Generally, reaction time (RT) was significantly increased in response to competition trials as compared to control trials, demonstrating that our experimental manipulation was indeed effective. In line with our predictions, we found significantly increased RTs for Stroop-incongruency trials compared to Stroop-congruency trials (t=2.319, p<.025), indicating the occurrence of Stroop-interference. Likewise, we found significantly longer RTs for Color-oddball trials compared to trials of the Oddball control condition (t=3.492, p<.004) as well as for Word-oddball trials compared to trials of the Oddball control condition (t=2.280, p<.026). The bar chart in Fig. 2 depicts the mean RTs and related standard errors for the experimental conditions [means±standard error: Stroop-congruency, 452 ms±4.3; Stroop-incongruency, 471 ms±5.2; Word-oddball, 478 ms±12.8; Color-oddball, 482 ms±6.1; Oddball control condition, 456 ms±5.3]. All subjects reached high accuracy levels, with correct responses amounting to 94.9% across all participants [wrong responses: 3.3%; missing responses: 1.8%].#

Basically, the current data do not corroborate the initial observation of a common activation pattern for Stroop-interference and oddball interference. On the one hand, the current oddball activations nicely match those found in prior investigations (e.g. Gruber and Goschke, 2004). However, we could not replicate the neuronal effects of Stroop-interference as observed by Zysset et al. (2001), who found activation in a frontoparietal network, broadly overlapping with the one we found associated with oddball interference in both the present and prior investigations. This finding suggests that the two contrasts focusing on Stroop-interference and interference from Word-oddballs in the present study map different subcomponents or aspects of attentional competition that refer to distinct neural mechanisms: a motor and an attentional component. (A) A motor component of competition. Stroop-incongruent trials and Stroop-congruent trials used exactly the same words (BIG and SMALL) that only differed in the particular color–size combination. Hence, word meaning (i.e. the word dimension) in both conditions is equally associated with the current task-set and can be expected to be equally able to distract attention from the currently relevant size dimension. In accordance with this assertion, Milham and Banich (2005) found enhanced activity in a posterior division of the ACC in relation to both incongruent and congruent Stroop trials. Based on this and prior findings (e.g. Milham et al., 2002), they also emphasized that attentional demands may be similarly increased on both incongruent and congruent trials due to competition between task-relevant and task-irrelevant information as task-irrelevant information in both conditions is semantically related to the current task-set. On the other hand, words of Stroop-incongruent trials introduce incompatible information at the response-level whereas words of Stroop-congruent trials do not. Taken together, the contrast “Stroop-incongruency vs. Stroop-congruency” subtracts out – at least to some extent – the attentional component of competition (i.e. equates for interference at the attentional level) and focuses on interference at the motor or response level. This conclusion is strongly supported by the activations revealed by this contrast which comprised dorsal and ventral premotor cortices, the pre-SMA, the cerebellum, as well as the basal ganglia and thalamus, regions that are well known to be implicated in the preparation and control of motor responses (e.g. Ikeda 1992; Wiese et al. 2004; Monchi et al. 2006). For instance, the dorsal premotor cortex has been described to play an important role in the mapping of sensory signals onto motor responses (Wise et al., 1996), and there is convincing evidence that this region is strongly involved in inhibitory motor control, i.e. controlled response selection (Praamstra et al., 1999). Similarly, an influential hypothesis assumes that the basal ganglia essentially contribute to motor control by inhibiting incompatible motor tendencies that (might) interfere with an actually intended motor action (Mink, 1996; Aron et al., 2003). Moreover, several studies implicated the thalamus conjointly with the basal ganglia in motor control during conflict situations (e.g. Huettel et al. 2001; Monchi et al. 2001; Aron et al. 2003). Given the reported findings, we presume that the activation pattern related to Stroop-interference primarily reflects strengthened motor control triggered by response competition (i.e. the co-activation of two incompatible response tendencies) which should prevent false responding.#

Moreover, there was significant activation in the left ventral somatosensory cortex related to Stroop-interference which appeared to be significantly stronger than in the Word-oddball contrast. Activation in this region has been repeatedly related to the processing of tactile sensations of the contralateral fingertip (e.g. Burton et al. 1999; Pleger et al. 2006) and beyond that has been shown to be boosted by increased attention to proprioception even in the absence of proper stimulation (Burton et al., 1999). Furthermore, as revealed by morphological investigations of the animal brain (Porter 1991 1997), the somatosensory cortex projects to primary motor areas and in this way may essentially contribute to motor preparation (see Pleger et al., 2006). In conclusion, we interpret the observed somatosensory activation as underlying strengthened proprioceptive/tactile attention to the responding fingers as further aspect of strengthened motor control efforts.#

(B) An attentional component of competition. Task-irrelevant word information in the Word-oddball condition gains saliency through its relative rareness of occurrence. At the same time, words of the Word-oddball condition are response-ineligible, so that response preparation in this condition should be widely unaffected. Consequently, competition emanating from Word-oddballs – which is also true for Color-oddballs – putatively occurs at an earlier processing stage, solely at an attentional level and not at the response or motor level. In line with this notion, Word-oddballs exhibited significant activation in a frontoparietal network which has been repeatedly related to enhanced attentional demands and beyond this has been interpreted as to reflect the exertion of top–down attentional control, i.e. the implementation of selective attention (e.g. Corbetta et al. 2000; Hopfinger et al. 2000; Corbetta and Shulman 2002; LaBerge 2005). Regarding the current study, frontoparietal activation arguably underlay the overriding of the orienting response to the oddball events (i.e. the oddball dimension) which has disrupted the task-appropriate attentional set.#

As already mentioned, we did not replicate the frontoparietal activation pattern associated with Stroop-interference as reported in the Zysset study. There are obvious differences between the studies that may account for the divergent findings. First, Zysset and colleagues employed another variant of the Stroop paradigm, the “Color–Word Matching Stroop Task”. Here, on each trial, subjects are presented with two words simultaneously while they have to match the color of the first to the meaning of the second. In this task version, interference takes place at a pure conceptual level and is (chronologically) separated from response preparation (see Zysset et al., 2001). Second, to define Stroop-interference, Zysset contrasted incongruent trials against neutral trials (consisting of a row of colored Xs), while the present study contrasted incongruent trials against congruent trials, a contrast that arguably equalizes for attentional components of competition processing (see above). On the other hand, the contrast incongruent against neutral trials in the Zysset study comprises substantial attentional components because neutral trials include no word meaning to attentionally interfere with the attended color.#

In summary, the present study's comparison between Stroop-interference and oddball interference did not define a core neural mechanism of attentional selection as was initially expected, but rather dissociated two functionally distinct and complementary control mechanisms that relate to different subcomponents of cognitive interference: a primary motor component and a primary attentional component.#

The two oddball conditions exhibited a broad overlap of neural activation, mainly in anterior regions (e.g. lateral and medial PFC) but also in the intraparietal cortex and other posterior processing areas, e.g. extrastriate visual cortices. Thereby, Color-oddballs exhibited the more extensive activation pattern relative to Word-oddballs and also additional unique activations, e.g. in the left IFG, the medial SFG, the precuneus and the left lingual gyrus. Activation in the left lingual gyrus has been previously implicated in processing of color (e.g. Corbetta et al. 1991; Zeki and Marini 1998), and therefore in our case substantiates the assumption that the oddball color has drawn special attention (i.e. has evoked an orienting response). Other unique activations of Color-oddballs may alternatively reflect quantitative differences, i.e. differences related to the degree of evoked interference rather than qualitative differences between the two oddball conditions. Interestingly enough, we found a double dissociation within the left and right TPC between the two oddball conditions. The TPC has been repeatedly implicated in conscious perception as well as semantic encoding and decoding of objects or object features (Markowitsch 1995; Mesulam 1984; Sewards and Sewards 2002; Damasio et al. 2004). The observed double dissociation putatively reflects the deviation detection in different visual attributes according to the dimension – color or word – in which the oddball occurred. The strong activation overlap that we found in prefrontal areas is in line with the assumption that an anterior prefrontal system modulates activation in posterior processing areas in order to select task-relevant over task-irrelevant information (e.g. Banich et al., 2000). Specifically, the posterior frontolateral cortex or IFJA is a candidate region to be strongly involved in this attentional control function (see below).#

In a recent review, LaBerge (2005) specifies neuroanatomical details of the network that reasonably underlies the exertion of top–down attentional control. Based on the triangular circuit theory (LaBerge, 1997), it is assumed that apical dendrite activity operating within cortico–thalamic–cortical circuits provides stable modulatory activity at the somas of pyramidal neurons that underlies the sustaining of attention over extended intervals of time. On the other hand, brief attentional operations (e.g. rapid attentional shifts) are assumed to emanate from direct connections between frontal areas of attentional control and parietal areas of attentional expression without intermediate thalamic involvement. In this context, an interesting question arose as to whether attentional selection works through boosting the processing of task-relevant information and/or through inhibiting the processing of task-irrelevant information. In the present study, the two oddball conditions showed – with few exceptions – a broad overlap of activation in posterior processing areas while they shared the same task-relevant information (size) and differed in the distracting task-irrelevant attribute dimension (word or color). This finding is compatible with the notion that boosting the processing of task-relevant information plays an especially important role in attentional selection (e.g. Wojciulik et al. 1998; Egner and Hirsch 2005). Banich et al. (2000) also investigated the influence of the processing domain of task-irrelevant information during conflict processing by comparing two different versions of the Stroop-task which differed in task-irrelevant information but not in the dimension which had to be attended. While Banich and colleagues likewise found highly overlapping activations within lateral PFC, in contrast to our data, they also report a strong influence of task-irrelevant information on posterior processing regions. This finding suggests that attentional selection also involves modulating the processing of task-irrelevant information. However, task-irrelevant information in the Banich study was semantically related to the task-relevant information, which is not true for the oddball conditions in the present study and which may account for the divergent findings. As outlined by Banich and colleagues themselves, selection of task-relevant information by lateral prefrontal regions may involve “alerting” all posterior brain regions that process information related to the current attentional set, even if this information is presented in an irrelevant dimension. Generally, when individuals direct their attention to one particular attribute of an item, increased activity is observed in the posterior brain region specialized for processing this visual attribute (e.g. human equivalent of MT or V5; O'Craven et al. 1997; Kastner et al. 1998; Martinez et al. 1999).#

Color-oddballs compared to Word-oddballs elicited more extensive neural activations as well as a stronger effect in the RT data. Thus, both behavioral and neuroimaging data indicated stronger interference emanating from Color-oddballs as compared to Word-oddballs which can be explained twofold: (a) color also occurred as task-relevant attribute during the experimental course within the color-task (see description of the experimental procedure below), whereas word meaning did not. As color had been previously attended to as task-relevant information, it hence may have been better able to attract attentional resources even as task-irrelevant attribute. (b) Independent of the task context, color is inherently a quite salient and conspicuous attribute dimension that apparently can be cognitively represented in a rather direct manner. Word meaning, in contrast, appears to be much less salient and requires mediating semantic decoding processes to be cognitively represented. Therefore, task-irrelevant deviances in color may be generally more outstanding and influential compared to task-irrelevant deviances in word meaning.#

Moreover, we cannot definitely rule out that differences in statistical power may have played a role in producing the relatively stronger and more extensive activations of Color-oddballs. The reason for this is that Word-oddballs had to be less frequent than Color-oddballs because prevalent word values were less frequent than prevalent color values (see Experimental procedures). Consequently, to ensure that Word-oddballs represent a sufficiently salient low-frequency event, we decided to include less Word-oddball trials and thereby accepted to loose some statistical power in the single subject analyses. Even though, this might have contributed to the differential magnitude of activations, we still prefer the two points outlined above as explanation for this finding. We do so in particular because reported results refer to random-effect analyses which can be considered relatively insensitive to trial frequency (as compared to fixed-effect analyses). Basically, random-effect analyses (Holmes and Friston, 1998) identify those brain regions that are consistently activated across different subjects rather than those regions that exhibit the strongest effects (t-values) within single subjects. Furthermore (and related to the latter), degrees of freedom in random-effect analyses exclusively depend on the number of considered subjects, i.e. are independent of the number of trials/scans.#

We observed strong activation related to both Color- and Word-oddballs in the inferior frontolateral cortex bilaterally extending along the posterior banks of the IFS, a cortical region that has been termed the inferior frontal junction area (IFJA) (Brass and von Cramon, 2004). A series of other studies also observed activations in the inferior posterior PFC under different attentionally demanding conditions, suggesting an important role of the IFJA in task-set management as well as in the selection of task-relevant over task-irrelevant information. For instance, this cortical region has been related to task-set preparation (Brass and von Cramon 2004; Gruber et al. 2006), cognitive set shifting (e.g. Konishi et al. 1998; Dove et al. 2000; Derrfuss et al. 2005), response inhibition (Konishi et al. 1999; Konishi et al. 2003), as well as to the processing of Stroop-incongruency (e.g. Leung et al. 2000; Zysset et al. 2001; Milham et al. 2003a; Derrfuss et al. 2005) and oddball events (e.g. Milham et al. 2003a; Gruber and Goschke 2004). In a study similar to ours, Milham et al. (2003a) sought to investigate prefrontal involvement in top–down attentional control. Thereby, they further wanted to determine whether brain areas activated by Stroop-interference overlap with those areas that are sensitive to oddball stimuli. As a result, both Word-oddball trials and Stroop-incongruent trials enhanced brain activity in the posterior frontolateral cortex corresponding to the IFJA. More anteriorly located areas referring to the mid-dorsal lateral PFC, on the other hand, were selectively activated by Stroop-incongruent stimuli. Based on these results, the authors proposed a functional dissociation along the anterior–posterior axis in attentional functioning of the lateral PFC. According to Milham and colleagues (see also Brass and von Cramon, 2004), the posterior inferior PFC in the vicinity of premotor cortex is primarily involved in manipulating posterior regions to ensure selection of task-relevant information. The latter assertion is additionally supported by the fact that the posterior lateral PFC is anatomically strongly interconnected with posterior processing regions (e.g. Barbas and Mesulam 1981; Petrides and Pandya 1984; Petrides and Pandya 1999). On the other hand, more anteriorly located sites within the inferior lateral PFC are thought to be primarily responsible for biasing maintenance and selection of task-relevant information in working-memory. In contrast to Milham's results, we found no activation in the IFJA in relation to Stroop-interference. Rather, the current data exhibited IFJA activation only related to Color- and Word-oddballs that we nevertheless assume to underlie the same function or process of selective attention as the IFJA activation in the Milham study. The fact that we did not observe IFJA activation in relation to Stroop-interference can be explained by differences between the Stroop-interference contrasts of the current study and the Milham study. While we contrasted “Stroop-incongruency” against “Stroop-congruency”, a contrast which at least partly equalizes for attentional processing (see above), Milham and colleagues contrasted response-eligible incongruent trials against response-ineligible incongruent trials. Since response-eligible incongruent color words are part of the task-set, arguably they are more salient (i.e. attention capturing) compared to response-ineligible words. Hence, the contrast “incongruent-eligible” against “incongruent-ineligible” in the Milham study apparently not only taps motor but also substantial attentional components of competition. We additionally computed the contrast “Stroop-incongruency” against “oddball control condition”. In line with the latter interpretation, in this contrast, Stroop-incongruity exhibited significant activation in the right IFJA. As word meaning on trials of the oddball control condition is not semantically related to the task at hand, the contrast can be expected to tap both attentional and motor components of competition. Accordingly, the contrast further showed significant activation in the left ventral premotor cortex.#

Taken together and in accordance with findings of prior investigations (e.g. Banich et al. 2000; Zysset et al. 2001; Milham et al. 2003a; Brass and von Cramon 2004), the current data emphasize a prominent role of the posterior lateral PFC in top–down attentional control. Particularly, the IFJA may provide a top–down executive mechanism for imposing an attentional set for task-relevant information by modulating processing in posterior neuronal perceptual systems.#

This putative role of the IFJA is also in line with the widespread assumption of a basically twofold neural organization of cognitive control comprising a monitoring instance represented by frontomedian cortices and an executive instance represented by frontolateral cortices. A series of prior studies consistently provide strong evidence for complementary roles of frontomedian cortex, particularly the ACC, and frontolateral cortex in conflict processing, wherein ACC monitors for the occurrence of conflict or interference and thereby signals the frontolateral cortex to strengthen control efforts to accommodate external demands (e.g. Carter et al. 1998; MacDonald et al. 2000; Botvinick et al. 2001; Kerns et al. 2004). These studies, however, consistently pointed to more anteriorly located prefrontal areas – referred to as ‘dorsolateral prefrontal cortex’ (DLPFC) – as the primary frontolateral region in the exertion of top–down attentional control. In contrast, the presented findings single out the IFJA and in that may give rise to the speculation that the primary role in the context of cognitive control attributed to DLPFC is owed to the fact that consistent activation in the posterior PFC has been neglected (see Brass et al., 2005).#

Twelve healthy and right-handed young adults participated in this study (6 men and 6 women; mean age 25,67; SD=1.88) after they had given written informed consent. All subjects had normal or corrected to normal vision. They received a monetary payment for participating. Participants were pre-trained a day before the fMRI session to ensure high accuracy levels in the task performance. Three participants had to be excluded from the statistical analysis due to uncorrectable motion artifacts in the fMRI data.#

Imaging was performed on a 3-T MRI scanner (Bruker Medspec 30/100; Bruker BioSpin MRI GmbH, Ettlingen, Germany) with a standard birdcage headcoil. Nineteen axial slices (voxel size 3×3×5 mm³, distance factor 0.2) were positioned in parallel to the AC–PC plane, covering the entire brain. Prior to the functional scans, anatomical MDEFT (modified driven equilibrium Fourier transform pulse sequence) slices and EPI-T1 (echo-planar imaging, t1-weighted) slices were obtained. These measurements were followed by three runs of a single-shot, gradient EPI sequence (TR 1.75 s, TE 30  ms, flip angle 90°, filed of view 192 mm, 64×64 matrix) each acquiring a total of 535 image volumes. Stimulus presentation was synchronized with functional image acquisition by means of ERTS (Experimental Run Time System, Version 3.11, BeriSoft Cooperation, Frankfurt am Main, Germany). In a separate session, a high-resolution structural scan (3D MDEFT) was obtained for each subject. Importantly, event duration (i.e. stimulus onset asynchrony) was not a simple multiple of the interscan interval (TR), so that the event onsets were systematically jittered with respect to the scan onsets, and the hemodynamic response was sampled at different time points (i.e. was oversampled). In particular, TR was 1.75 s and trial length was 2.25 s while we triggered every fourth trial so that the hemodynamic response was sampled at three different time points.#

Stimuli were presented using ERTS (Experimental Run Time System, Version 3.11, BeriSoft Cooperation, Frankfurt am Main, Germany). Participants were instructed to make a left response (corresponding to an index finger button press by the right hand) if the word stimuli appeared in big font size and a right response (corresponding to a middle finger button press by the right hand) if the target word appeared in small font size. It was explicitly pointed out that word meaning was irrelevant over the whole experimental run and that subjects would have to ignore it for good performance. Participants were further instructed to be fast but accurate. Because we planned to include a Color-oddball condition consisting of a task-irrelevant rarely occurring color (according to the oddball condition we created in prior investigations), we could not implement all of the experimental conditions of interest within the commonly used color Stroop task, in which color serves as response-indicating dimension. Therefore, we created a size task in which all conditions of interest could be implemented and compared. Nevertheless, we also included color Stroop trials to insert task switches in order to keep the overall attentional demands on a high level. In the beginning of each trial, a cue lasting for 500 ms signalized which task had to be performed on the upcoming stimulus. Targets appeared after a short delay of 250 ms and lasted for 750 ms. RTs were recorded within a time period of 1500 ms beginning with the target presentation until the onset of the next cue. Thus, the total trial duration was 2250 ms. The experimental stimulation was fully counterbalanced in that every dimensional value as well as possible value combinations occurred equally often within the experimental course. As an intended exception, we included low-frequency events to provide oddball conditions. Only occasionally we presented words without semantical connotation to size (e.g. German translated words for “COLD”, “LOUD”, “GOOD” and “NEAR”) which should serve as Word-oddball condition. We presented ten different oddball words that were matched for word length and syllable number with the prevalent word stimuli. Each oddball word was presented with a frequency of 1 in every 112 trials (∼0.9%) while each size word appeared with a frequency of 1 in every 6 trials (∼17%). Likewise, besides the prevalently occurring blue and yellow colored stimuli, we rarely presented words in red ink color as Color-oddball condition. Red colored stimuli appeared with a frequency of 1 in every 32 trials (∼3%) while blue and yellow colored words appeared on virtually every second trial (∼49%) each. Noteworthy, while oddball conditions occurred with a different frequency, the frequency ratio (i.e. relative frequency) of rare values (i.e. oddball values) to prevalent values was roughly identical for the color and the word dimension (∼1:17). Oddballs were distributed within the stimulation sequence in a pseudorandomized manner that ensured that no oddball trial followed another oddball trial, and furthermore that oddball trials were preceded by every other trial type equally often. During Stroop-congruency trials, word meaning matched the currently relevant dimension (i.e. the word “BIG” printed in big letters or the word “SMALL” printed in small letters) while during Stroop-incongruency word meaning denoted the opposed mapped value (i.e. the word “SMALL” printed in big letters or the word “BIG” printed in small letters). Trials of the Oddball control condition served as contrast condition for both Word-oddballs and Color-oddballs and included the response-ineligible color word “GREEN” that occurred equally often as the response-eligible size words. Also, all trials of the Oddball control condition appeared in the prevalently occurring ink colors blue or yellow. The stimulation sequence comprised congruent and incongruent trials as well as Color-oddball trials and trials of the Oddball control condition equally frequent, 36 times each, and trials of the Word-oddball 10 times. Important to note, all analyzed trials (with the exception of Color-oddball trials) were basically congruent, i.e. color and size were mapped to the same response. However, we also presented basically incongruent trials (e.g. color-incongruent trials within the size-task) in order to prevent subjects from focusing on the same dimension in both tasks. Overall, we included a total of 1168 trials which participants performed on three separate sessions. Conditions were presented in a pseudorandom, fully counterbalanced order. Experimental conditions and respective stimuli are depicted in Fig. 1.#