In the name of science

Telling your subjects that they can die will cause the less resolute ones to quit. Therefore, you wouldn't be testing a representative sample, because you will prompt half the sample to quit, therefore, you'd be only testing 'selected' people.

And you don't think the results will change? In Zimbardo's experiment, it's possible that if only the braver people remained, they may not incorporate the conditions at all, rather, they might consider them above that and just sit in the corner and refuse everything. (After all, the guards were not supposed to harm the prisoners.) Thus, this would cause the participants to fail, because they are not fully participating in the experiment.

If you're referring to Zimbardo's experiment, the results would probably not of changed. The testers chose 24 out of 70 who were the most psychologically stable and healthy, aka resolute.