Pengwuino said:
I actually think its an ok idea. You can realitically get a better grade then someone who knows less then you simply because you guessed better. I mean when it comes down to it, you're basically trying ot reward ignorance. When you're doing a multiple choice test, guessing shows nothing about your knowledge.
I've thought an interesting variation on the negative marking theme would go like this :
For every question, in addition to marking your answer, you have to mark out one of three circles to denote your level of confidence in your answer. The circles go from "Somewhat Unsure" to "Ambivalent"/"Moderately Sure" to "Completely Sure". In the case of complete lack of knowledge (pure guesswork) the circles are left blank (the answer can also be omitted, it makes no difference).
Grading goes like this : If the student marks the third circle "Completely Sure" for any question, he/she gets full credit of +3 if right and full negative credit of -3 if wrong. Similar considerations apply to the other two circles with the credits decreasing in magnitude (+2/-2 and +1/-1). Guesswork is neither rewarded nor penalised with a score of zero points.
The total score for a 100 question test can range from a minimum of -300 (all questions wrong with full (misplaced) confidence) to a maximum of +300 (all questions right with full (justified) confidence).
It seems time consuming to fill up these circles, but in practice it may not be so. Most people do the ones they're most comfortable with first and leave the iffy ones blank. After the first round, they can come back and tackle the iffy ones and shade either the first or second circles accordingly. Then for the rest of the questions they had already answered confidently at the start, just go down the line shading the third circle.
Just to give the student the benefit of the doubt, in a case where any of the confidence level circles are filled but no answer is given, the item is treated as a pure guess with no marks being given or taken away. And in cases where more than one confidence circle is filled in with an answer being marked, the level is taken as '2'.
I think this is a good system to implement, especially in somewhat subjective disciplines. I remember a Histology test I took in my first year with a rather green lecturer from India. Her lecture notes said something which my textbook flatly contradicted, but I did not have time to challenge her before the test. Of all things, that item came out on the test. I wasn't sure if she was going to be marking the test, so I didn't know whether to use her (wrong) version or the textbook's (correct) version. In the end, I just left that choice blank to get zero credit rather than risk negative marking.
If I had to indicate a level of confidence like in my testing model, I would've marked the correct (textbook) option and indicated a confidence level of '1' (somewhat unsure) for a possible credit of +/-1 rather than +/-3.
If 3 circles is too complicated, a 2 circle system may be feasible : "Mostly Unsure" and "Mostly Sure".
It's okay, I understand, your brain is a bit fried tonight. You were talking about biology, were you not? That's what I was referring to as well (but giving them, not taking them). 
Four weeks and two days till I leave for NM!
