This study aimed to quantify the intra-and inter-instrument reliability of the Actiwatch 4 accelerometer (AW4) in a mechanical setting. Twenty seven AW4 were attached to an isokinetic dynamometer and subjected to angular acceleration for 30 min at 50 deg/sec representing moderate intensity (MPA condition) and 200 deg/sec representing vigorous intensity (VPA condition), with a repeat trial conducted. Reliability was assessed using coefficient of variation (CV), absolute percent error (APE), and intraclass correlation coefficients (ICC). Mean AW4 activity counts displayed acceptable reliability according to CV in both conditions (ConMPA: CVintra = 4.6%, APEintra = 6.6%, CVinter = 6.4%, APEinter = 5.2%; ConVPA: CVintra = 3.9%, APEintra = 5.6%, CVinter = 5.9%, APEinter = 4.7%). When counts were re-categorised into minutes of MPA and VPA, lower CV values were observed (ConMPA: CVintra = 3.2%, APEintra = 4.5%, CVinter = 4.3%, APEinter = 3.2%; ConVPA: CVintra = 0.0%, APEintra = 0.0%, CVinter = 0.0%, APEinter = 0.0%). When activity counts were re-categorised as minutes of MVPA, excellent reliability was observed (CVintra, APEintra, CVinter, and APEinter = 0.0%) in both conditions. Mean AW4 activity counts exhibit reliability statistics comparable to other accelerometers. Reliability is improved when activity counts are re-categorised as time spent in MPA and VPA, with greatest reliability obtained when counts are recategorised as time spent in MVPA. As MVPA is the subcomponent of physical activity most associated with health benefits it would appear that the AW4 is reliable for measuring time spent in this health enhancing intensity category, at least from testing in a mechanical laboratory setting.