Single channel patch clamp recording allows modelling of ion channel behaviour. Thus, a single kinetic model can formally describe the complex interactions during poly-modal activation of the channels. For TRPM8, activation depends on temperature, voltage and chemical signalling (Voets et al., 2007), and such modelling provides the most comprehensive approach to the study of the channel. Here we show the methods used to derive the kinetic signature of the TRPM8 channel from a set of single channel recordings. Single channel (cell-attached patch mode) currents were recorded in HEK-293 cells stably expressing the TRPM8 channel. Data was recorded at room temperature and different voltages (from 40 to 140 mV). Currents were filtered (low pass Bessel) at 2 kHz and sampled at 10 kHz. Single channel transitions were identified by the half-amplitude threshold crossing criteria and time-course fitting (Colquhoun and Sigworth, 1995). Recordings were made from 7 different patches with at least 12000 openings in each recording (all values were expressed in means ± S.E.M.). Three steps were used to derive a kinetic model from single channel recordings: 1. The data: Dwell time 1D histograms were constructed from transitions >0.16 ms using the square root of the number of events per bin with a constant bin width expressed on a logarithmic time axis (ms, 20bins/decade). These histograms were fitted with a sum of exponentials using pClamp 9. The appropriate number of exponentials was assessed both by visual inspection as well as a statistical estimation of the ‘goodness of fit’ (Maximum Likelihood and Least Squares Errors). These methods pointed to the need of 2 to 4 open states (mean open time ranged from 0.2 ± 0.02 to 8.95 ± 0.35 ms) and 4 to 6 closed states (mean closed time ranged from 0.11 ± 0.01 to 671.04 ± 140.24 ms) to fully describe the data. 2. Modelling: There are two Freeware software packages for Markov analysis of single channel data: HJCFIT (http://www.ucl.ac.uk/) and QuB (http://www.qub.buffalo.edu/). They were used to fit single channel sequences to different models and solve for the most likely rate constants. In order to discriminate between models, dwell time 2D distributions (Rothberg and Magleby, 1998) and dependency and dependency difference plots (Magleby and Song, 1992) were also used. 3. Testing the predictions: Lastly, the correspondence between the predictions derived from the model and the values initially used to describe the experimental recordings was tested. This study provided evidence that the response of TRPM8 channel to voltage variations can be described in terms of a model that includes channel transitions involving 2 to 4 open states and 4 to 6 closed states. The most likely transitions that are voltage regulated have also been identified.