Owners spend a great deal of attention on what to reward with and very little on when. The second question does far more to determine what a dog learns.
The window is shorter than it seems
An animal linking an action to an outcome does so most reliably when the two are close together. As the gap widens, the link weakens quickly.
Within a second or two the connection is clear. By several seconds a dog has usually performed three or four other actions, any of which may be picked up instead.
The result is that a delayed reward does not teach nothing. It teaches something, just not the intended thing.
What actually gets rewarded
A common sequence runs: the dog sits, the owner reaches for a pocket, the dog stands, and the treat arrives to a standing dog.
What was rewarded is standing up. Repeated often enough this produces a dog that sits briefly and rises immediately, which owners then describe as an unreliable sit.
The same mechanism explains dogs that come most of the way and stop short. The reward arrived as the dog slowed, so slowing is what strengthened.
Why markers exist
The gap between the action and delivery cannot always be eliminated, so trainers bridge it with a signal delivered at the exact instant.
A short consistent sound — a word or a clicker — marks the moment. The reward can then arrive a few seconds later without ambiguity about what earned it.
The marker only works if it has been paired enough times to predict the reward reliably, and it stops working if it is sometimes used without one following.
Where value does matter
Reward value is not irrelevant; it determines whether a dog will work in a given situation rather than what it learns from the work.
A dry biscuit is sufficient in a quiet kitchen. The same biscuit competes poorly with a park, where something considerably more interesting is needed to hold attention.
Matching value to difficulty, rather than using the same reward everywhere, is what keeps a dog engaged as conditions get harder.
Rate of reinforcement in the early stages
New behaviours are built by many rewards in a short period rather than a few large ones spread out.
Small pieces delivered frequently keep the rate high without overfeeding, and a session of a few minutes with many repetitions outperforms a long session with few.
Anything used in training should be counted against the day's food. Treats displacing a meaningful share of a dog's intake is a common and avoidable route to weight gain.