The Discussion section takes a broader look at the results presented in your paper, and attempts to interpret them in the context of previous studies. It answers the question: What do these results mean, and how have they changed what we know about this field?
The first step is interpretation of the data. Do they support the hypothesis being tested, or do they falsify it? If the data support the hypothesis, do they provide strong support, or are there other hypotheses which would explain your results? If more than one hypothesis is being tested, or more than one set of results are reported, deal with each specifically rather than making broad generalizations.
The second critical portion of your discussion puts your results in the context of previous studies. Your Introduction outlined what we knew about the field going into your study. The Discussion examines what we know coming out of the study (Penrose and Katz 1998). Again, deal in detail with specific results from your paper, and specific papers which have previously been published. Are your results consistent with other studies or contradictory? What possible explanations are there for discrepancies?
One common mistake students make is to just blame their own poor experimental technique for any discrepancies between their results and others. It’s great to be honest about the limitations of your work, but usually there is something more interesting going on than just technical problems, and you should consider other possibilities (Pechenik 1997). That said, it’s sometimes useful to comment on specific techniques used in your study, how they might have affected the results, and how they could be changed or improved. Suggest what limitations need to be placed on your conclusions, given the limitations of your study. Are there questions that remain unanswered?
New questions may lead you to new hypotheses that are suggested by your results, or new experiments which might shed further light on the subject matter. Were there any unexpected or unusual results? Comment on them and speculate on the explanation.
And finally, if your results could be applied in some way for practical benefit, discuss how this might be approached.