We investigate the relationship between hours per worker and employment polarisation. Our core question is whether hours per worker follow the same polarisation patterns as previously observed for employment, measured by either heads or total hours. Using the occupational task index measures of Acemoglu and Autor (2011), we find large relative declines in hours per worker in routine manual jobs − precisely the occupations most negatively affected by employment polarisation from routine-biased technical change. We also find a lower relative decline in hours per worker for non-routine cognitive analytical jobs, which are growing through polarisation. At the same time, hours per worker declined significantly more than the trend for non-routine manual physical occupations. Instead of a polarisation pattern, we find that hours per worker have been declining more in manual jobs (routine manual and non-routine manual physical). These patterns are observed across age, gender and education groups, with few exceptions and changes in intensity. The decline in hours per worker occurred mostly within sectors. Using a wage ranking of occupations instead of occupational task indices, the decline in hours per worker is monotonically related to wages. The results are specific to the European countries and the same patterns are not found using data for the United States.