An exploratory study on the handwritten allographic features of multi-ethnic population with different educational backgrounds.

Ethnicity, native and/or foreign language knowledge, as well as the learned writing systems potentially influence the development of an individual's handwriting. The unique education system consisting of National schools, Chinese-medium vernacular schools, Tamil-medium vernacular schools, and Islamic religious schools in Malaysia may have established specific characteristic handwritten allographic features that deserve investigation within the intelligence context. This study was aimed to explore handwritten allographic features of handwriting samples from 120 subjects (30 writers from four different educational backgrounds mentioned above). Characteristic features which could be attributed to the study groups were statistically analyzed and identified. In this study, thirteen allographic features, including letters "A", "B", "D", "H", "p", "T", "t", "w", "X" and "x", were found to be discriminative. Such information could serve to indicate the primary education system undergone by a writer; enabling the comparison of different handwriting profiles and allowing characterization of writers to a specific group of people.

Introduction

Ethnicity, native and/or foreign language knowledge, and writing systems are factors that contribute to the development of one’s handwriting [1]. The possibility to detect habitual features in English handwritings to indicate culture could be due to the varying attributes of writing systems in western countries [2]. In the west, special letter shapes are retained in the handwritings of people who learned writing in Canada and the Netherlands, which are influenced by the copybook systems in respective countries [3]. Influences of taught writing styles were also observed in the handwritings between Polish and English [4], Indian and English [5, 6], as well as Vietnamese and Australian [7] writers in English written documents. However, only 1% of a population would continue to adhere to the copybook writing style through the examination of grapheme "th" [8]. Identification of a writer’s nationality based on copybook system was also found not successful, restricted by abundant variations in handwriting characteristics in certain countries [9]. Handwritings from England in English written documents were most likely to be confused with styles from other countries due to lack of emphasis on copybooks [9].

Compared to studies between people originating from different countries of varying native language knowledge, studies on handwriting class characteristics for individuals from a single country are limited. From the available literature, Cheng and co-researchers had identified six class characteristics of handwritings to characterize three major races in Singapore [10]. They suggested that the majority of Indians in Singapore have the habit of exhibiting curve stroke formation in the letter "X" and round top formation in the letter "A" while most Chinese in the country write the horizontal stroke first when writing letter "T". Additionally, the looping of "d" stem, looping on "f" lower stem, and straight stroke formation of letter ‘M’ and ‘N’ were found possible in differentiating different racial groups [10].

Writers of different nationalities, native language knowledge and the learned writing systems could have exhibited discriminative handwriting characteristics [6, 10]. Applicability of previous studies to Malaysian’s handwriting is unknown simply because the unique education system today was established during the British administration before Malaysia gained independence in 1957. Malaysia’s education system consists of four different primary educational systems based on the language or medium of instruction used in teaching and learning, namely: Malay-medium National schools, vernacular schools of Chinese and Tamil media, and Islamic religious schools. In addition to Malay and English languages as compulsory subjects in all schools, students also learn Chinese, Tamil, and Arabic languages in their respective schools. Hence, handwritten allographic features adapted during primary levels at different educational systems which is the copybook writing style, and their influence towards the formation of one’s handwriting known as class characteristics is of great interest for forensic intelligence pertaining to handwritten evidence. Specifically, knowledge on the influence of an individual’s educational background towards the development of handwriting could provide additional information in forensic handwriting examination.

Forensic handwriting examination which is part of questioned document examination (QDE) aims to identify the author of a questioned handwriting using defensible scientific methods and to determine whether the author has produced a series of documents or if the signature was falsified [1, 2]. In most cases, QDE involves the comparison of an unknown writing sample with samples from a known writer to establish similarities and/or differences between the writing of the two sources [1, 2, 11, 12]. Such comparison is mandatory during forensic investigation in criminal procedures or civil disputes, but it is only possible when comparable writing samples are available to forensic document examiners. In cases where there was no known suspect or victim or in the absence of comparable handwriting samples (such as an untraceable parcel sent to an individual or a written document received from an unknown sender), the application of conventional QDE comparison using known and suspect samples becomes impossible. In this situation, comparing the profiles of questioned handwriting to a group or database of handwriting can help narrow down the pool potential writers in document related fraud, forensic accounting and even in terrorism investigation [11].

The premise here is that the variations in academic backgrounds, or more specifically, the educational learning systems by Malaysians are believed to have influenced their handwriting formation and this remains unexplored. This study was aimed to identify and establish the characteristic handwritten allographic features corresponding to selected population samples from different primary educational backgrounds. It is hoped that the determination of these features coupled with examinations of each allographic feature would serve as initial data, enabling the comparison of different handwriting profiles within the intelligence context and allow for characterization of writers to a specific group.

Materials and methods

Study population

In this study, each member of specific sample populations (Malay-medium National schools, vernacular schools of Chinese and Tamil mediums and Islamic religious schools) had an equal chance of being included. The writer was chosen by chance for actual inclusion. A name list of university students was obtained upon institutional approval and a computer generated simple random sampling method was used to find potential participants. An email was sent to the potential writers with the details of the study, including the purpose of study, risk of providing handwriting samples, and the disposal of handwriting samples upon completion of the study. The determined venue and time for handwriting sample collection were given to participants, and their consents were obtained. Sample size was calculated through two proportion formula by considering the letter “U” with introductory stroke [4], 25% of the Polish (0% for English) possessed the handwritten feature. With 80% of power study, 5% of risk error, and 10% of non-response rate, 30 writers were collected from each group. This resulted in a total number of 120 writers who voluntarily participated in the current study. Only writers aged between 18 to 20 years old and well-versed in reading and writing in English Language were recruited. A qualification level of higher than pre-university level with no mental and physical disabilities were also considered as inclusion criteria for the writers. Ethical approval was obtained from the institutional Human Research Ethics Committee (USM/JEPeM/17120687).

Data collection

A data collection kit comprising of an information sheet, consent form, writer’s information form, a source document and three blank sheets, as well as a blue ballpoint pen were distributed to each writer. Information pertaining to the age, gender, race, handedness, occupation, education level, primary and secondary schooling systems were specified in the writer’s information form prior to collecting the handwriting samples. The writers were required to copy a source document, thrice, on blank A4 white papers using the blue ballpoint pen provided in the kit. The source document is an English text, similarly used by Srihari et al. [13]. The source text consists of 156 words capturing all letters and numerals, punctuation as well as distinctive letter and numeral combinations (graphemes of ff, tt, oo, and 00).

Data extraction and coding

Background data was organized in a Microsoft Excel® Spreadsheet (Redmond, WA). Individual spreadsheets were created for individual alphabets and a separate spreadsheet was created to record the writers’ background data, age, handedness, education level, and their respective learning systems. Writers with Malay-medium National schools background were designated as Group A whereas writers from Chinese medium and Tamil medium vernacular schools as well as the Islamic religious schools were designated as Groups B, C, and D, respectively.

All handwriting samples were scanned at 400 dot per inch (dpi) using Canon® ImageCLASS MF3010 scanner (Canon Inc., Ota City, Tokyo, Japan) and viewed at magnification 2:1 using Adobe® Photoshop® CC software (Adobe Inc., San Jose, CA) for detailed examination of the handwriting features. Allographic features of each alphabet were examined by investigating the (1) number of strokes, (2) sequence of strokes, (3) direction of crossbar stroke, (4) position of crossbar strokes relative to the average letter height, (5) connectivity of crossbar stroke to the adjacent letter, (6) length of crossbar strokes, (7) connectivity between strokes, (8) morphology of individual strokes, (9) apex design and/or (10) presence of hiatus and looping. Each said feature was assigned with a code value.

In this study, the number of writers with the features occurring repeatedly in their handwriting samples were counted and organized in the spreadsheet. Frequency of occurrence for each handwritten allographic feature was calculated and compared across the four groups of writers. Based on the comparison, alphabets which showed significant frequency of allographic features attributed to a specific study group were then identified.

Statistical analysis

IBM SPSS Statistics version 27 (SPSS Inc., Chicago, IL, USA) was used for data analysis. Data cleaning and descriptive analyses were performed to ensure there were no key in errors. All identified handwritten allographic features were analyzed using Pearson’s Chi-square tests to determine if there are associations between the four groups of writers from based on frequency of occurrence for each feature. A p-value <0.05 was considered statistically significant. The significant variables were treated as confounding variables and controlled in the regression model.

Multinomial logistic regression model was used to investigate each common handwritten allographic features associated with the educational backgrounds, adjusted with the confounding variables. The regression model was constructed as follows: where y was a nominal variable with four categories (types of educational background), α was intercept of equation, β1 was the coefficient of common handwritten allographic features, β2 was the coefficient of confounding variables μ was random effect of respondent level, with the variation σ2. A p-value <0.05 was considered statistically significant in the regression model. The common handwritten allographic features in one or more groups as compared to others were determined and evaluated.

Results

Data description

All the 120 samples collected in this study were written using the ballpoint pens provided in the kit with the aim to minimize variation which could have arisen from using different writing instruments. Both the original and scanned copies of the handwriting samples were examined to determine the characteristic allographic features by the research team which was also membered by a certified forensic document examiner of the country. During the initial stage, handwritten features of all the alphabets and numbers in each handwriting sample were examined and coded in two separate occasions independently. Subsequently, the coded results were compared and validated by the certified document examiner. In this study, only the validated data was considered for interpretation and statistical analyses.

Based on the handwriting samples, 18 allographic features were suggested as the characteristics which potentially occurred in one group as compared to other groups, covering 14 letters "A", "B", "D", "H", "I", "P", "p", "T", "t", "w", "X", "x", "Z" and "z". Four letters ("A", "B", "D", and "w") demonstrated two features to be tested at differentiating the handwritings of the four study groups. Description of the 18 allographic features is shown in Fig 1.

Fig 1

Description 18 allographic features suggested as the characteristic of at least one study group.

The demographic information of the participants from the four different educational backgrounds was assessed, as tabulated in Table 1. No association was found in gender (p = 0.485) and handedness (p = 0.217) in relation to primary educational backgrounds. However, age, ethnicity, and secondary schooling system demonstrated significant association with educational backgrounds (p <0.001). Due to the large proportion of writers having similar ethnicity as in their respective educational backgrounds, two confounding variables, i.e. age and secondary schooling system, were adjusted in the final model through multinomial logistic regression analysis.

Table 1

Description on demographic information of the participants (N = 120).

Variable		National school (n = 30)	Chinese vernacular school (n = 30)	Tamil vernacular school (n = 30)	Islamic religious school (n = 30)	p-value
Age	18/19	27(90.0%)	12(40.0%)	11(36.6%)	24(80.0%)	<0.001*
	20	3(10.0%)	18(60.0%)	19(63.4%)	6(20.0%)
Gender	Male	6(20.0%)	8(26.7%)	4(13.3%)	4(13.3%)	0.485
	Female	24(80.0%)	22(73.3%)	26(86.7%)	26(86.7%)
Ethnicity	Malay	25(83.4%)	0(0%)	0(0%)	29(96.7%)	<0.001*
	Non-Malay	5(16.6%)	30(100%)	30(100%)	1(3.3%)
Handedness	Left	6(20.0%)	2(6.7%)	2(6.7%)	2(6.7%)	0.217
	Right	24(80.0%)	28(93.3%)	28(93.3%)	28(93.3%)
Secondary Schooling System	National School	30(100%)	19(63.3%)	29(96.7%)	16(53.3%)	<0.001*
	Others	0(0%)	11(36.7%)	1(3.3%)	14(46.7%)

Pearson’s Chi Square test was performed.

*p-value<0.05

Frequency differences of individuals who exhibited the allographic features, and those who did not, were determined using Pearson’s Chi Square test (Table 2).

Table 2

Association of allographic features with educational backgrounds using Pearson’s Chi Square test.

Allographic Features	Group				χ2 (df)	p-value
	National school	Chinese vernacular school	Tamil vernacular school	Islamic religious school
Uppercase "A" with three individual strokes	17 (56.7%)	8 (26.7%)	10 (33.3%)	8 (26.7%)	7.937 (3)	0.047*
Uppercase "A" with crossbar exceeding the letter boundary	6 (20.0%)	7 (23.3%)	5 (16.6%)	19 (63.3%)	20.124 (3)	< 0.001*
Uppercase "B" with two individual strokes	9 (30.0%)	16 (53.3%)	17 (56.7%)	6 (20.0%)	11.955 (3)	0.008*
Uppercase "B" with no protruding initial stroke	14 (46.7%)	18 (60.0%)	24 (80.0%)	10 (33.3%)	14.411 (3)	0.002*
Uppercase "D" with two individual strokes	12 (40.0%)	23 (76.7%)	21 (70.0%)	10 (33.3%)	16.835 (3)	0.001*
Uppercase "D" with protruding initial stroke	21 (70.0%)	3 (10.0%)	7 (23.3%)	8 (26.7%)	27.768 (3)	< 0.001*
Uppercase "H" with commanding stroke written first followed by crossbar and vertical stroke	15 (50.0%)	25 (83.3%)	10 (33.3%)	11 (36.7%)	18.772 (3)	< 0.001*
Uppercase "I" written as single stroke (vertical stroke) without upper and lower serif	16 (53.3%)	9 (30.0%)	8 (26.7%)	16 (53.3%)	7.830 (3)	0.050
Uppercase "P" with one continuous stroke	13 (43.3%)	10 (33.3%)	8 (26.7%)	15 (50.0%)	4.089 (3)	0.252
Lowercase "p" with one continuous stroke	16 (53.3%)	12 (40.0%)	9 (30.0%)	25 (83.3%)	19.335 (3)	< 0.001*
Uppercase "T" with vertical stroke written first followed by arm	7 (23.3%)	5 (16.6%)	8 (26.7%)	15 (50.0%)	9.156 (3)	0.027*
Lowercase "t" with crossbar written first followed by vertical stroke	0 (0%)	25 (83.3%)	4 (13.3%)	0 (0%)	78.348 (3)	< 0.001*
Lowercase "w" with left and right sections apart	10 (33.3%)	20 (66.7%)	13 (43.3%)	22 (73.3%)	12.990 (3)	0.005*
Lowercase "w" with sharp left and right vertices	6 (20.0%)	10 (33.3%)	4 (13.3%)	18 (60.0%)	17.175 (3)	0.001*
Uppercase "X" with right to left diagonal stroke constructed first followed by left to right diagonal stroke	13 (43.3%)	11 (36.7%)	11 (36.7%)	27 (90.0%)	23.893 (3)	< 0.001*
Lowercase "x" with right to left diagonal stroke constructed first followed by left to right diagonal stroke	13 (43.3%)	8 (26.7%)	11 (36.7%)	25 (83.3%)	22.289 (3)	< 0.001*
Uppercase "Z" with rounded top arm-junction design	17 (56.7%)	8 (26.7%)	9 (30.0%)	11 (36.7%)	6.933 (3)	0.074
Lowercase "z" with rounded top arm-junction design	17 (56.7%)	8 (26.7%)	10 (33.3%)	11 (36.7%)	6.345 (3)	0.096

There was significant association between educational backgrounds and 14 individual allographic features from a total of 18 features. Letter "t" with crossbar written first followed by vertical stroke gave the highest discriminatory power (χ2 = 78.348, df = 3, p < 0.001) and was observed in more than 83% of the handwriting samples of writers from Chinese vernacular school group. By comparison, this feature was only found in approximately 13% of samples obtained from individuals from Tamil medium vernacular school background, and none of this feature was noted in the other two study groups.

Other discriminative features included the construction of letters "D" with protruding initial stroke (χ2 = 27.768, df = 3, p < 0.001), both "X" (χ2 = 23.893, df = 3, p < 0.001) and "x" (χ2 = 22.289, df = 3, p < 0.001) with right to left diagonal stroke constructed first followed by left to right diagonal stroke, as well as letter "A" with crossbar exceeding the letter boundary (χ2 = 20.124, df = 3, p < 0.001). Multinomial logistic regression analysis was subsequently carried out (Table 3 is referred). The statistical outcome was nominal variable with four categories of educational background with one of the educational backgrounds served as the reference group. Each allographic feature was treated as an independent variable.

Table 3

Probability of handwritten allographic features from writers of different educational backgrounds by multinomial logistic regression analysis.

Allographic Features	Group	Multinomial Logistic Regression
		OR	95%CI of OR		p-value
Uppercase "A" with three individual strokes	A (Ref) vs B	0.271	0.077	0.959	0.043 *
	A (Ref) vs C	0.34	0.103	1.122	0.077
	A (Ref) vs D	0.233	0.066	0.824	0.024 *
	B (Ref) vs C	1.254	0.389	4.041	0.704
	B (Ref) vs D	0.859	0.253	2.914	0.808
	C (Ref) vs D	0.685	0.189	2.482	0.565
Uppercase "A" with crossbar exceeding the letter boundary	A (Ref) vs B	1.367	0.336	5.566	0.663
	A (Ref) vs C	1.006	0.242	4.184	0.994
	A (Ref) vs D	5.779	1.571	21.123	0.008 * #
	B (Ref) vs C	0.736	0.189	2.861	0.658
	B (Ref) vs D	4.227	1.296	13.792	0.017 * #
	C (Ref) vs D	5.745	1.512	21.828	0.010 * #
Uppercase "B" with two individual strokes	A (Ref) vs B	0.971	0.267	3.597	0.976
	A (Ref) vs C	1.522	0.45	5.149	0.500
	A (Ref) vs D	0.26	0.058	1.159	0.077
	B (Ref) vs C	1.552	0.495	4.861	0.451
	B (Ref) vs D	0.265	0.074	0.947	0.041 * #
	C (Ref) vs D	0.171	0.042	0.699	0.014 * #
Uppercase "B" with no protruding initial stroke	A (Ref) vs B	0.629	0.177	2.24	0.474
	A (Ref) vs C	2.276	0.649	7.983	0.199
	A (Ref) vs D	0.316	0.087	1.149	0.08
	B (Ref) vs C	3.619	0.997	13.131	0.05
	B (Ref) vs D	0.503	0.155	1.628	0.251
	C (Ref) vs D	0.139	0.036	0.543	0.005 * #
Uppercase "D" with two individual strokes	A (Ref) vs B	1.813	0.492	6.681	0.371
	A (Ref) vs C	1.498	0.441	5.089	0.517
	A (Ref) vs D	0.354	0.094	1.334	0.125
	B (Ref) vs C	0.827	0.228	2.99	0.772
	B (Ref) vs D	0.195	0.056	0.677	0.010 * #
	C (Ref) vs D	0.236	0.062	0.906	0.035 * #
Uppercase "D" with protruding initial stroke	A (Ref) vs B	0.051	0.01	0.259	<0.001 * #
	A (Ref) vs C	0.186	0.052	0.656	0.009 * #
	A (Ref) vs D	0.146	0.039	0.543	0.004 * #
	B (Ref) vs C	3.619	0.726	18.033	0.117
	B (Ref) vs D	2.852	0.629	12.941	0.174
	C (Ref) vs D	0.788	0.198	3.138	0.736
Uppercase "H" with commanding stroke written first followed by crossbar and vertical stroke	A (Ref) vs B	5.341	1.326	21.515	0.018 * #
	A (Ref) vs C	0.51	0.157	1.658	0.263
	A (Ref) vs D	0.504	0.152	1.669	0.262
	B (Ref) vs C	0.096	0.026	0.352	<0.001 * #
	B (Ref) vs D	0.094	0.025	0.355	<0.001 * #
	C (Ref) vs D	0.988	0.281	3.474	0.985
Lowercase "p" with one continuous stroke	A (Ref) vs B	1.107	0.32	3.836	0.873
	A (Ref) vs C	0.662	0.202	2.164	0.495
	A (Ref) vs D	5.813	1.425	23.708	0.014 *
	B (Ref) vs C	0.598	0.184	1.943	0.392
	B (Ref) vs D	5.25	1.471	18.746	0.011 * #
	C (Ref) vs D	8.783	2.184	35.316	0.002 * #
Uppercase "T" with vertical stroke written first followed by arm	A (Ref) vs B	0.755	0.178	3.201	0.703
	A (Ref) vs C	1.33	0.362	4.887	0.668
	A (Ref) vs D	4.307	1.229	15.09	0.022 *
	B (Ref) vs C	1.76	0.471	6.585	0.401
	B (Ref) vs D	5.703	1.571	20.697	0.008 * #
	C (Ref) vs D	3.239	0.918	11.431	0.068
Lowercase "t" with crossbar written first followed by vertical stroke	A (Ref) vs B				-
	A (Ref) vs C				-
	A (Ref) vs D				-
	B (Ref) vs C	0.015	0.002	0.101	<0.001 * #
	B (Ref) vs D				-
	C (Ref) vs D				-
Lowercase "w" with left and right sections apart	A (Ref) vs B	4.64	1.322	16.291	0.017 *
	A (Ref) vs C	1.698	0.525	5.493	0.376
	A (Ref) vs D	5.315	1.523	18.551	0.009 * #
	B (Ref) vs C	0.366	0.119	1.123	0.079
	B (Ref) vs D	1.145	0.351	3.737	0.822
	C (Ref) vs D	3.129	0.917	10.684	0.069
Lowercase "w" with sharp left and right vertices	A (Ref) vs B	2.971	0.77	11.47	0.114
	A (Ref) vs C	0.826	0.187	0.3642	0.800
	A (Ref) vs D	7.99	2.518	29.587	0.002 * #
	B (Ref) vs C	0.278	0.072	1.076	0.064
	B (Ref) vs D	2.689	0.869	8.326	0.086
	C (Ref) vs D	9.677	2.359	39.694	0.002 * #
Uppercase "X" with right diagonal stroke constructed first followed by left diagonal stroke	A (Ref) vs B	0.622	0.175	2.206	0.462
	A (Ref) vs C	0.735	0.227	2.38	0.608
	A (Ref) vs D	10.984	2.381	50.66	0.002 * #
	B (Ref) vs C	1.182	0.376	3.716	0.774
	B (Ref) vs D	17.659	3.996	78.049	<0.001 * #
	C (Ref) vs D	14.937	3.163	70.546	0.001 * #
Lowercase "x" right diagonal stroke constructed first followed by left diagonal stroke	A (Ref) vs B	0.368	0.099	1.372	0.137
	A (Ref) vs C	0.696	0.214	2.26	0.546
	A (Ref) vs D	5.764	1.517	21.898	0.010 * #
	B (Ref) vs C	1.889	0.569	6.275	0.299
	B (Ref) vs D	15.645	4.055	60.354	<0.001 * #
	C (Ref) vs D	8.281	2.097	32.705	0.003 * #

*p-value<0.05

#Power of study ≥78%

OR = Odds Ratio; CI = Confidence Interval; Ref = Reference group; All models were adjusted with two confounding variables (age and secondary schooling system): Group A- National school; Group B- Chinese vernacular school; Group C- Tamil vernacular school; Group D- Islamic religious school.

Table 3 demonstrates the coefficients of the multinomial logistic regression model, with statistically significant coefficients highlighted. A p-value of less than 0.05 allowed for comparison between the two groups of participants in relation to their handwritten allographic features. In our case, a value of more than 1 in the odd ratio (OR) indicated a positive sign while OR less than 1 suggested a negative sign. Considering the feature of the uppercase "A" with crossbar exceeding the letter boundary, a positive sign indicated that it is more likely that the writer is from the Islamic School (Group D) than the National School (Group A- Ref group) if he/she had the feature in his/her handwriting (OR = 5.779, 95%CI: 1.571, 21.123, p-value = 0.008). On the other hand, a negative sign indicated that the writer is more likely to have originated from the Chinese Vernacular School (Group B) than the National School (Group A- Ref group) if he/she does not construct his/her uppercase "A" with three individual strokes (OR = 0.271, 95%CI: 0.077, 0.959, p-value = 0.043). The multinomial logistic regression model had allowed for the determination of specific handwritten features based on respective educational backgrounds.

This study served as an exploratory study to investigate the presence of significant differences among handwritings from individuals from the same country but with different educational backgrounds. The powers of study for each significant variable were tested, and those variable with more than 78% of power were highlighted (Table 3 is referred). This statistical power suggested that the test results were likely valid with low Type II error. Nonetheless, the handwritten allographic features should be further tested with greater number of samples. Handwritten feature with low powder of study (<70%) is suggested for further investigation with greater sample size, in this case the uppercase "A" with three individual strokes. For the current study, only those statistically significant features with adequate power (in this case greater than 78%) are discussed.

Discussion

Handwritten allographic features of Islamic religious school background

In general, individuals who had undergone their primary education in Islamic religious school (Group D), were more likely to carry three characteristic allographic features in their English handwriting as compared to the other groups. The three features are the: (1) uppercase "A" with crossbar exceeding the letter boundary [Fig 2(A)], (2) uppercase "X" [Fig 2(B)], and (3) lowercase "x" [Fig 2(C)] where right to left diagonal strokes were constructed first followed by left to right diagonal strokes. Islamic religious school students learn Arabic language as the main language for religion-related subjects. Unlike English, Arabic letters are written from right to left in constructing script. Shapes and construction of the script could vary based on their position in a word, whether it is written initially, medially, finally or as a separate letter; relative to its adjacent letter [14]. Arabic script is generally cursive and is usually written continuously to connect one letter to another while forming a word. In writing the uppercase "X" and lowercase "x", writers in Group D were prone to construct the right to left diagonal stroke first, most probably due to the habit adopted while learning to write Arabic scripts. The observation of significantly higher frequency of long crossbar exceeding boundary of letter "A" in this current study, could also be due to the characteristics of Arabic script which connects one letter to another in a word.

Fig 2

Characteristic handwritten allographic features from writers from Islamic religious schools.

English handwriting writers from Islamic religious school background were also characterized by the construction of respective letters "B" [Fig 2(D)], "D" [Fig 2(E)], and "p" [Fig 2(F)]. The construction of these letters is significantly different from handwritings produced by Chinese and Tamil medium vernacular school backgrounds individuals. As the Arabic scripts are mostly cursive, and the letters are usually connected, writers in Group D preferred to construct these letters in a single stroke. In other words, writers in Group B and Group C were more likely to form the letters "B", "D", and "p" with at least two individual strokes with the presence of at least one pen lift. There was no significant difference between writers in Group A and Group D for these three letters as they could have also learned Arabic language and Jawi—a writing with very similar formation in Arabic writing during formal schooling.

Significantly high frequencies of lowercase "w" with sharp left and right vertices [Fig 2(G)], as well as left and right sections apart [Fig 2(H)]; were evident in handwriting of individuals from Group D compared to those from Group A. Distinct and pointed vertices of the "w" had discriminated these two groups. Although Arabic script runs from right to left, it seems that there was no significant impact being imparted on the direction of letter formation in English handwriting of the writers from both groups. Clockwise or anti-clockwise direction in letter construction, such as "O" and "Q" were also reported to be not discriminative based on the examination of handwriting samples from these two groups.

Handwritten allographic features of National school background

Uppercase "D" with protruding initial stroke [Fig 3(A)] was found to be a characteristic feature in English handwriting of individuals from Group A. The writers of Group A also preferred to construct uppercase "B" with protruding initial stroke [Fig 3(B)], compared to handwriting from Group C writers who prominently construct "B" without the protruding initial stroke. This indicates that writers from the National school background preferred writing "B" and "D" with the second stroke extending from the base of the letter from the end of the first stroke instead of starting from the cap-height. Consequently, such writing habits led to the formation of protruding branch-like stroke at the initial start of the letters.

Fig 3

Characteristic handwritten allographic features from writers from National schools.

Handwritten allographic features of Chinese vernacular school background

While constructing a Chinese character with both horizontal and vertical strokes, writers are more likely to make the former followed by the latter in sequence. This was evident in the construction of letter "t" [Fig 4(A)] by Group B writers. This feature was also found to be the most prominently discriminating feature with the highest discriminating power. Letter "T" [Fig 4(B)] also demonstrated the preference of Group B writers to construct the character starting with the arm first followed by the vertical stroke, differentiating them from Group D writers, who prefer writing the other way. It was important to note that this feature concerning letter “T” did not appear distinguishable between the individuals from Groups A, B, and C. However, Cheng et al. [10] reported that this feature was found to be one of the class characteristics of Chinese writers, making their handwriting distinguishable from the handwriting of Malay and Indian writers.

Fig 4

Characteristic handwritten allographic features from writers from Chinese vernacular schools.

The basic rule in forming a Chinese character is through the sequence of strokes from the top to the bottom, and from the left to the right. The manner of stroke construction in writing Chinese characters had in a way dominated and influenced the handwriting of English letters, including letter “H” [Fig 4(C)] in this present study. Writers from the Chinese vernacular school were found more likely to construct the uppercase "H" by forming the commanding vertical stroke from top to bottom first, followed by the horizontal crossbar from left to right, and lastly a vertical stroke to complete the letter.

Handwritten allographic features of Tamil vernacular school background

Individuals from Tamil vernacular school use Tamil language as the medium for learning. Construction of characters in the Tamil writing system involves various combinations of vowel and consonant shapes. Strokes of Tamil syllables are generally curved and rounded. Some syllables have curls and loops at its terminal end. Additionally, the letters appear individually, and they are not cursive or connected to its adjacent letter as demonstrated in Arabic writing system. These letters are also written in the direction from left to right as in English writing.

In this study, it was found that only one allographic feature, namely the letter “B” with no protruding initial stroke (Fig 5), appeared as a characteristic in Group C writers that differentiated them from Group D writers. In the latter group, writers were more likely to generate this letter with a protruding initial stroke. Although other letters such as "D", "P", and "p" have similar stroke numbers and sequence as "B", the absence of protruding initial stroke did not appear consistently in these letters.

Fig 5

The only handwritten allographic feature from writers from Tamil vernacular schools.

Cheng et al. [10] reported that rounded apex of letter “A” and rounded vertex of letter "W" and "V" were exhibited in the handwriting of Indians. However, we have observed otherwise in this study. The design of apex of letter “A”, whether it is sharp or rounded, was not seen to be a feature which could distinguish between the groups herein. The said feature seems to randomly occur in the handwriting of all the writers regardless of their schooling background. Shah and Dahiya [6] found that rounded apex of letter "A" was an attribute observed among the Hindi writers in India instead of Tamilians. Most Indians in Malaysia are Tamilians and almost all writers who attend Tamil medium vernacular school do not have Hindi language knowledge. Letter "w" with rounded vertices were found significant to differentiate handwriting samples of Group C individuals from writers of Group D but not for the other two groups. In fact, this study observed that allographic feature representing Indian vernacular school characteristics was very limited.

This study adopted the detailed examination of individual allographs on a subset of 30 handwriting samples from four different school systems to identify if there is any characteristic and discriminative allographic features amongst writers. Based on the frequency of occurrence as the indicative of handwriting characteristics in a study group, 13 allographic features were identified. The number of features identified were greater than the six class characteristics identified by Cheng et al. [10] in the English handwriting of Singaporeans in a similar study. However, it is worth noting that Singapore has a limited language choice education system compared to the multi educational system in the current study.

Handwritings of Islamic religious school background individuals were found to have carried more unique characteristics, particularly in letters "A", "B", "D", "p", "w", "X" and "x". Writers from Chinese vernacular schools exhibited special features in letters "H", "T" and "t" as discriminative characteristics, especially the lowercase "t" which was observed in more than 80% of the handwriting samples from Group B and less than 14% in handwriting samples of other groups. Handwritings of individuals from National schools and Indian vernacular schools had less discriminative allographic features in their respective handwriting samples. Only "B" and "D", and "B" and "w" showed some discriminative characteristics in Group A and Group C, respectively.

In this study, the 13 allographic features revealed in Malaysians’ handwriting could serve as indicators of their primary educational background. Along with mother-tongue language learnt in the respective vernacular schools and Islamic religious schools, an individual also learns English and Malay languages as compulsory languages in Malaysian schools. Exposure to these languages may also influence one’s style of writing and result in specific handwriting characteristics. The findings herein suggest that the potential influence of taught-writing style and copybook systems in the English handwritings of writers, and that the formation of letters could be strongly associated to an individual’s primary educational background. It is also possible for forensic document examiners to consider these allographic features in giving opinions during handwriting examination and to provide insight into the likely identity of the writer, especially when a reference sample is not available [11, 15].

By further including the multinomial logistic regression model in this study, it allowed for the understanding of which primary education system has a higher probability of shaping specific handwritten allographic features, adjusted with confounding variables. Variations in school backgrounds, or more specifically, the unique learning systems reported herein were believed to have influenced the habitual formation of letters during writing [1, 2, 16].

Previous studies had suggested foreign influence toward the formation of handwriting, particularly in countries that have welcomed a huge number of immigrants in the past [4–7, 17]. This study had further proposed that even when individuals are from the same country, their handwriting could be differed, probably due to their respective taught writing styles. It was important to note that school background is not the only factor underlying handwriting style, as other factors such as the adequacy of standards [1, 2], health impairment [1, 2, 18–20], the influence of drugs and alcohol [1, 2, 21–23], as well as the concentration level during the writing procedure [1, 2] might also contribute to differences in writing. In this study, only writers who fulfilled the inclusion and exclusion criteria were recruited to minimize the effect of above-mentioned potential factors.

Since this study involved only the validated data as examined by the research team including a certified senior forensic document examiner, the inter-rater reliability was not investigated. Based on the identified features found in the current study, a greater number of observers coding the same handwriting samples is proposed to determine the inter-rater reliability among different individuals in future studies. It is also worth exploring the reliability between experienced examiners and individuals with no prior handwriting analysis experience to establish the validity of handwritten allographic features in grouping the writers.

The presence of handwriting allographic features common to a group of writers can be further expanded to other parameters, including gender, ethnicity, age group, handedness and any other relevant information contributing to forensic intelligence. Serving as an exploratory study, the examination of handwritten allographic features was restricted to 30 writers of limited specimens from each group of study. It is acknowledged that a greater population of handwriting samples could further explore the influence of copybook system and taught writing style during writers’ primary education.

Our on-going research on a larger population size would certainly allow for determination of allographic features in the handwriting of different groups coupled with analysis of each characteristic based on educational learning systems, with the hope that it will enable the comparison of different handwriting profiles for better forensic intelligence. The collation of information on a source or an author coming from a particular group or background education system through a forensic intelligence framework would be beneficial to the investigation of document related disputes and for use in criminal or civil court trials.

Conclusion

Learned writing systems is one of the factors that contribute to one’s writing style. This study investigated the common handwritten allographic features in English handwritings of writers from different primary education backgrounds. 13 allographic features were identified and they could be attributed to a specific primary educational background the writer had undergone. In routine handwriting examination, the comparison of a questioned sample with reference samples from a known writer is frequently conducted with the aims to identify the author or to detect a forged signature. However, there are instances where comparable writing samples are not available, commonly due to inability to trace a suspect. Nonetheless, the examination of questioned handwriting can still be expanded to retrieve useful information thus narrowing down the possible groups of potential writers, in this case, providing information on the educational backgrounds. It is hoped that determination of these features can then promote the handwriting examination to a wider application to trace or link a source or person for both criminal and civil forensic document investigation, especially when countries are populated with people from different educational backgrounds due to increases in cross border activities and migrations.

(XLSX)

Click here for additional data file.

9 Sep 2021

An exploratory study on the handwritten allographic features of multi-racial population with different educational backgrounds

PLOS ONE

Dear Dr. Abdullah,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

I would like to sincerely apologize for the delay you have incurred with your submission. It has been exceptionally difficult to secure reviewers to evaluate your study. We have now received five completed reviews; their comments are available below. The reviewers have raised significant scientific concerns about the study that need to be addressed in a revision.

Please revise the manuscript to address all the reviewer's comments in a point-by-point response in order to ensure it is meeting the journal's publication criteria. Please note that the revised manuscript will need to undergo further review, we thus cannot at this point anticipate the outcome of the evaluation process.

Please submit your revised manuscript by Oct 23 2021 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols.

We look forward to receiving your revised manuscript.

Kind regards,

Miquel Vall-llosera Camps

Senior Editor

PLOS ONE

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf.

2. In your Data Availability statement, you have not specified where the minimal data set underlying the results described in your manuscript can be found. PLOS defines a study's minimal data set as the underlying data used to reach the conclusions drawn in the manuscript and any additional data required to replicate the reported study findings in their entirety. All PLOS journals require that the minimal data set be made fully available. For more information about our data policy, please see http://journals.plos.org/plosone/s/data-availability.

Upon re-submitting your revised manuscript, please upload your study’s minimal underlying data set as either Supporting Information files or to a stable, public repository and include the relevant URLs, DOIs, or accession numbers within your revised cover letter. For a list of acceptable repositories, please see http://journals.plos.org/plosone/s/data-availability#loc-recommended-repositories. Any potentially identifying patient information must be fully anonymized.

Important: If there are ethical or legal restrictions to sharing your data publicly, please explain these restrictions in detail. Please see our guidelines for more information on what we consider unacceptable restrictions to publicly sharing data: http://journals.plos.org/plosone/s/data-availability#loc-unacceptable-data-access-restrictions. Note that it is not acceptable for the authors to be the sole named individuals responsible for ensuring data access.

We will update your Data Availability statement to reflect the information you provide in your cover letter.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #2: Partly

Reviewer #3: Partly

Reviewer #4: Partly

Reviewer #5: Yes

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: I Don't Know

Reviewer #3: Yes

Reviewer #4: No

Reviewer #5: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: No

Reviewer #3: No

Reviewer #4: No

Reviewer #5: No

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

Reviewer #3: Yes

Reviewer #4: Yes

Reviewer #5: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: Writing styles are culturally driven throughout writing learning and teaching. Graphic traditions for a singe writing system may push learners and writers toward common attitude and behavior in reproducing similar styles. The study is very much oriented toward this kind of consideration on individual variation; I would recommend to highlight better relevant applications of the authors' results.

Reviewer #2: General comments

1. In the introduction, the Rationale could be emphasized since the reader could not have any clue understanding of why it is important to study this question and Plos one is not a journal specialized in forensics. The discussion is clearer about the purpose for the study e.g. l300-301

2. The authors mentioned that “All relevant data are within the manuscript. “However, raw data and analysis scripts are not available? Why? Would it possible to publish them in Open Science framework for instance? That would improve replicability and transparency of this work, e.g. comparing with text written in other countries

e.g. Drotár, P., & Dobeš, M. (2020). Dysgraphia detection through machine learning. Scientific reports, 10(1), 1-11.

3. The methodology of selection of letters allographs is unclear “Based on the handwriting samples, eighteen allographic features were suggested as the characteristics which potentially occurred in one group as compared to other groups, covering 14 letters”

4. Were Level of education, sex and handedness data collected and could they

be confounding variables? Were the groups well balanced concerning these variables? Consider presenting a (demographic) table

e.g. Gargot, T., Asselborn, T., Pellerin, H., Zammouri, I., M. Anzalone, S., Casteran, L., ... & Jolly, C. (2020). Acquisition of handwriting in children with and without dysgraphia: A computational approach. PLoS One, 15(9), e0237575

5. What could be the role of new technology? Using scan or electronic tablets?

e.g. Yogarajah, P., & Bhushan, B. (2020, November). Deep Learning Approach to Automated Detection of Dyslexia-Dysgraphia. In The 25th IEEE International Conference on Pattern Recognition.

Or

Asselborn, T., Gargot, T., Kidziński, Ł., Johal, W., Cohen, D., Jolly, C., & Dillenbourg, P. (2018). Automated human-level diagnosis of dysgraphia using a consumer tablet. NPJ digital medicine, 1(1), 1-9.

6. L120 Results and discussions should be separated

7. Would it be possible to identidy the race of a writer from features identified in this paper? Would this method reliable? Would it be possible to compute a sensibility/specificity and intraclass correlation between raters? Would it be next steps and how?

Specific comments

8. L109 All identified handwritten allographic. How many were not identified? How many subjects were approached and eligible? Consider a flow chart. See for instance https://www.equator-network.org/reporting-guidelines/stard/

9. A p-value <0.05 was considered statistically significant. The methodology of such study could lead to false positives. Did authors have any priory hypothesis? Did they pre-registered them? Did they consider a Bonferroni correction?

10. Table 2. Tables should be readable alone. Consider describing the groups instead of calling them A,B,C and D

11. The authors did not report any limitation paragraph in their discussion

Typos :

Abstract and l8 : “to selected population” : not adding anything

L 5. “in the country of this study” - >in Malaysia

L 106. there was error-free ? ->were error-free.

L 161. From the results,” doesn’t had anything”

l 19 “Litterature suggest” doesn’t had anything.

L138. « was demonstrated » -> was reported

L285 “were found to have possessed“ -> had

L21 : in western countries

L91. Dpi : Dot per inch should be also explained in full text

Reviewer #3: The manuscript is interesting and of value in the context of forensic science. However, some revisions are needed:

1. The examination of handwritten samples: it is unclear who performed this examination? Forensic experts? or people without experience? Then, how it was performed and how it was coded; more information on procedure is needed.

2. There is no information on age, sex of participants, please report M and SD. How reading and writing capacities of participants were controlled/tested? Is it is possible that they have any visual and motor impairments? Is this was controlled? It is possible that they differ in intelligence level, is it was controlled? And how?

3. There is no limitation of the presented study. Please describe the limitations, it seems they are several.

Reviewer #4: The overall manuscript is readable and can be understood with ease. It though cannot hold on to the readers attention due to simple write up of results.

More comparative studies in the area can be citied so as to develop the introduction of the paper.

Some studies that can be looked at are:

1. van der Plaats, R.E., van Galen, G.P.

9534939400;57200608568;

Allographic variability in adult handwriting

(1991) Human Movement Science, 10 (2-3), pp. 291-300. Cited 6 times.

https://www.scopus.com/inward/record.uri?eid=2-s2.0-28144445689&doi=10.1016%2f0167-9457%2891%2990008-L&partnerID=40&md5=1c9c01fffdfaab36f1d7ebd1e49c5f05

DOI: 10.1016/0167-9457(91)90008-L

AFFILIATIONS: NICI, University of Nijmegen, Nijmegen, Netherlands

Demographic details other than the age of respondents (18-20) are missing in the paper.

2. Tan, G.X., Viard-Gaudin, C., Kot, A.C.

35300856200;9133978000;35588578100;

Online writer identification using alphabetic information clustering

(2009) Proceedings of SPIE - The International Society for Optical Engineering, 7247, art. no. 72470F, .

https://www.scopus.com/inward/record.uri?eid=2-s2.0-62249120403&doi=10.1117%2f12.805644&partnerID=40&md5=b4210fff315309017e4b63a0c5def5b8

DOI: 10.1117/12.805644

AFFILIATIONS: Centre for Information Security, Nanyang Technological University of Singapore, SINGAPORE, Singapore;

IRCCyN-UMR CNRS 6597, Ecole Polytechnique de l'Université de Nantes, France

The authors mention a researcher team but do not provide an interrater agreement score, so as to defend the choice of letters used in the study.

The number of samples for each study group does not seem adequate to conclude on the results.

The paper should have a table that shows examples of original handwriting of all four groups for highlighting each letter studied for comparing allographic features.

The conclusion can be more elaborate, may be include some examples of forensic cases wherein such allographic study was useful in identifying the criminal and the importance of such research in the country of study.

The manuscript misses data doi.

The authors can mention opportunities in future studies

Reviewer #5: The experimental design of this paper is rigorous, and there is also a scientific hypothesis test

method to some extent.

There are some problems. First, the article lacks a detailed description of data descriptive analysis,

and has not analyzed the statistical characteristics and distribution of the collected data.

In addition, the amount of data collected is only 120, which is relatively small in the experiment of a

large number of results to be verified. It will inevitably cause some test to show significant due to

small samples, distribution imbalance or other accidental factors.

Finally, the data is collected scientific and detailed, but the analysis method is single, which is a Chi

Square test, and the rest is multi-character analysis description and conclusions, lack of

effectiveness and innovation.

Overall, in terms of methods, the whole paper is not innovative enough. In short, the author just uses

Pearson's chi-square test to analyze the new data set collected.

In addition, some parts are confusing, which also significantly deteriorate the clarity and readability

of the paper:

1.Pearson’s chi-square test is used in the Statistical analysis section of this paper. However, the

expression of this method is not clear enough, and there is no detailed mathematical formula

introduction and citation description. For those who are not familiar with Pearson's chi-square test,

it is difficult to understand this method.

2.In table 2, it is not explained how the frequency characteristics are derived;

3.There is no explaination about what Group A, Group B, Group C and Group D represent in table 2；

4.The results for multiple comparison are introduced inTable 3, but there is no specific introduction

on how to perform multiple comparisons.

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: Yes: Prof. Barbara Turchetta

Reviewer #2: Yes: Thomas Gargot

Reviewer #3: No

Reviewer #4: Yes: Sheetal Thomas

Reviewer #5: No

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

12 Oct 2021

Dear Editors and Reviewers of PLOS ONE,

Thank you for your kind review and thoughtful comments. We appreciate the constructive feedback and have carefully taken the comments into consideration in preparing this manuscript. The following summarised our response to the comments:

Reviewer #1

Comment

Writing styles are culturally driven throughout writing learning and teaching. Graphic traditions for a single writing system may push learners and writers toward common attitude and behaviour in reproducing similar styles. The study is very much oriented toward this kind of consideration on individual variation; I would recommend to highlight better relevant applications of the authors' results.

Response

Thank you for the comment. The rationale and relevant application of the current study was included in the introduction (Line 55 – 68). In brief, handwriting examination frequently involves the comparison of an unknown writing sample with samples from a known writer. However, there were instances where suspect or victim is unknown or in the absence of comparable handwriting samples, that makes the application of conventional QDE methodology using known and suspect samples become impossible. Our study revealed that, through the examination of questioned handwriting, it could help in narrowing down the possible groups of potential writers and providing further information in document related fraud.

Reviewer #2

Comment 1.

In the introduction, the Rationale could be emphasized since the reader could not have any clue understanding of why it is important to study this question and Plos one is not a journal specialized in forensics. The discussion is clearer about the purpose for the study e.g. l300-301

Response 1.

Thank you for the suggestion. The rationale and importance of the study was included in the manuscript (Line 55 – 68). Although Plos One is not a specialised journal in forensic science; however, this study is transdisciplinary in nature where it could be explored from the perspectives of human sciences and digital sciences.

Comment 2.

The authors mentioned that “All relevant data are within the manuscript. “However, raw data and analysis scripts are not available? Why? Would it possible to publish them in Open Science framework for instance? That would improve replicability and transparency of this work, e.g. comparing with text written in other countries

e.g. Drotár, P., & Dobeš, M. (2020). Dysgraphia detection through machine learning. Scientific reports, 10(1), 1-11.

Response 2.

Thank you. For reference purpose, our raw data in the form of Excel file is supplied as supplementary file.

Comment 3.

The methodology of selection of letters allographs is unclear “Based on the handwriting samples, eighteen allographic features were suggested as the characteristics which potentially occurred in one group as compared to other groups, covering 14 letters”

Response 3.

Thank you for the comments. The procedure of examination and selection of the eighteen allographic features was described in the revised manuscript (Line 153-160).

Comment 4.

Were Level of education, sex and handedness data collected and could they be confounding variables? Were the groups well balanced concerning these variables? Consider presenting a (demographic) table

e.g. Gargot, T., Asselborn, T., Pellerin, H., Zammouri, I., M. Anzalone, S., Casteran, L., ... & Jolly, C. (2020). Acquisition of handwriting in children with and without dysgraphia: A computational approach. PLoS One, 15(9), e0237575

Response 4.

Thanks for suggestion. We have included the table describing the demographic factors (Table 2) and tested the differences among the groups. Those significant variables were adjusted as confounding variables in the final model.

Comment 5.

What could be the role of new technology? Using scan or electronic tablets?

e.g. Yogarajah, P., & Bhushan, B. (2020, November). Deep Learning Approach to Automated Detection of Dyslexia-Dysgraphia. In The 25th IEEE International Conference on Pattern Recognition.

Or

Asselborn, T., Gargot, T., Kidziński, Ł., Johal, W., Cohen, D., Jolly, C., & Dillenbourg, P. (2018). Automated human-level diagnosis of dysgraphia using a consumer tablet. NPJ digital medicine, 1(1), 1-9.

Response 5.

Thank you for the suggestion. The findings from this study would provide us the understanding on handwritten pattern among Malaysians. Subsequently, the collation of information on a source or an author coming from a particular group or background education system through a forensic intelligence framework would be beneficial to the investigation of document related cases. Additional information was included as future recommendation in the revised manuscript (Line 397-401, 412-414).

Comment 6.

L120 Results and discussions should be separated

Response 6.

As suggested by the reviewer, Results and Discussion were separated into two sections.

Comment 7.

Would it be possible to identify the race of a writer from features identified in this paper? Would this method reliable? Would it be possible to compute a sensibility/specificity and intraclass correlation between raters? Would it be next steps and how?

Response 7.

In this study, the race of a writer was demonstrated in Table 2. It is noted that majority of the writers from the respective school group is belong to one race group. For instance, 100% of respondents from Chinese and Tamil schools were non-Malay. Therefore, this variable was not tested for its possibility for discrimination. In this study, only two confounding variables (age and secondary schooling system) were adjusted in the final model (Line 174-177).

Procedure of the examination and coding of the handwriting features was described in the revised manuscript (Line 153 – 160). In brief, the handwriting samples were examined in two different occasions by two teams of researcher independently, and their results were validated by a certified forensic document examiner. As future recommendation, inter-rater reliability study involving multiple raters examining and coding the same set of handwriting samples shall be conducted, upon determination of specific allographic features useful for discrimination among the groups (Line 397-401).

Comment 8.

L109 All identified handwritten allographic. How many were not identified? How many subjects were approached and eligible? Consider a flow chart. See for instance https://www.equator-network.org/reporting-guidelines/stard/

Response 8.

The sampling method and subject recruitment of this study was included in the revised manuscript (Line 80-86). This study was aimed to identify and establish the characteristic handwritten allographic features corresponding to selected population from different primary educational backgrounds.

Comment 9.

A p-value <0.05 was considered statistically significant. The methodology of such study could lead to false positives. Did authors have any priory hypothesis? Did they pre-registered them? Did they consider a Bonferroni correction?

Response 9.

Thank you for the comments. We have included multinomial logistic regression model, adjusted with confounding variables in the revised manuscript. Our aim of the study is to investigate the probability of respondents of having that handwriting features are more likely from which particular group.

Comment 10.

Table 2. Tables should be readable alone. Consider describing the groups instead of calling them A,B,C and D

Response 10.

Thank you for the suggestion. The groups were named according to their educational systems (Table 3 in the revised manuscript).

Comment 11.

The authors did not report any limitation paragraph in their discussion

Response 11.

Thank you for the suggestion. A separate section on limitations was included in the revised manuscript (Line 397 –408).

Comment 12.

Typos :

Abstract and l8 : “to selected population” : not adding anything

L 5. “in the country of this study” - >in Malaysia

L 106. there was error-free ? ->were error-free.

L 161. From the results,” doesn’t had anything”

l 19 “Litterature suggest” doesn’t had anything.

L138. « was demonstrated » -> was reported

L285 “were found to have possessed“ -> had

L21 : in western countries

L91. Dpi : Dot per inch should be also explained in full text

Response 12.

Thank you for the comments. Grammatical and typo errors were corrected accordingly.

Reviewer #3

Comment.

The manuscript is interesting and of value in the context of forensic science. However, some revisions are needed:

Response.

Thank you.

Comment 1.

The examination of handwritten samples: it is unclear who performed this examination? Forensic experts? or people without experience? Then, how it was performed and how it was coded; more information on procedure is needed.

Response 1.

Thank you for the suggestion. The procedure of handwriting examination was included in the revised manuscript (Line 153-160).

Comment 2.

There is no information on age, sex of participants, please report M and SD. How reading and writing capacities of participants were controlled/tested? Is it is possible that they have any visual and motor impairments? Is this was controlled? It is possible that they differ in intelligence level, is it was controlled? And how?

Response 2.

Thank you for the suggestion. The information on age, gender and other demographic data was included as described in Table 2.

The inclusion criteria of the writers who had participated in this study was described (Line 90-93). In general, they were university students who are mentally challenged, no physical disabilities and well-versed in reading and writing in English Language.

Comment 3.

There is no limitation of the presented study. Please describe the limitations, it seems they are several.

Response 3.

Thank you for the suggestion. Limitations of the study were included in the revised manuscript (Line 397 –408).

Reviewer #4

Comment 1.

The overall manuscript is readable and can be understood with ease. It though cannot hold on to the readers attention due to simple write up of results.

Response 1.

Thank you. The manuscript is revised to include additional figures and table for the better understanding of readers.

Comment 2.

More comparative studies in the area can be citied so as to develop the introduction of the paper.

Some studies that can be looked at are:

1. van der Plaats, R.E., van Galen, G.P.

9534939400;57200608568;

Allographic variability in adult handwriting

(1991) Human Movement Science, 10 (2-3), pp. 291-300. Cited 6 times.

2. Tan, G.X., Viard-Gaudin, C., Kot, A.C.

35300856200;9133978000; 35588578100;

Online writer identification using alphabetic information clustering

(2009) Proceedings of SPIE - The International Society for Optical Engineering, 7247, art. no. 72470F

Response 2.

The introduction part was revised where the rationale and relevant application of the current study was added. Thank you for the suggestions on articles; however, the suggested articles might not be fitted into the current study as it more focused on the influence of educational backgrounds. The suggested articles shall be considered in our subsequent study which aims on individualizing the allographic features and establishing the algorithm model for discrimination.

Comment 3.

The authors mention a researcher team but do not provide an interrater agreement score, so as to defend the choice of letters used in the study.

Response 3.

Thank you for the suggestion. The description on the examination and coding of allographic features was included in the revised manuscript (Line 153-160). An inter-rater reliability study shall be proposed, perhaps to involve expert and non-experts to investigate the reliability of the method in the current study as a future recommendation in the revised manuscript (Line 398-401).

Comment 4.

The number of samples for each study group does not seem adequate to conclude on the results.

Response 4.

Thank you. This study serves as the exploratory study to identify the allographic features for the discrimination among the writers of different educational backgrounds. In the revised manuscript, the power of study was also tested where majority of the identified features achieved more than 78% of power. The number of writers in the current study was also included as one of the limitations of this study, which deserving further investigation in the future with greater number of participants.

Comment 5.

The paper should have a table that shows examples of original handwriting of all four groups for highlighting each letter studied for comparing allographic features.

Response 5.

Thank you for the suggestion. Table 1 in the manuscript describes the allographic features used to compare the handwriting samples among the writers of different educational backgrounds. To provide better understanding, images of the characteristic allographic features identified in this study from the writers of different educational backgrounds were included in the revised manuscript (Figure 1-4).

Comment 6.

The conclusion can be more elaborate, may be include some examples of forensic cases wherein such allographic study was useful in identifying the criminal and the importance of such research in the country of study.

Response 6.

Thank you for the suggestion. The examples of forensic cases wherein such determination is important was included in the introduction (Line 55-68). The conclusion was also revised to include the usefulness of data gathered from the current study for forensic science application (Line 421-426).

Comment 7.

The manuscript misses data doi.

Response 7.

The data is submitted as supplementary file.

Comment 8.

The authors can mention opportunities in future studies

Response 8.

Thank you for the suggestion. Future recommendation subsequent to the current study was included (Line 398-401, 404-408).

Reviewer #5

Comment 1.

The experimental design of this paper is rigorous, and there is also a scientific hypothesis test method to some extent.

Response 1.

Thank you for the comments. The manuscript was revised based on the comments below.

Comment 2.

There are some problems. First, the article lacks a detailed description of data descriptive analysis, and has not analysed the statistical characteristics and distribution of the collected data.

Response 2.

Thank you for the suggestion. Data descriptive analysis was included in the manuscript (Table 2). The statistical characteristics and distribution were also analysed, and the confounding variables were considered in the subsequent analysis.

Comment 3.

In addition, the amount of data collected is only 120, which is relatively small in the experiment of a large number of results to be verified. It will inevitably cause some test to show significant due to small samples, distribution imbalance or other accidental factors.

Response 3.

Thank you for the comments. Post hoc power analysis to check the sample size was included in the current study, and the results showed that the study carried adequate power to detect the statistically significant difference. The power of this study for each significant variables ranges from 78% to 100%, which indicated that our study has enough power to detect statistically significant (Line 235-239).

Comment 4.

Finally, the data is collected scientific and detailed, but the analysis method is single, which is a Chi Square test, and the rest is multi-character analysis description and conclusions, lack of effectiveness and innovation.

Response 4.

Thank you for the suggestion. In addition to the Chi Square, a multinomial logistic regression analysis was conducted, and the results was adjusted with all possible confounding variables.

Comment 5.

Overall, in terms of methods, the whole paper is not innovative enough. In short, the author just uses Pearson's chi-square test to analyse the new data set collected.

Response 5.

Thank you. A multinomial logistic regression model was proposed in the revised manuscript.

Comment 6.

In addition, some parts are confusing, which also significantly deteriorate the clarity and readability of the paper:

1.Pearson’s chi-square test is used in the Statistical analysis section of this paper. However, the expression of this method is not clear enough, and there is no detailed mathematical formula introduction and citation description. For those who are not familiar with Pearson's chi-square test, it is difficult to understand this method.

Response 6.

Thank you for the comments. Subsequent to previous comment, a multinomial logistic regression analysis was carried out. Description on the establishment of the model was included in the methodology (Line 138-149).

Comment 7.

In table 2, it is not explained how the frequency characteristics are derived.

Response 7.

Table 2 was revised (Table 3 in the revised manuscript).

Comment 8.

There is no explanation about what Group A, Group B, Group C and Group D represent in table 2.

Response 8.

Thank you for the suggestion. The groups were named according to their educational systems (Table 3 in the revised manuscript).

Comment 9.

The results for multiple comparison are introduced in Table 3, but there is no specific introduction on how to perform multiple comparisons.

Response 9.

Thank you for the suggestion. The table on the multiple comparison was replaced by the results from multinomial logistic regression analysis (Table 4). The detailed on the statistical procedure was also included in the revised manuscript (Line 138-149).

This manuscript is checked thoroughly. We really appreciate the efforts taken by the editor and reviewers to help us improve the manuscript, and we hope that we have addressed all your concerns in this manuscript.

Thank you.

Ahmad Fahmi Lim Abdullah, PhD

School of Health Sciences, Universiti Sains Malaysia, 16150 Kubang Kerian, Kelantan, Malaysia. Email: fahmilim@usm.my; Tel: +609-7677596; Fax: +609-7677515.

Submitted filename: Responses to Reviewers.docx

Click here for additional data file.

1 Nov 2021

12 Dec 2021

Dear Editors and Reviewers of PLOS ONE,

Thank you for your kind review and thoughtful comments. We appreciate the constructive feedback and have carefully taken the comments into consideration in preparing this manuscript. The following summarised our response to the comments:

Reviewer #3

Comment 1:

The authors have improved the manuscript. They inserted the data on examination of handwritten samples and on participants.

Response 1:

Thank you.

Comment 2:

However there are still some aspects that need revision. Please include the inter-rater reliability between forensic coders. This should be presented in the manuscript. The limitations are still not elaborated.

Response 2:

Procedure of the examination and coding of the handwriting features was described in the manuscript in which the handwriting samples were examined in two different occasions by two teams of researcher independently, and their results were validated by a certified forensic document examiner. In this study, only the data validated by the document examiner was considered in interpretation and data analysis (Line 161-162).

As for the inter-rater reliability among the coders as suggested by the reviewer, a greater number of coders shall be proposed to involve multiple raters examining and coding the same set of handwriting samples, and this was not conducted in the scope of the current study. The inter-rater reliability study would be more useful upon the determination of specific allographic features for discrimination among the groups as demonstrated in the current study, perhaps to involve also expert and non-experts as a future recommendation. Such limitation and future recommendation were included in the revised manuscript (Line 407-413).

Comment 3:

Many factors can differentiate the forms of letter examined by the researchers, not only education and nationality. These factors such as intelligence or perceptive/manual impairments were not controlled in the present study. The researchers should be aware of these potential factors, particularly in the situation while the sample size is small.

Response 3:

The inclusion and exclusion criteria of the subjects of this study was included in the revised manuscript (Line 92-94). In general, they were university students who are mentally challenged, no physical disabilities and well-versed in reading and writing in English Language.

The limitation on the sample size was also included in the revised manuscript (Line 240-241, 418-420), deserving further investigation in the future with greater number of participants.

Comment 4:

The list of references is very short. Please explain the impact of these potential factors on handwriting in Discussion and say why these factors were not controlled as well as enrich the literature on this topic.

Response 4:

Additional references with related information were included in the revised manuscript (Line 397-401).

The potential factors which could affect the development of handwriting were also discussed in the revised manuscript (Line 401-406).

This manuscript is checked thoroughly. We really appreciate the efforts taken by the editor and reviewers to help us improve the manuscript, and we hope that we have addressed all your concerns in this manuscript.

Thank you.

Ahmad Fahmi Lim Abdullah, PhD

School of Health Sciences, Universiti Sains Malaysia, 16150 Kubang Kerian, Kelantan, Malaysia. Email: fahmilim@usm.my; Tel: +609-7677596; Fax: +609-7677515.

Submitted filename: Responses to Reviewers.docx

Click here for additional data file.

15 Mar 2022

11 Apr 2022

Dear Editors and Reviewers of PLOS ONE,

Thank you for your kind review and thoughtful comments. We appreciate the constructive feedback and have carefully taken the comments into consideration in preparing this manuscript. The following summarised our response to the comments:

Editor

Comment:

1) Several reviewers previously raised methodology concerns, including sample size, statistical analysis, etc. Although this has been addressed in the study as being relevant for an exploratory study, please provide some additional justification for these issues.

Response:

Thank you for the comment. The justification on the sample size determination and choice of statistical analysis was included in the revised manuscript.

Line 88-91 - “Sample size was calculated through two proportion formula by considering the letter “U” with introductory stroke [4], 25% of the Polish (0% for English) possessed the handwritten feature. With 80% of power study, 5% of risk error, and 10% of non-response rate, 30 writers were collected from each group.”

Line 386-388 - “By further including the multinomial logistic regression model in this study, it allowed for the understanding of which primary education system has a higher probability of shaping specific handwritten allographic features, adjusted with confounding variables.”

Comment:

2) There are some outstanding language issues which, if rectified, would strengthen this submission. For example, some information is missing or unclear, e.g., lines 245-247, line 399, as well as others. Therefore, we highly encourage a thorough proof-reading prior to resubmission.

Response:

Thank you for the comments.

Line 245-247 was revised to include the percentage of the power of study and its description as follows:

“Handwritten feature with low powder of study (<70%) is suggested for further investigation with greater sample size, in this case the uppercase "A" with three individual strokes. For the current study, only those statistically significant features with adequate power (in this case greater than 78%) are discussed.” (Line 241-244).

Line 399 was also revised to “Previous studies had suggested foreign influence toward the formation of handwriting, particularly in countries that have welcomed a huge number of immigrants in the past.” (Line 391-392).

Comment:

3) Please also reconsider a rephrasing of your interchangeable use of 'race' and 'nationality' (some guidelines for this may be found here - https://jamanetwork.com/journals/jama/fullarticle/2783090).

Response:

Thank you for the comment. In this study, the subjects were from the three main ethnicity groups, and they are from the same country (same nationality). In the revised manuscript, the term “ethnicity” was used in the results and discussion.

Reviewer #4

Comment:

The conclusion can include more application areas for the study. The paper is well structured. Research methodology is adequately developed.

Response:

Thank you for the comment. The conclusion was added with “Nonetheless, the examination of questioned handwriting can still be expanded to retrieve useful information thus narrowing down the possible groups of potential writers, in this case, providing information on the educational backgrounds.” to support the application for the study.

This manuscript is checked thoroughly. We really appreciate the efforts taken by the editor and reviewers to help us improve the manuscript, and we hope that we have addressed all your concerns in this manuscript.

Thank you.

Ahmad Fahmi Lim Abdullah, PhD

School of Health Sciences, Universiti Sains Malaysia, 16150 Kubang Kerian, Kelantan, Malaysia. Email: fahmilim@usm.my; Tel: +609-7677596; Fax: +609-7677515.

Submitted filename: Responses to Reviewers.docx

Click here for additional data file.

9 May 2022

An exploratory study on the handwritten allographic features of multi-ethnic population with different educational backgrounds

PONE-D-20-35443R3

Dear Dr. Abdullah,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Avanti Dey, PhD

Staff Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

8 Jun 2022

PONE-D-20-35443R3

An exploratory study on the handwritten allographic features of multi-ethnic population with different educational backgrounds

Dear Dr. Abdullah:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Avanti Dey

Staff Editor

PLOS ONE