To modify RangerRick's suggestion for the Monk dropping his arm, I'd assume a fumble mechanic has 2 parts. A. Did a Fumble happen and B. what was the result.
If B. yields a nonsensical result (monk drops weapon) then reroll the result
I think people are confusing Critical Hits and their meaning with Fumbles and their meaning (or at least as I intepret them).
If we assume that in D&D3e, each die roll equals one sword swing (which is reasonable in a 6 second round, compared to 2nd Ed's 1 minute round), then we can discern the following:
For every HIT, there is a chance of it being a critical hit (i.e. hits a very sensitive spot)
Most people also seem to be coming to the following conclusion for the inverse:
For every MISS, there is a chance of it being a Fumble (i.e. you missed and hit a stone pillar)
I'd say they're half right. I think you can a fighter can fumble, regardless of whether he attacked, or hit or missed.
If a fighter stands there for 6 seconds blocking blows (fighting defensively? I forget what they call it), he's making no attack rolls, but getting a higher AC. What's undocumented in the mechanics is that essentially, he's still moving around in his 5' square and swinging his sword to deflect blows.
Logically, he could slip on some blood, drop his sword, break his sword in a mighty block.
My point is, it is bad design to put fumble detection on the Attack roll of the person making the attack. Since that means you only fumble when you attack.
My wording above leads us to another proposed solution, Attackers may cause defenders to fumble. Put a mechanic on the Attack roll, that checks to see if there is additional effect on the target.
This would support the following logic of non-fumbling:
if you are doing nothing, and nobody is trying to do something to you, you shouldn't fumble.
Just standing there is pretty safe, but if somebody is trying to hit me, I could slip, drop my weapon, or it could even break when you hit it.
If we were to assume that the attack rolls are NOT the only sword swings going on, than my argument is still supported. A 1st level PC making a Full Attack action would be waving his sword around just as much as a 20th level PC, it's just that the high level PC has more real chances of connecting a hit.
Basically, I'm against the most popular mechanic for determining a Fumble, which is, rolling a 1 for attacks triggers a Fumble check.
To sum up what a "realistic" fumble system would have to support, I propose the following criteria:
anybody can fumble
higher level character has less chance of fumble (overall, not just per attack)
I can fumble whenever I try to do something
I can fumble if someone is trying to do something to me
if I'm doing nothing, and nothing is happening, I won't fumble
doing mundane things has no chance of fumble
doing things under pressure has a chance of fumble
A fumble implies one or more of the following effects:
drop something
tangled in gear/clothes
break something
slip
slip and fall
hit the wrong thing (friendly fire, or scenery damaged)
locked blades (or arms on a monk)
I think the effects of a fumble is pretty trivial to figure out. That's just a table with your favorite effects, weighted towards the most likely. To each his own. But the mechanics for determing when a fumble effect occurs have been pretty weak.
I also think that 3.x's combat actions have been moving towards fumble effects, in terms of you actively take an action to induce the effect on your opponent. Rather than it being a purely random incident. Take the Book of Iron Might for instance. Lots of cool things to do to your enemy.
Janx