The definition of reinforcement is that it increases behavior--always. If behavior isn't increasing, it wasn't a reward. However, a lot of these studies are actually using something that's more like a "lure." It's the equivalent of training a dog to heel by holding a hotdog by your hip. You might get great heeling when you have a hotdog there--if there aren't any squirrels around.

When you stick the incentive out there in front of them and they know that's what they will get, they can then weigh how much they want that incentive vs other things they could be doing with the same amount of effort. The secret is to not tell people ahead of time you're planning to reward the various levels, and then just decide after the fact who met the criteria and reward them.