The Tortoise and The Hare … and we all know who won that particular race!

giant-tortoise-2516539_1920.jpg
Showing a clean pair of heels

It’s hard not to be impressed by the dog who speeds around the search area appearing totally engaged with the task in hand as he tries to locate his target scent. In comparison, the dog who moves more slowly is often overlooked in favour of his flashier counterpart. But, where scent detecting is concerned, speed might not be everything and selecting the right breed for the job may be more a matter of taste than a matter of science.

hare-3186347_1920.jpg
Speed isn’t everything

In the UK, the most commonly selected breeds for scent detecting work include (amongst others) Labradors, Spaniels, Border Collies, German Shepherds and Malinois. But which breed is best?

Gadbois and Reeve (2014) favour the Border Collie, basing their preference, amongst other things, on the breed’s apparent persistence and good sniffing ability. They propose that certain breeds have higher baseline levels of dopamine than others which, they state, “raises the intriguing possibility that baseline dopamine levels may have a direct impact on cognition, motivation, learning, and overall olfactory behaviour and performance”. They refer to this possibility as their “dopamine hypothesis” and link this to the notion of “work ethic”. Other breeds they single out as “dopamine dogs” include Belgian Malinois and Jack Russell Terriers. More about dopamine in a future blog.

27747868_1997985543807160_689171945931869202_o.jpg
Gadbois and Reeve (2014) favour the Border Collie. Floss in action.

All that said, Hall, et al (2015) and Rooney and Bradshaw (2004) argue that the choice of working dog may have far more to do with historical precedent than any real evidence of superior scent detecting ability. According to Johnen, et al (2017), good quality empirical studies remain thin on the ground and as Jamieson, et al (2017) warn, “continually selecting the same dog breeds, without inspecting other breeds, may reduce the effectiveness of detection dog programs”.

Interestingly, in McCulloch, et als (2006) study investigating the ability of dogs to detect early- and late-stage lung and breast cancer, it was a mix of pet dogs (rather than ‘working’ dogs) with only basic puppy training who (in a matter of weeks) were able to accurately identify breath samples from lung and breast cancer patients.

Hall, et al (2015) argue for the need for “direct behavioural measurement of assumed behavioural breed differences”. Contrary to all expectations, in their 2015 study comparing the scent detecting abilities of German Shepherds, Greyhounds and Pugs, the Pugs significantly outperformed both other breeds.

pug
The Pug … a surprisingly good scent detecting breed?

Whilst Hall, et als (2015) study may not have persuaded you to purchase your first scent detecting pug … well, not yet anyway … it does raise another important question. Did the Pugs succeed because of their superior olfactory ability or because of the training methods employed in the study?

In an attempt to address this question, Polgar, et al (2016) developed a simple strategy to measure differences in olfactory ability that did not require any pre-training. Raw turkey meat was placed under 1 of 4 ceramic pots. Five levels of difficulty were arranged by way of decreasing numbers of holes in the ceramic pots. Subjects were assigned to one of four groups; dog breeds selected for their scent detecting abilities, dog breeds selected for other purposes, dog breeds with short noses and hand-reared grey wolves. The dogs / wolves were led along the row of pots to identify, by scent alone, which contained the meat. Dog breeds selected for scent detecting work out-performed both the short nosed breeds and those bred for other purposes. At the most difficult level, wolves and the scenting breeds performed better than chance.

Interestingly, according to Maejima, et al (2007), “research with drug detection dogs failed to reveal any significant link between performance and theoretically relevant genotypes” however, Lesniak, et al (2008) argue that breed and heredity may have some influence on scent detecting ability … “there may be a relation between a certain genotype at a particular locus and the ability of more accurate scent detection of particular volatile organic compounds”. This raises the intriguing possibility that not only may some breeds (or individuals within a breed) have better olfactory abilities than others but that some breeds (or individuals within a breed) may have better olfactory abilities when tasked to detect particular target scents.

Similarly, we also show variation in our ability to detect certain odours. Odour sensitivity is heritable. One well referenced example of this “specific anosmia” relates to the ability to detect a sulphurous odour in urine following asparagus consumption (Pelchat, et al (2011)). Some of us can smell it, some of us can’t. Could our dogs also have “specific anosmias”?

asparagus-2538848_1920
Asparagus – Can you smell it?

Returning then to the original subject of this blog. Tortoise or Hare? Accuracy or speed? According to Helton (2009) “a dog may quickly find substances, but may miss targets because of haste. A dog may also be very slow to find substances, but based on a methodical approach may miss few or no targets”. Self-evidently, a dog whose performance is both accurate and fast seems hugely desirable and, as Helton (2009) suggests, “training and practice can markedly improve skill” … a combination of speed and accuracy, perhaps?

But, can training and practice really alter your dog’s general “approach” (or speed) to scent detecting work? Can you turn your slow and thoughtful Rottie into a speedy working Cocker? Experience would suggest not. Each breed (and every dog within that breed) comes with its own, highly individual, set of characteristics; a mixture of heredity and past life experiences. Far better to work with the dog you have in front of you than attempt to turn him into something he can never be. For me, one of the big enjoyments of working with so many different breeds is their very different styles of working. Each and every one of them brings something new to scent detecting and helps break down any stereotypical views of what different breeds might be capable of. Tortoise or Hare … I’ll take both please!

28938874_1552282231555834_1822055471_o
Slow and thoughtful Jim
28938468_1552288101555247_1345316862_o
Speedy Brook

 

 

 

 

 

 

 

 

So, a thought to leave you with. When Helton, et al (2009) asked a trainer which breed was best as an explosives detector dog, his reply was … “depends on where you live, you see here in the United States it is the Labrador Retriever, in the United Kingdom it is the Springer Spaniel, in Belgium it is the Belgian Malinois, in Germany the German Shepherd, and in Norway, they are pushing the Norwegian Drever, a dog most people have never heard of. Personally, I think it is all goofy, pick any dog you are comfortable with”.

And, if you think a Norwegian Drever might be the scent detecting dog for you … just follow this link … https://www.skk.se/en/NKU-home/nordic-dog-breeds/sweden/drever/

© Lesley McAllister – Scent : Detect : Find Ltd 2018

References / Further Reading

  1. Gadbois S and Reeve C (2014) Canine Olfaction: Scent, Sign, and Situation. IN: Horowitz A (Ed) Domestic Dog Cognition and Behavior. The Scientific Study of Canis Familiaris. London: Springer.
  2. Hall N J, Glenn K, Smith D W and Wynne D L (2015) Performance of Pugs, German Shepherds and Greyhounds (Canis Lupus Familiaris) on an Odor Discrimination Task. Journal of Comparative Psychology. 129(3). 237-246.
  3. Helton W S (2009) Overview of Scent Detection Work. Issues and Opportunities. IN: Helton W S (Ed) Canine Ergonomics. The Science of Working Dogs. London: Taylor and Francis Group.
  4. Helton W S, Feltovich P J and Velkey A J (2009) Skill and Expertise in Working Dogs. A Cognitive Science Perspective. IN: Helton W S (Ed) Canine Ergonomics. The Science of Working Dogs. London: Taylor and Francis Group.
  5. Jamieson L T J, Baxter G S and Murray P J (2017) Identifying suitable detection dogs. Applied Animal Behaviour Science. 195. 1-7.
  6. Johnen D, Heuwieser W and Fischer-Tenhagen C (2017) An approach to identify bias in scent detection dog testing. Applied Animal Behaviour Science. 189. 1-12.
  7. Lesniak A, Walczak M, Jezierski T, Sacharczuk M, Gawkowski M and Jaszczak K (2008) Canine Olfactory Receptor Gene Polymorphism and Its Relation to Odor Detection Performance by Sniffer Dogs. Journal of Heredity. 99 (5). 518-527
  8. Maejima M, Inoue-Murayama M, Tonosaki K, Matsuura N, Kato S, Saito Y, Weiss A, Murayama Y and Ito S (2007) Traits and genotypes may predict the successful training of drug detection dogs. Applied Animal Behaviour Science. 107. 3-4. 287–298.
  9. McCulloch M, Jezierski T, Broffman M, Hubbard A, Turner K and Janecki T (2006) Diagnostic Accuracy of canine Scent Detection in Early- and Late-Stage Lung and Breast Cancers. Integrative Cancer Therapies. 5(1). 30-39
  10. Pelchat M L, Bykowski C, Duke F F and Reed D R (2011) Excretion and Perception of a Characteristic Odor in Urine after Aspragus Ingestion: a Psychophysical and Genetic Study. Chemical Senses. 36. 1. 9-17.
  11. Polgar Z, Kinnunen M, Ujvary D and Gacsi M (2016) A Test of Canine Olfactory Capacity. Comparing Various Dog Breeds and Wolves in a Natural Detection task. Plos One. May. 1-14.
  12. Rooney N J and Bradshaw J W S (2004) Breed and sex differences in the behavioural attributes of specialist search dogs – a questionnaire survey of trainers and handlers. Applied Animal Behaviour Science. 86. 123-135.

Blank Searches. Yes or No?

The Blank Search – where your dog is required to search an environment that doesn’t contain his target scent – continues to be a common feature of many Scent Detecting Courses. If you’ve already spent many hours encouraging your dog to search for, and locate, his target scent, why might you want (or need) to train your dog to search an area when it isn’t there?

Two common responses to this question, include;

  1. Blank searches will increase your dog’s motivation to search.
  2. Future search areas may not contain your dog’s target scent. He needs to be ‘trained’ in preparation for this.

So, let’s examine each of these claims in turn and see just how much water they hold.

bucket-999973_1920

Do these claims hold any water?

Blank searches will increase your dog’s motivation to search.

In a study by Gazit, et al (2005), seven highly trained explosives detection dogs took part in a series of 4 experiments.

In the first experiment the dogs were trained alternately on two very similar paths. Path A always contained 5 hidden explosives whilst Path B contained no explosives. In just a few sessions the dogs demonstrated a significant decrease in search behaviour on Path B but not on Path A. This seems to offer the first, albeit small, piece of evidence that blank searches may actually reduce, rather than increase, a dog’s motivation to search.

In the second experiment the dogs were trained exclusively on Path B. 1 explosive was hidden on the path every 4 days. The dogs detection rates during this second experiment were found to be significantly lower than in the first experiment.

The third experiment was conducted on Path C. As in the second experiment, only 1 explosive was hidden on this new path every 4 days. Interestingly, both the probability of detection and motivation to search were significantly higher than in second experiment despite the number of hidden explosives being the same in both experiments.

In the fourth experiment the dogs were returned to Path B. Despite being trained for another 12 days, with 1 explosive being hidden on the path on each of the daily sessions, the dogs failed to demonstrate the levels of motivation that they had previously shown when searching any of the other paths that had usually contained explosives.

According to Gazit, et al (2005), these findings suggest that even a highly trained Explosives Detection Dog will quickly learn that a specific area does not contain explosives and, as a result, the dog will be “less motivated to search and will miss newly placed targets”.

As the Gazit, et al (2005) study demonstrates, the loss in search and detection performance following exposure to only a small number of blank searches, is ‘context specific’. In other words, the decline in the dogs performance is limited to the area in which they have been exposed to the blank search condition. These findings are further supported by Porritt, et als (2015) study into performance decline in search dogs. They state that “without the opportunity to find rewarded targets in repetitive search environments, scent detecting dogs will become ineffective after a short period, and this performance decrement is hard to reverse”.

So, what are the implications of these study results for your own scent detecting training? Because of the speed at which this decline in motivation to search is evidenced, I no longer set up any blank searches for my own dogs and actively discourage my clients from doing so too. Although this reported decline in motivation is said to be context specific, this may actually cause you extra difficulties. For many of you, choice of training area can be very limited and if you practice blank searches in your regular venue it seems likely that this will be to the detriment of any future training in this same area. As Gazit, et al (2005) conclude, “if the [search] behavior is extinguished in a specific context, it will be very difficult to restore that [search] behavior in that context”.

Some anecdotal evidence, provided by my own two dogs, may lend further support to my ‘Just say NO to Blank Searches‘ training stance. ‘L’ has experienced a number of blank searches in her scent detecting training career. ‘BB’, has no experience of blank searches. Although it has been many years since ‘L’s’ last blank search, her motivation to search is, in my opinion, markedly reduced. She has a tendency to slow down and eventually stop searching if she doesn’t locate her target scent within just a few minutes. In contrast, ‘BB’s’ motivation to locate the target scent increases (faster and more thorough) the longer the search continues. ‘L’ has been ‘fooled’ in the past – asked to search when the area was blank. ‘BB’ has never experienced this ‘deception’ – if asked to search all his experiences tell him that his target scent will be present, he just needs to continue searching (perhaps more thoroughly) until he can locate it … and he does!

It seems worth asking ourselves here whether it’s reasonable to cue our dogs to perform a  behaviour that we know, because of the set-up, they won’t be able to complete? One example that comes immediately to mind is the practice of pretending to throw a toy. The dog rushes off to find it, searches around for a while, looks back, ‘puzzled’, at the ‘thrower’ and, eventually, gives up. Personally, I dislike this practice and I’m pretty sure the dog isn’t too keen on it either!

dog-863529_1920.jpg

Pretending to throw a toy …

Similarly, with the blank search, the dog is cued to perform a very specific behaviour chain – search an area, locate the target scent and indicate its location. Except … it can’t. It’s been set-up to fail. There is no target scent to find. Not only does this practice run the very real risk of diminishing the dog’s motivation to search but I wonder what it does for the relationship between dog and handler? Worth thinking about?

Future search areas may not contain your dog’s target scent. He needs to be ‘trained’ in preparation for this.

For the pet, hobby or sports dog I can think of no convincing reason that they should ever have to search an area devoid of target scent. Unfortunately, because blank searches continue to appear in the competition scent-work arena, handlers may feel that they have little option but to train for this possibility. I would simply urge caution.

In contrast, blank searches are a very real, everyday, experience for the working dog team; those that earn their living by scent detecting. But does this have to be the case?

According to Porritt, et al (2015) “a co-trained, non-contraband odour, secreted in a dog’s working environment and contingently reinforced upon being found, acts to maintain performance in finding contraband target odours that would be rarely encountered during a dog’s working life”.

In Porritt, et als (2015) study, 21 labrador retrievers were trained to detect 3 ‘contraband’ odours (explosives) and 1 ‘non-contraband’ odour. They were then exposed to a number of different experimental search conditions. The group of dogs who had worked extensively with the non-contraband odour showed “little difference in their detection on non-explosive and explosive targets in the test phase. … maintenance of expectation to find any target was sufficient to maintain attention to all odour targets”.

So, what does this all mean? Put simply, it suggests a highly practical method by which the blank search problem can be managed. Working dogs can be trained to detect multiple scents (Williams and Johnston (2002) suggest that dogs can be trained to detect up to ten different scents) including one that can be placed in the working environment prior to a search to ensure that there is always a find for the dog. A highly practical solution to the blank search problem!

Blank searches. Yes or No? I think you know my answer.

 

© Scent : Detect : Find Ltd 2018

References / Further Reading

  1. Gazit I, Goldblatt A and Terkel J (2005) The role of context specificity in learning: The effects of training context on explosives detection in dogs. Animal Cognition. August. 143-150.
  2. Porritt F, Shapiro M, Waggoner P, Mitchell E, Thomson T, Nicklin S and Kacelnik A (2015) Performance decline by search dogs in repetitive tasks, and mitigation strategies. Applied Animal Behaviour Science. 166. 112-122.
  3. Williams M and Johnston J M (2002) Training and maintaining the performance of dogs (canis familiaris) on an increasing number of odor discriminations in a controlled setting. Applied Animal Behaviour Science. 78(1). 55-65.

False Indications, Clever Hans and You

In Berlin, during the late 1800’s, a horse named Hans started to demonstrate some amazing mathematical abilities. When asked a variety of computational questions, Hans would respond by stamping his hoof the correct number of times. According to his owner, Wilhelm von Osten, Hans could also tell the time, and successfully tackle more complex calculations (Jackson, 2005).

Initial investigations, by Professor Carl Stumpf, could find no evidence of cheating or trickery. Hans did, indeed, seem to be a very clever horse but not, perhaps, in the way that was first thought!

Hans with Osten
Hans and his owner, Wilhelm von Osten. (https://commons.wikimedia.org/wiki/File:Osten_und_Hans.jpg)

During 1907, further investigations, led by Professor Oscar Pfungst, reached the conclusion that, rather than being able to count, Hans was responding to the subtle postural and facial cues of his owner and other onlookers. Hans could only correctly answer questions that his questioner also knew the answers to. Additionally, Hans was unable to give a correct response if his questioner was hidden from view (Samhita and Gross, 2013).

What became known as the ‘Clever Hans Phenomenon’ or ‘Clever Hans Effect’ – “where a person or animal can be influenced by subtle and unintentional cueing on the part of a questioner” (Jackson, 2005) – led directly to the development of the “double blind” approach within scientific investigations; where neither the investigator nor the subject knows who is in the test or control groups.

So, what has the ‘Clever Hans Phenomenon’ got to do with you and your Scent Detecting dog? Well, just like Hans, your dog recognises and responds to the subtle (and not so subtle) cues delivered by those around him … including you, his handler!

This effect can be incredibly powerful and can influence your dogs actions in the most profound ways. In an interesting study by Szetei and Miklosi (2003), almost 50% of the study dogs would go to an empty bowl indicated to them by a person pointing to it rather than to a bowl in which the dog had previously seen and smelt food!

In another study by Lit, et al (2011), 18 drug / explosive detection dog and handler teams were given a number of detection tasks to complete. None of the tasks contained drug / explosive scent. In other words, they were all ‘blank’ searches. Any alerts / indications that the dogs gave would be incorrect, or, ‘false positive’ indications. Two of the detection tasks included a small piece of paper that falsely marked a scent location and two other detection tasks contained a decoy scent (toy / food).

In the detection tasks that included the false paper markers the handlers reported that their dogs alerted more at the marked locations than in other parts of the search area. According to Lit and her co-authors, when the handlers believed that there was scent present in the search area (paper markers) they were more inclined to claim that their dogs had indicated on a ‘target scent’; a ‘false positive’ indication. As Lit, et al (2011) state, this would seem to confirm that “handler beliefs affect outcomes of scent detection dog deployments”.

Interestingly, in the same study, when there were no paper markers present, there were more ‘correct’ (no indication) searches. It would seem then that we are as vulnerable to the ‘Clever Hans Effect’ as our dogs; taking cues from our environment and acting on our beliefs about these cues.

It’s worth noting here that, at the time of publication, the Lit, et al (2011) study received considerable criticism. If you’d like to know a little more about this debate, here are two interesting links to follow;

https://www.npr.org/2017/11/20/563889510/preventing-police-bias-when-handling-dogs-that-bite

http://swgdog.fiu.edu/news/2012/swgdog-response-to-lit-k9-study/swgdog_response_to_lit_study.pdf

So, how can you help guard against the ‘Clever Hans Effect’ in your own Scent Detecting work?

Practice more ‘Blind Searches’ – Find a training partner, someone who can hide your dog’s target scent for you. If you don’t know where it’s located you’ll be less likely to influence your dog’s searching behaviour. Remember, Hans could only ‘answer’ questions that his questioner could also answer. Your dog will now have to ‘answer’ the scent detecting question for himself.

Keep an OPEN mind – As Lit and her co-authors (2011) demonstrated, even a Blind Search doesn’t prevent you from starting to second-guess where the target scent might (or might not) be located. Something in the search area – for Lit, et al’s (2011) subjects it was a small piece of paper, for you it might be a chair that’s been moved – can cause you to encourage your dog  into, or away from, a certain area. Beware of any internal dialogue that tells you “He wouldn’t have put it there” because, guess what … he HAS put it there! I fell into this trap AGAIN just a few days ago. Believing I knew where the target scent wouldn’t be resulted in me directing my dog away from the correct search area. Ho hum. Time to follow my own advice!

pexels-photo-109998.jpeg

Keep an open mind

Allow your dog to work independently – I’m a big advocate of ‘Free Searching’. By this I mean letting your dog work away from you, choosing his own route around the search area without close-control and direction from you. Look out for a future Blog focusing specifically on this topic. Giving your dog space to work and, more importantly, resisting the temptation to start searching yourself – looking in boxes, opening cupboards, peering under furniture – helps alleviate any tendency that your dog might have to watch for cues from you rather than search the environment carefully himself. As the saying goes, “Why have a dog and bark yourself”? If you start to show interest in a particular part of the search area it’s highly likely that your dog will too. As soon as this happens, believing that your dog has located the target scent source, you’ll start to look harder yourself. A vicious circle with the almost inevitable result – a false indication from your dog prompted by inadvertent cues from you.

28643460_1537922602991797_1286099995_o

Poppy working independently

Consider your audience – Like Hans, your dog will respond to cues from any onlookers in the search area. Whenever possible, try to set up searches where your audience is also unaware of the location of the target scent. If they don’t know where it is they’re less able to inadvertently influence your dog’s search. Encouraging your audience to talk to one another can be helpful too. The more attention they’re giving one another during the search the less attention they’ll be paying to your dog. Less helpful is ‘audience silence’ and ‘breath-holding’ as your dog comes close to his target scent.

With “increasing evidence for an acute sensitivity to human gestures and attentional states” in our dogs (Udell, et al 2010), guarding against the Clever Hans Phenomenon can be problematic but an essential consideration in the development of a reliable scent detecting dog.

© Scent : Detect : Find Ltd 2018

References / Further Reading

  1. Jackson J (2005) The Clever Hans effect – a horse’s tale. Critical Thinking. http://www.critical-thinking.org.uk/pdf/clever-hans.pdf
  2. Lit L, Schweitzer J B and Oberbauer A M (2011) Handler beliefs affect scent detection dog outcomes. Animal Cognition. 14: 387-394
  3. Samhita L and Gross H J (2013) The “Clever Hans Phenomenon” revisited. Communicative and Integrative Biology. 6:6. November – December
  4. Szeiti V and Miklosi A (2003) When dogs seem to lose their nose: an investigation on the use of visual and olfactory cues in communicative context between dog and owner. Applied Animal Behaviour Science. Sept 5. Vol 83. Issue 2. 141-152
  5. Udell M A R, Dorey N R and Wynne C D L (2010) What did domestication do to dogs? A new account of dogs’ sensitivity to human actions. Biological Reviews. 2010. 85. 327-345