DISCUSSION

As indicated by these fMRI results, and as expected considering the complex, multifactorial nature of the task, multiple cognitive regions are implicated in VOT performance. The group fMRI mapping demonstrates an extensive bilateral response across all brain lobes, predominantly posterior, with a significant right hemisphere extent lateralization in parietal lobes. Of these localizations, the fMRI clusters in bilateral superior occipital and posterior superior parietal lobes demonstrate the most robust response. Presumably, these clusters reflect brain regions subserving the visuospatial processes that are the intended focus of the VOT spatial integration construct demands, and have been implicated in previous fMRI studies of visuospatial paradigms (Carpenter et al., 1999; Cohen et al., 1996; Gauthier et al., 2002; Grön et al., 2000; Jordan et al., 2002; Kraut et al., 1997; Ng et al., 2000; Sack et al., 2002; Shen et al., 1998; Sugio et al., 1999; Tagaris et al., 1996; Vannini et al., 2004; Weiss et al., 2003). Extensive clusters in bilateral regions of lateral occipital and posterior inferomedial temporal lobes are similarly robust, and follow the expected ventral visual stream subserving object identification and semantic retrieval (Carpenter et al., 1999; Gauthier et al., 2002; Haxby et al., 1991; Kraut et al., 1997; Shen et al., 1998; Sugio et al., 1999). The left frontal lobe cluster overlays the anatomical region of lateral inferior/middle precentral gyrus that is proximal to Broca's area, and likely represents the covert naming response for the fMRI paradigm. The small bilateral middle frontal gyrus clusters reflect activity in the frontal eye fields for this visual task, and the left anterior cingulate cluster likely represents a response to attention load.

The hemispheric lateralization indexes indicated a significant difference in comparison between parietal regions, with right parietal extent significantly great than left parietal. These lateralization index values are a measure of comparative fMRI spatial extent only, and do not necessarily indicate a segregation or difference between hemispheric functions; that is, it is not known from these results whether right and left parietal regions are involved in similar or different aspects of visuospatial processing. The left frontal region is probably a lateralized linguistic response related to the word-retrieval task demands, but the paradigm was not optimally designed to isolate specific functional responses. The lateralization index values are intended as a comparative measure of global–bilateral fMRI spatial extent.

These spatially extensive fMRI mapping results, at first inspection, could be interpreted as indicating at least a bilateral, and even a global network of functionality for VOT performance. Indeed, it would seem intuitive that without subserving networks of primary and visual association areas, VOT performance would be impaired. Object identification and expressive language are also expected networks of task involvement, since even a simplified multiple-choice or word discrimination paradigm design will still depend on a linguistic operative. However, some important aspects of fMRI assessment and the fMRI paradigm need to be considered.

Unlike lesion studies, or other functional mapping methods such as intraoperative cortical stimulation or transcranial magnetic stimulation, fMRI mapping by itself does not provide a hierarchy of functional organization and connectivity (Sarter et al., 1996). Interference with task performance from either a lesion or one of the mentioned stimulation techniques can be interpreted ad hoc as an indication that the affected region is a critical resource that must be invoked for effective task performance. Compared to these focal stimulation techniques, functional imaging modalities such as fMRI provide an advantageous means of spatially mapping multiple regions that demonstrate a hemodynamic response to a stimulus or cognitive task. But the fMRI indication that distributed cerebral interactions are involved in a psychometric task does not provide direct evidence that a focal region is imperative to the functionality that the test is intended to probe. For example, it is not known from these fMRI group results what the effects of a specific lesion or pathology would have on an individual subject's VOT performance. When assessed within these constraints, these fMRI results can offer evidence of the regions and pathways that are involved in the VOT and by implication the possible neural sources which may be indicated by a performance deficit. Previous studies have attempted to address this issue of differences between functional imaging and lesion studies. Ng et al. (2000) reported a bilateral posterior parietal fMRI participation in a group of normal volunteers performing a modified version of the Benton Judgment of Line Orientation visuospatial task. Comparison was made to a cohort of patients with either right or left parietal lobe damage. The right hemisphere lesion group was found to have a somewhat more severe performance deficit. Sack et al. (2002) reported superior parietal lobule involvement with fMRI mapping during performance of a visually presented spatial judgment task. Task performance was selectively impaired when volunteers were subjected to repetitive transcranial magnetic stimulation (rTMS) of the superior parietal lobules. Hemispheric differences or laterality in the parietal lobes were not reported in the Sack et al. study. Further research will be necessary to determine the correspondence between fMRI mapping results and degree of deficit from localized cerebral damage to performance of either the VOT or other visuospatial task paradigms.

The fMRI paradigm used in this study was an attempt to adapt the VOT method to the constraints of a MRI-compatible presentation and performance. The modification of the normally required overt naming response to a covert word generation and simplified word discrimination task allowed an effective minimization of confounding head movements associated with overt word production. This modification precluded an ability to record confrontational naming task accuracy, and the involvement of networks for overt speech. It is unknown to what extent the modification of the VOT paradigm to exclude overt verbal responses may have affected the fMRI mapping results. Presumably the decreased confrontational naming component of the fMRI paradigm may have affected the extent and magnitude of detected responses in functional language regions. This may have shifted the indices of lateralization or biased the results toward posterior regions. Debriefing of the subjects indicated that they were compliant with the paradigm instruction to covertly generate a word name for the VOT image presentations, and the accuracy of word discrimination responses demonstrated at least some measure of success to the naming performance. This word discrimination modification is similar in effect to the multiple-choice derivative proposed by Schultheis et al. (2000). Accuracy of word discrimination response was not applied as a covariate in the analysis, since the intent of the study was toward a general delineation of the regional responses to the VOT performance. As measured by the word discrimination scores there was a limited range in performance, with all subjects performing very well. Additional investigation with fMRI of subjects who perform poorly on VOT might yield differences for fMRI results in relation to intact subjects, or insights into lesion sites that directly affect VOT performance.

By modeling the fMRI paradigm after the VOT method, some restrictions are applied to the outcome specificity. As previously noted, the fMRI results indicate a general map of regional task involvement, but lack a specific focus to the critical task demands for visuospatial integration ability. The fMRI analysis is based on a simple comparison of (1) a visuospatial task performance condition, during the initial 3 s of each VOT image presentation; and (2) a presumed non-task condition, during the remaining 7 s between presentations. This presumption is based on subjects' post-fMRI debriefing indicating that they were usually able to identify each image during the first few seconds of its presentation. However, the selected control condition during which a reduced amount of visuospatial processing occurred was not devoid of visual stimulus or cognitive effort. The visual stimulus, working memory, and object name recognition processing that occurred during the control condition would be expected to minimize the effects of these operatives within the final contrast results. For example, the lack of primary visual area response mapping could be due to a relative lack of visual stimulus contrast between the conditions. The chosen conditional contrast could be expected to favor the visuospatial VOT demands, at the expense of underestimating the ongoing processes that occur throughout the entire 10 s of each trial.

Similar to the clinical application of the VOT, a self-paced paradigm and appropriate analysis might provide a more accurate representation of task performance timing and resultant hemodynamic response. A modification of the VOT method toward the design of a fMRI paradigm with a controlled non-task stimulus condition could provide additional specificity to the visuospatial task demands. An example of such a paradigm might be to contrast task conditions of the VOT with performance of the Boston Naming Test during the same scan. The similarity in line drawings and confrontational naming demands between the two tests could control for both visual presentation stimulus and language response. The resulting contrast could be presumed to indicate the specific regions involved in visuospatial integration. Alternatively, a paradigm design that contrasts VOT performance with non-stimulus periods of “rest” would be expected to more fully demonstrate visual stimulus and cognitive responses, but at the expense of visuospatial processing specificity. Both fMRI paradigm design approaches of (1) self-pacing, and (2) further stimulus condition control are beyond the scope of this current study, but may warrant additional investigation.

Previous studies showing gender-based differences in fMRI of visuospatial paradigms have been reported, utilizing a maze task (Grön et al., 2000) and mental rotation (Jordan et al., 2002; Weiss et al., 2003). Gender differences were not detected in this study at the groupwise FDR-corrected threshold of p < .005, so no gender effect is observed within the laterality index measures. At a reduced threshold of p < .00005 (uncorrected), the gender-based differences in this study are subtle. While not all the gender-related differences from these previous citation studies were seen with the VOT fMRI paradigm, the posterior cingulate region with male > female contrast (Grön et al., 2000) and right inferior parietal region with female > male contrast (Grön et al., 2000; Weiss et al., 2003) are in agreement. Other previously reported regions of gender-based differences in fMRI of visuospatial paradigms (e.g., right frontal lobe in female > male contrast) that are not observed in this study may be due to different task demands for the VOT performance compared to the tasks of mental rotation or maze navigation employed in the prior reports. It is also possible that the gender imbalance and relatively small sample size of 11 male and 8 female subjects in this study is insufficient to adequately define gender-based differences.

Tagaris et al. (1996) reported a quantitative correlation between mental rotation task difficulty and fMRI responses in superior parietal lobe. fMRI of increasing angular disparity in a mental rotation task has been reported to demonstrate increasing fMRI responses in parietal, frontal, and temporal regions (Carpenter et al., 1999). Gauthier et al. (2002) measured a fMRI increase in superior parietal lobe proportional to viewpoint disparity during mental rotation, but not during object recognition. The VOT task difficulty increases with successive presentations, but parametric analysis of signal increase related to time did not yield a fMRI response correlation. Inspection of averaged time courses from the 4 exemplary subjects in Figure 2 fails to indicate a consistent signal increase during the task duration. The failure to detect a fMRI signal change corresponding to VOT task difficulty may be due to inter-subject variability, insufficient degree of increasing difficulty, or insensitivity of the analysis model to subtle response differences.

In summary, this study is a report on the multiple regional hemodynamic BOLD responses to a fMRI-adapted paradigm of the VOT method. The global and bilateral nature of these mappings indicates that VOT performance involves a complex and extensive network of perceptual, linguistic, and integrative cognition. These findings represent an initial application of fMRI to the neuroanatomic correlates of the VOT, and further study can be expected to elicit enhanced understanding of the brain–behavior relationships specific to this neuropsychological test. Consideration of these fMRI results must be made within the context of potential limitations and differences between functional mapping methods and lesion studies. These results suggest that clinical VOT assessments should be made within the framework of a test battery that provides complementary probes of similar functional abilities.