OBJECTIVE: To investigate if heart rate variability (HRV) can be used to evaluate mental workload. METHODS: Forty-six volunteers were selected to simulate mental work by inputting the random number through keyboards. ECG was recorded by Holter 8800 during the performance and transferred into digital signal. Total power (TP), low frequency (LF), high frequency (HF) were analysed by Holter software. RESULTS: Along with time, the inputting speed increased but accuracy decreased, and was positively correlated with TP, LF (r = 0.851 and 0.827 respectively, P < 0.05), showing a distinct dose-effect relationship. HRV values at rest were higher than during mental work [TP: (2 745.4 +/- 1 301.6) vs (686.2 +/- 420.7) ms(2)/Hz; HF: (670.6 +/- 393.8) vs (202.6 +/- 171.7) ms(2)/Hz; LF: (870.5 +/- 553.7) vs (206.0 +/- 187.9) ms(2)/Hz] indicating that they were decreased remarkably as the mental work started. Then, they were gradually increased with the workload (intensity and time) increased, but still at lower level. CONCLUSION: HRV may be used as a parameter to evaluate mental workload but the change pattern of HRV needs to study further.