Aptitude tests: should you guess at random to scrape a few more marks?

We are asked this by email often enough that it deserves a few lines here. As a general rule, and without exception, the answer is no.
What is the situation, though? The candidate is sitting in front of the screen, watching the timer bar creep forward. A fair number of questions have been answered carefully, but the time left clearly will not be enough to finish the test.
The first thing to understand is that some tests are not built to be finished. Attention 2 is a good example: the top band sits somewhere around 45% correct answers. Hoping to work through it in full and with rigour is simply unrealistic. In any case, since the answers are typed in as numbers rather than ticked from a list, the expected gain from guessing is close to nil. Take the mental arithmetic test instead. Each question offers five answers, so in theory a candidate can expect 0.2 marks per question answered at random. On, say, ten remaining questions answered blindly, that is around two marks. All of this assumes that wrong answers are invisible and cost nothing.
Now think about the behaviour expected of a future airline pilot. Intellectual honesty is a phrase that comes up constantly, and it means objectivity, good faith and sincerity in the way you reason. A pilot is expected to be entirely honest in both words and actions. Bluffing, or spinning the wheel when a decision has to be made, is never an acceptable method. Picture an air traffic controller giving you a complex departure clearance on the ground. Reading it back is not a matter of repeating the sounds: you have to be sure the words leaving your mouth are clear in your own mind, and that you have built the right picture of the route to fly. What you say and do should come from genuine awareness of the situation, not from reflex.
Guessing your way through an aptitude test session works exactly the same way, and it is a strong deterrent for a selection board. A rough bluff score is easy to compute for each candidate: the ratio of correct answers to questions attempted on one side, and (correct answers minus wrong answers) divided by questions attempted on the other. When those two figures drift too far apart, the bluffing is plain to see.
So the rule of thumb is this: guessing at random will cost you more than it earns you, and the faintest suspicion of intellectual dishonesty is disqualifying in a selection process. The moral is a simple one. Do your best, and answer honestly.