If we are comparing two exercises with surface EMG on the same muscle activation, what increase in EMG would make a clinician choose the higher activation exercise over the other?
if you want to examine sEMG changes in several patients / people, you may be able to determine the MCID (minimal clinically important difference) based on your results using a distribution-based approach. Also see Copay et al. http://www.ncbi.nlm.nih.gov/pubmed/17448732
If you examine sEMG changes in a single person, you probably will have to rely on reliability studies. This resource will also be helpful: http://www.sportsci.org/ => assessing an individual.
I support the approach of Carsten to determine the MCID in your study. In addition, I think there are further points to consider. First of all, the question “how much” a muscle is active during different exercises cannot be answered by the original microvolt scaling, because the original electrical muscle activity is influenced by the local given detection condition that can greatly vary. Therefore the sEMG signals should be normalized (e.g. see http://www.ncbi.nlm.nih.gov/pubmed/20702112). Finally, if you are comparing two exercises with sEMG on the same muscle activation, you may bear in mind that muscular fatigue could occur and affect your EMG signal (normally both amplitude and frequency based analysis parameters show time domain changes due to muscular fatigue).
I would have to agree/support Dennis and Carsten regarding the within subject and across subject or day considerations you must consider when evaluating these data.
I would also add that such a clinical assessment of a higher or more active might muscle also require net activation through the range of motion achieved by the task. For example activation of a muscle during one task (e.g. leg press) might have a different peak amplitude, but might not require as much net activation through the range of motion achieved by the other task. Therefore, clinically choosing a task that involves activation through a greater range might be just as (or even more important) than just the peak.