For our third episode of our odor discrimination series, Kayla speaks with Laura, Morgan, and Jo from Conservation Dogs Collective and Emily from Auburn University about a fascinating project undertaken to detect New Zealand Mud Snails.
Science Highlight: None this week!
By Maddie L.
- Laura Holder, Executive Director of Conservation Dogs Collective
- Jo Lock, Canine Keeper at Conservation Dogs Collective
- Morgan, Canine Keeper involved in the project
- Emily from Auburn University, assisting in project design and analysis
New Zealand Mud Snails (NMS):
- NMS are tiny invasive creatures, approximately the size of a grain of rice.
- Native to New Zealand, they were discovered in the United States in 1987.
- NMS reproduce rapidly and lack natural predators in the U.S.
- They can easily hitchhike on fishing equipment, survive outside of water for weeks, and are resistant to digestion by fish.
- The project was initiated through a collaboration between Conservation Dogs Collective, the Wisconsin DNR, and the River Alliance of Wisconsin.
- The primary objective is to train dogs to detect NMS, but acquiring known positive and negative samples for training proved challenging due to the difficulty of guaranteeing the absence of snails.
Challenges in Samples:
- Challenges were faced in obtaining known positive and negative samples.
- Creativity was required to create known negative samples.
- The project partners adapted to develop an effective training strategy.
Training Phase 2:
- Phase 2 of the project introduced a new study design focusing on stimulus sets.
- Dogs were trained to build relevance for NMS scent and discriminate it from other stimuli.
- Training sets were created with varying combinations of positive and negative stimuli, increasing in complexity.
- Dogs had to achieve an 80% success rate before moving to the next set.
- This project stands out from traditional search tasks, focusing on signal detection.
- Dogs must identify a specific signal (snail odor) amid noise.
- Initial training showed promise, but blind testing revealed inconsistent results among the dogs.
- Ernie initially achieved 100% accuracy but dropped to 54%, raising concerns.
- A new approach to training was required, leading to a program pause for further evaluation.
- During training, samples are presented in different sessions, emphasizing the distinction between a “Yes” (snail odor present) and a “No” (snail odor not present) response.
Use of eDNA:
- While environmental DNA (eDNA) analysis is common for detecting invasive species, the project has shifted away from it due to backlogs, cost considerations, and concerns about reliability.
- The experts stress the value of embracing contradictory data in scientific research. Learning from what doesn’t work can be as valuable as confirming hypotheses.
- The project is ongoing, with plans for future training and data collection.
- You can stay updated by following the Conservation Dogs Collective on social media and Auburn University Canine Performance Sciences on Facebook.
Links Mentioned in the Episode:
Where to find Conservation Dogs Collective: Website
Where to find Canine Performance Sciences: Facebook
You can support the K9 Conservationists Podcast by joining our Patreon at patreon.com/k9conservationists.
Kayla Fratt 00:09
Hello and welcome to the K9Conservationists podcast, where we’re positively obsessed with conservation detection dogs. Join us every week to discuss detection, training, canine welfare, conservation, biology and everything that links those three. I’m Kayla Fratt, a co-founder of K9Conservationists, where we train dogs to detect data for land managers, researchers, agencies and NGOs. Today I have the pleasure of talking to Laura, Morgan, and Joe from Conservation Dogs Collective as well as Emily from Auburn University, about a fascinating project undertaken to detect New Zealand mud snails, we’re going to specifically be talking quite a bit about discrimination as part of our mini series on this topic, helping our dogs stay on target in the presence of other similar odors or generalizing to similar odors in the situation where your target odor is difficult to obtain or train on. Since we’ve got so many guests on this episode, we’re going to skip our science highlight and dive right in. So let’s start out with our introductions of who you are, what your role was in this project and the dogs that you share your life with. We’ll start with Laura and then head on over to Joe, Morgan, and Emily will round us out.
Laura Holder 01:12
All right, thank you, Kayla, for having us on. This is gonna be fun and nerdy. And a lot of really good sharing from what we’ve learned. So I’m Laura holder. I’m the executive director, and canine keeper with conservation dogs collective. I have two finders, Ernie and Betty White, who have been involved with this program in particular since 2020, which sounds so long ago. But we’ve been involved from the very beginning and through all the fun phases and learning that we’re going to be talking about today.
Jo Lock 01:44
Hi, I’m Joe. I’m also a keeper. With conservation dogs collective, I have find a holly. And my official involvement really only started with this project this year when we obtained a grant to continue the work and to get even deeper into the discrimination and detection task ahead of us.
I’m Morgan, I am also a canine keeper at conservation dogs collective. My involvement with a project started last year when we really started looking at the how are we going to turn this into research. We’ve got a great concept. Now let’s make it something really cool.
Hi, I’m Emily. As you mentioned, I work with Auburn University and their College of Veterinary Medicine. And I came on this project a couple years ago to help with some of the design and analysis of the project.
Kayla Fratt 02:41
Excellent. So why don’t we we’re going to zoom back or take a half step back now and just for anyone who doesn’t know, what is New Zealand mud snail, why do they matter? I mean, I’m sure we’ve got plenty of people at home who don’t know whether this is an endangered species that we’re trying to protect or an invasive species we’re working on managing and eradicating so someone give us a good lowdown on our New Zealand mud snails.
Laura Holder 03:04
Yeah, so New Zealand mud snails are native to surprise New Zealand in the United States. They are considered an invasive species. They’ve been here since about 1987. They were first kind of discovered in Idaho, where conservation dogs Collective is kind of rooted in the Upper Midwest region. They’re known invasive in multiple states in our area. And some of the concern about why there’s such nasty little invasives for our, our native ecosystems, they’re super tiny, they’re about the size of a grain of rice, when they’re fully grown. The females are born pregnant, and they give birth to live babies. And they can have you know, like 300 babies, which is just terrifying on so many levels, because they are so small that makes them really, really good at hitchhiking a ride on fishing equipment. So in Wisconsin, they’re found in the Driftless region of the state that’s along the Mississippi River. It’s a class one trout fishing like area. So fishers from all around the United States and even around the world come into that part of the state here and do fishing. So these little buggers will get themselves into the tread of waiters or fishing gear, or watercraft and all that. They can also bypass the digestive tract of fish so they have no known natural predators in the United States, which is also again terrifying. And then on top of that, they can also live outside of the water for several weeks. So just think about tiny grains of rice, you know, making their way everywhere all over the landscape and fishing systems, freshwater systems.
Kayla Fratt 04:57
Wow. Yeah. Thank you for that Laura and I grew up so my dad will was the county conservation Escalon and water conservationist for Ashland County. And I grew up with pamphlets about these aquatic invasives AISs, you know, in on our kitchen table and reading them, and I still don’t think I realized just how nasty these guys are. I mean, I know you said it, but again, it’s just stunning that they’re born pregnant, they can survive out of the water for several weeks. So cleaned, drained dry isn’t going to do the trick, and they’re the size of a grain of rice. That’s a huge challenge. Ken, I’m conservationist is thrilled to offer a self study on class. For those interested in joining the field of conservation dog professionals. This course includes 18 modules of video lecture assigned reading homework and quizzes. We have lectures from 10 amazing guest instructors, including fostering motivation and joy through high displacement training with Laura holder of conservation dogs collective safety training at assessments of dog teams with guests Fiona Jackson and Tracy Litton of Schuyler psychology, special considerations for insect and plant training samples with Arden Blumenthal of the New York New Jersey trail conference, and building networks acquiring clients with Paul bunker. Our alumni group is active and supportive, and we welcome students of all levels and backgrounds. The course is priced at $750 with generous financial aid and discounts available for Patreon members, learn more and sign up at Canine conservationist.org/class. So I think that gives us a really good foundation of why we might be paying attention to these New Zealand mud snails. So how did this project come to conservation dogs collective? And what were the initial goals of of this project?
Laura Holder 06:36
Yeah, the project kind of came to us from a connection we have with the Wisconsin DNR and the River Alliance of Wisconsin. So the River Alliance of Wisconsin is a nonprofit that’s dedicated to doing all sorts of really great freshwater education and programming. And they both learned about us in their backyard, so to speak, and they were like, Hey, there’s these little nasty New Zealand mud snails that are in our waterways? Can your dogs be trained to detect them? So it’s not uncommon for lots of conservation orders to get that question like, can your dogs, you know, detect this? And then, you know, so we started some conversations around, you know, where, and why, and all the how and all that stuff. And it came down to just kind of getting some funding to do an initial pilot study. We involved four dogs on the initial year in 2020, to start the program.
Kayla Fratt 07:35
So what were some of your concerns, initially, regarding dog performance on this project? Was there anything just even on that initial email, or as you started training that had you wandering or worrying about how the dogs were going to perform here?
Laura Holder 07:51
I would say in general, and I know everybody else in the room can probably attest to this in general, because there is no history of dogs being used to detect New Zealand mud snails anywhere in the world, we were somewhat, you know, creating this method for this particular species. So besides the like, Oh, my God, maybe my dog can’t smell a New Zealand once Neil thought that we all kind of have at times. A lot of it was around, how do we safely use the dogs first and foremost, so because they then New Zealand, mud snails are so easy to be transferred from the waterways, we decided early on that the dogs were not going to go into the field and do shoreline surveys or any kind of, you know, method application where they might be a vector for actually spreading the snails. So that was one thing right away where it was going to be more of a snails come to the dogs and more container or lineup setups like that. So in some ways, it was like, Ooh, that’s gonna be really easy, because the dogs aren’t traversing a bunch of terrain, but at the same time, the performance of the dogs is going to be for my two, they were so used to going out and sniffing in more natural areas that it was a consideration like, hey, well, this should be easy on them. But then as we learned, as the years went on, some of the performance outcomes of doing that type of work, we’re shedding some light on that method, and application differently.
Kayla Fratt 09:23
Yeah, definitely. That’s a variety of things that are quite different from a lot of the work that we see you all doing out on out in the big wide world. So what would the goal then be to be taking kind of water or soil samples from areas and having the dogs kind of thumbs up? Thumbs down whether or not they were detecting the snails?
Laura Holder 09:42
Yeah, there’s like 400 pages of notes I have, because in the original days, so in the original days, meaning in 2021, we’ve kicked off this program. The training started in early September of 2020. And we had about eight weeks of training the dog is on snails that’s a group of people painstakingly, like went with tweezers and isolate it out for us. So we could do initial target odor training and building the relevance around Target odor. And then from there, I was creating my own, I’ll say like hot samples that simulated fieldwork. So rocks and sediment that were pulled from bodies of water that did not have New Zealand mud snails in them, I’ve actually like made them New Zealand, mud snails and Mason jars and Ziploc bags. So I did all that training in about eight weeks, which was a pretty short amount of time, I was only doing one or two days a week of training, because of my schedule at that time. And then we went in to test day in mid November. And shortly thereafter, Emily joined us from all the results there. But in mid November, we it was in the middle of COVID to write so we met outside in a parking lot. And we ran Ernie and Betty White on five different sets of sediment and substrates that were pulled from the field. And so the DNR and some of their partners had collected those samples. But only a couple of them were from known water bodies with New Zealand mud snail populations, and then the vast majority were either unknown or known to be blank. I don’t know if that totally answers your question. Without going down too many details there. But we we didn’t have a ton of opportunity to train the dogs before testing day on known positives that were pulled from the field. There’s some something that I would share.
Kayla Fratt 11:43
Yeah, that’s always an important, important thing to remember. And certainly challenging. I know. Yeah, we talked about this all the time, how difficult Sample Acquisition can be in this field. So Emily, why don’t you tell us a little bit about how you came in on this project, and a little bit more about kind of when and how that became relevant for for your work.
Laura Holder 12:08
The whole reason that I kind of contacted Emily to was I got these results from the testing day in November. And we did a little second run in December as well. And I was like, the dogs are like indicating a lot like yes, this has snails in it. Yes, this also has snails and I was like this, I mean, great, right? Like, great. They’re getting a lot of rewards and everything. But I was also like, I don’t know if they’re actually correct, because nobody knew ahead of time, if there were a bunch of snails in there. And they just didn’t have time or funding to go back and sift through all the samples and see like, Okay, what was the infestation, you know, population, right, and some of these samples. So that’s what Emily and I started talking and like, I need some data here to really kind of address what the next phase of this program should be right? To really set the dogs up for success, and also the stakeholders for like, Okay, this is what we know now. But there were some big areas of improvement for how we did this first part of training. So
Well, I was contacted by Laura, after she’d gone through the first iteration of testing with the dogs. And so I took a look at her her data set and said, Okay, well, we have we have a lot of information here. But how would we want to analyze it in order to determine where we are in the both the training and testing? Do the dogs know the scent? What What can we get out of this data set? And so I took a look at it and and from the results that I saw, I made some suggestions for how it might, how it might work in a future iteration, so that we had all the data that we needed in order to properly analyze the results and have a full picture of what was happening. So I kind of stepped in to say we just we need to think about how how future whether it be lineups or any sort of testing scenario, we need to figure out what it is that we need to get out of this before we move forward.
Kayla Fratt 14:33
So yeah, Emily, as much as you’re able to share what were some of the recommendations that you made kind of for the team’s going forward to to help make sure that we could actually answer some of these questions that Laura wanted answered.
Well, a lot of times when we are trying to provide a an analysis of of the data like this we need we need to have a At least a set of known true conditions. So we need to know if a sample is positive or is negative, we’ve already mentioned that, truly, it’s difficult to know that when you have an unknown sample, so one of the things that I suggested was that we needed to have more known positives and no negatives. And so a lot of what what I did was help brainstorm, how can we create these types of samples so that we can have our known positives are no negatives, and that we can observe the behavior of the dogs, and determine if they’re learning on the positives and not on the negatives. So that allows us to do an analysis. And so this gives us a basis of understanding what the dogs are capable, capable of, in the context of these known samples. And it will generally allow us to say when we’re giving them unknown samples, what whether you know, with what probability are the dogs going to be right. So this probability of detection, and all of all of this information is background that you want to understand, you know, within the certain testing context, how the dog is operating before you go into the unknowns. So that’s, but as we will discuss here in a minute, it’s a lot trickier than it sounds.
Kayla Fratt 16:50
Yeah, that does sound really tricky. And so what was kind of the process that of chatting with the DNR and everyone about actually getting all of these known and unknown samples? You know, with an invasive that says prolific? How hard is it to even get a really good, known name negative sample that has, you know, like, I’ve got to imagine we’re thinking about substrate, we’re thinking about water quality, we’re thinking about all of these other covariates within the odor profile, what did that look like?
That was incredibly difficult. It was incredibly difficult it was Can anyone guarantee that there aren’t snails there? No, the answer was 100% of the time, no, we cannot guarantee that there are not snails in any of the watersheds. So that causes us to need to get very creative. And, and I mentioned before, that we’re trying to work within, you know, we give the dogs essentially, in a laboratory, we can set up these positives and negatives, but these are, you know, in some sense, they’re created, we can get a known, you know, a non positive from the, from the water from from the watershed. But yes, then no negative is, is very much trickier.
Laura Holder 18:15
I think we had, so internally with the group, the keepers and, Emily, we had, I swear it was probably six to 12 meetings just talking about like, but how can we try to make samples that are negatives, right, like with garden soil, or sand or whatever. And then we just kept going around in circles in a way. Meanwhile, the whole to answer like the client stakeholder side of the equation, Kayleigh, the DNR, and the River Alliance of Wisconsin had complete, open, honest conversations with us, but also like faith and the way that we were approaching the dog training side of things. So that was really something to, you know, be very respectful and mindful of, and also, like, that’s a great opportunity for us to really try to take the lead on, you know, this is how the dog should be utilized, which you get, but then at the same time, you’re like, Okay, now we have figured all this out. So Morgan, do you have anything to add to that? I mean, we can get into the the stimulus set breakdown, then probably, if you don’t,
Oh, my gosh, I don’t have a whole lot to add right here. But I will say this is like when I came in, and it made my head spin because there was so many different questions to answer. And you can’t answer every question in a single study or even just in like two or three studies, you really have to narrow your focus. So when I jumped in my perspective of things was like, holy cow, which direction are we going? It’s narrowed down a lot but you know, this is kind of a home Work of conservation work is this is all uncharted territory all the time. And it’s so hard to pick a direction and know which way to go.
Kayla Fratt 20:11
Yeah, absolutely. And yeah, Laura, I just want to underline what you said about it sounds like you had a really, really good supportive open project partners here and like hats off to Wisconsin DNR for that. That’s really great to hear. So yeah, let’s, let’s get into it, you know, we’ve got, we’ve got these results, these initial results that don’t quite make sense, we’ve got our Dream Team consultant brought on now what are we doing?
Laura Holder 20:32
So I call it select the 2022 to 20, or 2021 2022. phase two study. So we from the ground up, or from the paper up, we built a brand new study design. And where we ended up landing was we identified kind of three phases of training, which again, in hindsight, you’re like, should have just done maybe one, three phases of training, where we focus in on doing stimulus sets, and this is where I’m gonna pass the mic to Emily, because you’re the one that did all the kind of proactive research on like, here’s what we should probably try out with the dogs.
My thought on training the dogs would be that first, we really want to build the relevance of the mud snail, we want them to understand what is hot. But my thoughts also in this kind of second phase was that at some point, we’d also want our dogs to be able to demonstrate that they could discriminate the mud snails from many other common organisms, other stimuli that were in, in the water, the water itself, the chemicals that were floating in the water, the actual organisms. And again, this is where Yeah, our heads started to spin it, how many different variables, different, different scents, different, you know, live versus dead. upstream downstream, we were just the threshold of detection, we started going down all of these different paths. So we tried at first. So what I had thought was we should start with simple sets where this has a month now this doesn’t and have the dogs really understand that they need to give us an indication when a month now is present, and then slowly build up the complexity of, of the stimulus that was presented to them. So I kind of came up with a potential sequence of, of samples that the dogs would be trained on. So they would need they would need to find the mud snail in a simpler set, and then a more complex set, and then more complex set, and each set, the dog would need to be 80% successful before they could move on to the next set. How did that go, Laura?
Laura Holder 23:18
Oh, it’s fun. So my freezer was full of all sorts of other creepy crawlies because of this too. So in addition to just the Sample Acquisition and storage aspect, I want to touch on that because we decided to go this route. And then that meant, again, we had some really great volunteers that were like picking out native snails, you know, and then Canis fly and black fly larva and crawfish and weird plants that I don’t know the name of anymore, right. So we set up those, those sets that Emily was talking about, so we’d have snail in distilled water. And then we would have let’s say the crawfish would be in with the snails in that same jar, and then we would have a jar with just a crawfish in it. And that’s what we presented. So we had our control with just the crawfish and then our crayfish, right, if depending on where you live snails with the crawfish in the other jar. So at first, you know, like we did. Gosh, let me look at my my notes here, because we did, I think eight weeks of training again. And all of the training setups were known. So I knew like the snails are going to be in this one, you know, reward the dog for correct indication here. So things were going along really well. And that was from I think, October to December of 2021. And then we started doing some testing blind testing for me and the dogs in January. And Ernie was about 67% accurate, but he was 100% accurate in the first set of trials. And then we did a couple blind searches the next week, and then earnings went to 100 and then but he went to 54 sent in I was like what is happening? Right? So I’m going like what is happening? This is where Joe started to get my 911 calls of join you to talk to you about this. Morgan, Emily help like something’s going on with the dogs results, right. So I knew enough at that point, I’m like, have to pump the brakes with the approach. And that’s, that’s where we kind of just pause the program for a while. And again, like I told the DNR, the river lines like this is what’s happening. This is what we’re seeing with this study designed and the outcomes. So that opened a whole nother discussion on the training setup again, so you can see how like, we’re like, yes, things are going great. And then you hit another roadblock or a speed bump, or sometimes it felt like a brick wall, right? Because we’re all human. And we have feelings, as well. So
Kayla Fratt 25:53
yeah, gosh, that sounds really tough. And yeah, so Joe and Morgan, why don’t you just hop in now, with you know, what were some of your initial thoughts or hypotheses as you’re getting these 911 calls from Laura, from the field? And, you know, Emily, we’ll have you hop in as much as you’d like here, and then we’ll talk about Okay, now, what’s our next pivot?
Jo Lock 26:14
Yeah, thanks, sir. So, just like Morgan, my first thought was, gosh, this is hard. My head was spinning a little bit, too. I guess my, my first reaction is, has already been alluded to is that this isn’t really a search task. This isn’t really what we normally do. This isn’t really our bread and butter. This is a signal detection task. So what I mean by that is that we’re trying to detect a signal in a sample of dogs are not me. And we’re not really sure what that signal is, and how do we define what that signal is. And on top of that, the next level of complexity is that the dog needs to simultaneously be able to be very sensitive to that signal and learn the sensitivity of their signal because we need to catch the odor of these snails and the samples early so that they can bring in the management management tools. But plus, the dog also has to be very precise, because they can’t be distracted by any noise that’s in the signal. So in our teaching process, we have to eliminate as much bias as possible. And my concern about it was, as well as that, because it’s not a search task. And it’s not really what the dogs normally do, it could be kind of a bit boring and repetitive for them, as well as being quite taxing, because we’re asking a lot of them from their olfactory systems. So we need to find ways to mitigate any biases that might come in and trying to reduce the response cost for them and motivate them to do the task really, really well. And I guess we probably should define some of the terms that we’re throwing around. So signal detection has, when you detect a signal, there’s four possible outcomes. So you can have a hit, or a miss or a correct rejection or a false alarm. The four four possible outcomes of detecting a signal. And when you’re looking for improving sensitivity, you’re trying to determine the proportion of the samples that are correctly identified as containing the odor in relation to the number of samples that truly do contain it. So that comes back to the conversation about how really important it is to know if your your sample contains the odor or not. And then the precise element is to ensure that the proportion of the samples that are correctly identified as containing the odor. How many of those are there are in relation to the actual Yes, responses given by the dog? So? In other words, how well can the doc pick this nail odor out from the noise? So you’re looking at all the false alarms they give and the hits they give? And how many of those are hits. So what are the proportion of hits in relation to false alarms? Yeah, I don’t know if Morgan wants to jump in here.
Joe, I feel like you were reading my notes. You took the words right out of my mouth, I have really nothing to add. But I know that you probably have a whole lot to say about motivating dogs and influencing that bias through training decisions.
Jo Lock 29:30
One of the things that I worried about with this was that I think we can get a little bit tripped up specifically with this project. Because when we when we think about detection, typically we tend to view it as having a right and a wrong answer. So to throw some more technical terms in we think about there being a discriminative stimulus and SD and as an s delta. So yes, correct an incorrect response. But the problem with this particular project is that we want the dogs to detect Want the dogs to say, to be accurate in saying that the sample does contain the odor, but we also want them to be very accurate in saying that it doesn’t. So in some ways, there’s two correct answers. And we want to make sure the dog understands that. And so we started to talk about whether or not we needed to think about actually training that and specifically rewarding both of those things, because they’re both equally valid. So we got into some, we got deep into some of that stuff. And we’re still kind of kicking all around. So, you know, the, the normal, the normal way to approach some of this stuff, is is to, you know, Emily talked about earlier setting out a lineup. But I, with my background in cybernetics, I started to think about what okay, if we didn’t have dogs doing this job? I mean, actually, if there was somewhere, you know, magical, mechanical sensor that could do this job. How would we design that to work? And it’s unlikely that we would ask the sensor to run down a line of possible answers and tell us which one was correct. We wouldn’t, we wouldn’t present it in that way, we would actually just give one sample to the sensor. And we would say to the sensor, is this nail odor present in here a lot? Yes or no? So that sets up a completely different type of, of type of scenario for the dog. So in that situation, both a hit and a correct rejection would both be correct answers. So we wondered about whether or not we needed to train two different responses for the dog to give them the opportunity to actually tell us that. And if we do that, then we also have to make sure that the response cost in other words, the way that the dog tells us about that has to be similar. So it can’t be more costly and more effortful for the dog to say yes or no, otherwise, you build in some bias. And we also have to make sure that both of those responses are reinforced, and they have to be reinforced, in the same way with the same equal level of risk of free enforcement.
Kayla Fratt 32:11
I’m so glad you brought all of that up. And thank you for I know, we’re going back five minutes now. But all of the definitions and everything that’s also important with this, and yeah, really, I think it makes perfect sense to think about this. This is not your normal search task. The dog is not examining the environment for a trace of target odor and following a scent cone to actually find their target odor. It’s kind of a a yes, no go no go procedure more like, so. Yeah. What did that what did that how did that approach end up working? And is that is that the proach? You ended up ultimately taking?
We’re we’re working on it. Now. There’s, there’s no answer to that yet. But there will be asking in next year. Because because that’s pretty much present day.
Now. I want to interject here that I was speaking with one of the grad students the other day and saying a single, a single port, essentially versus multiple ports versus lineups. I just wanted to get some more thoughts on on that. And so we were discussing how costly it is, in terms of effort and how much how much effort goes into checking a fiver eight can line up that’s incredibly mentally taxing. And that a single port, you know, essentially yes or no, that that can work in your favor than but with complex sense. And I would imagine that we have a pretty complex sent picture for both the snail and everything around it that sometimes having a few other ports, maybe two or three where the dog can go back and forth, back can give them something to compare. And that so sometimes that actually provides an advantage to a single port. And so kind of science in in motion right now is you know, your your dog could do a single port where there’s a yes or no answer is one thing but also a yes here but no there. How would you train that? But would that give you functionally better results than a single port? I don’t know the answer to that. But but something to consider in the future of this project.
Kayla Fratt 34:52
That is really interesting. And I’m having flashbacks to all of the conversations that I’ve heard you As Dr. Hall having about, you know, their Olfactometer, 3 trillion or whatever it is that they’re on to now in their lab, and, you know, it really sounds like yeah, you guys have this project is such a big shift from a typical search. And I it’s just so interesting to think about trying to bring all of these things in. So, you know, so we’ve brought in the other two or three dogs. Now, Morgan, I’m not sure if you’re working with Sonny on this as well. Or if it’s just just Jesse at this point. But what have some of your considerations been as far as helping these dogs that are more used to arranging and doing more of a broad search actually get used to whether it’s this lineup or you do more of like the single port? go no go sorts of setups, I can imagine, Laura, for your labs that can be really challenging, are you using the same reinforcers like, let’s talk nitty gritty about what this actually looks like to get the dogs from searching acres and acres and acres to just checking, you know, a max of maybe eight pots.
You know, I actually have a lot to say about this one, because Jesse is a really challenging dog. He, I mean, he’s wonderful, but he’s challenging. He is so good at searching big areas for stinky decomposing things. When I have tried to switch him to other target odors, he does really good in new environments, like inside the house. But once I take him back outside, again, it all falls apart. And I truly think it’s because you know, that stimulus package goes back to like we’re searching outside, we’re looking for stinky dead things, and he just blows right past these odors that he’s learned really, really well inside, even even in tiny, tiny amounts. So for me, I think the biggest key for us is going to be really setting up a new environment, we’re not going to be able to be searching ports outdoors, because Jesse’s going to default back into like, we’re looking for dead stuff, we’re gonna have to do it like probably in a room and just make everything different. This is not a search task. Now we’re, we’re not doing the search trick. Right now we’re doing the, the like Port trick, we’re checking this hole, we’re checking this hole, good job, you’ve done it and just make it completely completely opposite of anything he knows. Otherwise, he defaults in his old behaviors, and they’re good behaviors, but they’re not what we need for this project.
Laura Holder 37:28
I know when we landed at one, we got funding, let me say that to continue to to support this program. And I’m very, very grateful for that, too. Because I feel like the stuff we’re learning along the way can apply across so many other projects that maybe haven’t even been discovered yet in the world of conservation detection. But when it you know that funding came in, I was like, Yeah, awesome. Here we go, we’re gonna try and figure out how to do this, you know, right from the get go. Given all of the information we have. And I mentioned something straight out to Joe, I remember we’re talking and just in our normal daily meetings that like, I don’t know, if my dogs are gonna like doing a single, like, yes, no thing, right. And then like, I sat on it and slept on a while I’m like, it’s you, Laura. Like, it’s not the dogs, they are incredibly smart. And they can figure out contextual clues and all that stuff. To understand, like, this is what you’re doing now. And at the end of the day, they’re going to be thrilled that they get reward, you know, for doing the correct task. So we’re still developing with this training protocol is going to look like across the organization, because we’ll have four dogs, you know, two of them live in my home, but the other two are in different homes with different handlers and really developing, you know, like, the dog gets the freeze dried cheese, regardless if it’s the yes or the No, you know, so you really help eliminate any kind of bias and it’s hard with labs sometimes, right? Because they’re like, all you do anything. So just trying to really put those parameters and framework around, like how can we do what we can to make this a clean task? Right, and Joe and Morgan and I and Emily are working on what that what that’s going to look like because we probably won’t kick off training until like August or September of this year. So we have some time to really figure figure this out.
Kayla Fratt 39:27
That’s really exciting. And you know, I’ve been meaning to say a couple times but congrats on the grant it’s really cool that you guys are going to be able to actually explore this and it’s not just one of those projects is going to fade away because of lack of funding i It breaks my heart think about how common that is. Yeah. And yeah, and I’m glad that you mentioned so i Laura, I gather from photos that normally your dogs are dogs that get toys in the field, but for this we’re thinking freeze dried cheese, which is kind of why I asked about reinforcers and I am I right about that are your dogs normally The toy reward guys kind of in the field,
Laura Holder 40:03
they It depends. It depends on the day, they typically get food reinforcers for me. But Betty, I’ll also mix in a little bit of toy play as well. For this project, in particular, I’ll probably use, I’ll say, for the transactional part of things when they, you know, do the yesno behavior, they’ll get a cookie, and then at the end of a session, like, given all done, we’re, you know, done now we’re gonna go play, then usually, that’s how I break up these kinds of tasks for them.
Kayla Fratt 40:33
That makes sense. And that was, again, that was kind of why I had asked that because I, I know from experience with barley, that trying to work on lineups, and yes, no sorts of things with him with toys involved works exponentially less well than doing search tasks with him with toy rewards involved. And yeah, when you kind of want this level of like focus and consideration and careful maneuvering, it seems like that can be pretty tricky. If you’ve got big feelings about toys involved,
Laura Holder 41:05
you can get a lot more repetitions and not that we’re gonna be aiming for 105 minutes or anything, but I feel like you know, if you wanted to do 10 repetitions in a couple minutes, I think that’s just easier to with the food like you’re, you’re done. Yeah.
Kayla Fratt 41:20
Yeah, well, and even just thinking about the logistics of having a good training space where you can give a toy reward in a way that is rewarding for the dog that can be challenging thinking about substrate and knocking over all of your, your, you know, your fitment fit pause stuff. If you’re even even if you’re lucky enough to have a training room that has your fit, pause. And it’s not you know, your laptop that you’re worried about the dog knocking over with a weird balance. So that makes perfect sense. So what, what else would you guys like to kind of circle back to or bring up as far as this project? What are you excited about for the future? Just anything else that we haven’t covered yet about this project and discrimination in mud snails?
Can I ask a question have you guys have you considered as you are preparing for this next wave the proportions of positive and negative samples that you present to the dogs so that so that the dogs get enough exposure to the positives, but then when you? Yeah, just have you guys consider that.
I would say it’s still under development. Because this is, this is year three of this project now. And every year, we’ve identified what didn’t work and what did work and then amended it. And so like, we’re at that stage where we’re looking back so that we can look forward. So it’s something that is on the list to consider, but we’re still working on the plans.
Jo Lock 42:52
And one of the things that I’ve been thinking about in relation to this is not so much about the proportion, but about the making sure that those samples are presented in the same session, so that the dog gets a clear contrast between what’s Yes, and what’s know in the same session, and really starts to be able to pick out the characteristics of what makes it a yes response. And what makes it a no response. And that, to me is like the real crux of this is making it really, really clear for the dog. So stimulus control is, you know, when the odor is present, I do this and when the odor is not present, I do this. And if you can, if we can make that very clear. And so the way to achieve that, I think would be to start with examples that are very simple in both of those categories. And then gradually increase the complexity. So that you get first of all, you start off with very, you know, very, obviously different. So very far out examples and non examples, and then you slowly start to bring the examples and non examples closer and closer in that you don’t start there. So we want to try and eliminate as many mistakes so that the dog never has to go through like too many extinction cycles and get confused. We need to try and keep the clarity as much as possible all the way through the training process.
Laura Holder 44:19
Just thinking of people that are listening are really new to this whole topic. Joe, when you say like a really out there example, in a task. Can you give an example of like what that order or item might be that the dogs would be presented with? That would be a no answer.
Jo Lock 44:39
I give an example on a project a different project. So a year or so ago, I was asked to do a project searching for a rare endangered plant. And so I had to teach. I had to teach Willow how to find this plant with a very small sub set of examples because obviously, it was very rare. So I think I had six of them or something. So that was challenging to start with. But one of the concerns was that there was a common variety that there was there was suspicion or concern that it because it was so closely genetically related, it would have a similar odor. And so my training approach for that was to start off, once once I started, once I built the odor relevance for the actual odor, I then took that odor out into the environment, but I started off putting environments where it was not going to come anywhere close to being that species. So I put it out in, in environments that had lots of plants, but no plants that were anywhere near that, that that genus, and then then I started to put it out into places where I knew there was going to be other millenniums, but not but not the common variety, not the autumn gnarly. And then I gradually moved it toward into environments where I got closer and closer to it until I actually put it right right next to the common variety. And by then, because she had enough examples of Yes, not not yes and not not Yes. She showed no interest in the common variety. So that’s that’s the kind of thing I mean, it’s like you, you, you, you, you think about it. It’s like mathematical set theory. So there’s one set is New Zealand muzzleloader, and one set is not New Zealand has that odor. And those two sets do not interact, they are not, there is no union or or intersection between those two. But there’s lots and lots of different everything that every basically there’s New Zealand, muzzleloader, and everything else. And so we wouldn’t probably start off with giving them a native snail or another mollusk as the as the non example. He would start off with something completely, completely different.
Laura Holder 46:58
Like an orange slice. Or whatever, right?
Kayla Fratt 47:04
Yeah, yeah. Yeah. Well, I think for anyone at home who’s curious about this, and would like to learn more that kind of foundational paper that underpins all of this, it would be discrimination learning with and without errors by hS terrorists from 1963. But it talks about training pigeons to click a specific color key. And they start with one key being much larger and much brighter and everything than the other key so that they’re functionally learning to ignore the stimulus that they’re supposed to ignore, by the fact that that stimulus is really, really ignorable. And again, we’ll link that paper in the show notes for anyone who’s curious about it. And we’ll also link a paper by Eileen and dogs are an article by Eileen and dogs that covers it in a little bit more kind of, quote, unquote, plain English for everyone. But Joe just gave a really lovely breakdown. I do have a question. And you all might not know the answer to this. But why or are they also using E DNA for this project? And if not, why not? I’m curious why Edna did not come into the picture as well as or instead of dogs potentially.
Laura Holder 48:10
They have used Edna for all sorts of aquatic invasive species, but in the state of Wisconsin, in particular, they got so backlogged in the labs doing the analysis. That was one reason why they have kind of moved away from that the cost involved with the E DNA analysis is also needs to be justified, right? So when we did our testing day, in November 2020, even though that’s what we’ve moved away from, you know, what the, the lineups the dogs were going down and back in like 30 seconds, and indicating, like it’s here versus here. And the people were like, that’s, you get the results like that, you know, like so just seeing the dogs capabilities was something really like, Okay. Also, DNA has known, has been known to be unreliable in some regards, as well. So there’s there’s multiple factors of why they might not be moving away completely from E DNA, but maybe being more selective with the usage of E DNA for this particular species.
Kayla Fratt 49:14
Excellent. Yeah. No, thank you for that. I just especially if we’ve got any biologists in the audience who are kind of sitting here screaming, why not Edna. All right. Well, is there anything else that we wanted to circle back to or other training things? This has just been so fascinating, and I’m so excited to see where this goes in next couple of years for y’all. We’ll have to have you back on and everything probably after it sounds like this is something that you’re thinking about going for publication with as well. Seems like a good fit for that. I’m getting a lot of nodding.
Jo Lock 49:45
I have one last thought. And it’s kind of like a soapbox, sometimes of mine. I just wanted to I was thinking about all of this. I’m just thinking that science is a process of discovery. and sometimes getting data that contradicts our hypotheses should be considered as just as valuable and sometimes even more valuable than getting a confirming our hypotheses. But that’s not always the case because of human nature. It’s quite hard for us to think that way sometimes. And we’re going to be probably experimenting with some some hypotheses that, you know, perhaps are not the kind of normal way that detection work is done. And so I have this quote on my wall from a son as a trainer called Hilary Hankey. She trains, birds, she’s a bird trainer. And she just recently did a course it titled how wonderful I was wrong. And I just thought that was just such a lovely, just such a lovely sentiment that I actually wrote it down and put it on a post it note and I have it on my wall. So I’m trying to embrace the how wonderful I was wrong. philosophy with with some of the things on this project.
Kayla Fratt 50:55
Yeah, that seems like a lovely note to end it on. And no, again, I think as much as we love finding all the places that detection dogs are the perfect, amazing solution for everything, I think, really highlighting the times where maybe they still will be but how hard it can be and how creative we have to be and the level of expertise needed to succeed with this is really important. And also potentially the times where detection dogs may or may not end up being the end all be all for any given project. So thank you all so much, Laura, why don’t you remind us where we can find conservation dogs collective and how we can support y’all and Emily, same for you with with your work.
Laura Holder 51:38
You can follow along and our adventures on social media. We’re on Facebook, Instagram, Pinterest, we have a YouTube channel as well. And then our website is a really great resource to learn all about us. That’s conservationdogscollective.org.
So I’m with Auburn University, and you can find canine performance sciences on Facebook.
Kayla Fratt 52:03
Yeah, and we’ll make sure to link all of that in the show notes. As always, for everyone at home. Thank you so much for listening. I hope you learned a lot and you’re feeling inspired to get outside and be a canine conservationist in whatever way suits your passions and your skill sets. You can find shownotes, donate to K9Conservationists, join our online course, join our Patreon, all that good stuff over at k9conservationists.org Until next time!