After each session, HRV data was downloaded to a computer using Polar® Software and converted into an ASCII file. R–R interval data was analysed using Kubios HRV software (Version 2.0 Biosignal Analysis and Medical Imaging Group (BSAMIG), Department of Physics, Univer- sity of Kuopio, Finland (http://bsamig.uku.fi)). Prior to analysis, R–R interval data was scanned manually and artefacts removed using Kubios' inbuilt ‘artefact correction’ feature. HRV parameters were calculated for three, 5 minute sections, selected at random, from each 90 minute recording session. The following time-domain variables were chosen for further analyses: Mean RR (μRR); Standard Deviation of R–R intervals (STDRR, ms); Mean Heart Rate (μHR); Root Mean Square of the Standard Deviation (RMSSD, ms); R–R interval Triangular Index (RRTI) and a number of successive R–R interval pairs which differ more than 50 ms (NN50) expressed as a percentage (pNN50, %). The only frequency-domain variable taken forward for further analyses was the ratio between the low frequency (LF) and high frequency (HF) band powers (LF/HF). In this study the nonlinear properties of HRV were