Scientific Understanding of Consciousness
Consciousness as an Emergent Property of Thalamocortical Activity

Hippocampus Logarithmic Sense of Space


Science 6 February 2015:  Vol. 347  no. 6222  pp. 612-613

Neuroscience: Our skewed sense of space

György Buzsáki

New York University Neuroscience Institute, New York University Langone Center, New York, NY 10016, USA.


The hippocampus is the brain region where spatial maps of our surroundings are encoded. A specific location will activate a set of neurons called place cells to represent the particular place. What happens as the number of environments encountered increases? Does the hippocampus continually create and store distinct independent “maps” for each locale, or can place cells be recruited for more than one map to generalize across locales? It appears that both mechanisms contribute in unique ways.

At any given position of space, a subset of hippocampal pyramidal cells is active (hence they are called “place cells”), and the firing fields of single neurons (“place fields”) can be regarded as units of spatial representation. Collectively, the active sets of place cells track the position of the animal in the environment, and thus they are hypothesized to provide a “code” for space. But the exact nature of this code is unknown.

Several overlapping stories emerged recently about the statistical structure of hippocampal neuronal activity (firing patterns) and its relationship to “coding” for the environment. The activity of place cells when an animal (rat) experiences a small, large, new, or familiar environment demonstrates that although the majority of these neurons have single place representation, a small minority can have many. Analyses of the observed skewed distribution of place fields and other log-like features of the firing patterns of hippocampal neurons offer a link between physiological organization and the long-known Weber-Fechner law of psychophysics, which describes our subjective perceptions on a logarithmic scale. Accordingly, when the stimulus strength is multiplied, the strength of our perception is only additive.

To examine the relationship between neuronal firing patterns in the hippocampus and the nature of representation of the environment, rats were tested in a familiar open field, a linear maze, and a radial arm maze. Although the majority of CA1 and CA3 pyramidal neurons had single place fields, a small fraction fired at multiple locations. Thus, both the majority and the minority of hippocampal neurons tiled the environments and the distribution of space coverage by individual place cells was strongly skewed. The within-place field firing rates of individual neurons were also skewed and followed a lognormal form (i.e., a bell-shaped distribution on a log scale). In turn, firing rates correlated with both the number and size of place fields.

In a given environment, only a fraction of pyramidal neurons are active. Will every neuron eventually become a place cell if the animal explores a large environment? Another recent study explored this question by training rats to run an expandable maze track with lengths of 3, 10, 22, and 48 m in the same large room. The number of fields formed by the CA1 pyramidal neurons was strongly skewed and showed a log-like recruitment. Nearly all hippocampal pyramidal cells would be active in an environment.

Although the hypothesis of completely independent (or “orthogonal”) representation is supported by the majority of place cells with single-room activity, the “heavy tail” of distributions containing 15% of the neurons active in multiple rooms suggests a more complex picture.

Overall, these studies demonstrate that the skewed distribution of place fields is a general rule, irrespective of the nature or size of the testing environment.

The higher mean firing rates of the active minority within their place fields correlate with their firing rates during sleep in the animal's home cage. Furthermore, the diligent minority fires synchronously with other neurons more frequently in all brain states during both sleep and waking than the slower-firing majority and critically, it exerts a relatively stronger and more effective excitation on its targets. The distribution of the magnitude of collective population firing pattern is also lognormal. The consequence of this population organization is that in the physiological time frames of theta oscillations and sharp wave ripples of neuronal activity, approximately half of the spikes emitted by the hippocampal neurons are contributed by the active minority; the remaining half are contributed by the great majority of neurons with single place fields. This mixed output is what the downstream observer-classifier neurons of the hippocampal output must use to generate action.

This emerging picture of hippocampal dynamics suggests that neurons at the opposite ends of the distributions may convey different but complementary types of information. The ever-active minority of place cells may be responsible for generalizing across environments and affords the brain the capacity to regard no situation as completely unknown because every alley, mountain, river, or room has elements of previously experienced similar situations. In many situations, this minority provides the “best guess” of the hippocampus and offers “good enough” solutions to get by. On the other hand, the majority of less active neurons constitute a large reservoir that can be mobilized to precisely distinguish one situation from another and incorporate novel ones as distinct.

The distribution of synaptic strengths, neuron firing rates, population synchrony, axon conduction velocity, and macroscopic connectivity of neuronal networks throughout the brain displays a skewed, typically lognormal form. The relationships among these multilevel skewed distributions need to be explored to better understand network operations that underlie brain function. An important practical implication of these recent studies is that analyzing physiological data by parametric statistics is a violation because most variables are skewed. The theoretical implication is that brain dynamics supported by lognormal statistics may be the neuronal mechanism responsible for Weber-Fechner (log) perceptions, including our sense of space.

[end of paraphrase]


Return to — Hippocampus