Do I discuss the lack of interaction if I already discussed lack of main effects? I conducted an experiment with 2 independent group variables (gender and language - language indicates either native speakers or 2nd language speakers) and ran a 2 way ANOVA. There were no main effects, and no interaction. My hypotheses had been that there would be main effects and interaction. So, I rejected all 3 hypotheses. I discussed what the reason may have been for having no main effects. Do I separately need to discuss the reasons for no interaction?
Also, even through the interaction did not reach significance, there was a 'trend' for males to perform better in a second language, while for females to perform better in their first language. Do I need to comment on that if I already mentioned it in results section?
 A: Yes, since it was one of your hypotheses, I would at least mention the lack of a significant interaction effect in the discussion, and probably refer back to your reasons for thinking that there would be an interaction effect. Interactions can be significant even if the main effects are not. And for instance you could discuss whether the lack of an interaction puts your results in conflict with other researchers', and if so what the explanation might be. If you suspect that, for example, lack of statistical power is the reason your results don't reach significance, then it might also be worth mentioning the direction and magnitude of the non-significant differences between the groups. It could be useful to check for outliers and whether there was more variation than expected within each of the four (2 x 2) groups, and if so why that might be. Large within-group variation could stop the between-group differences from being statistically significant, and might reflect the influence of an unobserved variable.
A: I don't think you need to discuss the lack of interaction. If you already evaluated that effect in the results section, I doubt there would be any good reason to continue to discuss the point. Perhaps the only exception I can think of is in the analysis of some Conditioned Place Preference experiments, where evaluating the statistical significance of the interaction effect is the primary analysis of interest. On your second point, unless there's something particularly interesting to you about the trend, I don't think you need to mention it again.
