Software system size provides a basis for software cost estimation and management during software development The most widely used product size metric for the specification level is Albrecht's Function Point Analysis (FPA). Symons has suggested an alternative for this metric called the Mark lI metric. This metric is simpler, more easily scalable, and better takes into account the complexity of internal processing. Moreover, it suggests different size values in cases where the measured systems differ in terms of system interfaces. One problem in using these metrics has been that there are no tools that can be used to calculate them during the specification phase. To alleviate this we demonstrate how these metrics can be automatically calculated from Structured Analysis descriptions. Another problem has been that there are no reliable comparisons of these metrics based on sufficient statistical samples of system size measures. In this paper we address this problem by carrying out preliminary comparisons of these metrics. The analysis is based on a randomly generated statistical sample of dataflow diagrams. These diagrams are automatically analyzed using our prototype measurement system using both FPA and the Mark II metric. The statistical analysis of the results shows that Mark II correlates reasonably well with Function Points if some adjustments are done to the Mark II metric. In line with Symons's discussion our analysis points out that the size of correlation depends on the measured system type. Our results also show that we can derive useful size metrics for higher level specifications and that these metrics can be easily automated in CASE tools. Because the obtained results are based on simulation, in the future they must be corroborated with real life industrial data,