statement about how imagination and memory formation are similar. The hippocampus had been associated previously with the formation of new memories. Clearly imaginative thinking has some aspect of it that mechanistically resembles the formation of new memories because they are both lost on lesioning of the hippocampus.

This is not surprising if you think about it because there is a lot of evidence that we construct our reality from sensations available to us. Memories are made; they are not born fully formed. Because reality is constructed, we can be fooled in a variety of predictable ways -- take visual illusions, for instance. Imagination must require a similar construction. Therefore, if the hippocampus is required for encoding new memories, it makes sense that it would be required for imagining novel ones.

The second reason I find this interesting is that the authors cling to the absence of spatial context as the explanation of the problems these patients have visualizing new experiences.

You get the sense of where they are going with this in the abstract:

Amnesic patients have a well established deficit in remembering their past experiences. Surprisingly, however, the question as to whether such patients can imagine new experiences has not been formally addressed to our knowledge. We tested whether a group of amnesic patients with primary damage to the hippocampus bilaterally could construct new imagined experiences in response to short verbal cues that outlined a range of simple commonplace scenarios. Our results revealed that patients were markedly impaired relative to matched control subjects at imagining new experiences. Moreover, we identified a possible source for this deficit. The patients' imagined experiences lacked spatial coherence, consisting instead of fragmented images in the absence of a holistic representation of the environmental setting. The hippocampus, therefore, may make a critical contribution to the creation of new experiences by providing the spatial context into which the disparate elements of an experience can be bound. Given how closely imagined experiences match episodic memories, the absence of this function mediated by the hippocampus, may also fundamentally affect the ability to vividly re-experience the past.

The hippocampus is clearly doing something spatial, and these patients are clearly lacking a spatial context for imagination. The problem that I have with this is that they are lacking in everything else necessary imaging new experiences as well. They don't just suffer an absence of space; they suffer a complete lack of contextual binding to anything. Nothing is cohering together to form a memory. From this perspective, the spatial problem may be an issue but it is largely secondary.

What fascinates me about this paper is that I feel like it is trying to stuff a square peg through a round hole. Surely the hippocampus is doing something spatial, but that isn't all it is doing. Why do these guys insist on viewing it only according to that narrow view?

Anyway, this may be a personal ax to grind because I think that a shift in understanding of hippocampal function is far overdue, but nonetheless I think this a very interesting paper.

Hat-tip: Faculty of 1000.