Report

Lateralized Cognitive Processes and Lateralized Task Control in the Human Brain

See allHide authors and affiliations

Science  18 Jul 2003:
Vol. 301, Issue 5631, pp. 384-386
DOI: 10.1126/science.1086025

Abstract

The principles underlying human hemispheric specialization are poorly understood. We used functional magnetic resonance imaging of letter and visuospatial decision tasks with identical word stimuli to address two unresolved problems. First, hemispheric specialization depended on the nature of the task rather than on the nature of the stimulus. Second, analysis of frontal candidate regions for cognitive control showed increased coupling between left anterior cingulate cortex (ACC) and left inferior frontal gyrus during letter decisions, whereas right ACC showed enhanced coupling with right parietal areas during visuospatial decisions. Cognitive control is thus localized in the same hemisphere as task execution.

Despite intensive effort, the what and how of human hemispheric specialization (“lateralization”) remain poorly understood. In a functional magnetic resonance imaging (fMRI) study, we investigated two unresolved issues that are central to understanding hemispheric asymmetry (14). First, there is conflicting psychophysical (17) and neurophysiological (811) evidence about how much lateralization depends on the nature of the processed stimuli per se and how much on the nature of the cognitive task performed. Second, there is psychophysical evidence that hemisphere-specific mechanisms may be involved in cognitive control (12, 13), e.g., in the voluntary guidance of complex actions through attentional top-down modulation of early sensory processing and integration of feedback (1416). The corresponding neurophysiological basis for these putative hemisphere-specific mechanisms of cognitive control, however, has not yet been clarified.

Male right-handed volunteers, 16 in number, performed two different tasks on an identical set of words consisting of concrete German nouns, each four letters in length (three letters in black and either the 2nd or 3rd letter in red) (17). In the letter-decision task the subjects had to ignore the position of the red letter and indicate whether or not the displayed word contained the target letter “A”; the visuospatial-decision task required them to ignore the language-related properties of the words and to judge whether the red letter was located left or right of the center of the word (18). During an additional baseline condition, subjects performed a simple reaction-time task on equivalent word stimuli, i.e., they responded as quickly as possible to the onset of each stimulus.

Varying the cognitive demands while keeping the stimuli constant across tasks provided explicit answers to the two questions of concern. First, we tested directly whether hemispheric specialization was processing-dependent rather than stimulus-dependent. If lateralization depends on the nature of the stimuli, no strong hemispheric differences should be observed between letter- and visuospatial-decision tasks because the stimuli and display conditions were identical. However, if lateralization depends on the nature of the cognitive processes needed to implement a particular task, clear hemisphere-specific activation patterns should occur. Second, by requiring different responses to identical stimuli, we ensured that the participants had to engage in supervisory processes underlying cognitive control, namely response selection, response inhibition, and conflict monitoring (1922). We characterized these top-down effects as context-dependent functional contributions of frontal candidate areas and tested whether they showed any hemispheric specificity.

Comparing letter with visuospatial decisions showed a significantly higher blood oxygen–level dependent (BOLD) signal in a number of areas in the left hemisphere (Fig. 1A), including Broca's region in the inferior frontal gyrus (IFG) (x = –42/y = 32/z = 0, tmax = 11.29), fusiform gyrus (–38/–52/–22, tmax = 9.25), lateral extrastriate cortex (–36/–84/14, tmax = 7.24), ventral premotor cortex and posterior IFG (–48/12/22, tmax = 6.51), anterior cingulate cortex (ACC) (–8/10/58, tmax = 8.54), and supplementary motor cortex (–8/2/70, tmax = 8.39). (Note: all clusters significant at P < 0.05, corrected, are reported; coordinates refer to local cluster maxima and tmax to the corresponding t-value.) The only activation that extended from the left into the right hemisphere was found in primary visual cortex (left: –6/–82/14, tmax = 6.43; right: 10/–78/12, tmax = 6.30; P < 0.001). In contrast, a significantly increased BOLD signal for visuospatial decisions as compared with letter decisions was found in the anterior (58/–24/44, tmax = 7.64, P < 0.019) and posterior part of the right inferior parietal lobule (46/–76/34, tmax = 5.92, P < 0.020), but not in any area of the left hemisphere (Fig. 1B).

Fig. 1.

Processing-dependence of hemispheric specialization. Brain areas with significant BOLD signal differences when comparing (A) letter with visuospatial processing and (B) visuospatial with letter processing, by using identical stimuli. Significant clusters of activation (P < 0.05, corrected) are displayed as a graded projection (i.e., medial activations appear less bright than lateral ones) on the normalized rendered brain of a single subject (5203) from this study.

These results showed a clear dissociation between the hemispheres: letter decisions led to strong left-hemispheric but no right-hemispheric activations except for the bilateral activation of primary visual cortex. Conversely, visuospatial decisions evoked strong right-hemispheric but no left-hemispheric activations. Because the stimuli were identical between conditions and subjects maintained fixation throughout the experiment, these differences can neither be ascribed to changes in the stimulus material nor to differential eye movements. Likewise, any potential bias at the level of visual input or motor output can be excluded because the stimuli occurred equally often in the left and right visual hemifield, and subjects responded equally often with their left and right hands. This functional dichotomy of the hemispheres may depend on cognitive control mechanisms that direct attention to specific stimulus features and guide the subsequent information processing. These processes could be similar to previously described mechanisms of attentional top-down modulation that are mediated through changes in effective connectivity (2325). We accordingly investigated whether a context-dependent modulation of functional influences from areas involved in cognitive control could be inferred from our data.

In order to determine candidate areas that exert cognitive control through attentional top-down influences in our experiment, we compared both letter- and visuospatial-decision tasks against the baseline condition. The rationale for these comparisons was that any such area should show higher activity during the attentionally demanding decision tasks as compared with the simple reaction task to equivalent stimuli during the baseline. We hypothesized that the source of this control should be found in either dorsolateral prefrontal cortex (DLPFC) or in the dorsocaudal part of the ACC (1922, 26). No significant differences were found in either left or right DLPFC when comparing both tasks together against the baseline. In contrast, the dorsocaudal ACC showed significant differences between tasks and baseline (P < 0.001, corrected), both in the left (–6/8/50, tmax = 8.74) and the right hemisphere (8/14/48, tmax = 8.36). Identical results, i.e., significant bilateral ACC activations and no significant activations in left and right DLPFC, were obtained when we compared the letter- and visuospatial-decision tasks separately against the baseline. The locations of the bilateral ACC activations corresponded well to previous studies of the role of the ACC in cognitive control (26).

These findings do not yet distinguish, however, whether or not there are differences between left and right ACC with regard to task-dependent changes of their functional interactions with the rest of the brain. Therefore, to characterize top-down influences from ACC directly in terms of functional interactions, we tested whether any region throughout the whole brain showed context-dependent changes in coupling with either left or right ACC, over and above any main effect of task or any main effect of ACC activity. This test was performed as an analysis of effective connectivity (17, 27). The results showed a dissociation between the two cerebral hemispheres. The left ACC increased its influence during letter decisions on a region in the anterior part of the left IFG (local maximum at –26/44/6, tmax = 5.61, P < 0.011, corrected) (Figs. 2 and 3A). No significant changes in the contribution of left ACC to any right-hemispheric area were observed during letter decisions, nor to any brain area at all during visuospatial decisions. The right ACC increased its influence during visuospatial decisions on areas in the anterior (42/–42/44, tmax = 4.63, P < 0.034, corrected) and posterior (28/–72/48, tmax = 4.49, P < 0.015, corrected) parts of the right intraparietal sulcus (IPS) (Figs. 3B and 4). In contrast, the right ACC did not show significant changes in its contribution to any left-hemispheric area during visuospatial decisions nor to any brain area at all during letter decisions.

Fig. 2.

Psycho-physiological interactions (PPI) of the left ACC. (A) All areas are shown that receive a significant context-dependent contribution from left ACC during letter decisions, projected on the same rendered brain as in Fig. 1. Left ACC significantly increased its influence on left anterior IFG during letter decisions (local maximum: –26/44/6, tmax = 5.61, P < 0.011, corrected). Note the specificity of this result: Even when the threshold was reduced to P < 0.05, uncorrected, no other significant clusters were found throughout the brain. (B) This schema summarizes the negative findings for left ACC: As indicated by the gray dashed lines, left ACC shows no context-dependent contributions to any right-hemispheric area during letter decisions and none to any left- or right-hemispheric area during visuospatial decisions.

Fig. 3.

PPI of left and right ACC for a single subject (5203). Measurements during the letter decision task (L), black circles; measurements during the visuospatial task (VS), red triangles. To take into account the hemodynamic lag of the BOLD signal when assigning data points to conditions, the onset of conditions was delayed by 6 s. Condition-specific regression slopes, bL and bVS. (A) Mean-corrected activity (in arbitrary units) in left IFG (voxel coordinates of subject 5203: –26/50/4) is displayed as a function of mean-corrected activity in left ACC (i.e., first principal component from a sphere of 4-mm radius, centered on –8/6/48). The difference between regression slopes constitutes the PPI (P < 0.001, t = 3.88). (B) Mean-corrected activity in the anterior right IPS (38/–38/36) is displayed as a function of mean-corrected activity in right ACC (i.e., first principal component from a sphere of 4-mm radius, centered on 6/10/50). This PPI was significant at P < 0.001 (t = 3.34).

Fig. 4.

PPI of the right ACC.(A) All areas are shown that receive a significant context-dependent contribution from right ACC during visuospatial decisions, projected on the same rendered brain as in Fig. 1. Right ACC significantly increased its influence on posterior (28/–72/48, tmax = 4.49, P < 0.015, corrected) and anterior parts of the right IPS (42/–42/44, tmax = 4.63, P < 0.034, corrected) during visuospatial decisions. Note the specificity of this result: Even when the threshold was reduced to P < 0.05, uncorrected, no other significant clusters were found throughout the brain. (B) This schema summarizes the negative findings for right ACC: As indicated by the gray dashed lines, right ACC shows no context-dependent contributions to any left-hemispheric area during visuospatial decisions and none to any left- or right-hemispheric area at all during letter decisions.

Our results suggest that cognitive control mediated by the ACC is localized in the same hemisphere as the areas involved in task execution. These findings do not contradict previous studies that propose a predominantly evaluative, feedback-integrating role for the ACC in cognitive control (19, 21). Although direct anatomical connections between the ACC and both inferior frontal and parietal regions have been described (28, 29), it is possible that the top-down influence of the ACC that we have shown in terms of effective connectivity is an indirect one, using the DLPFC (or other areas) as a relay station. However, as demonstrated by the contrast between tasks and baseline described above, such an additional executive area did not emerge from our analyses. Other tasks that require prefrontally mediated components of cognitive control may lead to similar hemisphere-specific couplings for the DLPFC as observed for the ACC in our study. A study of effective connectivity (25) that investigated two tasks with right hemisphere dominance demonstrated top-down effects that were specific for the right hemisphere, i.e., from the right middle frontal gyrus (area 46) on right extrastriate areas. It is thus likely that our findings generalize to other lateralized tasks. Although we cannot exclude lateralization contingent on stimulus type in some situations, our results are consistent with previous findings from split-brain patient studies (7) and positron emission tomography (11) showing hemispheric specialization based on task demands. Research on hemispheric specialization should move beyond analyses of asymmetric regional activations and focus more strongly on functional interactions within and between hemispheres.

Supporting Online Material

www.sciencemag.org/cgi/content/full/301/5631/384/DC1

Materials and Methods

References

References and Notes

View Abstract

Navigate This Article