Ever see a dog that when asked to do something, runs through his entire repertoire of tricks until he hits upon the one that gets rewarded? That's because none of those tricks (behaviors) are on cue. The dog just guesses as to which one you're asking for, and stops when he gets to the right one.
There's an easy solution to that problem: add a cue. Tell the dog what he's doing when he's doing it. And then only pay when the dog does the behavior you've asked for. So if you ask for a sit and Fido barks, gives you his paw, and then sits, don't reward that chain of behaviors. Only reward the dog when he gives you what you asked for, and nothing else. [I'll write a more detailed post about this in the coming days, as it's an important part of training.]
Here's a quick video of Talos learning about cues. In this clip, I'm asking him to either sit or back (back up a few steps). He has to listen to the word in order to get the behavior right. When he gives me a different behavior than what I've asked for, nothing horrible happens to him. I simply don't click him.
Watch as, at one point, he guesses – and guesses wrong! He tries to come toward me when I've asked for a back, and then tries sitting. I simply ignore him and get him up and moving. Then I simply re-cue the back, and voila – he gets it! Notice the enthusiasm with that last back – he gives me a little pony-like leap. If he continues to give me the leap, I'll stop reinforcing it (because I don't want him backing up and leaping at the same time). But for now, in the acquisition stage of learning about cues, I'll take it!