You are on page 1of 18

Australian Journal of Forensic Sciences

ISSN: 0045-0618 (Print) 1834-562X (Online) Journal homepage: http://www.tandfonline.com/loi/tajf20

Empirical investigation of biometric, non-visible,


intra-signature features in known and simulated
signatures

Tahnee N. Dewhurst, Kaye N. Ballantyne & Bryan Found

To cite this article: Tahnee N. Dewhurst, Kaye N. Ballantyne & Bryan Found (2016)
Empirical investigation of biometric, non-visible, intra-signature features in known
and simulated signatures, Australian Journal of Forensic Sciences, 48:6, 659-675, DOI:
10.1080/00450618.2015.1126637

To link to this article: http://dx.doi.org/10.1080/00450618.2015.1126637

Published online: 22 Jan 2016.

Submit your article to this journal

Article views: 133

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at


http://www.tandfonline.com/action/journalInformation?journalCode=tajf20

Download by: [200.32.85.10] Date: 07 November 2016, At: 02:12


Australian Journal of Forensic Sciences, 2016
Vol. 48, No. 6, 659675, http://dx.doi.org/10.1080/00450618.2015.1126637

Empirical investigation of biometric, non-visible, intra-signature


features in known and simulated signatures
Tahnee N. Dewhursta,b*, Kaye N. Ballantyneb,c and Bryan Founda,b,d
a
Department of Human Biosciences, La Trobe University, Victoria, Australia; bForensic Services
Department, Victoria Police, Victoria, Australia; cSchool of Psychology and Public Health, La
Trobe University, Victoria, Australia; dFaculty of Law, University of New South Wales, New
South Wales, Australia
(Received 23 August 2015; accepted 15 November 2015)

With the continued proliferation of electronic point of transaction signature record-


ing devices, research into biometric indicators of spurious handwriting is attracting
increasing interest. Whilst many investigations have focused on the static and
dynamic indications of known or spurious behaviour in handwriting, little empirical
research is available regarding the identication of such writings through the analy-
sis and comparison of non-visible, intra-signature, kinematic parameters. These fea-
tures are associated with segments within a signature formation where the pen is
momentarily lifted from the page, such as might occur between a rst and last
name, or when the pen is lifted from the page for the purpose of crossing a t or
dotting an i. It is postulated that this type of feature analysis may be of value in
support of examiners static examinations of disputed writings, and subsequent for-
mation of opinion as to genuineness or otherwise. To investigate this, 13 skilled
writers generated 195 known signature formations, which were then simulated 1560
times by eight simulators. All signatures, known and simulated, were simultaneously
captured both statically and dynamically. The presence of non-visible features in the
known signatures was recorded, analysed and compared with the prevalence of simi-
lar features in the simulation attempts. The duration, absolute size, straightness error
and jerk (disuency measure) of the extracted segments were examined and com-
pared, with the result that simulated signatures showed an increase in all the above
parameters, compared with the known signatures. Furthermore, the visualised repre-
sentation of the non-visible, intra-signature segments illustrated an overall gross pic-
torial disparity between the known and simulated signatures, which may be of use
during rst pass authenticity examinations.
Keywords: signatures; handwriting; simulation; forgery; behavioural biometrics;
validation; questioned document examination

Introduction
Historically, the vast majority of research regarding signature formations has concen-
trated on the examination of static off-line traces, predominantly focusing on key aspects
as documented by past authorities in the Forensic Handwriting Examination eld16.
With the advent and development of assistive technology, such as digitising tablets and
biometric analysis software, the eld has seen an increase in dynamic analysis, further
increasing the disciplines knowledge of key elements of both genuine and fraudulent

*
Corresponding author. Email: tahnee.dewhurst@gmail.com

2016 Australian Academy of Forensic Sciences


660 T.N. Dewhurst et al.

signing behaviours710. The research described here takes these approaches to analysis
one step further, by capturing and characterising fundamental behavioural biometric
indicators of signature formations that cannot be visualised with the naked eye. Little
empirical forensic research is available as to the characterisation of genuine or spurious
writings, via the analysis and comparison of dynamic features not visible on the static
trace. These non-visible, intra-signature, kinematic features are segments within a signa-
ture formation where the pen is momentarily lifted from the page within the signature,
such as might occur between a rst and last name, or when the pen leaves the page for
the purpose of crossing a t or dotting an i. It is hypothesised that these non-visual
segments are an inherent part of signature formations, as replicable as the visible traces.
Since it is difcult to imitate the dynamics of the visual components of a signature for-
mation, let alone the non-visual dynamic components, it is thought that this approach
may be useful in detecting simulations, in particular those simulations that by either luck
or practised effort are able to both pictorially and dynamically mimic a known signature
formation. It is postulated that not only may this type of dynamic feature analysis be of
value in a forensic capacity, by supporting examiners static examinations of disputed
writings and subsequent formulation of opinion as to genuineness or otherwise, but also
it may be useful in a biometric capacity in evaluating captures on electronic point of
transaction signature recording devices. Furthermore, it is empirical data such as that
collected and analysed here that continues to provide forensic handwriting examiners
with a resource on which to underpin their opinions in the casework environment and
solidify a more scientically robust platform for the discipline.

Method
Nineteen skilled, tertiary educated, adult writers were recruited to provide known model
signature samples. Each of the known model signature samples was assessed by the
experimenters to ensure that they were of sufcient complexity as to warrant examina-
tion and comparison within a forensic case work scenario, and were relatively represen-
tative of signatures found in the general population. Signature complexity was
determined using a statistical complexity model that has been previously reported11,12.
Each known model signature was then further examined and peer reviewed for clear
incidences of naturally occurring intra-signature pen lift(s). An intra-signature pen lift
was identied as a line break within the signature formation, where the pen had clearly
left the writing surface before being placed back on the paper to continue the signature.
Of these 19 writers, 13 were identied as having naturally occurring pen lifts within
their signature formations, such as might occur between a rst name and a surname
(Figure 1), an initial and a surname (Figure 2) or when lifting the pen from the page to

Figure 1. A signature formation illustrating a naturally occurring pen lift between the rst name
and surname. This study focused on movements occurring within these non-visible regions of
signature formations.
Australian Journal of Forensic Sciences 661

Figure 2. A signature formation illustrating naturally occurring pen lifts between the rst and
second initials and the surname.

complete components within the signature formation such as crossing a t (Figure 3).
Naturally occurring pen lifts were dened as those lifts that occurred as an artefact of
the writing behaviour as dened above, and that were always present, and as such
formed part of the motor memory for the handwriting movements.
The 13 individuals, identied as having naturally occurring pen lifts within their
signature formation, were asked to provide 15 repetitions of their signature in one sit-
ting, resulting in a pool of 195 known signature samples. Each signature was executed
using a Wacom Intuos Duo inking/digitising pen (a specialised inking pen that is com-
patible with the kinematic recording of the Wacom tablet) and written within a 118
38 mm printed box on a single A4 (REFLEX brand) 80 gsm sheet of paper placed
upon the digitising pad. Electronic data were captured using the Neuroscript Movalyzer
Ver. 6.6 movement analysis software. In this way, both visual and non-visual online
data of the signature formation could be captured simultaneously. What was of interest
was what was occurring during the non-visible, intra-signature component of each of
the signature formations, i.e. for example what the writer was doing in that break
between a rst name and surname. Further to the 13 known signature providers, an
additional eight individuals were recruited as simulators. These individuals were also
skilled adult writers. The simulators were provided with ve arbitrarily selected exam-
ples of each of the 13 model signatures on which to base their simulation attempts.
Simulators were allowed one minute to familiarise themselves with each model signa-
ture prior to commencing their simulations. Participants were not permitted to practice
prior to commencing their 15 attempts. Using the same method of collection as
described for the known model signatures, the eight simulators were directed to attempt
to simulate each of the 13 known signatures, 15 times, one signature per page. This
process resulted in a pool of 1560 signature simulations, comprising 195 simulation
attempts per simulator. Using the Movalyzer software, signature formations were
broken down into segmented components, using pen up, pen down occurrences as the

Figure 3. A signature formation illustrating a number of naturally occurring pen lifts, one of
which occurs due to the requirement of nishing an incomplete allographic formation, such as
crossing a t.
662 T.N. Dewhurst et al.

Table 1. Denition of the non-visible, intra-signature segment parameters measured in this


study.

Parameter Denition
Duration The movement time in seconds between two points
Straightness Normalised standard deviation from a straight line
Error
Size Absolute size of signature formation calculated from horizontal and vertical
vector differences
Jerk A disuency measure, the rate of change of acceleration for stroke size and
duration

segmentation parameter. In this way discrete, non-visualised intra-signature segments,


between a pen up and pen down event were able to be isolated, characterised and com-
pared with corresponding segments in the simulated signature formation. Four dynamic
parameters within the non-visible, intra-signature segments were measured: absolute
size, duration, normalised jerk and straightness error (Table 1). Statistical analysis of
the data was performed in SPSS v21.0 (SPSS Inc, Chicago). Owing to the nested,
repeated measures design, a generalised linear mixed model (GLMM) was utilised. For
each variable, two levels of subject were specied, with the known model signatures as
the highest level, and the simulator nested within each known signature. The trial num-
ber (115) was specied as a repeated measurement. For duration, straightness error
and absolute size a normal distribution with identity link was utilised. As normalised
jerk was signicantly non-normal, a log transformation was performed prior to analysis
with a normal distribution and identity link in the GLMM. The status of the signature
(known or simulation) was included as a xed effect, as was an interaction between
status and the complexity of the known signature and trial number. The known signa-
ture identity, the simulator identity, and the interaction between them were modelled as
random effects. A diagonal covariance type was specied for both subjects and random
effects, to indicate heterogeneous variances and a lack of correlation between matrix
objects. Post-hoc testing between groups was performed with sequential Bonferroni
correction.

Results and discussion


Known signature population
Non-visible, intra-signature segments within the 195 known signatures from 13 writers
were modelled for each of the four parameters measured; duration, straightness error,
absolute size and jerk, to determine the overall level of variation in a known popula-
tion. As would be expected, highly signicant differences were observed between the
13 writers for each of the selected parameters (Duration: F12,165 = 53.071, p < 0.0001,
Straightness error: F12,165 = 36.378, p < 0.0001, Absolute size: F12,165 = 108.499,
p < 0.0001, Jerk: F12,165 = 70.384, p < 0.0001 see Figure 4). These differences con-
rm that the non-visible, intra-signature components of the 13 different model signature
formations were as variable as their visible on-page components. Whilst conrming
inter-signature variability and providing further support for one of the underpinning
dogmas of handwriting identication, that no two people write exactly alike, the non-
visible components proved to be an integral and uid part of each of the signature for-
mations. As the pen is lifted within a signature formation, it does not merely rest, nor
Australian Journal of Forensic Sciences 663

Figure 4. Mean dynamic parameters and 95% condence intervals for genuine signatures. Panel
A = duration, B = straightness error, C = absolute size (cm), D = normalised jerk (cm s3).

necessarily take the shortest possible route from pen up to pen down, but traverses in a
pattern, as characteristic as the on page visual formation (Figure 5).
There were no notable trends across the known signatures between the four vari-
ables. For example, signatures with the largest size did not display the highest jerk or
longest duration of non-visible, intra-signature segments. For the segments examined

Figure 5. A known model signature formation, with the on-page static trace illustrated in blue,
non-visible, intra-signature component in red. The red line illustrates that pattern of movement
occurring between the pen up and pen down event, in between the rst name and surname of this
signature formation. Pen up, pen down incidences are marked by red circles.
664 T.N. Dewhurst et al.

Figure 6. Dynamic parameters of non-visual intra-signature segments through known model sig-
nature repetitions 115. Non-signicant comparisons are shaded blue, while signicant differ-
ences are shaded green. An overall mean was calculated for all signatures, and then each trial
was compared to that overall mean to determine if there was an increase or decrease. If there
was a linear trend, we would expect to see increases in trials 17 and decreases in trials 815.
Panel A = duration, B = straightness error, C = absolute size (cm), D = normalised jerk (cm s3).

within the known signature formations, there was no signicant practice effect observed
through replicates 1 to 15 for each of the variables (Duration: F14,165 = 1.022,
p = 0.434, Straightness error: F14,165 = 1.016, p = 0.440, Absolute size: F14,165 = 1.369,
p = 0.174, Jerk: F14,165 = 1.141, p = 0.326). Post-hoc testing was used to model the
deviation from the mean at each repetition (Figure 6).

Comparison between known model and simulated signatures


The relative intra-signature stability observed across all parameters for the non-visible,
intra-signature segments within the known model signatures provides an ideal point of
comparison for analysis with the contrasting segments within a simulated population.
The non-visible, intra-signature segments for simulated and known model signatures
were compared with the use of a generalised linear mixed model, where known and
simulated writer identity were included as random block effects, to allow personal inter-
cepts to be modelled for each combination of writers, and where trials were included as
a repeated observation to account for intra-individual variation. There was an overall
increase observed across all parameters investigated, for the simulated signature seg-
ments as compared with the known ones. This observation in itself is not surprising as
Australian Journal of Forensic Sciences 665

Table 2. Average values for each of the four parameters investigated, for known model and sim-
ulated non-visible, intra-signature segments.

Known average Simulated average


Variable (95% CI) (95% CI) Signicance testing
Duration 0.241 (0.2250.257) 1.71 (1.6361.778) F1,1663 = 69.3, p<0.0001
Straightness error 0.956 (0.0820.110) 0.146 (0.1380.154) F1,1663 = 5.846, p=0.016
Absolute size 0.960 (0.8961.023) 1.214 (1.1791.248) F1,1663 = 10.572, p=0.001
Normalised jerk 133.1 (101.7165.5) 32,252 (24,21240,292) F1,1663 = 88.988, p<0.0001

it has previously been reported that indicators of simulation in both static off-line and
dynamic on-line investigations include elements such as larger size and reduced line
quality7,13,14; however, it has not been previously reported as to whether these observa-
tions can also be said for the non-visible, intra-signature components of signature for-
mations. Overall, differences ranging from signicant to highly signicant were found
in all four measures, for the simulations as compared with the known model signatures,
with simulated signature segments recording longer duration, larger size, and increased
levels of disuency (Table 2). Jerk illustrated the most marked increase and, as we
know, this is one of the principal indicators for simulation behaviour in static off-line
examinations6. It was found that some simulators were able to mimic the size and
straightness measurements of the known model, non-visible segments, relatively well;
however, others showed marked variation from the known-model formation. None were
able to replicate the jerk, nor duration measurements as recorded within the non-visible
segments of the known model formations (Figure 7). Inter-individual variability was
accounted for within the known signature pool, in conjunction with intra-simulator abil-
ity within the simulated signatures, across trials 1 to 15. A practice effect within the
simulated signatures was detected for duration and jerk, but not for straightness error
and absolute size.
The complexity of the known model signature did not predict the extent of the
non-visible segments within either the known model or simulated signatures.
However, it did play a role in the simulators ability to effectively mimic non-visible
segments overall. For duration, size and jerk, simulators were able to replicate low
complexity non-visible segments more successfully than high complexity signatures
(Table 3).

Duration
Duration was measured as the time (ms) interval of the selected segment from start to
nish. Overall, a highly signicant effect (F1,1663 = 69.3, p<0.0001) was observed for
comparative duration length between the known and simulated signatures. The time
taken to execute the non-visible, intra-signature segments within simulations was found
to be longer than that observed in the comparative segments of the known model signa-
tures. In particular, non-visible segments of complex signatures were observed to have
a longer duration as compared with non-visible segments of low complexity signatures
in both simulated and known groups (Panel A, Figure 8). Clearly, measurement of
duration of these segments is a good predictor of simulation behaviour in the signature
formations examined here.
666 T.N. Dewhurst et al.

Figure 7. Non-visible segment parameter comparison between known and simulated signature
formations. Panel A = duration, B = straightness error, C = absolute size (cm), D = normalised
jerk (cm s3).

Straightness error
Straightness error was measured as the standard deviation from a straight line of the
selected non-visible, intra-signature segment. Overall, a signicant effect
(F1,1663 = 5.846, p=0.016) was observed for comparative straightness error determina-
tion between the known and simulated signatures. The calculated straightness error of
the non-visible, intra-signature segments within simulations was found to be larger than
that observed in the comparative segments of the known model signatures. Straightness
error was found to be similar in signature segments of both complex and simplistic sig-
natures, with intermediate signatures exhibiting less straightness error then the other
groupings (Panel B, Figure 8). Of the four parameters investigated, straightness error is
of the least determinative value when attempting to differentiate between simulated and
known segments. Whilst differences were observed between simulated and known sig-
nature formations, it is clear that straightness error is not as accurate a predictor of
Australian Journal of Forensic Sciences 667

Table 3. Effect of complexity level on non-visible, intra-signature segments, where K=known


signatures and S=simulated signatures.

Complexity High complexity Intermediate complexity Low complexity


level signatures signatures signatures
Duration Means K: 0.326 K: 0.300 K: 0.152
S: 1.408 S: 1.332 S: 0.886
Signicance t1663=8.837, t1663=8.471, p<0.0001 t1663=2.483,
test p<0.0001 p=0.013
Straightness Means K: 0.131 K: 0.069 K: 0.108
error S: 0.180 S: 0.097 S: 0.180
Signicance t1663=1.927, t1663=1.112, p=0.266 t1663=1.549,
test p=0.054 p=0.121
Absolute Means K: 0.857 K: 1.126 K: 0.620
size S: 1.164 S: 1.361 S: 0.795
Signicance t1663=3.923, t1663=3.010, p=0.003 t1663=0.915,
test p<0.0001 p=0.360
Normalised Means K: 98.4 K: 40.7 K: 8.93
jerk S: 5023.4 S: 2937.7 S: 1990.7
Signicance t1663=7.700, t1663=8.364, p<0.0001 t1663=4.327,
test p<0.0001 p<0.0001

simulation behaviour as the other three dynamic parameters (duration, size and jerk)
investigated in this trial.

Absolute size
Absolute size was calculated in centimetres using the horizontal and vertical vector dif-
ferences of the selected non-visible segments. Overall, a signicant effect
(F1,1663 = 10.572, p=0.001) was observed for comparative absolute size differences
between the known and simulated signatures. The absolute size of the non-visible seg-
ments within simulations was found to be larger than that observed in the comparative
segments of the known model signatures. Intermediate and highly complex signatures
were observed to have larger non-visible segments compared with the low complexity
signature segments in both the simulated and known groups (Panel C, Figure 8).

Jerk
Jerk was measured as the rate of change of acceleration for stroke size and duration of
the selected non-visible segment. Overall, a highly signicant effect (F1,1663 = 88.988,
p<0.0001) was observed for comparative jerk determination between the known and
simulated signatures. The calculated jerk of the non-visible segments within simulations
was found to be larger than that observed in the comparative segments of the known
model signatures. The non-visible segments of complex signatures were observed to
have a larger jerk compared with the non-visible segments within low complexity sig-
natures in both simulated and known groups (Panel D, Figure 8). Clearly, measurement
of jerk within these segments is a good predictor of simulation behaviour in these
signature formations.
668 T.N. Dewhurst et al.

Figure 8. Comparison of the overall duration (A), straightness error (B), size (C) and jerk (D)
of non-visible, intra-signature segments in 1560 simulated and 195 known signatures. High (1),
intermediate (2) and low complexity (3) signatures are represented by the black, blue and red
lines respectively.

Visual comparison
Fundamentally, the principal approach to the examination of disputed signatures has
varied little from the approach documented by historical authorities15,16 in that visual
examination, be it macro or microscopic, is still the primary method of investigation in
the vast majority of forensic signature examinations17. In forensic casework scenarios,
in the absence of dynamic data such as that described here, visual examination may be
the only examination type available. In this study, the dynamic data recorded for the
non-visible, intra-signature segments were converted to a visible representation of the
non-visible line trace, suitable for visual comparison (Figures 9 and 10). It is these
visual snapshots of the movements occurring between pen-up and pen-down intra-sig-
nature segments that may perhaps prove to be of the most forensic value. In the
opinion of the authors, the non-visible segments in the known model signatures were
clearly similar to each other across all 15 repetitions (Figures 11, 14 and 16). When
these segments were compared visually with the comparative segments within the
Australian Journal of Forensic Sciences 669

Figure 9. Known model signature A, illustrating the natural pen lift that occurs centrally, in
between the rst name and surname of the signature formation, depicted by the red circles which
indicate pen up and pen down events.

Figure 10. Known model signature A formation; the on-page static trace is illustrated in blue,
whilst the non-visible, intra-signature component is illustrated in red. The red line illustrates the
pattern of movement occurring between the pen up and pen down event, between the rst name
and surname of this signature formation.

Figure 11. Known model signature A with area of interest boxed. The cropped area of interest
of three alternative known model signature A formations is illustrated. Note the similarity of the
pattern of movement of the non-visible, intra-signature segments in red.

Figure 12. Simulation attempt of known model signature A. This illustrates the differences
observed between the non-visible, intra-signature component (depicted in red) of this simulated
signature formation as compared with the known model signature (Figures 10 and 11).

simulated signatures, gross overall pictorial disparity was observed (Figures 12, 15 and
17), as would be expected, according to the dynamic data discrepancies previously dis-
cussed. It is this gross pictorial disparity that may be of most interest to frontline foren-
sic document examiners or those called upon to make rapid authentication judgements,
such as may occur at border control points or nancial institutions. The value of the
segments investigated is twofold, in that known model signatures illustrate consistent
and repeatable behaviour within those segments investigated and that the contrasting
670 T.N. Dewhurst et al.

Figure 13. Known model signature B. Note natural pen lifts (2) and associated non-visible,
intra-signature pattern of movement occurring within the formation as indicated by the red line.

Figure 14. Three alternative known model signature B formations. Note the similarity
observed of the non-visible, intra-signature segments of this signature, depicted in red.

Figure 15. Two simulation attempts of the known model signature B. This illustrates the differ-
ences observed between the non-visible, intra-signature components (depicted in red) of these simu-
lated signature formations as compared with the known model signature (Figures 13 and 14).

Figure 16. Known model signature C with area of interest boxed. The cropped area of interest
of three alternative known model signature formations is illustrated; note the similarity of non-
visible, intra-signature segments in red.

segments within simulation attempts vary markedly from the known model signature.
Further underpinning the value of non-visible segments is the security of these features,
in that simulators are unable to observe or practise the segments. Furthermore, it is
thought that the gross disparity observed between known and simulated non-visible
intra-signature segments is such that the differences would still be able to be observed
at relatively low resolution, such as that used in the majority of electronic signature
capture devices.
Australian Journal of Forensic Sciences 671

Figure 17. Simulation attempt of known model signature C. This illustrates the differences
observed between the non-visible, intra-signature component (depicted in red) of this simulated
signature formation compared with the known model signature (Figure 16).

Overall, the dynamic analyses of the non-visible, intra-signature segment parameters


of duration, jerk, size and straightness error for both simulated and known signatures,
complements the historical and contemporary observations for both off-line and on-line
analysis of signature formations. Known model signatures are characterised by ballistic,
uent formations whilst simulations are characterised by both a lack of speed and u-
ency6,7,18. The signicant disparity observed between known and simulated formations
for the selected parameters indicates that these non-visual, intra-signature components
may be useful predictors of simulation behaviour in disputed signature formations. The
fact that the segments investigated cannot be visualised makes them innately difcult to
emulate and a useful reference tool in the examination of disputed signature formations.
It is further hypothesised that these segments may also be of benet in the forensic
examination of simplistic signatures. Simplistic signatures are those signature types on
which examiners may experience difculty in expressing opinions as to authenticity,
due to their low complexity and the perceived ease of reproducibility of the formation.
These signature types can generate more passable simulation attempts, pictorially, com-
pared with more complex signature formations; however, they still illustrate clear visi-
ble differences in the non-visible, intra-signature regions (Figures 1825).
Visual examination of the recorded simulations also revealed that some simulators
appeared to have their own idiosyncratic signature within their simulation attempts.

Figure 18. A known model signature formation. The on-page component is illustrated in blue.
The red line illustrates the above page pattern of movement occurring between pen up and pen
down events. Pen up, pen down incidences are marked by red circles.
672 T.N. Dewhurst et al.

Figure 19. The simulation attempt of the known model signature depicted in Figure 18. The
on-page component is illustrated in blue. The red line illustrates the pattern of movement occur-
ring between pen up and pen down events. Pen up, pen down incidences are marked by red cir-
cles. The on-page component of the signature formation (blue line) illustrates an overall pictorial
similarity to that of the known model signature (Figure 18), in marked contrast to the disparity
observed in the pattern of movement occurring within the above page, intra-signature sections of
the signature formation (red line).

Figure 20. A known model signature formation.

Figure 21. Simulation attempt of the known model signature depicted in Figure 20.

Figure 22. A known model signature formation.

That is, some forgers executed similar, non-visible, intra-signature patterns of move-
ment, within the same part of the simulation attempt (Figure 26). This observation war-
rants further research and may be of use in forensic cases that attract multiple
Australian Journal of Forensic Sciences 673

Figure 23. Simulation attempt of the known model signature depicted in Figure 22.

Figure 24. A known model signature formation.

Figure 25. Simulation attempt of the known model signature depicted in Figure 24.

simulators of the one target signature, such as art or memorabilia forgeries, or for those
signatories that allow multiple author use of their signature, such as might occur in
large companies or within families, etc.
Further studies will include the investigation of dynamic, non-visible, intra-signature
segments within disguised signatures as compared with their genuine counterparts. It is
hypothesised that the interruption of the abstract motor programme that occurs when
altering a well-practised signature formation may correlate with observable changes in
the non-visible, intra-signature segments. If this proves to be false, and non-visual,
intra-signature segments in disguised signatures remain stable, this method of signature
analysis may prove to be an excellent determinate for differentiating between disguised
and simulated signature formations, a task that up until now has proved difcult and
associated with error1921.
In addition to the investigation of disguised signatures, it is envisaged that further
studies involving more practised simulations may be benecial, to investigate whether
or not the non-visible, intra-signature segments in more practised simulation attempts
begin to approach those observed within the known model signatures. It is documented
674 T.N. Dewhurst et al.

Figure 26. Four simulation attempts of a known model signature by the one writer. It is
observed that following the pen up event (highlighted by the boxed area of these signatures) the
pen initially follows the same trajectory in all simulation attempts, executing a half circle in an
clockwise direction and crossing (virtually) over the on-page line previously executed.

that as simulation attempts becomes more uent, there are fewer points of hesitation
and an increased overall pictorial and dynamic similarity to the that of the known
model formations22. However, it is not known whether increased familiarity with the
execution of the model formation will result in more dynamically and pictorially simi-
lar, non-visible, intra-signature components of these simulation attempts.

Limitations
The study as described was subject to a number of methodological limitations that fur-
ther research and expansion of the basic construct will go some way to addressing. The
dataset used was limited in that only 13 model signatures were assessed by the authors
to have naturally occurring pen lifts, therefore, further studies will endeavour to collect
a larger data-set of signatures that qualify for this basic tenet. Additionally, it is
acknowledged that any testing regime may change the behaviour of the participants
involved, including signing behaviour in both the genuine and simulated signatures.

Disclosure statement
No potential conict of interest was reported by the authors.

References
1. Osborn AS. Questioned document problems. New York, NY: Boyd Printing Company; 1946.
p. 338.
2. Harrison WR. Forgery detection: a practical guide. London: Sweet and Maxwell; 1964.
p. 115.
3. Harrison WR. Suspect documents and their scientic examination. London: Sweet and
Maxwell; 1966. p. 324.
4. Hilton O. Scientic examination of questioned documents. Revised Ed. New York: CRC
Press; 1982. p. 185.
5. Ellen D. The scientic examination of documents: methods and techniques. New York: Ellis
Horwood; 1989. p. 4142.
6. Huber RA, Headrick AM. Handwriting identication. Boca Raton (FL): CRC Press; 1999.
p. 289.
Australian Journal of Forensic Sciences 675

7. van Gemmert AWA, van Galen GP. Dynamical features of mimicking another persons writ-
ing and signature. Amsterdam (The Netherlands): IOP Press; 1996. Handwriting and drawing
research: basic and applied issues, Simner ML, Leedham CG, Thomassen AJWN, editors;
459471.
8. Sita J, Rogers D. Changes in forgers handwriting pressure related to the original writers
dynamics. J Forensic Doc Exam. 1999;12:101112.
9. Franke K. Dynamic, on-line signature analysis: Trends and challenges for forensic investiga-
tion services. Salzburg (Austria); 2007. Proceedings of the seventh international congress of
the Gesellschaft fur Forensischen Schriftuntersuchung (GFS); Salzburg.
10. Mohammed L, Found B, Caligiuri M, Rogers D. Dynamic characteristics of signatures:
effects of writer style on genuine and simulated signatures. J Forensic Sci. 2015;60(1):8994.
11. Alewijnse LC, van den Heuvel CE, Stoel R. Analysis of signature complexity. J Forensic
Doc Exam. 2011;21:3749.
12. Dewhurst T, Found B, Rogers D. The relationship between quantitatively modelled signature
complexity levels and forensic document examiners qualitative opinions on casework. J.
Forensic Doc Exam. 2007;18:2140.
13. van Galen G. Structural complexity of motor patterns: a study of reaction times of handwrit-
ten letters. Psychol Res. 1984;46(1-2):4957.
14. Franke K. Analysis of authentic signatures and forgeries. Computational forensics. Proceed-
ings of the Third International Workshop on Computational Forensics (IWCF 2009); 2009
Aug 1314; The Hague (The Netherlands): Springer Verlag, 2009.
15. Hagan WE. Treatise on disputed handwriting. Albany (NY): Banks & Brothers; 1894.
16. Mathyer J. The expert examination of signatures. J Crim Law and Criminology. 1961;52
(May-June) Article 12: 122133.
17. Lewis J. Forensic document examination, fundamentals and current trends. Oxford: Academic
Press; 2014.
18. Leung S, Cheng Y, Fung H, Poon N. Forgery I - simulation. J. Forensic Sci. 1993;38:
413424.
19. Mohammed L, Found B, Caligiuri M, Rogers D. The dynamic character of disguise behavior
for text-based, mixed, and stylized signatures. J Forensic Sci. 2011;56(S1):S136S141.
20. Bird C, Found B, Ballantyne K, Rogers D. Forensic handwriting examiners opinions on the
process of production of disguised and simulated signatures. Forensic Sci Int. 2010;195
(13):103107.
21. Bird C, Found B, Rogers D. Forensic document examiners skill in distinguishing between
natural and disguised handwriting behaviors. J Forensic Sci. 2010;55(5):12911295.
22. Dewhurst T, Found B, Rogers D. Are expert penmen better than lay people at producing
simulations of a model signature? Forensic Sci Int. 2008;180(1):5053.

You might also like