Sedentary behaviour (SB) is associated with a number of adverse health outcomes. Studies that have used ActiGraph monitors to define sedentary time tend to use a threshold of <100 counts per minute (cpm) for classifying SB; however, this cut-point was not empirically derived for adults. It is not known whether ActiGraph cut-points for SB differ depending on the context in which it occurs. We aimed to: (1) empirically derive an optimal threshold for classifying SB, using the cpm output from the ActiGraph GT3X+, compared to the sedentary classification from the activPAL3™; and (2) ascertain whether this varied by day of the week and in working time versus non-working time. A convenience sample of 30 office-based university employees (females (66.67%); age 40.47 ± 10.95 years; BMI 23.93 ± 2.46 kg m-2) wore the ActiGraph GT3X+ and activPAL3™ devices simultaneously for seven days. Data were downloaded in 1 min epochs and non-wear time was removed. Generalised estimating equations were used to make minute by minute comparisons of sedentary time from the two devices, using sedentary minutes (when all 60 s were classified as sitting/lying) from the activPAL3™ as the criterion measure. After data reduction participants provided on average 11 h 27 min of data per day. The derived cut-points from the models were significantly higher on a Saturday (97 cpm) compared to weekdays (60 cpm) and Sunday (57 cpm). Derived cpm for sedentary time during working time were significantly lower compared to non-working time (35 (95%CI 30-41) versus 73 (54-113)). Compared to the 100 cpm and 150 cpm thresholds, the empirically derived cut-points were not significantly different in terms of area-under-the-curve, but had lower mean bias for working and non-working times. Accelerometer cut-points for SB can depend on day and also domain, suggesting that the nature of sitting differs depending on the context in which sedentary time is accrued.