This is a repost from my personal Blog talking about the potential for multiple types of pattern separation…
Ooh Ooh Ooh! They Finally Did It!
So there has been a long discussion going at PatternSeparation.com about what role the hippocampus and nonhippocampal cortices play for pattern separation. A new paper is available that adds an interesting wrinkle in this discussion by revealing a novel type of nonspatial patern separation in the hippocampus of rodents. The data also suggest there may be two types of patern separation. depending upon the demands of the behavioral task.
The manuscript discussed in this post is “The Role of the Ventral Dentate Gyrus in Olfactory Pattern Separation” by Christy Weeden in Ray Kesner’s lab at the University of Utah. The article can be found in the journal Hippocampus.
Disclaimer: I am in no way unbiased when it comes to singing the praises of anything from Ray Kesner’s laboratory. My first research job was with Ray and I have been an active collaborator with Ray for almost 15 years now. I also helped teach Christy the ventral dentate gyrus lesions they use in the study.
What is Being Studied and Why?
It has been pointed out recently, there are at times differences between the role of a brain region for pattern separation of incoming sensory stimuli and a role for the same brain region for pattern separation when a memory component is added to the task (this was proposed first by Brad Aimone et al., 2011). Simply stated, the memory or task requirements placed upon the individual or animal fundamentally change the nature of encoding – and thus change the manner by which the brain processes individual stimuli and relationships among stimuli.
To test this, Weeden and colleagues designed a behavioral test to evaluate olfactory pattern separation in rats, to see if the ventral dentate gyrus was involved in pattern separation with a memory component, as opposed to pattern separation without a memory component.
An Innovative Approach
Pattern separation at the sensory/perceptual level
An example of pattern separation processes being performed directly upon incoming sensory stimuli is the recent work in the olfactory system. Donald Wilson and co-workers have demonstrated in a number of different tasks that in mixtures of odorants, there is a clear pattern separation effect when a single odorant is replaced with another Link. This effect takes place at the level of the olfactory bulb, suggesting that at this earliest stage of processing stimuli are already processed in an orthogonal manner so far as possible. This means that pattern separation for odors happens at the olfactory bulb!
Pattern separation and completion at the mnemonic level
In parallel with the example given for perceptual pattern separation using olfactory cues, it can be shown that the hippocampus is critical for pattern separation for olfactory cues when a mnemonic component is included. A recent report from Ray Kesner’s lab (that I was involved in) demonstrated that for a series of straight carbon chain alcohols that form an aliphatic series, the ventral, but not dorsal, hippocampus is critical for solving a task requiring the rat choose which of two odors was previously encountered during a working memory paradigm Link. Importantly, rodents performed increasingly poorly on this task as the number of carbons separating the odors was reduced, suggesting increased interference among the olfactory cues – presumably along the domain of carbon chain length. To verify there was not a deficit at the sensory/perceptual level, rats trained to discriminate odors separated by as little as one carbon were able to make the discrimination, so long as the mnemonic contribution to task performance was minimized (i.e., no delay or memory demand was present that required flexible use of olfactory information).
To follow up on this, Weeden and colleagues assessed the role for the ventral dentate gyrus for olfactory pattern separation. This task used a matching-to-sample for odors paradigm. Odor separations of 1, 2, 3 or 4 were selected for each choice phase and represented the carbon chain difference between the study phase odor and the test phase odor. Once an animal reached a criterion of 80–90% correct across all carbon chain separations over 16 trials, rats received a control or ventral dentate gyrus lesion and were retested on the task after a 7 day recovery period.
On postoperative trials, there were no deficits at 15 second delay for either the controls or the ventral dentate gyrus lesioned rats (15 seconds was chosen as it was top speed for the experimenter to switch out the odors). This means that the ventral dentate gyrus is not involved for olfactory pattern separation if there is not a memory component (assuming the rats can easily remember odor information across a 15 second delay).
However, when the delay was increased to 60 seconds rats with ventral dentate gyrus lesions were significantly impaired at short carbon chain separations, but their performance improved linearly as the difference in carbon chain length increased. The performance of rats with ventral dentate gyrus lesions matched control rats at the largest odor based separation. The graded nature of the impairment and the significant linear improvement in performance as a function of increased separation illustrate a deficit in odor pattern separation. Based on these results, it was concluded that lesions of the ventral dentate gyrus impair odor based pattern separation when there is a memory component (in this case a 1 minute delay before the test). The data suggest that the ventral hippocampus, especially the ventral dentate gyrus, but not the dorsal hippocampus, support pattern separation for odor information along the domain of carbon length in aliphatic series.
The addition of this mnemonic component to a task was sufficient to completely change the nature of pattern separation performed required to perform well on the task. If the rats used by Weeden and colleagues showed deficits for the perceptual pattern separation similar to those reported previously by Don Wilson’s group, the rats would have been unable to perform the control task – as the interference among the stimuli would have been insurmountable. Similarly, deficits for the type of olfactory pattern separation being measured by Weeden and colleagues would be irrelevant for performance of the tasks reported by Wilson’s group as these experiments did not require flexible use of the orthogonal representations to guide task performance.
Take Home Message
Weeden and colleagues proposed that it is not only important to functionally dissociate pattern separation across the perceptual and mnemonic levels of processing, but also essential to a more complete understanding of how pattern separation and pattern completion processes influence behavior. As described above, one can see a clear dissociation of mnemonic and sensory/perceptual pattern separation so long as tasks are designed to emphasize one process over the other. These examples illustrate the importance of accounting for task design in studies into pattern separation.
What may appear at first blush to be a paradigm requiring intact pattern separation at a memory level may actually be subserved by pattern separation processes at a more basic perceptual level. Importantly, such data further emphasize that pattern separation exists across all sensory modalities to facilitate orthogonal coding of behaviorally relevant stimuli even at the level of primary sensory cortex–without the assistance of the hippocampus.
However, when a memory load is required such as the requirement that these representations be held in memory for 1 minute between sample and test phases, the hippocampus may be recruited to assist in pattern separation across multiple attributes or domains, and not just for spatial processing.
TL;DR
TL;DR The dentate gyrus is a pattern separator for olfactory information if there is a memory requirement for the task. The olfactory bulb can do the same thing when no memory requirement is made. Future study is needed to know if the computational mechanisms are the same across these brain areas, but I suspect they are.