I don't know how many places this experiment's been carried out in, but this example is from a sociology class (as far as I remember) in UCD, Ireland.

Students were handed out questionnaires asking them to fill in their birth dates, places, times, and other astrology-relevant information. Personalised character profiles were produced for each of them, and delivered a few weeks later. Each student was then asked to rate their personalised profile based on its accuracy, etc.

More than 90% of students rated their personalised character assessment as 'good' or 'very good'. Based solely on astrological information provided by them, researchers had been able to identify character traits, and write an accurate report on that person.

The problem was that all the 'individual' character profiles were the exact same. The researchers had written up a basic profile, consisting of information that everyone wanted to hear. I can't remember the details, but it probably consisted of statements such as 'You are well-liked by most people' and 'You consider yourself a fairly good worker'.

The fact that most students saw this single profile as relating to them personally highlights the care that must be taken when generating and reading horoscopes, and could even be seen as evidence deriding them completely.