Enhance service canine human scent matching/nonmatching task acquisition, accuracy, and reliability for investigative and forensic operations.
Human scent matching/nonmatching service canines can be used during both lineup detection and missing person mantrailing operations. Their task is to choose from among novel alternatives the novel individual unique human scent information that matches the individual unique information on a previously presented scent sample or indicate the absence of a matching scent. Human scent matching/nonmatching mantrailing canines have an additional task of following a missing person’s scent along their line of travel while discriminating between nonmatching alternatives. They can be used when simpler solutions (other than human scent matching/nonmatching) are not possible to solve the problem. Such as, when the missing person is a stranger to the dog (has not been used during any prior training or testing) and the search area has been contaminated with other human scents from strangers, when the dog cannot be put on a specific line of travel but instead must smell the odor on a scent sample and choose from among alternatives a matching human scent trail or indicate its absence, when a number of possible locations must be searched, and in order to narrow the search area, cover long distances, and shorten search time when a matching trail has been found, scent matching/nonmatching service canines can be used to hop-scotch ahead via a squad car to other probable locations. Alternatively, lineup detection human scent matching/nonmatching service canines can be used to indicate whether or not the human scent on an item found at a crime scene matches the scent of a suspected perpetrator.
Scent matching (matching-to-sample, MTS) is a more complex task than simple discrimination between the presence and absence of an odor, such as a single track, or simple discrimination between similar odors, such as discrimination between two or more human scents. It is a conditional discrimination task involving a conditional cue that must be used in order to solve matching problems from one trial to the next; otherwise it would not be conditional discrimination. In conditional MTS discrimination procedures, a cue is required to signal which alternative is correct on a given trial and as such, constraints are incorporated within the procedure to control against alternative solutions that do not involve use of sample stimuli to solve the problem. Human scent matching service canines must not only discriminate between two or more human scent choice alternatives (scent sample comparisons), moreover they must learn to use the scent sample to signal which alternative is correct.
In addition, olfactory MTS tasks can be solved either by learning associative stimulus-response chains or by relational solution strategies.*1 However, MTS associative stimulus-response chain solution strategies cannot transcend training stimuli (see e.g. Carter & Werner, 1978). Alternatively, service canines that have learned a relational MTS solution strategy can reliably choose from among all stranger scented alternatives the comparison that matches the previously presented scent sample. Thus, there can be a significant difference between learning to solve the MTS task and learning a general solution strategy that can transcend training stimuli. Because during real world operations human scents are from strangers, involving no prior reinforcement history, human scent matching service canines must learn a relational solution strategy. In order to control against MTS associative stimulus-response chain solution strategies, service canines must be MTS tested with all novel human scents prior to operations and test results should be limited to first trial performance.
Furthermore, in order to avoid response to a nonmatching scent on nonmatching trials in which there is no matching scent present, such as when a suspect is innocent or when there is no matching trail present in a given search area, human scent matching service canines must not only learn a matching relational solution strategy, they must also learn a nonmatching relational solution strategy. To reliably solve both matching and nonmatching problems, they must learn two response rules; such as, “choose the comparison that is the same-as the sample, but if all the comparisons are different-from the sample, respond in such a way to indicate they are all different”. Yet, when dogs are only trained on matching problems, novel human scent MTS only problems can be reliably solved with a single solution strategy. The simpler solution involves discrimination between more and less familiar novel choice alternatives in which the novel matching human scent became familiar during the scent sampling process. When dogs are trained and tested on MTS only problems, with no nonmatching trials, they do not need to learn two rules to solve the problem; that is, one rule to respond to the matching alternative and another rule to suppress response to nonmatching alternatives. A stronger disposition to respond to a matching alternative than to a nonmatching alternative is sufficient to solve MTS only problems (Premack, 1978). However, a single MTS solution strategy does not enable dogs to perform accurately and reliably on nonmatching trials (see e.g. Schoon, 1996). The only way in which service canines can reliably solve both matching and nonmatching novel human scent problems during investigative and forensic operations is to learn two rules that can transcend training stimuli; “choose the comparison that is the same-as the sample, but if all the comparisons are different-from the sample, respond in such a way to indicate they are all different”. Same-as and different-from are concepts.*2 Concept learning involves higher-order cognitive processes. Not only is conditional discrimination a more complex task than simple discrimination and not only can there be a significant difference between learning to solve the MTS task and learning a general solution strategy that can transcend training stimuli, the same-as/different-from relational solution strategy service canines must learn in order to be reliable during both matching and nonmatching operations is more complex, involving higher-order cognitive processes. Because novel human scent matching only problems can be reliably solved with a single solution strategy, prior to operations service canines should be tested with both novel human scent matching and nonmatching test trials to determine whether they can accurately and reliably choose from among novel comparisons the novel individual unique human scent information that matches the individual unique information on a previously presented scent sample or indicate the absence of a matching scent. See https://k9scentdiscrimination.com/canine-human-scent-matchingnonmatching-conditional-discrimination/ for service canine human scent matching/nonmatching test criterion.
Matching-to-sample (MTS) is a procedure comparative learning and cognition researchers use to study complex learning in animals. Thus, the procedure has been extensively investigated. Prior to MTS training, subjects are first trained to make a particular response or responses. Since this phase does not involve MTS training in which a cue is required to solve the problem, subjects are not presented a sample stimulus at the start of each training trial; they are only required to learn responses that will be pertinent during subsequent MTS training. After response training, subjects are trained MTS. During MTS training, subjects are first exposed to a single sample stimulus. However, in order to control against a systematic simple discrimination solution that does not involve using the sample stimulus as a cue to signal which comparison is correct on a given trial, the sample stimuli used during training are varied randomly over trials. That is, the sample stimulus presented in the previous trial is not the same as the sample stimulus presented in the next trial and over trials the sample stimuli used during training are not presented in a systematic or predictable order. After response to the sample stimulus, it is removed and two or more choice alternatives (comparisons) are presented, one of which is the same as the sample.*3 The subject’s task in MTS is to choose from among alternatives the comparison that matches the sample. The reinforcement contingency is, if the subject chooses the comparison that matches the sample the subject is rewarded, but if the subject chooses a comparison that does not match the sample the trial is terminated without reward. In addition to randomly varying sample stimuli over trials, the comparison stimuli and the presentation positions of the comparison stimuli relative to one another are also randomly varied over trials in order to control against subjects learning some systematic presentation pattern discrimination rather than learning about the matching relationship holding over trials between sample stimuli and matching comparisons.
Matching-to-sample is a conditional discrimination task that necessarily involves constraints within the procedure to ensure subjects use the sample stimulus as a conditional cue to signal which choice alternative is correct from one trial to the next. Without the constraints, the procedure would not be a conditional discrimination procedure; subjects could learn from prior reinforcement history a simple discrimination solution strategy rather than conditional discrimination in which subjects must use the sample stimulus to solve the problem. Although the MTS procedure has been extensively investigated by highly competent objective animal learning researchers who understand the importance of the constraints within the procedure, human scent matching training protocols typically involve pseudo conditional discrimination procedures; that is, pseudo-matching-to-sample. Pseudo-MTS involves scent sample presentation at the start of every trial but omitted are the MTS constraints meant to control against alternative simple discrimination solutions. Therefore, over pseudo-MTS training dogs do not have to use the scent sample as a conditional cue to signal which choice alternative is correct from one trial to the next. Once a simple discrimination solution not controlled against has been learned, scent sample presentation at the start of every trial is uninformative. In pseudo-MTS arrangements, the scent sample signals nothing that is not already perfectly predicted from prior reinforcement history. Thus, over pseudo-MTS training dogs can learn to ignore scent sample information, which can retard subsequent MTS training in which they must learn about the individual unique matching relationship holding over trials between scent samples and matching comparisons in order to respond accurately and reliably during operations.
For example, prior to MTS lineup detection training, some pseudo-MTS training stages involve stacked trials in which the scent sample and matching alternative is from only one person over trials. After presentation of the scent sample, lineups consist of that person’s scent presented among unscented blank alternatives. The reinforcement contingency is to respond to the only human scented alternative in the lineup, rather than to sample to odor on the scent sample and then choose from among alternatives the comparison that matches the sample. Considering that simple discrimination is learned significantly faster (within a few trials) than conditional discrimination (involving hundreds of trials), it’s reasonable to assume dogs simply learn to discriminate between the presence and absence of scented items, which is a simple discrimination solution strategy, rather than learning to choose the human scent that matches the sample, which is a conditional discrimination solution strategy. Other pseudo-MTS stages involve presentation of the same person’s scent used during all prior pseudo-MTS training, except in place of one or all of the blanks, decoy human scents are presented. Thus, the lineup consists of one positive (S+) human scent alternative, which predicts reinforcement from prior reinforcement history, and negative (S-) human scented alternatives. Although dogs must now discriminate between human scents, the simplest solution still only involves simple discrimination between S+ and S- as opposed to using the scent sample to signal which discriminative stimulus is correct from one trial to the next. Finally, subsequent lineup detection training protocols may involve MTS or a mix of pseudo-MTS and MTS conditional discrimination in which dogs can learn to solve scent-matching problems from the MTS arrangement (for some lineup detection training protocols see e.g. Kaldenbach, 1998; Schoon & Haak, 2002). However, if the function of service canine human scent matching training is to enable dogs to detect the individual unique matching relationship holding over trials between scent samples and matching comparisons, it seems counter productive to initially train with pseudo-MTS arrangements in which dogs can learn to ignore scent sample information. Once dogs learn to ignore scent sample information during pseudo-MTS training, acquisition of subsequent MTS will be retarded compared to dogs that have not learned to ignore scent sample information prior to MTS training.
Although more research is needed, it is proposed here that not only does prior pseudo-MTS training fail to enhance subsequent MTS acquisition, with enough pseudo-MTS training it can retard subsequent MTS acquisition compared to control groups not initially trained with pseudo-MTS arrangements prior to MTS. Such experimental testing would involve an experimental group and control group of dogs. Alternatively, another test that does not necessarily require experimental and control groups, involves recording changes in the response elicited during scent sample presentation over the course of pseudo-MST training. If novice dogs initially attend to and sniff the scent sample, but over the course of pseudo-MST training come to turn away from the scent sample when it is presented, the change in response to the scent sample is a measure that pseudo-MTS arrangements produce dogs that learn to ignore the scent sample, which can retard subsequent MTS acquisition. Conversely, acquisition about the individual unique matching relationship holding over trials between scent samples and matching comparisons can be, in part, enhanced during MTS training by increasing attention and associability of the individual unique component of human scent prior to MTS training, in the absence of pseudo-MTS arrangements.
Mantrailing is not exempt from pseudo-matching-to-sample training arrangements in which dogs can learn to ignore scent sample presentation over training. Tracking and mantrailing service canines can be divided into two categories; tracking and trailing canines that do not need to scent match and mantrailing canines that do need to scent match in order to solve the problem. Non scent matching tracking and trailing police service canines can solve a host of real world working problems by (1) discriminating between the presence and absence of a track or trail when there are no other competing tracks or trails present. They can also (2) discriminate between a fresher target track/trail and older competing tracks or trails. A third solution that enables police service canines to follow a stranger laid track under more contaminated circumstances involves discrimination between the saliency of a target track and competing cross tracks that differ in saliency regardless of whether cross tracks are fresher, provided dogs have enough opportunity to become familiar with the target track prior to encountering competing tracks (see e.g. Most, 1954). There is also a fourth non scent matching solution strategy that enables dogs to very accurately and reliably follow a target trail laid by someone with prior reinforcement history over hard surfaces in contaminated urban environments when competing stranger laid decoy trails are fresher, the same age, or older than the target trail. However, this non scent matching solution strategy is training stimulus bound and does not enable dogs to solve stranger laid trailing problems in contaminated urban environments involving fresher or same age competing trails. The solution involves simple discrimination between scent with prior reinforcement history (that has come to predict reinforcement from prior stimulus-reinforcement pairing) and competing decoys that have no prior reinforcement history (that do not signal reinforcement). Because this solution strategy is training stimulus bound, requiring prior reinforcement history, during real world urban environment operations involving all strangers laid trails with no prior reinforcement history, dogs that have simply learned to follow the scent that has come to predict reinforcement from prior reinforcement history cannot accurately and reliably single out a target stranger laid trail from among competing decoy stranger laid trails of the same age because there is no prior reinforcement history under those circumstances.
Notice, if a scent sample is presented prior to working the types of problems that can be solved without scent matching, the arrangement is a pseudo-MTS arrangement and accurate performance could make it appear as though dogs have learned to scent match when they have not. In order to determine if dogs have learned to scent match, alternative solutions need to be controlled against. Yet, appropriate controls during training and testing are too often overlooked. For example, FBI canine program manager Stockham reported (trailing email list on bloodhounds.net, March 28, 2005) that two Law Enforcement Bloodhound Association (LEBA) instructors were shocked at the test results from an appropriately controlled scent matching/nonmatching trailing test Stockham recommended. When the choice alternatives consisted of three stranger laid blind trails (two older trails laid at the same time plus one fresher decoy trail) most of the supposed scent matching dogs tested did not match the smell from the scent pad to the matching trail or indicate its absence, but instead incorrectly chose the fresher trail that did not match the odor on the scent pad. Although a scent sample was used during training (and possibly during prior inappropriately controlled testing), rather than leaning about the individual unique matching relationship between the scent sample and matching alternative, the dogs learned a simpler solution not controlled against during pseudo-MTS training, which was to follow the freshest track.
Stockham’s matching/nonmatching test involved four unfamiliar volunteers, A, B, C, and D. Scent pads were collected from A, B, and D and stored separately [in glass jars; not plastic bags]. A scent pad was not collected from volunteer C because C was the decoy who laid the freshest trail. From a start location, volunteers A and B walked side by side for approximately 100 yards at which point A and B turned left and right, walking away from each other another 75 to 100 yards and then hide out of sight. 20 to 30 minutes later, C laid a third trail along a parallel line, starting from the same start location as A and B. However, at the juncture where A and B split and walked away from each other, C continued walking straight for another 75 to 100 yards and then hid out of sight. After the trails were laid, the dogs were scented with either A, B, or D scent pads at random (for the current test, see Scientific Working Dog Group and Orthogonal Detector Guidelines http://swgdog.fiu.edu/, approved guidelines, “SC9 Scent Dogs: Pre-Scented Canine Aged Trail”).
This site is a work in progress. Not all chapters have been posted yet and posted material is subject to change. However, in an effort to enhance acquisition, accuracy, and reliability of human scent matching service canines, and perhaps prompt further research, the material to be reviewed here regards how animals come to learn about the predictive relationships between events, stimulus selection, various types of discrimination procedures, and the early history of police tracking in which it was widely believed police tracking dogs readily learned to scent match when alternate solutions were not controlled against. Another chapter will review how dogs come to attend to and discriminate between the individual unique components of human scent at the expense of other stimulus information, such as gender, perfumes, medications, drugs and so on. Pseudo-MTS arrangements involving the various solutions strategies dogs can learn will also be discussed. Finally, the various ways in which comparative psychologists have experimentally found to enhance same-as/different-from relational solution strategies will be reviewed with regard to their application to human scent matching/nonmatching service canine training.
Considered effort has been made not to mix scientific evidence with personal opinions that have no scientific basis and whenever possible the information reviewed is referenced. Readers are encouraged to read the referenced material and contact me with any further questions or comments. I am happy to share information and thoughts that might promote the advancement of human scent matching dogs.
*1. The term strategy does not mean subjects contemplatively choose a strategy. When different solutions can be learned, the term strategy refers to what is learned or how the problem is solved. Although certain species and individuals may be predisposed to learning a particular strategy, it is without question that specific constraints imposed during training to control against alternative solutions can significantly affect which strategy is learned.
*2. Operationally speaking, learning about the matching relationship holding over training trials between the scent sample and matching comparison on matching trials and the nonmatching relationship holding over training trials between the scent sample and comparisons on nonmatching trials provides a better account of matching/nonmatching problem solving than does an appeal to the notion of same-as different-from concept formation. Although we can maintain, human scent matching/nonmatching service canines must learn domain specific, individual unique human scent, same-as/different-from concepts in order to be reliable during human scent matching/nonmatching operations, for training and testing purposes what human scent matching/nonmatching service canines must learn is better understood from an operational perspective.
*3. Choice alternatives are termed comparison stimuli by cognition researchers interested in the ability of various species to form or use same-as abstract concepts because subjects must compare the choice alternatives with the memory of the previously presented sample stimulus in order to determine the matching comparison is the same as the sample.
Carter, D. E. and Werner, T. J. (1978). Complex learning and information processing by pigeons: a critical analysis. Journal of Experimental Analysis of Behavior, 29, 565-601.
Kaldenbacj, J. (1998). K9 Scent Detection: My Favorite Judge Lives in a Kennel. Detselig Enterprises Ltd., Calgary, Alberta, Canada.
Most, K. (1954). Training Dogs: A Manual. London: Popular Dogs Publishing Company.
Premack, D. (1978). On the abstractness of human concepts: why it would be difficult to talk to a pigeon. In S. H. Hulse, H. Fowler, and W. K. Honig (Eds.), Cognitive Processes in Animal Behavior, (423-451), Hillsdale, NJ: Erlbuam.
Schoon, G. A. A. (1996). Scent identification line-ups by dogs. (Canis familiaris) trained in a tube-retrieving method: Experimental design and forensic application. Applied Animal Behaviour Science, 49, 257-267.
Schoon, G. A. A. and Haak, R. (2002). K9 Suspect Discrimination: Training and Practicing Scent Identification Line-ups. Detselig Enterprises Ltd., Calgary, Alberta, Canada.