The use of handwriting examinations beyond the traditional court purpose

Accepted Manuscript The use of handwriting examinations beyond the traditional court purpose

Anna Agius, Kylie Jones, Rochelle Epple, Marie Morelato, Sébastien Moret, Scott Chadwick, Claude Roux PII: DOI: Reference:

S1355-0306(17)30057-6 doi: 10.1016/j.scijus.2017.05.001 SCIJUS 667

To appear in:

Science & Justice

Received date: Revised date: Accepted date:

15 December 2016 8 May 2017 12 May 2017

Please cite this article as: Anna Agius, Kylie Jones, Rochelle Epple, Marie Morelato, Sébastien Moret, Scott Chadwick, Claude Roux , The use of handwriting examinations beyond the traditional court purpose. The address for the corresponding author was captured as affiliation for all authors. Please check if appropriate. Scijus(2017), doi: 10.1016/j.scijus.2017.05.001

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT Journal: Science & Justice Proposition of: Professional article

The use of handwriting examinations beyond the traditional Court purpose Anna Agius 1, Kylie Jones 2, Rochelle Epple2, Marie Morelato1, Sébastien Moret1, Scott

PT

Chadwick1, Claude Roux1 1 University of Technology Sydney, Centre for Forensic Science, PO Box 123,

RI

Broadway, 2007, Australia

Australia

NU

* Corresponding Author:

AC C

ED

EP T

Author‘s contact details: Anna Agius: Kylie Jones: Rochelle Epple Marie Morelato: Sébastien Moret: Scott Chadwick: Claude Roux:

MA

Prof. Claude Roux University of Technology Sydney Centre for Forensic Science PO Box 123, Broadway 2007 Australia Phone: +61295141718 E-mail: [email protected]

SC

2 Forensics, Specialist Operations, Australian Federal Police, G.P.O. Box 401, Canberra, ACT,

[email protected] [email protected] [email protected] [email protected] [email protected] [email protected] [email protected]

ACCEPTED MANUSCRIPT Abstract Traditionally, forensic science has predominantly focused its resources and objectives on addressing court related questions. However, this view restricts the contribution of forensic science to one function and results in lost opportunities as investigative and intelligence roles are often overlooked.

MA

NU

SC

RI

PT

A change of perspective and expansion of the contributions of forensic science is required to take advantage of the benefits of abductive and inductive thought processes throughout the investigative and intelligence functions. One forensic discipline that has the potential to broaden its traditional focus is handwriting examination. Typically used in investigations that are focused on both criminal and civil cases, the examination procedure and outcome are time consuming and subjective, requiring a detailed study of the features of the handwriting in question. Traditionally, the major handwriting features exploited are characteristics that are often considered individual (or at least highly polymorphic) and habitual. However, handwriting can be considered as an information vector in an intelligence framework. One such example is the recognition of key elements related to the author’s native language. This paper discusses the traditional method generally used around the world and proposes a theoretical approach to expand the application of handwriting examination towards gaining additional information for intelligence purposes. This concept will be designed and tested in a future research project.

AC C

EP T

ED

Keywords: Forensic Science; Intelligence framework; Document examination; Intelligenceled policing

ACCEPTED MANUSCRIPT 1. Introduction

NU

SC

RI

PT

A document may be broadly defined as any material displaying marks, symbols or signs that communicate a message to another. The majority of documents consist of information written on paper by mechanical or electronic means, for example a typewriter or printer, or by hand using a writing instrument, like a pen or pencil [1]. However, many other objects and instruments can be employed to convey messages (e.g. rocks, leaves, graffiti on walls or lipstick on mirrors) or brandings (e.g. drug packaging, stamped or printed information on counterfeited objects) [2]. The examination of documents is a discipline that stemmed from the courts’ and investigations’ need for assistance in understanding evidence relating to the origin and history of questioned documents [1, 3]. The purpose of the analysis is to answer questions related to, but not limited to, the authorship, authenticity and/or source of the document. Questioned documents are those suspected of being false or with a disputed or unknown origin. Depending on the questions being asked, the examiner seeks not only to identify the author of the handwritten and/or printed matter, but additionally, analyse inks, papers and other elements of a document, as well as recognise additions or substitutions, restore erased writings and detect indentations [1].

AC C

EP T

ED

MA

A significant share of investigative questions relating to document examinations involves handwriting examinations. The purpose of this analysis is to identify the author of the handwriting on a questioned document, detect forged signatures and determine whether a common author produced a series of documents [4]. Handwriting is a complex perceptualmotor task, requiring the conscious and repetitive formation of letters and words in order to practice and reinforce the skill [1, 3]. Once this ability is developed, the writer redirects most of their conscious thought to the subject matter rather than the writing process. This skilled motor process results in the introduction and repetition of a writer’s own personal habits to the handwriting trace [5]. These writing characteristics were traditionally divided into two types; class characteristics and individual characteristics. The former can be defined as traits common to a group of writers that may be influenced by the writing system learnt and geography, while the latter represents the unique elements that discriminate writers [1, 3]. A person’s writing style stems from the adaptation of the prescribed writing system studied and it has been theorised that an understanding of the class characteristics induced by the teaching methods and systems taught, may be able to indicate the nationality and academic background of the writer [1]. However, due to the relatively recent shift towards electronic communication, paralleled with the lack of concern placed on penmanship, class characteristics have become less distinct and identifiable in present day handwriting [3]. Therefore, for the purpose of handwriting examinations, a person’s writing is comprised of a combination of individual characteristics. These characteristics are influenced by individual preferences, natural skill level and the amount of practice an individual undertakes to build and maintain the trace [1]. This results in differing combinations of handwriting characteristics among writers, which is referred to as the principle of individuality and has led to the assumption that no two people write exactly the same and no one person writes

ACCEPTED MANUSCRIPT the same word identically twice. Generally, these two fundamental ideals of handwriting are acknowledged when providing expert opinion with respect to the source of a piece of writing [1, 5].

EP T

ED

MA

NU

SC

RI

PT

The efficiency of forensic science as a whole has recently been called into question [6-8]. Generally, forensic case data from many disciplines of forensic science falls prey to poor integration into the investigation and intelligence processes [9]. This is largely due to the existing paradigm of forensic science focusing on court processes, which has been categorised as slow, susceptible to subjective opinions and inconsistent in terms of the interpretation of case data [8, 10]. The role of forensic science should thus be expanded and rather than using forensic case data solely to assist in prosecution, it can be analysed systematically to obtain knowledge about criminal activity and disrupt and prevent crime [11]. This approach is referred to as forensic intelligence [12], and is the “accurate, timely and useful product of logically processed forensic case data for investigation and/or intelligence purposes,” as defined by Ribaux et al. [9]. This new paradigm aims to move away from the current case-by-case, reactive law enforcement response, towards proactive intelligence-led policing [13, 14]. The transversal, adaptable nature of the forensic intelligence approach across different forensic science sub-disciplines is evidenced by its application in the area of document examination and illicit drug profiling [13, 15]. This involved the categorisation of false identity documents, such as identity cards, passports and stolen blank documents [14, 16-18]. Visual features, such as UV features, watermarks and the machine-readable zone, were extracted and used to establish a document profile. A comparison procedure was then developed and implemented to identify links and trends between cases. At this stage, we ought to ask the question: could this approach also be expanded to handwriting analysis? If this is possible, a redirection and extension of the contributions and goals of traditional handwriting examinations towards the provision of intelligence could also be possible.

AC C

Applying the forensic intelligence approach to handwriting examinations stems from an operational need. It is believed that handwriting can provide another layer of information about the object from which it originates. For instance, a security problem that this could help address is the “low volume-high frequency importation” of illicit drugs using the international mail system [19]. In Australia, the Australian Federal Police (AFP) is responsible for upholding the Commonwealth law and contributing to protecting Australians both domestic and overseas from criminal threats [20]. In 2012, it launched its National Forensic Rapid Laboratory (Rapid Lab). The Rapid Lab “allows different forensic disciplines to work on a seized item at the same time,” allowing for a quick turnaround time when analysing parcel post seizures [19]. The packages are triaged and examined to obtain useful information, including fingermarks, handwriting, chemical profile of the drug/s and potential place of origin. This information is then stored and compared with other seizures allowing the detection of potential links and trends between cases, uncovering the extent and volume of organised crime [19]. With respect to handwriting, analysts may have the potential to not

ACCEPTED MANUSCRIPT only link cases that appear to have been written by the same person, but also infer their native language, which may be useful as this intelligence product can be fed back to the originating country’s authorities. In conjunction with other traces, such as the physical characteristics of the parcel’s packaging or the chemical profile of the illicit substance within, it has the potential to provide important intelligence information. This may further differentiate a series of crimes or associate series of crimes, which were previously believed to be unrelated.

SC

RI

PT

As a way to progress the debate, this paper provides an overview of the traditional methods for handwriting examination, discussing the current limitations of this court-oriented discipline. A theoretical expansion for handwriting analysis is then proposed, which aims to present the advantages of expanding this practice’s objectives towards intelligence-led policing. This approach will be practically tested and pres ented using data in future research.

NU

2. Traditional role of handwriting examination

AC C

EP T

ED

MA

The process of handwriting examination is concerned with identifying the author of a questioned document where the court systems are the primary stakeholders of the opinion/s produced [4]. The identification of the source of an unknown (questioned) handwriting involves the comparison of its elements with the features of known material (referred to as specimen) obtained from suspect writers [2]. In Australia and New Zealand, the generalised procedure used to examine handwriting is maintained and updated by an editorial group overseen by the Australian and New Zealand Documents Specialist Advisory Group. It is based on an initial publication by Found et al. [21] the ‘Documentation of Forensic Handwriting Method: A Modular Approach’, and is generally accepted by practitioners in policing and government laboratories as the procedure for examinations undertaken in casework. Internationally, a number of procedures have been published as standards and used within specific regional areas or jurisdictions. These include publications by the European Network of Forensic Science Institutes (ENFSI), the ‘Best Practice Manual for the Examination of Handwriting’ [22], and the standard previously published by the American Society for Testing Materials (ASTM), the ‘Standard Guideline for Examination of Handwritten Items’ [23], which is currently under review for publication by the National Institute of Standards and Technology (NIST) Organisation of Scientific Area Committee (OSAC) for Forensic Document Examination. While different standards exist internationally, it should be noted that the procedures for the comparison of both handwriting and signatures are generally comparable. As with many forensic science disciplines based on comparative examination, the systematic and thorough approach to handwriting examination and identification stems from the broader analysis, comparison, evaluation and verification (ACE-V) framework [24]. In this approach, the strength of forensic findings is evaluated through the absence of meaningful differences and degree of correspondence between features of the documents at stake and

ACCEPTED MANUSCRIPT the probability of observing those features in another source. The conclusions are finally reported according to an opinion scale described by either; one of the discipline specific documents cited above, the non-discipline specific Australian forensic standards 5388.2 and 5388.3, or to the ENFSI guideline for evaluative reporting in forensic science, which promotes a likelihood ratio approach [22]. The next paragraphs will review the traditional process in more detail. 2.1 Analysis of handwriting

NU

SC

RI

PT

The analysis of handwriting involves employing the ‘general to the particular’ approach when observing the features of the questioned and specimen material. Firstly, the overall quality and quantity of all the material is assessed to ensure the two groups contain sufficient handwriting characteristics to enable the examination process to proceed and a proper comparison to take place [21, 22]. For example, initials or short signatures may not display enough features to enable the examiner to undertake an adequate comparison and reach a reliable opinion [21]. Fundamental, pictorial comparability between the questioned and specimen material should also be considered at this stage prior to the examination commencing.

AC C

EP T

ED

MA

As discussed by Found et al. [21], both the specimen and questioned material should then be assessed to ensure that only one writer is represented. Where extraneous entries that may have originated from a secondary writer (referred to as contamination from another writer) are identified, these should be excluded from the examination. From here, the intra writer variability of the material is studied and the scope of the variation is recorded for each piece of writing. Specimen material provided for comparison to the questioned handwriting should preferably be written within a comparable timeframe and be produced in a form (style, content and/or format) that is similar to that of the questioned document. In the profession of handwriting examination, letters are referred to as graphemes (an abstract representation of a letter) and words are composed of a string of graphemes. An allograph is a style of writing, in particular cursive, script and lettered [3]. It is paramount that the same allographs for a particular grapheme are compared (such as the printed character ‘g’) as the comparison of different allographs, like ‘G’ and ‘g’, is not possible as they symbolise two different motor memories [21]. Optimum questioned and specimen writings have common words or sentences within them as people often employ various forms of the same allograph depending on its position in the word. A range of characteristics, including pictorial and structural features are used to describe the handwriting [21]. The features of both the questioned and specimen material that may be assessed include: structural formation(s) used to produce each grapheme, the method of connecting letter to letter, writing fluency and line quality (indicative of skill level), slant, letter and word spacing, size and baseline alignment [22, 25, 26]. Therefore, once the examiner has concluded that a reasonable amount of comparable features between the

ACCEPTED MANUSCRIPT specimen and questioned writings are present, especially at the allographic level, the assessment of any significant similarities and/or differences is performed [5]. 2.2 Comparison of questioned and specimen handwriting

MA

NU

SC

RI

PT

The examiner progresses to comparing the various characteristics of the questioned and specimen handwriting. The comparison must not only seek to identify the presence of a combination of shared attributes, which is compulsory, but also an absence of fundamental differences, which is paramount [1]. Huber et al. [3] describes the characteristics that comprise the population of features that can be considered and categorised them into four types; elements of style, elements of execution, consistency and lateral expansion and word proportions. The following are some examples of features compared from a summary by Found et al. [21], including: placement of text, uniformity of margins, interline spacing, parallelism of lines, use of numerals and symbols in monetary amounts, class of allograph (cursive, script, lettering), connections between graphemes, feature proportion, slant, use of abbreviations, presence of embellishments, legibility, pen stops and lifts and line quality. Commonly, the examiner does this comparison task visually with the aid of a microscope. However, these features are subject to variation, both between writers (inter-writer variability) and between writings from the same author (intra-writer variability) [21]. Technologies exist and research has been conducted around the world into the possible use of databases, software, algorithms and codebook approaches to assist with this process and improve the efficiency of examiners [25, 27-29].

AC C

EP T

ED

One of the motivations behind the development of technologies to assist handwriting examiners stems from the theory of inter-writer variation. This notion suggests that given a group of handwritings were randomly selected from the general population, the chance of finding writings that display exactly the same combinations of features is very low [21]. However, there is a lack of experimental data to support this proposal, in addition to the absence of quantitative comparison techniques. This has resulted in criticism of the comparison stage, as it is heavily based on the examiner’s training and experience, with the result being that of their opinion [21]. This principle of individuality derived from the concept of inter-writer variation was studied by Zhang et al. [28] by testing over 12,000 word images from 3000 handwritings from 1000 participants in the US. The authors selected four different handwritten words, ‘been’, ‘Cohen’, ‘Medical’ and ‘referred’, and extracted the gradient, structural and concavity features from images taken of the words. The results indicated that 83% of the writers were correctly identified when all four words were considered. Using the four words was found to be more discriminating than using a single word or individual characters. Numerous studies have reported the features that were found to be the most discriminating between individuals. A survey conducted by Hanlen et al. [30] sought to compose a list of the handwriting features used by forensic handwriting examiners during comparisons and their discriminatory power. From seventy-six respondents, a list of the top eleven features

ACCEPTED MANUSCRIPT

AC C

EP T

ED

MA

NU

SC

RI

PT

was formed, which represented those characteristics that more than 94% of the responders stated were always used and possessed a medium to high discriminatory power. These included: letter design/formation, connecting strokes, beginning/ending strokes, line quality, writing quality (legibility), hesitation, pen lifts, intra-writer variation, size relative within and between characters and spacing between characters. Ling [25] employed image processing software to perform detailed measurements of letters and spacing. From ten different measurements taken from the handwriting specimens, the most discriminating dimensions were the height of the middle zone (the main body of the letter), height of the letter ‘o’ and the slope. The shape of the loop of the grapheme, o, was another feature of the letter targeted for analysis by Marquis et al. [31, 32], which prompted further research into the loop shapes of other characters, including a, d and p [32]. A method based on Fourier descriptors was applied to the contours of the characters, which provided a morphological characterisation of the enclosed contours. This information was extracted using an automated image analysis procedure. The authors displayed, in a quantitative and objective way, that the loop shape of the letter d was the most discriminant characteristic amongst the thirteen participants that were tested, with a correct classification rate of 82.4%. Given the correct classification rates achieved, this study indicated that carrying out an analysis of a small portion of an individual’s handwriting containing a limited number of features is possible [32]. The combination of ‘th’ was also investigated by Pervouchine et al. [33], in addition to d, y and f, however, to decrease variation, d and y were only examined when they were at the terminal position of words. The features of the ‘th’ combination, like height, width and distance between ‘t’ and ‘h’, provided a significantly greater discriminating power than the other characters. Indeed, features such as final stroke angle, fissure angle and slant of f-stem were deemed irrelevant for writer discrimination. However, given the absence of such quantitative comparison techniques in practice, the examiner must provide an opinion regarding the similarities and dissimilarities between the questioned and specimen writings [21]. This presents some limitations that are described further below. 2.3 Evaluation and verification of questioned and specimen handwriting The evaluation process involves determining the significance of the similarity or difference of each feature between the questioned and specimen materials and proposing an explanation for their occurrence. The examiner’s aim is to weigh up the opposing propositions of observing similarities/differences between the questioned and specimen materials given that they were written by the same author or by different authors [3, 34]. The subjective nature of the process does attract some criticism, from those who are sceptical of the reliability of the practice [3]. As Found et al. [21] stated, handwriting examination is “subjective as it is based on the perceptual and cognitive processes of an examiner” and the results of the examination are conveyed as opinions. Research testing whether forensic handwriting examiners have superior skills at expressing opinions than laypersons have been conducted [35], confirming the concept that handwriting examination

ACCEPTED MANUSCRIPT is a field of expertise. Nonetheless, the call for empirical evidence to substantiate the validity of the examiner’s opinions has been expressed [36]. Further, in our opinion, research in this area should go beyond studies testing the expertise or quantifying error rates in handwriting examination, but should also investigate handwriting features and their statistical variability within and between writers. This is supported by Johnson et al. [37] who highlighted that handwriting examination is complex to analyse statistically due to the limited studies about background frequencies and occurrences of features .

ED

MA

NU

SC

RI

PT

With the advances in computing technology, some researchers have developed methods for quantifying handwriting features, which have the potential, in the future, for identification, evaluation and verification tasks to be achieved [27, 31, 34, 38]. Recently, statistical techniques have been applied to the relevant handwriting features to assess their evidential value, which were selected due to the development of standardised and automated extraction programs. For example, Taroni et al. [39] proposed a likelihood ratio approach to communicate the value of the handwriting features. Several measurements of the loops of handwritten characters were taken from the questioned (recovered data) and known (control data) specimens, and then applied this two-level model, which assumed nonconstant intra-writer variability. The questioned handwriting was compared with both the, writings of the author known to have written the questioned document, and specimens taken from authors whom it was known did not write the questioned material. Upon examination of the distribution of the likelihood ratio generated following the comparison of characters written by the same or different writers, they could conclude that their results correctly advocated the hypothesis of interest. Hence, this approach may be useful as a way of assessing the value of the evidence from a comparative examination point-of-view [39].

AC C

EP T

At the end of the evaluative step, the confidence of an examiner’s opinion is conveyed via an opinion or conclusion scale [21, 22, 25]. Broadly speaking, the scale used by practitioners is made up of at least two levels of opinion either side of an inconclusive result. One side describes different degrees of confidence supporting the proposition that the questioned material was written by the writer of the specimen material, and the other side supports the proposition that the questioned handwriting was written by another author. In addition to the traditional evaluative level, Taroni et al. [39] studied another objective of forensic science, which they referred to as the investigative level. The investigative stage involves providing police investigators with information that supports more general propositions regarding the events that led to the detected trace. For example, in a case where there are no specimens, the evaluation process cannot occur. However, potential, broader hypotheses were proposed by the authors that may be tested instead, such as ‘the author of a questioned document is a male (or a female)’ or ‘the author of a questioned document is (or is not) a left handed writer’ [39]. The two-level model employed in this study reflects an expansion in the role and objectives of forensic science disciplines from the

ACCEPTED MANUSCRIPT production of evidence for purely court purposes to intelligence-led policing, referred to as forensic intelligence [12].

SC

RI

PT

At the conclusion of a handwriting or signature comparison, preferred practice is that, a verification process takes place. This involves a quality check of the examination and opinion/s by another independent, qualified assessor, whereby a second ACE examination is carried out [24]. However, as undertaken in some other forensic disciplines such as fingermark comparisons, due to the size and complexity of handwriting comparisons, a complete ACE examination is seldom performed during this verification step. Instead, the extent to which the verification process is fulfilled varies between jurisdictions and at the very least, most laboratories have defined quality assurance processes in place, which include technical reviews of casework and an expectation that examiners keep their knowledge and skills up-to-date with continual training and proficiency testing [22]. 3. Possibility of using handwriting examination for intelligence purposes

NU

3.1 The paradigm shift towards forensic intelligence

AC C

EP T

ED

MA

Margot proposed a definition of forensic science that is quintessential - “Forensic science is the study of crime and its traces. These are silent witnesses that need to be detected, seen, and understood to make reasonable inferences about criminal phenomena, investigation or demonstration for intelligence, investigation and court purposes” [40]. This definition conveys the broad range of informational processes that traces can contribute to in order to assist decision-making, such as from an investigative and forensic intelligence perspective, and provide the courts with information that they may perceive as evidence [12, 40]. However, across the board, the default position of forensic science is to employ or evaluate particular traces in a manner that meets the expectations of the courts only. This continuous demand of judicial authorities appears to have exhausted the majority of forensic laboratory resources, which has prompted a tendency for forensic science to distance itself from participation in the intelligence and investigation aspects [12]. Similarly, the issues plaguing the current structure of policing are consequences of the structure of traditional law enforcement, such as the case-by-case focus, the division of information and responsibilities between departments and the lack of uniformity in data management [13]. This indicates that a change is required to expand the role of forensic science beyond the production of evidence for the courts. Forensic case data from many disciplines of forensic science falls prey to poor integration into the investigation and intelligence processes. Hindsight has resulted in the realisation that the information needed to guide the investigation was often in the case files and had it been exploited, would have led to the proactive and timely resolution of the case or a better understanding of the criminal activity [9]. For example, Ribaux et al. [41] analysed the modus operandi and time/space data from 10,000 burglaries in four adjacent regions in Switzerland using an intelligence-based approach. A new series of burglaries was detected

ACCEPTED MANUSCRIPT by linking several seemingly individual cases based on corresponding shoe marks and modus operandi. The information describing this series was compared with that of other, previous series already inputted into a memory. This provided a tenuous link between the new series and a previous one, whereby the previous series had already a list of possible suspects, which could then have been used as a selection of suspects for the new series [41].

AC C

EP T

ED

MA

NU

SC

RI

PT

Morelato et al. and Baechler et al. [13, 15] developed a general and multi-commodity model that could guide the use of any forensic case data in an intelligence-led perspective. Its transversal nature is thus relevant to this study. Indeed, the model accommodates all types of traces, including handwritten or printed documents. A trace is defined by Margot [40] as a “remnant of a source and/or activity that produced it”. From the perspective of a forensic scientist, the trace (or effect) must be analysed and deciphered to determine its cause, which may be either a source or act that led to its existence. A concept introduced by Morelato et al. [13] referred to as the forensic intelligence rationale evolves this ideal a step further. It assumes that, if “similarities between the comparable features of different objects” are detected, it may indicate that these features are “repetitive effects of the same cause or type of cause” [13]. In order to assume a common cause, it relies on the development of a hypothesis that is capable of accounting for the similarities between the objects. Previously, the traditional, probative approach focused on seeking differences between the features of two objects and attempting to quantify these differences, whereas, the modernised, intelligence approach aims to seek the proximity between two objects or a deviation in data [13]. Granted, there is the possibility that multiple hypotheses are equally adequate in explaining the cause of the effect, which then introduces a degree of uncertainty, as the conclusion cannot be considered definitive. However, this uncertainty does not pose a significant problem from an intelligence perspective as it may from an investigative or court perspective. For example, similarities between the features of false identity documents may either be a result of their production by the same counterfeiter and equipment used or by different criminals using the same manufacturing method [13]. Further information on this approach to false identity documents can be found in the literature [14, 16-18]. Following the model developed by Morelato, Baechler et al. in 2014 and 2015 [13, 15], if we consider the handwritten document an object1, the characteristics of the handwriting (e.g. spatial and construction characteristics) can be extracted to transform the object into an entity called a profile. Once representative, reliable and reproducible characteristics are extracted to form the profile, comparison metrics can then be used to compare the profiles. Groups or trends can then be inferred after analysis. This handwriting data must then be combined with alternative information to produce findings, hypotheses, suggestions or recommendations and communicate them to decision makers in order for them to have an impact on the criminal environment [13, 15]. It is hypothesised that forensic intelligence has 1

An object is any type of trace related to a crime or security problem [13].

ACCEPTED MANUSCRIPT the potential to expand the contributions of handwriting examination from purely the production of evidence for the court to also include crime analysis processes. 3.2 Expanding handwriting examination towards forensic intelligence purposes

MA

NU

SC

RI

PT

Sometimes the quality of the trace and/or circumstances of the case are such that the trace cannot be exploited as the sole evidential proof of identity. However, this is not a necessity in an intelligence context whereby the trace may be used simply to obtain complimentary, non-identification information about the author, such as their potential country of origin. In particular, gain knowledge about the country of origin of the writer through their English handwriting, given they have learnt English as a second language. To understand the potential of this approach and to test it further, it is necessary to acquire some knowledge about alphabets and handwriting learning systems. Huber et al. [3] provided an overview of the formation of the alphabet over time and suggested multiple theories of the origin of the first systematic method for written communication. It is speculated that the first indications of an alphabet arose out of Egypt and Mesopotamia around 1600 B.C.E. Over thousands of years of adaptation, the Romans adopted a derivation of the Greek alphabet, which became the current twenty-six-letter alphabet [3]. Once the alphabet became established, the Romans began developing new letter forms, the original was called lapidary capitals, and since then hundreds of different writing systems have been created and taught throughout different English speaking countries [3].

AC C

EP T

ED

For the purpose of this paper, Australian writing systems taught in schools will be the focus. A list of different systems taught in North America can be found in the publication by Huber et al. [3]. In Australia, there are currently five different handwriting systems taught; New South Wales (NSW) and the Australian Capital Territory (ACT) use the NSW Foundation Style; Victoria, Western Australia (WA) and the Northern Territory (NT) use Victoria Modern Cursive; Queensland uses Queensland Modern Cursive; South Australia (SA) uses SA Modern Cursive and Tasmania uses Tasmania Handwriting Style [42]. Hilton referred to the features that differentiate the writing systems, inducing class characteristics within the writing of the students as national characteristics [1]. The product of this is the potential ability to distinguish between the attributes of writing systems of other countries [3]. Hilton suggested that “writing systems [were] as nationalistic as languages,” proposing that foreign ‘accents’ were present in the writings of individuals who received their early education in their native land yet no longer lived there [1]. For example, a specimen of the English writing of a German educated writer displayed a method of crossing and connecting the ‘t’ to the subsequent letter, which reflected the characteristic forms of German writing systems [1]. As a consequence, the ability to detect foreign features in English handwriting that are indicative of a particular nationality may be possible [1, 43, 44]. However, nowadays there are numerous writing systems taught in the different states or regions of the one country. Therefore, it may be more accurate to refer to the class characteristics within a group of people that have been taught from the same copybook system as regional characteristics.

ACCEPTED MANUSCRIPT

ED

MA

NU

SC

RI

PT

The idea of detecting foreign accents in English handwriting (or any other second language) is not a new concept as research into this has been published in the literature. Accents in handwriting are defined by Ramaiah et al. [43] as, “the influence of a writer’s native script on his/her writing style in another script”. Cheng et al. [44] studied the different accents present in the English handwriting of the three main racial groups in Singapore: Chinese, Malay and Indian. They identified six class characteristics that were significant in differentiating between the racial groups as their presence could be attributed to the habits of writing their own native language. For example, in constructing Chinese characters, fundamental rules dictating the sequence of the strokes are adhered to, which results in the strokes being written in a descending fashion. The study found that all the Chinese writers consistently wrote the horizontal stroke of the ‘T’ prior to the vertical stroke. This particular letter formation is influenced by the conventions of the construction of Chinese characters. Ramaiah et al. [43] proposed a method for accent detection in English handwriting that involved distinguishing between native and non-native writers. The authors investigated two populations, one group that could only write English and the other whom were multilingual writers and could write English. The multilingual group was composed of writers whose native Indian script varied from writer to writer. They first gained an understanding of the different Indian writing styles and characterised their features by analysing contour direction distribution and fractal, structural and concavity features. This pilot research into accent detection revealed strong indications of the existence of accents in handwriting. However, further studies are required to overcome the problem of the variability between writing systems and progress to accent identification to recognise the writer’s native script from their English handwriting.

AC C

EP T

The development of a handwriting analysis procedure for intelligence purposes, whereby the English handwriting of people is examined to detect their native language, may not involve the comparison of unknown specimen writings with known documents . Therefore, the parameters of the traditional comparison stage require adjustments to become less focused on individual, habitual features and directed more towards extracting features common amongst a group of people (i.e. regional characteristics). A parallel can be drawn from the study by Baechler et al. [15], which presented a forensic intelligence model that was transversal across different trace types. The authors demonstrated its transversality by investigating illicit drugs and false identity documents. Although they are entirely different traces, the same criteria can be used to select the relevant features to extract from handwritten entries. The aim is to identify the features that enable the examiner to associate entries/documents from the same source and differentiate those from different sources. In order to determine the appropriate handwriting features to extract, the typical ‘general to the particular’ ideal of forensic science provides a good basis. As this notion is relatively embryonic, all possible features must be considered, even those that appear largely generic. A range of features should be considered including those of the writer (e.g. their native

ACCEPTED MANUSCRIPT

MA

NU

SC

RI

PT

language) and those of the handwriting, such as spatial (e.g. slant, height relationships, number of letters per line and area of the text) and construction characteristics (e.g. pen direction and number of strokes used to form a particular character). Indeed, a consideration of the potential class characteristics that may have been induc ed from the native alphabet of the writer may be required. A study by Turnbull et al. [45], led to the identification of the class characteristics in the English handwriting of Polish people. Using these class characteristics that were purported to have been influenced by the English copybook system taught in Poland, they were capable of distinguishing Polish writers from English. This research involved acquiring extensive knowledge of the English copybook systems taught in Poland. It provided insight into the patterns of adherence and departure from taught writing systems, and approaches to extracting this alternate information about the author for intelligence-led policing frameworks. It is recognised that theorising the potential country of origin of an unknown author is based on abductive thought processes. Abductive reasoning infers some conclusions about the particular case using observed handwriting features and knowledge about writing systems, hence a greater margin of uncertainty is a by-product [46]. Nonetheless, as explained by Hilton, this additional information must be regarded as an approximation and an investigative tool to be employed in another role of forensic science [1].

EP T

ED

Practically, handwriting from people of different countries, who have learnt English as a first or second language should be obtained. Features of interest can then be extracted to form a document profile. These profiles should be compared to see if they can reflect and characterise the origin of the writer. The outcomes of extracting and comparing profiles would provide a preliminary indication of the possibility of obtaining knowledge of the writer’s native country from their English handwriting within an intelligence context. In order to provide valuable information, a large writer population that contains a number of countries should be analysed. For this approach to be considered, the feature extraction process should become automated.

AC C

This type of information could be exploited at mail triage centres, such as the AFP’s Rapid Lab, where a parcel or letter is identified as containing illicit material or upon receipt of handwritten, threatening correspondence. The Rapid Lab has the capacity to analyse hig hvolumes of parcel post using a multi-disciplinary, real-time approach [47]. The handwriting combined with other forensic traces (such as fingermarks and drug profiles) may help to greatly improve authorities’ understanding of the criminal activity and its potential structure. An example of this was discussed by Horne et al. [47], whereby the AFP were investigating Silk Road drug traffickers using the international mail stream. They deconstructed and examined letter-sized envelopes that contained controlled substances that were coming from Canada. The information recorded ranged from the physical properties of the packaging (i.e. packaging stamps from a single post office), the method of drug concealment, fingermarks and the handwriting (i.e. addresses and consignment notes). Using the combination of these traces, links were made between multiple packages, which

ACCEPTED MANUSCRIPT suggested a criminal syndicate was orchestrating the shipments. The AFP were able to provide the Royal Canadian Mounted Police (RCMP) with a forensic intelligence product in a timely manner to aid in the identification of the syndicate and their location, whilst also locating the purchasers in Australia [47]. As a consequence, modifying the systematic use of handwriting to determine the country of origin of the writer for operational and strategic intelligence would benefit a rapid laboratory style set-up and law enforcement agencies in general. The two approaches are illustrated in Figure 1.

ED

MA

NU

SC

RI

PT

This expansion of the discipline has the potential to transform the contribution of handwriting analysis, which can be characterised as a time consuming and subjective process that is generally carried out during an investigation for the purpose of judicial proceedings. This expanded function would be complementary to traditional authorship examinations, providing intelligence information about the involvement of particular groups of peoples during the investigation or crime analysis. Indeed, collating information about the origin of a writer combined with other forensic or circumstantial information could provide a strategic view of organised crime. If this is shown to be possible, this type of approach could focus law enforcement resources towards specific cases, people or groups of people, which would be beneficial to the investigation. In turn, a traditional handwriting examination could then be applied to identify a particular individual once specimen material has become available. It is believed such an expansion is necessary to help improve forensic science effectiveness in the end-to-end process. Research is currently underway to deliver usable statistical data that aims to demonstrate the practical aspects of the proposed approach.

4. Conclusions

EP T

<>

AC C

The current situation in handwriting examination is described above. More importantly, an additional use of handwriting trace is proposed and discussed. The traditional handwriting analysis procedure (i.e. ACE-V process) was developed in response to the needs of the court and is focused on extracting individual characteristics to identify the author of a piece of writing. Whilst this evidence can be of high value in regards to providing probative information to validate or refute a hypothesis, a full examination is generally not conducive to quick turnarounds and specimens (or comparison materials) are mandatory. From an intelligence-led policing perspective, the generation of information in a timely manner is paramount for the information to be valuable and actionable (i.e. influence decisions). As the traditional comparison stage of the ACE-V process is seen as excessive in an intelligence setting, the duration of a handwriting examination can be reduced and in part, automated. However, this is only possible if handwriting is recognised as an information vector that can feed a variety of processes within a policing and security system. It would thus be

ACCEPTED MANUSCRIPT interesting to gain an understanding of the depth of information obtainable from handwriting, including the plausibility of identifying the origin of a writer from their Engl ish handwriting. Research in this direction is underway in our organisations and will be published in the future. Overall, at a time when the effectiveness of forensic science is being questioned, it is believed the proposed expansion is a valued addition to the forensic end-to-end process. Why shouldn’t we explore it further?

PT

5. Caption

RI

Figure 1 Illustration of the traditional and proposed approaches of handwriting examination. 6. References

5.

6. 7.

8.

9.

10.

11. 12.

13.

SC

NU

MA

4.

ED

3.

EP T

2.

Hilton, O., Scientific Examination of Questioned Documents. 1993, c1992, Boca Raton, FL: CRC Press. Kelly, J.S. & B.S. Lindblom, What is Forensic Document Examination?, in Scientific Examination of Questioned Documents. 2006, CRC/Taylor & Francis: Boca Raton, FL. Huber, R.A. & A.M. Headrick, Handwriting Identification: Facts & Fundamentals. 1999, Boca Raton, NY: CRC Press. Ellen, D., Handwriting: The Variations Between Normal Writings, in Scientific Examination of Documents. 2005, CRC Press. p. 11-34. Harrison, D., Burkes, T.M. & Seiger, D.P., Handwriting Examination: Meeting the Challenges of Science and the Law. Forensic Science Communications, 2009. 11(4): p. https://archives.fbi.gov/archives/about-us/lab/forensic-sciencecommunications/fsc/oct2009/review/2009_10_review02.htm (last accessed November 16, 2016). Robertson, J., Forensic science, an enabler or dis-enabler for criminal investigation? Australian Journal of Forensic Sciences, 2012. 44(1): p. 83-91. Julian, R.D., Kelty, S.F., Roux, C., Woodman, P., Robertson, J., Davey, A., Hayes, R., Margot, P., Ross, A., Sibly, H. & White, R., What is the Value of Forensic Science? An Overview of the Effectiveness of Forensic Science in the Australian Criminal Justice System Project. Australian Journal of Forensic Sciences, 2011. 43(4): p. 217-229. Committee on Identifying the Needs of the Forensic Sciences Community and National Research Council, Strengthening Forensic Science in the United States: A Path Forward. 2009, National Academies Press: Washington, DC. Ribaux, O., Walsh, S.J. & Margot, P., The contribution of forensic science to crime analysis and investigation: Forensic intelligence. Forensic Science International, 2006. 156(2): p. 171181. Ribaux, O., Roux, C. & Crispino, F., Expressing the value of forensic science in policing. Australian Journal of Forensic Sciences, 2016: p. 1-13. http://dx.doi.org/10.1080/00450618.2016.1229816. Ribaux, O., Crispino, F., Delemont, O., & Roux, C., The progressive opening of forensic science toward criminological concerns. Security Journal, 2015: p. 1-18. Ribaux, O., Baylon, A., Roux, C., Delemont, O., Lock, E., Zingg, C. & Margot, P., Intelligenceled crime scene processing. Part I: Forensic intelligence. Forensic Science International, 2010. 195(1): p. 10-16. Morelato, M., Baechler, S., Ribaux, O., Beavis, A., Tahtouh, M., Kirkbride, P., Roux, C. & Margot, P., Forensic intelligence framework - Part I: Induction of a transversal model by

AC C

1.

ACCEPTED MANUSCRIPT

15.

16.

22. 23. 24. 25. 26. 27. 28.

29.

30.

31.

NU

MA

21.

ED

20.

EP T

19.

AC C

18.

SC

RI

17.

PT

14.

comparing illicit drugs and false identity documents monitoring. Forensic Science International, 2014. 236: p. 181-190. Baechler, S., Ribaux, O. & Margot, P., 2012 Student Paper: Toward a Novel Forensic Intelligence Model: Systematic Profiling of False Identity Documents. Forensic Science Policy & Management: An International Journey, 2012. 3(2): p. 70-84. Baechler, S., Morelato, M., Ribaux, O., Beavis, A., Tahtouh, M., Kirkbride, K. P., Esseiva, P., Margot, P. & Roux, C., Forensic intelligence framework. Part II: Study of the main generic building blocks and challenges through the examples of illicit drugs and false identity documents monitoring. Forensic science international, 2015. 250: p. 44-52. Baechler, S., Terrasse, V., Pujol, J. P., Fritz, T., Ribaux, O. & Margot, P., The systematic profiling of false identity documents: method validation and performance evaluation using seizures known to originate from common and different sources. Forensic science international, 2013. 232(1-3): p. 180-90. Auberson, M., Baechler, S., Zasso, M., Genessay, T., Patiny, L. & Esseiva, P., Development of a systematic computer vision-based method to analyse and compare images of false identity documents for forensic intelligence purposes-Part I: Acquisition, calibration and validation issues. Forensic science international, 2016. 260: p. 74-84. Talbot-Wright, B., Baechler, S., Morelato, M., Ribaux, O. & Roux, C., Image processing of false identity documents for forensic intelligence. Forensic science international, 2016. 263: p. 67-73. Australian Federal Police (AFP), New initiative brings rapid results, in AFP Platypus Magazine. 2013: Canberra, Australia. Australian Federal Police (AFP). Our organisation. 2016 1/5/2017]; Available from: https://www.afp.gov.au/about-us/our-organisation. Found, B., Rogers, D. & Bird, C. (Eds.), Documentation of Forensic Handwriting Method: A Modular Approach. Journal of Forensic Document Examination, 2012. 12: p. 1-90. European Network of Forensic Science Institutes (ENFSI), Best Practice Manual Forensic Examination of Handwriting. 2015. American Society for Testing Materials (ASTM) E2290-07a, Standard Guide for Examination of Handwritten Items. Withdrawn 2016, ASTM International: West Conshohocken, PA. Tierney, L., Analysis, Comparison, Evaluation, and Verification (ACE-V), in Encyclopedia of Forensic Sciences, J.A.S.J.S.M. Houck, Editor. 2013, Academic Press: Waltham. p. 69-73. Ling, S., A preliminary investigation into handwriting examination by multiple measurements of letters and spacing. Forensic Science International, 2002. 126(2): p. 145-149. Ahola, N.M., Classification and Frequency of Occurrence of Specific Number Styles. Canadian Society of Forensic Science Journal, 2000. 33(1): p. 13-22. Bensefia, A., Paquet, T. & Heutte, L., A writer identification and verification system. Pattern Recognition Letters, 2005. 26(13): p. 2080-2092. Zhang, B. & S.N. Srihari. Analysis of handwriting individuality using word features. in Document Analysis and Recognition, 2003. Proceedings. Seventh International Conference on. 2003. Khalifa, E., Al-maadeed, S., Tahir, M.A., Bouridane, A. & Jamshed, A., Off-line writer identification using an ensemble of grapheme codebook features. Pattern Recognition Letters, 2015. 59: p. 18-25. Hanlen, R., Manzolillo, P., Muehlberger, R. & Sperry, G., Survey of Forensic Document Examination Habit Areas: Degree of Use and Discriminatory Power. 1999, Pacific Northwest National Laboratory. p. 1-40. Marquis, R., Schmittbuhl, M., Mazzella, W.D. & Taroni, F., Quantification of the shape of handwritten characters: a step to objective discrimination between writers based on the study of the capital character O. Forensic Science International, 2005. 150(1): p. 23-32.

ACCEPTED MANUSCRIPT

38. 39.

40. 41.

42. 43. 44.

45.

46. 47.

PT

RI

SC

37.

NU

36.

MA

35.

ED

34.

EP T

33.

Marquis, R., Taroni, F., Bozza, S. & Schmittbuhl, M., Quantitative characterisation of morphological polymorphism of handwritten characters loops. Forensic Science International, 2006. 164(2): p. 211-220. Pervouchine, V. & G. Leedham, Extraction and analysis of forensic document examiner features used for writer identification. Pattern Recognition, 2007. 40(3): p. 1004-1013. Marquis, R., Bozza, S., Schmittbuhl, M. & Taroni, F., Handwriting Evidence Evaluation Based on the Shape of Characters: Application of Multivariate Likelihood Ratios. Journal of Forensic Sciences, 2011. 56(1): p. 238-242. Found, B. & D. Rogers, The probative character of Forensic Handwriting Examiners' identification and elimination opinions on questioned signatures. Forensic Science International, 2008. 178(1): p. 54-60. Risinger, D.M., Denbeaux, M.P. & Saks, M.J., Exorcism of Ignorance as a Proxy for Rational Knowledge: The Lessons of Handwriting Identification "Expertise". University of Pennsylvania Law Review, 1989. 137(3): p. 731-792. Johnson, M.E., Vastrick, T.W., Boulanger, M. & Schuetzner, E. , Measuring the Frequency Occurrence of Handwriting and Hand-Printing Characteristics, in NIJ Award 2010-DN-BXK273, N.A. 2010-DN-BX-K273, Editor. 2010, University of Central Florida. Plamondon, R. & G. Lorette, Automatic signature verification and writer identification — the state of the art. Pattern Recognition, 1989. 22(2): p. 107-131. Taroni, F., Marquis, R., Schmittbuhl, M., Biedermann, A., Thiery, A. & Bozza, S., The use of the likelihood ratio for evaluative and investigative purposes in comparative forensic handwriting examination. Forensic Science International, 2012. 214(1): p. 189-194. Margot, P., Forensic science on trial - What is the law of the land? Australian Journal of Forensic Sciences, 2011. 43(2-3): p. 89-103. Ribaux, O. & P. Margot, Inference structures for crime analysis and intelligence: the example of burglary using forensic science data. Forensic Science International, 1999. 100(3): p. 193210. Australian School Fonts. School Handwriting Styles - Australia & New Zealand. 2016 1/2/16]; Available from: http://www.australianschoolfonts.com.au/styles/. Ramaiah, C., Porwal, U. & Govindaraju, V., Accent Detection in Handwriting based on Writing Styles. 2012, Tenth International Workshop on Document Analysis Systems: Buffalo, NY. Cheng, N., Lee, G.K., Yap, B.S., Lee, L.T., Tan, S.K. & Tan, K.P., Investigation of Class Characteristics in English Handwriting of the Three Main Racial Groups: Chinese, Malay and Indian in Singapore. Journal of Forensic Science, 2005. 50(1): p. 1-8. Turnbull, S.J., Jones, A.E. & Allen, M. , Identification of the Class Characteristics in the Handwriting of Polish People Writing in English. Journal of Forensic Science, 2010. 55(5): p. 1296-1303. Morelato, M., Forensic drug profiling: a tool for intelligence-led policing. 2015, University of Technology Sydney, PhD Thesis. Horne, N., Edmondson, K., Harrison, M. & Scott, B., The applied use of forensic intelligence for community and organised crime. Australian Journal of Forensic Sciences, 2015. 47(1): p. 72-82. http://dx.doi.org/10.1080/00450618.2014.916755.

AC C

32.

ACCEPTED MANUSCRIPT Figure 1

Detect & collect trace No k nown comparison material available

Known comparison material available

Profile questioned

specimen handwriting

handwriting

Compare questioned &

Compare profiles of

specimen handwriting

questioned handwritings

MA

NU

SC

RI

PT

Analyse questioned &

Provide opinion

Links & trends between

questioned handwriting Traditional handwriting

AC C

EP T

examination

ED

regarding authorship of

handwriting

Intelligence & Investigation

ACCEPTED MANUSCRIPT Highlights

PT RI SC NU MA ED EP T

 

Expansion of the application of handwriting examination beyond the Court purpose Handwriting as an information vector feeding many processes in policing and security Potential recognition of key elements related to the author’s native language Improved effectiveness in the forensic end-to-end process

AC C

 