This paper examines the manner in which labour services are modelled in the aggregateproduction function, concentrating on the relationship between numbers employed and averagehours worked. It argues that numbers employed and hours worked are not perfect substitutesand that conventional estimates of total factor productivity which, by using total hours workedas the measure of labour services, assume they are perfect substitutes, will be biased whenthere are marked changes in average hours worked. The relevance of the theoretical argumentis illustrated using data for the United States and the United Kingdom.