I actually think the first 5 questions on quant have a pretty dramatic impact on your score.
The GMAT is designed so that no single question will tank your score. If you got the first question wrong, and proceeded to get the following 5-6 right, the impact would be minimal.
However, if you get the first question wrong, second question right, third question wrong, fourth question right, fifth question right, then I think it would have a fairly significant impact.
If you do any reading about CAT tests, the whole idea is that it can very quickly determine your ability level. Once it brackets your ability level, it measures whether or not you are actually at that level by keeping you at a 50% failure rate, meaning you get one right, one wrong in sequence consistently. The idea is that your 50% failure rate is the point where if it gets any harder, you always get it wrong, and if it gets any easier, you always get it right.
However, the GMAT is an imperfect CAT. A perfect CAT would keep giving you questions until it is 99% confident that it has successfully bracketed you. So if there are variations from the 50% failure rate, it would keep fine tuning until you got there.
The GMAT, however, only has 37 quant questions in which to determine your ability level. Of these 37, approx. 1/4 are unscored questions that GMAC is testing out in order to ascertain their difficulty level.
What this means is that, in reality, there are only 27 questions that the algorithm has to work with in order to determine your score. This is not a large sample set. So the algorithm has to quickly determine your estimated ability level. This means that early questions do count for quite a bit on the GMAT.
Anyway, getting the first question wrong is not a deal breaker. However, it does make the mountain that much harder to climb, and requires that you be much more diligent in subsequent questions.
My feeling is that the first question is always pretty straightforward math. They don't ever throw curve balls on #1. It is almost always a problem solving question type, and conforms generally to basic concepts, except they are usually of above average difficulty.
What this tells you is that they do rely on it pretty heavily to separate test takers. #1 is a litmus test where the most advanced test takers should get it right every single time. The least advanced should get it wrong every single time. And the middle of the field might be hit or miss.
