Treats, the clicker, and fooling the dog
Using food in dog training is a no-brainer to most modern, motivational trainers. Dogs
are highly motivated by food (they’re predators, after all!), it’s easy to use food
lures to manipulate them into various positions, the dog generally enjoys the training
more so is more motivated to work, the human finds it easier so is more inclined to
persevere, and over time the association between food goodies and the owner becomes
very strong so the dog/owner relationship tends to improve. (Dogs like people who give
them treats. People often like other people who give them chocolate!)
But many traditional trainers still object to the use of food in training. Nine times
out of ten, the reason for this is that they have only ever seen food being mis-used in
training. (And of course, a few still cling to the quaint notion that the dog should
do it “for love”. As Jean Donaldson points out in The Culture Clash, these are usually
trainers who are heavily reliant on some or other form of collar correction; in fact,
the dog’s primary motivation for working is avoiding pain. A few dogs will work
exclusively for praise and bonding, but most won’t. You may like your employer and
spend time outside of work with him or her, but that doesn’t mean you’re prepared to
work for free! Nothing strange here.)
Why are there still so many misconceptions around about how to train using food?
First of all, many people confuse the training process with the end result. The other
day, a friend of mine said she would prefer to use the Koehler method for teaching a
reliable drop at a distance “because you might not have your clicker handy and the dog
might be about to run under a bus, so how would you get his attention?” (Of course,
she might not have her choke or prong collar handy, either, particularly if the dog is
off the lead!)
There are two problems with her statement. Firstly, the clicker (or any other
conditioned reinforcer) is not an attention-getting device. It’s an event marker which
tells the dog he has just done something right, and that a treat, a game or some other
primary reinforcer is on its way.
And secondly, the clicker is a training tool like any other. Eventually, for any given
behaviour, the dog has to become independent of the training device, i.e. the drop at a
distance should be obeyed even if there isn’t a clicker (or a prong collar) within a
The training progression for training a drop at a distance might look a bit like this
1) train the dog to lie down, either by luring or capturing downs, and then clicking
and treating every time the dog does. (In the early stages of teaching something new,
we use a continuous schedule of reinforcement, which just means that the dog gets
reinforced every time it does the right thing.)
2) Depending on your requirements, you might use shaping techniques to get the dog
downing faster, straighter, folding back or sitting first. If you used a food lure,
you would also start fading this so the dog doesn’t need to see the food up front.
3) Once you’ve got the dog reliably offering you downs, you add the voice cue, e.g.
“down!” just as the dog starts to lie down. Gradually start saying “down!” earlier and
earlier. You would also stop clicking for downs which the dog offered without the cue.
4) Once the dog is doing the down the way you want it and it is pretty reliably on cue,
you would move to an intermittent schedule of reinforcement. We usually use a variable
ratio schedule, which simply means that you click and treat some of the time, but not
all of the time, and never in a pattern which the dog can work out. (N.B. Whenever you
click, you follow up with a treat – that’s the deal. But you don’t always click.)
Once you’ve reached this stage, the dog now knows that the voice cue “down!” means: if
I lie down, there’s a good chance that I’ll get a treat. He also knows that the
presence of the clicker and treats doesn’t guarantee a treat. What he still needs to
learn is that the absence of the clicker and treats doesn’t necessarily mean he won’t
get a treat, either. Some clicker trainers accomplish this by hiding clickers and
bowls of treats around the house so they can get to them quickly if the dog does
something they like or obeys a cue. My housekeeping isn’t up to this; if I hid some
bits of sausage in my CD drawer, I probably wouldn’t discover them again until they got
up and climbed out by themselves!
So what I do is install a second marker, or conditioned reinforcer, this one verbal.
(I call this phase ‘fooling the dog’.) I use a word/pronounciation combination that
I’m not likely to use in normal conversation, in my case, an eldritch screech of
“treeeeeeet!” I condition this slightly differently from the clicker: I chop up some
treats and hide them in the fridge (safest place!), and then while I have both dogs in
the kitchen I screech “treeeeeeet!” and then race to the fridge, open it and then give
them each a treeeeeeet, I mean a treat. I repeat this several times until they clearly
understand that “treeeeeeet!” means “race to the fridge and get a treat”, and then I
move around the house, into the garden and so on, screech “treeeeeeet!” and race for
the kitchen. Of course I could condition both dogs separately as well, and probably
should have, as when one does something deserving of a treat, the other one expects one
I can now use my verbal reinforcer to mark good behaviours. I normally only use this
with behaviours that are already partially trained and on cue, so in the case of a
down, I would say “down!” at random moments, and if I get a down, I screech
“treeeeeeet!” and race for the kitchen…well, you get the picture.
And then I start putting randomly cued behaviours on an intermittent schedule as well.
The net result is that the dog becomes a gambler. If I say “down!”, he knows that
there is a pretty good chance of him getting a treat if he complies, whether or not
there is a clicker and treats around. And variable ratio schedules have been shown to
be extremely successful in maintaining behaviour; the more care I put into this phase,
the more likely the dog is to persist in downing on cue, even in the absence of a
5) I then generalize, or proof, the dog’s down by asking him to down under
circumstances of increasingly high distraction – in the garden, in the street when it’s
quiet, in the street with another dog at a distance, in the street with another dog a
bit closer, and so on. At this stage, I expect his down to fall apart a bit, and I
simply re-teach under the new conditions. While I’m re-teaching, I go back to the
clicker and a continuous schedule (every correct down gets a click and treat) and as
soon as he’s got it, I go back to intermittent reinforcement at that level of
distraction. This might sound laborious, and it is at first, but the more the dog
learns and the more different conditions you proof under, the faster the proofing goes.
Each successive situation is easier, and so is each new behaviour!
6) I’m now in a position to start adding distance and time (which both simply involve
delaying the click until I have the distance and duration that I want). I might start
by putting the dog on a lead, doing some downs close to me, then taking a step back and
doing some more, then taking another step back and doing a few more, and so on (it
helps a lot if your dog can sit and stand on cue as well!) Again, this is something
new to the dog, so I go back to clicking and treating for every correct response, and
only move to intermittent reinforcement when the dog is reliable at a particular
I increase the distance little by little, and as I start working at fairly long
distances, I might use a tie-out and attach the dog to a fence-post or pillar. For
each correct down I would click and throw the treat within reach.
Eventually I would move to doing this off the lead, and again, I would go back to a
short distance and increase it gradually. I would also need to teach the dog to down
The end result – the fully-trained behaviour – would be a drop at a distance which I
would only need to reinforce the dog for occasionally, and which would be pretty
reliable regardless of whether I had a clicker handy, assuming I had done a good job.
Now I’m not trying to suggest that this is an easy or basic behaviour to train – far
from it. It’s one of the more difficult things in the dog trainer’s repertoire. And
I’m also not suggesting that you allow your dog to wander around without a leash in
potentially dangerous situations! But the point here was that one shouldn’t confuse
the training process with the end result, and this is something that critics of reward
-based training often do. (They also often take shortcuts and end up punishing the dog
for non-compliance at a level he hasn’t been proofed at yet, but that’ll have to wait
for another article!).
So to rephrase what we’ve said so far: when teaching the dog to do something new, you
use treats almost constantly. As the dog’s knowledge and experience increases, the
food rewards become intermittent and ultimately it shouldn’t be necessary for you to
have a clicker and treats on you in order to get compliance.
You should, of course, continue to reinforce occasionally even for well-known
Another area of confusion is the difference between a lure/bribe and a reinforcer. We
often use food lures to teach the dog something new or to get him into a new position,
but we make a point of fading (gradually getting rid of) these as soon as possible.
Why is that?
Well, luring really does make the dog dependent on seeing the food up front. I’ll sit
if you wave a treat over my head so my nose goes in the air, but if you don’t show me
the treat first, then no way!
The problem with a lure is that it comes before the behaviour we want. First the dog
sees the lure and then he does the sit. And for many people who experiment with food
when teaching their dogs tricks, that’s as far as it gets. (And for many conventional
trainers, that’s the extent of their experience in using food to train!)
A reinforcer, on the other hand comes after the behaviour we want. First you sit and
then I give you the treat. In other words, by the time the food is produced, the dog
has already done what you wanted him to. A totally different training bargain – and
the one on which all operant training techniques are based.
A third area of dispute is the idea that if you use food in training, your dog will be
more inclined to beg at table or when you’re eating. Well, if you give him food from
the table or when you’re eating, yes, of course he will, but this is a training
problem, not a failing of food training. If you don’t want the dog to beg, then never,
and I mean never feed him from the table, and never share your snacks with him. In
fact, if he learns that he gets treats when you’re training him, but never when you’re
eating, he will if anything be less likely to beg (dogs are great at making this sort
of distinction). You get what you reinforce. If you don’t want your dog to beg, don’t
reinforce him for begging.
A fourth, and final ‘problem’ with food training is that the treats are bad for your
dog and he might get fat. This is the only one which actually has any validity, but
it’s certainly not insurmountable. Most of us, myself included, just don’t get to
spend enough time training to make this a serious issue, but if your dog has a tendency
to pick up weight, then go for high-taste, low-fat treats (baked liver treats are
good), and cut down his meal rations. There’s a very good case to be made for getting
the dog to work for at least half his food. Vary the treats as much as possible so he
doesn’t get an overdose (or underdose) of any particular ingredient. Cut your treats
up as small as possible and give one (not a handful) when you click. And never give
your dog chocolate, especially expensive chocolate. It contains theobromine, which is
toxic to dogs and can be fatal.
Until next time, all the best, and above all, have fun!
Leave a Reply
You must be logged in to post a comment.