CLPEX Chat Board

From Ashton

For example; when we attempt to analyze the unknown we examine a series of factors in a “bottom up” process. We examine the substrate, the matrix, elements of pressure distortion, clarity, etc. The analysis process requires us to examine the latent and interpret the information we see.

I have seen in several books and paper’s the principle of “Bottoms Up” Analysis. I have been told that it is this way because of a cognitive psychology concept that by examining things from the top down could lead to errors in the analysis and possibly to the conclusion of the examination so we must conduct our analysis in method leading from the bottom to the top (Substrate—Matrix—Deposition—Anatomical Features). Not being a Cognitive Psychiatrist I cannot speak of this principle but often wonder if it refers to a mental sequencing of events or does it actually represent a physical application of events (i.e. counting money we move from the larger denominations to the smaller).

The following quotes are presented not as all aspects of the ‘bottom up’ or ‘top down’ principle but are only presented to perhaps spur those not familiar with the concepts to research the topic on their own.

Top-down and bottom-up processes in learning and thinking
In cognitive psychology and cognitive science, top-down and bottom-up processes refer to processes that flow from either the top or the bottom of the information processing hierarchy, respectively (Lindsay & Norman, 1972).The top of the hierarchy is assumed to contain high-level, abstract, and encompassing knowledge representations such as concepts, mental models, and schemata. On the other hand, the bottom of the hierarchy is assumed to contain low-level, concrete, and specific knowledge representations such as visual features, lexicons, and propositions (e.g., Bruning, Schraw, & Ronning, 1995; Kintsch, 1998). Cognitive psychologists and cognitive scientists believe that both top-down and bottom-up processes are indispensable to both human and machine thinking.
Bottom-up processing is said to occur when one draws from some particular examples, instances, cases, or events to a generalization, rule, or law to capture the commonality between the examples, instances, cases or events (e.g., Brown,Collins, & Duguid, 1989). Similarly, top-down processing is said to occur when one infers from a generalization, rule or law to conclude something about a particular example, instance, case, or event. Induction is an example of bottom-up processing, and deduction is an example of top-down processing.

(From ‘Top-down, Middle out, and Bottom-up Processes: A Cognitive Perspective of Teaching & Learning Economics’ by Ling-po Shiu, Chinese University of Hong Kong and Chor-yiu Sin, Hong Kong Baptist University)

Top-Down Processing and Pattern Recognition
bottom-up processing or data-driven processing:
• stimulus information arrives from the sensory receptors (the bottom level of processing)
• the combination of bottom-level features allows us to recognized more complex, whole patterns
• feature analysis
top-down processing or conceptually-driven processing:
• our knowledge (memory) about how the world is organized helps in identifying patterns
psychologists believe both bottom-up and top-down processing are necessary to explain pattern recognition
(From ‘www.alleydog.com/cognotes under pattern recognition)

From my POV the supposition that the assessment of latent prints must take the approach of Substrate—Matrix—Deposition—Anatomical Features has been at odds with the current way of doing assessments or analysis. From my POV most of the latent prints that I was analyzing the conclusion would not be affected by the substrate or matrix determination. It must be pointed out that yes there were some where the matrix, substrate or deposition was considered in the analysis but from my POV this might only account for a small number of latent prints (perhaps 3—4 %). I wanted to see if other LPE might be using the approach I was accustomed to or were they actually using the S-M-D-A approach to their assessment of latent prints.

I had selected 6 latent prints that I assigned a level of difficulty as Simple, Moderate and Complex (2 of each). Over a period of several months I would take one of the latent prints and approach one of six latent print examiners and hand them the latent print and asked that they look at it (the total of examinations were 6 latent prints X 6 latent examiners equals 36 assessments). I noted what they looked at (the substrate information were on the opposite side of the lift card). I would then ask them what they looked at first. All responded with the ridge detail first. However I did notice in their examination that out of 36 assessments the substrate information was considered in only 4 assessments but only after looking at the friction ridge detail first.

Since in my POV ‘distortion effects’ only impacted a small percentage of latent prints being analyzed I then wanted to determine some type of hierarchy in the types of ‘distortion effects’ (disruption of friction ridges [scars, injuries, etc] were not considered at this time) that might be present. I selected 50 latent prints that were moderate or complex and determined that the majority dealt with deposition effects (movement, pressure, multiple impressions), with the matrix being the next most prevalent and last with the substrate being the least common (numbers varied as a number of latent prints were not representative of a single type of distortion effect but consisted of a combination and to a degree my interpretation of the effects). My ‘distortion effect’ hierarchy consisting of deposition, matrix, and then substrate.

The following proposal is a different perspective of how latent prints are assessed during the 1st phase of an examination. First one should image a pyramid. An information pyramid with the majority of the information is at the base and that information that is less prevalent moves up the pyramid. Keep in mind that just because information is less prevalent does not mean it is not important.

Taking from Dr. P.L. Kirk and H. Tuthill I suggest the first few bits of information should proceed from the class (gross) features or ridge flow and then next move on to that information that is needed to individualize. In other words observations move from the class to the individual. If during this assessment of class and individual features that one observes that the ‘effects of distortion’ is present and present in such a form that might impede a comparison then the ‘cause’ of the ‘effects’ of the distortion will be examined to determine what information may be derived from the ‘cause and effect’ of distortion. Moving up the information pyramid (hierarchy) one would examine the latent print for the effects of deposition, and then next to the matrix, and lastly the substrate. Keeping in mind that the majority of latent prints will not have to move up the information pyramid as comparison and decisions reaching a conclusion can be formed upon the majority of latent prints while only a few will require a full examination of the pyramid.

By approaching the analysis in this form one can still maintain the “bottom ups” approach since the hierarchy first derives from the anatomical features of class (ridge flow) to individualization features (2nd and 3rd LD), then moves to a consideration of distortion effects starting with Deposition, then to Matrix, and finally to Substrate.

In any event do you as a Latent Print Examiner start your analysis with determining the substrate and then moving to the matrix, and then to deposition and finally with the anatomical features on all of your examinations???

(Note: for the sake of this short presentation I am not including ‘Ridge Disruptions’ scars, injuries, disease or genetic aberrations as these are on the friction ridges before contact with a substrate. Also for the sake of brevity I am only referring to ‘Distortion Effects’ of which the causes are deposition (movement, pressure, multiples, etc); matrix (dirty, contaminants, amount, etc) or substrate (various). I acknowledge that others may have more varied and definitive terms.)

Charles,

I’m confused and I feel like people are talking about 2 different things when it comes to top-down and bottom-up processing.

Some people associate bottom-up with the order that we look at data, or the order that information is placed (the substrate being on the bottom, then the matrix, and then the ridges). Top-down would be looking directly at the ridges and then considering the matrix and the substrate (or ignoring it all together).

Others think of bottom-up processing as using objective data like ridge flow and ridge characteristics. I think the matrix and substrate are also part of bottom-up (objective data). Top-down processing would be using subjective data like case type, previous arrest record, and anything emotional that weighs on our conclusion. I don’t know how people label distortion. I see some distortion as being objective but other distortion could be very subjective.

From what I’ve seen, I think people are using words from psychology to describe the physical order of data. I could be wrong but I think using these words to describe the sequence isn’t how they should be used.

Michele Said

I’m confused and I feel like people are talking about 2 different things when it comes to top-down and bottom-up processing.

Michele, not as confused as I am on this subject! First I would like to set some terms as I see them so that we might have a common language and hope you can straighten me out or the very least we continue the discussion. .

If we take what is already written and look at a latent print from the edge the bottom is the substrate, the matrix, deposition, and then the ridges just like you described. With bottom up processing the substrate is looked at first and then the others in the order of the data. Top down would be the ridges (ridge flow and characteristics), then deposition, matrix and last substrate.

You did clear something up in that some others think of top-down processing is using case type, previous arrest record or anything that is emotional that weighs on our conclusion. I have never seen that description before for top-down and have never considered those criteria as top-down information but they are peripheral information that does not come into consideration in the examination. However if people do believe that then I see why they do not consider top-down processing as very safe. I would agree if that was the standard definition of top-down processing.

I look at bottom-up and top-down as more in line with the first quote I made on this posting. One is Inductive and one is Deductive. One is analysis and one is synthesis.

But now I am in a pickle! If I take what the cognitive psychologist who are on the peripheral edge of this disciple and some of the more vocal LPE that say top down is bad, bad, bad and that I am doing it wrong because I start at the top with the friction ridges and move downwards through deposition, matrix and then substrate with the last ones being considered if the examination demands it.

So if Top-Down is bad (apparently some Cognitive Psychologist do not agree—read the first two quotes) then I have to alter my way of conducting the examination by looking at the substrate first on all latent print examinations or change my understanding as to what is the actual bottom of the information available. Is it the substrate or friction ridge detail? Since the substrate does not figure in on the majority of latent print examinations it kind of rules itself out as the base of the information heap. If it is not considered in a particular examination then it certainly cannot be considered as the base.

We build houses, cars, bridges from the base up. The very foundation (base) of the examination process is the ridge flow and then moving to ridge features, then the matrix, on to deposition and finally the substrate. The case and subject data are peripheral and should not be considered in any examination.

I reject the position that by looking at the friction ridge detail first will lead to errors in the conclusion. Misinterpretation of any of the data can lead to an error and not the order in which it is examined. I would also say that looking at friction ridge detail first, then to deposition, followed by matrix and then substrate is also a structured and defined analysis.

Michele Said

From what I’ve seen, I think people are using words from psychology to describe the physical order of data. I could be wrong but I think using these words to describe the sequence isn’t how they should be used.

Could you elaborate further? What is your definition and criteria for Bottom-Up and Top-Down processing?

When you pull a lift card out of the envelope do you first look at the substrate information whithout even looking at the ridge structure??

Charles,

I think the first time I saw this word was when Dror used it to describe influencing people emotionally in “The effect of contextual top-down processing on matching fingerprints”.

I can read several things into the second quote. It refers to processes that flow from either the top or the bottom of the “information processing hierarchy”. I don’t think this is necessarily the same as the sequential order of where data sits. This quote refers to top-down as abstract and bottom-up as concrete, this could mean that top-down is more subjective and bottom-up is more objective, but that’s just an interpretation.

If I take what the cognitive psychologist who are on the peripheral edge of this disciple and some of the more vocal LPE that say top down is bad, bad, bad and that I am doing it wrong because I start at the top with the friction ridges and move downwards through deposition, matrix and then substrate with the last ones being considered if the examination demands it.

I have heard that bottom-up is the preferred way of doing things. I don’t necessarily agree with this but which cognitive psychologist stated the information in your quote? I can see that abstract information may bias people but I also think that there are times when this information is useful information. For instance, if we’ve ID’d a latent from a bottle to a suspect, then I’d say that it’s more likely that the others on the bottle will also be his. I don’t use this information to make the ID but I use it to aid in my searching.

I agree with you that there is no harm in looking at the ridges first.

Michele

Michele, you are correct. The Top-Down is subject/case data as well as past experiences and I think current information. I just read Dror and Fraser-Mackenzie piece in the weekly detail. Here all this time I was thinking it was perceptual when it was informational data.

So scrap all that I said previously and see if the following works:

Bottom Up Model: A systematic approach to the analysis of distortion in complex prints.

I think that's good for people who already know what bottom-up is but those that don't know may need more of an explanation.

By the way, your post came at a perfect time for this weeks detail. Coincidence?

Charles,

A couple of thoughts. Some of this may be not so much a conflict between bottom-up and top-down as it is which direction is up and which is down. It may be worth thinking about. Stepping away from the usual up/down assumptions, why is substrate the bottom? Is it appropriate to try to force the whole process into one or the other mode? We can speak of a purely analytical model in which the unknown is meticulously analyzed from broad issues of such as substrate and progressing through to fine detail. But we have to ask if the model can reasonably be enforced in practical application. If the first examiner's task is limited to producing an analysis, the model works well. It would be as if the task was to prepare data for an automated matching system that would next be applied. But that's not the task. The task includes comparison. And I think we confuse ourselves when we imagine that every aspect of comparison can be corrupted if we don't strictly exclude any reference to the known in the analysis stage, or we imagine that analysis and comparison can ever and always be kept isolated from each other. There IS circularity in an examination. If the requirement was that the latent and known images overlay each other exactly, circularity could be eliminated. But it's not like that. Analysis has not ended when comparison begins. SOME analysis is properly ended and locked in. There is a high risk when the absolute determinations from analysis that are required for certainty in individualization are revisited during comparison without extraordinary caution. Other analysis of the sort that determine if a perceived difference is artificial will not even be prompted until comparison is underway. And the will therefore certainly not fit into any strict up or down analysis.

When we talk about up or down, are we talking about analysis or comparison or both? We just looked at analysis. Comparison, taken alone, is normally most efficient when done from global to specific. If Level I is out of agreement, we're finished and finished quite quickly. If Level II is out, we're done, without squinting at Level III. And, of course, agreement at a higher level prompts comparison at a lower level. But that seems to me to be so obvious that it's outside our top/bottom discussion. The real question seems to me to be: Is the nature of latent print analysis, bound as it is to a larger program of comparison, amenable to any strict interpretation as up or down, or is the concept of up or down merely a worthwhile consideration in maintaining awareness of the integrity issues involved? In other words, are we just fishing for another formula because we think it makes our fishing look more like ichthyology.

I wonder if I can apply an idea from my crime scene teaching to this. I introduce students to the concepts of the global and discrete. We first view the scene globally, judging it's apparent nature and extent. This phase extends through the early operations of scene survey or walk-through and general recording. Individual evidentiary objects are approached discretely, each one being handled without presumptions driven by other evidence. It's documented and collected to the same high standard, no matter how the global view seems to place it in the overall situation. The process becomes global again in a partly free-associating and partly logically analytical combination. But it cannot be so strict that potential is lost. Things that might be observed but not recognized as evidentiary in isolation become objects of documentation and collection at any stage because of other, more clearly evidentiary, things. And, of course, within the stage of handling discrete objects, there is a general hierarchy based on interactions between objects and the intruding investigator/analyst, largely preventing loss or damage to one in the course of dealing with another. It's a structured approach for a specific global task, and the peculiarities of the that task makes it not amenable to strict application of any universal principle.

These issues are about principles. They are about protective and validating parameters, not about programmed machine thinking. And mistakes along the way do not preclude corrections when the reason for the principles are kept high in the examiner's awareness.

http://www.visionlab.harvard.edu/member ... MITECS.pdf

http://users.ecs.soton.ac.uk/id/ACP%20e ... 0ident.pdf

The above may help to clarify things a little.

cheers


Dave

Dave,

Useful orientation on the issue. I would first, though, play the critic and hold that the subject bias study is less significant that the designer hopes. The conclusion acknowledges that competent examiners may or may not be more or less vulnerable to the kind of top-down influences used and give reasons for the alternate speculations. But it is a very great leap from the circumstances of the experiment to actual examination. The subjects in the study were not doing anything resembling fingerprint analysis, except that they happened to be looking at fingerprints. They might as well have been looking at leaves or zebras or ink blots, except that the study required some extraneous information that could be provided that could both relate to fingerprints and induce two different levels of emotional response. In other words, fingerprints were well-suited to the needs of the study, but the study says little about fingerprint examination, and suggestion that it has much to do with it, aside from confirming a quite predictable response is an overstatement. A probably correct conclusion that top information is unlikely to bias bottom-up judgments does not necessarily strengthen the applicability to fingerprint examination.

The first article reminds that experience and especially training are hugely important factors. We can identify the nature of an object with high accuracy from the merest detail, because we have considerable experience viewed such objects. I would guess that few people in advanced cultures would fail to recognize an automobile from a photograph of a very small detailed portion, perhaps a smoothly shaped, typically finished area of body abutting a typically curved corner of windshield. A body shop technician might be expected to do even better. A naive observer will not perceive as an experienced observer. We see the effect in many things. Having been a serious and professionally trained photographer, I always see the shadow on concrete under a clear sky as blue, in life, or just as it was recorded and reproduced in a photograph. A naive viewer, one who is not familiar with the phenomenon nor ever charged with dealing with the effect, will see it according to their expectation. Concrete is gray. That is their "training." It does not matter that, even when directly observed, it is blue under that condition of skylight and will be found to be blue when the color is evaluated by a photo-electronic sensor. They are not trained to see is as it is and have no reason through training to expect to see as anything else. Nor will they perceive that the face of the Caucasian person as green, which if not corrected, is it, as a result of blue skylight falling on an essentially yellow skin.

A naive viewer of two fingerprints is prone to simply not interpreting clear differences between them, because they have no conception of what constitutes sufficient similarity or difference. Even if they make an effort to move to bottom level details, they have no basis for judging what is difference and can say little with confidence unless the two are virtual matching overlays. Perhaps I should more accurately say they can say little with justifiable confidence and will become far less confident if told that the differences are hard to detect. The naive viewer of the photograph or even the live scene will say with considerable confidence that the concrete is gray, and the skin is yellowish tan, but when the phenomenon is explained to them, they can easily see that one is blue and one is green, and they will now confidently describe them correctly, even more so if they are further trained by comparing a version of the photograph with the blue corrected to gray.

Gerald,

A naive viewer of two fingerprints is prone to simply not interpreting clear differences between them, because they have no conception of what constitutes sufficient similarity or difference.

This is a good point. M. Kam, et al. showed this (in a series of studies) with trained document examiners (FDEs) v. novices (lay people). He found lay people were better at finding and assessing similarities than recognizing and evaluating appropriate differences. In other words, it's easier for lay people to make correct matches than correct exclusions. Both the strategies and thresholds for these two processes (matching v. excluding) may be different.

I found the same effect when testing fingerprint experts v. novices (forthcoming JFS study, accepted for publication for 2009 May). Novices performed better when examining same source images than when comparing different source images.

We also have found differences in experts in biasability and accuracy with respect to same source judgments v. different source judgments. So the effect is not just limited to laypeople, only more pronounced with laypeople.

I think as a profession we can explore some interesting dichotomies between "identifications" and "exclusions". I know many of us think of these two things as polar opposites on a scale or balance (I often teach that way using Pat Wertheim's scale analogy). It's convenient and intuitive to think of them that way. But, I believe at this time, research will show vast differences in both our approach (strategy) to identifications v. exclusions, and differences in the threshold of information needed to reach the conclusion.

Food for thought.

g.

g. wrote:I think as a profession we can explore some interesting dichotomies between "identifications" and "exclusions". I know many of us think of these two things as polar opposites on a scale or balance (I often teach that way using Pat Wertheim's scale analogy). It's convenient and intuitive to think of them that way. But, I believe at this time, research will show vast differences in both our approach (strategy) to identifications v. exclusions, and differences in the threshold of information needed to reach the conclusion.

It's natural to come to think of them that way because they're products of the same examination, and both are presented as absolute conclusions. But they are different, I think. I think it's quite possible to determine that an unknown (1) is an impression of a finger, (2) is insufficient for individualization, and (3) conclusively not that of a known source. The simple case would be a poor impression with enough to say it's a finger and enough to orient it and determine that a specific clear detail does not appear in the known.

I think we learn something by thinking about how we use impression and distortion interpretations. Once we reach the threshold for individualization, we presume that areas of very poor clarity would not contain excluding details that we could see if the impression was better. We have established a threshold for that impression and would demand proof in the form of a contradiction to invalidate the threshold. This points up the necessarily rather fluid nature of the identification threshold, in that we would extend our search for differences into those areas if they were available to us. We admit that possibility of difference by seeking it (and further agreement) in latents with large areas of high clarity but deal presumptively with equally large areas of poor clarity. I think we would answer the obvious criticism by pointing out that a responsible examiner uses all available detail. While we would still hold that it is our firm opinion that the identification would be correct without so large an area available, we would be forced by the lack of a sharply defined standard threshold to say that we would accept a clear difference in that additional area as reason to exclude, albeit that it would be so significant a finding as to make it appropriate for publication. In other words, the answer to the question: "Were you to find that in that poorly impressed area a clear discrepancy, how would it affect your conclusion?" An answer that depends on a variation on the belief that the sufficiency of the clear area is convincing is not an entirely bullet-proof response. Perhaps someone can provide a better one.

But as to the scale analogy, I think the nature of the two present as a scale that starts with some weight on one side that requires a bit more than that on the other to tip. These are what we might call individualizers and non-individualizers. But there are also anti-individualizers that can drop out of the sky at any time, and they have infinite mass and not only tip the scale but destroy it, eliminating even the concept of relative weight for the problem of that pair of impressions.

Gerald, I finally got back to this one:

Gerald Said

A couple of thoughts…. which direction is up and which is down…. why is substrate the bottom? Is it appropriate to try to force the whole process into one or the other mode? We can speak of a purely analytical model in which the unknown is meticulously analyzed from broad issues of such as substrate and progressing through to fine detail. …..I think we confuse ourselves when we imagine that every aspect of comparison can be corrupted if we don't strictly exclude any reference to the known in the analysis stage….. There IS circularity in an examination….. Analysis has not ended when comparison begins. SOME analysis is properly ended and locked in….Other analysis of the sort that determine if a perceived difference is artificial will not even be prompted until comparison is underway..... (some) will therefore certainly not fit into any strict up or down analysis.

I agree that some parts of the analysis are static while some are certainly circularity (back and forth). Also I think that having such a model for those images that have been determined complex can also be problematic.

Gerald Said

When we talk about up or down, are we talking about analysis or comparison or both? We just looked at analysis. Comparison, taken alone, is normally most efficient when done from global to specific…..it's outside our top/bottom discussion. The real question seems to me to be: Is the nature of latent print analysis, bound as it is to a larger program of comparison, amenable to any strict interpretation as up or down, or is the concept of up or down merely a worthwhile consideration in maintaining awareness of the integrity issues involved?

I think the later is appropriate as being worthwhile in maintaining some type of integrity with the observations to be made.

Gerald Said

….the concepts of the global and discrete. We first view the scene globally, judging its apparent nature and extent. This phase extends through the early operations of scene survey or walk-through and general recording. Individual evidentiary objects are approached discretely, each one being handled without presumptions driven by other evidence……

Moving from a global to discrete or class to individual is a big part of most analytical approaches.


Gerald Said

these issues are about principles. They are about protective and validating parameters, not about programmed machine thinking. And mistakes along the way do not preclude corrections when the reason for the principles are kept high in the examiner's awareness.

I agree that principals are the guiding rules of an analytic process but the process by its very nature back and forth by observing, collecting the information and collating with past experience, knowledge of oneself or perhaps in some cases others.


Gerald, this whole thing started with me in 1999 when I read a popular book that mentions about systematically starting at the substrate. I did not think much about it at the time other than if that person wants to start at the substrate so be it. As for me I will start at the ridge detail going from class to individual and during my assessment if I observe problems then try to resolve them by cause and effect. Here is the effect now what is the cause. Then I read where the bottom up thing was to be done on complex prints. I said fine if someone wants to do that but it seems out of sync since you have to look at the ridge detail to determine if you have a complex print. But then a couple of years ago there was a comment about if the number of errors could be attributed to those persons taking a “top down” approach vs. the “proper” ‘bottom-up’. Now that bothered me. Why would it work that way? Why would someone think of it that way?

It was not until I read Michele’s piece, then Prof. Dror’s piece and of course David C. put that link in to another article that when the others are talking about ‘top-down’ they are talking about a whole different thing. They are not talking about a vertical ‘slice’ of a latent print but they are talking about the case dynamics (how gruesome), or the number of arrests a subject had or that they had spit on their mother at age 17. All that stuff I use to think of as ‘peripheral information’ since I do not use that as a part of my examination.

So now when I hear or read the terms ‘top down’ or bottom up’ I have to stop and figure out if they are talking about ‘biasing’ information or are they talking about the physical vertical layers of a latent print.

I think the book from 1999 was talking about the physical layers which I still think that such a structured model leaves something to be desired. If you have a complex print and recognize that the effect is from a poor matrix are you going to stop and study the substrate? From my POV when one is dealing with a complex print then one needs to be flexible within the stated principles as to the effect and cause of the complexity of that image.

I think authors are driven to lay out their presentation with a lot of structure. Part of this is to help the book flow in a way that makes sense and can be readily followed. Some of it may be viewing the book as a learning/teaching text, and students will get confused by complex interactions, and they don't have the experience or orientation to work globally when appropriate. And there's always the factor of "the system." There's a lot of potential in coining. Consider "Ridgeology." Such books can attract more attention and provide the author with more related earning activities when it's seen as "a new approach" or so-and-so's system. It's not bad. A lot of advancement in any number of areas of practice has come about because being hailed as a step forward moved the potential practitioners to learn it and apply it. And an expert practitioner reading the book will take in the information, knowing it can be applied without entirely abandoning current practice.

The expert practitioner can appropriately attend more closely to one aspect of analysis or largely pass over that aspect. The novice cannot safely step outside the structure. The novice print examiner would no doubt be well advised to attend to substrate and matrix, because they can affect aspects of the image in ways that he cannot know are important. He's still at the experience stage of asking how much he believes in what he sees. (He is also gaining experience in making judgments about substrate and matrix and therefore should be attending to them in every analysis.) An experienced examiner can confidently observe that there is nothing peculiar about the impression that would require other analysis. But even the experienced expert must have been trained in the value of analyzing substrate and matrix and must be mindful of its potential. It must always be in his toolbox, although he may not always apply it.

Some years ago, my patrol car went stone cold dead one day. It was towed to the dealership where technicians poured over it, following the structured diagnostic approach and getting nowhere. They were talking about bringing in a specialist from out of town for this new evolution of police-package electrical system when an old retired mechanic walked by the door. They called him in and described the problem. He called for a jumper cable and grounded the negative battery cable to the frame and started the car. Experience told him that was the most likely problem and that looking for something complex wasn't productive. He also knew enough to know that it couldn't hurt anything to try it.

The trick, of course, and what makes an individual expert reliable, is knowing what's more or less dangerous and having the discipline to manage the risks.

My apologies folks,

This is nothing but a shameless bump-up to keep this thread close to the top for a while until such time as I can study it properly or print it out.