“When can I stop using treats?” I’ve seen this question asked and answered in a couple different places recently, and thought I might add another layer to the discussion. The glib answer is “never,” and I often also see “would you keep working if you didn’t get paid?” Some more nuanced answers convey the idea that you will need to keep using treats when you are asking the dog to do a behavior that isn’t “natural” or that the dog does not “prefer.”
I want to suggest that an even more useful approach might be to think about where the reinforcers are coming from. (When I say reinforcers, I mean consequences that increase or maintain the frequency of a behavior, so if what you think should be a reinforcer isn't doing the job, it's not a reinforcer.) We can only infer what dogs “prefer” from what they do, and they do what they do (largely) because of the outcomes it produces in their environment. We do know that behavior that is not reinforced is likely to weaken; that is the adaptive nature of behavior.
Very often, when we are teaching a new skill, we use treats because we are breaking a bigger, more complex behavior down into small steps that don't produce any other clear benefit to the dog. As the skill is built, however, the dog can reach a point where they are able to do enough behavior to directly preoduce reinforcement.
So, my answer is: You can stop using “treats,” if that’s what you used to get the behavior started, when the behavior is big and strong enough to carry the dog to reinforcers that are more likely to occur as a typical product of the behavior.
Here's an example: with Rody and Zuzu, who jumped, barked, and spun while their meals were being prepared, I broke down staying on the floor into small steps, reinforcing with bits of their dinner or other treats, until they could stay put long enough for me to prepare dinner and serve it to them. (Note: the videos are labeled first, second, third session etc. but there might've been a couple others. I basically just trained with almost every meal for a few days.)
That's kind of an easy one, because the final reinforcer was still food. But what about behaviors that ultimately might not be done for food if you weren't "training"? For instance, many dogs will initially try to escape or investigate any strange thing you touch them with, and in order to be able to even start to bring a brush, a stethoscope, or a needle toward them, you may need treats to reinforce small amounts of holding still.
But when you teach a dog to hold still long enough for a couple of brush strokes, they may learn that the feel of the brush going down their back is kind of nice. That is the naturally occurring reinforcer for holding still for brushing. We might say that dog “likes” brushing, but it is more helpful to recognize when holding still for brushing is reinforced by the feel of the brush going through the hair or down the back.
This is less likely to happen with a needle poke, so you probably will still want to use treats for the same behavior of holding still when the outcome is a sharp jab in the butt.
Another example: You teach a dog to walk on a loose leash using treats. When the dog can walk on a loose leash long enough to get from a bush to a garbage can to a fire hydrant, access to the things he wants to sniff can likely take over as reinforcers.
TLDR: When the behavior will not bring the dog into contact with other reinforcers, keep using treats.