Weird interaction results - need someone's help!

#1
I ran some hierachical multiple regression in SPSS, and plotted my significant interactions (I have all continuous predictors and outcomes). One of the interactions doesn't make sense. The slopes look normal, but the problem is the Y axis. My outcome variable is a teacher rating of children's literacy, and it's on a 1-4 scale. However, when I calculate the regression lines for the interaction, I get all negative numbers - in other words when I plot it, all of the Y's are negative. See attached plot.


I've pretty much ruled out the following things:
  1. my predictors are centered correctly
  2. I have calculated the regression lines both by hand and by computer, and get the same negative values
  3. I (and my advisor) have checked my regression output to the values I used to plot, and I didn't use the wrong values
  4. Missing data values aren't messing up my variables


I have figured out that the problem is probably with the negative constant in the regression equation. I don't think I should have a negative constant after I've centered my predictors, but I do. Oh, and I didn't accidently center the teacher literacy outcome, so it's not that.

Any ideas? :)