Two alternative mechanisms have been proposed to account for the redundant target effect. “Statistical facilitation” posits that the observed facilitation in reaction time to redundant targets occurs because redundant targets activate multiple, independent, parallel processing channels. Each channel processes one of the redundant targets, and its speed varies from trial to trial as a stochastic process. Consequently, reaction time in a given trial reflects the processing time of whichever channel happened to be faster on that trial, causing the average reaction time to multiple redundant targets to be shorter than the average reaction time for any of the two channels alone. By analogy, statistical facilitation is sometimes described as a ‘horse race’, where the winner initiates the response [6]. This mechanism assumes that there is no interaction among the channels. “Coactivation models”, on the other hand, posit that engaging parallel channels results in a multiplicative activation, or interaction of channels, prior to response initiation [7], [8]. This pooled activation, thus, yields a faster initiation of the motor response. The original coactivation model was abstract and did not take into consideration the underlying neural architecture. Recently, however, coactivation has been typically interpreted as indicating neural summation [9]–[14].

Miller [8] proposed a mathematical test (see Methods) to differentiate between statistical facilitation and coactivation accounts of the redundant target effect. His equation establishes the maximum difference between reaction times to single versus redundant presentations for which statistical facilitation can adequately explain the redundant target effect. In practice, this limit is exceeded on some trials, evidence that some other mechanism must be responsible for response facilitation, at least in those trials. [See, for example, 2,3,15,16]. It is important to note that when this limit is exceeded, statistical facilitation is ruled out as an explanation of the redundant target effect. However, when the limit is not exceeded, coactivation cannot be ruled out.

Theoretically, the functional locus of the redundant target effect may occur at a sensory, central (cognitive), or motor stage of processing. The empirical data are mixed. Many studies have ruled out that it occurs at either very early perceptual or late motor stages of processing. The redundant target effect is typically greater with bimodal stimuli (for example, visual-tactile) than with unimodal stimuli (for example, visual-visual) [15]. These instances provide evidence that the effect occurs after early sensory processing, when information from different modalities is integrated (Miller, 1982). On the other hand, two event-related potential studies reported an early locus of the redundant target effect [10], [17]. In both studies, earlier peak P1 latencies were observed for redundant visual stimuli compared to single visual stimulus trials. Another event-related potential study used redundant audio-visual stimuli and reported early audio-visual interactions consistent with sensory processing [18]. Likewise, Cavina-Pratesi et al. [19] addressed whether the redundant target effect occurs as late as a motor stage of processing, using a task where subjects had to withhold responses on trials with stop-signals. Redundant stop-signals were more effective than single stop-signals in inhibiting motor responses. Similarly, responses to redundant stimuli were more difficult to inhibit compared to single stimuli. The effects of redundant signals on motor responses in these two stop-signal experiments suggest that the redundant target effect occurs at a late, pre-motor, stage, prior to late ballistic motor output [19].

As for the anatomical locus of the effect, reports have suggested that it occurs in either extrastriate or premotor regions, in line with information processing accounts of the effect. The event-related potential data suggest that the redundant target effect is detectable in the extrastriate cortex, but the poor spatial resolution of the event-related potential technique makes it difficult to precisely identify sources of influence [20]. A single-trial fMRI study, on the other hand, found increased blood oxygen-level dependent (BOLD) signal in the left and right dorsal premotor cortex and right intraparietal sulcus for redundant compared to single stimulus targets [16]. The premotor activations reported in that study support a later, motor, stage of processing.

Given the conflicting reports in the literature, the critical brain regions associated with parallel processing of stimuli remain a matter of investigation. Importantly, previous studies have only considered redundant versus single target conditions without distinguishing between performance explained by statistical facilitation and coactivation. Therefore an investigation of the neural locus of coactivation must look at these special trials separately.

The bilateral display used in this paradigm introduces an interhemispheric component to the task. Somewhat paradoxically, split brain and acallosal subjects often exhibit redundant target effects much larger than those in normal subjects which often exceed the boundary predicted by statistical facilitation [9], [11], [14]. These results suggest, counterintuitively, a greater degree of interhemispheric interaction in the absence of the corpus callosum and that, in the normal brain, the corpus callosum may serve to inhibit interhemispheric interaction [11]. Analysis of the functional connectivity of brain regions associated with the redundant target effect could prove useful in determining the role of interhemispheric connections in mediating it. Functional connectivity analyses allow us to examine the temporal cross-correlation of brain regions associated with activity in a seed region and are presumed to reflect structural connectivity between functionally related regions [21]. This analysis is complementary to task activation maps because it describes regions that follow the temporal sequence of information processing rather than the regions that engage simultaneously.

Using this novel approach, we observed selective increase in BOLD signal centered in the right angular gyrus (temporo-parietal junction) for coactivation compared to non-coactivation redundant target trials. Recall that these trials are visually identical yet on some trials, the speeds of responses are so fast that they exceed the upper limit of statistical facilitation. Our results fit well with the larger body of work on the redundant target effect which concluded that the functional locus of the redundant target effect is post-perceptual, but prior to ballistic motor output [10], [16], [19].

The right temporo-parietal junction is an established sensorimotor region important for spatial attention. From an information processing point of view, the right temporo-parietal junction is situated in a central ‘cognitive’ stage of processing. Damage in the region is associated with visuospatial neglect [22], [23] and transcranial magnetic stimulation induces hemiextinction when applied over the right temporo-parietal junction [24]. In terms of functional anatomy, the right temporo-parietal junction has been implicated as part of a larger right-lateralized ventral fronto-parietal system that also includes the middle and inferior frontal gyri and is activated during detection of behaviorally relevant, salient, unattended stimuli [25]. Indeed, our functional connectivity analyses revealed activity in the homologous, albeit a more restricted temporo-parietal junction region in the left hemisphere as well as the right middle temporal gyrus, right middle frontal gyrus and right inferior frontal, largely consistent with the proposed network.

The question occurs whether there exist trials where coactivation occurs, but which do not exceed the limit for statistical facilitation. Figure 4 suggests that the answer is “no”, at least in the right temporo-parietal junction. Non-coactivation trials do not activate this region at all (relative to rest). This indicates that the right temporo-parietal junction is active only during coactivation trials that exceed the limit for statistical facilitation. However, our data cannot exclude that coactivation may occur in other brain regions for trials not exceeding the limit of statistical facilitation.

How do identical stimuli result in such differences in reaction time? A schematic model of how this might occur is presented (Figure 6). Under single target conditions, the reaction time is determined by the processing speed of a single activated channel (Figure 6A). However, in redundant bilateral targets conditions, one of two possible scenarios occurs. In the typical redundant targets case, a statistical facilitation occurs where the faster of two independent processing streams determines the speed of response, resulting in faster average reaction times compared to the single target conditions (Figure 6B). On some redundant trials, however, a reaction time advantage beyond that predicted by statistical facilitation occurs (coactivation). On these trials, the two parallel processing streams operate at just the right delay for coactivation (Figure 6C), as suggested by previous data on callosal patients [14]. This would be due to intrinsic properties of oscillatory systems, as cortical neural systems tend to be (see ref. 14 for a full explanation of the model.) The probability of such a delay is much greater in the split-brain than in the normal brain.

Split-brain patients often exhibit a redundant target effect that exceeds statistical facilitation [9], [11], [14], [26]. Reuter-Lorenz et al. (1995) proposed an ‘and-or’ model to explain the paradoxically enhanced redundant target effect in simple reaction time that exceeds statistical facilitation observed in split-brain patients. They posited that coactivation occurs at a response selection stage and acts as an ‘and’ gate that requires input from both hemispheres. Under this scheme, a redundant target effect is due to the release of a chronic inhibition on motor pathways under bilateral redundant stimuli conditions. This model was ruled out on empirical grounds [14]. Corballis et al. [26] offered an alternative account where the corpus callosum normally serves to inhibit interhemispheric signals, while in the split brain, interhemispheric inhibition is released. Thus, in the healthy brain, performance that exceeds the limit is rare and intermittent, presumably due to removal of inhibition during rare and limited fluctuations in interhemispheric inhibition. By contrast, in the split brain, this left hemisphere inhibition is constantly released because the corpus callosum is sectioned, resulting in the paradoxical ‘hyper’ redundant target effect.

Our own data showing the right lateralized activation for reaction times that exceed the statistical facilitation limit and the right lateralized functional network observed here for bilateral trials indeed suggest a role for the corpus callosum in mediating the redundant target effect. The dynamic modulation of interhemispheric inhibition may account for why healthy subjects' performance is typically explained by statistical facilitation, while there are also infrequent, yet clear, instances of coactivation in the healthy brain.

A few studies have investigated the redundant target effect in visuomotor integration of redundant stimuli using fMRI. However, even though these studies used the same general paradigm, they investigated different aspects of the visuomotor transformations required by the task. For instance, Iacoboni et al. [14] studied individual differences in the redundant target effect in patients with callosal agenesis. They interpreted their findings as suggesting different forms of cortico-subcortical interactions in callosal agenesis patients with and without performance that exceeds statistical facilitation. Iacoboni & Zaidel [16] used single-trial fMRI to look at the neural correlates of the redundant target effect in the healthy brain. Their results revealed increased BOLD signal for responses to bilaterally redundant stimuli compared to unilateral single stimuli in the precentral gyri bilaterally, left postcentral gyrus, and right intraparietal sulcus. Here, we investigated yet another aspect of the performance at this task, namely those instances of coactivation in the healthy brain. Different brain regions seem associated with these different aspects of the performance at the task, suggesting a fair amount of regional specialization for the visuomotor transformations that occur.