I do not think that comes to the opposite result. The only variant they test that has Cancel, OK in that order also is the only one where the rightmost button is far away from the other content. How far way, the report does not state. Being a web study, this could have been dependent on window width.
The study also doesn't go into the consequences of various error rates. It is just about “Please complete the form fully and accurately". It does not test average performance if, say, users want to cancel out of the input in 5% of cases. It does not even ask me whether they feel confident that they can do so if they want.
The study also doesn't go into the consequences of various error rates. It is just about “Please complete the form fully and accurately". It does not test average performance if, say, users want to cancel out of the input in 5% of cases. It does not even ask me whether they feel confident that they can do so if they want.