Contrary to prior assumptions, matched groups differed from each other nearly as much as they did from the unmatched group. This has implications for the ability to generalize from interspousal data even when comparison groups are matched on several traditional variables. (Author)