Cerebral Lateralization During Handwritten and Typed Word Generation: A Functional Transcranial Doppler Ultrasound Study in Left‐Handers and Right‐Handers

ABSTRACT

Cerebral lateralization of written language, much like oral language, is predominantly left lateralized. However, handwriting has been the primary focus in lateralization studies. The cerebral lateralization of typing—a widely used method of writing—remains unexamined. This preregistered study aimed to explore the cerebral lateralization of typing versus handwriting and to investigate possible handedness‐related differences. We hypothesized that (i) cerebral lateralization would not differ between the two writing methods after movement correction and (ii) both handwriting and typing would show weaker lateralization in left‐handers compared to right‐handers. To investigate this, we used functional transcranial Doppler (fTCD) ultrasound, a reliable method for assessing cerebral lateralization during language tasks that remains unaffected by movement artifacts, such as those generated by handwriting and typing. A total of 24 left‐handers and 30 right‐handers participated, performing written word generation through both handwriting and typing on a computer keyboard while undergoing fTCD assessment. We applied a Bayesian framework for our analysis, as it enables us to demonstrate the absence of a difference (i.e., no difference between two variables), which is not possible with the use of p values (estimated under the frequentist framework). Our results provided evidence supporting the absence of a difference in cerebral lateralization between handwriting and typing after movement correction. However, we found no conclusive evidence to either support or refute a difference in lateralization between left‐handers and right‐handers, suggesting that more research is needed to clarify the role of handedness in cerebral lateralization for different writing methods.

We investigated the cerebral lateralization during handwriting and typing in a computer keyboard, for both left‐handers and right‐handers. We used functional transcranial Doppler ultrasound to measure the blood flow toward the left and the right brain hemispheres. Greater blood flow velocity corresponds to greater activation for that hemisphere. We used a Bayesian approach in our analyses to detect possible evidence of an absence of difference.

List of Abbreviations

BF

Bayes factor

CNTR

control

COR

corrected

EHI

Edinburgh Handedness Inventory

fMRI

functional magnetic resonance imaging

fTCD

functional transcranial Doppler ultrasonography

H

handwriting

LH

left hand

LI

lateralization index

MCA

middle cerebral arteries

PC

personal computer

POI

period of interest

QHPT

quantification of hand preference test

RH

right hand

SBF

sequential Bayesian factor

SD

standard deviation

T

typing

WGT

word generation task

WPM

words per minute

1. Introduction

The two most prevalent forms of language are oral and written. Oral language is speculated to have developed along with the emergence of our species as early as 250,000 years ago (Mcbrearty and Brooks 2000). Written language is estimated to have evolved much more recently, with evidence of its existence in Ancient Greece (Yunis 2003), Mesopotamia (Bottéro 1995), Ancient China (Wang 2014), and Ancient America (Rodríguez Martínez et al. 2006). Interacting with written text became more prominent with the invention of the printing press in 1450 ad (Rees 2005) and, closer to our time, with the invention of the personal computer (PC; Blankenbaker 1973). It was, however, the widespread adoption of the internet that established written language as a mainstream form of communication (Crystal 2006). A report from 2015 showed that in 16 out of the 19 countries investigated, at least half of the population used the internet, with at least 85% of the population doing so using a computer (USC Annenberg School 2016). Given the fact that writing—and, even more so, keyboard typing—is a very novel invention in our evolutionary history, studying writing would allow us to investigate the neural underpinnings of skills that are not biologically hardwired (Purcell et al. 2011). Typing also deserves investigation, as it is a more efficient method of writing. For instance, Mueller and Oppenheimer (2014) observed that using a keyboard to transcribe lecture notes resulted in increased transcription fluency (i.e., the ability to transcribe thoughts as fast as they come to mind) compared to handwriting. However, the importance of writing and the widespread use of PCs are not reflected in the literature on the neural underpinnings of language in general and on the cerebral lateralization of language, in particular . Rather, the majority of studies on the cerebral lateralization of language focus on oral language (e.g., Bishop, Woodhead, and Rutherford 2020), and within those investigating the cerebral lateralization of written language, only one used typing (Krueger et al. 2020), which is the focus of the present preregistered study.

Oral language is left‐lateralized (meaning that there is a relative dominance in the activity of the left hemisphere compared to the activity of the right hemisphere during oral language production), with lesion studies showing this as early as 1866 (Berker, Berker, and Smith 1986). For example, people who experience strokes in the left hemisphere are more likely to lose their ability to speak compared to people with strokes in the right hemisphere (Pedersen, Vinter, and Olsen 2004). More evidence comes from patients undergoing diagnostic tests for the treatment of epilepsy who are typically found to exhibit predominantly left‐hemispheric lateralization of language processes (Benke et al. 2006; Loring et al. 1990; Rasmussen and Milner 1977). Studies in healthy individuals have further established that the majority of individuals show left cerebral lateralization of oral language (Khedr et al. 2002; Loring et al. 1990; Petit, Badcock, and Woolgar 2020).

A similar left‐hemispheric lateralization has been observed for written language (Menon and Desmond 2001; Planton et al. 2013). Lesions in the left hemisphere give rise to apraxic agraphia, a condition that results in an inability to write (Beeson et al. 2003; Roeltgen 1993). Unlike oral language, however, the investigation of the cerebral lateralization of written language in the intact brain presents the challenge of distinguishing between the activation related to linguistic and motor components. Studies have therefore tried to isolate the linguistic component of writing by using motor control conditions. Areas in the left frontal lobe have been shown to have a linguistic‐specific activation (Planton et al. 2013; Tam et al. 2011), including Broca's area, which has been shown to increase activation during writing compared to drawing (Potgieser, van der Hoorn, and de Jong 2015). Additionally, areas in the left temporal lobe increase activation during writing compared to drawing (Planton et al. 2013; Potgieser, van der Hoorn, and de Jong 2015; Tam et al. 2011). Left‐hemispheric areas are engaged in the motor component of writing similarly to its linguistic component. The left parietal lobule is involved in the sequential execution of writing (Karimpoor et al. 2018), and, along with areas of the left sensory motor cortex, the left parietal lobule promotes the interaction of areas subserving language comprehension with motor regions during writing tasks (Karimpoor et al. 2018; Segal and Petrides 2012). Areas in the left somatosensory and premotor cortex have also been associated with the execution of writing (Karimpoor et al. 2018; Menon and Desmond 2001) and the acquisition of handwriting (Palmis et al. 2017).

The above‐summarized literature on the cerebral lateralization of writing is limited by the fact that findings have been derived from studies using functional magnetic resonance imaging (fMRI). Despite fMRI being an efficient way to localize specific areas of the brain, this method can be challenging when studying writing as the participant has to lie on their back and write in an unnatural way (natural being sitting on a chair). Moreover, typically, fMRI studies include small sample sizes due to the cost of the technique and often report mean group effects, which are calculated using variable statistical significance thresholds that are fixed at a group level (Bradshaw, Bishop, and Woodhead 2017; Johnstone, Karlsson, and Carey 2021; Wilke and Lidzba 2007). An alternative to fMRI is functional transcranial Doppler ultrasound (fTCD).

FTCD does not have the spatial resolution that would allow for the localization of areas involved in writing, but, when it comes to investigating cerebral laterality, its results are comparable to fMRI (Chilosi et al. 2019; Deppe et al. 2000; Somers et al. 2011). More specifically, cerebral lateralization results were similar between fTCD and fMRI in similar word generation tasks of healthy participants (r = 0.95, p < 0.0001; Deppe et al. 2000), with another study reporting a high correlation (r = 0.75, p < 0.001) in the lateralization indices derived from fTCD and fMRI in identical word generation tasks (Somers et al. 2011). Another recent study investigated 18 participants with damage in their left hemisphere and 18 controls and showed that the categorization of the participants as having left, right, and bilateral lateralization was highly correlated (r = 0.70, p < 0.001) between fTCD and fMRI during two language tasks (Chilosi et al. 2019). Similarly to fMRI, fTCD is noninvasive but is considerably less expensive, allowing for larger sample sizes (Badcock and Groen 2017) and easier follow‐up assessments (Lohmann et al. 2005). Finally, an advantage of fTCD over fMRI for investigating cerebral lateralization during writing is that participants can perform the tasks while sitting on a chair, which is natural for writing, instead of lying down in a scanner.

Another limitation of fMRI studies on the cerebral lateralization of writing is the systematic exclusion of left‐handers, with only one fMRI study to date sampling left‐handers (Zaman, Wartolowst, and Roberts 2002; a one‐page brief report comparing normal and mirror writing with dominant and non‐dominant hands). Left‐handers constitute about 10% of the general population (Papadatou‐Pastou et al. 2020), a proportion that is too great to be ignored (Bailey, McMillan, and Newman 2020; Willems et al. 2014). The difference between left‐handers and right‐handers extends further from mere hand preference to brain function with around 90% of right‐handers, but only 70% of left‐handers, being left lateralized for language production tasks (Carey and Johnstone 2014) and with left‐handers exhibiting weaker left lateralization as a group and on the individual level even if they are left‐lateralized (Johnstone, Karlsson, and Carey 2021). This difference has even led to the suggestion to treat left‐handers as comprising two distinct groups: a group with typical and a group with atypical cerebral lateralization (Johnstone, Karlsson, and Carey 2021). It is important to note that the above‐mentioned findings are based on oral‐language tasks. Written language requires hand movement, which results in left‐hemispheric activation for right‐handers and right‐hemispheric activation for left‐handers. Given the effects of handedness on the cerebral lateralization of oral language and taking into account hand usage during writing, investigating handedness differences in cerebral lateralization in the context of the written language is important.

FTCD works by having ultrasound probes placed over the temporal windows on both sides of the head and measuring the blood flow velocity of the middle cerebral arteries (MCAs). Targeting the MCA is appropriate given their role in supplying blood to the brain areas involved in language production described above (Gibo, Carver, Rhoton, Lenkey, and Mitchell 1981), with a recent study showing that the areas that the MCA supply are even wider than previously thought (Kim et al. 2019). More specifically, MCAs are supplying blood to brain areas associated with writing including the left superior frontal sulcus, left intraparietal sulcus (Planton et al. 2013), and the graphemic motor area (also “Exner's” area; Planton et al., 2019). By comparing the blood flow velocity in the left and the right MCA, we can derive a lateralization index (LI), which provides information about which hemisphere is more active during a cognitive task.

Two recent fTCD studies focused on the cerebral laterality of written language and included left‐handers (Kondyli et al. 2017; Papadatou‐Pastou et al. 2022). Kondyli et al. (2017) studied 31 left‐handers and 29 right‐handers, who performed an oral and a written word generation task (i.e., participants were asked to produce as many words as possible starting from a cue letter in a set amount of time). Left‐handers showed a greater decrease compared to right‐handers in left‐hemispheric activation during writing compared to silent word generation. This was attributed to writing potentially activating a more widespread language network, including right‐hemispheric areas, in left‐handers compared to right‐handers. However, the result was alternatively explained by the activation of the right hemisphere by left‐hand motor action during writing. To address this issue, Papadatou‐Pastou et al. (2022) administered the written word generation task in 23 left‐handers and 31 right‐handers, but in this study, they isolated the motor from the linguistic component of writing by using symbol copying as the motor control task (repeatedly copying a cued symbol). Right‐handers had significantly more left‐hemispheric activation during written word generation compared to symbol copying, a finding that failed to replicate in left‐handers. This could be consistent with the hypothesis of a broader language production network or with a greater variability in cerebral laterality patterns in left‐handers.

As far as the means of writing is concerned, available data rely heavily on handwriting. Only one study has investigated the cerebral lateralization of typing (Krueger et al. 2020). This study of 30 right‐handers examined the neural correlates (using fMRI) of typing prose (e.g., filling up the blank part of a sentence or a longer response, which resembles a function of natural language) versus typing code (e.g., filling in a blank in a program with a code snippet, which, other than the use of letters, holds little resemblance to natural language) and showed that prose typing exhibited left lateralization (Krueger et al. 2020), consistent with handwritten language production (Papadatou‐Pastou et al. 2022). However, this study did not sample left‐handers, and participants completed the tasks lying down. Moreover, the two tasks did not allow for the isolation of the linguistic component of typing, as prose typing was contrasted with typing code and not a motor control task.

In the present study, we investigated the cerebral laterality of writing using two different means of writing: handwriting and typing. We also investigated possible handedness differences in the cerebral laterality of these two writing methods. We built our hypotheses on the assumption that there is one language production network and that this network is activated regardless of whether language is produced by handwriting or by typing. We expected the laterality of the motor component of writing to be different between left‐handers and right‐handers (as they activate their contralateral motor cortices), as well as between handwriting (employing one hand) and typing (employing two hands), but we only investigated the laterality of the motor component in an exploratory way, because it was outside the scope of the preregistered part of this study. We formed the following hypotheses:

Hypothesis H1

There will be an absence of difference in the cerebral laterality of the linguistic component of writing as assessed during handwriting versus typing.

Hypothesis H2

Right‐handers will exhibit stronger left‐hemispheric lateralization compared to left‐handers for the linguistic part of (i) handwriting and (ii) typing, as is the case with oral language.

2. Materials and Methods

2.1. Sampling Plan

We preregistered to initially recruit 32 healthy adult volunteers (16 left‐handers and 16 right‐handers according to their Edinburgh Handedness Inventory (EHI; Oldfield 1971) score) as a starting point for evaluating the evidence, a sample size which also ensures adequate counterbalancing (for the four versions of the experiment). We aimed for a BF10 that exceeds 6 (or a BF10 that is smaller than ⅙), corresponding to “moderate evidence” (Lee and Wagenmakers 2014), for all hypotheses (H1 and H2i, ii). Bayes factors (BF10) quantify the strength of evidence for one hypothesis over another. Here, values greater than 6 indicate moderate evidence for the alternative hypothesis, and values smaller than ⅙ indicate moderate evidence for the null hypothesis. For reference, BF10 values greater than 3 (or smaller than ⅓) suggest weak evidence for the alternative (or null) hypothesis, and values greater than 2 (or smaller than ½) indicate anecdotal evidence. However, the literature recommends treating BF10 values between ½ and 2 as uninformative, as they do not provide strong evidence in either direction. Using a sequential Bayesian factor (SBF) with maximal n design (Schönbrodt et al. 2017), two extra participants were added in each handedness group until our aim was reached or until an upper limit of 48 participants in total (24 participants per group) was reached, given time and resources constraints. This upper limit is also in line with Schönbrodt et al. (2017), according to which, 24 participants per group for the SBF Design is an average sample size without wrong inference related to false‐negative evidence, where true effects may not be detected due to insufficient power. Participants were recruited through advertisement in social media and in the lab and project websites. Moreover, university students were recruited in return for course credit. No monetary incentives were provided. Recruitment was conducted using the following criteria:

2.1.1. Inclusion Criteria

1. Participants had to be self‐reported healthy adults, monolingual, and native speakers of the Greek language (we excluded participants that were systematically exposed to another language before the age of 6 years), and they had normal or corrected‐to‐normal vision.

2. Participants had to be able to type on a computer keyboard (with the standard Greek letter arrangement that matches the English QWERTY keyboard) using both of their hands at the same time. This was assessed using a word composition task on a computer keyboard where we assessed their words per minute (WPM) typing speed. Participants were asked to type any words that came to mind as fast and accurately as they could in a blank Notepad document within a 60‐s time window. The number of words correctly composed (all the letters and their order need to be correct, although we did not take into account the stresses) was then calculated, and this was the participants' WPM score. The estimated average for this word composition task is 19 WPM for the general population (Karat et al. 1999; see Badcock et al. 2018), which translates roughly in five words per 15 s. As we will describe later in Section 2, the participants in our experiment had 15 s to compose as many words as they could starting from a cue letter. Thus, for the inclusion criteria, we set a boundary of 12 WPM, which translates to three words per 15 s.

3. Participants had to have an EHI score of 0%–40% or 60%–100%.

2.1.2. Exclusion Criteria (by Self‐Report)

1. Prior diagnosis of dyslexia or dysgraphia. Presence of neurological problems or other problems affecting the mobility and normal function of the hands.

2. Consumption of medication that could affect the central nervous system in the previous 6 months or current use of illicit drugs or other substance abuse.

2.2. Assessment of Linguistic Lateralization

To calculate the cerebral LIs, we administered modified versions of the Word Generation task. The task was originally  developed by Knecht et al. (1998) and is considered to be the gold‐standard test for assessing cerebral laterality of language using fTCD. The tasks were designed in a way to regress out the motor component of writing so that what is left is the linguistic component. The layout of each trial was as follows (for a schematic representation, see Figure 1):

1. Thirty‐five seconds of rest while a Greek translation of the text “try to clear your mind” was presented on a gray background. A cuing tone of 440 Hz lasting for 0.5 s was presented in the beginning of this period.

2. A cueing tone of 440 Hz lasting for 0.5 s, followed by a 4.5‐s pause. The cueing tone was used to help focus participants' attention on the upcoming task.

3. A capital letter of the Greek alphabet (in arial font with white‐colored letters) was presented in the center of the screen (gray background) for 2.5 s. At this time, a marker was sent from the stimulus PC to the fTCD device for synchronization.

4. A 12.5‐s production period followed for a total of 15 s of production including the presentation of the letter.

FIGURE 1.

Example of an fTCD task trial.

During this 15‐s period, participants were asked to perform one of four tasks (corresponding to four conditions):

1. “Handwriting” (pen‐and‐paper written word production): In the handwriting task, the participants were instructed to write down as many words as possible starting with the cue‐letter using pen and paper.

2. “Handwriting control” (nonlanguage task as a control for the handwriting task): In the handwriting control task, participants were instructed to copy the cue‐letter as many times as possible during the given time period.

3. “Typing” (computer keyboard written word production): In the typing task, the participants were instructed to type as many words as possible starting with the letter appearing on the screen using a computer keyboard; and

4. “Typing control” (nonlanguage task as a control for the typing task): In the typing control task, participants were instructed to press random keys at about the same speed as if they were typing words.

For the writing and typing conditions, participants were instructed to produce words the way that they would naturally. Specifically, they were free to use capital or lowercase letters and to stress the words as they would normally. Misspelling was not taken into consideration, because orthography is uninformative to the aims of the present study.

There were 80 trials in two sets of 40 trials (separated by a break), with 10 trials of each condition per set (20 overall). For practical purposes, the handwriting trials were performed immediately before or after the handwriting control trials, counterbalanced across participants, and the same was done with the typing conditions. Each participant had the order of the tasks randomly assigned. The second set of 40 trials had the same order as the first.

Twenty (of 24 possible) letters from the Greek alphabet were used as cues (each presented four times, once for each task), identified via a pilot procedure described in Kondyli et al. (2017) to allow participants to produce the maximum number of words. The words generated were recorded for each trial.

We calculated an LI for each condition (see below for more details on how LIs were calculated): LI_handwriting for the handwriting condition, LI_handwriting_control for the handwriting control condition, LI_typing for the typing condition, and LI_typing_control for the typing control condition. Two additional LIs were calculated: (1) representing the difference between the handwriting condition and the handwriting control condition, i.e. the linguistic component of writing (LI_handwriting_corrected = LI_handwriting ‐ LI_handwriting_control) and (2) representing the difference between the typing condition and the typing control condition, i.e. the linguistic component of typing (LI_typing_corrected = LI_typing ‐ LI_typing_control).

For the typing conditions, we recorded the number of keystrokes per trial to evaluate whether the motor demands, as quantified by the keystroke count, were comparable between the control and production conditions. This ensured that the control trials closely matched the production trials in terms of movement characteristics, allowing for a valid comparison of conditions. We performed a Bayesian t test using a prior with an r = 0.707 (which has been proposed as an appropriate value for effects with an expected moderate effect size; Morey and Rouder 2011) to assess whether the number of keystrokes during the control condition matches the number of keystrokes during the production condition. We preregistered that if the average number of keystrokes was not different between the two conditions, we would calculate the LI_typing_corrected with all the trials that met the fTCD criteria. If the t test showed a difference in the keystroke number between the two conditions, then we would filter out the control trials for which the average number of keystrokes would be 3 standard deviations away from the average number of keystrokes for the mean word production trials and ran the t test again. If the remaining trials were less than 10 for each given participant, we would exclude the participant.

Participants were excluded at this stage for the following reasons:

1. Inadequate ultrasonographic penetration of the skull by the ultrasound beam, rendering ultrasonography impossible.

2. Noisy data (i.e., not meeting the criterion of 10 accepted epochs per condition).

Two 2‐MHz robotic transducer probes of a commercially available Doppler ultrasonography device (Delica EMS‐9F) were fitted to an agile headset and situated on the left and the right temporal windows of the participant. We performed the ultrasonography at an optimal depth for each participant (45–56 mm) and at angles that provided the optimal signal density in each participant. We used the PsychoPy open software (Peirce 2007) to present visual and auditory stimuli (i.e., the letters and the cueing tone) and to send trigger marks to the Delica system to annotate the preparatory cueing tone. For the presentation of the stimuli, we used an HP 250 G8 laptop that has a 15.6‐in. display and built‐in speakers, set at 100% brightness level and at a 60% loudness level, which is an appropriate level for the audio cues to be audible but not distracting. We then extracted the spectral envelope of the Doppler signal at a frequency of 125 Hz and saved it for offline analysis.

We performed the data analysis using the MATLAB based toolbox DopStep (https://github.com/nicalbee/dopStep; Badcock and Groen 2017; Badcock et al. 2012). The data were trimmed to remove unnecessary data from the start and end of the recording. Each epoch started 18 s prior to the cueing tone and lasted until 36 s after its appearance. We removed the variability caused by heartbeats (Deppe et al. 1997; Knecht et al. 1998; Meyer et al. 2014) using a linear correction (see Badcock et al. 2018). In case of extreme values (beyond −3 to 4 SD from the mean, affecting less than 5% of the data, assumed to reflect minor signal dropout), we used a linear interpolation from 1.5 s either side of the extreme value to correct them. The blood flow velocity of the left and right channels was normalized to a mean of 100 on an epoch‐by‐epoch basis. Epochs containing cerebral blood flow velocity values outside the range of 50%–150% (registered as 70%–130% in Stage 1; see Section 2.7) of the mean or an absolute left‐minus‐right channel difference of 20% multiplied by each individual's interquartile range, affecting more than 1% of the data, were rejected. A baseline correction (−10 to 0 s relative to cueing tone) was performed on an epoch‐by‐epoch basis. The final data were averaged, and the LI was calculated as the average left‐minus‐right channel difference within the period of interest (POI). The POI was 7 to 24 s after the cueing tone because this is when the maximum activation was expected (Papadatou‐Pastou et al. 2022). This POI is preferred over a POI centered around the peak activation because the former produces a distribution that resembles more closely the normal distribution (Woodhead, Rutherford, and Bishop 2020) and is statistically unbiased when calculating LIs derived from fTCD measurements (Petit, Badcock, and Woolgar 2020).

The split‐half reliability was measured for the consistency of mean LI across trials for each task by performing a correlation between odd and even trials (Bishop, Watt, and Papadatou‐Pastou 2009). Analysis proceeded only for those tasks that provide a split‐half reliability coefficient greater than or equal to 0.5 (“moderate reliability”; Koo and Li 2016; Parsons, Kruijt, and Fox 2019).

2.3. Assessment of Handedness

Handedness was estimated by self‐reported hand preference using the Greek version of the EHI (Oldfield 1971). Participants were instructed to indicate which hand they prefer to use for 10 activities: writing, drawing, throwing a ball, using scissors, using a toothbrush, holding a knife to carve meat, holding a spoon, holding a broom (upper hand), striking a match, and opening the lid of a box. We also included two additional activities referring to foot and eye preference: kicking a ball and looking with one eye, which was not used in the registered analysis but was included in our openly shared raw data set. Participants were asked which hand, foot, or eye they used when they performed each of those activities. The possible responses were “always left,” “usually left,” “both equally,” “usually right,” and “always right.” A value of 0 was given to “always left” responses, 1 to “usually left” responses, 2 to “both equally” responses, 3 to “usually right” responses, and a value of 4 to “always right” responses. The total score of each participant was then divided by the maximum score (40) and multiplied by 100. This laterality index (LI) ranged from 0% (extreme left‐handedness) to 100% (extreme right‐handedness). Individuals were classified as left‐handers if their scores were 40% or below and as right‐handers if their scores were 60% or above.

For the typing and typing control tasks, we measured left‐hand and right‐hand preference for each letter keystroke by having participants type a Greek sentence that covers all the letters (Pantogram) before the start of the fTCD data collection. Using a phone camera, we noted which hand is preferred for each keystroke (letter) for the sample sentence. We then used that as a baseline to estimate keystrokes per hand during the two tasks and calculate a left–right hand utilization index for each task. The left‐right hand utilization index was calculated by the formula [(LH−RH)/(LH + RH)]*100, where RH is the number of keys the participant typed using the right hand and LH the number of keys the participant typed using the left hand. This index was not used for the registered analyses but for exploratory analyses only.

We also collected further handedness measures for inclusion in our raw data set, thereby providing handedness data in a format comparable to studies that might have used different handedness measures. Indeed, the optimal handedness measurement is still a topic of debate. In fact, large studies (e.g., Annett et al. 1979; Cornish and McManus 1996; DeLisi et al. 2002; Gorynia and Müller 2006; Groen et al. 2013) and large‐scale meta‐analyses (e.g., Papadatou‐Pastou et al. 2020) have suggested the assessment of both hand preference and hand skill. The additional handedness measures include:

• 1Pegboard: Annett's pegboard task (Annett et al. 1979) was employed to measure relative hand skill. We used a 32 × 18 cm wooden board, which consisted of two attached wooden pieces with 10 holes drilled along their length. The distance between the two wooden pieces was 15 cm and the diameter of each hole was approximately 1.2 cm. Each peg was 7.0 cm in length and 1.0 cm in width. Once seated in front of the pegboard, the participants were asked to move all 10 pegs as quickly as possible from the full row to the empty row beginning on the side of the pegboard ipsilateral to the hand being used to perform the task. Trials for all participants started with the right hand and then the left and right hands alternated. The task was repeated three times for each hand. If a participant dropped a peg, the trial was repeated. Participants were instructed not to talk while carrying out the task, as talking might have caused interference. The time that lapsed between the touch of the first peg and the release of the last one was recorded for each participant for each trial (three trials for each hand) using a stopwatch. An LI was calculated using the formula: LI = [(LH‐RH) / (LH+RH)]*100, where RH is the time needed to move the pegs using the right hand and LH is the time needed to move the pegs using the left hand. A negative score was representative of a right hand superiority, while a positive score was representative of a left hand superiority (Figure 2).

FIGURE 2.

Example of wooden pegboard.

• 2Quantification of Hand Preference Test (QHPT): Hand preference was observed (rather than self‐reported, as is the case of the EHI) using the QHPT (Bishop et al. 1996). Seven positions were marked on a table, each at a distance of 40 cm from the midpoint of a baseline, at successive 30° intervals. Three cards were placed at each position, resulting in a sum of 21 cards. The participants were asked to stand in front of the table with their arms resting at their sides and to pick up a named card and place it in a box in front of them. The order of the cards was random but identical for all participants. The hand chosen to pick up each card was recorded. No points were assigned in the case that the left hand was used to place the card into the box, 1 point in case of changing hands, and 2 points if the right hand was used. The total points for each participant were then divided by the maximum score (42) and multiplied by 100 to calculate an LI. This LI varied from 0% (extreme left‐handedness) to 100% (extreme right‐handedness) (Figure 3).

FIGURE 3.

Representation of the spatial planning of the Quantification of Hand Preference Task (QHPT).

• 3“Seek time” metric: The “seek time” is a metric of the finger movements on the keyboard when a person is typing. It was first described by Pentel (2018) who showed that “seek time” predicts hand dominance with 95% accuracy. “Seek time” is the time between releasing one key until pressing the next one. Keys that are being operated by the dominant hand have on average a shorter “seek time”. The timing of the keystrokes was collected using a custom code while the participants were performing the typing task. An LI was calculated using the formula: LI = [(LH−RH)/(LH + RH)]*100, where RH is the “seek time” for the keys that were pressed using the right hand and LH is the “seek time” for the keys that were pressed using the left hand. A positive score is representative of a right‐hand superiority, whereas a positive score is representative of a left‐hand superiority.

2.4. Procedure

An online prescreening questionnaire was sent to potential participants (confirming normal vision, that they do not use illicit drugs, etc.). An information sheet was sent 72 h prior to participation to allow enough time to decide if they wish to take part in the study. Upon arrival at the laboratory, the study was explained to the participants, and they were encouraged to ask questions. They gave their informed written consent but were explicitly told that they were free to leave at any time, without providing a reason for doing so. The participants were seated individually in a quiet room. The fTCD experiment then took place. To prepare for fTCD, they were asked to sit in front of a computer, and they were given the option to watch the first few minutes of a video while the ultrasound robotic probes were being fitted. The data collection procedure ensued. After the first half of the experiment (first 40 trials of 80 total), we collected the handedness measures and returned for the second half. Participants were debriefed after the completion of the study.

2.5. Analysis Plan

We tested Hypothesis H1 by performing a Bayesian dependent (paired) samples t test for the LI_handwriting_corrected and LI_typing_corrected. Thus, the LIs are the dependent variables and the method of writing the independent variable. The prior for the t test is described by a Cauchy distribution with a width parameter of 0.707 (preregistered as 0.2 in Stage 1; see Section 2.7), which is the default prior for effects with an expected moderate effect size (Morey and Rouder 2011; Rouder and Morey 2012).

Hypothesis H2 was tested by performing two Bayesian independent samples t‐tests for the LI_handwriting_corrected and LI_typing_corrected (dependent variable) between left‐handers and right‐handers (independent variable). The Bayes factor implies the likelihood of the alternative (difference between handedness groups in the linguistic component of writing) over the null hypothesis. The prior for the t test is described by a half‐Cauchy distribution (Rouder et al. 2009) and with a width parameter of 0.94 (Schmalz, Manresa, and Zhang 2021). The value of the width parameter is equal to Cohen's d for the difference in lateralization of the linguistic component of writing (using copying symbols as an active baseline to writing words) between left‐handers and right‐handers in Papadatou‐Pastou et al. (2022).

2.6. Ethical Considerations

Ethical approval for the study was granted from the National and Kapodistrian University, School of Education (Protocol Number: 1936/25.2.2020), and the Biomedical Research Foundation of the Academy of Athens' Ethics Committee (Protocol Number: 65/25.05.2020).

2.7. Deviations From the Preregistered Protocol

The only deviations from the preregistered protocol were the following:

1. Epochs were excluded if they contained cerebral blood flow velocity values outside the range of 50%–150% of the mean instead of the preregistered 70%–130%. This was done due to anecdotal observations during the data collection, as the original range was found to be too conservative, potentially excluding valid trials. This decision was also supported by recent publications (Gutteridge et al. 2020; Petit, Badcock, and Woolgar 2020).

2. For Hypothesis H1, a prior of r = 0.707 was used instead of the preregistered r = 0.2. This was done following recent large simulations demonstrating that the BF can be disproportionately smaller in cases where the null hypothesis is true (compared to when the alternative is true), because of the large density assigned on the null (δ = 0) value by the Cauchy prior (Phylactou et al. 2024). Due to this disproportionate nature of the BF when it comes to null effects, a narrow prior set at r = 0.2 (such as the one originally registered) makes it impossible to detect adequate evidence for a true null effect, even with large sample sizes (i.e., up to 200 participants in each group; Phylactou et al. 2024). What we opted to use instead was a prior of r = 0.707 (√2/2), which is the standard prior for effects that are expected to have a moderate effect size (Morey and Rouder 2011; Rouder and Morey 2012).

3. We had preregistered that we would perform the pegboard and the QHPT tasks after the fTCD measurements had concluded. Instead, we performed the tasks in the break after the first half of the fTCD measurements. This was done because performing the handedness tasks at this point could not affect the outcome of the fTCD measurements in any way, and it helped fill the break time with something interesting for the participants.

4. Our sample included more right‐handers (n = 30) than the maximum number of right‐handers preregistered (n = 24). This was due to the fact that one of the ways that we recruited participants was by offering course credit to university students. Because we could not deny some students course credit, we tested  a total of 35 right‐handers (and ended up using data from 30). We then invited as many left‐handers as possible, and we succeeded in recruiting 27. However, upon realizing that we could use data from 23 of the  left‐handers, we recruited one additional left‐hander to reach 24, a sample size that aligns with the maximum preregistered number of participants in each handedness group (see Section 3 for a detailed explanation on why participants were excluded).

3. Results

Out of the 62 participants (35 right‐handers with mean age = 23.88 years and SD = 7.48, mean EHI score = 88.92 and SD = 9.59; 27 left‐handers with mean age = 30.82 years and SD = 7, mean EHI score = 16.51 and SD = 12.6) that came to our lab, we successfully obtained data from 58 (33 right‐handers with mean age = 23.09 years and SD = 6.55, mean EHI score = 88.4 and SD = 9.61; 25 left‐handers with mean age = 30.96 years and SD = 7.23, mean EHI score = 15.19 and SD = 12.04). The remaining four were dismissed due to failure of the ultrasound beam to penetrate their skull, an exclusion rate of 6.45%, which is lower compared to previous studies (18.9% in Kondyli et al. 2017; 19% in Papadatou‐Pastou et al. 2022) possibly due to our use of robotic probes. In addition, five more participants were excluded (8.6% exclusion rate) due to noisy data (i.e., not meeting the criterion of 10 accepted epochs per condition). This 8.6% exclusion rate is again lower than previous studies (Papadatou‐Pastou et al. 2022 and Papadopoulou et al. 2024; 19.5% and 14.3% respectively), but, of note, those studies used a 70%–130% threshold, whereas we opted for a threshold of 50%–150%. Twenty‐three trials from different participants were manually marked as nonvalid during data collection for various reasons, such as the participant not performing the correct task (e.g., performing the word generation task instead of the control task) or momentarily being distracted and not seeing the letter on the screen, and were therefore excluded from further processing.

Following this data preprocessing, our sample consisted of 23 left‐handers and 30 right‐handers. We had preregistered an upper limit of at least 24 participants per group in the case that the BFs for both hypotheses failed to reach the preregistered threshold (⅙ < BFs < 6), as was the case here. Therefore, we recruited one more left‐hander, and our final sample consisted of 30 right‐handers (two males, mean EHI score = 87.03 and SD = 9.8, mean age = 23.26 years and SD = 6.85) and 24 left‐handers (five males, mean EHI score = 22.59 and SD = 28.29, mean age = 31.37 years and SD = 7.33).

The preregistered split‐half reliability criterion was met for all tasks. We performed Bayesian Spearman's correlations using a uniform prior that assigns equal probability to all possible correlation values (stretched Beta [1, 1]), and the results are the following: LI_handwriting: r = 0.88, BF = 3.4 × 10¹⁶; LI_handwriting_control: r = 0.86, BF = 9.53 × 10¹⁴; LI_typing: r = 0.66, BF = 8.12 × 10⁵; LI_typing_control: r = 0.51, BF = 647.92. Hence, all tasks were included in the analysis. Moreover, the average number of keystrokes was not different between the two conditions (r = 0.707, BF = 0.33) (Figure 4).

FIGURE 4.

Average alterations in blood flow velocity for both left‐handed and right‐handed individuals under different conditions. (a) Average alterations in blood flow velocity for both left‐handed and right‐handed individuals representing the differences between handwriting and typing conditions. (b) Average alterations in blood flow velocity for both left‐handed and right‐handed individuals specifically for the typing conditions.

The LIs for the different handedness groups across the different conditions are presented in Table 1 together with the BF of the t test for the LI of the condition when compared to zero, using a prior of r = 0.707. Handwriting and typing were predominantly left‐lateralized for right‐handers. For the left‐handed participants, engaging in handwriting resulted in a rightward lateralization, whereas the average LI for typing did not significantly deviate from zero. Furthermore, the pattern of cerebral lateralization observed in left‐handed individuals when copying letters was rightward, as was the case with handwriting. For right‐handed participants, the LI for letter copying was positive, but the evidence of it differing from zero was weak (a Bayesian t test showed a threefold likelihood of it being non‐zero; BF = 3, r = 0.707). Further exploratory analyses were performed to investigate the underlying properties of this finding and are presented further in the text.

TABLE 1.

Descriptive statistics for cerebral lateralization in the different functional transcranial Doppler ultrasound experimental conditions.

Task	Handedness	Mean	SD	Epochs	Min	Max	BF
Handwriting	Right	3.35	2.54	18.76	−3.64	11.03	244,618.70
	Left	−2.89	2.89	18.66	−8.05	6.01	422.64
Handwriting control	Right	0.99	2.11	18.73	−4.72	5.87	3.06
	Left	−4.53	3.56	18.83	−16.05	3.57	8468.91
Typing	Right	2	1.77	18.76	−0.71	5.87	18,408.84
	Left	0.31	1.76	18.37	−4.37	3.17	0.299
Typing control	Right	−0.25	1.91	18.43	−4.19	3.79	0.246
	Left	−1.42	1.72	18.7	−6.51	1.45	66.22
Handwriting corrected	Right	2.36	1.64		−0.45	7.73	1,259,389
	Left	1.64	3.22		−6.84	9.5	2.72
Typing corrected	Right	2.25	1.42		−0.15	5.45	7,784,893
	Left	1.73	1.87		−4.03	4.92	192.93

Abbreviations: BF: Bayes factor of the t test for the LI of the condition when compared to zero using a prior of r = 0.707; min/max, minimum and maximum scores observed in each handedness group for the condition; SD, standard deviation.

3.1. Registered Analyses

Using the standard prior for an expected finding of moderate evidence (namely, √2/2), r = 0.707, we get a BF = 0.148, which gives more than six times the likelihood that there is an absence of difference in the cerebral lateralization during handwriting and typing. A robustness check was further performed for a wide range of priors (0–1.5) and showed that evidence of at least a threefold magnitude (BF = ⅓) in favor of the null hypothesis was found with a Cauchy prior set at r = 0.3 (BF = 0.334). Further, a Cauchy prior with r = 0.632 resulted in a BF = 0.167, indicating a sixfold ratio in favor of the null hypothesis. This robustness analysis indicates that with the various expectations of the potential true effect size, the null hypothesis is between three to six times more likely to be true than the alternative. Notably, these results also echo simulation results, indicating that there is below chance probability (< 50%) to reach evidence of at least a threefold magnitude in favor of the null hypothesis with very narrow priors (i.e., r < 0.3; Phylactou et al. 2024).

Using a prior of r = 0.94, we get a BF_typing_corrected = 0.386 and a BF_handwriting_corrected = 0.358, which both indicate that it is less than three times more likely that there is an absence of difference in the cerebral lateralization during writing between left‐handers and right‐handers. This is anecdotal evidence in favor of the null hypothesis that left‐handers and right‐handers do not show a difference in their linguistic component of writing lateralization for both means of writing. When performing a robustness check for a range of priors (r = 0–1.5), we find evidence of at least a threefold magnitude (BF = ⅓) for typing_corrected using an r = 1.14 (BF_typing_corrected = 0.332) and for handwriting_corrected using an r = 1.04 (BF_handwriting_corrected = 0.331). Similarly, we find evidence of at least a twofold magnitude above an r = 0.65 for typing_corrected (BF_typing_corrected = 0.499) and an r = 0.59 for handwriting corrected (BF_handwriting corrected = 0.495). Using an r = 1.5, we get a BF_typing_corrected = 0.263 and BF_handwriting_corrected = 0.224, both of which are higher than BF = ⅙. This robustness analysis suggests that, across a range of plausible effect sizes, the null hypothesis is either uninformative or is two to four times more likely than the alternative. However, it does not reach a sixfold likelihood, which would indicate moderate evidence for the null.

3.2. Exploratory Analyses

3.2.1. Differences Between the Lateralization Indices for the Linguistic Component of Handwriting and Typing in Each Handedness Group

Following Papadopoulou et al. (2024), we evaluated the potential differences in the linguistic component between handwriting and typing in each handedness group and observed weak evidence supporting the absence of a difference in both left‐handers (r = 0.707: BF = 0.218) and right‐handers (r = 0.707: BF = 0.202). A scatterplot of the LIs for the two handedness groups can be found in Figure 5.

FIGURE 5.

Lateralization indices for the different conditions and handedness groups. Note: Lateralization indices for the different writing conditions. The handedness of participants, categorized according to the EHI, is represented in the graph with: left‐handers in squares and straight lines; right‐handers in circles and dotted lines. Bold squares and circles represent the mean for left‐handers and right‐handers, respectively, and gray symbols depict individual participant lateralization indices. The bars represent confidence intervals of 95%.

3.2.2. Hemispheric Dominance

The letter copying task can also be considered a positive control task, because it is meant to isolate exclusively the motor component of handwriting, which is expected to show a cerebral lateralization contralateral to the writing hand. Using jack‐knifed standard deviations (as per Papadatou‐Pastou et al. 2022 and Papadopoulou et al. 2024), we determined that, among right‐handers, 22 individuals (74%) and, among left‐handers, two individuals (8.3%) showed a leftward lateralization when copying letters. This suggests that although letter copying is expected to primarily activate the motor component of writing, leading to a contralateral lateralization, 26% of right‐handers displayed an atypical lateralization pattern, compared to 8.3% of left‐handers. Despite these variations in the lateralization for letter copying across handedness groups, the linguistic component of handwriting, obtained by subtracting the LI for letter copying from the LI for handwritten word generation, was left lateralized for both groups. Utilizing the same method of jack‐knifed standard deviations, we found that for the linguistic component of handwriting, 19 left‐handers (79.2%) and 29 right‐handers (96.7%) were classified as left‐lateralized. Similarly, for typing, 21 left‐handers (87.5%) and 29 right‐handers (96.7%) were classified as left‐lateralized.

3.2.3. Differences in Cerebral Lateralization During Writing Among Participants With Typical Lateralization

To address potential concerns regarding the influence of individual variability in lateralization, we conducted additional analyses excluding participants with extreme LIs and handedness scores. First, we excluded one highly left‐lateralized left‐handed participant (female), whose LI of 9.5 was notably higher than others (see Figure 6). This analysis showed minimal impact on our conclusions for Hypothesis H1, with the Bayes factor remaining consistent with the original findings (BF = 0.16 with r = 0.707), suggesting that the exclusion of this participant did not substantively affect our interpretation of lateralization differences between handwriting and typing. For Hypothesis H2, we reevaluated typing and handwriting data after excluding the highly left‐lateralized left‐handed participant, yielding BFs of 0.43 and 0.79, respectively, which were less informative than those of the initial analysis. Furthermore, we repeated our analyses while excluding left‐handers with extreme EHI scores below 5% (six females and three males) and again found that BFs remained consistent with our original interpretations (BF = 0.18 for Hypothesis H1 and BFs of 0.26 and 0.34 for handwriting and typing, respectively). When performing an analysis to identify whether the extreme left‐handers have a different lateralization than the nonextreme left‐handers, we get a BF = 1.05, which is as close as a result can be to being uninformative. In sum, these additional analyses confirmed that the inclusion of extreme left‐handers did not substantially alter our findings.

FIGURE 6.

Scatterplot of the lateralization indices for the linguistic component of handwriting and writing using a keyboard (typing) for each handedness group (left panel: left‐handers; right panel: right‐handers). Note: The error bars correspond to the 95% confidence intervals for the LI of each participant (horizontal: LI_handwriting_corrected; vertical: LI_typing_corrected). Participants were classified in handedness groups using the EHI. The red line corresponds to the linear trend of the correlation coefficients surrounded by their standard deviation (gray area).

3.2.4. Exploration of Cerebral Lateralization Among Left‐Lateralized Individuals

To explore potential differences in lateralization among left‐hemisphere dominant participants, we reanalyzed the data, excluding right‐hemisphere dominant individuals from both handedness groups. The new sample consisted of 19 left‐handers (mean age = 31.00 years, five males) and 29 right‐handers (mean age = 23.31 years, two males). This aimed to assess whether left‐dominant left‐handers in our sample exhibited weaker left lateralization than right‐handers, as reported by Johnstone, Karlsson, and Carey (2021) for the cerebral lateralization of oral language, and to determine if the inclusion of right‐dominant participants had influenced our main findings. For Hypothesis H1, which tests the absence of a difference between handwriting and typing, the Bayes factor was 0.39 (r = 0.707), compared to 0.15 in the original sample. For Hypothesis H2 (handwriting), testing lateralization differences between left‐handers and right‐handers, the reanalysis yielded a BF of 0.28 (r = 0.94), slightly stronger evidence toward the absence of a difference than the original BF of 0.36. For Hypothesis H2 (typing), comparing left‐handers and right‐handers during typing, we obtained a BF of 0.25 (r = 0.94), indicating stronger evidence for the absence of a difference compared to the original BF of 0.39.

4. Discussion

This study was designed to (1) investigate the relationship between the lateralization indices of the linguistic component of handwriting and typing and (2) examine differences in these indices between left‐handers and right‐handers. We measured relative hemispheric activation in 24 left‐handed and 30 right‐handed participants through fTCD as participants engaged in four tasks: handwritten word generation, copying letters, word generation using a computer keyboard, and pressing random keys using a computer keyboard. The LI for the linguistic component of each mean of writing was determined by subtracting the LI during a control task (letter copying for handwriting and random key presses for keyboard typing) from the LI observed during the corresponding written word generation task (handwriting and typing, respectively). It is important to note that the selection of letter copying as a control task for handwriting aimed to refine the methodology of Papadatou‐Pastou et al. (2022), who used symbol copying as a control condition. Symbol copying was suggested to induce right‐hemispheric activation due to its higher attentional demands, stemming from the novelty of symbols compared to letters. Adequate reliability for all the tasks was confirmed.

Our first hypothesis posited that the linguistic component of handwriting and typing would not differ with regards to cerebral lateralization. Our findings, indicating a greater than six times likelihood of no difference between these modalities (BF = 0.148), support the notion that the cerebral lateralization of the linguistic component of writing remains consistent across different means of writing. This aligns with the theoretical expectation that the neural circuits involved in the linguistic component of written language production should not vary with the means of writing, opening the way for future research to potentially utilize writing using a computer keyboard as a viable means for studying the linguistic component of writing. Studying written language using writing on a computer keyboard, as opposed to handwriting, facilitates the collection of additional data, such as keystroke analytics, allowing for more nuanced analyses.

Our second hypothesis was that the linguistic component of handwriting and keyboard typing would exhibit stronger left lateralization in right‐handed individuals compared to their left‐handed counterparts. The hypothesis was based on previous findings on the cerebral lateralization of written language by Papadatou‐Pastou et al. (2022) and the general trend of left‐lateralization in oral language production among neurotypical individuals (e.g., Assaneo et al. 2019; Petit, Badcock, and Woolgar 2020), with right‐handers showing stronger leftward lateralization compared to left‐handers (Bruckert et al. 2021; Knecht et al. 2000; Kondyli et al. 2017). Although we observed left‐lateralization for the linguistic component of writing in both handedness groups and for both means of writing, the evidence was not strong enough to conclusively support either the presence or the absence of a difference between left‐handers and right‐handers; only anecdotal evidence emerged for an absence of a difference (BF_typing = 0.386; BF_handwriting = 0.358). This suggests that the neural underpinnings of written and oral language might differ, with only oral language being differentially lateralized between left‐handers and right‐handers. Our findings could also have been influenced by our inclusion criteria of accepting left‐handers with an EHI score of 0%–40% and right‐handers with an EHI score of 60%–100%. With stricter criteria, our results might have differed, as there is evidence of greater likelihood of rightward cerebral lateralization for extreme left‐handers (Knecht et al. 2000). Too few extreme left‐handers were included in our sample (nine out of 24 left‐handers had an EHI score equal or smaller than 5%) to test this hypothesis with our current data set. Alternatively, the fact that only anecdotal findings emerged for this comparison might be due to the nature of the control tasks that we implemented to isolate the motor component of writing. In particular for typing, use of both hands may have led to a more balanced hemispheric activation across different handedness groups. Nonetheless, the fact that Hypothesis H1 was confirmed, namely, that there was evidence of an absence of lateralization differences in the linguistic component between handwriting and typing (i.e., after controlling for the motor component of writing), suggests that this interpretation might be less tenable.

With respect to the lateralization for the letter copying task, 26% of the right‐handers were not left‐lateralized. Assuming that letter copying primarily isolates the motor aspect of writing, right‐handers were expected to show leftward lateralization (i.e., lateralization contralateral to their writing hand). Conversely, only 8.3% of left‐handers displayed an unexpected leftward lateralization for the letter copying task. These findings mirror those of a recent study from our laboratory, where 37% of right‐handers and 7% of left‐handers showed a lateralization pattern during letter copying that was ipsilateral to their dominant hand (Papadopoulou et al. 2024). The consistency of these results across two separate studies reduces the likelihood of our sample being nonrepresentative. We thus join Papadopoulou et al. (2024) in suggesting that the letter copying task may potentially implicate right hemisphere functions. Letter copying and written word generation tasks both require muscle coordination and working memory (Palmis et al. 2017). However, letter copying is simpler than written word generation and might engage the right hemisphere more due to its repetitive nature (Chang 2020). Lesion studies have shown that, when the left hemisphere is damaged, the right hemisphere is able to mediate language processing using a letter‐by‐letter method of reading (Behrmann, Plaut, and Nelson 1998; Tröster et al. 1996). This letter‐by‐letter method of processing words could be associated with the letter copying task that we employed and explain why a larger than expected percentage of participants exhibited a rightward cerebral lateralization. Another interpretation could be that high‐level linguistic processing might occur when copying letters (for a discussion, see Papadopoulou et al. 2024).

The exploratory analysis, excluding right‐hemisphere dominant participants, provided further insights into the lateralization differences in left‐hemisphere dominant individuals. For Hypothesis H1, the results continued to support the absence of a difference in lateralization between handwriting and typing, though the evidence was less conclusive, likely due to the reduced sample size. For Hypothesis H2 (handwriting), the findings showed slightly stronger evidence for the absence of lateralization differences between left‐handers and right‐handers, likely reflecting the narrower focus on left‐dominant participants. Although these results did not align with Johnstone, Karlsson, and Carey's (2021) report of greater left lateralization in left‐dominant right‐handers, they provide a closer comparison to their study. Similarly, for Hypothesis H2 (typing), the reanalysis aligned with findings for handwriting, offering further support for the absence of a lateralization difference between left‐handers and right‐handers. Overall, these additional analyses affirm the robustness of our findings, although we did not replicate the findings of Johnstone, Karlsson, and Carey (2021), who observed greater left lateralization in left‐dominant right‐handers compared to left‐dominant left‐handers in oral language. This discrepancy may stem from task differences, as written language tasks such as handwriting and typing likely engage distinct neural mechanisms compared to oral language. Additionally, differences in sample characteristics, smaller sample size after excluding right‐dominant participants, and methodological differences (fTCD vs. fMRI) may have contributed. Lastly, lateralization patterns may vary more across tasks than previously assumed, potentially explaining the divergence in findings.

Our additional analyses addressing the influence of participants with extreme laterality indices and handedness scores replicated our findings from our preregistered analyses. Excluding these individuals did not substantively alter the evidence supporting the absence of differences in cerebral lateralization between handwriting and typing or between left‐handers and right‐handers. This underscores that the observed patterns are not driven by outliers.

The current study has several strengths. It is the first study to investigate the cerebral lateralization of modern, technological means of writing (i.e., typing using a computer keyboard) while employing a neurophysiological technique specifically designed to assess lateralization (i.e., fTCD) and further doing so using robotic probes. It also introduces the use of random key presses on a computer keyboard as a control condition for typing, offering a novel approach, which employs the use of both hands, for isolating the linguistic component of typing. For reference, in the 2020 study of Krueger et al., prose typing was compared to typing code. Whereas typing code does involve motor aspects, it also engages higher order cognitive functions like reasoning, problem‐solving, and working memory. This makes it challenging to isolate only the motor aspects, as both prose typing and code typing may activate cognitive processes beyond language production.

Moreover, the adoption of the Bayesian framework for our analysis allowed us to discern between the absence of evidence and evidence of absence, providing clarity in the interpretation of our findings. The limited spatial resolution of fTCD, although it only provides information at the hemisphere level and does not allow for further localization, lends itself well to addressing our study's primary aim of investigating cerebral lateralization differences during writing. However, future research employing higher resolution techniques like fMRI could provide more detailed insights into the specific brain regions within each hemisphere that contribute to the linguistic and motor components of writing. This could further add to our understanding of the neural mechanisms underlying cerebral lateralization for writing. We would also like to note that our participants were predominantly female (seven males and 44 females). However, our aim was not to investigate sex differences, as previously a meta‐analysis of functional brain imaging studies (Sommer et al. 2008) has shown no consistent evidence of sex differences during language tasks. More recently, a meta‐synthesis showed that, when present, sex differences account for only about 1% of the variance in brain structure or functional laterality (Eliot et al. 2021; also see Eliot 2024).

One limitation of this study is the relatively, albeit larger than preregistered, moderate size sample (24 left‐handed and 30 right‐handed participants). Although our Bayesian analyses provided moderate evidence for our first hypotheses, it did not give conclusive results for the second hypothesis. Moreover, the sample size could still have limited the precision of estimates and the ability to detect more nuanced or smaller effects, particularly when considering variability within handedness subgroups. Future studies with larger and more diverse samples are necessary to replicate and extend these findings across handedness subgroups.

In summary, our research robustly showcases that the linguistic component of handwriting and keyboard typing share their cerebral lateralization patterns. It further showed that the linguistic component of writing is left‐lateralized in both handedness groups. However, anecdotal evidence of an absence of differences in lateralization between left‐handers and right‐handers was found, contrary to what is the case for oral language lateralization. The intricate nature of written language, its interaction with handedness, and the methodological challenges in isolating its linguistic component highlight the need for further investigation in this area.

Author Contributions

Christos Samsouris: data curation, formal analysis, investigation, methodology, writing – original draft, writing – review and editing. Konstantina A. Papadopoulou: formal analysis, investigation, writing – review and editing. Nicholas A. Badcock: formal analysis, software, writing – review and editing. Filippos Vlachos: writing – review and editing. Phivos Phylactou: formal analysis, methodology, writing – review and editing. Marietta Papadatou‐Pastou: conceptualization, funding acquisition, methodology, project administration, supervision, writing – original draft, writing – review and editing.

Conflicts of Interest

The authors declare no conflicts of interest.

Peer Review

The peer review history for this article is available at https://www.webofscience.com/api/gateway/wos/peer‐review/10.1111/ejn.70013.

Data Availability Statement

All materials, code and data have been uploaded in the Open Science Framework online repository (https://osf.io/9qmge/?view_only=62cd4305f2d24ea8aae8810bb659e407).