In Why my mom didn’t buy me a slot car track for Christmas in 1978, even though I told her that’s what I wanted, Dave Munger looks at a study of couples and how they predict each other’s taste:
Lerouge and Warlop recruited 35 couples to participate in the study. Each partner was placed in a separate cubicle for the duration of the study, and had no contact with his or her mate. They then were shown pictures of 30 different sets of bedroom furniture and asked to indicate their impression of it (positive or negative). The next part was the key to the study:
Each partner was shown 30 new furniture sets — the sets the other partner had already rated. In each case, they were asked to predict how another person would feel about the furniture. Half the participants were told they were predicting their own partner’s preference, and the other half were led to believe they were choosing for a stranger. In fact, all participants were predicting their own partner’s preferences. After each prediction, the partner’s actual preference was revealed, so as the study progressed, presumably each participant would better understand his or her partner’s preference, and gradually make better predictions.
Overall, participants were better at predicting their partners’ tastes when they believed they were predicting the tastes of a stranger. When they knew they were predicting their own partners’ tastes, the accuracy of predictions depended on how similar their own tastes were to that of their partners.
Here’s the key finding: When partners’ tastes were different from their own, then they were better at predicting each others’ tastes if they believed they were predicting the tastes of a stranger.
[...]
When you know someone well, it appears, you begin to assume they have the same tastes as you do.