There's an interesting short in this week's Economist about people and their dogs. Volunteer observers were correctly able to match dogs to their owners, in individual photographs stripped of other information. They were able to do this at a significantly higher-than-chance rate. This only seems to work with pedigreed dogs- a second study (this is a hot topic!) found that volunteers were not able to reliably match mongrels with their owners.
The explanation offered is that humans tend to form emotional bonds on an "assortive" basis, most prominently in the choice of mate, but also apparently in choice of pets. Mongrels are less reliably matched, either because no one can predict what they'll look like, or because they get adopted under a wider range of circumstances compared to purebreeds.
They offer an explanation for the evolutionary basis for assortive bonding, which I found a little thin (avoid incest but optimise outbreeding). But it has indeed been documented that married couples can be identified in just this sort of test.
The scientific authors think this phenomenon might affect a number of "gut decisions," for instance choice of business partner. Of course, no one can yet explain Macintosh usage.
Subscription link (Journal of Ethology) is here. I was sorely tempted to break copyright and reproduce the figure 1, showing the lineup. There's some great hair going on there.
UPDATE: Oh hell, if you want to laugh, look here.