Spousal characteristics such as age, height, and earnings are often used in social science research to infer social preferences. For example, a "male taller" norm has been inferred from the fact that fewer wives are taller than their husbands than would occur with random matching. The large proportion of husbands out-earning their wives has similarly been cited as evidence for a "male breadwinner" norm. This paper argues that it is difficult and potentially misleading to infer social preferences about an attribute from observed marital sorting on that attribute. We show that positive assortative matching on an attribute is consistent with a wide variety of underlying preferences, including "female taller" or "female breadwinner" norms. Given prevailing gender gaps in height and earnings, positive sorting implies it will be rare for women to be taller than, or earn more than, their husbands – even if there is no underlying preference for shorter or lower-earning wives. In an empirical application, we show that simulations which sort couples positively on permanent earnings can largely replicate the observed distribution of spousal earnings differences in US Census data. Further, we show that an apparent sharp drop in the distribution function at the point where the wife begins to out-earn the husband results from a mass of couples earning identical incomes, a mass which we argue is not evidence of a norm for higher-earning husbands.