Conclusions, Dos and Don'ts

Conclusions

Only constructive criticism works ultimately; learn to give and take constructive advice in writing and reading reviews.
Work toward a better hypothesis, thesis or theory.
Decide what theory you are using and are NOT trying to endanger. Work with hypotheses derived from that theory. Then check to make sure that there is something intermediate in scope between your theory and your individual hypotheses (that might be endangered if your work is convincing enough).
To have a good hypothesis (little prediction from a bigger idea), you need a thesis (an overarching idea).

Develop multiple working hypotheses that can be falsified by data.
Beware especially of single hypotheses that have a person's name attached, especially your own.
Avoid hypotheses that cannot be falsified (by observation).
Maintain healthy skepticism of data as well as hypotheses.
Realize that any acceptance of a hypothesis is tentative and impermanent if a better hypothesis comes along or if the data prove to have been faulty.
Avoid collecting data and then making up a story about them (an ad hoc theory or hypothesis) without the intent to test the validity of that same story on new data.
Work toward or from predictions; take observations after making specific predictions.
Don't confuse empirical predictions based on observed statistical regularities with predictions based on mechanistic understanding. Empirical models do not predict except in the hollow sense that future data will behave like past data. Empirical models themselves are still useful for empirical estimation (assuming that observed relations in past data will hold in future data) and as sources of inspiration for hypothesis generation.