Who is right, the statistician or the surgeon? Consider the case described below, from Peacock (1972). This passage seems to imply the young statistician is making a smart, correct statement. 
But is he?

 A: As the statistician did not make any statements, he cannot be wrong. He just asked two questions: 1) Did you have controls? and 2) Which half?
The surgeon is clearly wrong, unless a) Every patient he treated survived and b) No patient who was not treated would survive (or, of course, vice versa). 
Both the surgeon and the statistician are making good points. 
A: This sounds a lot like that story about one of the sons in the fourth generation of the Pearson family, the one that became a paramedic. He used to not help half of his patients with a cardiac arrest in order to test whether helping or not helping was significantly helpful in order to get the heart beating again. 
A grand child of Joan Fisher and Joerge Box is currently doing a project for the final exam as air traffic controller. He is testing on half the pilot whether they will fly better and crash less often if he is not speaking to them.
Do you think they are right to do so?
A: The statistician sounds like a frequentist, and he's correct if we view things in terms of measures of evidence. In particular, at this point we have no direct evidence in regards to the effectiveness of the surgeon's effectiveness. 
Maybe surprising to most statisticians, the surgeon is taking more of a Bayesian perspective. That is, because of his advanced knowledge of medicine, he is very strongly convinced that his procedures are helping his patients. He's human, so he must realize that he does know exactly how effective his treatments are, but he also is so confident that it's positive that the long-term benefit is better for him to treat every patient than it is to collect controls, who will with very high probability be worse off than if they were treated only to collect data that confirms what he already knows. So while collecting data on controls may be informative, it is dangerous to the controls and not likely to make any differences in future decisions. Therefore, it is quite logical for him to not use controls. 
Who's correct? Well, the statistician is certainly correct that we don't have any data that demonstrates that the surgeon's methods are effective.
But the lack of evidence doesn't mean the surgeon is wrong! Assuming the surgeon is not over-confident, the surgeon is also correct that collecting data on controls is not the ethical thing to do. What it all comes down to is: do you trust the surgeon's confidence? 
A: The surgeon is right. 
The people who suffered or died because they did not get this operation serve as a control group. It would be better to formalize this and quantify the improved performance (e.g. 70% mortality rate vs 10%), but we do have a group to which we can compare.
Now...if the surgeon is claiming that his treatment saved lives, yet the patients tended to do just fine without the procedure, then the success of the treatment is not so remarkable. However, quite the opposite is implied.
The "which half" line is wrong. Nothing suggests that the surgeon's procedure causes death. Perhaps it doesn't help compared to a control group, but it certainly sounds like most patients survive. Operating on a patient certainly doesn't suggest that they are doomed to die in the OR.
