Sunday, July 19, 2009


A couple of days into the search, things were not looking so good for our subject.

For one thing, there was his physical condition. Diabetes, a kidney transplant, and what the family described as “near-blindness” all argued against him getting very far or surviving very long from where and when he’d mired his quad [1].

For another, we had the results of our efforts to date — or lack of results. We’d plastered the immediate area around that damned quad, with several dog teams and grid teams sequentially covering the same ground. I believe I took one of the last of these tasks — I think it was a hasty that walked us past the ATV, though it’s been a while and may just have been one of those “start at the quad and spiral outward” deals. On this task I was the sidekick for my first SAR partner and ever-reliable anchor, Lilly, and a Civil Air Patrol cadet who’d already demonstrated himself to be a sharp navigator and all-around handy fellow [2].

Lilly showed some interest immediately upon encountering the ATV — but that didn’t mean much to me, as so many people had already covered the area, and Lilly was not “scent-discriminating,” identifying the scent of a specific person. She searched for all human scent. I was marginally more interested when she put her nose down, trailing along the dirt road into and through a clearing up ahead, where the jeep trail ended in a T intersection.

Again, I wasn’t sure what to make of it — so many people had been through the area. Still, it wasn’t impossible that it could be our guy Lilly was following, so I tried to keep an open mind.

Something caught my walker’s eye; he paced into the intersection, knelt down — and called me over.

Lying on the ground was a Virginia Slims butt.

Now, middle-aged, male, western Pennsylvania kidney recipients don’t smoke Slims …

“The daughter,” I said aloud.

We’d been less than thrilled by the family’s insistence on going out on their ATVs on a private search the previous night — both because untrained searchers can inadvertently destroy evidence and otherwise interfere with the organized search effort and, frankly, because the noise had kept us from sleeping well. In particular I’d been peeved by the incident command staff’s inability to put their foot down and stop the freelancing — though, with a heck of a lot more experience under my belt and without a young man’s impatience, today I understand why they just couldn’t do it.

Be that as it was, I realized right away how profound a scent-contamination problem we had surrounding that quad — and how unlikely it was that we had anything left to discover in the immediate area. I told the debriefing officer as much; while I have no doubt that my report didn’t carry any greater weight than anybody else’s, it is a fact that, soon after, the plans section folks started writing up tasks farther away from the ATV — on the third of these, a grid team found this nearly crippled, diabetic, blind subject over a mile away, awaiting help next to the creek that had kept him hydrated while we were looking for him [3].

The guy walked himself out.

When I tell this story — as, in fact, with a lot of SAR war stories having to do with contaminating scent — I know that it automatically begs a question: if it’s possible to ask a dog to search for a specific person’s scent, why would you not do that every time?

There are a few reasons, actually. For one thing, uncontaminated “scent articles” — things that the subject has recently worn, carried, sat upon, touched, etc. — are harder to come up with than you may think. Generally everybody in the family has handled it before you get hold of it — remember, the daughter’s contaminating scent was an issue here. And that’s if you’re lucky, and the entire police and fire departments haven’t passed it among one another. Good scent-discriminating-dog handlers have some tricks, and some training background, that helps them deal with a small amount of contamination (at least, if they know about it); but you get to a point where you’re essentially searching for all human scent anyway.

For another reason, an airscent dog doesn’t need to scent discriminate. If you’re following a scent trail on the ground, you’re inevitably going to encounter other trails, and you need to stay on the right one. But if you’re searching an area in the hope of encountering the wind-borne scent coming off a person at that moment, you can just follow the scent to its source and, if it isn’t your guy, you’ve at least got someone you can ask: “Have you seen him?” Usually, all you’ve lost is a few minutes [4].

Ultimately, though, there’s an even better reason for us air-scent handlers to engage in discrimination cautiously, if at all: it’s flippin’ hard to do right, and if you screw it up it’s worse than useless.

Without taking away the slightest from what trailing handlers do, the tactical, navigational, and team leadership demands on an air-scent handler require a pretty serious, ongoing time commitment to acquire and maintain; the challenges of working with and communicating with a dog who’s working at liberty aren’t trivial either [5]. By no means is it impossible to do this alongside training for scent discrimination. But it ain’t easy, and again, if you don’t do it at a very high level of accuracy indeed, you’re by definition sucking at it.

Enter item number one, shared by th’ wife: a lawsuit against the only dog handler in Texas who uses scent lineups to identify crime suspects. Seems that at least two of the folks his bloodies fingered — so to speak — turned out to be innocent, and now they’re out for, um, blood.

Now, I don’t know this guy or his training methods. His dogs may be 99 percent accurate — but that still leaves one suspect out of a hundred who’s screwed, if the courts accept the dog’s nose as evidence.

More to the point, I doubt that anybody has put a numerator to his denominator and found out, for training or in the field, what his accuracy actually is [6]. And that’s a big part of the problem. Get this, from a lawyer with the Innocence Project of Texas:

“This is junk science. This isn’t even science. This is just junk.”

Now, in my book the Innocence Project folks are among the Good Guys. It hurts to hear one of them say this; it hurts even worse that, by the current standards of practice, he’s probably right.

The problem is that it’s so hard to work with a dog without cueing her as to what you want her to do. For us airscent dog handlers, it’s relatively simple: just don’t tell me where the training subject is hidden; either I find him or I don’t, and in theory at least it should be simple to demonstrate whether having the dog with me increases my search effectiveness or not [7].

For scent discrimination the task is harder. You can do a certain amount of screwing up without messing up a non-discriminating air-scent dog’s training; but every time you reward a discriminating dog, you train her; every time she self-rewards without being corrected for a faulty identification, you train her. Most discriminating handlers, therefore, spend most of their training time unblinded; they know what the answer is, so they can jump in and reward the right and correct the wrong.

The problem is, sometimes, without realizing, we jump in before the dog has actually committed to a choice, even by a fraction of a second: and thus we train the dog to tell us what we want or expect rather than the real answer. Good discriminating handlers proof themselves periodically with a blinded problem; but I worry that a lot of practicing handlers haven’t done that, or if they have, haven’t kept track of how often they mess up.

What we want and expect matters to our dogs. It seems undeniable to this working handler that our dogs read us far more effectively than we will ever be able to read them. The slave knows the master better than the master knows himself, as Heather is fond of abstracting the otherwise inabstractable Hegel.

The end result is that, even among the best scent-discrimination handlers, I don’t know how well established the standards of practice are for high-stakes lineup identification — not that you’re not doing well, just that you don’t have a handle on the actual number, and so can’t be sure.

Perversely, as we see from a review in Science, when the stakes are highest is exactly when we’ll do what we don’t want to. A phenomenon called the “ironic monitoring process” stems from the fact that intending not to do something requires us, at some level, to focus on it — which in turn makes us more likely to do it.

People do have the ability not to “mention the war:”

Our brains can, in fact, achieve the tricky balance of concentrating just enough on something to avoid it without concentrating enough to dive straight into it. But it’s difficult, and if we’re tasked with other stuff — whether actual distractions or the distraction of stress — we’re more likely to blurt out something about Hitler.

Particularly interesting in this context is a task in which one person who can see four shapes — say, a small triangle, a circle, a heart, and a large triangle — is asked to point out the small triangle to someone who can’t see the large one. It’s not at all uncommon for the first person to refer to the target shape in a way that tips off what the hidden one is: for example, by saying “look at the  small triangle.” And they’re more likely to do that if you tell them to keep the large triangle a secret than if you don’t.

There is, for us dog handlers, another way: in the early days, defense attorneys were more than leery about DNA forensics. They anticipated, quite correctly, that if we didn’t hold the field to high standards, it would result in bullshit convictions. But the guys in the DNA forensics field did their homework, refined their numbers, and today I think that in general defense lawyers love DNA testing: in the cases for which it’s relevant and doable, and when the meaning of its results are accurately reported to the jury, it is the quickest, easiest way to get innocent defendants back to their lives.

We can do this in the dog community. We can do our homework, put numbers to our abilities, and find out exactly what we can and can’t do reliably; and at that point, a jury can take our evidence for exactly what it is.

A few years back I did an article for Advanced Rescue Technology about SAR dogs and olfaction. I had the opportunity to interview Larry Myers, one of the few scent researchers who’s consistently interacted with working dog handlers, for that piece, and along the way we got into a discussion about the reliability of dog scent evidence in legal proceedings. The editor decided to let me do this as a sidebar to the scent article; my only regret is that I didn’t think of proposing it as a more in-depth, stand-alone article, but you can read it here.

The short version is that, for the level of certainty of probable cause — the gauge of reasonable suspicion a cop, once he’s stopped you for a traffic violation, needs to have in order to search your car for contraband — Larry thinks that dog evidence is OK; but for beyond reasonable doubt — what you need to convict someone of a crime — not so much. For what it’s worth, I tend to agree with him.

Nor should we feel like second-class citizens in being asked to do this. Reasonable doubts have begun to pop up with a number of common identification procedures, ranging from polygraphy to eyewitness identification [8] to, of all things, fingerprints. The common factor of these methods: they came into use before the courts started looking at the scientific bases behind methods, and so got enshrined based on “common sense” or, maybe even worse, the opinions of “experts,” itself a slippery term that I tend to try to avoid. Quantitative evidence is the wave of the future: we hold it at bay at the risk of bringing about our own irrelevance.

[1] Yes, it occurred to me too. But it’s not the strangest thing I’ve encountered regarding a search or a search subject.
[2] Briefly: on an earlier task the previous day — a hasty along the top of a ridge to the east of the ATV — a ranger, because of a miscommunication with the operations folks, left us off at a completely different location than our intended jump-off point. In those pre-GPS days, it amounted to leaving us in the middle of the woods without the slightest idea of where we were. Efforts to align our surroundings to the terrain proofed fruitless until my walker noticed that we were standing next to the confluence of two creeks — and our map only showed one such intersection. Once we knew our spot, it was simply a matter of shooting a bearing to where we’d intended to start, and then following through with the task. We never had to ask for help, and completed our assignment — took us a few extra hours, but we did it.
[3] If you don’t have anything to eat, you don’t need your insulin.
[4] At least once, I did have reason to believe we’d missed a subject on a real search because my dog had been distracted by finding someone else in my area. But that’s more an argument for not letting people — either civilians or other searchers — get into each other’s assigned areas than for scent-discrimination per se.
[5] For the record: as I train to become a trailing handler, I find the level of handler-dog communication necessary to work trailing to be daunting. But it’s a distinct and different type of communication from that of an airscent team.
[6] As some of you may know, we’re in the process of doing something much like this with our own dogs’ performance. The preliminary results are encouraging in terms of showing something objective; but it’s a long-term project, and I don’t know how long before we get results that I’d subject to peer review. Stay tuned.
[7] Again, stay tuned.
[8] Wikipedia also has an article that encompasses the “anti” argument fairly well, though it has been tagged for non-neutrality.

No comments: