An assumption underlying the use of human scent matching service canines is that individuals have unique odor profiles that remain constant over time, which dogs can detect and discriminate between. Another assumption is that dogs can not only detect and discriminate between the individual unique odor profiles of human scent, they can also scent match individual unique odor profiles of human scent while discriminating between other nonmatching human scent alternatives during investigative and forensic operations and if there is no match, reliably indicate the absence of a matching scent. That is, sample novel individual unique human scent information on a scent sample and match it to a matching scent in a lineup or along a trail, involving more than one alternative, while discriminating between other nonmatching unfamiliar human scents, or indicate when there is no matching scent.
There is good evidence to support the assumption that individuals have unique odor profiles that remain constant over time, which dogs can detect and discriminate between (Kalmus, 1955; Nicolaides, 1974; Wallace, 1977; Green et al., 1984; Halpin, 1986; Ferstl et al., 1988; Hepper, 1988; Ferstl et al., 1992; Sommerville et al., 1993; Schoon & De Bruin, 1994; Settle et al., 1994; Singer et al., 1997; Eggert et al., 1999a; Eggert et al., 1999b; Singh, 2001; Harvey & Harvey, 2003; Zavazava et al., 2005; Boehm & Zufall, 2006; Harvey et al., 2006; Penn et al., 2007). However, human scent matching/nonmatching is a conditional discrimination task, involving more than simple discrimination between the presence and absence of human scent or simple discrimination between human scents. Conditional discrimination, involving a conditional cue (the scent sample) is a more complex task than simple discrimination. In conditional discrimination, a cue is required to signal which, if any, of the discriminative stimuli (choice alternatives or sample stimulus comparisons) is correct on a given trial. For example, a typical simple discrimination procedure involves exposing subjects to a positive (S+) and a negative stimulus (S-) in random presentation order over trials. Response to S+ is always followed with reinforcement, whereas response to S- is always followed with the omission of reinforcement. Typically, simple discrimination is learned very rapidly, sometime within just a few training trials (see Discrimination training in The History of Police Canine Tracking, https://k9scentdiscrimination.com/discrimination/ ). Alternatively, in the matching-to-sample (MTS) conditional discrimination procedure used to investigate complex learning in animals, a trial begins with the presentation of a sample stimulus, after which two or more comparison stimuli are presented; one is the same as the sample while the other differs from the sample. Over trials the sample stimuli, comparison stimuli, and presentation positions of comparisons are varied randomly to control against subjects learning a simple discrimination solution, as opposed to a conditional discrimination solution. The reinforcement contingency is; response to the stimulus that matches the sample is reinforced, whereas reinforcement is omitted after response to a nonmatching comparison. During MTS training, even in species that spontaneously learn a solution strategy that can transcend training stimuli, it typically takes hundreds or more trials before subjects learn to solve the MTS task. And dogs are no exception when training and testing involves appropriate constraints to control against alternative solutions that do not involve scent matching. Furthermore, human scent matching/nonmatching service canines have an additional task of learning to reliably indicate the absence of matching individual unique human scent information during randomly presented nonmatching trials. In order to respond accurately and reliably during both novel human scent matching and nonmatching trials, service canines must learn a rule to choose from among comparisons the individual unique human scent information that is the same as the individual unique human scent on the previously presented scent sample, but if all individual unique human scents are different from the individual unique human scent on the scent sample, respond in such a way to indicate they are all different. Service canines must learn a relational solution strategy involving same-as/different-from concepts, rather than associative solutions, in order to respond accurately and reliably to stranger’s scents during operations.
In addition to taking hundreds or more training trials to learn to solve the MTS task, depending on species learning bias or processing capacity and the arrangement of MTS training, animals can learn to solve MTS tasks associatively, relationally, or a more or less mix of both (and sometimes configurally*1). However, when associative MTS solutions are learned, subjects cannot match novel stimuli that have no features in common with training stimuli. Associative MTS solution strategies cannot transcend training stimuli. Thus, dogs that have learned an associative solution during training cannot accurately and reliably choose from among novel human scents the novel human scent that matches the scent sample or indicate its absence. Alternatively, an arguably more complex cognitive achievement involves learning to respond on the basis of a relationship holding over training trials between stimuli, such as learning that the individual unique human scent information on the matching comparison is the same as the individual unique human scent on the sample. The benefit of a relational human scent matching solution is that it releases subjects from direct control of stimulus-response-reinforcement history and permits flexible adaptive response when stimuli to be matched are novel. Thus, relational solution strategies do enable subjects to transcend training stimuli. If a relational MTS solution strategy has been learned, service canines can respond as accurately and reliably when all comparisons are from strangers as they can with training human scents.
Associative MTS solutions, which animals can more readily learn, involve learning multiple stimulus-response chains specific to each training sample (Skinner, 1950). For example, analogous to learning during training that “sit” signals when the response of sitting will be reinforced, over training sample stimulus A comes to signal when response to alternative A will be reinforced (A->A->R), sample B comes to signal when response to alternative B will be reinforced (B->B->R), and so on. Learned stimulus-response chains are equivalent to sample stimulus specific if-then rules (Carter & Werner, 1978); if sample A then respond to alternative A, if sample B then respond to alternative B, etc. However, associative MTS solution strategies are training stimulus bound.*2 In order to learn a set of associative stimulus-response chains/if-then rules, dogs must have prior learning experience with each specific sample and correct alternative reinforcement contingency. Each associative chain learned during training is specific to each sample stimulus and cannot transcend the training stimuli outside of common features training and test stimuli share. Thus, dogs who learn associative MTS solutions during training cannot reliably scent match during operations when scent is from strangers.
Alternatively, dogs that learn during training about the matching relationship between scent samples and matching comparisons (learn a relational solution strategy) can scent match individual unique human scent information during operations when the human scent to be matched is from strangers. Relational MTS solution strategies, such as a rule to choose the comparison that is the same as the sample, are acquired by comparing sample stimuli with choice alternatives (sample stimulus comparisons*3) and detecting that reinforcement is contingent upon choosing the comparison that is the same as the sample. Rather than associating the absolute physical properties of stimuli with reinforcement, dogs may compare the comparisons with the previously presented scent sample and detect that the individual unique information on the matching comparison is not only the same as the sample, but that over trials the individual unique information on the sample and matching comparison are always the same. They may learn that response to the comparison that is the same as the sample predicts reinforcement. That is, learn a general MTS rule to choose the comparison that is the same as the sample. However, in order to detect the same as relationship holding over trials and learn a general MTS rule, dogs must compare the matching stimulus information on a given training trial, detect the matching sample-comparison information is the same, and treat the relational concept ‘same as’, as a stimulus. Furthermore, they must not only compare the matching stimulus information on a given training trial, they must also compare trials with one another in order to detect the matching relationship holding over trials, which is why multiple training trials within a session are advisable. Moreover, classifying stimuli as being the same involves higher-order representation, one that transcends perceptual features of individual stimuli. Thus, same as relational MTS solution strategies are more complex, involving higher-order cognitive processes than simple associative solutions.
The advantage of general MTS rules is that they can transcend training stimuli, which enables subjects to respond as accurately to novel stimuli as they can to training stimuli with prior reinforcement history. Unless human scent matching service canines have learned a general MTS rule, they cannot reliably solve human scent matching problems with strangers.*4 However, although all mammals can readily learn associatively when associative solutions are possible, in many species relational MTS solutions are not as readily learned when circumstances also permit alternative associative solutions. For example, comparative cognition research shows that when animals are trained MTS with the standard MTS procedure, after conditioning some species (such as chimpanzees) show spontaneous transfer when subsequently tested with novel stimuli, whereas other species (such as pigeons) require more explicit training (modifications of the standard MTS procedure ) before they show good transfer to novel stimuli. Although humans and a few other species readily learn relational solutions when associative solutions are not controlled against, under the same circumstances other species do not readily learn relational solutions and tend to be more or less biased toward learning simpler associative solutions. Therefore, unless clear evidence is provided that a more complex cognitive process is involved in the solution learned, it must be supposed that a simpler process accounts for matching accuracy. When training stimuli (or stimuli with prior experience) are used to test and assess human scent matching reliability, due to the existence of simpler associative accounts, reports of high matching accuracy are not sufficient to conclude behavior is mediated by a general rule that can transcend training stimuli, enabling dogs to reliably solve matching problems with unfamiliar human scents during operations. To assess whether performance is independent of prior reinforcement history, accurate and reliable first trial MTS performance with novel or unfamiliar human scents is necessary.
Although, there is good evidence to show dogs can solve human scent matching tasks (Hepper, 1988; Schoon & De Bruin, 1994; Settle et al., 1994; Schoon, 1996; Schoon, 1997a; Schoon, 1998; Schoon, 2001; Schoon & Haak, 2002; Harvey et al., 2006), aside from a few studies (Schoon and De Bruin, 1994 and Schoon, 1996), scent matching with unfamiliar human scents has not been reported and there have been no studies explicitly reporting first trial novel human scent matching performance. Therefore, it is not clear whether accurate performance in the current studies was partly mediated by previously learned training stimulus specific solution strategies (prior reinforcement history) or by a general rule. Furthermore, service canine human scent matching certification tests often do not specify that all test stimuli must be novel and test results should be limited to first trial performance. Without appropriately controlled novel stimulus transfer tests, designed to test the processes involved in transfer, it is not possible to know whether accurate performance is mediated by a general rule that will enable service canines to transcend the training stimuli and respond accurately and reliably to unfamiliar human scents during operations.
Because service canines are required to scent match and discriminate between the scents of strangers during investigative and forensic operations, dogs should be transfer tested with human scents from strangers that have never been used during any prior training or testing to determine whether what they have learned during training transfers to new circumstances, that is, generalizes to novel human scents. The criterion for testing whether animals have learned a general rule that can transcend training stimuli has been outlined by numerous investigators (Premack, 1978; Cook, 2002; Katz, Wright, Bodily, 2007; Bodily, Katz, Wright, 2008). The criterion for transfer testing human scent matching/nonmatching service canines is summarized as follows:
- Dogs should be double-blind tested according to the random control matching-to-sample procedure used for scientific investigation so the alternative solutions are controlled against.
- All human scents during both matching and nonmatching transfer tests should be novel, not used during any prior training, testing, or familiarization processes. Thus, novel human scents should not be reused in subsequent test trials.
- Results of transfer tests should be limited to the first test trials.
- Transfer trials should not combine both familiar and novel human scents. All scent samples and both matching and nonmatching comparisons should be novel on all test trials because prior experience with the testing stimuli may confound the results. (However, sometimes tests are added, involving randomly presented nonmatching comparisons from a familiar person among novel matching and nonmatching comparisons, to expose an incorrect solution strategy to always choose the most familiar scent regardless of whether it matches the sample.)
- Matching and nonmatching comparisons should be collected at the same time and be scented for the same time period in order to control against simple discrimination between fresher and older scents. (However, sometimes tests are added, involving randomly presented fresher or stronger smelling nonmatching human scent among the same age matching and nonmatching comparisons, to expose an incorrect solution strategy to always choose the freshest scent regardless of whether it matches the scent sample.)
- Care should be taken that scent samples and comparisons are not cross contaminated with human scent form others.
- There should be enough matching and nonmatching transfer test trials to determine dogs are not responding by chance. Thus, a number of novel test stimuli are required in order to provide sufficient statistics to evaluate transfer performance.
- Matching transfer test trials should be randomly and equally mixed with nonmatching transfer test trials; they should not be systematically (predictably) mixed with one another, such as nonmatching trials always first.
- During both matching and nonmatching transfer trial tests, correct responses should be reinforced according to the matching-to-sample/nonmatching reinforcement contingency used during training in order to avoid changing the way dogs have learned to solve the problem (see below the matching/nonmatching relational learning procedure for a nonmatching reinforcement contingency).
- In order to establish dogs have fully learned a general rule that can transcend training stimuli, transfer performance to novel human scents should be equal or above baseline performance (training accuracy before transfer) and both baseline trials and transfer test trials should be well above chance.
Transfer to novel human scents provides good evidence that dogs have learned a relational solution strategy that can transcend training stimuli. Thus, if high baseline accuracy is maintained over transfer to novel human scent matching/nonmatching test trials, it can be assumed that accurate performance is mediated by a general rule; that what was learned during training generalizes to novel human scents. However, when dogs are tested and transfer performance with novel human scents is not equal to baseline training performance, it indicates dogs have learned a solution strategy other than a general rule to solve the task or have only partially learned a general rule and will not perform reliably during human scent matching/nonmatching operations involving strangers. For example, in a study to evaluate whether specially trained Dutch police human scent matching dogs could match scents collected from different parts of human bodies, Schoon and De Bruin (1994) found significant differences in test results between familiar and stranger suspects. They reported that “When the dog knew the ‘suspect’, 73% of the trials were correct, when the ‘suspect’ was a familiar scent (often used in training situations) 67% were correct, and when the ‘suspect’ was a complete stranger to the dog 25% of the trials were correct”.
Nonmatching general rule
During operations, scent evidence may be found at a crime scene that does not match the scent of a suspect. Therefore, in addition to learning a matching rule that generalizes to novel human scent, human scent matching service canines must also learn a nonmatching rule that generalizes to novel human scent. In order to avoid falsely accusing an innocent person or wasting valuable search time, human scent matching/nonmatching service canines must be as or more reliable at indicating when the individual unique human scent information on a scent sample does not match any comparisons as they are at indicating when the information does match and both matching and nonmatching accuracy rates should be well above chance. Furthermore, the response used to measure when there is no matching scent present should be equally likely, or more, to occur as the response used to measure when there is a matching scent present.
Although it is important that human scent matching/nonmatching service canines learn a general nonmatching rule in addition to a general matching rule, MTS training alone does not require animals to learn two rules; both a matching and nonmatching rule. Subjects do not need to learn a nonmatching rule in order to respond accurately and reliably to novel stimuli during MTS alone trials. During MTS training, responding is differentially reinforced, response to the matching alternative is reinforced and response to a nonmatching alternative is not reinforced, while suppression of response, by nature of the MTS procedure, is never followed with reward; not differentially reinforced. Thus, there is nothing in the MTS procedure to inform subjects when suppression of response is correct. Subjects can learn about the circumstances in which a response will be reinforced, but cannot learn about the circumstances in which suppression of response is correct. Moreover, in order to reliably solve matching problems, subjects do not need to learn two solution rules, a rule to respond to matching alternatives and another rule to suppress response to nonmatching alternatives. A stronger disposition to respond to matching alternatives than to respond to nonmatching alternatives, which is established through differential reinforcement of response during matching alone training, is enough to assure successful matching performance (Premack, 1978). Therefore, no response to a nonmatching scent during matching control trials, in which a suspect’s scent is a nonmatching alternative, does not establish dogs will suppress response during nonmatching investigative or forensic trials when an innocent suspect’s scent does not match the scent on an item found at a crime scene.*5 The absence of an identification response to nonmatching alternatives on matching trials does not establish dogs have learned a nonmatching rule.
From the standpoint of a matching only rule learned during MTS only training, nonmatching test trials are impossible to solve. When animals are first trained with one procedure, MTS, and then tested with another procedure, go/no-go in which the measure of learning is response on matching trials and suppression or latency of response on nonmatching trials, there is no match present and no reinforcement predictable from the initial MTS alone training on subsequent no-go nonmatching test trials unless subjects lower their choice criterion and respond incorrectly in an attempt to earn reward. That is, the only way in which matching rule only animals can attempt to predict reward on subsequent go/no-go nonmatching test trials is to lower their choice criterion and respond, which is what experimental evidence indicates they do when animals are trained one way, MTS only, and are then tested another way, go/no-go, (see e.g. Urcuioli & Nevin, 1975; Wright & Sands, 1981; Hepper, 1988; Lu et. al., 1993, Schoon, 1996). Thus, although latency of response is a more reliable response measure than suppression of response to indicate the absence of a matching alternative on nonmatching test trials, latency of response also does not establish subjects have learned a nonmatching rule. Furthermore, because response is differentially reinforced according to the MTS reinforcement contingency, while suppression of response is never reinforced during MTS only training, MTS only training increases a disposition to respond relative to suppression of response. Thus, after MTS only training, both latency and suppression of response are unequal measures relative to response.
In addition, although many trainers and agencies are aware of the importance of both matching and nonmatching training and testing, matching/nonmatching training and testing typically (traditionally) involves a conditional go/no-go procedure in which correct response on go (matching) trials are reinforced and correct suppression of response on no-go (nonmatching) trials are not reinforced. In other words, responding is differentially reinforced on go matching trials, in that correct response to a matching alternative is reinforced and incorrect response to a nonmatching alternative is not reinforced, yet on no-go nonmatching trials, suppression of response is not differentially reinforced. On no-go trials reinforcement is omitted both when subjects incorrectly respond to nonmatching alternatives and correctly suppress response to nonmatching alternatives. Thus, on no-go nonmatching trials there is no match present and no reinforcement predictable. There is no information provided in the conditional go/no-go procedure that informs subjects when suppression of response is correct.*6
As with MTS only training, conditional go/no-go training procedures increase a disposition to respond because response is differentially reinforced and suppression of response is not differentially reinforced. Once again, response and suppression of response (including response latency) are unequal measures. After conditional go/no-go training, accuracy rate on on-go nonmatching trials should be low relative to go trials. In other words, dogs are likely to respond, indicating that there is a match when there is no match on no-go nonmatching trials, which is what experimental evidence shows.v
Schoon (1996) compared the standard “negative check” scent identification lineup design used in the Netherlands, involving a conditional go/no-go procedure (both matching and nonmatching test trials), with three other designs that only involved matching test trials and found that the standard negative check design had the lowest percent of correct responses and the highest present of incorrect responses. 65.5% of incorrect responses (failure to suppress response) occurred on no-go nonmatching negative trials. This is not acceptable for forensic and no place-last-seen matching/nonmatching mantrailing operations. A disposition to indicate a match should not be greater than a disposition to indicate there is no match; if anything for forensic operations, it should be the opposite. Furthermore, it’s worth repeating that the absence of response to a nonmatching scent during matching only control trials (which is what the other three designs involve) does not establish dogs will suppress response during nonmatching trials when a suspect is innocent.
Comparative studies with various species, modalities, and stimuli show that when animals are trained MTS alone and conditional go/no-go they do not completely suppress response when tested on no-go trials (Urcuioli & Nevin, 1975; Wright & Sands, 1981; Hepper, 1988, Lu et. al., 1993; Schoon, 1996). Both procedures have the disadvantage of providing two different response measures of unequal sensitivity. After both MTS only and conditional go/no-go training, the predominant disposition is to respond. Alternatively, by differentially reinforcing not just matching trials but also nonmatching trials in the matching/nonmatching relational learning procedure and requiring dogs to respond in the same manner but to different stimuli, rather than mixing requirements to respond and not respond, the problem of unequal response measure can be avoided.*7 In the matching/nonmatching relational learning procedure advocated here, the reinforcement contingency is as follows: on matching training trials, response to the comparison that matches the sample is reinforced, but reinforcement is omitted after either response to a comparison that does not match the sample or response to stimulus X (that should be present during both matching and nonmatching trials); on nonmatching training trials, response to stimulus X is reinforced (which indicates there is no matching comparison present on that trial), but reinforcement is omitted after response to a nonmatching comparison. Over training and testing, matching and nonmatching trials are varied randomly, as well as the sample stimuli, comparison stimuli, and the presentation positions of the comparisons (see, matching/nonmatching relational learning procedure in The History of Police Canine Tracking).
It should not go without mention that the same/different (S/D) procedure may be more suitable for same-as/different-from relational learning and forensic application than the matching-to-sample (MTS) alone procedure (see Schoon, 1997b, for some preliminary investigation*8). In the same/different procedure, ‘same’ and ‘different’ trials occur separately and are equally but randomly distributed over trials. In a given trial, subjects are simultaneously presented two comparison stimuli that are either the same or different from one another. If the comparisons are the same, subjects are required to make an arbitrarily designated ‘same’ response to obtain reward, such as respond to one of the matching stimuli. If they are different, a response to stimulus X, present on both same and different trials, is required for reinforcement, which indicates they are ‘different’.
Some advantages of the same/different procedure over MTS alone training are; (1) subjects explicitly receive both same and different training trials involving a reinforcement contingency for both (the opportunity to receive reward after a correct response on both same and different trials), (2) subjects must learn two rules in order to solve both same and different problems accurately and reliably, and (3) the designated same-different responses are the same but are directed to different stimuli so the response measure is equal in both same and different trials. These three advantages can also be found in the matching/nonmatching relational learning procedure advocated here, rather than MTS alone.
However, although these three advantages can be found in the matching/nonmatching relational learning procedure, another S/D significant advantage involves simultaneous presentation of the to-be-compared stimuli in the S/D procedure, rather than successive presentation of the to-be-compared stimuli during MTS or matching/nonmatching trials. In order to learn a same-as or different-from relationship, stimuli must be compared. When the discrimination between stimuli is difficult (involving complex stimuli, such as human scent presented on various items involving a compound of different olfactory information) and the to-be-compared stimuli are presented successively (as they are during MTS and matching/nonmatching trials when the sample stimulus is presented at one time and the comparison stimuli are presented at another time), a same-as or different-from relationship may be less obvious to subjects than they would be when the to-be-compared stimuli are presented simultaneously (as they are in the S/D procedure). In the MTS or matching/nonmatching relational procedure the comparison must be made between one stimulus (the matching comparison) and the memory of the other (the scent sample). For example, in both human scent MTS alone and matching/nonmatching training, the memory of the previously presented sample stimulus must be compared with subsequently presented comparisons in order to detect whether any of the individual unique information on the comparisons is the same as or different from the individual unique information on the sample. Both human scent MTS alone and matching/nonmatching training has the disadvantage of involving memory of the sample stimulus until a matching alternative can be located because the to-be-compared stimuli are presented successively. Furthermore, when the training set is small, memory of a matching comparison from a previous trial can interfere with memory of the sample stimulus on a current trial (proactive interference) and also the number of comparison stimuli within a trial can interfere with memory of the sample (retroactive interference). In addition, human scent is a complex stimulus, involving a variety of information such as genetics, gender, diet, smoking, hygiene, and cosmetic information. Thus, during both MTS alone and matching/nonmatching tasks canines must observe the olfactory information on the scent sample, encode the individual unique information present in compound with additional irrelevant information, retain the individual unique information over a search delay, compare the encoded individual unique information with choice alternatives without disruption of memory, identify the same-as or different-from relationship, and respond correctly. Alternatively, because the comparison stimuli are presented simultaneously in the S/D procedure, the likelihood that dogs will compare the to-be-compared individual unique human scent information is greater, which can significantly increase the likelihood that they will learn a same-as/different-from relational solution that can transfer to novel human scents. With the S/D procedure, canines can compare scents and make a same-as or different-from determination without the addition of memory processes, the need to successively compare encoded information, and competing stimuli that may disrupt memory.
Experimental evidence indicates that when discrimination is difficult, simultaneous presentation of stimuli results in significantly faster relational learning than does successive presentation, although not necessarily if it is easy (see e.g. MacCaslin, 1954). Presumably, this is so because when the discrimination is difficult, involving stimuli close in similarity or complex stimuli, the simultaneous presentation of discriminative stimuli makes evident relational cues (such as, same-as and different-from cues) that may not be perceptible when stimuli are presented successively. The opportunity for simultaneous comparison makes available relational cues that can be used in addition to the absolute features of the discriminative stimuli, which greatly increases the incidence of relational solutions, whereas in successive discrimination, only absolute cues may be used. When the discrimination is difficult, but not necessarily when it is easy, simultaneous presentation of discriminative stimuli fosters solutions based on the relative relationships (relative to one stimulus, the other is the same-as or different-from), whereas successive presentation fosters solutions based on the absolute characteristics of stimuli.
Alternatively, some of the disadvantages from successive presentation of sample and comparison stimuli during matching/nonmatching conditioning can be overcome by increasing attention and associability of individual unique human scent information prior to matching/nonmatching training. This involves overtraining simple human scent discrimination prior to scent matching/nonmatching training (or S/D training). Researchers have found that dogs trained to discriminate between human scents, do not respond on the basis of gender information, race, smoking habits, or cosmetics (see e.g. Schoon & Haak, 2002). Although, research indicates dogs trained to discriminate between human scents respond on the basis of the individual unique human scent profiles of human scent at the expense of other information present in compound with the individual unique information, the reasons why dogs learn to use individual unique human scent information remain largely unexplained. A subsequent chapter will offer a theoretical explanation that is supported by empirical evidence and widely accepted by the comparative psychology academic community. For now however, it is worth noting that overtraining simple human scent discrimination prior to matching/nonmatching conditional discrimination training can enhance same-as/different-from relational learning during subsequent matching/nonmatching training by increasing attention and associability to the individual unique component of human scent. Not only can overtraining simple human scent discrimination increase the likelihood that dogs will notice the individual unique matching relationship between the scent sample and matching alternative during subsequent MTS training, overtraining simple human scent discrimination can also eliminate a stimulus presentation response habit prior to MTS, increase the significance of scent samples prior to MTS, and by increasing attention and associability to the individual unique component of human scent at the expense of irrelevant stimuli, simple human scent discrimination overtraining can help overcome some of the memory drawbacks during subsequent matching/nonmatching training. Other ways to overcome memory problems during subsequent matching/nonmatching training involves training with a large training set (ideally all strangers) and having only two comparisons in each trial during the acquisition phase of training.
*1. A third way animals can learn to solve MTS problems is by memorization of visual configural patterns. When the set or number of training stimuli is small, animals can learn to solve visual MTS problems by learning the specific configural patterns of the stimulus displays (Carter and Werner, 1978). Configural learning involves memorization of the global presentation pattern of each training display and learning which alternative to choose based on the global pattern. Subjects that have learned to solve MTS problems using configural cues cannot solve matching problems on transfer tests with novel stimuli because prior learning experience is necessary for memorization of each specific stimulus pattern. Because the configural solution strategy involves memorization of visual stimulus patterns, as long as olfactory MTS training and testing involves scent items that are all visually the same configural pattern learning need not be a concern.
*2. Another associative solution strategy subjects may learn during MTS training involves memorization of visual display patterns. If the set or number of training stimuli is small, animals can learn to solve visual MTS problems by learning the specific configural patterns of the stimulus displays (Carter and Werner, 1978). Configural learning involves memorization of the global pattern of each training display and learning which alternative to choose based on the global presentation pattern. Subjects that have learned to solve MTS problems using configural cues cannot solve matching problems on transfer tests with novel stimuli because prior learning experience is necessary for memorization of each specific stimulus pattern.
*3. In comparative cognition, MTS choice alternatives are termed comparisons under the assumption subjects compare the alternatives with the previously presented sample stimuli.
*4. In the next chapter, an easier to learn MTS solution strategy (involving discrimination between more and less familiar novel human scents) will be reviewed, which enables dogs to reliably solve human scent matching problems with strangers, but does not enable them to solve nonmatching problems involving innocent suspects.
*5. More will be reviewed about matching alone training in the next chapter.
*6. The human scent matching/nonmatching go/no-go procedure is termed the conditional go/no-go procedure to distinguish it from the matching/nonmatching relational learning procedure.
In the conditional go/no-go procedure, go (matching) and no-go (nonmatching) trials are successively presented in random order to control against dogs learning a systematic pattern of response. For example, if a nonmatching no-go trial is always presented first before matching go trials, dogs can simply learn to always suppress response on the first trial, rather than learn about the nonmatching relationship between the sample and alternatives. Over trials, sample stimuli, choice alternatives, and the presentation positions of the alternative stimuli are also randomly varied. On matching go trials, a trial begins with the presentation of a scent sample. After response to the sample, it is removed and dogs are required to choose between alternatives the individual unique human scent that matches the individual unique human scent on the scent sample. The reinforcement contingency on go trials is; if dogs choose correctly the matching odor, they are rewarded, but if they choose incorrectly a nonmatching alternative, reinforcement is omitted.
On nonmatching no-go trials, a trial begins with the presentation of a scent sample. After response to the scent sample, it is removed and two or more nonmatching human scent alternatives are presented. The dog’s task on no-go trials is to suppress response. However, there is no reinforcement contingency on no-go trials to inform dogs the task requirement to suppress response; reinforcement is omitted regardless of whether dogs correctly suppress response or incorrectly respond to a nonmatching alternative. In conditional go/no-go procedures, subjects can only obtain reinforcement on go trials. On no-go trials, reinforcement is always omitted regardless of response.
*7. The problem of unequal response measure can also be avoided with the S/D procedure. The S/D procedure involves the same response (rather than response and suppression of response) but to different stimuli on ‘same’ and ‘different’ trials and both randomly and equally distributed ‘same’ and ‘different’ trials are differentially reinforced.
*8. The procedure Schoon termed odd-even was a same/different procedure in that dogs had to make either a same-as or different-from determination in order to solve the problem and other than the pair of stimuli to-be-compared there were no additional comparison stimuli (as there are in MTS). However, in Schoon’s odd-even procedure the to-be-compared stimuli were presented successively, rather than simultaneously as they are in the standard same/different procedure.
Bodily, K. D., Katz, J. S., and Wright, A. A. (2008). Matching-to-sample abstract-concept learning by pigeons. Journal of Experimental Psychology: Animal Behavior Processes, 34, 1, 178-184.
Boehm, T. and Zufall, F. (2006). MHC peptides and the sensory evaluation of genotype. Trends in Neuroscience, 29:2, 100-107.
Brisbin, I. L. and Austad, S. N. (1991). Testing the individual odour theory of canine detection. Animal Behavior, 42, 63-69.
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.
Cook, R. G. (2002). Same-different concept formation in pigeons. In M. Bekoff, C. Allen and G. M. Burghardt (Eds.), The Cognitive Animal, (pp. 229-237), Cambridge, MA: MIT.
Eggert, F., Luszyk, D., Haberkorn, K., Wobst, B., Vostrowsky, O., Westphal, E., Bestmann, H. J., Wolfgang, M., Ferstl, R. (1999a). The major histocompatibility complex and the chemosensory signaling of individuality in humans. Genetica, 104:3, 265-273.
Eggert, F., Muller-Ruchholtz, W., Ferstl, R. (1999b). Olfactory cues associated with the major histocompatability complex. Genetica, 104, 191-197.
Ferstl, R., Eggert, F., Muller-Ruchholtz, W. (1988). Major histocompatibility complex-associated odours. Nephrol Dial Transplant, 13, 1117-1119.
Ferstl, R., Eggert, F., Westphal, E., Zavazava, N., and Muller-Ruchholtz, W. (1992). MHC-related odors in humans. In R. L. Doty and D Muller-Schwarze (Eds.), Chemical Signals in Vertebrates, (Vol. 6, pp. 205-211), Plenum Oress, New York.
Green, S. C., Stewart, M. E., and Downing, D. T. (1984). Variation in sebum fatty acid composition among human adults. Journal of Investigative Dermatology, 83, 114-117.
Haplin, Z. T. (1986). Individual Odors among mammals: Origins and functions. Advances in the Study Behavior, 16, 39-70.
Harvey, L. M. and Harvey, J. W. (2003). Reliability of bloodhounds in criminal investigations. Journal of Forensic Science, 48:4, 811-816.
Harvey, L. M., Harvey, S. J., Mom, M., Perna, A., and Salib, J. (2006). The use of bloodhounds in determining the impact of genetics and the environment on the expression of human odortype. Journal of Forensic Science, 51:5, 1109-1114.
Hepper, P. J. (1988). The discrimination of human odour by the dog. Perception, 17, 549-554.
Kalmus, H. (1955). The discrimination by the nose of the dog of individual human odours and in particular of the odour of twins. The British Journal of Animal Behaviour, 3, 25-31.
Katz, J. S., Wright, A. A., and Bodily, K. D. (2007). Issuses in the comparative cognition of abstract-concept learning. Comparative Cognition & Behavior Reviews, 2, 79-92.
Lu, X-C. M., Slotnick, B. M., and Silberberg, A. M. (1993). Odor matching and odor memory in the rat. Physiology and Behavior, 53, 795-804.
MacCaslin, E. F. (1954). Successive and simultaneous discrimination as a function of stimulus similarity. American Journal of Psychology, 67, 308-314.
Nicolaides, N. (1974). Skin lipids: their biochemical uniqueness. Science, 186, 19-26.
Penn, D. J, Oberzaucher, E., Grammer, K., Fischer, G., Soini, H. A., Wiesler, D., Novotny, M. V., Dixon, S. J., Xu, Y., and Brereton, R. G. (2007). Individual and gender fingerprints in human body odour. Journal of the Royal Society of Interface, 4, 331-340.
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, & W. K. Honig (Eds.), Cognitive Processes in Animal Behavior, (pp. 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. (1997a). The performance of dogs in identifying humans by scent. Thesis, University of Leiden, Netherlands.
Schoon, G. A. A. (1997b). Scent identification by dogs (Canis familiaris): A new experimental design. Behavior, 134, 531-550.
Schoon. G. A. A. (1998). A first assessment of the reliability of an improved scent identification line-up. Journal of Forensic Science, 43:1, 70-75.
Schoon G. A. A. (2001). The effect of experimental setup parameters of scent identification line-ups on their reliability. Probemy Kryminalistyki, 236, 43-49.
Schoon, G. A. A. and De Bruin, J. C. (1994). The ability of dogs to recognize and cross-match human odours. Forensic Science International, 69, 111-118.
Schoon, G. A. A. and Haak, R. (2002). K9 suspect discrimination: Training and practicing scent identification line-ups. Detselig Enterprises Ltd., Calgary, Alberta, Canada.
Settle, R. H., Sommerville, B. A., McCormick, J., and Broom, D. M. (1994). Human scent matching using specially trained dogs. Animal Behavior, 48, 1443-1448.
Singer, R. A., Beauchamp, G. K., Yamazaki, K. (1997). Volatile signals of the major histocompatability complex in male mouse urine. Proceedings of the National Academy of Sciences of the United States of America, 94, 2211-2214.
Singh, P. B. (2001). Chemosensation and genetic individuality. Reproduction, 121, 529-539.
Skinner, B. F. (1950). Are theories of learning necessary? Psychological Review, 57, 193-216.
Sommerville, B., Settle, R., Darling, F., and Broom, D. (1993). The use of trained dogs to discriminate human scent. Animal Behaviour, 46:1, 189-190.
Urcuioli, P. J., and Nevin, J. A. (1975). Transfer of hue matching in pigeons. Journal of the Experimental Analysis of Behavior, 24, 149-155.
Wallace, P. (1977). Individual discrimination of humans by odor. Physiology and Behavior, 19, 577-579.
Wright, A. A., and Sands, S. F. (1981). A model of detection decision processes during matching to sample by pigeons: performance with 88 different wavelengths in delayed and simultaneous matching tasks. Journal of Experimental Psychology: Animal Behavior Processes, 7, 191-216.
Zavazava, N., Wobst, B., and Ferstl, R. (2005). Soluble MHC class I molecules in human body fluids. Journal of Clinical Laboratory Analysis, 8:6, 432-436.