We have used a Galactic sample of OB stars and associations to test the performance of an automatic grouping algorithm designed to identify extragalactic OB associations .
The algorithm identifies the known Galactic OB associations correctly when the search radius ( 78 pc ) is defined by the observed stellar surface density , which suggests that the sample of Galactic OB associations constitutes a reasonably uniformly-identified sample .
Roughly 25 % of the groups identified automatically which contain more than 10 stars actually comprise two or more unrelated OB associations , which raises the concern that the largest extragalactic associations identified via this algorithm may also represent multiple associations .
Galactic OB associations identified with a 78 pc search radius have diameters that are \sim 3 times larger than OB associations identified with a 22 pc search radius in M33 .
Applying the smaller search radius to the Galactic data matches both the sizes and the number of member stars between the two galaxies quite well .
Thus , we argue that this and similar algorithms should be used with a constant physical search radius , rather than one which varies with the stellar surface density .
Such an approach would allow the identification of differences in the giant molecular cloud populations and star formation efficiency under most circumstances .