The present study demonstrated that the LIFG is critical for suppressing prepotent responses to simple letter stimuli in a Go/NoGo task. Patients with lateral PFC lesions that included left posterior IFG and frontal opercular regions made more false alarm errors than controls, particularly when response inhibition was more difficult. Conversely, similar to controls, they showed faster RTs on error trials than on correct trials and a low rate of misses, suggesting that impulsive responding, rather than a failure to comply with task instructions or to maintain task set, can account for the increase in NoGo errors. This deficit in response inhibition was not initially predicted by the neuroimaging literature, which has focused almost exclusively on RIFG. The meta-analysis shown in Fig. (see also [40
]) and numerous fMRI experiments [3
] suggest that activation in right inferior and middle frontal gyri is associated with response inhibition to a greater extent than the corresponding left hemisphere regions. Our results provide a relatively unique example of how neuropsychological data can constrain models of cognitive function developed mainly from fMRI data.
An intriguing new possibility, however, also emerged from our ALE map, which is the most comprehensive meta-analysis of motor response inhibition tasks to date: the importance of bilateral anterior insular regions in Go/NoGo and Stop-Signal tasks. Wager, Nee, and colleagues have made this observation as well, based on prior experimental [25
] and meta-analytic [43
] evidence from tasks of interference resolution. Although a common finding in neuroimaging studies, the importance of the insula in response inhibition has not been widely discussed in the literature, nor has it been reported in previous lesion studies. The maximal overlap in the current patient group includes portions of both left IFG (especially pars opercularis and rolandic regions) and left insula, so we cannot distinguish the relative contributions of each.
Nonetheless, the current lesion study has yielded a deficit in patients that was not generally predicted by the neuroimaging literature. While the present study does not question the importance of RIFG in response inhibition, it does draw attention to the complementary nature of results from neuropsychology and neuroimaging [44
] and shows that functional imaging results should not become the sole source for generating hypotheses in cognitive neuroscience. Shallice [47
] has noted some of the pitfalls of comparing lesion and neuroimaging results, particularly for cognitive processes that are poorly understood. Another caveat is that the degree of lateralization in neuroimaging studies is often relative and not absolute. Thus, a unilateral lesion may not produce a deficit predicted by the neuroimaging data [45
] if the spared hemisphere can compensate (or vice versa). Our results suggest that current imaging techniques may not identify every brain area that makes a significant contribution to a particular function, although the possibility of type II error ("false negatives") in fMRI analyses cannot be overlooked.
Aron and Poldrack [48
] have argued that response inhibition is right lateralized, which receives support from the quantitative meta-analysis presented in Fig. . A model of fronto-striatal loops implementing motor inhibition is quite plausible [48
], but it does not explain why fronto-striatal loops of the right hemisphere are dominant for inhibitory control. The RIFG does not appear to have privileged access to the indirect fronto-striatal pathway and is not likely to have direct projections to motor cortices. Thus, no specific anatomical asymmetries between left and right inferior frontal cortices can account for why LIFG should not play a role in inhibitory control at all. So, extant anatomical knowledge alone would probably not hint at a unique role for RIFG in inhibitory control.
The clear performance deficit in LIFG patients suggests that response inhibition processes are represented bilaterally in IFG. Nevertheless, our results do not preclude the possibility that RIFG patients would be even more impaired on this task. However, the contribution of LIFG to inhibitory control is more than minor, since the spared RIFG was not sufficient to compensate for the effect of the LIFG lesion. Previous studies included fewer left unilateral PFC patients than the current experiment, and did not employ direct comparisons between patient groups [27
]. One of these studies did not find a correlation between Stop-Signal RT and LIFG lesions [27
] and the other found a correlation with lesions in left BA 6 [29
]. The lesion locations in individual patients were not presented in these papers, complicating a direct comparison to the present results.
Another potential explanation for this discrepancy is that different tasks were used. Are there cognitive and/or motor differences between Stop-Signal (SSRT) and Go/NoGo (GNG) tasks that would recruit different regions (or different hemispheres) in PFC? In general, the extent and laterality of IFG activations reported in neuroimaging studies do not differ between the two tasks. Only two studies have administered GNG and SSRT to the same groups of subjects. Zheng et al. [50
] implicated right middle frontal gyrus as a key region in both tasks. However, Rubia and colleagues [22
] reported that although overlapping PFC regions were activated in GNG and SSRT tasks, the former had more L hemisphere involvement, the latter more R hemisphere involvement.
Very recently, some theorists have proposed that the Go/NoGo task and the Stop-Signal task measure different aspects of response inhibition (Aron and Poldrack [48
]; Eagle et al. [51
]). Eagle, Bari, and Robbins [51
] divided "action inhibition" into different subtypes with distinct neuroanatomical and psychopharmacological correlates. Following Schachar et al. [52
], they distinguished between action restraint – inhibition of a motor response before
the response has been started, and action cancellation – inhibition of a motor response during
its execution. This model of response inhibition views the Go/NoGo task as an example of action restraint, whereas the Stop-Signal task is an example of action cancellation. Furthermore, the GNG task is thought to be dependent on serotonin (SSRT is not), while SSRT might be dependent on norepinephrine, although this latter point was not entirely clear [51
]. The GNG task also seems to be influenced by norepinephrine, implying that the two tasks share some of the same neural substrates.
The paper by Schachar et al. [52
] is notable here, because it is the first to test the same group of subjects on the standard SSRT (cancellation) and a new version that is similar to GNG (restraint). The participants were children with and without ADHD. Interestingly, performance on the restraint and cancellation variants was significantly correlated in the control children, suggesting that the two tasks assess a common latent inhibition construct and share cognitive and neural resources. Furthermore, children with ADHD were impaired in both versions of the task, and their performance did not show a correlation between the two tasks, suggesting less sharing of resources in ADHD [53
Robertson and colleagues [54
] have argued that in addition to motor response inhibition, the Go/NoGo task is a measure of sustained attention. Both motor response inhibition and/or sustained attention deficits can produce high NoGo error rates. Two versions of the Sustained Attention to Response Task (SART), a variant of the Go/NoGo task, were developed to target this ambiguity [54
]. In the random SART, subjects withhold responses to the digit "3" (randomly intermixed with other digits 11% of the time), but in the fixed SART, the numbers always proceed in numerical order. In the random SART, lapses of attention, perseveration, and failures of inhibition can all lead to false alarm errors, whereas with completely predictable NoGo trials in the fixed SART, false alarms are primarily due to lapses of attention. TBI patients were impaired in both, but disproportionately so in the fixed SART [54
In our experiment, the 10% NoGo blocks might be comparatively more monotonous than the 50% blocks, so sustained attention is required to a greater degree in the former. LIFG patients showed a larger difference between RTs in the two probability conditions than controls. This alone would be consistent with the sustained attention account, in which speeding up in the 90/10 condition can be attributed to entering "autopilot" mode. However, the 10% NoGo condition differs from the fixed SART in that the NoGo stimuli are unpredictable. Importantly, the LIFG patients showed increased false alarm rates in both conditions. Although the percentage of error trials is higher in the 10% condition, the absolute number of errors is similar. Thus, another possibility is that the subjects responded on a small percentage of trials without considering the Go/NoGo signal at all. This type of error was increased in the LIFG group, exemplifying an important form of impulsive responding. Therefore, an inhibitory control deficit remains the best explanation for the LIFG patients' performance.
Further work is required to elucidate the precise nature of response inhibition in both the GNG and the SSRT tasks. For example, there is clear evidence that motor preparation occurs on both Go and NoGo trials [56
] so to some extent this task can be considered not only in the light of action restraint, but also as a form of action cancellation. Moreover, recent conclusions based on the SSRT task, with respect to the nature of inhibitory control, may not be definitive at this point. Along these lines, a unique aspect of the SSRT task is that some versions involve switching attention across modalities, from a visual target to an auditory stop-signal. Therefore, alternative interpretations of SSRT results are possible, incorporating both response inhibition processes and the ability to switch attention to the stop-signal tone [57
]. Future neuropsychological and neuroimaging studies should test the same groups of subjects on both tasks.
Evidence against a highly specific link between inhibition and RIFG has been accumulating. Impairments in response inhibition have been reported in patients with dorsomedial frontal damage [9
]. A recent fMRI study associated greater activation in left superior frontal gyrus (BA 8) with more efficient response inhibition [58
]. Importantly, a new meta-analysis [59
] classified Go/NoGo tasks as either simple (the NoGo stimulus was always the same) or complex (the NoGo stimulus changed depending on context). Common to both task types was greater activation in the pre-supplementary motor area (SMA) during response inhibition (see also Fig. ), but activation in right dorsolateral PFC was observed only in the complex tasks, which made demands on working memory. As a new theoretical framework incorporating these findings develops, the emerging emphasis is likely to be on a well-circumscribed but anatomically distributed frontal lobe inhibitory control network. A core element in this network includes pre-SMA circuits, with recruitment of additional frontal (and posterior) regions perhaps varying according to task demands [35
Returning to the idea of a unified hypothesis of LIFG function, a key commonality involves restraining alternatives in a given context that includes motor, semantic, mnemonic, or linguistic alternatives. Semantic selection [30
] involves inhibition of unselected alternatives; speech production has both cognitive and motor control components, possibly tapping into general purpose selection/inhibition mechanisms [60
]; vocal control for speech might share evolutionary origins with manual motor control for gesturing [61
]; left hemisphere dominance for action might have implications for motor response control [62
]; rejection of new items in a recognition memory task might involve inhibition of any tendency to generate a yes response [63
]. The present data add the inhibition of dominant motor response tendencies to this roster. Another possibility to consider, for posterior LIFG at least, is that subvocalization is actually a critical aspect of many complex cognitive activities, as speculated in a review article on the role of inner speech in self-reflective processing [64
]. Although beyond the scope of this particular paper, ongoing research is investigating a parcellation of LIFG cognitive control functions along the anterior-posterior dimension [65
Interestingly, OFC patients did not commit a greater number of false alarm errors, contradicting a general characterization as impulsive in all behavioral domains. This lends a degree of anatomical specificity to the LIFG inhibitory control impairment. On the other hand, all of the OFC patients suffered head trauma, and this finding diverges from some results in TBI patients [54
], but not others [66
]. This latter study did not find a deficit in the random SART in a group of 26 TBI patients [66
]. While the differences in the time post-injury and differences between standard Go/NoGo and SART procedures may account for the spared performance in OFC patients, the current finding is of theoretical interest in relation to OFC function.
The present findings have significance from a clinical standpoint as well. A number of different psychiatric disorders have been described as dysfunctions of "frontal" inhibitory processes that involve increases in impulsive behavior, motivating investigators to explore which frontal areas might be dysfunctional in various psychiatric conditions. The Go/NoGo task has been used by various researchers to investigate the biological basis of motor impulsiveness [67
], mainly relying on neuroimaging and electrophysiological data [68
]. Human lesion studies with precise neuroanatomical characterization of the PFC regions underlying these different types of disinhibition can contribute to a better understanding of the neurobiological correlates of disorders such as ADHD, alcoholism, drug abuse, schizophrenia, and obsessive-compulsive disorder.