Ten right-handed healthy adults (1 female, 9 males), aged 22–30 years old, participated in experiment 1. Ten adults (1 female, 9 males), aged 23–30 years old, participated in experiment 2. Ten adults (2 female, 8 males), aged 23–30 years old, participated in experiment 3. In the present study, the inclusion criterion was an age of 20–40 years, while exclusion criteria were a history of neurological and psychiatric diseases, neurological surgery, and substance abuse. Some subjects participated in two or three experiments in a random order with at least a one-month interval between experiments. The present study was approved by the Ethics Committee, Graduate School of Comprehensive Human Sciences at the University of Tsukuba.

Electrocutaneous stimuli (square wave, constant current pulse) of a 0.2-ms duration were presented to the left thumb (40%, standard) or middle finger (10%, deviant) and right thumb (40%, standard) or middle finger (10%, deviant) in a random order through ring electrodes attached to the first (anode) and second (cathode) interphalangeal spaces. The stimulus intensity was adjusted to approximately 2.5-fold the subject’s sensory threshold and was never reported as painful and uncomfortable. Interstimulus intervals varied randomly between 700 and 900 ms for 11 steps (mean 800 ms).

Subjects were seated comfortably in a chair in an electrically-shielded room, placed their hands with the palms down on a wooden board, and performed several attention conditions in each experiment. They were instructed to look at a crosshair 1.5 m in front of them and not to look at their hands during performance of the task. In all experiments, each condition consisted of 4 runs of approximately 200–300 stimuli, resulting in 1000 stimuli (400 left and 400 right standard stimuli and 100 left and 100 right deviant stimuli). The order of conditions was randomized among subjects. The interval between conditions was approximately 3 min, and that between runs was about 1 min. In all experiments, a control condition was performed, where subjects were instructed to relax and look at a crosshair 1.5 m in front of them and had no task. The control condition allowed us to assess whether ERPs were facilitated on the attended side or were suppressed on the unattended side (Alho et al., 1987; Desmedt and Tomberg, 1989, 1991; Garcia-Larrea et al., 1995; Kida et al., 2004b).

Electroencephalograms (0.5–100 Hz) were recorded at a sampling rate of 500 Hz with Ag/AgCl electrodes from 5 locations on the scalp: Fz, Cz, Pz, C3, and C4 (SYNAFIT, Nihon Denki San-ei Corp., Japan). All the electrodes were referenced to the average of earlobes. Impedance was carefully balanced and maintained below 5 kohm. Electrooculograms (EOG) were recorded bipolarly from the right outer canthus and suborbital region to monitor eye movements or blinks. The analysis time was 600 ms, including a 50-ms prestimulus baseline. Trials exceeding ±80 μV (EOG and EEG amplitudes) were automatically excluded from averaging, and trials with eye blinks and eye movements were excluded manually. Trials with omission and commission errors were also excluded from further analyzes. Grand-averaged waveforms were filtered using a low-pass Butterworth filter with a cut-off frequency of 40 Hz.

N140 amplitude was increased more by focused attention than by the other conditions examined. Furthermore, N140 amplitude in the divided attention condition was intermediate between the focused and unattended conditions. P300 amplitude was lower in the divided attention condition than in the focused attention condition. Detailed information is provided below.

The effects of focused and divided attention on ERP amplitude were similar to those in experiment 1 (Figure 4). P300 latency was longer in the divided attention condition than in the focused attention condition. Hand position did not affect N140, but changed P300 amplitude and latency. Detailed information is provided below.

Figure 8 shows grand-averaged ERP waveforms in experiment 3. The effect size of focused attention on N140 amplitude was changed by crossing the forearms. Forearm posture affected P300 amplitude and latency. Detailed information is provided below.

In the analysis of single-trial P300 latency using the ACF technique, 38.6 and 39.9% of all trials were identified as good estimates in experiments 2 and 3, respectively. In experiment 2, the two-way ANOVA (3 attention conditions [unilaterally-focused/hands-separated, unilateral focused/hands closely-placed, and divided] and 2 target sides [left-and right-hand targets]) of single-trial P300 latency showed a main effect of the attention condition (F (2, 18) = 4.4, p < 0.05; η_p² = 0.33), similar to the analysis of the averaged waveform, with latency being longer with hands closely-placed than separated (Table 4). Divided attention also resulted in slightly longer latency (Table 4) than unilaterally-focused attention. This was the same pattern as that observed in the results on averaged waveforms. The single-trial analysis of RT and P300 latency showed that RT preceded P300 latency in 46.0, 36.9, and 36.6% of good trials in the focused/hands-separated, focused/hands closely-placed, and divided attention conditions, respectively, for left-hand targets, and in 54.6, 43.1, and 39.2%, respectively, for right-hand targets. Therefore, right-hand targets had more trials with RT preceding P300 latency than left-hand targets in all conditions tested, and also hands closely-placed and divided attention decreased the preceding ratio of RT (or the delayed ratio of P300), resulting in a decrease in the difference between left-and right-hand targets. The variabilities (SD) of single-trial P300 latency and RT were not significant in ANOVA with the attention condition and target side. The correlation of single-trial P300 latency with RT was moderate in the unilaterally-focused attention/hands-separated condition (correlation coefficient R = 0.52) and low in the unilaterally-focused attention/hands closely-placed condition (R = 0.38) and divided attention condition (R = 0.38), all of which showed correlations (Table 4). The former showed a stronger correlation than the latter two.

In experiment 3, the two-way ANOVA of (2 forearm positions and 2 target sides) single-trial P300 latency showed a main effect of forearm posture (F (1, 9) = 6.1, p < 0.05; η_p² = 0.41), with latency being longer when the forearms were crossed than when they were uncrossed irrespective of the target stimulus hand (Table 5). Therefore, the same pattern of results was observed as averaged waveforms. The single-trial analysis of RT and P300 latency showed that RT preceded P300 latency in 42.5 and 50.7% of good trials in the uncrossed-and crossed-forearms conditions, respectively, for left-hand targets, and in 59.3 and 51.3%, respectively, for right-hand targets (Table 5). Therefore, right-hand targets had more trials with RT preceding P300 latency in both forearm positions, and crossing the forearms decreased the RT-P300 latency difference between left-and right-hand targets. The variabilities (SD) of single-trial P300 latency and RT were not significant in ANOVA with forearm position and the target stimulus hand. A correlation of single-trial P300 latency with RT was observed in the uncrossed-forearms condition (R = 0.37, significant), but not in the crossed-forearms condition (R = 0.23). Furthermore, a significant difference was observed in the correlation between the two conditions (z = 1.98) with a small effect (q = 0.15). When it was examined separately for each target side, the correlation was significantly higher in the uncrossed-forearms condition for right-hand targets (R = 0.45) than in the crossed-forearms condition (R = 0.23) (z = 2.29, p < 0.05) with a small effect (q = 0.25), whereas left-hand targets showed no significant difference between the crossed- and uncrossed-forearms conditions.

Another result on the correlation of single-trial P300 latency with RT was observed when left-and right-hand target stimuli were separately analyzed. The single-trial P300 latency-RT correlation in most of the conditions examined in experiments 2 and 3 was slightly stronger for right-hand target stimuli than for left-hand target stimuli (the only exception was the crossed-forearms condition), whereas no significant difference was observed when it was tested separately in each condition (Tables 4, 5). To increase the statistical power and examine correlation patterns general to all attention conditions, we concatenated latency and RT data from all attention conditions for each target side and then compared correlations between different target sides. The analysis revealed correlations for both target sides (R = 0.31 for left-hand target, R = 0.41 for right-hand target) and a significant difference in the correlation between left-and right-hand target stimuli (z = 2.28, p < 0.05) with a small effect (q = 0.12). Collectively, correlation patterns showed that right-hand attended targets (and responses) produced a stronger P300 latency-RT correlation with a small effect than left-hand targets, which was decreased by crossing the forearms.

The present study showed that the amplitude of N140 was modulated by directing attention to the unilateral hand, and was higher for attended stimuli than for unattended stimuli and identical stimuli in the control condition. Therefore, we successfully replicated previous findings on the effects of somatosensory attention on the amplitude of N140 (Desmedt and Robertson, 1977; Desmedt et al., 1984; Michie et al., 1987; Desmedt and Tomberg, 1989; Garcia-Larrea et al., 1995; Eimer and Forster, 2003a; Kida et al., 2004b, 2012a,b).

N140 amplitude in the divided attention condition was intermediate between those elicited by attended and unattended stimuli during the focused attention condition. Previous studies in audition showed that the amplitude of N1 in the divided attention condition was intermediate between those elicited by attended and unattended stimuli during the focused attention condition (Hink et al., 1977, 1978; Parasuraman, 1978), which was consistent with the present results. Therefore, dividing attention between the hands may be controlled similarly to auditory divided attention, which may be explained by a capacity model of attention (Hink et al., 1977). Some researchers have also suggested that the attentional modulation of early ERP components is associated with the perceptual resource (Kok, 1997; Kida et al., 2012a,b) in the framework of multiple resources, including perceptual, central, and response resources (Wickens, 1991; Wickens and McCarley, 2008).

Regarding the distribution of attention, previous studies demonstrated a spatial gradient of attention in different modalities using visual (Mangun and Hillyard, 1988; Wijers et al., 1989; Heinze et al., 1994), auditory (Teder-Salejarvi and Hillyard, 1998; Teder-Salejarvi et al., 1999), and somatosensory ERPs (Heed and Roder, 2010), suggesting modality-independent patterns of the distribution of attention. In addition, visual and auditory ERP studies reported that attention forms a unitary zone that may expand to multiple relevant locations, but also includes the area between them (Mangun and Hillyard, 1988; Wijers et al., 1989; Heinze et al., 1994; Teder-Salejarvi et al., 1999). In contrast, a previous study on touch reported that when attention was directed simultaneously to non-adjacent fingers within one hand, ERPs in response to stimuli delivered to spatially and anatomically intervening fingers showed no attentional modulations (Eimer and Forster, 2003b). This study concluded that, in contrast to vision, the focus of somatosensory attention may be split and directed simultaneously to non-adjacent areas, thereby excluding spatially and anatomically intermediate regions from attentional processing. An MEG study also reported that somatosensory attention has a gradient and may also be divided into non-adjacent areas (Kida et al., 2018), supporting ERP results. These electrophysiological findings indicate a somatosensory-specific pattern of attention. Based on the present and previous findings, we speculate that somatosensory attention may be split between non-adjacent fingers and also between the hands. The modality specificity of the distribution of attention needs to be examined in more detail in future studies.

P300 was found for attended infrequent stimuli (targets) in both the focused and divided attention conditions, but not for identical stimuli in the control condition or unattended infrequent stimuli in the focused attention condition. This pattern is a common feature of the appearance of P300. In addition, P300 amplitude was lower when attention was directed simultaneously to both hands than when it was unilaterally focused on one hand. Previous studies reported that P300 amplitude reflects the amount of the modality-non-specific, perceptual-central resource allocated to a given task (Wickens et al., 1983; Kramer et al., 1985; Sirevaag et al., 1989; Kok, 1997, 2001; Kida et al., 2004a, 2012a,b). The gradual change observed in P300 amplitude in the present study (i.e., focused attention >divided attention >unattended or control condition) was consistent with the resource allocation view of P300 amplitude. Therefore, the decrease in P300 amplitude in the divided attention condition was assumed to be caused by the division of the modality-non-specific perceptual-central resource between the hands. The common pattern of P300 amplitude to mental counting and motor response tasks shows that the allocation of the modality-non-specific resource to both hands was independent of whether the response was covert (mental) or overt (motor).

In the divided attention condition, target probability was two-fold or target-to-target interval (TTI) was half that in the focused attention condition. P300 amplitude has been shown to decrease with high target probability and a short TTI (Kok, 2001; Polich, 2007). Therefore, it is unclear whether the decrease in P300 amplitude in the divided attention condition was due to resource division or changes in task difficulty following changes in target probability and TTI. However, the resource limitation explanation may account for the potential relationship of target probability and TTI with P300 amplitude. When target stimuli are presented more frequently (a higher target probability or shorter TTI), more resources are consumed in a given amount of time than with less frequently presented stimuli, and P300 amplitude is small. When stimuli are presented more infrequently (lower target probability or longer TTI), the structures involved in the generation of P300 may recover more fully and P300 amplitude is large (Gonsalvez and Polich, 2002; Gonsalvez et al., 2007). This resource limitation explanation accounts for the interaction between task difficulty and target stimulus probability (Kramer et al., 1986; Ruchkin et al., 1987; Polich et al., 1988). Therefore, even if the decrease in P300 amplitude by dividing attention to both hands is associated with higher target probability or shorter TTI, it may also be explained by the resource allocation view.

In contrast to the common pattern in P300 amplitude, P300 latency was longer in the divided attention condition than in the focused attention condition in the motor response task, but not in the mental counting task. This result suggests the functional dissociation of the amplitude and latency of P300. More specifically, the modality-non-specific resource may be divided between the hands without affecting the stimulus evaluation speed during the mental counting task, whereas dividing attention between the hands decreases the evaluation speed during the motor response task. It remains unclear whether motor response demands affect P300 latency, with some studies reporting no effects (Kutas et al., 1977; McCarthy and Donchin, 1981; Magliero et al., 1984; Doucet and Stelmack, 1999). In contrast, other studies noted the significant effects of motor response demands on P300 latency (Ragot and Renault, 1981; Ragot, 1984; Ragot and Renault, 1985; Pfefferbaum et al., 1986; Ragot and Lesevre, 1986; Christensen et al., 1996; Leuthold and Sommer, 1998). However, the effects of motor response demand on P300 latency have not been investigated under divided versus focused attention conditions in the somatosensory modality. In the motor response task, attention needed to be divided between the hands for both stimulus and action processes. Therefore, attentional demands required for action may be sufficiently high to decrease the stimulus evaluation speed in the motor response task to lower than that in the mental counting task with no motor response demand.

Hand position, such as closely-placed hands, did not significantly affect the attentional modulation of N140 amplitude. This result shows that the attentional modulation of N140 largely depended on the anatomical space rather than the physical space. If an attentional effect on N140 amplitude is exclusively based on the physical (external) space, it is expected to disappear with closely-placed hands because spatial attention will operate equally on both to-be-attended and not-to-be-attended closely-placed hands. In contrast, if an attentional effect on N140 amplitude is based on the anatomical space, it is expected to appear in both separated-hands and closely-placed hands conditions. In addition, changes in hand position follows changes in arm posture. Therefore, a postural difference in the arms with closely-placed hands does not affect the attentional modulation of N140 amplitude.

In contrast to closely-placed hands, crossing the forearms reduced the attentional increase in N140 amplitude for standard stimuli. In a selective attention task, the ERP amplitude for standard stimuli generally reflects a pure selective attention effect, whereas that for target stimuli may contaminate a target-related effect or potential (Garcia-Larrea et al., 1995; Kida et al., 2004b, 2018). Therefore, the attentional increase in N140 amplitude in the present study represents the effects of somatosensory selective attention, which may have operated less efficiently at this stage when the forearms were crossed. Since positional and postural changes to the hands and forearms with closely-placed hands did not affect the attentional increase in N140 amplitude as discussed above, crossing the forearms may be a specific hand position and arm posture leading to changes in attentional somatosensory processing at this stage. A possible explanation for this result is based on the experimental condition that the hands and arms were placed on an unhabitual side with an unhabitual posture. This crossed-forearms condition will produce an incongruency between a representation of actual stimulus (and response) sides and a mental image of an internal space. This incongruency may result in the suppressive effect of crossing the forearms on the attentional increase in N140 amplitude. Previous studies demonstrated that attentional enhancements in early ERP amplitudes were smaller for crossed forearms than for uncrossed forearms (Eimer et al., 2001; Kennett et al., 2001; Eimer et al., 2003a; Gherri and Forster, 2012a,b), supporting the present results. Therefore, the result suggests that the attentional modulation of stimulus processing reflected by N140 amplitude is not only based on the anatomical space, but also the congruency between real and internal spaces depending on the hand position and arm posture. A psychophysical study reported that somatosensory attention was dependent on the physical space, but not on the anatomical space (Lakatos and Shepard, 1997). In contrast, ERP studies provided evidence to show that somatosensory attention was associated with an incongruency between different spatial coordinates (Eimer et al., 2001; Kennett et al., 2001). The present finding supports the latter ERP evidence. This pattern of attentional modulation associated with different spatial codes is also consistent with generators of N140, which originates from modality-specific and multisensory areas, including the second somatosensory, anterior cingulate, and prefrontal cortices (Allison et al., 1992; Tarkka et al., 1996; Waberski et al., 2002; Inui et al., 2003; Tanaka et al., 2008).

In contrast to the early N140 component, P300 amplitude was affected by both hand position and forearm posture. The behavioral measures, RT and RA, paralleled a decrease in P300 amplitude and increase in P300 latency in both the crossed-forearms and hands closely-placed conditions. P300 amplitude has been associated with equivocation or post-stimulus uncertainty (Sutton et al., 1965; Johnson, 1986; Kok, 2001) as well as resource allocation, whereas P300 latency was related to the stimulus evaluation time (Kutas et al., 1977). Therefore, crossing the forearms may decrease the resolution of post-stimulus uncertainty and increase the stimulus evaluation time through a congruency of real and learned spaces, whereas closely-placed hands exert the same effects by overlapping the attentional range at targets and non-targets.

Crossing the forearms significantly decreased the correlation of single-trial P300 latency with RT. As discussed above, crossing the forearms may produce an incongruency between a representation of actual stimulus (and response) sides and a mental image of a learned physical space. A previous study suggested that following accuracy maximizing instructions, subjects hesitated before pressing the button because the task used reduced their confidence of a correct response, thereby decoupling P300 latency from RT (Pfefferbaum et al., 1983). Similarly, an incongruency between actual and learned spaces may be associated with this hesitation before responding, thereby resulting in the decoupling of stimulus- and response-related processing. Crossing the forearms also decreased P300 amplitude and increased P300 latency as already discussed. Therefore, we speculate that crossing the forearms caused the decoupling of stimulus- and response-related processing, decreased the resolution of post-stimulus uncertainty, and reduced the stimulus evaluation speed concomitantly through an incongruency between real and learned spaces.

The present and previous studies using the ACF technique showed the task-dependent nature of the correlation of single-trial P300 latency with RT. A historical study on P300 latency using ACF found that speed-maximizing instructions resulted in a weaker correlation of single-trial P300 latency with RT than accuracy-maximizing instructions, suggesting the loose coupling of stimulus- and response-related processing in the former instructions and motor command output before the stimulus has been fully evaluated (Kutas et al., 1977). In contrast, another study reported the reverse effect, i.e., a weaker correlation of P300 latency with RT under accuracy-maximizing instructions than speed-maximizing instructions (Pfefferbaum et al., 1983). The latter study suggested that the difference in tasks and strategic differences for task requirements explain the discrepancies in the findings obtained as discussed above. We also previously used the ACF technique to demonstrate stronger stimulus–response coupling by the performance of a dual task than a single task (Kida et al., 2012b). Two explanations for this result, a snap decision strategy in the single-task and the lack of the resource allocated to the eliciting task during the dual-task performance, were suggested because stronger stimulus–response coupling by the dual-task performance was associated with a lower P300 amplitude and smaller number of trials with RT preceding P300 latency. However, the ratio of trials with RT preceding P300 latency was even lower in the present selective attention task (36.9–59.3%) than during single-task performance (76.8%) and was in the same range as that during the dual-task performance (53.7%). This is because the present study employed a type of selective attention task where somatosensory stimuli were presented to many fingers in bilateral hands and subjects had to discriminate one or two target stimuli, i.e., task demand was higher than the simple somatosensory oddball task used in the previous study where two types of somatosensory stimuli were presented to unilateral fingers. RT was earlier in the previous study (349.5 and 336.6 ms in experiments 1 and 2, respectively) than in the present study (ranging between 363.5 and 394.7 ms depending on the task condition, Tables 2, 5), supporting the selective attention task used herein being a more demanding task than the oddball task used in the previous study. Therefore, the snap decision is not the main cause for the changes observed in the coupling of stimulus- and response-related processing in the present study. Another suggestion, the lack of the perceptual-central resource, is also not straightforward to explain our previous findings and the present results because the two studies showed the reverse pattern for P300 amplitude and the correlation of single-trial P300 latency with RT. Therefore, the present and previous studies showed the importance of task features in interpreting the correlation of single-trial P300 latency with RT.

In the present study, the correlation between single-trial P300 latency and RT was stronger for the right hand than for the left hand in right-handed subjects. Previous studies reported a hand preference (i.e., a difference between the dominant and non-dominant hands) in motor tasks and sensorimotor tasks (Beste et al., 2009), whereas others showed no hand preference in a tactile perceptual task (Finlayson and Reitan, 1976). Regarding neural activation, movements with the non-dominant hand were associated with stronger and more extended brain activation than those with the dominant hand (Leocani et al., 2001; Potgieser et al., 2015). In contrast, an ERP study found that theta power related to handwriting was higher with the dominant hand than with the non-dominant hand (Pei et al., 2021). In addition to these findings from psychophysics and neuroimaging, the present study provides evidence for a hand preference for stimulus–response coupling using the combined measure of behavioral and neural response times. This hand preference of stimulus–response coupling was eliminated by crossing the forearms, but not by placing the hands close together or dividing attention between the hands. As already discussed, crossing the forearms may produce an incongruency between real (external) and learned (internal) spaces, which may eliminate the hand preference in the stimulus–response translation process.

In the present study, N140 amplitude was higher for deviant stimuli than for standard stimuli in all conditions, including attended and unattended stimuli. This amplitude increase was associated with somatosensory mismatch negativity (s-MMN), which has been reported to occur in the range of this latency in passive tasks (Kekoni et al., 1997; Kida et al., 2004c; Restuccia et al., 2007, 2009; Chen et al., 2014; Shen et al., 2018a; He et al., 2020). MMN is considered to reflect the automatic detection of stimulus changes in the sensory environment (Naatanen, 1992; Naatanen et al., 2007) and has recently been associated with predictive coding (Kimura, 2012; Stefanics et al., 2014). The first demonstration of MMN was in both the unattended and attended channels in an auditory selective attention task (Naatanen et al., 1978), suggesting the automatic nature of MMN. However, s-MMN was previously detected in passive tasks, such as reading (Kekoni et al., 1997; Restuccia et al., 2007, 2009) and video watching (Chen et al., 2014; Shen et al., 2018b), both of which are general MMN recording procedures. In contrast, the present study observed s-MMN in both attended and unattended deviant stimuli in a selective attention task, suggesting that the generation of the response is independent of attention. Similar attention-independent responses to stimulus onset, offset, and change have been demonstrated in both active and passive tasks (Yamashiro et al., 2008, 2009). Previous studies found s-MMN in frontal (Kekoni et al., 1997; Kida et al., 2004c) and parieto-occipital sites (Restuccia et al., 2007, 2009). We also detected s-MMN in frontal and central sites, supporting previous ERP findings. Placing the hands close together and crossing the forearms had no impact on the higher amplitude for deviant stimuli than for standard stimuli, thereby supporting the primarily automatic nature of MMN.

Since we recorded ERPs from a limited number of electrodes, we were unable to perform a source-level analysis or current source density analysis. However, N140 and P300 have both been extensively examined and an abundant amount of information has been obtained on their generators (Allison et al., 1992; Tarkka et al., 1996; Valeriani et al., 2000, 2001; Tanaka et al., 2008). Therefore, it is possible to speculate about the brain regions involved in the modulation of ERP components. Furthermore, we did not perform the divided attention condition when the hands were closely located or when the forearms were crossed. These conditions may provide insights into the interaction between different spatial coordinates in somatosensory attention. Another limitation is the small sample size; therefore, we performed a post-hoc power analysis (Supplementary Material). The results obtained showed that most of the significant differences observed in ANOVAs, MANOVAs, and t-tests remained and, thus, our suggestions are effective. However, some analyzes showed low statistical power, suggesting higher type II error in some results, which was at least partly due to the small sample size. Therefore, further studies are warranted.