Frank M. Howell wrote:

> In survey research, it is often the case that an instrument uses > "reverse-coding" to reduce "nay-saying" and "yea-saying" bias during > data collection. These codes tend to get translated into the raw data set > "as is" during the data-entry process. The analyst, however, can forget > that these items should all be scored in the same "direction," > semantically, before conducting index-construction (RELIABILITY, etc.) in > SPSS.

Good point, Frank. I used to work as a graduate statistics consultant at the University of Cincinnati, and this was one of the most common errors that I found when graduate students would bring their work to me.

Dave