The Performance Dip is a newly characterized behavioral phenomenon, where, paradoxically, a weaker task-irrelevant visual stimulus causes larger disturbances on the accuracy of a main letter identification task than a stronger stimulus does. Understanding mechanisms of the Performance Dip may provide insight into unconsciousness behavior. Here, we investigated the generalization of the Performance Dip. Specifically, we tested whether the Performance Dip occurs in a motion-related Simon task, and if so, whether the Performance Dip involves the same brain region, that is, the dorsolateral prefrontal cortex (DLPFC), previously implicated in the Performance Dip, or the supplementary motor area (SMA) and pre-SMA, implicated in a motion-related Simon Task. Subjects made manual directional responses according to the color of stochastic moving dots while ignoring the global direction of moving dots, which could be either congruent or incongruent to the response appropriate to the main task. We found that weak incongruent task-irrelevant stimuli caused a Performance Dip, in which the SMA and pre-SMA, rather than DLPFC, played critical roles. Our results suggest a possible common brain mechanism across different neural circuits, in which weak, but not strong, task-irrelevant information is free from inhibition and intrudes into neural circuits relevant to the main task.
Human performance is influenced by various factors including task-relevant and task-irrelevant information. We previously reported that task-irrelevant visual motion stimuli more greatly disrupted accuracy for a rapid serial visual presentation (RSVP) task when the motion coherence was low than when the motion coherence was larger or zero. Based on this, we defined the Performance Dip as a lowered accuracy of a main task with presentation of a parathreshold task-irrelevant stimulus. We also found that the dorsolateral prefrontal cortex (DLPFC) accounts for the Performance Dip, in that the DLPFC failed to detect and therefore to suppress very weak coherent motion signals. A similar Performance Dip has also been reported with low coherence task-irrelevant motion when the central task required lexical decision on moving words (Meteyard et al. 2008). These studies identified important aspects of how the brain suppresses task-irrelevant visual stimuli that could distract an observer from performing on a task-relevant visual feature and/or cause a conflict with the task-relevant signals, particularly when the available attentional resource is limited (Miller and Cohen 2001).
However, the underlying neural mechanisms for the Performance Dip, and its generality to other tasks, is largely unknown. The previous studies investigated the Performance Dip in tasks where the interfering task-irrelevant visual motion stimuli (Tsushima et al. 2006; Meteyard et al. 2008) were largely distracting to discriminating the identity of linguistic stimuli such as letters or words. This may suggest the possibility that the Performance Dip is a particular phenomenon confined to letter or words. To test whether the Performance Dip is also generated in the directional ‘motor’ response presented with task-irrelevant visual motion stimulus, here we used a modified version of the Simon effect paradigm (Simon and Small 1969), which utilizes random dots (Bosbach et al. 2004, 2005; Wittfoth et al. 2006). We asked the subjects to make a manual directional tilt in either the left or right direction by the lever depending on the color of the moving dots. At the same time, we manipulated the coherence of the random dots so that the perceived global direction of the dots was either congruent or incongruent to the required tilt direction, but task irrelevant.
A question arises as to whether the DLPFC is the unitary system that involves a Performance Dip. In Experiment 1, we tested whether a Performance Dip occurs in a directional motor task and confirmed that it did when the task-irrelevant coherent motion was incongruent to the main task. In Experiment 2, we conducted an functional magnetic resonance imaging (fMRI) analysis to test which of the following two models, the DLPFC model or the pre-SMA/SMA model best accounts for the Performance Dip. The DLPFC model is predicted by the previous Performance Dip study (Tsushima et al. 2006), where the DLPFC plays a key role in the inhibitory modulation of inputs to the visual cortex (Knight et al. 1999; Tsushima et al. 2006; Rossi et al. 2009). In contrast, several other frontal regions including bilateral anterior insula (Wager et al. 2005), the right ventral lateral prefrontal cortex (VLPFC) (Forstmann et al. 2008), and anterior cingulated cortex (ACC) (Lau et al. 2006; Fan et al. 2008) have been regarded as candidates to process response conflicts stemming from stationary visual signals and motor response. However, in the present study, the conflicting information is between the directions of coherent motion as a task-irrelevant feature and the direction of the lever tilt response appropriate to the color cue. It has been reported that pre-SMA, but not the DLPFC, is involved in a conflict between the directions of moving dots and motor response (Wittfoth et al., 2006). Thus, as opposed to the DLPFC model, we built the pre-SMA/SMA model, where the SMA and pre-SMA, which usually suppress motor plans (Brass and von Cramon 2002; Crone et al. 2006; Isoda and Hikosaka 2007; Sumner et al. 2007; Imamizu and Kawato 2008), fail to suppress weak task-irrelevant signals that lead to conflicting motor plans. In this model, the failure of pre-SMA and/or SMA to suppress task-irrelevant information causes a Performance Dip.
We describe the predictions of the two models below after the brief summary of the previous study. Suppose there were three levels of coherence for the task-irrelevant visual motion: zero (totally random), very weak, and very strong, and the Performance Dip occurred when the coherence of the task-irrelevant visual motion was very weak. The results of the previous study (Tsushima et al. 2006) indicated two things. First, activation of middle temporal areas (MT+) did not increase monotonically as a function of the coherence of the task-irrelevant motion signal. MT+ was most highly activated when task-irrelevant motion coherence was very weak, when the Performance Dip occurred, and activation of MT+ was lower for weaker and stronger coherent motion signals, for which performance was higher. Second, no significant difference was found between blood oxygen level–dependent (BOLD) signals in the DLPFC between zero coherence and low coherent motion stimuli, but DLPFC was significantly activated for strong motion signals. The previous study (Tsushima et al. 2006) suggested that the weak coherent motion signal was below the threshold of perception, resulting in a failure of the DLPFC to detect and inhibit the MT+ response to this distracting visual signal.
If the Performance Dip in the present study is subserved by the same mechanism as in the previous study (Tsushima et al. 2006), that is, if the DLPFC hypothesis is true, then it is predicted that activation in MT+ will be suppressed for strong motion signals relative to weak coherent motion signals and that DLPFC will be significantly activated for strong motion signals compared with weak or zero motion signals.
In contrast, the pre-SMA/SMA hypothesis provides a different prediction. This model predicts that the dip will be associated with the low activation of the pre-SMA/SMA. In addition, since there are no known direct connections between MT and pre-SMA/SMA (Desimone and Ungerleider 1986; Ungerleider and Desimone 1986; Luppino et al. 1993), and the pre-SMA and SMA have not been reported to send inhibitory modulations to the visual cortex, the activation of MT+ may not be suppressed in our task, in contrast to the previous study (Tsushima et al. 2006). Rather, MT+ activation in our task will increase monotonically as a function of the coherence of the task-irrelevant motion signal.
Thus, in Experiment 2, we conducted an fMRI experiment to measure the brain activation in the four specified regions: MT+, DLPFC, pre-SMA, and SMA, to test which of the aforementioned hypotheses is more likely to account for the Performance Dip.
Materials and Methods
A total of 28 subjects participated (12 females, age range: 18–35 years) in the two experiments: 13 subjects (8 females, age range: 18–26 years) and 15 subjects (4 females, age range: 24–35 years) participated in Experiments 1 and 2, respectively. The number of right-handed subjects was 10 in Experiment 1 and 13 in Experiment 2. All the subjects gave written informed consent. The present study was approved by the Institutional Review Boards at Massachusetts General Hospital, Boston University, and Keio University.
We utilized a modified Simon effect paradigm (Simon and Small 1969) to manipulate the strength of task-irrelevant visual stimuli. While the classical Simon effect paradigm requires subjects to choose a right or left-side button to indicate the color of a stimulus, where the button’s location is irrelevant to the task, we asked subjects to respond to the color of moving dots by tilting a lever while ignoring the dots’ motion direction. It has been reported that the Simon effect occurs with this type of manipulation (Bosbach et al. 2004, 2005; Sheremata and Sakagami 2006; Wittfoth et al. 2006).
Figure 1 illustrates stimulus presentation and timing for manual responses in Experiment 1. A stimulus consisted of randomly moving and coherently (in the same direction and speed) moving dots from one frame to another. The visual stimulus had 3 key visual features: The color of the moving dots, the direction (rightward or leftward) of coherent motion, and the level (strength) of the coherent motion. Only the color was task relevant. The direction and level of coherent motion was task irrelevant. The direction of coherent motion (rightward or leftward) was congruent with the direction of a directional motor response (lever tilting to the right or left) in 50% of the trials and was incongruent in the remaining 50% of the trials.
The level of coherence (0%, 7%, 15%, 20%, 25%, 30%, 50%, 75%, and 100%) was varied in a random order from trial to trial. The higher the level of motion coherence, the more salient the perceived global direction becomes (Britten et al. 1992, 1993, 1996). A motion display consisted of approximately 200 moving dots (speed 12 degree/s) presented within an aperture with 28 degrees diameter for 500 ms. A one-degree diameter fixation disk was presented throughout the trial. Moving dots were not presented within one degree of the fixation disk. The direction of coherent motion (leftward vs. rightward) was varied from trial to trial in a randomized order.
Prior to the experiments, we measured each subject’s isoluminant level of blue and green colors (Cavanagh et al. 1984) and used the level for the experiments.
Each trial started when subjects set the joystick lever to the upright position. 200 ms after the onset of the initial fixation point, the motion stimulus came on with white dots for 250 ms. Then the color changed either to blue or to green for 150 ms (Fig. 1) and then turned back to white for the last 50 ms of the presentation. Subjects were instructed to tilt the joystick (Logitec 3D Pro Joystick), using their dominant hand, to the right if the color of the dots was green and to the left if the color was blue, while ignoring the coherent motion direction. The tilt threshold, for recording subjects’ responses, was 10 degrees. When blue dots moved leftward or when green dots moved rightward, the required motor response was congruent with the coherent motion direction (congruent condition). On the other hand, when blue dots moved rightward or when green dots moved leftward, the required motor response was incongruent with the coherent motion direction (incongruent condition). Subjects received no feedback as to the correctness of their responses.
The total number of trials was 1440 with 18 conditions (congruence/incongruent × 9 coherence levels × 80 repetitions). Note that one-third of the 1440 trials were randomly assigned to catch trials (no-go trials): the subject was instructed not to respond if the fixation disk turned red at the same time as the onset of the moving dots (otherwise the color of the fixation dot was black). The purpose of these catch trails was to test how well the subject fixated. If the subject mistakenly responded in a catch trial, a warning beep was provided.
Experiment 2 used the same stimuli as in Experiment 1, but only 5 conditions: 1) 0% motion coherence condition, 2) congruent low (25%) motion coherence condition, 3) congruent high (75% or 100%) motion coherence condition, 4) incongruent low (25%) motion coherence condition, and 5) incongruent high (75% or 100%) motion coherence condition. We used an event-related fMRI paradigm to present the 5 conditions with randomized interstimulus intervals for maximized statistical efficiency (Dale 1999). Given the limited scan time, we presented a smaller number of conditions in this fMRI experiment than in Experiment 1 so that we could obtain a sufficient number of repetitions for each condition; 75% was used as a high coherence for 9 subjects, and 100% was used for 6 subjects, while 25% was used as a low coherence for all subjects. There were no catch trials.
Stimuli were generated on a Macintosh G4 computer and presented via LCD projector. The stimulus configuration and timing were similar to those in Experiment 1 except for the following deviations. The subject’s responses were measured with a joystick, which was fMRI compatible. Responses had to be completed within 1550 ms of the offset of the moving dots. The visual stimulus subtended 22 degrees in visual angle in diameter.
Each fMRI session consisted of at least 9 experimental runs (up to 15 runs for some subjects). Each run lasted 330 s and consisted of 25 trials × 5 conditions. Each trial lasted 2 s. Due to instrumental restriction, subjects’ responses were registered when the joystick was tilted more than 20 degrees either leftward of rightward from the starting position.
We used a 3T MR scanner (Allegra, Siemens) at Martinos Center, Massachusetts General Hospital, and a 3T MR scanner (Trio-Tim, Siemens) at Keio University. A head coil was used throughout the experiments. Three anatomical T1-weighted MR images (magnetization prepared rapid gradient echo) were acquired for each subject (time repetition [TR] = 2.351 s, time echo [TE] = 3.28 ms, flip angle = 7°, time inversion = 1100 ms, 256 slices, voxel size = 1.3 × 1.3 × 1.0 mm3) for further reconstruction of the cortical surface (Dale et al. 1999; Fischl et al. 1999) to identify brain regions and conduct subsequent data analysis.
Functional MR images were acquired using gradient echo EPI sequences (TR = 2 s, TE = 30 ms, flip angle = 90°) for measurement of BOLD signal. Thirty-five contiguous slices (3 × 3 × 3.5 mm3) oriented parallel to the AC-PC plane were acquired to cover the entire brain. All functional data were registered to the individual anatomically reconstructed brain (Dale et al. 1999; Fischl et al. 1999).
We selected the following 4 regions of interest (ROIs) for the main analysis; MT+, SMA, pre-SMA, and DLPFC. We defined these ROIs for each subject by functional or anatomical criteria, as was done in the previous study (Tsushima et al. 2006). First, the areas of bilateral MT+ were localized functionally in a separate fMRI session with low contrast moving versus static stimuli (Tootell et al. 1995). The bilateral MT+ was included in the analysis. The other ROIs were defined anatomically as follows. Using an individual automated anatomical parcellation method (Fischl et al. 2004), DLPFC was defined as the posterior half of the middle frontal gyrus, contralateral to the dominant hand. The pre-SMA and SMA were defined anatomically according to previous papers (Hikosaka et al. 1996; Crosson et al. 1999; Picard and Strick 2001): We first set three planes perpendicular to the AC-PC line, one through the rostral-most point of the genu of the corpus callosum (vgcc), one through the posterior margin of the anterior commissure (vac), and one through the posterior commissure (vpc). The bilateral medial part between vgcc and vac was defined as pre-SMA (Liu et al. 2002), and the part between vac and vpc, contralateral to the dominant hand, was defined as SMA (Lehericy et al. 2004).
Data were analyzed with FS-FAST and FreeSurfer software (http://surfer.nmr.mgh.harvard.edu). All functional images were motion corrected (Cox and Jesmanowicz 1999), spatially smoothed with a Gaussian kernel of 5.0 mm (full-width half-maximum), and normalized individually across scans. A finite impulse response model (Burock and Dale 2000) was used to estimate the hemodynamic response functions. The fMRI time course for each ROI was computed to percent signal changes, at 20 1-s interval time points, by averaging all voxels within each ROI. The percent signal change in each ROI was calculated by using peak value (4–6 s) of hemodynamic response. All trials including incorrectly responded trials were included in the analysis.
Our main aim was to test whether the Performance Dip is observed in the modified Simon task and to examine how performance on go-trials was affected by the congruence and signal strength of the task-irrelevant motion stimuli. Subjects successfully withheld responses in 89 ± 7% (SD) of no-go trails. These results suggest that subjects fixated well at the central disk and responded appropriately to the color of the fixation point. Among go-trials, we calculated accuracy (proportion of correct trials) separately for each coherent motion level in each of the congruent and incongruent conditions (see Fig. 2A). While the direction of dot motion was task irrelevant, its congruency with the correct motor responses impacted task performance. The results of a 2-way repeated measures ANOVA (congruency × coherence) indicated significant main effects of both factors (congruence, F(1,12) = 20.04, P < 0.001; coherence, F(7,84) = 2.41, P = 0.027) with a significant interaction between the factors (F(7,84) = 3.64, P = 0.002). Note that we did not include accuracy with 0% coherence in this analysis because there is no congruence at 0% coherence.
Since the interaction term in the above 2-way ANOVA was significant, we next conducted one-way repeated measures ANOVA on the difference of accuracies between the congruent and incongruent conditions with a factor being coherence (Fig. 2B), now including 0% conditions. The result of this analysis revealed a significant main effect of coherence (F(8,26) = 4.26, P < 0.001). Importantly, further post hoc analysis revealed the significant difference between 0% versus 25% (t(12) = 3.62, P < 0.05, with Bonferroni correction) and 0% vs 30% (t(12) = 3.46, P < 0.05, with Bonferroni correction). The significant differences were caused by the lower accuracy of the task when the coherence was about 25–30% in the incongruent condition, indicating the Performance Dip. Additionally, a significant difference was found in 0% versus 100% conditions (t(12) = 3.43, P < 0.05, with Bonferroni correction). No significant difference was found in any other pairs.
Reaction time often reflects interference from task-irrelevant stimuli (Bosbach et al. 2004, 2005). In the present study, reaction time was significantly longer in the incongruent condition than in the congruent condition (a 2-way ANOVA with factors of congruence and coherence, a significant effect of congruence (F(1,12) = 7.24, P = 0.020)), indicating effects of congruency on motor responses (Fig. 2C). These results show that the Performance Dip did not involve a speed-accuracy trade-off. However, the differences in reaction times between congruent and incongruent conditions were relatively small, and we found no correlate with the accuracy Performance Dip in the reaction time data. Based upon these observations, we used accuracy as our dependent measure of the Performance Dip in Experiment 2.
The results of Experiment 1 thus confirmed that the Performance Dip on a directional motor task was induced by the presence of weak incongruent task-irrelevant coherent motion.
Figure 3 shows the average accuracy for the motor responses recorded in the fMRI session. The Performance Dip in Experiment 2 visibly matches that of Experiment 1, although the overall accuracies are slightly higher in Experiment 2 (Fig. 3). The results of Experiment 1 clearly demonstrated that the Performance Dip occurs in the directional motor task. Thus, in Experiment 2, where we tested whether the Performance Dip also occurs with the same task but in the fMRI environment, we ran statistical tests on the following specific hypothesis: that accuracy for the incongruent 25% condition is lower than 1) the 0% condition and 2) the incongruent high coherence conditions such as the 75% and 100% conditions. Those tests in Experiment 2 were done by one-tailed t-tests because the Performance Dip is expected to occur in Experiment 2. As there was no significant difference behaviorally between the group of 75% coherence and of 100% coherence (P = 0.81), their performance with 75% and 100% coherence was lumped and tested as high coherence condition. The results confirmed that the above 2 predictions were the case. The accuracy in the incongruent low (25%) coherence condition was significantly lower than in the 0% condition (t(14) = 3.1, P < 0.05, with Bonferroni correction), and the incongruent high coherence condition (t(14) = 2.14, P < 0.05, with Bonferroni correction). Thus, Experiment 2 replicated the Performance Dip that was found in Experiment 1.
Next, we tested which of DLPFC or pre-SMA/SMA hypothesis is more likely by examining brain activations in the 4 ROIs (MT+, pre-SMA, SMA, and DLPFC). The aforementioned two hypotheses predict different activation patterns in the 4 ROIs. If the DLPFC hypothesis holds, then the following patterns of brain activation are predicted. First, the activity in MT+ with low (25%) incongruent coherent motion should be larger than 0% and high (75 or 100%) incongruent coherent motions. Second, the activity in the DLPFC with low (25%) incongruent coherent motion, where the Performance Dip was found in this study, should not be different from with the 0% coherent motion but should be smaller than with the high (75 or 100%) incongruent coherent motion. Third, the pre-SMA and SMA should be independent of the coherence level of the incongruent condition.
On the other hand, if the pre-SMA/SMA hypothesis holds, the following three things are predicted. First, the activity in MT+ should monotonically increase as the coherence of the incongruent motion is increased as has been found in the study in which coherent motion was task relevant (Rees et al. 2000). Second, the activity of the DLPFC should be independent of the coherence level of the incongruent motion. Third, the pre-SMA/SMA activity with the low (25%) incongruent coherent motion should not be different from that with the 0% motion coherence but should be smaller than with the high (75 or 100%) incongruent coherent motion.
For each of ROIs, we first tested whether there was any activation related to the present task. We tested whether each ROI showed enhanced activation during the task compared with the no-task condition in which only a fixation point was presented at the center of the display (baseline condition). If this test showed a significant activation, then we combined the groups of 75% coherence and 100% coherence as high coherence conditions, after confirming there was no significant difference between the degrees of fMRI activation in the 75% and 100% conditions (see below). We then tested whether the activation of each ROI was significantly different in the following 2 comparisons by 2-tailed t-tests: 1) 0% versus incongruent 25% condition and 2) incongruent 25% versus incongruent high because those 2 comparisons were sufficient to determine which of the DLPFC and pre-SMA/SMA hypotheses is more plausible.
The overall activation of MT+ during the task was significantly higher than in the baseline condition (t(14) = 11.69, P < 0.001). Since there was no significant difference between the degree of activation in the 75% and 100% conditions ((F(1,13) = 0.007, P = 0.933), repeated-measures ANOVA with 2 within factors (congruence and coherence) and 1 between factor [condition]), we combined the data from the 75% and 100% conditions as a high coherence group. The MT+ activation linearly increased as a function of coherence (shown in Fig. 4A), which was in accord with the pre-SMA/SMA hypothesis. We found significant differences both 1) between the 0% and 25% incongruent conditions (t(14) = 3.44, P < 0.05, with Bonferroni correction) and 2) between the 25% and high incongruent conditions (t(14) = 6.47, P < 0.05, with Bonferroni correction).
The overall activation of both the pre-SMA and SMA were significantly higher than in the baseline condition (t(14) = 17.43, P < 0.001). Since there was no significant difference between the degrees of the fMRI activation in the 75% and 100% conditions ((F(1,13) = 0.18, P = 0.680), repeated-measures ANOVA with 2 within factors (congruence and coherence) and 1 between factor (condition)), we combined the data from 75% and 100% conditions as a high coherence group. The multiple t-tests revealed that the activation patterns of pre-SMA and SMA correspond to the patterns predicted by the pre-SMA/SMA hypothesis (Fig. 4B). While the DLPFC hypothesis predicts that the pre-SMA/SMA activation is independent from the coherence level, the pre-SMA/SMA model predicts greater pre-SMA/SMA activation in the incongruent high than in the incongruent 25% condition and no difference between the activation in the 0% and incongruent 25% conditions. The activation of the pre-SMA/SMA was significantly higher in the incongruent high condition than the incongruent 25% condition (t(14) = 4.19, P < 0.05, with Bonferroni correction), while there was no significant difference between the 0% and incongruent 25% condition (t(14) = 1.04, NS).
In contrast to the above ROIs, the DLPFC (Fig. 4C) failed to show any significant activation during the task compared with the baseline condition (t(14) = 0.74, P = 0.47). This suggests that DLPFC is not actively involved in the task, clearly inconsistent with the DLPFC hypothesis.
In the present study, we tested whether the Performance Dip (Tsushima et al. 2006; Meteyard et al. 2008) takes place in a task, which does not involve identification of letters or words. The main task of the present study required subjects to respond to a color with manual directional responses. The results clearly demonstrated that the Performance Dip also occurred in a non-letter or word identification task when incongruent weak task-irrelevant visual stimuli were presented. Thus, our result suggests that the newly characterized Performance Dip generalizes to different task-types.
Results of Experiment 2 showed a different neural circuit for the Performance Dip in the modified version of Simon effect than that found by Tsushima et al. (2006) for the RSVP letter task. It is implicated that the pre-SMA/SMA, rather than the DLPFC, in the failure of inhibition of task-irrelevant visual information, leads to the Performance Dip. Both in the present study and in the previous study (Tsushima et al. 2006), the task-irrelevant stimuli employed were coherent moving dots. These results suggest that the critical neural circuit involved in the dip in the performance depends on the main task, not the type of task-irrelevant stimulus. In the present study, since the main task involved conflicting motor plans, motor controlling area such as pre-SMA and SMA (Brass and von Cramon 2002; Crone et al. 2006; Wittfoth et al. 2006; Isoda and Hikosaka 2007; Sumner et al. 2007; Imamizu and Kawato 2008) played a role in the Performance Dip. Thus, the DLPFC is not a unitary system that accounts for the Performance Dip.
Although involved brain regions are different, the same principle seems to have resulted in the Performance Dip in the non-letter task used in the present study as well as in the letter and word identification tasks of previous studies (Tsushima et al. 2006; Meteyard et al. 2008). Usually, task-irrelevant visual information is first processed in visual areas and then suppressed by top-down attentional processing (Del Cul et al. 2007). However, there is a tricky range of strength of the task-irrelevant visual information, which is above the threshold in the visual area to allow for initial processing, but not to exceed the threshold for higher areas to provide top-down suppressive processing, resulting in failure of task-irrelevant information within the range to be suppressed by top-down processing. The present study suggests that this failure system is not specific to the DLPFC. The other regions such as pre-SMA/SMA seem to have also failed to control the task-irrelevant information leading to the Performance Dip. That is, the Performance dip could occur with the same principle implemented by different sets of brain regions in different tasks.
While the same kind of task-irrelevant information (coherent motion) was used both in our previous study (Tsushima et al. 2006) and the present study, the point of Performance Dip in the previous study was at 5% coherent motion that was just below the threshold of coherent motion perception, whereas it is at 25% coherence in the present study. What caused the difference? This may be caused by differences in the experimental procedures between the two studies as follows.
One difference in the experimental procedure was in the spatial relations between the task-relevant and task-irrelevant features. In the case of Tsushima et al. (2006), task-irrelevant motion stimuli were presented in an annulus surrounding the task-relevant letter stimulus and were thus spatially isolated from the task-relevant stimuli. In this case, spatial attention could have operated to suppress the motion stimuli (Somers et al. 1999). On the other hand, in the present study, the task-relevant color was a feature of the motion stimuli, and thus, the motion stimulus need not be suppressed via a spatial mechanism. In addition, in the present study, the direction of task-irrelevant coherent motion could raise a conflict with a direction of lever tilting that is processed at a motor control stage. Motor control stages occur later than the visual stage in the neural processing and may process signals from visual areas in a format that is suitable for motor commands and does not necessarily keep as high spatiotemporal resolutions as in outputs from visual areas. These factors might make the signal-to-noise ratio of coherent motion signals higher in motor control stages than in visual stages. If so, pre-SMA and SMA might not detect coherent motion whose coherence level is even higher than 5%, causing the Performance Dip location to shift to a higher level of motion coherence.
So far we have focused on the Performance Dip, lower performance in the incongruent condition at a low level of task-irrelevant motion coherence. In addition to the Dip, in Experiment 1, significantly higher performance at the highest level of coherence in the congruent condition than in the 0% condition was observed (Fig. 2B). Thus, such a positive effect of congruence of the task-irrelevant information can be seen in some conditions. We have a few notes on this congruence effect, which is out of focus in the present article. First, the congruence effect in Experiment 1 was not replicated but only the dip in the performance was replicated in Experiment 2. Thus, we propose that a congruency effect and the Performance Dip may be governed by different mechanisms. Second, we speculate this congruence effect may have cancelled out the performance dip in the congruent condition. As shown in both Experiments 1 and 2, Performance Dip was only observed in the incongruent condition, not in the congruent condition. A neutral stimulus with respect to the main task in the Tsushima et al. study (2006) also caused the Performance Dip. Congruent task-irrelevant stimuli, however, may contain a facilitatory effect by itself with respect to the main task, unlike incongruent and neutral task–irrelevant stimuli, not only at the high coherence conditions but also at lower coherence conditions. Thus, a negative effect caused by the nature of task-irrelevant stimuli in general might be cancelled out by the facilitatory effect by the congruent information, yielding no Performance Dip. Since the focus of the present study was to investigate the Performance Dip, future studies need to address the underlying mechanism for the congruency effect.
The present study demonstrated that the Performance Dip is a general phenomenon across tasks and occurs when a relatively weak task-irrelevant visual motion signal is presented in conjunction with a task. The neuroimaging data showed that the pre-SMA/SMA played a critical role in producing the Performance Dip for the modified Simon task, not in the DLPFC that plays a role in the Performance Dip for letter identification task (Tsushima et al., 2006). These results suggest that the Performance Dip in different tasks can be caused by the common principle in which weak task-irrelevant signals fail to be detected and thus to be subject to suppression. However, this common principle can be implemented in different brain areas.
National Institutes of Health (NIH, R01EY015980, EY019466, AG031941, MH091801); the National Science Foundation (BCS-0345746, BCS-0549036, BSC-0946776); the Human Frontier Foundation (RGP 18/2004); the NIH National Center for Research Resources (P41RR14075, S10RR021110); grants-in-aid for Scientific Research (KAKENHI-22830081, 23680028); the Mental Illness and Neuroscience Discovery Institute; Athinoula A. Martinos Center for Biomedical Imaging; Massachusetts General Hospital; The ministry of education, culture, sports, science and technology, Japan (MEXT) (Tamagawa global COE program, Grant-in-aid for Scientific Research (A), Grant-in-aid for Scientific Research on Innovative Areas); Japan Science and Technology Agency (CREST); and the ERATO Shimojo Implicit Brain Function Project for their support on this project.
We would also like to thank Theresa Cook, Eric Chen, Dan Welch, and Jonathan Dobres for their helpful comments on the manuscript. Conflict of Interest : None declared.