DM: The Baron tells you, “I haven’t seen the Duke.”
Player 1: I roll Insight to see if he’s lying. Crap, rolled low.
Player 2: I roll then. Mine is pretty high.
Player 3: My bonus isn’t very high, so I roll to assist.
This happens for many kind of skill rolls. An idea occurs to one player, who rolls a skill, which then prompts several other players to also roll. If a check is declared, then everybody rolls who can roll.
You can’t blame the players: after all, rolling dice is fun, and if you roll high, you get the spotlight, and probably advance your quest. (Succeeding in a skill roll is rarely a bad thing in D&D, since there’s no cost involved.) However, the end result is that the average skill roll result is inflated, probably past what the designers intended. It also means, especially in groups with lots of players, NPCs don’t stand a chance. There can potentially be one chance per player to roll high against an NPC’s skill who only gets one check or DC. It also serves to take a bit of the emphasis off of a PC’s particular trained skill.
So what are some options? First, given the scenario, you can always just restrict who can roll in a given situation. Maybe only the person actively engaged with the conversation can roll an Insight check (or only the person actively observing and not talking is the only one who can roll.) The same can be applied to assisting others- my personal pet peeve is one PCs try to assist on knowledge checks to recall something specific. While the argument could be made that they’re sharing information with each other, I always picture one PC massaging the brain of the other one.
Skill Challenges, depending on how they’re handled, can be a good way to limit this. Instead of one roll determining success or failure (and thus encouraging everyone to focus on the same roll), a series of skill rolls, with PCs having to choose where to direct their efforts. PCs can still assist or try to do the same thing as another PC, but there is a two-fold cost: opportunity cost on the round and the possibility of incurring a failure.
Finally, you could create staggered amounts of information for each skill roll. The person who rolls the highest on a knowledge check gets one key piece, the second highest gets another piece, and so on. (For added fun, insert bad pieces of data. This often happens when someone critically fails, though is likely to be dismissed as a joke.) This also won’t work in the initial example, where the answer tends to be yes or no.
It’s an issue that I have yet to fully tackle in my game, having gotten in this paradigm of allowing as many rolls as people who speak up. The more I want to create nuances in my campaign of information gathering though, the more it annoys me, and the more I’ll try and restrict it.
How about you? Any suggestions on this problem? Is it a problem in your games?