1876120

Statistical Adventures in Handwriting

Sun, 12 May 2024 04:45:15 GMT

This is a subtitle for your new post

Steven Strach gave a presentation at the Australasian Society of Forensic Document Examiners (ASFDE) Inc. ( www.asfdeinc.org ) meeting in Canberra on 10 April 2024. It was titled “Statistical Adventures in Handwriting” and covered the area of determining the expected frequency distribution of similarities and differences in basic classified features of some letters and numerals written by 83 people. This results in a large number of possible combinations of comparison pairs (3403) and for the initial work this was reduced by considering the comparisons for 23 of these people (with 253 pair comparisons). Useful feedback was received as requested from attendees on efficiently dealing with the statistics of large numbers of comparison pairs.

Historical Association of Dr Steven Strach with the development of the ESDA

steven@strachforensic.com.au (Steven Strach) — Sat, 20 Apr 2024 12:19:45 GMT

This is a subtitle for your new post

This describes the background to Steven’s role in bringing to light research on electrostatic imaging and in taking actions that caused new funding for the development of the prototype Electrostatic Detection Apparatus (ESDA) as a machine for the detection of writing impressions in paper.

Steven first met Bob Freeman and Doug Foster at the London Printing College (now the London College of Communication) in January 1977 when he was on a two day printing technology course there. At that time Steven was working as a Questioned Document Examiner in the then London Metropolitan Police Forensic Science Laboratory (MPFSL). The head of the document examination section (of what was then considered to be the most prestigious forensic science laboratory in the world) at that time was David Ellen, subsequent author of Scientific Examination of Documents Methods and Techniques (CRC/Taylor & Francis [4th edition 2018]; first published 1988). Steven was encouraged especially by David to meet Messrs. Foster and Freeman who were researching electrostatic methods for producing images of fingerprint patterns on materials and paper which might provide improved images compared with the chemical methods for developing such fingerprint deposits on paper (predominantly employing ninhydrin development in those days).

Doug and Bob enthusiastically showed Steven some recent test results on what, to the unaided eye, was blank paper. They had applied fingers to the apparently blank paper surfaces in order to test out the effectiveness of various proposed electrostatic development methods. The most promising method was an early inception of the now familiar ESDA method of placing thin Mylar film over the document, charging the film with a corona discharge unit and developing images with toner powder such as was used in the then increasingly common, dry toner photocopiers. As well as developing images of the fingerprints, to their surprise the electrostatic process also developed images of what were clearly writing impressions caused by writings on pages which had earlier been removed from the otherwise blank pad of test paper. They regarded the development as an interesting curiosity but their focus (as was that of their various research grant providers) was on developing improved methods of fingerprint detection. Although it worked in limited circumstances, the electrostatic method did not provide as good images of the test fingerprints as the chemical methods and never caught on as a regular technique.

These surprise writing impression results made a huge “impression” on Steven. After conveying his enthusiasm for the significance and great potential of these results to Doug and Bob, he went back to the laboratory with tales of a possible revolution in forensic document examination techniques taking place just down the road. MPFSL very shortly took over some of the funding of their project with a changed emphasis on developing the technique to detect writing impressions. This was just as well as their original grants were about to run out and the two research scientists were going to be looking for jobs elsewhere. The eventual result, after significant funding went into the product development, were the first commercial sales of the ESDA. By 1978, the machines were being manufactured in the garage of a private residence in Grays, Essex. Foster+Freeman Ltd has of course become a highly successful company developing, producing and marketing a large number of instruments for use in forensic science.

Conclusion Scale used in Forensic Document Examination

Wed, 21 Feb 2024 04:43:07 GMT

This is a subtitle for your new post

In forensic science, results of analyses are often expressed in terms of standard wording in a scale of conclusions. For some branches of forensic science (such as DNA analysis) there is statistical data available allowing a quantitative conclusion and results may or may not additionally be expressed qualitatively.

In forensic document examination and handwriting/signature analysis, statistical information is generally not available and it is important that the examiner’s qualitative opinion be expressed as clearly as possible using a scale of conclusions.

Conclusions on the results of handwriting/signature examinations undertaken at Strach Forensic take the following form (here expressed for signature examinations with a positive conclusion).

“There is {qualifier} support for the proposition that the questioned signature was written by the writer of the specimen signatures rather than for the alternative proposition that the questioned signature was not written by this person.”

Levels 1 to 4 have the following {qualifier} descriptions:

1. extremely strong;

2. strong;

3. moderate; and

4. limited.

For negative conclusions, the wording is changed round to read:

“There is {qualifier} support for the proposition that the questioned signature was not written by the writer of the specimen signatures rather than for the alternative proposition that the questioned signature was written by this person.”

With levels 9 to 6 having similar {qualifier} descriptions (in the negative sense):

9. extremely strong;

8. strong;

7. moderate; and

6. limited.

Where an examination was undertaken but none of these positive or negative conclusions could be reached (e.g. in a signature comparison case where a combination of similarities and differences was found or there were too few comparable components in the signatures), the result is described as “inconclusive” (level 5) with explanatory wording:

“That is, I cannot distinguish between the qualitative levels of support for the propositions that the questioned signature was or was not written by the writer of the specimens”.

The scale of conclusions is always provided in the report so that the conclusion(s) reached in that report can be seen in the context of that scale.

The conclusion levels have their origins in those branches of forensic science for which statistical information is available and relate to the general magnitude of the likelihood ratio (LR). This comes from a Bayesian statistical analysis of the data, the details of which will not be discussed here. However, it is useful to describe the LR and how it may be assessed qualitatively (without statistical data). The LR can be thought of as a scale (either numerical or verbal) which can support one proposition over another (the alternative) proposition. For example, if proposition A is “the questioned writing was produced by the writer of the specimens” then the alternative proposition B is “the questioned writing was not produced by the writer of the specimens” and the LR derives from the ratio of the probability of finding such evidence as has been observed given proposition A to the probability of finding such evidence as has been observed given proposition B. Such probabilities can either be statistically determined (as a numerical value between 0 and 1 or in percentage terms between 0 and 100%) or qualitatively assessed. In handwriting and signature comparison cases, such probabilities are typically qualitatively assessed. It is useful to have an idea of the numerical value that the LR might have if the statistical data were available for the various conclusion levels. Thus a level 1 positive conclusion is associated with LR values of the order of 10 ⁶ (i.e. 1,000,000) or greater, level 2 with LR values of the order of 10 ⁴ to 10 ⁵ , level 3 with LR values of the order of 10 ² to 10 ³ , and level 4 with LR values of the order of 10. For negative conclusions at levels 9 to 6, the LRs would be the inverse of those indicated for the positive conclusions at levels 1 to 4, respectively. For inconclusive results, the LR would be of order of (i.e. about) 1. These are not hard and fast numbers to be rigidly adhered to but orders of magnitude indicators of the expected ranges of LRs.

The forensic document examiner qualitatively assesses the probabilities (i.e. likelihood) that such (a type of) evidence as has been observed would occur given the one proposition (e.g. “written by”) and would occur given the alternative proposition (e.g. ”not written by”). Each qualitative probability must be the result of combining all plausible contributions from each sub-proposition within each main proposition. For example, within the “written by” (the writer of the specimens) main proposition may be the sub-propositions written by in normal writing style; written by in a disguised style; written by whilst affected by adverse writing conditions or by the effects of drugs/alcohol/medication; written by in an alternative writing style. The ratio of the probability of the observations in the main proposition to that in the alternative proposition gives the likelihood ratio all of which can be thought of qualitatively or estimated quantitatively.

The importance of this process is that all realistic explanations for the observations in both of the competing propositions are considered and assessed so that a reasoned conclusion (essentially on the assessed strength of the evidence) can be arrived at after starting at a neutral viewpoint. Reasons for the conclusion reached are expressed in the report and if required upon testifying in court.

This approach to reaching conclusions is described as the logical approach or evaluative method. The particular scale used by Strach Forensic closely parallels that used by the European Network of Forensic Handwriting Examiners (ENFHEX) ( https://enfsi.eu/about-enfsi/structure/working-groups/handwriting/ ) .

The main other scale of conclusions is that mostly used in North America (known as the SWGDOC [Scientific Working Group on Forensic Document Examination] scale ( https://swgdoc.org/ )). This is a nine point scale, with for example the strongest level 1 conclusion being worded along the lines of “Document A was written by the writer of the specimens” and level 2 “there is a strong (or high) probability that document A was written by the writer of the specimens”. From the perspective of the logical or evaluative approach to reaching conclusions, such expressions of conclusions on the probability that the document was written by a particular person are not ultimately justifiable as they would require an assumption of the prior probability which in most cases cannot be made. The prior probability is the probability that document A was written by the writer of the specimens before considering the handwriting evidence. This prior probability might be estimated from the population of possible writers but this would be a major assumption.