Dog-friendly Anti Bark Training Device

[Aidan3]

Even in a random environment, an FI schedule means that a dog doesn't get rewarded any more if it is doing an unrelated "bad" thing than when it isn't doing it, so it's highly unlikely that the "bad" thing will be reinforced.

If you look at what pigeons do when they are put in a Skinner box and non-contingently reinforced on an FI schedule, you'll see what I mean. Skinner called it "superstitious behaviour".

Going off studies of this paradigm (FI with a response cost contingency) I suspect you'll be more likely to end up with superstitious behaviours than reduction of unwanted behaviour. You could probably test your hypothesis with a Manners Minder, hiding out inside to trigger the machine on a schedule. See if it works or not.

It's great that you've managed to help a lot of people with barking (I'm the sort of person who tends to be pretty good with the dogs... less so with the people). But the people you are helped are those that have come to you (a trainer? Have they even paid for your help? I don't know your background so don't know if you're talking about friends or clients)

They've all come looking. I am a trainer but I've written for various publications (mostly Karen Pryor's websites, KPA, and the APDT), and that information is free. I also had a free, interactive "Help With Your Dog's Barking Problem" course on my website for a long time, for which I received a lot of feedback. So I know that people can learn this sort of thing even by correspondence (as I have said, it's really very simple), but they would all at least know what a clicker is.

[Aidan3]

It depends a lot on the sort of schedule you need to employ. Off the top of my head (without looking up studies) you could probably go FI(5min) which means one piece of kibble every 5 minutes (not initially though!) so too much food would not be an issue. 96 pieces of kibble in an 8 hour work day, not really a big deal. Whether you could get to that point I have my doubts about.

Edited July 1, 2011 by Aidan2

[critters]

On the Quirky page you mention that the visual cue indicates to the dog that the device is inactive - how will the dog pair this visual cue with their barking? ... Granted, you do mention that there will be an audible tone to indicate when the device is inactivated, however if the dog is not looking at the device then how does it pair the tone with the light with the lack of food treats?

This would (hopefully) be achieved in the initial training phases where the treats are coming out almost constantly. The dog is going to be close to the device, and thus will hear and see the cues. It might be that you need to set the dog up to fail (eg, if they speak on command, ask them to speak. Otherwise ring the door bell or show them a cat or anything else that will set them off barking while they are in close proximity to the device) a few times during this initial phase to help them make the association.

Also, if the dog was barking excessively at the other end of the yard then it is quite possible it wouldn't even hear the tone unless you made it quite loud, which would be an added annoyance to your neighbour (and let's face it, the tone also has the potential to stimulate the dog to bark).

Unless you have a few acres, a dog with normal hearing is going to hear it! There's no need for it to be an unpleasant and annoying noise, and certainly no worse than barking, so neighbours aren't going to be too upset even if they can hear it too.

You also state in one of your answers that initially the treats would be dispensed at quite a rapid rate, with only a few seconds between each treat. Now, even for the largest dog, you are going to get through their daily food quota pretty quickly, and for small dogs I suspect you wouldn't even make it to 30 minutes (and that's being very VERY conservative - based on the general size of dog kibble and the daily portion of food for a small dog).

Thanks for bringing this up as it is an important point that I wasn't clear about. The initial training phase sessions would need to be quite short, and the owner should be present to ensure it is working as intended. Hopefully you'd then be able to move quite quickly onto a longer time interval between treats to stretch them out across the whole day.

Also, you say that it is based on negative punishment - how do you know that negative punishment is less aversive than positive punishment? Are there studies on this? Punishment is punishment - that is, it is an undesirable consequence that weakens a behaviour. How can you say which is preferable to the dog? Have you done preference tests?

That's an interesting idea, and I don't know of any studies that have looked into it (I'm inclined to go look), but it is VERY intuitive that -ve punishment is preferable to +ve. What would you prefer, that I give you $1 every hour until you say something that disagrees with me, or that I give you nothing and when you disagree with me I yell at you/zap you/spray you with something foul?

Another problem I foresee with this is that the treats are dispensed regardless of any other behaviour other than barking. You asked "why do you not want a dog to get treats all day long if it is being good?" (sorry, don't know how to quote as I very rarely write on DOL - I'm more of a reader than a contributor). What if I have a quiet dog that spends all day digging? Is it not possible (and in fact I would suggest likely) that your device would inadvertently reinforce other 'problem' behaviours in my dog? (Hypothetical dog, btw, so don't give me solutions on solving digging, its just an example).

This is the same concern as the other posters, and so is obviously an important one. In your hypothetical, though, I would ask why on earth you purchased an anti-bark device for a quiet dog with a completely different behavioural problem :p

Since the dog gets the same level of reward when it is digging as when it is not digging, it isn't actually that likely to be reinforced. You could equally well argue that a warm, sunny, pleasant day or any other basic comfort reinforces the dog's digging, if they happen to coincide.

[aussielover]

It depends a lot on the sort of schedule you need to employ. Off the top of my head (without looking up studies) you could probably go FI(5min) which means one piece of kibble every 5 minutes (not initially though!) so too much food would not be an issue. 96 pieces of kibble in an 8 hour work day, not really a big deal. Whether you could get to that point or not I have my doubts about.

I'm kinda thinking that even my food obsessed lab would not find one piece of kibble every 5 mins an overly exciting reward!

I do not understand how the dog is meant to know that not barking for 5 mins is the desired behavoiur to produce the "treat". There are a lot of other behaviours that could happen in the 5 minutes that could be inadvertently reinforced instead.

Am I right in thinking that at least initially, the machine would need to reward for a relatively short period of silence.

[critters]

Even in a random environment, an FI schedule means that a dog doesn't get rewarded any more if it is doing an unrelated "bad" thing than when it isn't doing it, so it's highly unlikely that the "bad" thing will be reinforced.

If you look at what pigeons do when they are put in a Skinner box and non-contingently reinforced on an FI schedule, you'll see what I mean. Skinner called it "superstitious behaviour".

Going off studies of this paradigm (FI with a response cost contingency) I suspect you'll be more likely to end up with superstitious behaviours than reduction of unwanted behaviour. You could probably test your hypothesis with a Manners Minder, hiding out inside to trigger the machine on a schedule. See if it works or not.

OK... can we come up with a solution to this? What if the reward was contingent FI - the dog triggers something to get the reward, but it doesn't work if the dog has recently barked? Admittedly this is probably more likely to result in obsessive behaviours, though again most dogs should decrease their response on an FI schedule until a reward is due. What if it was a fully interactive device like a treat ball, that could stop working when the dog barks? Tricky when mixing dogs and electronics... what if the device is mounted out of reach and has a home-alone-style-toy hanging from it, and as the dog pulls on the toy treats are dispensed from up top (unless, of course, the dog has recently barked)?

[aussielover]

That's an interesting idea, and I don't know of any studies that have looked into it (I'm inclined to go look), but it is VERY intuitive that -ve punishment is preferable to +ve. What would you prefer, that I give you $1 every hour until you say something that disagrees with me, or that I give you nothing and when you disagree with me I yell at you/zap you/spray you with something foul?

I would prefer whatever method worked fastest to help me understand what it was that you wanted....

Personally, if I make a mistake or piss soemone off I would prefer them to tell me directly.

[Aidan3]

It depends a lot on the sort of schedule you need to employ. Off the top of my head (without looking up studies) you could probably go FI(5min) which means one piece of kibble every 5 minutes (not initially though!) so too much food would not be an issue. 96 pieces of kibble in an 8 hour work day, not really a big deal. Whether you could get to that point or not I have my doubts about.

I'm kinda thinking that even my food obsessed lab would not find one piece of kibble every 5 mins an overly exciting reward!

I do not understand how the dog is meant to know that not barking for 5 mins is the desired behavoiur to produce the "treat". There are a lot of other behaviours that could happen in the 5 minutes that could be inadvertently reinforced instead.

I think what critters is proposing is quite different to this. Imagine you get $2 every 5 minutes so long as you are just going about your business quietly, but if you speak, a tone sounds and at the end of the period your $2 fails to materialise. That would be a response cost, which in dog training we normally refer to as negative punishment. So the consequence for barking is to lose the reward.

If you really want to geek out, look up the inverse of this, which is the Sidman Avoidance Procedure. Learning can happen in some crazy ways.

[persephone]

Critter- I haven't voted, but would like to wish you well... and would also like to pat you on the back for thinking up some sort of different idea

the fact that the idea has generated some interesting discussion is a bonus! :D

[Dogs rock]

Ok:

Even in your initial training phase with the dog near the device, how can you guarantee that the dog will be looking at it when they bark to associate the visual cue with the auditory one, and then the consequence? I would also think that sun position would have an effect on the intensity of the visual cue (as an example, have you ever come across a traffic light where the sun is hitting it at just the wrong angle and you can't tell which light is on?)

For a dog with normal hearing, yes, they COULD hear it, but a dog that is barking (particularly one in 'full-on bark mode') would be just as likely not to hear it, or not take any notice of it as it is irrelevant at the time.

With regards to the interval of treat delivery, you can't 'hope' that the dog will be able to cope with a (lets be honest) dramatic increase in interval between treats in a short space of time, and working to get to this point is labour intensive for the owner (I know, they should care, but lets be honest, owner compliance is a big problem in fixing behaviour 'problems' in dogs).

Is negative punishment less aversive than positive punishment? Well, wouldn't that depend on my emotional state and the value that I place on the behaviours and consequences involved? It is very anthropomorphic to think it is intuitive to assume what a dog would prefer based on our own feelings (which is a big problem in the dog training industry, for example the belief that 'shock' collars are 'nasty' because people think of electrocution not mild stimulation). I know that it is not the case in your device, but 'time out' is an example of negative punishment and I believe that there are dogs that would prefer a quick 'no' than being isolated from their social group.

I realise that your 'why on earth would you buy an anti-bark device for a dog with different behaviour problems' comment is tongue-in-cheek, but if my hypothetical dog was being problematic due to boredom, then it is quite possible that he both barks AND digs holes (but I'm not going to get in trouble with the council for the digging so I address the barking). But you are correct, it may not associate the treat with the digging, but the provision of the treat in this form will do nothing to help with the digging, whereas providing treats in a treatball/Kong etc will be far more likely to be effective in treating both the digging AND the barking. However, if my dog either barks or digs (e.g. barks for a bit, then digs for a bit) then it is possible that the positive 'you're not barking tone' will occur when my dog is digging, thus reinforcing the behaviour.

[Aidan3]

Even in a random environment, an FI schedule means that a dog doesn't get rewarded any more if it is doing an unrelated "bad" thing than when it isn't doing it, so it's highly unlikely that the "bad" thing will be reinforced.

If you look at what pigeons do when they are put in a Skinner box and non-contingently reinforced on an FI schedule, you'll see what I mean. Skinner called it "superstitious behaviour".

Going off studies of this paradigm (FI with a response cost contingency) I suspect you'll be more likely to end up with superstitious behaviours than reduction of unwanted behaviour. You could probably test your hypothesis with a Manners Minder, hiding out inside to trigger the machine on a schedule. See if it works or not.

OK... can we come up with a solution to this?

http://www.sciencedirect.com/science/article/pii/S0376635705000288

[critters]

Even in a random environment, an FI schedule means that a dog doesn't get rewarded any more if it is doing an unrelated "bad" thing than when it isn't doing it, so it's highly unlikely that the "bad" thing will be reinforced.

If you look at what pigeons do when they are put in a Skinner box and non-contingently reinforced on an FI schedule, you'll see what I mean. Skinner called it "superstitious behaviour".

Going off studies of this paradigm (FI with a response cost contingency) I suspect you'll be more likely to end up with superstitious behaviours than reduction of unwanted behaviour. You could probably test your hypothesis with a Manners Minder, hiding out inside to trigger the machine on a schedule. See if it works or not.

OK... can we come up with a solution to this?

http://www.sciencedirect.com/science/article/pii/S0376635705000288

Interesting. What exactly are you suggesting with that article?

I'm picturing an array of lights, which are pre-conditioned to be token reinforcers, and each time a dog barks one goes out. When owner comes home, they give the dog as many treats as there are lights still on (thus reducing the possibility of rewarding the wrong behaviours, and the total amount of food needed to last the day). Perhaps the lights gradually come back on after periods of silence, so that the dog doesn't use them all up and then go bark-mad.

Obvious issue is maintaining the effectiveness of the token system. Are you suggesting the article because you have used or know of work with dogs using token reinforcers?

[Aidan3]

I'm picturing an array of lights, which are pre-conditioned to be token reinforcers, and each time a dog barks one goes out.

That's the sort of thing, except you wouldn't make them wait all day, it would need to be more salient. Maybe an hour? I can only guess really.

I like dogs to get their crazies out when I'm home and to sleep during the day. So maybe you have lights going on on an FI schedule for going into a kennel, and lights going out if they bark. That way there is no real penalty for coming and going from the kennel as needed, but there is a reward for seeking it out and spending time in there. It would take a bit of thinking about to come up with something suitable.

You could still use environmental enrichment, treat balls etc

Are you suggesting the article because you have used or know of work with dogs using token reinforcers?

Tokens are fairly well studied, not so often used with dogs. Nina Bondarenko, who trains service dogs in the UK now, was having a problem with pups getting too much food when they were being trained intensively so she went to a bridging system very similar to the Syn Alia "Intermediate Bridge".

Removing tokens as a penalty is something you more usually see with humans (monetary fines, for e.g), I linked to the pigeon study because it was the first animal model I found using Google Scholar.

[critters]

It sounds like a plan with a lot of promise, but there's a lot of work that needs to be done to work out details. The big one would be finding the most effective way to condition and maintain tokens - and preferably methods simple enough that the average dog owner is able to use them correctly. There's also the issue of convincing people that it works- I imagine a lot of people thinking it a ridiculous idea that personifies dogs to the extreme.

[mumtoshelley]

I too don't see how this is going to work. I see problems already. It may stop the barking but in a way a dog is smart and will find way eg it will stop barking to get the treats then once all the treats are gone it will go back to barking. So in a way the collar will always have to have food in it to work. Like some dogs will not do a command unless there is food involved. I do not train using food at all, I would if i had a strong headed dog that needed encourgement. I also see this collar reforcing bad behaviours, Like the dog is quiet but is digging a hole then a treat comes out bam its just been rewarded for digging, same goes with clothes pulled of the line,chewing up stuff.

[Aidan3]

I also see this collar reforcing bad behaviours, Like the dog is quiet but is digging a hole then a treat comes out bam its just been rewarded for digging, same goes with clothes pulled of the line,chewing up stuff.

These things are already self-reinforcing, so there would be little to no risk of making them any worse to be honest.

[poodlefan]

It sounds like a plan with a lot of promise, but there's a lot of work that needs to be done to work out details. The big one would be finding the most effective way to condition and maintain tokens - and preferably methods simple enough that the average dog owner is able to use them correctly. There's also the issue of convincing people that it works- I imagine a lot of people thinking it a ridiculous idea that personifies dogs to the extreme.

I don't think its a riduculous idea. I simply doubt its efficacy.

Would I buy one? No.

I fitted jet spray collars (not citronella) to my barking boys for a few months after a barking complaint. They haven't worn them in a very long time but I've not had a complaint since the original one.

One collar is now with someone with a Pharoah Hound that's been barking up a storm - and again its working. They wouldn't work on some 'tougher' breeds but they sure did the job on these dogs.

The idea of having to have this device fuelled up to dispense treats every day for the rest of the dog's life doesn't seem to be a better proposition to me.

The aversive tackled the problem. This device won't. You can already buy a device that can be timed to cough up rewards. Its called a manners minder.

I do wonder how obsessive some dogs could become about it.

Edited July 3, 2011 by poodlefan

[critters]

Sorry, I wasn't that clear. I think that a token system that Aidan and I briefly discussed is what would come across as "ridiculous" to some people, even though it could be a better training system than my original idea.

The original device is specifically designed to be phased out; as barking improves, the time between treats and the time after a bark til the treats resume increases, until the dog is just getting a few treats a day and none at all if it barks, and then no treats at all. Leave the device on the wall for a couple of weeks after it stops spitting out treats, and then take it away - really much the same technique as phasing out food in any training method.

I believe this device would tackle the problem just as well if not better than a spray collar. How does a spray collar fix the problem? My understanding of them is that they simply distract the dog from barking, which eventually leads to a habit change but does not tackle any underlying reason for barking. My device does similar, but can also act to relieve anxiety which may (or may not) be causing the barking.

I am quite curious about spray collars, though. All the spray collars I have seen for sale advertise that the spray is unpleasant for the dog and thus punishes barking - but I have heard Ian Dunbar say this is a misunderstanding and that the dogs actually like the spray (and that the collars work even better if it is a nice smell like liver or cheese) and, because they need to stop barking to sniff, are rewarded for stopping barking. If he's right (and I'm inclined to believe him since he has done a lot of research in canine olfaction!) then I would be quite keen on promoting cheesey livery spray collars. But if that's correct, why are there only citronella and lemon type collars on the market and why do they claim to be aversive?

What scent was the collar you had success with?

[poodlefan]

What scent was the collar you had success with?

It had no scent. The cold spray was delivered onto the dog's neck, not near its nose. It was an interrupter that was mildly aversive.

They added lemon a while later so that people could tell whether or not the collar had been triggered.

If this new device scales down to only delivering a few treats a day, how much incentive is that for a dog not to bark? Is the expectation that the dog will hang around the device all day waiting for it to deliver a treat?

[critters]

If this new device scales down to only delivering a few treats a day, how much incentive is that for a dog not to bark? Is the expectation that the dog will hang around the device all day waiting for it to deliver a treat?

Just as with any training method that uses food rewards (correctly), you gradually ask the dog to perform better and better for the same amount of reward, and it works. Once the desired behaviour becomes habit, and the dog enjoys the desired behaviour for its own sake, then it is quite happy to perform really well for only a tiny reward. You wouldn't jump straight to that level of behaviour, though - you start with treats every second, then every 5 seconds, 10 seconds... ...10 minutes... ...2h, 3h, 4h, 6h, 8h, none. I can't guess at how long this will take most dogs and how big the time steps can be without testing it out, but that's the general idea.

In theory the dog will be used to the treats coming out at whatever time interval it is up to, and thus go off to do other things while it waits (but knowing that if it barks, then the next treat won't come).

[sas]

I wouldn't want my dog recieving treats at random intervills because what happens if my dog is doing something I don't want it doing at the moment it gets reinforced for?