A total of 90 adults aged ≥50 years old participated in this study. All CI participants were diagnosed with a degenerative disease, of whom 26 were diagnosed with AD (male:female = 10:16) and 31 were diagnosed with aMCI (male:female = 8:23), based on neurological examination and medical history at the general hospital. Additionally, 33 adults with normal cognitive function (HC) (male:female = 9:24) were included. Aβ deposition was diagnosed by visual evaluation by a nuclear medicine physician.

The selection criteria for each group are as follows:

The AD group included (1) those whose Aβ denatured protein accumulation was confirmed as positive (+) on PET, (2) those whose standard score was less than 16th percentile based on age and years of education as a result of Seoul Neuropsychological Screening Battery-2, SNSB-II (Kang et al., 2012), (3) Clinical Dementia Rating scale (CDR) of mild dementia, (4) those with no visual, hearing, or hand-motor impairments that affected task performance, (5) right-handed individuals, and (6) adults with educational attainment of primary school or higher.

The aMCI group included (1) those whose Aβ denatured protein accumulation was confirmed as positive (+) on PET, (2) those whose SNSB-II (Kang et al., 2012) result, delayed recall score of the Seoul Verbal Learning Test, or delayed recall score of the Rey Figure Test was less than 16th percentile, (3) those with familial history of cognitive impairment, (4) those with no visual, hearing, or hand-motor impairments that affected task performance, (5) right-handed individuals, and (6) adults with educational attainment of primary school or higher.

The HC group included (1) those whose Aβ denatured protein accumulation was confirmed as negative (−) on PET, (2) those whose SNSB-II (Kang et al., 2012) results, age, and years of education are equivalent to 16th percentile or higher in all areas of cognitive function, (3) those with no history of mental or neurological disease after conducting a health screening questionnaire (Christensen et al., 1991), (4) those who scored <8 points on the Short form Geriatric Depression Scale (SGDS) and had a normal level of depression, (5) those with no visual, hearing, or hand-motor impairments that affected task performance, (6) right-handed individuals, and (7) adults with educational attainment of primary school or higher.

The Korean-Mini Mental State Examination (K-MMSE) (Kang, 2006) and the Korean Version of the Montreal Cognitive Assessment (MoCA-K) (Lee et al., 2008) were used to check the overall cognitive level of the participants before the test. There were no significant differences among the three groups in terms of years of education [F_(2,87) = 1.865, p = 0.161], age [F_(2,87) = 0.284, p = 0.753] and SGDS scores [F_(2,87) = 2.415, p = 0.095] (Table 1). A total of 67 participants (21 in the HC group, 25 in the aMCI group, and 21 in the AD group) were tested for grip strength, which is thought to be related to writing tasks, and the results showed no difference among the three groups [F_(2,64) = 0.194, p = 0.824].

Data collection for this study was conducted on patients and healthy individuals who visited the neurology department of a general hospital in Seoul. Before conducting the writing test, the cognitive function evaluation (SNSB-II) results of all participants who had consented to participate in the study were checked on the electronic medical record form. The collected data included the results of the cognitive function evaluation conducted within the last year, a health screening questionnaire, and an abbreviated geriatric depression scale (SGDS) conducted as screening tests prior to the main test. To check the cognitive abilities of the participants, the Korean version of the Mini-Mental State Examination (K-MMSE) and the Korean version of the Montreal Cognitive Assessment (MoCA-K) were administered for 10 to 15 min. After the screening test, the participants were given a short break and afterward the main test was conducted. The main test was a word dictation task and was conducted in the order of regular word, irregular word, and non-words. Cognitive ability tests and dictation tasks require quiet environments with ample space because these tasks require listening to the stimuli presented in the auditory sense and answering or dictating them in spoken words. Therefore, the test was conducted in a conference room located in the hospital at a 1:1 participant-to-facilitator ratio. The succeeding test was the grip strength measurement and the additional copying task was conducted on select participants. The total examination time, including the screening test and additional tasks, took approximately 40 min. All procedures in this study were conducted after receiving prior approval (No. 2021-12-068) from the Institutional Review Board of the general hospital.

All writing tasks were performed by the participants on a tablet laptop screen while sitting on a chair with the correct posture and holding the stylus as if holding a pencil. Since the stimulus for the word dictation task can be affected by the volume or speaking speed of the examiner, only one examiner conducted the study. All words were presented with a volume of about 55–60 dB SPL, and the presentation speed was presented in spoken language by checking the average speed (about 510 ms) calculated in syllable units after the examiner called each word once through the speech analysis program Praat. Before starting the dictation task, the examiner asked the participants to listen to the words presented auditorily and speak along to check whether the words were recognized. However, since one-syllable irregular words can be misinterpreted as meaningless non-words if presented without context, meaning clues were presented together but limited to one-syllable irregular words [e.g., (/gamasot/means “iron pot”) so (/sot/) should be written]. In the case of non-words, low familiarity and difficulty in accessing the meaning can lead to questioning of the task performance. Additionally, auditory misperceptions may occur during the task, such as hearing and writing down the phonemes of word stimuli as similar sounds [e.g., (/konna/) as (/honna/)]. To prevent this, the examiner provided a disclaimer that stated, “I will now play some strange words that do not exist in Korean. Please repeat what you hear, and if it is correct, we will proceed with the dictation task in order.” In the case of non-words, auditory misperceptions may also occur for high-frequency consonants such as, //, // and, //, as well as diphthongs. When meaning clues are provided [e.g., When writing (/nuhal/) as (/nuhwal/)], correct responses may be influenced by the utilization of the semantic system. To exclude this, rather, meaning clues on the syllable that showed the wrong response was provided and clarified the target syllable was not the same to the wrong syllable. For the copying task, the instruction given was, “Please copy the characters presented on the screen in the space below.”

The posture for measuring grip strength was based on previous studies (Guerra et al., 2017), in which the participants sat on a chair with their arms fixed at a 90° angle to the side of their body. The handle was set at Level II, regardless of hand size. The measurement order began with two measurements of the left hand’s grip strength, recording the higher value. Then, two measurements of the right hand’s grip strength were taken, recording the higher value. The participants were instructed, “We will measure the grip strength twice on each side, left and right. Please hold the grip strength meter with your hand and place your elbow at the side of your body in contact.”

The word dictation task was analyzed in terms of linguistic, visuospatial, and kinematic aspects. In the linguistic aspect, a correct response was scored 1 point when all graphemes that make up the word were spelled accurately. If a linguistic error was made in any of the graphemes, 0 points were given. A total of 60 points were assigned, with 20 points each for regular words, irregular words, and non-words.

The types of linguistic errors were analyzed as follows:

Visuospatial aspects were analyzed based on the shape of letters or the combination of grapheme. Visuospatial errors refer to characters that cannot be seen in existing Korean alphabets and syllables, or awkward combinations. Therefore, all words were divided into “stroke level” and “grapheme level” and their frequency and type were analyzed. At the stroke level, it was divided into addition and omission errors. Stroke addition and stroke omission were regarded as visuospatial errors only when they were not replaced by another grapheme of Hangeul. At the grapheme level, grapheme combination errors were analyzed. A combination error of grapheme means an error in which each grapheme constituting one syllable maintains the form of a Korean grapheme, but the combination is incorrect and is combined in a form that does not conform to the rules of Hangeul syllable (Supplementary Table 2).

In terms of kinematics, pen pressure writing speed was checked. Pen pressure meant the pressure applied when writing a letter on a tablet notebook, and was measured as the average value of changing pen pressure from the point the pen touched the screen to the point the pen was removed to write a single stroke. The range of pen pressure was measured within the range of 0–4,096 levels, which is the range that can be measured on a tablet notebook, and the measured value was automatically converted into the range of 0.00001–1 levels in the software and used. The pen pressure of “word dictation” was measured as the average value of the pen pressure values measured for each stroke while writing the target word. Writing speed was measured by taking into account the length of each stroke (length) and the time taken to write it (duration), which are automatically measured by the software. The speed (mm/ms) was calculated by dividing the length of each stroke (measured in mm from the point the pen touched the tablet notebook screen to the point it was lifted) by the time taken to write it (measured in ms), for each stroke required to write a single character (Lee et al., 2020). The speed of “word dictation” was measured by the speed value obtained by adding the total time required while writing the target word and the total stroke length.

To find out the interaction (to see whether the performance aspects or patterns among groups differed depending on the word type), repeated one-way analysis of variance (ANOVA) was performed using the SPSS (version 26) program. ANOVA and Bonferroni post-hoc tests were performed to examine the differences among groups. A paired t-test or repeated one-way ANOVA was performed for within-group differences. To control for type I error across multiple testing, the significance probability was corrected with Bonferroni correction.

In this study, we identified linguistic, visuospatial, and kinematic aspects through a word dictation task that could examine the decline in cognitive-linguistic-motor function in MCI and AD with Aβ deposition with high sensitivity.

First, in a word dictation task, we observed that the AD group showed lower linguistic performance compared to the HC and aMCI groups in regular words. Regular words were defined as words with lexical properties and regularity aspects, where the sound and letters correspond to each other in terms of grapheme-phoneme correspondence. Unlike irregular words and non-words, both lexical and phonological routes can be accessed, making it a relatively easy word condition (Lee et al., 2020). Therefore, it can be interpreted that while the performance is maintained until the aMCI stage, a decline in performance is observed from the AD stage owing to the difficulties in accessing not only the lexical route but also the phonological route (Sung, 2010). Irregular words are words that carry meaning like regular words but have a grapheme-phoneme inconsistency where the sound and letters do not correspond. Therefore, to perform irregular words, one must rely solely on the lexical route rather than the phonological route (Penniello et al., 1995; Sung, 2010; Yeon et al., 2018). In irregular words, we observed a staircase-like decline in performance as cognitive abilities gradually decreased from the aMCI stage to the AD stage. Here, we added a control for individuals who show cognitive impairment and decreased abilities in daily life, as well as positivity for Aβ deposition. Therefore, the results of this study suggest that damage to the lexical route can be caused by the aMCI stage with Aβ deposition, and it can be interpreted that it becomes more difficult to utilize the lexical route as one enters the dementia stage (Roeltgen, 2003). Non-words do not have a familiar meaning like regular or irregular words, and they have the characteristic of matching graphemes to phonemes, so they require activation of the phonological route rather than access to the semantic system (Yeon et al., 2018). In non-words, we observed that performance was lower in the aMCI and AD groups compared to the HC group. Traditionally, the processing of the phonological route is associated with the temporo-parietal lobe, Wernicke’s area, left fronto-temporal, Broca’s area in the left hemisphere, and perisylvian region (Rapcsak et al., 2009). Therefore, individuals with Aβ deposition and cognitive decline starting from the aMCI group may have difficulty utilizing the phonological route, and in the early stages of AD, writing processing via the phonological route is maintained at the level of the aMCI group.

Second, when examining the error types in the word dictation task, substitution errors gradually increased as the severity of cognitive impairment progressed from aMCI to AD, reflecting the overall decline in cognitive function. Addition errors were significantly more frequent in both aMCI and AD groups with cognitive impairment compared to the HC group, while omission errors were significantly more frequent only in the AD group. Specifically, substitution errors accounted for the highest proportion in all groups, and these results may reflect the unique characteristics of Hangeul. As Hangeul has a form of group writing in which each grapheme is gathered and combined, and the location of the initial, middle, and final consonant within a syllable is fixed, it has more visuospatial characteristics than the English alphabet arranged in a row (Yoon et al., 2010, 2012). Therefore, even if one grapheme is missing from a syllable, it is easier to recognize that the shape of the syllable has changed compared to the alphabet. Specifically, the patients may have recognized the position of the letter in the target word, but may not have known exactly what the letter was, and thus attempted to replace it with a letter they knew in order to maintain the shape of the target word, leading to more substitution errors compared to other error types. In this context, omission errors, which fail to maintain the shape of the syllable, may have been observed significantly more often only in the AD group, which had the most impaired cognitive function among the three groups.

Third, in irregular words, PPE errors increased from the aMCI stage, while PIE errors significantly increased as AD progressed. These results suggest that in the aMCI group, individuals utilized residual phonological pathways to perform writing tasks, whereas in the AD stage, difficulties with phonological pathways were exacerbated, leading to errors unrelated to the phonology of the target word. In non-words, lexicalization errors increased from the aMCI stage. This may be interpreted as a pattern in which incomplete lexical pathways lead to the production of meaning words that remain within the internal lexicon during the processing of non-words.

Fourth, as a result of comparing the frequency of correction between groups during word dictation, the aMCI group showed significantly more self-correction than the HC and AD groups (Supplementary Table 5). Considering the proportion of participants who made corrections, the NC group had 9 out of 33 participants (27.28%), the AD group had 10 out of 26 participants (38.46%), while the aMCI group had 25 out of 31 participants (80.65%), indicating a more than two-fold higher rate than the other two groups. This can be explained in relation to the attitudes and meta cognition of the participants shown in the word dictation test. In contrast to the HC and AD groups, the aMCI group was observed to constantly confirm the accuracy of their responses with the examiner during the task. This implies that the aMCI group is capable of monitoring their performance and asking for feedback, which is one of the meta cognition abilities related to executive control. Self-monitoring, a cognitive strategy related to higher cognitive abilities, involves systematically observing and recording one’s own behavior (Shapiro and Cole, 1994). In writing, it can be seen as checking whether one understands the target word they are writing. Therefore, self-correction during word dictation could reflect the use of goal recognition and cognitive strategies to resolve errors (Shapiro and Cole, 1994).

Fifth, visuospatial errors occurred significantly more frequent in the AD group. Damage to neurons in the brain areas including the temporal-parietal areas, which are responsible for visuospatial organization functions, can result in damage to the pathways responsible for these functions in the AD group (Zillmer and Spiers, 2001). Especially, the dorsal pathway located in the posterior parietal areas is associated with determining where the target is located and integrating information on spatial perception such as spatial configuration between targets and self-motion caused by the movement of the body (McFie et al., 1950; Trick and Silverman, 1991; Gilmore et al., 1994; Johnson et al., 2002; Kavcic et al., 2011). When this pathway is damaged, individuals may not be able to accurately depict an object and may transform the central axis or draw the object inaccurately or at a different location with intersecting lines. Additionally, they may rotate lines or parts of the object while drawing (McFie et al., 1950). A study analyzing the handwriting of individuals with CI reported that there was a possibility of changes in letter shapes (Cilia et al., 2021a). Therefore, the AD group’s ability to perceive and analyze spatial information may be impaired, resulting in the addition or omission of strokes during character processing. An error in the combination of grapheme means that a grapheme that exists in Hangeul is used, but the combination between grapheme is wrong and is expressed in the form of a syllable that does not exist in Hangeul. Grapheme combination errors were the most frequent in words composed of diphthongs (e.g., regular word “,” non-words “”). In terms of phonetics, vowels can be more subtle than vowels of other sounds in hearing and distinguishing sounds compared to consonants (e.g., “,” “”). In particular, there were a lot of phoneme combination errors in singer vowels “” and “” (“” single vowel “” was changed to “” and “” single vowel “” was changed to “”). In a study (Byun, 2020) that confirmed the perceptual characteristics of // and // in Korean with normal young adults in Korea, the difference between // and // vowels is not large in terms of oral opening and they are close, so it is difficult to distinguish them after hearing them. Therefore, a combination error of grapheme that was combined in a form that does not possibly conform to Korean rules was observed.

Sixth, as a result of comparing motor performance, the speed of the CI group (aMCI group and AD group) was slower than that of the HC group. Writing speed is influenced by various factors involved in performing movements, as well as cognitive processes involved in analyzing stimuli and organizing and planning responses during word dictation tasks (Van Galen, 1991; Jung, 2015). In tasks of writing a sentence dictation and repeated strokes (Delazer et al., 2021), mild AD patients exhibited slower writing speed and less automated movements compared to the HC group. They suggested that this could be caused by the deficits on not only central levels of processing but also the motor execution. In contrast, there was no significant difference in pressure between the groups, which is consistent with the previous study (Delazer et al., 2021). The pressure applied to the screen may be related to the basic strength of the hand. As a result, when grip strength was measured, no significant difference was observed between the groups, consistent with previous studies (Kluger et al., 1997). In the CI group, movement disorders may not be prominent in the early stages of the disease, and participants who participated in this study did not show major kinematic problems such as tremors or rigidity of body parts that could affect pen pressure. It can be interpreted that there was no difference between the groups.

Additionally, when comparing the results of motor performance according to syllable length (1 syllable, 2 syllables, 3 syllables, and 4 syllables), the motor performance according to syllable length differed according to the groups. In terms of pen pressure, a trend was observed in both the HC and AD groups that pen pressure increased as syllable length increased. The tablet screen used in this study has a lower coefficient of friction compared to paper, which can give the impression of slipping when writing (Gerth et al., 2016; Guilbert et al., 2019). Therefore, to compensate for the greater decrease in frictional force as the syllable length increases, the tension of the hand’s forearm muscles, shoulder, elbow, and muscles may have increased while performing writing tasks, leading to an increase in pen pressure (Bara and Gentaz, 2011; Kim et al., 2017). However, in the aMCI group, no significant difference in pen pressure was observed according to syllable length. This may be because frequent self-corrections during dictation tasks could cause interruptions in performance and may not reflect efforts to compensate for the decrease in frictional force. Therefore, it is possible that the absence of differences in pen pressure according to syllable length in the aMCI group, compared to the HC and AD groups, who wrote characters without frequent corrections, is due to the frequent corrections made by the aMCI group during writing tasks.

Notably, in terms of writing speed, the aMCI group was observed to have faster writing speed as syllable length increased. This can be attributed to the characteristics of the writing to dictation task. Unlike reading, in which visual cues are continuously provided, the dictation conducted in this study was a task in which the examiner listens to and writes down the contents. Accordingly, the dictation task, in which the audibly presented letters are temporarily stored and calculated by hand through the planning and programming stages of writing, required more working memory ability than reading (Kim et al., 2017). Since working memory is impaired in aMCI (Jang and Kim, 2014), the aMCI group may have compensated for their reduced working memory capacity by increasing the speed of hand movements to write longer syllables during the process of processing auditory presented words. However, in the case of the AD group, as the cognitive ability deteriorated further, the compensatory action for this did not occur properly; thus, it is assumed that the pattern of increasing the speed as the syllable lengthened was not observed. To confirm this assumption, we conducted an additional test in which the participants were asked to copy the same words as the ones used during dictation. Copying requires less working memory ability than dictation because the task allows the participants to continuously visualize the word while imitating it. Therefore, rather than confirming spelling knowledge, it can be said that it is a task that reflects more the visual compositional ability and motor ability required when performing writing tasks. Consequently, there were no changes in speed of the copying task even if the syllable length of words increased among all groups, including the aMCI group (Supplementary Table 4). The results of the copying task support the hypothesis that the demands on working memory in the spelling task may have affected writing speed with longer syllables.

In future studies, it is necessary to analyze the neuroimaging and neuroanatomical correlations related to writing performance in the cognitively impaired group and healthy individuals. Moreover, the additional task was conducted among 20% of the total number of participants; thus, it is necessary to supplement the results through a full survey based on the total number of participants. While this study checked the grip strength as a basic motor ability of the hand related to pen pressure, it is necessary to also check basic motor abilities of the hand in relation to speed. Lastly, future improvement is needed as our data also reveal a gender bias, as reported that females have a higher prevalence of AD-type dementia compared to males (Gao et al., 1998; Kim and Han, 2012).

In summary, the results of this study demonstrate that compared to those in the HC, patients in the aMCI group exhibited challenges in utilizing the lexical and phonological routes at an early stage, leading to a decline in performance with irregular words and non-words. As the disease progresses to the early stages of AD, the degree of impairment in phonological route utilization remained at the aMCI level, while the utilization of the lexical route became even more difficult. In terms of visuospatial aspects, patients with AD showed difficulties in adding or omitting strokes and combining graphemes. In terms of kinematic ability, both aMCI and AD groups showed a slower writing speed. This study is noteworthy and meaningful because it provided comprehensive results related to the writing performance of CI groups as “the First digital-based writing analysis study” that objectively and multidimensionally analyzed data collected using digitized software for patients with Aβ deposition. Our findings can be utilized as one of the cognitive-behavioral symptoms that can discriminate cognitive impairment caused by Aβ deposition. Furthermore, this study can serve as a clinical foundation for comparing and analyzing language, visuospatial, and motor abilities in other disease groups that show similar patterns of cognitive impairment as the CI group in the future.