One of the key elements in the logical approach to evidence evaluation are the propositions used for the evaluation. They are, in a certain sense, the most important part of the whole process. At the same time, they are also one of the least understood.

Today’s post explores the concept of propositions. I will attempt to describe what they are, how they are used, why we don’t change them once set and why they matter so much, among other things… all from the perspective of forensic document examination (though also applicable to other forensic disciplines).

This post is long (sorry about that). I have written it in the following sections which you can jump to using the following links:

  1. Introduction: the bit just below this.
  2. What are the propositions and what do they represent?
  3. What are the basic requirements for propositions?
  4. The Hierarchy…
  5. How are propositions ‘set’?
  6. Is there a ‘standard set’ of propositions for handwriting examinations?
  7. Sub-propositions — Why FHE is so complex
  8. How are propositions used?
  9. Why shouldn’t propositions be changed?
  10. Some additional thoughts… explanations versus propositions


This topic can be both complex and confusing. So before we head down a rather deep rabbit hole let’s begin with a quick review of the requirements for proper forensic evaluation of evidence (discussed more fully here).

Briefly, all evaluation of evidence occurs:

  1. Within a framework of information pertaining to the matter at hand,
  2. With consideration for at least two competing and mutually-exclusive propositions, and
  3. With a focus on the evaluation of evidence given each proposition; not the evaluation of the propositions given the evidence.

This post will touch on each of these points but it will focus on the concept of “propositions”.

What are the propositions and what do they represent?

As Evett and his co-authors explained in 1998,1 propositions are statements “which represent, respectively, prosecution and defence allegations in a criminal trial.” In other words, they will reflect different arguments that will be made during a trial. In one sense, they can be considered theories or hypotheses about what has happened to produce the evidence in question. In the legal context those theories and beliefs are the basis of an overall argument relating to the definition for a particular legal charge or claim of liability. The ‘ultimate’ arbiter in all such matters is a judge or trier-of-fact who, after listening to the various arguments made by the parties, reaches a final decision regarding guilt or innocence. While this is true many investigations never get to court due to a decision made by some other party. There are, in fact, several decision-makers involved along the way (not that each of them reaches a decision about guilt — for example, sometimes the decision relates to whether or not an effective case, for one point-of-view or the other, can even be made).

First, there is an investigator who must sort through various possibilities about what may have happened. In that role they often make their own decisions about what did or did not happen. When satisfied they have enough evidence to support some particular theory or belief about what happened, the matter proceeds to the first ‘legal’ stage of the process. The information (and the theory and the charge) is reviewed by a lawyer who must decide if the matter is well-enough established to move it further. Again, at that stage, the lawyer must weigh the evidence and reach their own decision about it. Through each of these steps the parties will undoubtedly be considering, to some degree, alternative possibilities or theories. In most cases, only when they feel one theory in particular has been ‘established’ will they move forward. This may or may not involve any input from the suspect or their legal representative who would, almost certainly, have some contrary position in mind, though most courts have no requirement for the defence to present an alternative argument. As a minimum, however, there is always the generic ‘alternative’ where “my client is innocent and knows nothing of what happened” whether that is ever expressed or not. In the end, if the matter fulfils the criteria for a prosecution or civil suit it then advances to a trial/hearing of some sort.

At that point, the propositions of interest are best defined in terms of the actual allegations and arguments to be made during the trial. However, it should be obvious that a number of parties are involved whose beliefs, and arguments, may differ or change over the course of the investigation. It should also be clear that, therefore, there may be a large number of different propositions to consider. And that the relevant propositions may differ at different points in the process.

What are the basic requirements for propositions?

There are three aspects of propositions that come out of the earlier statement regarding the general nature of the evaluation of evidence. They are:

  1. Two or more — evidence can be assessed in terms of a single proposition but that information alone provides minimal probative value for the trier-of-fact. Probative value comes, literally, from the ability to differentiate between two or more positions/theories/beliefs. Consider, for example, a situation where the evidence is ‘exactly’ as would be expected when a particular event occurred. On the face of it and taken as-is, we might say that the evidence strongly supports that particular theory. However, what does this mean for a trier-of-fact? Well, it’s interesting and potentially very useful information. But, taken on its own, it is pretty much useless information. Why? Because the evidence could be very significant or completely meaningless in terms of ‘explaining’ what may have actually happened. The actual value of the evidence to the trier depends not on its ‘absolute’ value but rather its ‘relative’ value in terms of other possible theories. The trier needs to know how much the evidence might support some other competing proposition or theory. If the evidence provides little or no support for any other proposition, then it would clearly provide relatively strong support for the first theory over those other propositions. However, if the evidence happens to provide equally strong support for some other proposition, then there will be no differential support for the first over the other proposition. Thus, in the latter situation the actual probative value of the evidence, in terms of differentiating between those particular propositions, would be nil. This issue is closely related to the concept, sometimes overlooked, of the need for ‘balance’ in the evaluation process. At least one other alternative is required to achieve any semblance of balance. Please see also my comments at the end of this point about the distinction to be made between an ‘explanation’ and a ‘proposition’.
  2. Competing — as the above comments indicate the different propositions are competing in the sense that they represent different theories about what may have happened to produce the evidence. In order to actually evaluate evidence and provide some sense of its probative value to the trier, there must be at least one other alternative theory against which the evidential value can be weighed.
  3. Mutually-exclusive — it should be obvious that propositions need to be mutually-exclusive, meaning that both propositions cannot be true (or occur) at the same time. If the propositions are not mutually-exclusive then it will be impossible to differentiate between them no matter what evidence is found/present.

The ENFSI Guideline for Evaluative Reporting in Forensic Science (2015) sums all of this up nicely:2

Propositions are statements that are either true or false, and that can be affirmed or denied (Anderson et al. 2005). Propositions should be formulated in pairs (e.g., views put forward by the parties to the cases) and against a background of information and assumptions. Moreover, they should be amenable to a reasoned assignment of credibility by a judicial body and be useable for rational inference.

Edit: I should make a comment about one other facet of propositions — the concept of exhaustivity. People often that the set of propositions must be exhaustive and cover all possibilities that may apply, but that is not true. While exhaustivity might be nice from the point-of-view of the trier who wants to understand the ‘truth’ of the matter, there is no requirement for it. Why? Because the propositions, as noted above, should reflect the “views put forward by the parties to the cases” and there is absolutely no need for those views to be exhaustive and complete. In fact, in many instances the arguments made in court are very specific, limited and far from exhaustive.

Edit 2: The fact that the proposition set does not have to be exhaustive has some implications worth noting. First, the set under consideration may or may not include the actual event that produced the evidence being evaluated. This is significant because the end result may favour one proposition over any other, and even do so strongly, but that proposition may not represent the ‘actual’ causal event. This reality may bother some examiners (and others) but it is essentially a reflection of how our justice system works. Second, from a mathematical point-of-view this means that the resulting ratio of likelihoods (the LR) will be the form of relative, and not absolute, odds.

The Hierarchy…

If you read the (plentiful) literature on this topic you will encounter the ‘hierarchy of propositions’. Like all good things in life (wines, beer, and many other things) we can place propositions into a hierarchy. In this case, one reflects the type of question being asked of the examiner and of interest to the trier.

At the base are “source propositions” so-called because the question relates to the “source” of some trace or sample. A good example is a blood stain or a piece of glass. Well, that was the bottom in the original conceptualization of the hierarchy. At the time forensic scientists didn’t have the ability to examine ‘nothing’. Okay, it’s not really nothing — but the ability to detect and analyse ‘trace DNA’ did not exist. Now it does. Which means one can swab a surface with no apparent material on it and get a DNA result. To deal with this unanticipated (but logical) reality the term “sub-source” was coined. It’s the new ‘base’ rung of the ladder.

Source, and sub-source, propositions focus on from whence evidence comes. Literally, what is the source of the sample? Such information can be very useful to the trier but it’s often not as useful as questions posed at higher levels in the hierarchy.

The next level are “activity propositions” which try to address the issue of what happened when the trace was deposited or made, rather than focusing on the source of a stain, trace sample or material. The issue here is how did that evidence come to be there, or how did it manifest? These are much more interesting to the trier (as a rule) because now the issue being addressed is some aspect of what actually happened. Examples might be “the questioned sample was written by the suspect” (versus “the questioned sample was written by someone other than the suspect”), “the window was broken by the suspect using a hammer held in his hand” (versus “the windows was broken by some other person”), or “the semen was deposited during non-consensual sex” (versus “the semen was deposited during consensual sex”).

Now, is it even possible to address the latter question? It would not be easy and it clearly requires information beyond the question of source. It might be possible but it really depends on the nature of the evidence and whether or not it helps to differentiate between the two propositions. However, leaving that issue aside, this type of question is set at the ‘activity’ level. It’s very easy to see how it differs from the more basic (and arguably less useful) issue of source. At the same times such things are not always clear. In some situations, for example, questions about handwriting authorship, the distinction between ‘source’ and ‘activity’ is debatable.

The highest level on the hierarchy are “offence propositions” which, quite frankly, are beyond the scope of the examiner to address. These derive literally from the questions the trier-of-fact must address. These would be things like “the suspect committed forgery” or “Mr. X raped Ms. Y”.3 Obviously, there are things involved in the assessment of such statements that an expert cannot possibly address.

Evidence is important for evaluations undertaken at any level in the hierarchy but, for the expert, only those below ‘offence’ level can be addressed in a meaningful way. The top level remains the exclusive domain of the trier.

I should note, however, that traditional conclusions (framed in terms of posterior odds of the propositions) often speak to this type of proposition either directly or indirectly. Beyond this, the reality is that the expert will be asked these questions all too often.  After all, both the trier and the advocate want answers at this level.  They will, as a result, often ask questions at this level which the examiner should defer from answering.

How are propositions ‘set’?

As noted above, propositions ideally represent competing theories or hypotheses about “what happened”. As such they will obviously and necessarily vary according to the specifics of a case or investigation. The only person(s) in a good position to express such beliefs (for one side, if not both) will be the submitter. That person can, as a minimum, express their belief about what has happened which would be the basis for the main proposition in the competing set. But that is only ½ of what is needed. The alternative position is equally, if not more, important to the matter for reasons given above.

In an ideal world, the ‘other side’ would simply state their position in the matter and provide the examiner with the alternative to be considered. But we live and work in the real world where such an ideal is a rarity. In many jurisdictions there is no requirement for a defence to state their position at all. Some have taken this to mean you cannot use the logical approach to evaluation because, quite simply, it requires an alternative to work. The counter-argument is that there is always an implicit alternative to any proposition — that being the negation of the proposition. In fact, the negative is traditionally what most examiners have used, albeit often unknowingly. This can be seen in purpose statements like ‘to determine whether or not…’ where the alternative is just the ‘not’. Any time the defence opts to not provide a specific counter-argument, or whenever one is not provided for any reason, then this is a simple and logical choice for the alternative. However, it may not be the best or ideal counter-position. In which case, a more specific alternative proposition can, and should, be specified and used. The fact is that very often examiners are not given the main proposition, let alone the alternative, so they must formulate all of the propositions to be used.

When specific propositions are not provided the examiner must formulate a set of working propositions as best they can. This can be done by according to policy or based on past experience. Often a reasonable set of working propositions can be derived from the basic question asked by the submitter. No matter how it is achieved, the alternative should be set to “favour” the defendant as much as possible and, as a minimum, to represent an acceptable alternative, mutually-exclusive point-of-view.

In the end it is important to declare the propositions used and, again ideally, to have that set approved by the submitter before proceeding. It is also important to let everyone know what can be expected should different propositions be considered or used. To that end people should be informed that any change to propositions, information or evidence may result in a different conclusion or outcome. This can be done in various ways but a simple disclaimer may suffice. I should note that many examiners get upset with the idea that their conclusion is not final and ‘set in stone’. But that is an inevitable reality deriving from the conditional nature of our opinions. If the conditions at play change, then the evidence must be interpreted accordingly. That doesn’t mean the ultimate opinion will change, but it might. It is the need for transparency in all we do that requires us to let people know about this aspect of our work.

Is there a ‘standard set’ of propositions for handwriting examinations?

No, not really. Any set of propositions (main and alternative) of interest to the trier is possible — at the same time this should not be taken to mean the examiner will be in a position to address every possible proposition posed to them.

Having said that there is, in reality, a fairly obvious set of propositions that one might consider in any given handwriting comparison aimed at assessing potential authorship. They are as follows:

  • H1: the writer of the specimen samples wrote the questioned handwriting
  • H2: someone other than the writer of the specimen samples wrote the questioned handwriting

This set of propositions reflects, quite closely, the approach most examiners take in a traditional examination. It derives from the question originally posed which is usually along the lines of “did my suspect write the questioned writing?”

This seems pretty basic and simple. But there is more here than meets the eye.

Writing is an interesting type of evidence. Unlike many other forms of evidence the static questioned trace is the product of a dynamic human behaviour.4 Thus, the production of writing is subject to a huge number of sources of variation; some internal to the writer, others external. I won’t delve too deeply into this as it is laid out quite well in the literature.

The bottom line, from the perspective of authorship evaluation, is that the examiner must make inferences about the original writing action based upon information present in the static trace. This requires careful and extensive sampling of any given suspect’s writing to inform inference about their actual habits, extensive knowledge of intrinsic and extrinsic factors that may affect a person’s writing, plus extensive knowledge of writing in general (e.g., what one can expect from people in the population at large when writing for a number of different purposes) to inform inferences falling under the alternative propositions.

Thinking in terms of possible propositions one might encounter in any given case we can start the process by considering the variety of different ‘ways’ in which a given questioned writing might come into being. Please note that the relevant propositions for a given case must be defined by the specifics that apply for that matter. The present exercise is simply that; an exercise to explore the nature of propositions more fully.

Sub-propositions — Why FHE is so complex

The set outlined above can be thought of as ‘top-level’ or ‘main-level’ propositions. However, each of these top-level propositions can be decomposed into a number of sub-propositions that reflect the manner in which handwriting might be produced.

This idea has been presented before; see, for example, Biedermann et al’s discussion in Learning about Bayesian networks for forensic interpretation: An example based on the ‘the problem of multiple propositions’.5 Their approach was different from mine, but presented a similar propositional decomposition. The breakdown I use is similar to one outlined by Stephen Day in Handwriting and Signatures, Interpretation of Comparison Results, World Encyclopedia of Forensic Science (2009).6

As a starting point, consider a ‘decomposition’ of the main proposition, H1 (i.e., the writer of the specimen samples wrote the questioned handwriting), into the following sub-propositions:

  • H1a: the writer of the specimen samples wrote the questioned handwriting, in a completely normal and natural manner
  • H1b: the writer of the specimen samples wrote the questioned handwriting but as some form of unnatural writing

Attention is given here to two very broad ‘types’ of writing called natural (i.e., written in a normal reflexive manner without conscious control or any other influence at play) and unnatural (i.e., written while exercising some degree of conscious control over the writing and/or while being affected by some internal or external factor).

Now, the latter might entail a number of sub-types that can be expressed in further decomposition:

  • H1b1: the writer of the specimen samples wrote the questioned handwriting, attempting to disguise their identity
  • H1b2: the writer of the specimen samples wrote the questioned handwriting, after having a few drinks (or, more generally, under the influence of alcohol)
  • H1b3: the writer of the specimen samples wrote the questioned handwriting, under the influence of medication
  • H1b4: the writer of the specimen samples wrote the questioned handwriting, while riding in a vehicle on a bumpy road

… and so on.

In general, each of the ‘qualifying terms’ added to the base proposition describes a form or type of writing that might be produced by the suspect (since this is H1 we’re talking about), writing either naturally or unnaturally for a variety of reasons or conditions.

But this is just the start. Something similar can be done for the alternative proposition, H2, that someone other than the writer of the specimen samples wrote the questioned handwriting.

Let’s start with:

  • H2a: someone other than the writer of the specimen samples wrote the questioned handwriting, with no attempt to mimic the suspect (i.e., a spurious writing)
  • H2b: someone other than the writer of the specimen samples wrote the questioned handwriting attempting to mimic the suspect

Now, let’s expand these in a manner similar to what was done earlier. We might end up with something like:

  • H2a1: someone other than the writer of the specimen samples wrote the questioned handwriting, with no attempt to mimic the suspect and in their normal and natural manner
  • H2a2: someone other than the writer of the specimen samples wrote the questioned handwriting, with no attempt to mimic the suspect but attempting to disguise their writing


  • H2b1: someone other than the writer of the specimen samples wrote the questioned handwriting attempting to mimic the suspect but while attempting to disguise their identity
  • H2b2: someone other than the writer of the specimen samples wrote the questioned handwriting, attempting to mimic the suspect’s writing habits using a tracing method
  • H2b3: someone other than the writer of the specimen samples wrote the questioned handwriting, attempting to mimic the suspect’s writing habits using a freehand simulation method

… and so on.

Again, each of the ‘qualifying terms’ serves to describe a form of writing that might be produced by someone other than the suspect, writing either naturally or unnaturally for various reasons.

It should be clear that the situation can very quickly become complicated. That will come as no surprise to anyone familiar with handwriting examination. This type of work is, to be blunt, a difficult and challenging task. Everything I have just described is routinely done by examiners — though they may not be cognizant of it and they may not express it in this manner.

It is important to note that the examples provided here are not intended to be complete or exhaustive though they represent many of the scenarios examiners would consider as a matter of course. In any given situation it may be that some of these possibilities would be ‘ruled out’ by other information provided to the examiner in which case the set can be reduced. When this happens, an explanation should be provided in the report as to why that sub-proposition was removed or not considered. More important, in any given situation a specific sub-proposition (one of these, or some other) may be the focus of a particular matter.

Finally, it is possible to take these decompositions further forming even more precise sub-propositions. However, that is an exercise to be addressed at another time.

How are propositions used?

This is the meat of the matter. All evaluation of evidence is done in terms of the specific propositions under consideration, and within a framework of other information.

Once the propositions have been determined and agreed to, and before any analysis or examination occurs, it is possible for the examiner to outline in good detail what they would expect to see under each of the propositions of interest. This might sound strange, but it really is not that hard to do. After all, examiners are trained in what to expect when writing is natural versus traced versus simulated, etc. This is a fundamental part of expertise. Each of these types of writing is expected to produce certain features. Such expectations are not absolute in nature, rather they are probabilistic, meaning there are things we expect to see more often with some types of writing than for others.

Any competent examiner should easily be able to express probabilistic expectations for each of the various propositions. Whether the examiner chooses to express their beliefs in actual numeric terms is optional though there are good reasons for doing so in my opinion. Nonetheless, it is possible, and perhaps advisable, to include this information in the examiner’s notes.  How that is done depends on the protocols used in a particular lab or office.

I know that most examiners do not do this in any formal way even though a fully trained, competent examiner is capable of doing so. The value in making this type of ‘pre-declaration’ of expectations rests in the fact that the examiner cannot be accused of fitting their expectations to the data if they have declared those expectations before they ever looked at the data in a given case.

At any rate, once expectations have been outlined, the examiner should proceed to the analysis of the evidence. In an ideal world, a complete and full analysis should precede any of the evaluation. However, things don’t work that way in practice. In reality, examiners begin to evaluate the submission as soon as a case arrives in their office. Every effort should be made to limit biasing non-evidential factors from influencing the examination and evaluation. To that end, various protocols can be invoked but, for our present discussion, let’s assume all those protocols have been followed.

The formal analysis of the questioned and specimen samples discloses certain findings.7 Findings are made up of all the features, similarities and dissimilarities (the latter I prefer to call ‘divergences in form or habit’), observed in the examination and comparison between the sets.

Evaluation of evidence is always subjective and personal in nature. It is important to remember that this evaluation is being done by a highly trained, competent and skilled “expert”. That facet of the matter is the same for all types of evidence and in every forensic discipline. The opinion of the expert, whether or not hard quantified data is available to inform their opinion or belief, will always be subjective and personal.

The assessment should be done for each of the outlined propositions. The ultimate objective is to determine which, if any, of the main propositions is supported by the evidence more than the other. In addition, the examiner must assess how much more support is provided by the evidence for one rather than the other. Strictly speaking, the final determination should be a direct reflection of the magnitude of the likelihood-ratio that applies for the evidence and those propositions. In most instances, having already outlined the expectations for the material it is a relatively simple matter to determine whether or not (and how well) the material matches those expectations.

Now, using formal mathematical symbols the likelihood-ratio (LR) is: $$\dfrac{p(E|H_{1})}{p(E|H_{2})}$$ In this formula we can clearly see the two parts of the process that must be assessed.8 The numerator involves the assessment of the evidence under the main proposition while the denominator involves the assessment of the evidence under the alternative. A key element in that evaluation is the understanding that the numerator relates to the suspect while the denominator does not. When considering the denominator the suspect (and the specimen samples that exist) do not matter very much at all —they are of interest only under alternative sub-propositions involving tracing or simulation.

Some examiners are comfortable with the idea of probabilistic belief and have no difficulty in expressing numeric values for the denominator and numerator; that is, the probability of observing the evidence under each of the specified propositions. Those individuals can form a likelihood-ratio directly based on the probabilities they assign. Another, slightly different but equivalent, approach focuses on the expectations outlined earlier in the process and assesses how well (in a probabilistic sense) the evidence meets those expectations. Thus, if the expectations are met ‘completely’ then the probability would be close to 1 but, if those expectations are not met, then the probability would be close to zero.

After the examiner has determined the magnitude of the LR they can translate it into the appropriate opinion to be expressed in their report. There are various guidelines for this purpose, including the specific policy of their lab/office.

Now, an astute reader will have noticed that little explanation has been given about the role of sub-propositions in this process. The examiner’s opinion will address the evidence in terms of the main propositions, H1 and H2. The sub-propositions outlined above serve to clarify aspects or elements of each of the main propositions; specifically the main proposition under which any given sub-proposition falls. Thus, when an examiner is trying to evaluate the probability of the evidence given H1 they must consider each of the sub-propositions to get a sense of the probability of the evidence under each of those sub-propositions. After doing this for each of the sub-propositions, the examiner must ‘consolidate’ or ‘integrate’ the set to form their belief for the main proposition.

To help in this, the sub-propositions should be stated in mutually-exclusive terms (e.g., written naturally versus unnaturally) whenever possible. Doing so makes the assessment and integration back into the main proposition much easier. However, as the decomposition goes deeper this is not possible. In other words, the evidence stops being effective in terms of differentiating at some sub-level of decomposition. At that point, the examiner must not try to interpret the evidence at the lower level. They must, instead, do the integration keeping that limit in mind.

One other point is worth noting relating to what happens if any of the sub-propositions can ‘ruled out’ by the framework or by declaration of the submitter. Any time a sub-proposition can be dismissed from consideration the entire process becomes simpler and easier to manage.9 An obvious question is how or when this might be possible?

It might happen in a situation where the prior odds of one or more of the sub-propositions are extremely low. For example, situations where there is absolutely no reason why a suspect would disguise their own signature because the sole purpose behind signing is to ‘identify’ that person. Clearly any final decision about the dismissal of a sub-proposition should not be made by the examiner as it must be based upon factors external to the evidence itself.  Therefore, this should be discussed with the submitter during the initial definition of the propositions permitting that party to make up their mind about it. If the submitter agrees (and they are willing to explain the reasons to the trier-of-fact) then that particular sub-proposition may be excluded from the evaluation. I strongly recommend that any such exclusion also be clearly explained in the report.

The evaluation process for sub-propositions is discussed more fully here.

Why shouldn’t propositions be changed?

The issue underlying this is the need to assess the evidence GIVEN the propositions. Propositions must never be ‘adjusted’ to conform to the evidence.

That does not, however, mean that propositions must never be changed. A legitimate change might occur when the key issues in the case or key conditioning information in the framework have changed. But such changes should not be made by the examiner.

Some additional thoughts…

People often think in terms of the ‘best’, ‘correct’ or ‘only’ explanation for the evidence. This isn’t how things work in a court of law. What matters, first and foremost, are the arguments being made to the trier-of-fact. Hopefully, the final outcome will be ‘best’ and ‘correct’. But only rarely will there exist ‘one and only one’ possible explanation for the evidence. This applies at every level of the evaluation, including that being done by the examiner. Even when the evidence provides only limited differential support for one proposition over another it may have value to the trier-of-fact.

It is also important to understand the distinction between an ‘explanation’ and a ‘proposition’. Consider the definitions for each of these taken from the 2015 ENFSI Guideline:

Propositions are statements that are either true or false, and that can be affirmed or denied (Anderson et al. 2005). Propositions should be formulated in pairs (e.g., views put forward by the parties to the cases) and against a background of information and assumptions. Moreover, they should be amenable to a reasoned assignment of credibility by a judicial body and be useable for rational inference. Propositions should be distinguished from explanations that do not have the aforementioned properties.


In the context of a forensic science evaluation, an explanation has been recognised as an intermediate consideration for use when exploring less formal alternatives. A key characteristic of explanations is that they are generated after the forensic findings have been obtained. While an explanation has the potential to account for particular observations, it does not qualify as a formal proposition because — often — it may be a statement of the obvious, speculative or fanciful. Moreover, an explanation can be offered provided that parties have presented no exclusive alternatives.

In historical terms, forensic examiners have tended to think in terms of ‘explanations’ for the evidence, rather than in terms of ‘propositions’ and the probability of the evidence given those propositions. This is one of the key differences in the logical approach to evidence evaluation.


  1. Evett et al. A Hierarchy of Propositions: deciding which level to address in casework. Science & Justice; 38(4):231-239.
  2. ENFSI guideline for evaluative reporting in forensic science: Strengthening the Evaluation of Forensic Results across Europe (STEOFRAE). June 9, 2015. Available online from the Université de Lausanne website.
  3. It should be obvious that propositions which are set at the ‘offence’ level speak directly to the issue of guilt versus innocence. The information may not be sufficient on its own to warrant such a verdict in a trial but nonetheless they speak directly to that issue.
  4. Ultimately, the examiner is interested in the underlying habitual motor patterns that produce a given static trace.
  5. Biedermann et al. (2012) Learning about Bayesian networks for forensic interpretation: An example based on the ‘the problem of multiple propositions’. Science & Justice, 52:191-198.
  6. Another, more recent, discussion can be found in Kadane, Joseph B., and Anders Nordgaard. “Using Bayes Factors to Limit Forensic Testimony to Forensics: Composite Hypotheses “. Australian Journal of Forensic Sciences 0, 0 (12 Feb 2024): 1-17.
  7. I am (slowly) adopting the terminology preferred in the ENFSI guide. The results of an examiner’s analysis and comparison constitute ‘findings’ rather than the more common the term ‘evidence’ with the latter being reserved for information disclosed and discussed in court. I am trying to use this terminology consistently, but I didn’t do so in this post.
  8. A more complete formula would include a parameter for the ‘framework of information’ that is omnipresent in the evaluation. It looks like this: $\dfrac{p(E|H_{1},I)}{p(E|H_{2},I)}$
  9. It can, in fact, completely change the opinion (or at least the strength of it).

3 thoughts on “Propositions — key to the evaluation process

  1. This is most interesting and definitely helps me to understand what is going on with propositions. It isn’t quite as simple as I thought. Thanks.

    1. From one ‘old’ QD guy to another I can see your perspective on this. It hasn’t been easy for me to adjust to this way of thinking. But the effort is worth it and it gets easier the more you do it.

      It is not at all simple to see how ‘propositions’ function in the context of logical evaluation of evidence. At least, not at first.

      At the same time, and from a purely practical point-of-view, none of this is that hard to do once you really understand how it all works. I firmly believe experts must understand such subtleties in order to apply this approach correctly. And I believe it is something we should all be doing.

      Thanks for the comment!


Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.