Two-point discrimination is measured as an indicator of cortical reorganisation in musculoskeletal medicine. Nevertheless, data are lacking for the reliability of this measure in patients with non-specific chronic low back pain (NSCLBP). We aimed to quantify the intra and inter-observer reliability of a novel protocol for measuring two-point discrimination in these patients. 35 participants (12 males, 23 females, mean age 52, SD 15years) with NSCLBP were recruited. Three clinicians made 14 consecutive measurements of two-point discrimination with callipers. One of these clinicians repeated the assessment protocol within 7 days. During each measurement, the calliper width was widened in 5-mm increments until participants could consistently identify two points. Intra- and inter-observer agreement was quantified using mean difference, within-subject SD and limits of agreement (LOA). After using the first measurement for familiarisation, the mean of measurements 2 to 5 within an assessment resulted in the optimum compromise between clinic time constraints and acceptable intra-observer reliability; the within-subjects SD being 7.5 mm (LOA: 20.8 mm). Inter-observer reliability was generally poorer; requiring the mean of measurements 2 to 9 within an assessment for a similar within-subjects SD of 8.6 mm (LOA: 23.7 mm). It was estimated that these within-subjects SDs were small enough for a clinically-important change to be detected with a feasible sample size in future studies. The intra-observer reliability of our assessment protocol is acceptable for detecting a clinically relevant difference in two point discrimination for future research purposes. Nevertheless, individual patient measurement variability is relatively high, especially between different clinicians.