Little is known about the lying behaviour of sheep despite its high value in supporting productivity and welfare in other livestock. The aim of this experiment was to investigate the lying behaviour of pregnant sheep and test for the effects of biological and physical factors on lying behaviour. Data for 96 Mule ewes managed to lamb indoors, and 80 predominantly Welsh Mountain ewes managed to lamb at grass were used for the study. Acceleration values were collected for the two flocks from HOBO Pendant G accelerometers (Onset Computer Corporation, Bourne, MA) fitted vertically to the outside of the rear right leg and set to record at 1-min intervals for at least 14 d prior to parturition.Ewes were simultaneously recorded using video equipment to identify lambing and to verify predictions of lying (total lying time, mean lying bout duration and mean numbers of lying bouts) using 10 randomly selected ewes from the indoor flock on day -10 prior to lambing. Linear regression was used to evaluate predicted behaviours with video footage. Measures of lying (mean daily lying time, mean lying bout duration and mean daily lying bouts) were calculated for all ewes using averages taken across days -10, -9 and -8 prior to lambing and linear regression was used to test for effects of independent variables (pregnancy scan result (single-or twin-bearing), ewe age, ewe BCS, lambing ease, lamb sex and lamb birth weight) on each measure of lying. Predictions of total lying time (R 2 ≥ 0.99; P > 0.05 for slope = 1, intercept = 0), mean lying bout duration (R 2 ≥ 0.99; P > 0.05 for slope = 1, intercept = 0) and mean number of lying bouts (R 2 ≥ 0.98; P > 0.05 for slope = 1, intercept = 0) were strongly associated with video footage (P < 0.001), demonstrating that a 1-min sampling interval provides reliable estimates of ewe lying behaviours. Significant associations (P < 0.05) were found between measures of lying and pregnancy scan result, ewe age, sex of singleton lambs and twin birth weight for housed, Mule ewes. Only ewe age and twin birth weight were significantly associated (P < 0.05) with measures of lying for Welsh mountain ewes managed at grass. This information could help guide further research on sheep behaviour for management purposes (e.g., to optimise stocking densities and welfare for pregnant ewes). Further work should also consider evaluating measures of lying as proxies for imminent parturition.