How can the interaction between theoretical neuro-scientists and their experimental counterparts be improved? This article discusses a number of suggestions relating to the presentation of data in experimental studies. In particular, published data should account for the diversity of response properties encountered, rather than concentrating on the 'representative' response, as well as emphasizing the stochastic nature of neurons by routinely including raw, unprocessed data from individual trials, which show the degree of variability prior to averaging.