Rail cost function analysis has been a popular topic in the (empirical) economics literature over the past decades. Most studies find increasing returns to density for rail companies. The results can, however, be quite diverse. Results for the Dutch National Railway company (NS), for instance, indicate very strong increasing returns to density (Andrikopoulos and Loizides, 1998) or decreasing returns to density (Preston, 1994). Using the, to our knowledge, most comprehensive dataset for the NS, this paper estimates a translog variable cost function for the NS. While the returns to density parameter shows increasing returns, like so many other papers_new, the calculated standard errors shows that the null hypothesis of constant returns may not be rejected.