Analysis of parallel spike trains
The brain is composed of billions of neurons, the elementary units of neuronal information processing. The neocortex, which is critical to most higher brain functions, is a highly complex network of neurons each of which receives signals from thousands of other neurons and projects its own spikes to thousands of other neurons.
In order to observe neuronal activity in the active brain a large variety of recording techniques are being employed: intra-cellular recordings of the membrane potentials of individual neurons, extra-cellular spike recordings from one or more individual neurons, and recordings of signals that measure the activity of populations of neurons either locally as the local field potential, or from larger brain regions via the EEG, MEG or fMRI. Any particular choice of the recording technique reflects the hypothesis the researcher has in mind about the mechanisms of neuronal processing.
The focus on spike recordings from individual neurons implies that one strives to understand the elementary units of neuronal processing. Early electro-physiological experiments were bound to record from single neurons only. The resulting insights are now the basis for the “classical” view of sensory coding: firing rates are modulated in a feed-forward hierarchy of processing steps. Signals from sensory epithelia are assumed to eventually converge to cortical detectors for certain combinations of stimulus features. Highly specific percepts would be represented by the elevated firing of a single nerve cell or a small group of neurons. Due to a number of conceptual shortcomings, however, It has been seriously questioned, whether such a scheme qualifies as a universal method for representation in the brain.
It was Donald Hebb (1949) who first demonstrated the conceptual power of a brain theory based on cell assemblies. Inspired by Hebb’s influential writings, and driven by more recent physiological and anatomical evidence in favor of a distributed network hypothesis, brain theorists constructed models that rely on groups of neurons, rather than single nerve cells, as the functional building blocks for representation and processing of information. Despite conceptual similarities, such concepts of neuronal cooperativity differ in their detailed assumptions with respect to the spatio-temporal organization of the neuronal activity.
To understand the principles of coordinated neuronal activity and its spatio-temporal scales it is obligatory to observe the activity of multiple single-neurons simultaneously. Due to recent technological developments in recording methodology, this can easily and regularly be done. Coordinated activity of neurons is only visible in (wide-sense) correlations of their respective spike trains, which typically admit no simple interpretation in terms of fixed synaptic wiring diagrams. Rather, it became evident that the correlation dynamics apparent in time-resolved multiple-channel measurements reflect variable and context-dependent coalitions among neurons and groups of neurons. Thus, the analysis of simultaneously recorded spike trains allows us to relate concerted activity of ensembles of neurons to behavior and cognition. Different analysis approaches are thereby relevant to distinguish different or even complementary spatio-temporal scales. The analysis of parallel spike trains is a necessary step to understanding neuronal mechanisms underlying information processing in the brain. The book on “Analysis of parallel spike trains” summarizes concepts of analysis approaches to such data and provides related software.