Re-analysis of data from Salisbury (1974) shows no significant relationship between seed size and habitat conditions, a relationship for which this paper is frequently cited as support. The discrepancy between the results reported here and those found in the original paper is due to Salisbury's ‘evidence’ being a table of congeneric pairs collected specifically because they possessed the target relationship. This paper emphasizes particularly the need for an unbiased sampling of species pairs in comparisons attempting to define ecological relationships. The analysis presented here represents the fourth such instance in which previously assumed robust examples of the relationship between seed size and habitat type are shown not to be so when the appropriate methods are used to account for relatedness among species.