The Reliability of Drug Sniffing Dogs

You may also like...

14 Responses

  1. JoeJP says:

    This seems like a good area for some experimentalization in the state courts using local constitutional standards when appropriate.

  2. Ken Rhodes says:

    You wrote “The use of a drug detection dog by itself is not considered a “search” that implicates Fourth Amendment protections.”

    If I were a defense attorney, I would be pushing very hard for the court to establish a presumption of 100% accuracy, with an assumption of NO false positives. Because in that case I think there would be an ironclad argument to overturn the principle in your sentence. The dog, in that case, would be just like a powerful x-ray machine that can see inside the trunk of your car without opening it. The contraband is not in public view, and the cops are using an instrument to conduct a search of your private space. They are only entitled to see the inside after they’ve obtained their warrant, not to use their “x-ray vision” to get the warrant.

  3. Steven Lubet says:

    This is a great post, but I wonder about your parenthetical concern: “(Complicating this issue is the fact that a dog may alert to the residual odor of a drug that is no longer present in the car, and may not even belong to the car’s owner. Should this be considered a false positive, since no contraband is presently in the car?)”

    How could that be anything other than a false positive? The dog alerts; no drugs are found. The possibility of residual odor must be purely speculative, and it could never be ruled out. Thus, the assumption would be self-fulfilling: the dog must have alerted on something. But that is the very proposition that has to be proved.

    In any event, the purpose of a dog inspection is to find existing drugs, not to sniff out vanished ones.

  4. Erica Goldberg says:

    Thanks for your comments.

    JoeJP- Do you mean local constitutional standards as in state constitutional law, or local experimentation with federal constitutional law?

    Ken Rhodes- I also think drug detection dogs perhaps should be considered a search, but they are not because they reveal only whether or not the car contains contraband, and the Court thinks that we have no legitimate expectation of privacy in contraband. The Court is going to have to revisit this underlying assumption when it hears argument in Florida v. Jardines- about whether dogs can sniff at your doorstep.

    Steven Lubet- Thanks, and agreed.

  5. steph says:

    Hey, it’s Steph! Totally never knew that dogs varied so much in terms of reliability–definitely something that courts should worry about. I worry, though, about your solution–that the Court decide upon a maximum false positive rate.

    NOTE I AM TOTALLY NOT a probably cause expert. We administrative law folks don’t often have to deal with that. :) But from the doctrinal level, I wonder how that fits in with the whole case and controversy requirement. By setting a threshold limit, wouldn’t the court be addressing a situation that isn’t really before it–like if the dogs in this particular case posed no concerns related to the Court’s preferred “threshold.” (See, e.g., the last para of Orin S. Kerr, The Limits of Fourth Amendment Injunctions, 7 J. Telecomm. & High Tech. L. 127 (2009), which I looked up because I wanted to see if my instincts were totally off, because my constitutional focus is generally kind of elsewhere.)

    My more legal policy concern is that if the Court could, constitutionally, establish some kind of numerical level and then did so, then we might have a situation where courts enshrine into law a threshold hold that we might later–because of developing understandings or developing science–want to change, and then it would be much harder to do so than if the court articulated some kind of more values-based standard and then researchers could flesh out those standards. I mean, I totally get why it’s frustrating because then this means stuff needs to get hashed out on a case-by-case basis each time, but is that worse than if a court set a bad and hard-to-change standard that we might in a few years want changed but are unable to do so?

    I’m totally open to persuasion otherwise, just that these were the things that jumped out to me (because there similar concerns –(a)separation of powers, and (b) institutional capacity and expertise *are* raised in contexts with which I’m more familiar.

  6. Orin Kerr says:

    Erica,

    I’m curious if you have thoughts on my contrary argument, that it is important to keep probable cause from being quantified. The article is Why Courts Should Not Quantify Probable Cause, and the abstract is here:

    Probable cause is one of the fundamental concepts of Fourth Amendment law, but the Supreme Court has refused to quantify it. The Court has described probable cause as a “fair probability,” but it has declined to explain just how likely a “fair” probability might be. Does a “fair probability” mean a 50% likelihood? A 40% likelihood? And why won’t the Justices say? Are they just afraid of math?

    This essay argues that courts should not quantify probable cause because quantification would produce less accurate probable cause determinations. The core problem is that information critical to probable cause is often left out of affidavits in support of warrants: Although affidavits say what techniques police tried that added to cause, they generally leave out both what the police tried that did not add to cause and what techniques the police never tried. Determining probable cause accurately often requires this information, however. By leaving probable cause unquantified, current law enables judges to use their intuition and situation-sense to recognize when missing information is likely important to assessing probable cause. Quantification would lead to less accurate probable cause determinations by disabling those intuitions, creating the false impression that the information provided in the affidavit is the only relevant information. Cognitive biases such as the representativeness heuristic and anchoring effects would allow the government to create the false impression that a low-probability event was actually a high-probability event. To ensure accurate probable cause determinations, then, probable cause should remain unquantified. The result is counter-intuitive but true: Knowing less about probable cause improves how the standard is applied.

    As to your statement that the dog results give us a numerical probability, I disagree: The numerical probability always depends on the base rate, and we can’t tell what the base rate is in any given use of the dog. Richard Myers had a nice article on this in the George Mason Law Review a few years ago.

  7. JoeJP says:

    “Do you mean local constitutional standards as in state constitutional law, or local experimentation with federal constitutional law?”

    The former but if the latter was possible, that would be intriguing.

  8. Erica Goldberg says:

    Responses to Steph and Orin Kerr, whose perspectives I greatly appreciate and have given me pause.

    With respect to the Article III issue, I think the Court can give guidance on what maximum false positive rate renders a dog unreliable in the process of ruling that false positive rates must be admitted into evidence as part of the probable cause inquiry. (Some courts focus on accuracy rates, but it seems to me that only false positive rates are relevant.) Technically, this guidance would be dicta, but in the Crim Pro context, the Court has certainly reached farther than this in illuminating issues not strictly before it.

    With respect to the merits, I think there is a way to add numerical precision to this inquiry without overly mathifying it. I’d personally like the Court to explicitly hold that probable cause is like the summary judgment standard, which is whether a reasonable juror could rule in favor of the plaintiff by a preponderance of the evidence. Unlike the preponderance of the evidence standard, probable cause remains undefined—so we don’t know whether the reasonable officer would have to believe there is an 80%, 51%, or even 30% likelihood that a suspect committed a crime in order for there to be probable cause. I’d like the Court to explicitly define this benchmark as at least greater than 50%, but perhaps higher, so that the Court’s fair probability assessment is not too deferential to the police. So, the standard would have some squishiness in it, as summary judgment does, but at least we’d have some benchmark to judge when an officer is reasonable.

    If we could precisely determine the false positive rate of a particular dog, the Court could say that, given that the dog has a 45% false positive rate, a reasonable officer could not have thought it was more likely than not that there were drugs in the car. The Court need not set a minimum false positive rate, because Florida established a totality of the circumstances analysis, but could set a ceiling on how unreliable these dogs can be and still yield probable cause. This might ease Steph’s concerns that the law would not allow for malleability and evolution. If you check out the opposition to certiorari, you’ll see that courts already have to contend with whether certain false positive rates are sufficient to establish probable case.

    Professor Kerr, I completely agree about numbers creating cognitive biases that obscure deeper truths/intuitions. That’s why I think, if a false positive assessment can be made, the drug detection dog context is the perfect place to start quantifying probable cause. Other solutions, however, are to require police to be more detailed in their affidavits, and to still use totality of the circumstances so that judges cannot simply base their decisions on numbers. I worry that too much fuzziness in the standard not only allows judges to use their intuitions, but allows them to manipulate the standard and enshrine their own personal prejudices. I think there is some legal legitimacy to be gained by forcing judges and the police to be more transparent about their reasoning.

  9. Orin Kerr says:

    Thanks for the response, Erica. I hope you won’t mind me pressing you a bit more.

    First, how can you create a false positive assessment without knowing the base rate of how often drugs are present? The dog is only brought out in some circumstances: I would think that the false positive rate depends on the nature of those circumstances.
    See generally http://tinyurl.com/82bs43z

    Second, a recent study about how handler perceptions influence dog alerts raises difficult questions about whether a dog alert is proof that the dog smells drugs or that the dog thinks the handler wants the dog to alert:
    http://www.npca.net/Files/SWGDOG/LIT%20Study.pdf

  10. Erica Goldberg says:

    Orin Kerr,

    Please, press on.

    I do see Myers’ problem. The probability of an alert, given innocence, is not the same as the probability of innocence, given an alert. Because we cannot know when a dog does not alert and there are no drugs in the vehicle (a true negative), we cannot accurately measure the false positive rate. What if false positives had to be measured in simulated situations in training, and the Court established a uniform standard and maximum false positive rate before a dog could graduate from a training program?

    The second study, if true, would require the Court to really consider whether the use of drug sniffing dogs can ever give rise to probable cause. This gets into the difficult questions of institutional competency raised by Steph, and, as far as I know, no one on the Court is a statistician/ social scientist. Yet, it is the Court’s responsibility to safeguard the Constitution and decide whether these dogs’ alerts can give rise to probable cause. I don’t think the study militates against establishing a maximum (and only a maximum) false positive error rate. It simply sets a floor on reliability before a dog alert can establish probable cause. Sure, we can use the totality of the circumstances, but if we can never trust an alert, then why use dogs at all? The alert is not more accurate because there are circumstances that corroborate the police officer’s suspicion.

    Another solution, although perhaps too provocative, is simply to say that the use of drug sniffing dogs DOES constitute a search. This would require a lot more dismantling of the jurisprudence but is, in some ways, cleaner. I think most people find the drug sniffs intrusive, and the fear of a false positive endangers legitimate expectations of privacy in those not carrying contraband (assuming those are the only individuals entitled to a legitimate expectation of privacy) in the same way that checkpoints do. At the airport, the drug sniffs would be reasonable because we give up so much privacy.

    Prior to the Florida Supreme Court’s ruling that false positive rates should be examined, lower state courts in Florida were accepting proof of training to establish the reliability of a drug detection dog, even when training programs vary widely. This seems like unacceptable deference to me.

  11. steph tai says:

    Sorry for starting this conversation and then bailing for a bit — oh, the trials of sharing a household with someone trying to run four fantasy baseball teams at once. Anyway, I’m so glad Orin Kerr joined in to discuss the Article III issue — so much better a person to bring that up than I.(*) It’s really great to see both of you go back and forth on this.

    As for the substantive part, I do like the idea of at least defining probable cause according to other standards — that, to me, would not raise the institutional competence problem so much (given that that’s well within the sort of stuff that courts do all of the time.) I might even be okay with a sort of factor-based ceiling on the unreliability of the dogs. But the closer the court gets to quantifying that ceiling, the more I start to worry about the leaking of this sort of approach into other contexts. I’m thinking of Commerce Clause issues (quantifying how much “substantial effect” there needs to be on interstate commerce), due process and punitive damages issues (quantifying the permissible size of punitive damages awards), etc. And to me it seems that the Court does seem to be inching towards (but never actually) quantifying those things, and–to the extent that it is doing that inching–I find that troublesome.

    And maybe to explain why, I’d have to get into some of my personal inclinations towards Constitutional law and informational development. To me it seems that by *not* setting quantitative limits (frustrating as it might be! I really do understand that!), we at least open room for dialogue between all of the different branches regarding the developing scientific support for evaluating each of the subfactors that go into those constitutional evaluations. Whereas setting a hard limit would not only raise the concerns of quenching malleability and evolution, but also reduce incentives for other parties (legislators, police officers, the defense bar, criminal justice scholars) to further develop that information. And I guess in more technical areas–which I see the evaluation of the reliability of drug sniffing dogs to be–I think that’s important.

    p.s. On a personal note, though I haven’t been blogging in awhile, it’s discussions like these that do make me more psyched about blogging. I think my first article written at UW was inspired by my stint as a guest blogger here.

    (*) And I’m also glad to see that even though I’m just an administrative person, I wasn’t totally off the mark in even raising those concerns!

  12. Paul Pietz says:

    Very good article, though my concern with this issue is the claim that when a dog alerts but there is no contraband then it must be from a residual scent. In fact, the officers in my State that run K9′s don’t consider these to be false positives, they assume that the dog is detecting a past odor, so they claim a near 100% reliability.

    Additionally, I fail to see how claiming that the dog must be detecting a residual odor benefits the State’s position. They are basically admitting that the dog may detect contraband that is no longer present, so the dog is not, as Justice Stevens put it, detecting the presence or absence of contraband.

    Finally, to truly determine if there is a search or not, there should be some scientific evidence that the odor the dog is smelling is actually outside of the vehicle. There are numerous references in cases to “free air” searches, but at a dogs sensitivity level is is possible he is smelling the air inside the car? I think there is an unfounded assumption that the scent is outside the vehicle. Courts have routinely upheld suppression of evidence obtained by technical means that invades beyond the threshold of entry (e.g. use thermal imaging to detect grow lights inside a house). A dog is just a organic machine, as are humans. The problem with organic machines is that they are fallible.

    Consistency is the true issue. Does a number really work here? Isn’t that just incentive to report numbers in a favorable way, explaining away false positives? And are everyone’s rights protected equally if the dog is having an off day one day, and a great day the next? His percentage may should some reliability, but what if you are the person who falls into that “false positive” percentage and are detained for 3 hours while they search and dismantle your vehicle (yes, this happens), and then find nothing? What is your remedy if you missed a flight, missed your daughter’s ballet recital, or got fired because you missed an important meeting? None. This is what the 4th Amendment was designed to prevent. Dogs, some with reliability below 30%, are actually worse at determining fair probability of the presence of contraband than flipping a coin.

    Of course none of this addresses the officers that use the dogs solely for the purpose of searching, claiming any number of different signs by the dog is an alert.

  13. Doug Cullen says:

    I think Paul’s last point is one that requires some oversight. Someone ought to be taking handler/dog teams (and both parties are part of the process) to task when they have a significant number of “positive” indications that yield no drugs. These teams need to be re-evaluated for their training and methodology. Best case is that poorly performing teams are the result of poor training/retraining regimes and that the supervising dept needs to make changes. Worst case is that some handlers are using dogs to validate search of suspects they “know” are handling drugs. Frankly, handler/dog teams with poor records should be dismissed as a team and perhaps retrained with different partners. Handlers that have repeated poor results should be removed from k9 service. This would create some accountability for officers who may be tempted to use K9 as a tool to “get to an arrest”

  14. Very interesting opinion, nice to read such interesting post on this site.