응용 반려견 후각 처리 과정: 학습 이론을 넘어 트레이너가 알아야 할 사항.

uh while i'm actually thanking people i want to talk thank all of my my students volunteers and mostly the dogs in my lab because we are the dogs in the lab and the ones we use in the field as well we consider that they are research assistants and which is a nice way to say that they they help us figure out stuff and they they are more than just a tool i hate that when people say uh yeah the tool you're using that's not a tool it's it's a dog it's a dog that you know we have a relationship and a bond with and and that's what we do it with dogs and not rats or other animals but that topic will come back later as you will see okay so um by the way i made a comment yesterday that i will clarify here the thing is that you know you can't improvise knowledge in something that may sound as simple as training sniffer dogs and i will clarify what i meant by this yesterday like i said the problem with working with scent is that you're dealing with something that is intangible that is invisible and a lot of people that get into this work assume that that's not really a problem but remember we are very much of visual species and this is why we work with dogs because we think the human dog team especially when we do wildlife tracking and trailing is the best possible because you have the visual animal that's us uh you know higher from the ground looking for stuff like snakes for instance as you will see later and then you have like the the quadruped next to you with a fantastic nose closer to the ground it's the best team ever that you can imagine and that's really why that that works so um dogs are there to help us with these kinds of things try to understand this chemical world around us these molecules that we sometimes detect and sometimes don't but the thing though is that if you really want to do this kind of work you're fooling yourself if you think that you can't that you can avoid getting into some kind of obscure areas actually of science and it's all over the place and i'm still struggling with some of them by the way uh fluid dynamics is one of them by fluid by the way in physics chemistry we really talk about fluid meaning air but also water it's not just uh you know liquids air is a is a fluid psychophysics obviously we'll talk about some of that today uh microclimatology and macro meteorology especially at ground level all right because sometimes we don't realize for instance that it may be like i don't know like cool at 12 degrees oh sorry i will use celsius here i have no idea what's in fahrenheit uh so 12 degrees celsius which is kind of a fall day in canada and but at the ground level because of the sun in the moss

it's uh it's at 25 27 degrees which is what around 80 maybe i don't know hot warm anyway much warmer than you would actually expect so basically the temperature you experiencing at this level is quite different from the one at ground level but that influences scent in a major way okay among many other things and obviously some knowledge in analytical chemistry actually another person that's helping us at dallas university is dr wenzel he's an analytical chemist and helps us understand all the stuff that katherine and i my student forgot many many years ago or forgot recently from many years ago okay look as you know i've worked with kanids and uh we started actually descent processing work in 2006 roughly so it is relatively recent and what we started with actually were projects with what we call forensic canines and there's many different kind of applications here and i will get back to that in a minute and more recently in fact probably our most successful program has been the conservation and wildlife canines looking for species at risk mostly reptiles as you will see but also the work helping us with the work with coyotes in capricorn since uh we have a conservative right-wing government in canada now and they're shutting down everything in terms of funding that has to do with research and conservation in the environment so if you do research in conservation environment like i used to uh you're you're out of luck uh we we don't have actually any funding left for this anymore unfortunately we've turned to uh invasive forest uh insects like the emerald ash borer uh the brown spruce long-run beetle which is a real problem in nova scotia uh gypsy moth etc so detection of basically invasive species that really damage north american forests and we we have funding there and it obviously does a maybe economic impact but certainly an environmental one as well and as you will see uh in more detail to uh later today with catherine catherine reeve my phd student we are doing a little bit of biomedical stuff uh hypoglycemic detection and youth with type 1 diabetes is one of them and she will tell you about some of the challenges of that research the problem with a lot of the sniffer dog list richard it can be cancer dogs you know cancer detection dogs or the police work with with tracking dogs is that there's a lot of debate out there about is this junk science or not and trying to figure out why people say this is not always easy but one of the things that we we know is that first of all there's been some pretty bad science done on this kind of stuff over the years but also this kind of a disconnect between what the practitioners do

and and what the science uh would would like uh uh things how science would like things to be done and also the most amazing thing is most of the time when people come to me and say these guys don't know what they're doing it's actually dog trainers they say this group in that state they're completely wrong they're using wrong methodologies they don't have control groups etc etc so it's not necessarily the scientists that are saying there's something not going on here by the way so that's not what i meant yesterday it's actually uh dog trainers and dog handlers criticizing each other and and often based on on good signs but you know again not always so is the science really bad well maybe that's part of it but there's also an issue with how it's done in the field the practitioners okay so clearly we need more empirical work in this we need to talk about things like you know we do talk about this stuff in science quite a bit things like validity and reliability and accuracy and accuracy can be defined at least the way we do it often in psychophysics in terms of sensitivity and specificity and then we get into mathematic mathematical models like signal detection theory all that kind of stuff great stuff but it's not just for anybody it takes my honor students for instance quite a while to understand signal detection theory but when you understand it you realize how great tool it is to really understand what the dog is actually doing and also the nature of the stimulus as well so in a way uh in terms of standardizing things for this kind of work especially the forensic canines europeans are actually ahead of us here especially the dutch group schoon schoon and hack some of you may actually know that book on on this kind of work the german the polls etc are actually doing quite well here and here in north america we talk a lot about it but there's nothing unified yet and no real consensus on how it should be done but people talk about it which is good sign eventually we'll get there so um like i said earlier i think what you need to get into this world of applied work with sniffer dogs is know about animal learning principles and a lot of you are there already most of you in fact and that's great but behind this there's another kind of very important theoretical background that you need to really understand in psychophysics and that gets a little bit more complicated there's a number of mathematical models there and you really have to understand the sensory stimuli that you work with so with olfaction obviously it's it's uh it's nor it's chemical essentially okay it's molecules and it's uh for that reason has its own little

challenge but like i said before you need a few other things as well and i won't repeat this again but you get the idea that you need to know about other things going on now this is just a comment uh in relation to how people often think about the work of sniffer dogs the wrong way uh they either say nah they can do it or they tend to have the exact opposite attitude which is sure they will well it's really not that simple like usual it's not black or white and it's in the shades of gray we've had great success with some of our sniffer dog programs and close to failures with others so they're not good at everything and also we tend to forget the conditions and the context right so you know you would not go do bird watching at night or even actually uh well unless you're looking for owls obviously uh uh in thick fog not a number of reasons why first of all the visual conditions are not so good but i see sometimes people do uh this kind of work with dogs that seem to expect that their dog is actually going to perform at minus 30 celsius which is apparently -22 that i remember doing that um at 10 humidity pretty dry thank you in high wind you know what the olfactory system basically shuts down in those conditions even in walls okay they really have a hard time finding their lunch in those kinds of conditions so it's not because they're well trained that they can actually thrive in those conditions obviously they don't so we have to keep this in mind okay why use dogs and actually this is an interesting question because i will actually challenge the idea that dogs are the perfect tool here oh i said tool sorry the perfect research assistance maybe not uh okay why well first of all they have that thing that we call the nose and olfactory system i'll talk about later that's the primary reason obviously and i i go through this with engineers sometimes and people in the biomedical field because they really engineers for mind detection for instance or people in medical diagnosis they really like their machines and technology and they really are resistant to this idea that maybe a biological system can be better than the technology think about some of the scent detection or cancer detection work with dogs actually where they are clearly better than any diagnosis tool out there if i ever get tested for my prostate i want a dog not any of the other diagnostic tools that currently exist okay so i showed them this and they kind of laugh but really they don't kind of fully get it so they have also a system that allows you to tell them what to do and help them do their work it's quite remarkable if you think about this

and they also have a way to tell you if they feel good or not if they're quote unquote calibrated that day or if you actually need to maybe give them a little bit of a rest and i know this is starting to sound like real science fiction but it's also a mobile unit they move around by themselves you don't have to carry them and uh they can hear commands right uh so they see the ear commands uh it's it's quite remarkable and this is why we can use dogs and there's something really interesting about this is that every 10 or 20 years here in the u.s not so much in canada but sometimes in europe or even israel actually i'm thinking of robert lubao here they will actually re-question this and then they come up with those huge projects let's try to see what's going on with dogs if they're really good at this and this is actually a study from 1985 from the us army where they actually the conclusion of the report after spending a few million dollars on this is this the dog is a mine detection system which is capable of operating in a fast expanse of climatic and topographical environment it is a highly adaptable high sensitivity high specific detection system which is relatively inexpensive reasonably durable readily reproducible and immediately available wow a few million dollars later but we kind of already knew this anyway but they're doing this again and they will probably do it in under 20 years so i don't know why we're so resistant to this idea that they can actually do the job now like i said i will challenge part of this later towards the end of the talk so first of all the first thing i want to do is get a little bit into the scent detection stuff some of the basic concepts here what a lot of people people forget is that there's actually different kind of scent work it depends what you're asking the dog to do okay first of all there's pure detection usually this is a stimulus versus the background okay and background can be background noise we use the word noise and psychophysics as i mentioned yesterday as meaning interference everything else that is not actually the target that you're looking for that's the most basic form i'm not saying it's easy i'm just saying that that's you know stimulus versus background then there's discrimination technically discrimination simply means discriminating between stimulus one versus stimulus two okay and it can get more complicated than this and you can add a number of stimuli to uh discriminate all right then there's identification now this actually comes a little bit from human psychophysics actually you'll say well it's the same thing as discrimination in

some ways it is but with human task usually in identification it's often naming tasks where you show a stimulus and the subject or the participant has to say it's this or it's that it's a cat it's a dog okay so in other words you are really in a way here checking that there's an understanding maybe more cognitive in some ways from the animal or human that oh no i know what this is it is this individual or that scent okay so there's a number of ways to do this but as you will see later matching tasks or matching to sample tasks are often the ones that will be used in that context then slightly different is the idea of scaling that's actually hard to do in a way but scaling has a lot to do with the quantity of the stimulus how much of it is there so i'll give you an example which actually we mentioned in our book chapter uh in the horowitz book is uh we were asked a number of years ago although we've never done that project it's it's been put on the back burner for a number of reasons to help beekeepers uh looking for varroa and nostema which are two of the internal and external parasites that seem to be a factor in the beehive colony collapse okay but the idea is is this is that unfortunately when you talk to beekeepers they will tell you this every beehive at least in nova scotia and most of north america is infected by nossema and varola it's that's not the issue we don't want a dog that will say yeah no i said we want a dog that will be able to say beyond this threshold tell us that there's a problem below that threshold it's still acceptable for us so that's a kind of a discrimination on a gradient if you want that's really actually hard to train okay it takes a while to get there especially if you have low silences stimuli okay then by far my favorite as an ethologist as somebody working in the field looking for coyotes and wood turtles and and ribbon snakes is searching but that's really hard to do because it's very dynamic you're working with the dog in the field it's tracking or trailing and it's really hard to study but we're slowly finding ways to actually do this especially with the current gps technology for instance that will allow us to actually work a lot on this idea of reliability and validity etc but that's fascinating work and this is where you see the dog really in its element trying to really understand that world that we have no idea how it actually works but that's to me the most fascinating part of it unfortunately it's hard to do in lab conditions for obvious reasons so sometimes you have to make compromises and it's a real challenge to study but that's why i'm really very much

interested in it so stay tuned because hopefully we'll get do more and more of that in the future so i just want to talk about the forensic applications here for a minute first of all there's a lot of the work that you can do that is here lab based and a lot of schoon and hack are talking about in her book uh on forensic canines is basically scent matching of subjects so the idea uh suspects sorry that are basically a potential criminals and the idea is that sometimes they will have touched object or piece of clothing and for some reason you don't have fingerprints so then the question is can you actually match this object that was found on the crime scene to uh the potential perpetrate the perpetrator so that's essentially the idea so that's one of the things that we'll discuss and obviously does the field part of this which is a police department's using tracking trailing dogs to actually find people that are escaping justice for instance okay so usually this idea is that again you have an object that is often a weapon that may have belonged to the perpetrator you want the dog to actually be able to match it with that potential individual and usually historically what people do here is a lineup and that's interesting because you need to understand that it's historical why people adopted a lineup in those conditions it's because that's what we do with humans okay we'll put a lineup of six people and then we'll tell the uh the witness do you recognize any of those individuals and that is a big debate in cognitive psychology actually about how to do this if it should be a photographic lineup very sequential or if you you should look at all the individuals and then make your choice problem with this is often individual will say they will pick up the person that looks the most like the perpetrator there's all kinds of problems and it's back and forth in the literature just recently there was a paper actually that's re-questioning a lot of the ideas that we had about how to do this with humans but one of the questions here is is actually why would you do this because in this kind of work the way i described with a dog the dogs are actually not eyewitnesses they're not trying to remember somebody they saw six months ago do something bad so the lineup actually is intrinsically a mnemonic task meaning that it's a memory task but actually what we're asking the dogs to do in the pres the kind of procedure i described to you is not based on memory as much as their ability to discriminate between scent or actually pick up scent that are very similar so that is a very serious problem in cognitive psychology if you think that way

so before i go into more details i'll explain later exactly why the lineup is not the right procedure to use uh i'll i'll just mention something about what people call call in forensic science the law of fry and this is actually used by many countries and it's basically standards or principles used by juries and it's the idea that juries if you want your evidence to be accepted in court will require a few things for instance is it verifiable if if it was tested it was was it actually peer-reviewed and published uh is there a known error rate that's actually very important psychophysics can produce a lot of that information is the procedure standardized and a lot of them are not and is it generally accepted and that's the idea that science likes to have consensus obviously doesn't always work although i have to say so far so good for this conference i think you realize that we are keeping saying about all the same things so that's good so there is consensus in science often with exceptions okay so um you know one of the questions here obviously is are there some individual orders can you really distinguish individuals you know you and you and you from each other and we kind of know now that's actually really that's that's that that works and there's all kinds of reasons why uh wolves raccoons frogs whatever could actually do this theories in biology based on can selection for instance these theory actually work mostly if there's kin recognition and that means that indeed they can discriminate between each other between sex status of the immune system status of the endocrine system meaning hormones all these kinds of things so it makes sense that animals can actually do this that they distinguish us not just by sight but also by scent so as long as you agree with this from the forensic dog perspective now what you're thinking about probably is what is there now is this idea that uh it's almost like for dogs we have an olfactory fingerprint okay so we have our own order signature and that's why forensic canines in principle can do the work very well so sent identification at that level in that context is basically biochemical fingerprinting now the methods and this is where the problems start with people that actually do some of this kind of training is that you need actually some of these good practices that we just hinted at first of all double blinds and controls like control conditions uh very basic and most of you know what i'm talking about won't get into the details here but double mind just to make sure i need to explain what it is and it's this idea that especially in the maybe not super early stages of training

but certainly fairly quickly after you start training the animal the handler and trainer can't possibly actually know what the dog is supposed to flag as a target why well james actually just talked about the clever hands effect that is certainly one of the reasons and by the way you don't have to cognitivize this you just have to realize that they're just good at picking up on other cues they're conditioned to something that is not the target but other cues that they pick up on say oh those two things are associated and there you go they figure out what's going on okay so um there's other issues too that arise when you do this kind of work should you have single scent dogs this is something that people ask me quite often i don't think so some of our dogs are really specialized my dog zila that unfortunately died recently was actually doing almost everything she did coyote she did wood turtle she she did ribbon snake as you'll see later you'll see pictures of her in a video of her working and many other things and she was good at all of them probably one of the most consistent also including across some of those things i mentioned earlier like detection discrimination identification tracking trailing and all that okay but no i agree not all dogs can do all of this and and be good with all kinds of odors uh age well she started at eight years old and she died recently at 15 and she was working still last year and still one of her best dogs so age matters yeah but you know police forces and military that actually retire dogs at four or six years old i think they're missing on on so many years of good work actually there's no reason especially if the dog still wants to work why you should you should actually retire retire them this this early or this young the training methods obviously can matter breed well we talked about yesterday i won't go back there but you know what i think about this and there's this issues also about what kind of responses you can actually train which by the way is a real problem with the forensic canines and a lot of the work actually that's been criticized by dog handlers and trainers here in the u.s in relation to some of their colleagues is that sometimes they say but what is the response what's the alert you don't know and then the antler will say well i know my dog i can read my dog. i know what they're saying but in some cases especially when you go in court you want a dog that that clearly gives a clear response yep that's it that's the one no no no this right here right but in some of the videos that you will see i've done doing tracking trailing in forensic context

they walk around i personally don't see the cues showing that they actually know what they're doing but somehow the handler says they know and that's where you get into these kinds of mysterious clever hunts like effect back and forth between the dog and the handler and the handler and the dog it's it goes both ways in strange ways okay and how you interpret interpret those behavior is very important and actually this is one of those real good cases where i will disagree with a lot of scientists that say that the experimenter or the handler whatever should not be close to its subject if you really want to be good at this you really need to know your dog and your the dog that works with you needs to really know you so forget about all those cute principles that it should be no objective just because no no no in this case you want that good relationship you want to be able to read your dog your dog read you and it that's how it's going to work well uh the worst mistake you can do is give your dog to a new antler and say go and figure it out by yourself because they will make up stuff i guarantee you both the dog and the handler whoops okay uh we also need procedures that will get into what we call inter dog confirmations okay comparing between dogs we often work with two dogs in the field if a dog flags something we'll call the other dog and say come come check this and i'm click yeah yeah that's it or sometimes you disagree which is quite interesting and then we have to make a decision which one we trust right introduct confirmation as well in controls i mean they're not always consistent they have bad days remember you know when i showed the tale earlier it's just to make a comment that you need to read your dog that way too maybe maybe uh fido doesn't want to work today and and you need to acknowledge this there's no you waste your time you make it very aversive to the dog if you force them to work in conditions that or conditions and i mean external conditions or their own condition if they don't want to if they don't feel good if they you know if they skip breakfast that's a bad sign maybe maybe it's not the time to work today so if you do the lineups for instance there's a number of things that you can do a lineups and i'll show you a line up in a minute but i think most of you know the principle is that you would have like for instance stainless steel jars or glass jars with uh foils and a specific target in there and we call it zero runs in my lab but it's this idea that essentially when you give a line up to a dog you should have what we call a blank lineup and this is a com a very common

mistake and what i mean by this is you have to give them the opportunity to be able to go through a lineup get at the end of the line and say sorry but there was nothing there and a lot of trainers strangely don't do this and then you have a dog that's basically convinced that a lineup means there will be something there so they get at the end of the lineup they go like oh what happened here i didn't find anything so they go back and then they try they'll flag something they'll fight anything so you have to give them the opportunity to say no sorry there's nothing here and make sure that they give a clear response of that as well you would be surprised how many times that's not done also you need you need to do what we call the two target lineup it's a lineup where there's not just one target but two why because often the dogs will run the lineup and uh if let's say that the the the stimulus uh the the target is on position three so that the third jar what they'll do is they'll go oh i'm done okay and then they disconnect from the rest of the lineup the problem is there are three more jars there and maybe if they had investigated them they would have realized that the fifth one actually is the target or is closer to the target so by giving sometimes not all the time two potential targets you actually force the dog to go like okay i found this one let's finish that line up to see if there's more okay and uh and that that is absolutely essential and there's another thing that humans do that's very strange is that you have to use a dice to do this to do it really random because the problem is actually dogs will often run to the fifth of six jars and it's called the position five bias why because when humans think that they're being uh random in their assignment of the target to a specific jar they always tend to put it on average not always more often in position five than of the others not too early in the lineup not right at the end five sounds like a good you know spot and there you go and the dog gets in the room they don't even sniff one two three four they go to five right away they check at first like oh okay it's not that one they're smart aren't they amazing okay um now the to go back to the forensic canines there's a number of things also that are real problems like when you have like the scent of perpetrators the question of gender race yeah race is actually interestingly a factor here age diet uh health conditions all these things are actually more noise that's added to this and there's a problem of saliency and actually catherine later this afternoon will actually discuss this which is really interesting is this idea

that when you train dogs to do this to identify specific humans the problem is that you're asking them to do something that you think is obvious but is not why because all of their lives so far they've been bombarded completely over overtaken overflown by human sent and they've basically kind of learned to ignore it and now suddenly out of nowhere you're telling them no no it's relevant but you need to really help them understand this because at first they'll say yeah that smells like somebody and that smells like somebody too and that too and what do you all need to find here and they'll actually start looking for other things that have nothing to do with the order of the person actually so and you'll see a great example of this with what catherine will talk about later so the significance of this stimulus and the whole familiarity effect again that you have to tell them no no now i'm really telling you this is really important it is important that you can distinguish between peter johnny and michelle okay uh i will actually skip this uh this is just more considerations about how damn complicated the whole thing is uh again you can read schoon and hack actually and they they they talk a lot about these things i'll just make a comment about the genetics of things when i said actually earlier that they match people quite well there's actually uh recent data showing that not just uh fraternal twins but also identical twins they can actually differentiate which on from a biological point of view that's actually quite surprising and it was not always that study said that by the way uh it was actually a question for quite a while there's the pink pink spartas study actually 2011. so relatively recent okay now dogs and handlers actually make mistakes so james talked about the cleverness effect i will get that in a minute that's the importance of actually doing double blind uh the positional bias i already talked about uh there's also the question of other cues that can be present especially around the lineup if you use the lineup hand washing although here you have to be a little bit careful because that can also make things worse in some situations there's a very interesting novelty effect the odd man out effect sometimes that they react to but some people agree that maybe it's not such a factor in fact the other one is more the question of familiarity that i just mentioned and there's something really interesting that schoon and hawk discuss which is rare preferences and dislikes sometimes you'll see dogs running a lineup with human scent and they will literally consistently avoid a specific scent of a specific person they

will literally make a large detour around the vial we don't really know why but they just don't like it okay so now i'm going to discuss a little bit of data i'll talk about lineup procedures and why they may not be the best tool matching to sample procedures the saliency training is something that catherine will talk about later and by the way we use other techniques that i won't mention today like habituation this habituation although you have to be very careful with that one we use an adaptation or of error-less discrimination training that's also in the horowitz book so i won't get into those details now so a lineup as i said earlier is basically something that looks like this this is a graphical representation when you see s minus this is just a stimulus you don't want them to flag and the s plus that happens to be in bold there is the target like i said sometimes there's two sometimes there's none just to make sure that they can actually uh figure out that it's not always very predictable so the way we do it actually is about if we have a session with 10 trials one trial will be a blank lineup no target one trial would be two and uh the rest of the trials the other eight uh will have actually won uh one target and by the way we adjust some dogs tend to really have a problem with signaling there's nothing in this lineup in which case we will increase actually the number of blank lineups so they can really learn more quickly that no no this is a relevant response we want you to tell us if you think there's nothing there you have to tell us so this is a by the way very early training okay this is what we never do this in further steps you will see here that the handler is really showing the dog what i want you look at this and then this and then this so don't worry it's not done like that later this is in early training and later they can run the lineup by themselves they don't have to be harnessed leashed or anything like this and they will do it very systematically okay so this is early training this is pete so okay now i need to explain something you just saw what what she did there is a piece of paper the sample actually in this case is in a piece of paper and there's a number of reasons why we tried to do this because i said stainless the stainless the steel glass sorry stainless steel glass gyres we actually are experimenting with this for a number of reasons um i won't get into those details now but this is actually a procedure that has worked out with us quite a while so this is actually this was one of the setups that we use with six uh with six items uh one being um the target okay

so basically like i said earlier the lineup tends to be the the standard for the training us and dogs out there and especially with forensic canines i was telling you that it may not be the best way to go simply because again they're not really eyewitnesses they are expert witnesses but not eyewitnesses so they're not relying on memory in fact they should not rely on memory okay it's not a memory task it is a perceptual task and that's where you need to change the procedure and i'll explain why see this is a little experiment we did a number of years ago with two dogs two of our quote-unquote expert dogs and basically what we did is we looked at their performance rate percentage correct based on the position of the target in the lineup and what's interesting here is what you're seeing is that when the target is in position one or two or actually even three they tend to be above ninety percent detection but when you start having the target at position four it goes down below but it's still actually in one case at least in one of the two dogs here a little bit above 80. okay but then by the time that you when you have targets at position five and six their performance is really crappy now why is this so the way typically that you do it is that you will present the sample to the dogs they sniff this okay now go find okay so now they have the this stimulus that they just processed and there's a memory component here because they go to one then two three by four this thing like wait a minute what was that again they've accumulated a lot of interference here five i don't remember that's six i don't know so they're starting to guess more and more as they go another analogy is this is actually a research assistant of mine uh jennifer corning that came up with this analogy she has a lot of fish at home and sometimes would actually transfer her fish from one tank to another and you have to match when she would clean the the aquaria i guess and she had to match the water temperature so she would go and put her hand in his tank okay is it the same temperature at this one yep this one and already she says you know by the time you add the third tank you don't really remember what exactly what the first one was like so you see that that's kind of that mnemonic thing that you put there you make it a memory task and remember what i said those dogs were not under crime scene right they they're not trying to remember what happened six months ago you just want them to match directly is this what i'm showing you now in there so it doesn't have to be a memory task so and you then you you remove that problem of a degradation of performance as the line

um you know gets longer okay so the alternative way to do this is to make it simpler to meet in a way it's a shorter lineup and it's to do a matching procedure with two or we think uh based on our data about three maximal maximum choice uh so you reduce the working memory load here this is a cognitive psychology term which is this idea that when you work when you do work where you're in action you are actually memorizing things okay i just did that now i have to do this now i'm going to do this now that okay but working memory is fascinating thing to study especially in what we call ergonomics because it's really a problem in some cases where uh you know if you've ever been a bartender for instance you know that your working memory is working like crazy okay you what drink you want this one you have to one okay you go get the drinks you remember the change all that working memory big time okay but you see especially early in training this is very taxing on the cognitive system on the brain it's very demanding so why do this if you can avoid it is essentially what i'm saying and also you tend to create a longer delay between the presentation of the stimulus and you know the the potential target especially if it's far down the the lineup you understand this yes is it clear okay all right so we came up with uh and i think maybe catherine will explain no actually she doesn't sorry we came up with what i call the bucket procedure and there's another reason for this is that often people would say well how do you know that they're sniffing by the way this is often a problem uh the the schoon and huck group will actually tell you that sometimes when they have the sample uh you know in in the jar and they present it to the dog what they actually do is shove the nose of the dog in the jar to force them to take to take a breath we don't want to do this but a lot of the time we present to them and especially if they starting to get bored okay they won't sniff it and the problem is how can they pick up the right thing than if they don't sniff at first so we're working on ways to actually make sure that they actually sniffing the sample and then sniffing all of the of the other alternatives so um this is actually so i'll come back to that in a minute for some reason i thought i had something else here okay all right this is actually one of the the matching tests that we had we were not working with essential oils anymore because we had evidence that some dogs are very sensitive to them we used to dilute them a lot by the way now we use uh things like herbs you know tarragon estragon things like that or tea green tea whatever and different

dilutions of it you will see this with catherine actually and that that's good enough for most of the work that we want to do here they in this task they were actually learning eight different odors so the sample that you present to them can be one of eight and then they have to go and find a match and all of the others the foils are actually at first nothing then the same distractor and then all the five positions of foils will be different orders eventually so you make the task harder and harder and this is still uh but you see with the lineup it becomes actually difficult so the two afc tasks here at alternative force choice this is from a few years ago we don't do this setup anymore and you will know why in a few minutes this is milo we used to call him einstein so sample okay tell me which one is the match okay he kept his nose there for a few seconds that actually got better with time i asked the students to actually get the dog to keep the nose on the sample for at least five seconds so you count one thousand two thousand three thousand four thousand just to make sure that he's committed to that choice essentially but you will notice in a second oh and this dog responded very well to food reinforcement no clicker sorry uh but it worked really well for him and there were other reasons why we did this you will see i think with catherine the clicker being used later okay now so he pokes you know when he's pretty sure but you will notice eventually he will actually make a mistake and then what you'll realize what we're doing and i forget maybe this is the one or i may have missed it already i think that's the wrong one so then we yep sorry no we make him wait just a tiny little bit and he goes like oh damn it i missed that okay all right and then sometimes what we also do is that we'll show uh an absence of match so that's not the one i think it may be the next one oh he got it wrong again oh no okay i think so uh no actually i think it's the next one so my working memory is a little bit um i'm processing too many things at the same time okay so this is actually here what you just saw is he said sorry there was no match there and he was actually right okay in other words the sample was presented didn't have a match in the two alternatives that were that were presented okay so that's it for milo he seems to be doing what he's doing you know quite well it goes on and on all right just remember him oops that's not going forward okay so uh this is actually his data and look at his performance over time there's a few details here that we don't need to go get into but you know he performed uh pretty much

around if not above 90 pretty quickly um there is actually a dip at one point and there was a new handler that day and that throw him off that's very common with dogs so you know we we thought this is great it's fantastic looks looks fantastic this is actually rosalind rosalind took a lot longer to get there but you know and more inconsistencies she she's much better now she was quite a young border collie at the time and she actually got better with time now the thing is this is a little bit more typical of what we would have actually expected and uh oops i'll go back for a second so the problem is this this cute little boy here when that honor student left brittany we had a break and sometime in may we started doing the work again but she was done and she's from toronto so left for toronto when we start running those trials again he went back to chance a lot and consistently and he never really picked that up again so then i kind of start talking to now what you don't see by the way you're thinking well this is not double blind but what you're not seeing is brittany is sitting right there next to brittany behind a screen there's a handler or somebody or sorry an assistant that will give her the jars and when the dog makes the the right decision normally what she should have been doing is say to brittany yeah that was the right one or we have a more subtle system than that that the dog can't pick up on and then brittany can give the the the reinforcer now the thing is after drilling that research assistant for a while really insisting she eventually told me no actually it was not double blind sorry brittany knew which one was the right stimulus in other words we think that milo never really knew what he was doing but picking up on something else and i spent hours looking at those videos trying to figure out what was the cue we had another camera facing i could see brittany's face looking at you know as soon as i thought like it's the movement of the eye then something else would happen okay it's not that rewind the tapes not start again what else you know every part of her face and we're not exactly sure we think it may be actually the way she was pushing back when he was actually giving his answer so now we don't use this procedure anymore for that reason and i insist that the students tell me did you do this double blind at least later in training you know initially it may be faster and easier not to do so but you can't wait too long before you start having problems and the dog is learning the wrong things essentially so that's a great example of a potential clever hands effect so another way to do this especially if you

want to show how well our good dogs at doing this in terms of of their detection capabilities we use matrices like this uh three by three four by four or five by five where again you have to target among many different distractors and at first we train them on an open matrix where they actually just scan like this and when they find the right jar they put their nose on it and stay for five seconds and that's their way of saying it's like pointing that's the right one and then when they they're okay with this what we do is we put a hood on top of it and now it becomes a real detection task where we can use signal section theory if some of you know what it is and what they'll do is that they'll come in the room sniff over the whole of the the funnel basically and then they just have to say if the scent is there or not and obviously the larger the matrix becomes the harder it is because there's more and more interference and what's interesting is it depends on the stimuli we use obviously but matrix uh the the three by three and the four by four they're doing okay with the five by five with most similarly we use it starts to get really really difficult and probably not really representative of what they experience in the field unfortunately now the bucket procedure is an interesting one because here what you'll see is we tell the dog to pick up the sample okay can i stop this no yeah okay let me sorry so i'll explain it i'll mimic it for you so um the now the sample is a bucket to our container and there's like three crumbled pieces of paper you've noticed right that's why we use paper it's just easier for them to pick it up so to verify that they're actually sniffing the sample in that container there's like two crumbled pieces of paper that have nothing on them and one that has the the sample or the the potential match so they have to go there and they have to pick the one that has the scent and that's our way to know that okay they're really sniffing it they know that's the one okay and then they go to the buckets okay so again okay that's okay that's the one that's the smell i'm going to look for now i'm going i'm looking in this bucket no it's not there i'm looking in this bucket and then within that bucket they pick the target so this is our way to verify that perceptually they really understand what's going on okay so the two buckets first of all which one has the odor but then when they they choose the right bucket we we ask a little bit more of them it's like okay you know it's in there good but now tell us exactly which of those pieces of paper has the sent and that's a great way to verify that

they really quote unquote know what they're doing so um usually actually the dogs did quite well oh actually sorry this is a i made a mistake here i did last minute changes on this and clearly the sequence is not right i apologize for this but actually this is the the matrix data uh and actually basically it what it's showing is that the the uh the performance uh with the size of matrix as you see uh uh it a lot more difficulty with the larger matrices i don't actually have right now data on this we're working on it it's in progress with the kind of work specialized work that we do it takes a long time to get data from unfortunately very few dogs but it's it's coming very nicely so we have a high hopes for this uh i will skip this because this is a little bit technical um and i'll just say this and it's not really the conclusion of my talk because i'm getting into something else in a minute we know that dogs actually can do this but we can do it better and let's do it in other words you and us scientists can work together on this to do real good scent work training and uh you know also train good handlers that will do it the right way with all the right control conditions and uh because otherwise if we don't do this it's ruining this ruining this industry and it gives bad rap to you to us and to the dogs and actually there's no reason why we can't fix this so i'll tell you a little bit about the olfactory system now i said i was not going to do the neuroscience hat that much but like i can't help myself so just very quickly i'll just say look uh first of all they have a big nasal cavity there they're they just have the you know they have the hardware for it although i did say yesterday remember that hardware is not everything okay the software part is important as well this is where the dopamine breathes as i call them coming to the fold in the sense that uh they they have the the right neurotransmission going on basically dopamine being an extremely important transmitter in the olfactory bulb in the olfactory system in general so uh are breeds better than others sure now this is us humans by the way we tend to say we're bad at this well you know i mean we have a fairly good system it's not that bad it's it's a little bit smaller it's not as big it's you know but it it works except that you know it's at the software level again in terms of how our oil factory system works that you know dogs are far superior than we are we are after all microsomatic as opposed to macrosmatic now a little note and i think it's uh ray ray coppinger actually touched on us earlier there's actually two olfactory systems in most mammals in fact in most

vertebrates one is what we call the main or primary olfactory cortex and that's the one that tends to process most odors neutral odors but some significant biological orders as well and then we have the accessory or secondary olfactory cortex which is basically the vomeronasal organ that or complex that ray was mentioning that is also called the jacobson's organ and that's the one that's usually associated with processing pheromones in other words smells that usually have their source from hormones okay that are secreted but just make it out of the body through sweat for instance breath etc etc so that system interestingly is like the other one connected also to the um the main processing system including the limbic system what we actually don't really know is how they complement each other and most of the time we assume that dogs work with the first one not so much with the second one with most of the tasks that we do although uh for again for what we do i mean obviously if if a male is looking for a female it's going to be the second one the vno that will be activated uh but actually it's not really clear that they're not kind of working with each other and complementing each other we have a dog for instance once that started giving that fleming response that actually ray was uh mentioning cats will do this bringing the molecules right to the the opening of the pallet that that you can actually touch by the way right you put your your finger inside the the amount of your dog you'll find a little bump right here that's that's the opening that i will describe actually in a minute and but she started doing this uh my dog zila like like this and she was drooling and everything and what was interesting is the only time she ever did that and she got on more like hunting mode it's the only time that we had to actually restrain her away from that snake but we realized when we caught the snake it was actually bleeding it had already been injured so interestingly that activated the vno system so they can work together sometimes and there's good evidence with other models like hamsters for instance that actually they are often uh not as separate from each other as we think some species can process pheromones actually with the primary factory system and humans are a great example of this makeling talk effect we do have a vno by the way you can see with neuroimaging the only thing is it's not connected to the brain or to the olfactory system so it's kind of useless okay i'll skip this one and uh that that duct or opening that i was talking about inside the mouth that ray actually showed from a skeletal perspective it's this right here okay

that's from uh actually that's a rat not uh not uh okay so uh i will skip this and this is your primary olfactory system there's a lot more in the orowitz book i'm really running out of time here sorry this is the first one i'm really misjudged i'm clearly if you want to read more about this you're welcome to look at that book chapter it's a relatively complex system actually but it's it's kind of an interesting one i talked about this yesterday the difference between smelling and sniffing just something to keep in mind i will skip this as well i will skip this sorry okay in in the book one of the thing that one one of the things that we do is that we distinguish actually between the what and aware system and it's an interesting idea that actually come from came from people in neuroscience and psychophysics perception as well arguing that actually in the visual system there is two components one which is really focused on identifying what is this stimulus and the other one which is really based on where is it so there's a grizzly bear oh it's a grizzly bear but now where is it in comparison should i run should i what okay so a very clear adaptation basically of you know uh in the brain to respond to uh potentially threatening stimuli and basically you can see that a lot of the tasks that i described earlier will fit within what catherine and i described as the what and where system and potentially how much as well though that may be related to wear which is the question of scaling so very quickly here i'm going to show you one of our most successful program which ran out of funding because of our government let's be political for a second hopefully they listen and they will do something about this so please mr harper remember that conservation in the environment is important okay uh so we do a lot of reptile research with ribbon snakes wood turtles i'm not interested in reptiles uh for the sake of reptiles myself anyway i won't get into that but anyway um and the dogs have been fantastic here we've used them as a great outreach tool as well and what i mean by this is that interestingly uh sometimes the kids are not really interested in the reptiles per se but they love seeing the dogs doing the work and then they get into the reptiles after so it's great to work in those kind of context and it helps us with a lot of different things with that kind of field work and i will probably have to come back to talk about some of that stuff later this is a ribbon snake this is the kind of environment we work in so the grass gets really really high and you can't really see what they are and they're very well

camouflaged and there's water a lot of people at dnr the department of natural resources actually told us there's no way dogs can actually do this in this kind of environment it's marshy there's a lot of smells that can interfere with the olfactory system and like i said actually again one of our most successful program they did fantastic that's again a little bit more of the environment this was a day here where there was actually a lot of flooding going on so they were literally tracking in water and doing quite well you see a ribbon snake on top there there's a this is my dog zela just behind oh by the way you see a choke collar there don't panic it was not used as a choke collar that's just okay all right in fact i don't even know why she had that around us but anyway so uh there's a ribbon snake there can you see it oh you're good uh well anyway she doesn't see it yet but she's tracking it so that's you know what i was saying earlier with great research team uh we combined two it's right there i don't know if you see the bands so you see that camouflage is remarkable especially when thank you i'm out of time when the you know the grass is really long it's really really hard to see this is actually more of the environment two of my former students uh this is actually zela actually looking for one of those snakes that went and hide under the those stumps and i believe hopefully this is going to work okay so this is her actually looking for it and you'll see her she's struggling a little bit we are there we're not seeing anything but clearly she's onto something and she's starting to do these kind of concentric circles here and look at the tail is that because i know my dog i can tell you that i knew i knew she was on it plus the sniffing rate was really strong you could hear it and the tail like that means yeah now it's here it's i know i know let it give me a second i'll find it i'll find it i'll find it i'll find a guarantee and she's doing this for a while and then unfortunately whoever was filming this i forget stopped the camera just as somebody said snake and then so you don't actually see the snake but trust me she actually found it it took a little while it was a tiny little thing about this big and about thin like a pencil it was really hard to see but she she totally found it okay i'll just skip this you get the idea anyway they can do this in those kinds of conditions oh by the way the harness on the leash this is because parks canada which is our equivalent of parks you call it here park service yeah uh required the dogs to be leashed so they don't uh start apparently uh you know

chasing wildlife and slaughtering them this is my english shepherd uh dog flynn having a stair context a contest with a wood turtle that he just found very proud of himself it was his first actually that was it's very important picture for us and they are very good at finding also the tiny little turtles they're about the size of a toonie oh god you don't know what this is a tunis are a two dollar two dollar piece it's metal right so it's about this big uh but you can see the size anyway and when you you look for this in the woods among leaves and rocks everything's impossible to see but the dogs are fantastic at finding those little babies and that's zila actually sniffing it they love sniffing they can go there for five minutes ten minutes we don't know what is it that it turns them on that much about the baby wood turtles but they just love them and that was a great day with zela and flynn and the team and you can see most of the students there actually have turtles in their hands uh we that day actually we couldn't keep up the dogs we're finding them like this and we were processing them at the same time taking measurements and everything and we had turtles all around us just fantastic so um can i take a one more minute two minutes but or you no okay so i apologize if i go over time okay now this is the thing um are dogs the best tool for scent work now you're going to start hating me not necessarily and what made me realize this is working actually with some agencies that really got excited about some of these projects and then when they ask us how much this is going to cost you give them you know a price and we don't charge much because i don't have to pay my students you know it's that's the way it works unfortunately um so i don't pay a trainer and but you know um they don't even like the amount i give them which is just base for equipment commuting costs and these kinds of very basic things but the thing is sniffer dogs cost anything as you know between five and fifty thousand dollars depending on what you're asking them to do and uh when you actually tell people these kinds of things they say oh no this is way too expensive and um for cancer detection the same thing is you know if they're so good at it why is it that they've not been implemented somewhere in canada in the u.s or europe and they haven't yet tons of studies showing how good they are at doing this so we have to convince what the or who the fda health canada what are the problems here no but again it's it's the time it takes three six nine months sometimes more to train some of the dogs to do some of that specialized work

um and the market value like i said is i said fifty fifty five thousand dollars actually that's us that's not a lot of money but really i mean in the real dog market went by done by trainers it's anything between fifteen and fifty thousand um and there's another problem with dogs is all the stuff we've been talking about you know the last few days they have intelligence affect sentience personality and that sometimes gets in the way so the thing is this is that if you are looking for a mobile unit as i said earlier if you're doing tracking and trailing fantastic you'll never use anything better than a dog and again like many people will tell you this uh and uh you know if you have read robert lubos lubow's book war animals which is actually kind of really hard to read but he makes great points of the use of animals including tracking dogs actually has been great very efficient kind of systems and they've tried many other species in israel and the us here and they always come back to the dogs anyway so one of the issue again is the deployment issues um i have colleagues that are interested in doing work with the emerald ash borer and recently they realized that people from the usda and some state departments are actually saying okay good but when those companies are saying okay fine but one dog is 35 000 they go like oh but we need like four of those just for the state of minnesota for instance and then they start saying that maybe this is too much so one solution is to do remote sensing where you sample things in the field including those trees infected with potentially infected with some of those parasites or those insects and then bring those samples to the lab and ask the dogs then to do the the detection or discrimination identification in the lab and if it's well documented then you can say you know sample number that tree yeah infected gokart and destroy it kind of thing and that would be less costly so this is one of the things that we're experimenting with now is that remote sensing which actually came directly from people working with demining dogs in europe and uh but again the question is well if they're so costly if it takes so long why use dogs and i will finish sorry uh very quickly by saying that actually there's a number of groups out there that have used other systems you may have heard of the bees doing explosive detections or using doing cancer detection you think what bees well here the thing is that you don't actually work with operant instrumental conditioning you actually work with classical conditioning and there's a number of projects including here in you in the us that

used wasps for instance fruit flies more recently actually were used so patients with lung cancer or potentially having lung cancer can blow for instance in one of those kind of weird vials there that they use weird system and if there's actually volatiles associated with cancer and they were trained on these odors before then the insects will aggregate in a very specific part of that system and that the way to know that something is going on and there are other invertebrate models that actually we are looking into because and i will skip this unfortunately because of lack of time it's showing you should look at this on on youtube it's the the in sentinel group in uh in in the uk um the study at the bottom there by reigns that have trained bees to do explosive detection it's really impressive to see how quickly they learn just by associating the food with the explosive basically and they have what we call the proboscis response which is how they extend the proboscis what they eat with when they actually smell the explosive because they've associated with the food before and insects there's problems with insects there's a question of handling especially bees or wasps it's weird that we go with those models husbandry issues because now you need a hive system and everything and the problem too is longevity most insects live at adult stage what about three months and that's it so you constantly have to train those animals now again the thing is if you look at those videos you'll see that actually it takes about five minutes to train those animals with classical conditioning which is quite remarkable so what we are planning to do basically is to look at another invertebrate model but i won't tell you what it is and maybe in the future i'll have a chance to talk to you about it thank you very much you