A Cognitive Look Into Simulations of Hig

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 4

A Cognitive Look into Simulations of High and Low Complexity Signatures

Avni PEPE, Douglas K. ROGERS and Jodi C. SITA


Handwriting Analysis and Research Laboratory,
School of Human Biosciences, La Trobe University,
3086, Victoria, AUSTRALIA
a.pepe@latrobe.edu.au, d.rogers@latrobe.edu.au, j.sita@latrobe.edu.au

Abstract. The difficulty with which a signature can be simulated has been considered to be related to
elements associated with its complexity, with more complex signatures assumed to be harder to simulate as
compared to less complex ones. In this study, eye movements and handwriting dynamics were examined
during subjects’ attempts at simulating two model signatures of different complexities. It was found that
significantly more fixations were made on the model signatures than the simulated signatures for both the low
and high complexity simulations. Subjects made significantly more fixations on the high complexity model
signature, compared to the low complexity model signature and these fixations had longer durations. The
duration of fixations on the simulated signatures did not differ between the complexity conditions. When
asked to nominate which model signature would be harder to simulate, during pre-simulation, the majority of
subjects nominated the high complexity signature, however, only a few retained this view post-simulation.
Although gaze data did lend support to contemporary views as to how signature complexity may be modeled,
it appears that there are other factors contributing to complexity that have yet to be incorporated into the
model tested.

1. Introduction
Eye-tracking techniques can be used to explore issues regarding the cognitive processes that may occur during
the real time production of graphomotor output. Such techniques have contributed to the appearance of various
models (see Alamargot & Chanquoy, 2001) that attempt to explain the cognitive processes that occur throughout
the act of writing extended pieces of text. Text production is a complex activity and involves the management of
four main processes in working memory: planning, formulation, execution and revision (Alamargot et al., 2006).
It is thought that these processes can occur sequentially, or in parallel, as long as working memory capacity is
not exceeded by the process demands (Kellogg, 1996). However, whether or not the same mechanisms are
involved in signature production, is unclear. To date, there has only been one study (Sebastian et al., 2009) that
has used eye tracking to explore the cognitive processing that occurs during signature production.
It has been proposed that the complexity of a signature relates to the difficulty that forgers have when
simulating that signature (Found & Rogers, 1998). Given that eye movements occurring during goal directed
hand movements are driven primarily by top-down processes, that is, by goals of behavior rather than salient
image properties (Land & Tatler, 2009), it is possible to explore factors that influence, or govern signature
simulation difficulty. Only one study has used eye-tracking methods to explore the processes involved in
signature forgery (Dyer et al., 2006b).
The only other studies using eye-tracking related to forgery were by Dyer et al. (2006a, 2008) who
tracked the eye movements of control subjects and forensic document examiners during signature identification
tasks (determining whether signatures were genuine or forged). They found that document examiners
incorporated global processing of stimuli, as well as local feature extraction and that subjects’ inspection of
signatures appeared to be mediated by a bottom-up search strategy. They concluded that the difference in
accuracy between examiners and controls regarding the authorship of a signature was due to a difference in
cognitive processing of the visual information they extract from the images.
The process of perceiving whether a signature is forged is different to the process of producing a forgery
itself, which can be similar to a copying task. Miall et al. (2009) looked at gaze patterns and brain activity of
subjects copying drawings of faces to gain insight into the cognitive processes involved during copying tasks.
They proposed that copying involves visually capturing the target visual image, holding this in memory,
executing the drawing action and visually inspecting the drawn line.
In this study, in addition to making novel observations regarding eye movements and inferred cognition
during the process of signature simulation, we were interested to see whether eye movement patterns would
provide support for a published complexity classification theory (Found & Rogers, 1998).This theory would
predict that subjects would rate the high complexity signature as being more difficult to simulate than the low
complexity signature. Of interest is whether predicted signature complexity would be supported by gaze data.

2. Methods
17 healthy right-handed adults volunteered for this study. The mean age of subjects was 29 years. All subjects
underwent a visual evaluation and verbally confirmed the absence of any known eye movement problems or the
use of any medications that are known to affect eye movements. Subjects all had a corrected visual acuity level
of at least 6/12 and corrected near sight visual acuity level of at least 6/6.
A Panasonic NV-GS17 Digital camera attached to a camera stand was used to capture the handwriting
movements and written trace of the subjects while they wrote on white 7x15cm pieces of paper. A PTZ-1230
Wacom Intuos 3 digitising tablet recorded handwriting dynamics, while a Tobii X-50 eye-tracker simultaneously
captured eye movements of subjects. The sampling rate of the eye-tracker was 50 Hz with an accuracy of 0.5°
and the digitising tablet recorded at 200 Hz with an accuracy of 0.25mm in the x and y direction. Tobii
Clearview 2.7.1 software, in conjunction with the X-50 system provided a qualitative view of the areas that the
subjects looked at and a quantitative analysis of gaze data. Handwriting dynamic data was analysed and
extracted using Movalyzer 5.4 software.
Eye fixations were defined by the eye remaining within a 50 pixel area for a minimum time of 100ms. The
resolution of the camera was 768x576 pixels.
The measures included were: average number of fixations, average duration of fixations, average pen
pressure and average writing velocity. Subjects were also surveyed as to their perceptions of signature simulation
difficulty, pre- and post-simulation.
After familiarisation with the study setup and a verbal description of the equipment, subjects were seated
at the table and calibrated to the eye tracker. This required them to look at a set of 9 predetermined points on a
piece of paper, which lay flat in front of them on top of the digitizing pad. Subjects were first asked to produce
their own signature six times, each time on a separate piece of paper. Subjects were instructed to produce their
signature after they heard an auditory cue and to notify the experimenter upon completion of each trial. After
genuine signatures were collected, subjects viewed both the high and low complexity signature. Upon viewing,
they were asked to answer the question; "of these two signatures, which do you think would be harder to
simulate?” The subjects’ responses were noted pre- and post-simulation attempts. Subjects were then informed
that the "simulation trials will now begin" and that they were to “start their simulation attempt following the
auditory cue sound”. Subjects were instructed to reproduce, as accurately as possible, both signatures, and
informed that they could take as much time as they needed. For each subject, there was a minimum of 6
simulation attempts of each model signature. The simulation order of the two model signatures was
counterbalanced between subjects.
Student t-tests were performed to compare eye gaze and handwriting dynamics between and within the
high and low complexity simulations.

3. Results
3.1 Within and between high and low complexities. Subjects’ eye gaze data was compared and analysed
within and between the high and low complexity signatures. Figure 1 is an example of two gaze plots taken from
the low and high complexity simulation trial, revealing the number, location and relative duration of fixations
made on both the model (a. low complexity, b. high complexity) and simulated (c, d) signatures.

a. b.

c. d.
Figure 1. Example of a gaze plot relating to a simulation trial of the low complexity (a.) and high complexity
(b.) signature. (Circles indicate fixation location and numbers indicate fixation order. Size of the fixation is
proportional to its duration).

Looking within the complexity conditions, significantly more fixations were made on the model
signatures than the simulated signatures for both the low (t(16)= 3.26, p< 0.01) and high (t(16)= 5.59, p< 0.0001)
complexity simulations. Additionally, significantly longer fixation durations were made on the model signatures
than the simulated signatures for both the low (t(16)= 2.65, p< 0.05) and high (t(16)= 4.45, p< 0.001) complexity
simulations. In a comparison between the high and low complexity simulations, we found that subjects made
significantly more fixations on the high complexity model signature, compared to the low complexity model
signature (t(16)= 2.36, p< 0.05) and significantly longer fixations (t(16)= 4.14, p< 0.001) were made on the high
complexity model signature compared to the low complexity model signature. Interestingly, there was no
significant difference between the duration of fixations (t(16)= 1.44, p>0.05) made on the simulated signatures
between the high and low complexity simulations.

3.2 Handwriting measures. The two handwriting measures analysed included pen pressure and writing
velocity. Significantly more pen pressure was applied during the high complexity simulations (t(16)= 3.93, p<
0.01). However, the low complexity simulations were completed with a significantly greater writing velocity
(t(16)= 3.05, p< 0.01).

3.3 Subjects opinions. The majority of subjects’ opinions about simulation difficulty of the two model
signatures changed from pre- to post-simulation. Pre-simulation, 15 of the 17 subjects believed that the high
complexity signature would be harder to simulate, however, post-simulation, 12 of the 17 subjects thought that
the low complexity signature was harder to simulate. Chi-square testing revealed that the change in subjects
opinions regarding simulation difficulty was significant (X2 = 5.0, p = 0.025).

4. Discussion
The present study has taken a novel approach to indirectly investigate cognition during the simulation of
signatures by tracking the movement of the eyes. In addition, we have used what is currently known about eye
gaze to investigate the validity of the complexity classification model developed by Found and Rogers in 1998.
For the gaze measure, we found more fixations where made on the model signatures than the simulated
signatures, regardless of signature complexity. This finding is suggestive that there is a greater need to gather
visual information from the model signature than to check the output of writing. It is possible that once the motor
action is planned and motor execution begins, the eyes only serve to guide spatial aspects of output, which is of
lesser priority than refreshing visual memory of the model signature in the process of simulation. This makes
sense in light of the finding that the duration of fixations made on the simulated signatures did not differ between
the high and low complexity simulations. These fixations, which were consistent in length, may be
representative of the reoccurring process of spatial ‘checking’ - a behaviour which remained uninfluenced by the
properties of the graphomotor output. This notion is supported by Miall et al. (2009) who report that drawing
from memory activates areas in the brain consistent with encoding and retention of spatial information.
However, frequent fixations made on the simulated signature may also occur to prevent degradation of
visuomotor memory in the absence of visual feedback when gaze is directed at the model signature. Vaillancourt
and Russell (2002) found after 0.5 to 1.5 seconds without visual feedback, constant finger pressure declined
exponentially compared to when visual feedback was available. A similar effect has been found in aiming tasks,
whereby aiming accuracy is compromised following 2 seconds of visual occlusion (Elliott et al., 1991). In
relation to our study, this could explain why subjects generally employed a very frequent and reoccurring back
and forth gaze pattern between the model and simulated signature.
Smyth and Silvers (1987) found that in the absence of visual feedback during writing, more errors were
made and the spatial arrangement of written words along the horizontal was distorted. It would appear that
kinesthetic and haptic feedback alone are unable to provide the information necessary for maintaining these
aspects of writing, as they may operate in conjunction with a visual feedforward mechanism (Ghez et al., 1990).
Past research into eye movements has led to the observation that fixation duration can be influenced by
task difficulty and the amount of cognitive processing occurring at the time, with longer fixations usually
associated with increased task difficulty and a higher cognitive demand (Epelboim et al., 1995). Given that
longer fixations where observed on the high complexity than the low complexity model signature, consistent
with the complexity classification model, this is suggestive that the high complexity signature was either more
difficult to simulate, or demanded a greater amount of visual loading and retention. However, the association
between task difficulty and fixation duration must be taken with caution, given that fixation duration can be
greatly influenced by how fixations are defined. It is possible that slower writing speeds of the high complexity
signature combined with closely positioned lines are responsible for this finding.
Interestingly, most of the subjects changed their opinion about which model signature would be/was
harder to simulate, after completing the simulations. This shift in opinions suggests that initially, subjects may
have used a bottom-up assumption to determine simulation difficulty, but later became aware of other less
extrinsic influential factors (including pen pressure and velocity). It is possible that given the physical nature of
the low complexity signature, a greater amount of visual loading combined with the need for faster graphomotor
execution was required for accurate reproduction. In relation to the complexity classification model, whether or
not subjects’ opinions about simulation difficulty actually matched their performance would require further
assessment. Regardless, the opinion that the low complexity signature was harder to simulate could also mean
that other factors contribute to simulation difficulty other than the number of turning points, intersections and
retraces, which are the only predictors currently used in the complexity classification model.
Future work into the patterns of gaze behaviour between expert and novice simulators and a look into the
visual properties that govern simulation difficulty is planned.

References
Alamargot, D., & Chanquoy L. (2001). Through the models of writing. Dordrecht, The Netherlands: Kluwer.
Alamargot, D., Chesnet, D., Dansac C., & Ros, C. (2006). Eye and Pen: A new device for studying reading during writing.
Behaviour research methods, 38(2), 287-299.
Dyer, A. G., Found, B., & Rogers, D. (2006a). Visual attention and expertise for forensic signature analysis. Journal of
Forensic Sciences, 51(6), 1397-1404.
Dyer, A. G., Found, B., & Rogers, D. (2008). An insight into forensic document examiner expertise for discriminating
between forged and disguised signatures. Journal of Forensic Sciences, 53(5), 1154-1159.
Dyer, A., Found, B., Sita, J., & Rogers, D. (2006b). Eye Movement Evaluation of Signature Forgeries. In Proceedings of the
Australasian Ophthalmic and Visual Sciences Meeting, December 2-4. John Curtin School of Medical Research,
Australian National University.
Elliott, D., Carson, R. G., Goodman, D., & Chua, R. (1991). Discrete vs. continuous control of manual aiming. Human
movement science, 20, 393-418.
Epelboim, J., Steinman, R. M., Kowler, E., Edwards, M., Pizlo, Z., Erkelens, C. J., & Collewijn, H. (1995). The function of
visual search and memory in sequential looking tasks. Vision Res, 35, 3401-3422.
Found, B., & Rogers, D. (1998). A consideration of the theoretical basis of forensic handwriting examination. International
Journal of Forensic Document Examiners, 4, 109-118.
Ghez, C., Gordon, J., Ghilardi, M. F., Christakos, C. N., & Cooper, S. E. (1990). Roles of proprioceptive input in the
programming of arm trajectories. Cold Spring Harb. Symp. Quant. Biol, 55, 837–847.
Kellogg, R. T. (1996). A model of working memory in writing. In C. M. Levy & S. Ransdell (Eds.), The science of writing:
Theories, methods, individual differences and applications (pp. 57-72). Mahwah, NJ: Erlbaum.
Land, M. F., & Tatler, B. W. (2009). Looking and Acting: Vision and Eye Movements in Natural Behaviour. Oxford: Oxford
University Press.
Miall, R. C., Gowen, E., & Tchalenko, J. (2009). Drawing cartoon faces - a functional imaging study of the cognitive
neuroscience of drawing. Cortex, 45, 394-406.
Sebastian, J., Rogers, D. K., & Sita, J. C. (2009). Some Relationships between Eye Movements and Handwriting Movements.
In Proceedings of the 14th International Graphonomics Society. Dijon (France), 211-215.
Smyth, M. M., & Silvers, G. (1987). Functions of vision in the control of handwriting. Acta Psychologica, 65, 47-64.
Vaillancourt, D. E., & Russell, D. M. (2002). Temporal capacity of short-term visuomotor memory in continuous force
production. Exp Brain Res, 145, 275-285.

You might also like