Recent empirical evidence based on extensive databases shows that firm size distributions (FSD) vary with the sample. This paper analyses the effect of sample size on the FSD of Spanish manufacturing firms for the years 2001 and 2006. We use a comprehensive dataset that has two measures of firm size: sales and employment. Our database shows a skewed FSD to the right which there are numerous small firms and a few large firms. Applying a rolling regression to control for sample size developed by Peng (2010), we show the existence of a non-constant power-law distribution that depends on the sampling size. Furthermore, the FSD of employees is more sensitive to firm age than the FSD of sales.