1.

Introduction

Cognitive control is a key characteristic of the human cognitive system. It refers to the ability to attend to goal-relevant information, ignore distracting information, overcome conflict, and select the appropriate response. Cognitive control over conflict usually relies on conflict detection and resolution, and it has been widely studied using the Stroop task. In the classic Stroop task,¹ subjects are instructed to identify the color in which a word stimulus is printed. Usually, behavioral responses to incongruent stimuli (when the word and the ink color are incongruent, e.g., the word “blue” shown in green ink) are slower and less accurate than responses to neutral or congruent stimuli (when the word and the ink color are the same). The behavioral difference between incongruent and neutral or congruent stimuli is called the Stroop effect, which has been used as an index of cognitive control. Because the Stroop task is also useful for studying neurological/psychiatric disorders,^2,3 investigating its neural correlates can both improve our understanding of conflict processing and provide knowledge for practical uses.

A number of functional neuroimaging studies have investigated the neural correlates of the Stroop task and have identified that the prefrontal cortex (PFC) and the anterior cingulate cortex (ACC) are involved in Stroop conflict tasks.^4–7 The PFC has been proven to play a predominant role in cognitive control.^8,9 The PFC implements top-down attention control to bias the neural processing of task-related information and resolve conflict.^9–11 Regarding conflict detection, the conflict monitoring theory asserts that the ACC detects the occurrence of conflict and then sends conflict-related information to the PFC, which resolves the conflict.^6,12 However, some studies have questioned the ACC’s role in conflict detection,^9,13 and an increasing amount of research has suggested that the PFC also plays a role in conflict detection.^9,14,15

Regarding electrophysiological measures, the previous event-related potential (ERP) studies have revealed two primary ERP markers in the Stroop task: N450 and the late positive complex (LPC).^16–20 N450 is more negative in amplitude for incongruent stimuli than for neutral or congruent stimuli. It has been generally considered to be related to conflict detection and resolution,^16,21 although some studies have suggested that N450 only reflected conflict detection.^22,23 The LPC amplitude is more positive for incongruent stimuli than for neutral or congruent stimuli. The cognitive processes reflected by the LPC are even more ambiguous than those reflected by N450. Some studies have suggested that the LPC was related to the additional processing of word meaning,^16,17 while others have proposed that this component was linked to conflict resolution.^22,23 Some studies have also suggested that the LPC reflected the conflict adaptation process.^24,25

Although the previous neuroimaging studies have provided insight into the neural correlates of the Stroop task, most research has studied the hemodynamic and electrophysiological signals separately and the relationship between these two types of signals is still poorly understood.^7,26 Considering that an analysis of the relationship between hemodynamic and electrophysiological signals could inprove our understanding of human brain function,^26,27 it is necessary to study the relationship between these two types of signals to better understand the neural correlates of conflict processing.

Near-infrared spectroscopy (NIRS), a noninvasive optical functional neuroimaging method, can measure the concentration changes of oxy-hemoglobin ($\mathrm{\Delta }[{\mathrm{HbO}}_2]$) and deoxy-hemoglobin ($\mathrm{\Delta }[\mathrm{Hb}]$) to reflect relative regional brain activity. Previous studies using NIRS have successfully investigated the neural correlates of the Stroop task^28,29 and have improved our understanding of the PFC’s role in conflict processing.^5,30 In addition, NIRS can be conveniently integrated with simultaneous electroencephalography (EEG) recordings.^26,31 Thus, it is appropriate to investigate the relationship between conflict-related hemodynamic and electrophysiological signals by combining NIRS and EEG. Our previous study has combined NIRS and ERP to investigate the relationship between hemodynamic and electrophysiological signals in a Stroop task and obtained some significant results.²⁶ However, the NIRS data were collected from only four detector channels with our portable brain function imager,³² and the ERP data were recorded from only four frontopolar electrodes, which might overlook the presence of the N450 Stroop effect. Such limited measurement of the NIRS and ERP signals restricted the analysis of the relationship between hemodynamic and electrophysiological signals in conflict processing.

In this study, we aimed to examine the relationship between the conflict-related hemodynamic signal and the electrophysiological signal by combining NIRS and ERP. Since a color-word matching Stroop task primarily activates the lateral PFC,^5,13 it is more suitable for NIRS due to its low penetration depth.³¹ We employed a Chinese color-word matching Stroop task and measured the hemodynamic signal from the PFC and the electrophysiological signal simultaneously during the task. Compared with our previous study, the NIRS signal was recorded from a larger prefrontal region using our newly developed brain function imager,³³ which emits and detects light with optical fibers. The ERP signal was recorded from anterior to posterior cortical sites.

2.

Methods

2.2.

Materials

The color-word matching Stroop task was used with a block design adapted from the previous studies.¹³ Each stimulus consists of two Chinese characters, and the subjects were instructed to judge whether the color of the upper character corresponded to the meaning of the lower character [Fig. 1(a)]. If the answer was “match,” the subjects pressed a button with the index finger of their left hand; if the answer was “unmatch,” they pressed another button with the index finger of their right hand. There were two stimulus conditions: neutral and incongruent [Fig. 1(a)]. For neutral stimuli, the upper characters were noncolor words (涂, 贯, 华, 球, 奖 meaning “scrawl,” “pass through,” “China,” “ball,” and “prize”) presented in red, yellow, blue, green, or purple, and the lower characters were color words (红, 黄, 蓝, 绿, 紫 meaning “red,” “yellow,” “blue,” “green,” and “purple”) printed in white. For incongruent stimuli, the upper characters were color words shown in a disparate color. The subjects were instructed to attend to the upper character first and then make their decision. This instruction was used to prevent subjects from ignoring the upper character. For each stimulus condition, the color of the upper character was the same as the meaning of the lower character in half of the trials, whereas the color of the upper character differed from the meaning of the lower character in the other half of the trials. Moreover, the “same” and “different” trials were semi-randomly mixed within each block to avoid the consecutive appearance of more than three “same” or “different” trials.

Fig. 1

Experimental design. (a) Trial examples for the neutral and incongruent conditions. The subjects were instructed to judge whether the color of the upper character corresponded to the meaning of the lower character. The correct answer was “match” for the upper two trial examples and “unmatch” for the lower two trial examples. 华 means “China” and 蓝, 红 corresponds to “blue, red.” (b) The task sequence in each run. Rest, 1-min rest period before and after each run; B, 15-s baseline between blocks within each run; N, 30-s neutral task block; I, 30-s incongruent task block.

2.4.

NIRS Recording

The NIRS data were acquired using a continuous-wave NIRS system developed by our laboratory.³³ The system uses two wavelengths (785 and 850 nm) to determine the concentration changes of ${\mathrm{HbO}}_2$ ($\mathrm{\Delta }[{\mathrm{HbO}}_2]$) and Hb ($\mathrm{\Delta }[\mathrm{Hb}]$) based on the modified Beer–Lambert Law. The NIRS probe was supported by a flexible plastic base and held one source and eight detectors, providing eight detector channels. (The channel was defined as the midpoint of the source-detector pair.) Two such NIRS probes were used to cover the left and right PFC. The right probe was placed so that Channel 3 (Ch 3, the midpoint of the NIRS source and Detector 3) was over the F4 electrode, and the line through the source and Detector 3 was parallel to the midline (i.e., the arc from the nasion to the inion). The left probe was placed symmetrically (Fig. 2). The distance between the source and the detector was 3 cm. To estimate the measured brain region, we first determined the corresponding (i.e., the nearest) electrode in the 10–5 system³⁴ for each NIRS detector channel. Then, based on the standard electrode coordinates for the 10–5 system, we projected all corresponding electrodes to the standard Montreal Neurological Institute (MNI) space using the probabilistic estimation method.³⁵ Finally, we estimated the measured brain region according to the MNI coordinates of all these corresponding electrodes. Thus, we presumed that the NIRS probes covered part of the bilateral dorsolateral prefrontal cortex (DLPFC) and the ventrolateral prefrontal cortex (VLPFC) in this study. The time resolution was 70 Hz.

Fig. 2

Schematic of near-infrared spectroscopy (NIRS) measurement channel locations on the head. The red circle indicates the NIRS light source and the blue rectangle indicates the detector. The number (1 to 16) denotes the NIRS detector channel (midpoint of the source-detector pair). Channels 3 and 11 were over F4 and F3, respectively.

3.

Results

3.1.

Behavioral Results

Figure 3 displays the average RT and error rate for each run. Regarding RT, the repeated-measures ANOVA revealed a significant Condition effect [$F(1,14)=41.133$, $p<0.001$, ${\eta }_p^2=0.746$], suggesting a longer RT for the incongruent condition compared with the neutral condition. There was a significant Run effect [$F(1.964,27.493)=23.62$, $p<0.001$, ${\eta }_p^2=0.628$]. Follow-up multiple comparisons showed that the RT order was $\text{Run}1>\text{Run}2>\text{Run}3\approx \text{Run}4$. Regarding the error rate, the repeated-measures ANOVA disclosed a significant Condition effect [$F(1,14)=42.857$, $p<0.001$, ${\eta }_p^2=0.754$], indicating a higher error rate for the incongruent condition. The ANOVA also revealed a significant Run effect [$F(3,42)=8.567$, $p<0.001$, ${\eta }_p^2=0.38$]. Follow-up multiple comparisons showed that the error rate was higher for Run 1 than for Run 2 and Run 4. Although no significant interaction was observed for either RT or error rate ($p>0.05$), a paired t-test was used to determine whether the Stroop effect was significant for the RT and error rate in each run. The results showed a significant Stroop effect in each run (RT: $t>4.84$, $p<0.001$; error rate: $t\ge 3.062$, $p\le 0.008$), indicating that the Stroop interference effect remained significant across the entire experiment.

Fig. 3

Response time and error rate for each run. The data are shown as $\text{mean}\pm \text{standard error}$ (SE).

3.2.

NIRS Results

Figure 4(a) shows the grand average time courses of the ${\mathrm{HbO}}_2$ responses to neutral and incongruent tasks at one typical channel (Ch 9). The statistical analyses revealed significant Stroop effects in all eight channels for the left PFC (Ch 9 to 16) and in three channels for the right PFC (Ch 2/6/7; $p<0.05$, FDR correction), indicating greater ${\mathrm{HbO}}_2$ concentration increase for the incongruent task than for the neutral task. Figures 4(b) and 4(c) show the maps of $t$-values from the Stroop effect tests for the left and right PFC, respectively. According to the effect size analyses, the Stroop effects were larger in the left PFC than in the right PFC, with the largest Stroop effect at Ch 9 (Cohen’s $dz=1.667$).

Fig. 4

(a) The grand average time courses of ${\mathrm{HbO}}_2$ responses to neutral (blue) and incongruent (red) tasks at typical Channel 9 (Ch 9). The gray region indicates SE. The black lines at 0 and 30 s indicate the start and end of the task blocks. (b) and (c) show the $t$-values from the Stroop effect tests for the left and right PFC, respectively.

3.3.

ERP Results

Figure 5 shows the average ERP waveforms for neutral and incongruent stimuli at the six midline electrodes.

Fig. 5

The grand average of the event-related potential (ERP) waveforms for neutral (blue) and incongruent (red) stimuli at the six midline electrodes. The gray bars at Pz indicate the time windows used in the analysis of variance for N450 (left) and late positive complex (right).

The mass univariate analyses revealed significant differences between incongruent and neutral stimuli during two distinct time windows (422 to 680 and 870 to 1000 ms) in the six regions. The ERP elicited by incongruent stimuli was more negative during the 422 to 680-ms period and more positive during the 870 to 1000-ms period compared with the neutral stimuli.

3.3.1.

N450 (450 to 650 ms)

The ANOVA revealed a significant Condition effect [$F(1,14)=16.001$, $p=0.001$, ${\eta }_p^2=0.533$], indicating a more negative N450 response for the incongruent condition. The $\text{Condition}\times \text{Region interaction}$ was evident [$F(1.446,20.248)=3.963$, $p=0.047$, ${\eta }_p^2=0.221$]. Further simple effect analyses showed that the Stroop effect was significant at each region (Table 1). No other interactions involving Condition were significant ($p>0.05$).

Table 1

The statistical results of the N450 Stroop effect at each region.

N450	$F$	$P$	${\eta }_p^2$
Frontal	7.41	0.017	0.336
Frontocentral	8	0.013	0.364
Central	10.71	0.006	0.434
Centroparietal	18.2	0.001	0.565
Parietal	21.67	$<0.001$	0.607
Occipitoparietal	12.49	0.003	0.471

3.3.2.

LPC (880 to 1000 ms)

The ANOVA disclosed an evident Condition effect [$F(1,14)=10.609$, $p=0.006$, ${\eta }_p^2=0.431$], indicating a more positive LPC response for the incongruent condition. The $\text{Condition}\times \text{Laterality interaction}$ was significant [$F(2,28)=3.751$, $p=0.036$, ${\eta }_p^2=0.211$]. Simple effect analyses revealed that the Condition effect was significant for the left, middle, and right cortical sites ($p<0.05$), with the largest Condition effect at the left cortical sites indicated by the largest effect size (left: ${\eta }_p^2=0.474$; middle: ${\eta }_p^2=0.400$; right: ${\eta }_p^2=0.370$). No other interactions involving Condition were significant ($p>0.05$).

3.4.

Correlation Between HbO₂ and N450

Since it is difficult to determine the neural generator source location from the ERP scalp voltage distribution, the largest N450 Stroop effect was correlated to the ${\mathrm{HbO}}_2$ Stroop effect. The larger effect size (${\eta }_p^2$) indicated that the N450 Stroop effects were greater in centroparietal and parietal regions than in the other four regions. Thus, the N450 in the centroparietal and parietal regions was used to correlate with ${\mathrm{HbO}}_2$. The N450 Stroop effect in the centroparietal (parietal) region was obtained by averaging the data from the three electrodes in that region, and then correlated with the significant ${\mathrm{HbO}}_2$ Stroop effect (left PFC: Ch 9 to 16; right PFC: Ch 2/6/7).

The N450 Stroop effect in the centroparietal region was significantly negatively correlated with the ${\mathrm{HbO}}_2$ Stroop effects in the left PFC (Ch 9/10/11/16; r ranged from $-0.541$ to $-0.654$, $p<0.05$). The N450 Stroop effect in the parietal region was significantly negatively correlated with the ${\mathrm{HbO}}_2$ Stroop effects in the left (Ch 9; $r=-0.537$, $p<0.05$) and right PFC (Ch 2; $r=-0.523$, $p<0.05$) (Table 2). Due to the negativity of N450, this negative correlation indicated that the larger the difference in the N450 amplitude between these two stimulus conditions, the greater the difference in the ${\mathrm{HbO}}_2$ signal. To explore whether the ${\mathrm{HbO}}_2$ responses in the left and right PFC were correlated with the N450 at different time ranges, the correlation between the ${\mathrm{HbO}}_2$ Stoop effect and the ERP Stroop effect for each 20-ms bin from 400 to 700 ms after the stimulus onset was further examined. These analyses revealed that the ${\mathrm{HbO}}_2$ Stroop effect in the left PFC (Ch 9/10/16) was negatively correlated with the N450 Stroop effect during the early phase (from 440 to 580 ms; Fig. 6), and the ${\mathrm{HbO}}_2$ Stroop effect in the left (Ch 11) and right PFC (Ch 2) was negatively correlated with the N450 Stroop effect during the later phase (from 600 to 680 ms; Fig. 6).

Table 2

Correlation between HbO2 Stroop effects and N450 Stroop effects in the centroparietal and parietal regions.

	Centroparietal	Parietal
Ch 2	$-0.48$b	$-0.523$a
Ch 6	$-0.148$	$-0.241$
Ch 7	$-0.38$	$-0.448$b
Ch 9	$-0.654$a	$-0.537$a
Ch 10	$-0.592$a	$-0.449$b
Ch 11	$-0.541$a	$-0.41$
Ch 12	0.131	0.221
Ch 13	$-0.166$	$-0.14$
Ch 14	$-0.113$	$-0.144$
Ch 15	$-0.174$	$-0.206$
Ch 16	$-0.57$a	$-0.453$b

^aSignificant correlation (p<0.05).

^bMarginally significant correlation (p<0.1).

Fig. 6

(a) Correlation coefficients between ERP Stroop effects for each 20-ms bin within 400 to 700 ms at the centroparietal region and the ${\mathrm{HbO}}_2$ Stroop effects at Ch 9 (left top), Ch 10 (left middle), and Ch 11 (left bottom). (b) Correlation coefficients between ERP Stroop effects for each 20-ms bin within 400 to 700 ms at the parietal region and the ${\mathrm{HbO}}_2$ Stroop effects at Ch 2 (right top), Ch 7 (right middle), and Ch 10 (right bottom). The red rectangle indicates the time window during which there was a significant correlation between these two Stroop effects ($p<0.05$).

In addition, the N450 Stroop effect in the centroparietal region was marginally significantly negatively correlated with the ${\mathrm{HbO}}_2$ Stroop effect in the right PFC (Ch 2; $r=-0.48$, $p<0.1$). There were also some marginally significant negative correlations between the N450 Stroop effect in the parietal region and the ${\mathrm{HbO}}_2$ Stroop effects in the left (Ch 10: $r=-0.449$, $p<0.1$; Ch 16: $r=-0.453$, $p<0.1$) and right PFC (Ch 7; $r=-0.448$, $p<0.1$; Table 2). Further analyses of the marginally significant correlation between ${\mathrm{HbO}}_2$ and N450 also proved that ${\mathrm{HbO}}_2$ in the left PFC (Ch 10/16) was correlated with the early phase of N450 (Fig. 6), and ${\mathrm{HbO}}_2$ in the right PFC (Ch 2/7) was correlated with the later phase of N450 (Fig. 6).

4.

Discussion

In this study, we examined the relationship between the conflict-related hemodynamic signal in the PFC and the electrophysiological signal during a Chinese color-word matching Stroop task. There were significant Stroop effects in behavioral (RT, error rate), NIRS (${\mathrm{HbO}}_2$), and ERP (N450, LPC) data, consistent with the previous studies.^5,16,17 Furthermore, the ${\mathrm{HbO}}_2$ Stroop effect was negatively correlated with the N450 Stroop effect. More specifically, the ${\mathrm{HbO}}_2$ Stroop effect in the left PFC (Ch 9/10/16) was negatively correlated with the N450 Stroop effect during the early phase (440 to 580 ms), while the ${\mathrm{HbO}}_2$ Stroop effect in the bilateral PFC (left: Ch 11; right: Ch 2/7) was negatively correlated with the N450 Stroop effect during the later phase (600 to 680 ms).

The color-word matching Stroop task was used because this task mainly activates the lateral PFC.^5,13 In the present study, the ${\mathrm{HbO}}_2$ response in the bilateral PFC was stronger for the incongruent stimulus task than for the neutral stimulus task, which is consistent with the previous studies reporting the involvement of the bilateral PFC in the Stroop task.^5,40,41

The N450 amplitude was more negative and the LPC amplitude was more positive for incongruent stimuli compared with the neutral stimuli, which was consistent with the previous reports.^16,17,21 The significant N450 Stroop effect lasted from approximately 420 to 680 ms after the stimulus onset, and the significant LPC Stroop effect began at approximately 870 ms. The start of the evident LPC Stroop effect was somewhat earlier than the behavioral response to incongruent stimuli in Run 1, but later than the behavioral response to incongruent stimuli in Runs 2 to 4. LPC has been associated with conflict resolution,^22,23 conflict adaption,^24,25 or the re-activation of word meaning.^16,17 Because conflict needs to be resolved before a behavioral response occurs,^20,38 the significant LPC Stroop effect occurring after the behavioral response indicates that the LPC could not be associated with conflict resolution before a manual response. Additionally, the conflict adaption might be greatly diminished under the conditions of our experimental design.⁴² Taken together, we suggest that the LPC could be related with word meaning processing.^21,22 Combined with the previous studies, we infer that N450 is related to conflict processing.^16,21 In addition, N450 Stroop effects during the early and later phases were negatively correlated with ${\mathrm{HbO}}_2$ Stroop effects in the left PFC and the bilateral PFC, respectively. This dissociation of the early and later phases of N450 suggests that there are two stages in N450. Our study supports that N450 reflects both conflict detection and resolution.^16,21

The negative correlation between N450 and ${\mathrm{HbO}}_2$ indicates that the larger the difference in the N450 between these two stimulus conditions, the greater the difference in the ${\mathrm{HbO}}_2$ signal. Previous animal studies have suggested a tight coupling between the hemodynamic response and the local field potential (LFP).^43,44 The LFP is known to be the basis of scalp EEG/ERP,⁴⁵ implying that the spatiotemporal data integration can be obtained by examining the correlation between the hemodynamic response and the scalp EEG/ERP in most situations. In addition, human studies have also shown that these two types of measures were correlated during brain function.^26,39 Our results fit well with the previous studies that show a coupling between these two measures. The correlation suggests that N450 and activity in certain PFC regions reflect related processing.

The ${\mathrm{HbO}}_2$ Stroop effect in the left PFC (Ch 9/10/16, located approximately in the left DLPFC) was negatively correlated with the N450 Stroop effect during the early phase. The early phase of N450 corresponded with conflict detection. Although some studies have suggested that the ACC detected conflict,^6,12 others have suggested that the ACC was associated with response-related processing, but not conflict-related processing.^13,46 Given that the ACC activity was not observed in the color-word matching Stroop task,¹³ this correlation indicated that the left PFC might be involved in conflict detection. Considering the bidirectional connection between the DLPFC and posterior brain regions,⁴⁷ and the involvement of the left middle frontal gyrus in Chinese semantic processing,⁴⁸ it is reasonable to conclude that the left DLPFC plays a role in detecting conflict. Furthermore, there is increasing evidence supporting the assumption that the DLPFC can detect conflict. Milham et al. found that practice had different influences on the ACC and the left DLPFC.⁸ While the ACC activity existed only during the first two cycles, the left DLPFC activity persisted even until the last two cycles (Cycles 5 to 6). Milham et al. suggested that the DLPFC could detect the nonresponse conflict independent of the ACC.⁸ Morishima et al. reported that the left DLPFC activity covaried with the neutral activity of sensory conflict for incongruent trials in a face-word Stroop task, suggesting the left PFC’s role in detecting conflict.¹⁴ Some lesion studies in patients and animals also suggested that the ACC was not necessarily involved in conflict processing,^15,49 while DLPFC was involved in conflict detection.^9,15,50 Given our findings and those of the previous studies, we suggest that the left PFC is involved in conflict detection.

The ${\mathrm{HbO}}_2$ Stroop effect in the bilateral PFC was negatively correlated with the N450 Stroop effect during the later phase. This correlation supports the generally accepted viewpoint that the PFC is involved in conflict resolution. The PFC exerts the cognitive control to resolve conflict by favoring the processing of task-relevant information and/or inhibiting the processing of task-irrelevant information.^51,52 Participation of both the left and right PFC in conflict resolution has been proven in many previous studies.^11,40,41

NIRS measures the hemodynamic response and indirectly reflect neural activity at a reasonable spatial resolution, while ERP measures the electrical response and directly reflect neural activity at a high temporal resolution. Our study shows that the analysis of the relationship between these two complementary responses can not only provide information about the processing stage in which one brain region is involved, but also distinguish the different processing stages in the ERP response with the help of brain activity.

This study examined the relationship between hemodynamic and electrophysiological signals to investigate the neural correlates of conflict processing in a Chinese color-word matching Stroop task. The significant LPC Stroop effect occurring after the behavioral response to incongruent stimuli suggests that the LPC is associated with the re-activation of word meaning. The evident N450 Stroop effect occurred before the behavioral response to incongruent stimuli. Besides, the N450 Stroop effects during the early and later phases were negatively correlated with ${\mathrm{HbO}}_2$ Stroop effects in the left PFC and in the bilateral PFC, respectively. These results indicate the following: (1) there are two stages in N450: conflict detection and resolution; (2) the left PFC may be involved in conflict detection and the bilateral PFC is engaged in conflict resolution. Our study promotes the understanding of the neural correlates of conflict processing and shows that an analysis of the correlation between hemodynamic and electrophysiological measures can be an effective method for studying human brain function.