Heart rate variability (HRV) from the frequency domain and three calculations of entropy (En), approximate (ApEn), sample (SampEn) and Lempel-Ziv (LZEn), were assessed for intra- and inter-visit reproducibility from short-term recordings in persons with tetraplegia and neurologically intact controls. The intraclass correlation coefficient (ICC) was calculated for multiple comparisons to determine the reproducibility in both groups during a 4 h visit and across three visits. By ICC in both groups, ApEn, SampEn and LZEn possessed excellent intra- (>or=0.87) and inter-visit reproducibility (>or=0.90). In contrast, frequency domain measures were collectively less concise. In general on intra- and inter-visit comparisons for both groups, high frequency (HF) measures were more reliable than low frequency (LF). In control subjects relative to other units of expression, normalized units (nu) of LF had the best intra-visit reliability across all comparisons. This was not the case on inter-visit comparisons where absolute (>or=0.74) and natural log (ln) (>or=0.66) representations of LF were more reliable. In the group with tetraplegia, LFln was the most reliable for comparisons up to or including 180 min (>or=0.60) and two visits (>or=0.63). Thus, calculations of En and HF HRV appear to be more reliable than LF HRV and less confounded by small sample sizes.