i'am currently reading " theoretical neuroscience by dayan abbott" and it says

" If we ignore the brief duration of an action potential (about 1 ms), an action potential sequence can be characterized simply by a list of the times when spikes occurred. For n spikes, we denote these times by $ti$ with i = 1, 2,... , n. The trial during which the spikes are recorded is taken to start at time zero and end at time T, so 0 ≤ $ti$ ≤ T for all i. The spike sequence can also be represented as a sum of infinitesimally narrow, idealized spikes in the form of Dirac δ functions . "

enter image description here

then it says we average ρ(t) s that we get during the trials .

" We use angle brackets, <> , to denote averages over trials that use the same stimulus, so that for any quantity z is the sum of the values of z obtained from many trials involving the same stimulus, divided by the number of trials. The trial-averaged neural response function is thus denoted by <ρ(t)> "

how can we average ρ(t) . it consists of many infinity long spikes . so summing them and then dividing them by number of trials . is just like summing them . and does not give us the sense of average spikes .

the book further explains :

" In any integral expression, the neural response function generates a contribution whenever a spike occurs. If instead, we use the trial-average response function this generates contributions proportional to the fraction of trials on which a spike occurred. "

the materials I'm referring to are from pages 7 and 8 .


1 Answer 1


ρ(t) in Equation 1.1 cannot be summed and averaged because the idealized spikes are infinitesimally narrow and the chance of two spikes from different trials occurring at the same time t is zero.

In Equation 1.2, the idea of convolving each idealized spike with a "well behaved function" h(t). If you use a rectangular function around zero as h(t), with a reasonable width (say 1ms), then Equation 1.2 will give you a timeseries that can be averaged. This is because the chances of two nonzero values from different trials occurring at the same t will become a more reasonable (nonzero) value.

If you have enough trials, then your trial-averaged neural response function using this more reasonable representation of spikes will give you a good sense of the average spike density.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.