Totally agree that rewards should be used for training but it doesn't have to be food. It's just that food is usually perceived as a reward by probably 99% of the canine population .

You can use other rewards but you need to ensure that your dog really perceives whatever you respond with as a reward. If you praise your dog with a "good boy" or pat on the head ~ are you really sure that your dog understands & appreciates verbal praise &/or petting? If you are, then fine, but if not then you need to teach your dog what praise means (by pairing "good boy" or "well done" or whatever you use, with a food treat).

With regard to the other points you made, what you are really talking about is reinforcement schedules in instrumental learning (or operant conditioning if you prefer that term). There are differing schedules of reinforcement (e.g. continuous, fixed ratio, variable ratio, fixed or variable interval, response rate, response type etc.). When teaching a new behaviour you will need to use a continuous reinforcement schedule, i.e. every correct response is followed by a reward. The dog will respond at a steady rate with brief unpredictable pauses but the response will gradually slow as the dog becomes full up with the food reward ~ so train in short sessions only.

Once the dog has learned the required behaviour then you can move on to a variable schedule, e.g. the dog is only rewarded after say 5 responses, or 2 or 8 etc. This will means that the behaviour is not forgotten, and the dog will keep reponding, as sometimes it gets a reward but sometimes it doesn't, so it's worth sitting as there may be a treat in it!. Partial reinforcement of the response tends to make the response even stronger.

You should also vary the value of the reward you give ~ sometimes you can give a low value treat ( a dry biscuit or a "good boy") or sometimes a whole handful of roast chicken (a jackpot). This will keep the dog responding as it will never know what reward it will get (much like fruit-machine addicted gamblers in Las Vegas forever chasing their jackpot).

You can also start only rewarding the quality of the response, e.g. only reward an immediate "down", not one that takes several seconds before the dog responds.

But the key to positive reinforcement is to ensure that whatever you offer to reinforce the behaviour you want, is truly perceived as a reinforcer by the dog. If you are offering a dry biscuit in order to ask your dog to come back to you, when it is having a wonderful time rough & tumbling with it's best friend in the park, then what you have to offer compared to what the dog is currently doing, is clearly no reward!!

I always reward - even if it`s just a `Good Girl!`. Why not? How else is the dog to know it`s done what you wanted.
Sometimes it is a sequence of moves of course - like a big fuss and play after an agility round. But I think a good action deserves to be marked with a reward.

I asked my trainer this question recently and she told me to still treat but not every time (4 in 10 she suggests) and to vary the treats (high/low value) so the dog is never sure when and what he'll get so he's still willing to work for it. I was told that the behaviour will phase out if no longer treated.

LOTS!

Keisha just started doing agility training last week, I packed a pack lunch box full of sardine cake, cheese, ham and sausages (she's IS quite a big dog) and ended up using the lot!!

I was encouraged to constantly treat throughout, it was the first time she had seen all the agility eqpt, so at the end of the weave poles treats were left, at the other side of jumps treats were left, treats in the tunnel etc etc!...when she came to me, treats were given... She was a superstar and deserved them all. She responds very well to high reward treats and will do anything for them!

I'm not going to give her any tea next time!