- Email: [email protected]
Spatial and temporal dissociation in prefrontal cortex during action execution
www.elsevier.com/locate/ynimg NeuroImage 23 (2004) 1186 – 1191
Spatial and temporal dissociation in prefrontal cortex during action execution Michael D. Hunter,a,* Russell D.J. Green,a Iain D. Wilkinson,b and Sean A. Spencea a
Sheffield Cognition and Neuroimaging Laboratory, Academic Clinical Psychiatry, Division of Genomic Medicine, University of Sheffield, Sheffield, UK b Academic Unit of Radiology, University of Sheffield, Sheffield, UK Received 5 April 2004; revised 2 July 2004; accepted 12 July 2004 Available online 7 October 2004
It is widely accepted that dorsolateral prefrontal cortex (DLPFC) is activated at the time of action generation in humans. However, the previous functional neuroimaging studies that have supported this hypothesis temporally integrated brain dynamics and therefore could not demonstrate when DLPFC underwent activation relative to the emergence of voluntary behavior. Data that are time-locked to the instant of voluntary action execution do not reveal DLPFC activation at that moment. Rather, activated foci are seen at the frontal poles. We investigated this apparent conundrum through three differentially constrained experiments, utilizing functional magnetic resonance imaging to identify those prefrontal areas exhibiting functional change at the moment of spontaneous action execution. We observed profound functional dissociation between anterior and dorsolateral regions, compatible with their involvement at different points during the temporal evolution of action: bilaterally the frontal poles activated at the moment of execution, while simultaneously (and relative to a prior activation state) left DLPFC ddeactivated.T D 2004 Elsevier Inc. All rights reserved. Keywords: Dorsolateral prefrontal cortex; Functional neuroimaging; Timelocked data; Willed action
Introduction A characteristic of higher organisms is their ability to select which behaviors to execute (or withhold) and when to make such selections. In humans, this capacity has equated with the dwillT (Dilman, 1999) and is subserved by executive centers in the prefrontal cortex (Spence et al., 2002). A body of functional neuroimaging work has implicated dorsolateral prefrontal cortex * Corresponding author. Sheffield Cognition and Neuroimaging Laboratory, Academic Clinical Psychiatry, Division of Genomic Medicine, University of Sheffield, The Longley Centre, Norwood Grange Drive, Sheffield S5 7JT, UK. Fax: +44 114 226 1522. E-mail address: [email protected] (M.D. Hunter). Available online on ScienceDirect (www.sciencedirect.com.) 1053-8119/$ - see front matter D 2004 Elsevier Inc. All rights reserved. doi:10.1016/j.neuroimage.2004.07.047
(DLPFC) in the free selection of response behaviors including random number (Jahanshahi et al., 2000) and letter (de Zubicaray et al., 1998) generation, word stem completion (Desmond et al., 1998), verbal fluency (the generation of word sequences constrained by orthographic or semantic rules; Frith et al., 1991; Phelps et al., 1997; Spence et al., 2000), and willed motor action (the selection of which movements to perform from a set of exemplars; Frith et al., 1991; Hyder et al., 1997; Spence et al., 1998). When brain imaging data are integrated over minutes, DLPFC is also seen to be activated when subjects self-initiate motor action (i.e., when they choose when to execute movements; Jenkins et al., 2000). In an ecologically valid model of spontaneous motor action, we have previously allowed these dwhichT and dwhenT components of response selection to coexist simultaneously; our minimally constrained, event-related, functional magnetic resonance imaging (fMRI) experiment identified prefrontal regions that were activated at the moment of action execution (Hunter et al., 2003). In that experiment, our data were time-locked to the execution of an explicit action and revealed activated foci located at the frontal poles, in Brodmann area (BA) 10. We were initially surprised not to find activation of DLPFC. However, we reasoned that our approach differed from previous studies in a critical respect: the utilization of event-related fMRI to study motor action which is entirely self-initiated (c.f., Rowe et al., 2000). Previous studies had used dblockedT fMRI (series of individual events modeled as a single, prolonged, block event; Hyder et al., 1997) or positron emission tomography (PET; not involving a time-series model; Frith et al., 1991; Jenkins et al., 2000; Spence et al., 1998) to investigate the dwhichT or dwhenT aspects of response selection. They would have identified those brain regions activated within a block (fMRI) or scanning session (PET), regardless of the temporal relationship between such activation and individual, intrablock (session), motor-action events. Crucially, such studies could not reveal when DLPFC underwent activation relative to the moment of action execution. The apparent discrepancy between our event-related work (DLPFC activation absent) and the blocked fMRI/PET work of others (DLPFC activation present) suggested a new hypothesis: that dorsolateral prefrontal cortex is involved
M.D. Hunter et al. / NeuroImage 23 (2004) 1186–1191
during the production of spontaneous motor action, but its involvement is pivotally before the act, so that it undergoes a relative ddeactivationT at the moment of action execution (relative to an earlier, activated state). Materials and methods Subjects We used event-related fMRI to investigate healthy, right-handed (Oldfield, 1971), male volunteers. Six subjects participated in an initial experiment, and six different subjects participated in two further replication experiments. Subjects were aged 25 F 3 years (Experiment 1) and 29 F 5 years (Experiments 2 and 3). No subject had any clinically significant history of medical disorder. All subjects gave informed consent to participate in the study, which was approved by the North and South Sheffield Research Ethics Committees. Motor-action paradigms Inside the scanner, subjects (with their eyes closed) were required to generate sequences of spontaneous motor action by deciding when to perform button-pressing movements with the digits of their right hand. In Experiments 1 and 2, subjects were also free to choose, for each motor-action event, which of two adjacent buttons to press with their right index or middle fingers. Experiment 1 differed from Experiment 2, however, in that we requested that subjects not produce motor actions more frequently than every 10–12 s, while emphasizing that the decision regarding when to act was that of the individual subject (and that they should not estimate time by counting). Our rationale for the temporal dpart-constraintT was to maximize statistical power (on functional image analysis; below) by maintaining an implicit physiological baseline in the fMRI time series. Because Experiments 2 and 3 sought to replicate findings in those anatomical regions of interest identified by Experiment 1, their statistical power requirements were reduced. Hence, we were able to increase the ecological validity of these experiments by allowing temporally dunconstrainedT responding. Experiment 3 differed from Experiment 2 in that subjects were restricted to utilizing a single button and their index finger to produce sequences of motor action. Hence, Experiment 1 contained a dwhichT and dwhenT choice, subject to temporal dpart-constraint.T Experiment 2 also contained a dwhichT and dwhenT decision, but was temporally dunconstrained.T Experiment 3 contained a dwhenT choice only and was temporally dunconstrained.T The intrascanner response box was optically connected, via an interface (New Micros Inc., Dallas, TX), to an Apple Macintosh G3 computer. All responses were automatically recorded. In Experiments 1 and 2, for each subject, we calculated an information score based upon the frequency with which the four possible second-order diads (i.e., button A followed by button B; AA; BB; BA) appeared in the sequences that were generated (Baddeley, 1966). The information score describes the amount of information, in dbits,T that is transmitted by a sequence. This score was then transformed into percentage redundancy, where a truly random sequence is 0% redundant, and a stereotypic sequence is 100% redundant. In the temporal domain, interresponse interval data were also calculated.
1187
Functional image acquisition Event-related fMRI was performed on a 1.5-T system (Eclipse, Philips Medical Systems, Cleveland, OH) at the University of Sheffield. At each of 600 time points (Experiment 1) and 450 time points (Experiments 2 and 3), 10 noncontiguous transverse images were acquired, covering most of the frontal and parietal cortices (gradient-recalled echo-planar imaging [EPI]; repetition time = 1 s; echo time = 40 ms; slice thickness = 6 mm; interslice gap = 1 mm; field of view = 240 mm; in-plane matrix = 128 128). Experiment 1 lasted 10 min, Experiments 2 and 3 lasted 7 min 30 s; we utilized less functional time points in the latter experiments because of storage considerations resulting from the acquisition of consecutive, large data sets from the same subjects on the same day. Functional image analysis Group analyses were carried out using statistical parametric mapping in SPM99 (http://www.fil.ion.ucl.ac.uk/spm). Images were realigned, spatially normalized (Friston et al., 1995) and smoothed with a Gaussian kernel of 6-mm full width at halfmaximum. Motor-action events were modeled by a canonical hemodynamic response function (HRF) and its temporal derivative. High and low pass filters were applied to the blood oxygenation level dependent (BOLD) response data. SPM99 combines the General Linear Model and Gaussian Field Theory to draw statistical inferences from BOLD response data regarding deviations from the null hypothesis in threedimensional brain space. We used a fixed-effects model to assess the BOLD response associated with motor-action events. Such a model can allow qualitative inferences to be drawn regarding normal functional anatomy, typical of the population being studied (Friston et al., 1999). In this case, we were solely interested in a qualitative factor: the direction of an effect (activation or deactivation). The identification of brain areas where the hemodynamic series correlated with our model (baseline was implicit) produced a parametric brain map in the stereotactic space of the Montreal Neurological Institute (MNI; Evans et al., 1993). In Experiment 1, we used a conservative statistical threshold ( P b 0.05, corrected for multiple comparisons in the whole brain) that does not take into account prior anatomical hypotheses regarding the likely location of activation/ deactivation maxima. Experiments 2 and 3 explicitly tested hypotheses generated by Experiment 1 and were, therefore, appropriately thresholded at P b 0.001, uncorrected (the conceptual principle that allowed us to permit temporally unconstrained responding in those experiments, with consequent increased convolution of the dbaselineT and reduced statistical power). In view of such varying power between the experiments, we emphasize that these experiments were designed to stand alone as a series of replications and not to be contrasted across any particular experimental factor. In addition to group average analyses and to permit population level inference, we performed conjunction analyses across subjects, thresholded at conjoint P b 0.001, uncorrected, identifying those foci common to each subject in the three experiments (results marked with an asterisk in Table 1). According to Friston et al. (1999), the probability of a conjunction across n subjects given that the proportion (c) of the population exhibiting the effect is less than a certain proportion (c c) that defines typicality (i.e., the null hypothesis;
1188
M.D. Hunter et al. / NeuroImage 23 (2004) 1186–1191
Table 1 Similar neuroimaging and behavioral findings in separate, differentially constrained experiments Neuroimaging data Activation Left frontal pole (Experiment 1)* (Experiment 2) (Experiment 3)* Right frontal pole (Experiment 1)* (Experiment 2)* (Experiment 3) Supplementary motor area (Experiment 1)* (Experiment 2)* (Experiment 3)* Left primary motor cortex (Experiment 1)* (Experiment 2)* (Experiment 3)* Deactivation Left dorsolateral prefrontal cortex (Experiment 1)* (Experiment 2)* (Experiment 3)*
Talairach co-ordinates (x, y, z)
t Statistic
38, 55, 12 32, 50, 20 28, 50, 20
6.75 3.52 4.08
34, 55, 17 38, 50, 21 42, 53, 14
7.13 3.97 3.72
2, 12, 44 2, 6, 51 2, 6, 44
10.47 8.90 10.41
32, 14, 63 34, 13, 60 30, 18, 64
14.35 10.07 13.46
53, 28, 13 50, 24, 12 40, 32, 19
7.34 4.39 3.81
Behavioral data
Interresponse interval (s)
Redundancy (%)
(Experiment 1) (Experiment 2) (Experiment 3)
19.05 F 5.04 8.28 F 4.96 10.41 F 5.38
3.80 F 6.15 5.32 F 4.49
Group average data are presented. Two-tailed statistical thresholds are P b 0.05, corrected for multiple comparisons in the whole brain (Experiment 1) and P b 0.001, uncorrected (Experiments 2 and 3). * Foci also present in conjunction analyses, across all subjects, in each experiment (conjoint P b 0.001, uncorrected). The average dactivationT data and behavioral data (mean F standard deviation) in Experiment 1 are from Hunter et al., 2003, cited to demonstrate replication.
in our case that less than half of the population shows the effect and c c = 0.5) is
Fig. 1 demonstrates the clear functional dissociation between this area and the bilateral activation foci seen at the frontal poles (BA 10), which we observed in our original analysis (Hunter et al., 2003). Fig. 2 shows averaged hemodynamic series from the three main regions of interest (right and left frontal poles, and left DLPFC) in a time frame encompassing the period between 7-s preaction and 4-s postaction. The deactivation of left DLPFC is evident. Experiments 2 and 3 also demonstrated deactivation of left DLPFC (BA 46) and replicated our earlier findings (Hunter et al., 2003) of activation in the frontal poles (BA 10), supplementary motor area and primary motor cortex (Table 1). Thus, in separate, differentially constrained, experiments we were able to demonstrate a consistent functional dissociation at the moment of spontaneous motor-action execution between anterior (activated) and dorsolateral (deactivated) prefrontal regions. We also observed behavioral consistency (Table 1). In the spatial (dwhichT) domain of response selection, Experiment 2 yielded redundancy data comparable with our previous findings (Hunter et al., 2003); subjects utilized buttons A and B to generate sequences of high randomness content (Baddeley, 1966). In the temporal (dwhenT) domain, the mean interresponse interval and its variance were similar in Experiments 2 and 3 (both temporally dunconstrainedT). Modeling ddeactivationT as prior dactivationT In Experiment 1, we were able to further deconstruct the time course of left DLPFC because the mean interresponse interval approached 20 s. Fig. 2 shows that the mean hemodynamic response in left DLPFC peaked at 4 s before action execution. The canonical HRF used in SPM99 models peak hemodynamic response at 5 s following the deventT (with a degree of variation conferred by the first temporal derivative). Hence, we reasoned that if deactivation of left DLPFC was an expression of its prior activation (rather than a negative deviation from a flat dbaselineT), then the same area that was deactivated in our model (time-locked to the moment of action execution) should be activated in an alternative model: time-locked to an event defined as occurring 9 s before action execution.
½ad ð1 cc Þ þ bd cc n Where a is the specificity (i.e., [0.001]1/6 in the current example, as the conjoint P threshold in SPM99 is 0.001 and there are six subjects) and b is the sensitivity (paradoxically, the most conservative approach is to assume high sensitivity, because it is very unlikely that an insensitive test could detect an effect in n subjects). Hence, conjoint P = 0.001 in a group of six corresponds to P = 0.05 in the population for the effect being typical, that is, occurring more often than not, if sensitivity = 90%. For the purposes of reporting and neuroanatomical labeling, the coordinates of significant areas of activation were transformed from MNI space into the stereotactic space of Talairach and Tournoux (1988). Results In Experiment 1, we observed deactivation of left DLPFC (BA 46), at the moment of spontaneous motor-action execution.
Fig. 1. Functional dissociation at the moment of spontaneous motor-action execution. Group average data from Experiment 1 are projected on the SPM99 dsmooth brainT (www.fil.ion.ucl.ac.uk), viewed from the front and left. Significant clusters ( P b 0.05, corrected) are color-coded. Areas that demonstrated a significant positive correlation with the hemodynamic model of neural activation, at the moment of motor-action execution, are shown in red (activation); negatively correlated areas, the most extensive of which was left DLPFC, are shown in blue (deactivation).
M.D. Hunter et al. / NeuroImage 23 (2004) 1186–1191
1189
Fig. 2. Temporal response dynamics in the periaction period. Averaged hemodynamic time series from selected regions of interest in Experiment 1: right frontal pole (34, 55, 17), left frontal pole (38, 55, 12), and left dorsolateral prefrontal cortex (53, 28, 13). These filtered data represent the first eigenvariate from spheres of radius 5 mm centered upon the activation maxima.
Group average analysis revealed that this was the case (BA 46; activation coordinates: 51, 30, 14 [c.f., deactivation: 53, 28, 13]; peak t statistic = 5.98; 66 voxels exceeded threshold P b 0.05, corrected for multiple comparisons in the whole brain). We also identified those areas that were ddeactivatedT at 9 s before action execution. The largest involved primary motor cortex (BA 4; 32, 14, 63; t = 7.97; 117 voxels exceeded P b 0.05, corrected). This was the same region that activated at the point of action execution (Table 1). Thus, the left DLPFC and motor cortex were effectively dmirror imagesT of each other. At the moment of action execution, motor cortex activated (while DLPFC deactivated), and at 9 s before execution (the approximate midpoint of the mean interresponse interval), DLPFC activated (while motor cortex deactivated). Discussion We have used event-related fMRI to demonstrate functional dissociation between distinct prefrontal regions at the moment of spontaneous motor-action execution. Specifically, we have shown that, while frontal pole regions are activated at the point of action execution, left DLPFC deactivates, a finding crucially different from those of previous studies. The area of left DLPFC deactivation that we observed, in Experiment 1, was maximal at a relatively inferior location, close to premotor cortex (BA 45/46). In Experiment 2, the deactivation maximum was located at almost exactly the same coordinates (Table 1), perhaps reflecting the methodological similarity between Experiments 1 and 2 (both of which permitted dwhichT and dwhenT response selection). In Experiment 3, the deactivation maximum was more anteromedial, close to some of those coordinates for left DLPFC activation maxima which have been reported in the previous literature (Frith et al., 1991; Hyder et al., 1997; Spence et al., 1998). Nevertheless, we are cautious in comparing such coordinates in view of the many potential sources of variation between respective studies: individual subject anatomy, resolution
of imaging technique, and data preprocessing, not least image smoothing. However, the qualitative points remain: first, that previous work described DLPFC activation, whereas our eventrelated experiments demonstrate deactivation at the point of action, and second, that the provision of a dwhenT choice in the psychological paradigm was sufficient to elicit this effect. The existence of a dwhichT response selection parameter may modulate the deactivation process but is not a prerequisite. In this paper, the term ddeactivationT means reduced fMRI (BOLD) signal. For several reasons, the term ddeactivationT has been controversial in fMRI research. First, the BOLD signal is always a relative measure (Arthurs and Boniface, 2002). In the current study, we found left DLPFC to be deactivated at the moment of motor-action execution relative to activated areas at the frontal poles and relative to its own, prior activation. Secondly, fMRI utilizes an indirect (vascular) marker of neural activity. We must, therefore, have some understanding of the neurovascular coupling between neuron and BOLD signal to interpret both increases and decreases of the latter. Simultaneously acquired fMRI and in situ electrophysiological data have helped to elucidate the nature of this relationship (Logothetis et al., 2001). On the basis of such work, the BOLD signal is currently thought to represent a positive correlate of the neuronal local field potential (a measure of synaptic activity). In straightforward terms, increased BOLD signal corresponds to increased synaptic activity (which, itself, may correspond to increased neuronal firing). However, neural dactivityT at this level may represent excitation or inhibition (at the neurotransmitter level). It has been suggested that inhibition is associated with lower metabolic demand than excitation, because there are fewer cortical inhibitory neurons, which are, additionally, more efficient than their excitatory equivalents (Arthurs and Boniface, 2002). Supporting evidence for this hypothesis comes from a combined fMRI/ transcranial magnetic stimulation paradigm, which demonstrated that, unlike excitation, inhibition of the motor cortex did not evoke measurable change in the BOLD signal (Waldvogel et al., 2000). Although it has also been argued (Tagamets and Horwitz, 2001) that
1190
M.D. Hunter et al. / NeuroImage 23 (2004) 1186–1191
inhibition might raise the BOLD signal, in the context of minimal local excitation, one interpretation of our results could be that relative deactivation of left DLPFC, at the moment of action execution (relative to a previously activated state), reflects reduced excitation in that region at that moment. The bilateral activation at the frontal poles that we observed was almost identical to the result of a PET study examining prospective memory (Burgess et al., 2001). In that study, subjects were expecting to perform a cued motor task at infrequent intervals. Critically, by comparison with a baseline condition, it was the expectation of action execution (the online maintenance of dwhat to do,T whether or not execution followed) that was associated with such activation. Hence, it is possible that, at the moment of action execution, anterior prefrontal activation is related to the monitoring of motor intention (dpress a button at willT) and might serve to drefreshT prospective memory of the required motor behavior. In psychological terms, such activity might therefore correspond to an dimplementation intentionT (Sheeran, 2002). However, and more importantly, we found evidence for deactivation of left DLPFC at the moment of action execution, relative to its prior activated state. This suggests early involvement of DLPFC during the temporal evolution of action initiation and execution, an obligate effect congruent with the region’s functional anatomy: its role in the free selection of response behaviors (Petrides and Milner, 1982) as a component of the supervisory attentional system (Shallice and Burgess, 1996). Although DLPFC also supports working memory (Postle et al., 1999), its involvement during action selection is distinct and dissociable (Rowe et al., 2000). Unlike Rowe et al. (2000), we did not find evidence of DLPFC activation at the moment of action execution. However, that group’s paradigm involved cued dselection from memoryT (c.f., spontaneous motor-action execution), and their DLPFC activation foci were in the right hemisphere (c.f., our deactivation foci in the left, at a relatively inferior location, maximal adjacent to premotor cortex). These differences are likely to reflect functional specificities within regions of the prefrontal cortex, differentially probed by distinct neuroimaging paradigms. Hence, we have a hypothesis regarding the dcausalT direction of willed action, namely, that the supraordinate precedes the subordinate (in this case, that left DLPFC activates before premotor and motor cortices). This hypothesis is supported by electrophysiological and magnetoencephalographic data that indicate prefrontal and premotor activation early in the evolution of motor behavior (Goldman-Rakic et al., 1992; Nishitani and Hari, 2000). Yet, we are also demonstrating temporal dissociation within higher, executive regions, since left DLPFC is activating earlier than the frontal poles and supplementary motor area. Also, we have previously reported that, among those executive regions activated proximal to the moment of action execution, the left frontal pole (BA 10) and rostral supplementary motor area (BA 6) have significantly shorter BOLD response latencies than primary motor cortex (Hunter et al., 2003), a difference that could correspond to the fMRI equivalent of the electrophysiological readiness potential but which does not amount to the profound difference between deactivated DLPFC and other activated brain regions that we report here. Hence, using eventrelated fMRI, we have discerned a sequence from the activation and subsequent deactivation of left DLPFC through the activations of the frontal poles and supplementary motor area, to the activation of motor cortex—an evolution of volition: voluntary behavior. Regarding the biological significance of these changes in signal, Ball et al. (1999) have used high-resolution EEG, combined
with fMRI, to show that the expected readiness potential originating in the supplementary motor area during the emergence of voluntary action actually undergoes a sharp reduction immediately before its execution. Those authors argued that such a reduction in signal may trigger the motor act by releasing the primary motor cortex from putative inhibition. It is conceivable (though conjectural) that our observed reduction in DLPFC activation in the moments before action could similarly act as a release signal for premotor cortex. Regarding subjective phenomenology, we acknowledge that we could not know what our subjects were dthinkingT while they performed our experiments, nor do we know the latency between their decisions to act and the actions themselves. However, we need not invoke a direct correspondence between the physiology of motor execution and the phenomenology of intention (as our experiments are designed to address the biology of action). Nevertheless, it has been suggested that there may be a robust coupling of the experience (phenomenology) of will and specific, identified behaviors on the part of the agent even when they are temporally dissociated by as much as 5 s (Wegner, 2002). Also, crucially, conscious intention has been shown to follow electrophysiological evidence of action initiation (Libet et al., 1983). Hence, our data inform the physiology of action, although not necessarily its subjective phenomenology. We should not equate left DLPFC activation with the moment of perceived subjective choice (even if it is the moment of neural choice). While our results are entirely compatible with earlier PET and blocked fMRI studies that identified the involvement of DLPFC in response selection (Frith et al., 1991; Hyder et al., 1997; Jenkins et al., 2000; Spence et al., 1998), they significantly refine their interpretation. In those studies, DLPFC activation was revealed because of its temporal integration over the scanning window (i.e., it occurred at some point during the action tasks involved). Hence, a parsimonious (but incorrect) interpretation might have been that DLPFC was activated at the moment of action execution. However, crucially, our studies demonstrate that this is not the case and that, at the moment of spontaneous motor-action execution (relative to its prior state), DLPFC deactivates. Acknowledgments M.D.H. is supported by the Wellcome Trust. An investigatorled award to S.A.S. from Cephalon (UK) supported R.D.J.G. We thank our radiographer colleagues at the Academic Unit of Radiology, University of Sheffield, for assistance during image acquisition. References Arthurs, O.J., Boniface, S., 2002. How well do we understand the neural origins of the fMRI BOLD signal? Trends Neurosci. 25, 27 – 31. Baddeley, A.D., 1966. The capacity for generating information by randomisation. Q. J. Exp. Psychol. 18, 119 – 129. Ball, T., Schreiber, A., Feige, B., Wagner, M., Lucking, C.H., KristevaFeige, R., 1999. The role of higher-order motor areas in voluntary movement as revealed by high-resolution EEG and fMRI. NeuroImage 10, 682 – 694. Burgess, P.W., Quayle, A., Frith, C.D., 2001. Brain regions involved in prospective memory as determined by positron emission tomography. Neuropsychologia 39, 545 – 555.
M.D. Hunter et al. / NeuroImage 23 (2004) 1186–1191 de Zubicaray, G.I., Williams, S.C., Wilson, S.J., Rose, S.E., Brammer, M.J., Bullmore, E.T., et al., 1998. Prefrontal cortex involvement in selective letter generation: a functional magnetic resonance imaging study. Cortex 34, 389 – 401. Desmond, J.E., Gabrieli, J.D.E., Glover, G.H., 1998. Dissociation of frontal and cerebellar activity in a cognitive task: evidence for a distinction between selection and search. NeuroImage 7, 368 – 376. Dilman, I., 1999. Free Will: A Historical and Philosophical Introduction. Routledge, London. Evans, A.C., Collins, D.L., Mills, S.R., Brown, R.D., Kelly, R.L., Peters, T.M., 1993. 3D statistical neuroanatomical models from 305 MRI volumes. Proc.-IEEE Nucl. Sci. Symp. Med. Imaging Conf. 108, 1877 – 1878. Friston, K.J., Ashburner, J., Frith, C.D., Poline, J.B., Heather, J.D., Frackowiak, R.S.J., 1995. Spatial registration and normalization of images. Hum. Brain Mapp. 2, 165 – 169. Friston, K.J., Holmes, A.P., Worsley, K.J., 1999. How many subjects constitute a study? NeuroImage 10, 1 – 5. Frith, C.D., Friston, K., Liddle, P.F., Frackowiak, R.S.J., 1991. Willed action and the prefrontal cortex in man: a study with PET. Proc. R. Soc. London, B. 244, 241 – 246. Goldman-Rakic, P.S., Bates, J.F., Chafee, M.V., 1992. The prefrontal cortex and internally generated motor acts. Curr. Opin. Neurobiol. 2, 830 – 835. Hunter, M.D., Farrow, T.F.D., Papadakis, N.G., Wilkinson, I.D., Woodruff, P.W.R., Spence, S.A., 2003. Approaching an ecologically valid functional anatomy of spontaneous bwilledQ action. NeuroImage 20, 1264 – 1269. Hyder, F., Phelps, E.A., Wiggins, C.J., Labar, K.S., Blamire, A.M., Shulman, R.G., 1997. bWilled action:Q a functional MRI study of the human prefrontal cortex during a sensorimotor task. Proc. Natl. Acad. Sci. U. S. A. 94, 6989 – 6994. Jahanshahi, M., Dirnberger, G., Fuller, R., Frith, C.D., 2000. The role of the dorsolateral prefrontal cortex in random number generation: a study with positron emission tomography. NeuroImage 12, 713 – 725. Jenkins, I.H., Jahanshahi, M., Jueptner, M., Passingham, R.E., Brooks, D.J., 2000. Self-initiated versus externally triggered movements. II. The effect of movement predictability on regional cerebral blood flow. Brain 123, 1216 – 1228. Libet, B., Gleason, C.A., Wright, E.W., Pearl, D.K., 1983. Time of conscious intention to act in relation to cerebral activity (readinesspotential). The unconscious initiation of a freely voluntary act. Brain 106, 623 – 642. Logothetis, N.K., Pauls, J., Augath, M., Trinath, T., Oeltermann, A., 2001. Neurophysiological investigation of the basis of the fMRI signal. Nature 412, 150 – 157.
1191
Nishitani, N., Hari, R., 2000. Temporal dynamics of cortical representation for action. Proc. Natl. Acad. Sci. 97, 913 – 918. Oldfield, R.C., 1971. The assessment and analysis of handedness: the Edinburgh inventory. Neuropsychologia 9, 97 – 113. Petrides, M., Milner, B., 1982. Deficits on subject-ordered tasks after frontal- and temporal-lobe lesions in man. Neuropsychologia 20, 249 – 262. Phelps, E.A., Hyder, F., Blamire, A.M., Shulman, R.G., 1997. FMRI of the prefrontal cortex during overt verbal fluency. NeuroReport 8, 561 – 565. Postle, B.R., Berger, J.S., D’Esposito, M., 1999. Functional neuroanatomical double dissociation of mnemonic and executive control processes contributing to working memory performance. Proc. Natl. Acad. Sci. U. S. A. 96, 12959 – 12964. Rowe, J.B., Toni, I., Josephs, O., Frackowiak, R.S.J., Passingham, R.E., 2000. The prefrontal cortex: response selection or maintenance within working memory? Science 288, 1656 – 1660. Shallice, T., Burgess, P., 1996. The domain of supervisory processes and temporal organization of behaviour. Philos. Trans. R. Soc. London. Ser. B. 351, 1405 – 1412. Sheeran, P., 2002. Intention–behaviour relations: a conceptual and empirical review. In: Stroebe, W., Hewstone, M. (Eds.). Eur. Rev. Soc. Psychol. vol. 12. John Wiley and Sons Ltd, Chichester, UK, pp. 1 – 36. Spence, S.A., Hirsch, S.R., Brooks, D.J., Grasby, P.M., 1998. Prefrontal cortex activity in people with schizophrenia and control subjects. Evidence from positron emission tomography for remission of bhypofrontalityQ with recovery from acute schizophrenia. Br. J. Psychiatry 172, 316 – 323. Spence, S.A., Liddle, P.F., Stefan, M.D., Hellewell, J.S.E., Sharma, T., Friston, K.J., Hirsch, S.R., Frith, C.D., Murray, R.M., Deakin, J.F.W., Grasby, P.M., 2000. Functional anatomy of verbal fluency in people with schizophrenia and those at genetic risk. Focal dysfunction and distributed disconnectivity reappraised. Br. J. Psychiatry 176, 52 – 60. Spence, S.A., Hunter, M.D., Harpin, G., 2002. Neuroscience and the will. Curr. Opin. Psychiatry 15, 519 – 526. Tagamets, M.-A., Horwitz, B., 2001. Interpreting PET and fMRI measures of functional neural activity: the effects of synaptic inhibition on cortical activation in human imaging studies. Brain Res. Bull. 54, 267 – 273. Talairach, P., Tournoux, J.A., 1988. A Stereotactic Co-planar Atlas of the Human Brain. Thieme, Stuttgart. Waldvogel, D., van Gelderen, P., Muellbacher, W., Ziemann, U., Immisch, I., Hallett, M., 2000. The relative metabolic demand of inhibition and excitation. Nature 406, 995 – 998. Wegner, D.M., 2002. The Illusion of Conscious Will. MIT Press, Cambridge, MA.
Spatial and temporal dissociation in prefrontal cortex during action execution Michael D. Hunter,a,* Russell D.J. Green,a Iain D. Wilkinson,b and Sean A. Spencea a
Sheffield Cognition and Neuroimaging Laboratory, Academic Clinical Psychiatry, Division of Genomic Medicine, University of Sheffield, Sheffield, UK b Academic Unit of Radiology, University of Sheffield, Sheffield, UK Received 5 April 2004; revised 2 July 2004; accepted 12 July 2004 Available online 7 October 2004
It is widely accepted that dorsolateral prefrontal cortex (DLPFC) is activated at the time of action generation in humans. However, the previous functional neuroimaging studies that have supported this hypothesis temporally integrated brain dynamics and therefore could not demonstrate when DLPFC underwent activation relative to the emergence of voluntary behavior. Data that are time-locked to the instant of voluntary action execution do not reveal DLPFC activation at that moment. Rather, activated foci are seen at the frontal poles. We investigated this apparent conundrum through three differentially constrained experiments, utilizing functional magnetic resonance imaging to identify those prefrontal areas exhibiting functional change at the moment of spontaneous action execution. We observed profound functional dissociation between anterior and dorsolateral regions, compatible with their involvement at different points during the temporal evolution of action: bilaterally the frontal poles activated at the moment of execution, while simultaneously (and relative to a prior activation state) left DLPFC ddeactivated.T D 2004 Elsevier Inc. All rights reserved. Keywords: Dorsolateral prefrontal cortex; Functional neuroimaging; Timelocked data; Willed action
Introduction A characteristic of higher organisms is their ability to select which behaviors to execute (or withhold) and when to make such selections. In humans, this capacity has equated with the dwillT (Dilman, 1999) and is subserved by executive centers in the prefrontal cortex (Spence et al., 2002). A body of functional neuroimaging work has implicated dorsolateral prefrontal cortex * Corresponding author. Sheffield Cognition and Neuroimaging Laboratory, Academic Clinical Psychiatry, Division of Genomic Medicine, University of Sheffield, The Longley Centre, Norwood Grange Drive, Sheffield S5 7JT, UK. Fax: +44 114 226 1522. E-mail address: [email protected] (M.D. Hunter). Available online on ScienceDirect (www.sciencedirect.com.) 1053-8119/$ - see front matter D 2004 Elsevier Inc. All rights reserved. doi:10.1016/j.neuroimage.2004.07.047
(DLPFC) in the free selection of response behaviors including random number (Jahanshahi et al., 2000) and letter (de Zubicaray et al., 1998) generation, word stem completion (Desmond et al., 1998), verbal fluency (the generation of word sequences constrained by orthographic or semantic rules; Frith et al., 1991; Phelps et al., 1997; Spence et al., 2000), and willed motor action (the selection of which movements to perform from a set of exemplars; Frith et al., 1991; Hyder et al., 1997; Spence et al., 1998). When brain imaging data are integrated over minutes, DLPFC is also seen to be activated when subjects self-initiate motor action (i.e., when they choose when to execute movements; Jenkins et al., 2000). In an ecologically valid model of spontaneous motor action, we have previously allowed these dwhichT and dwhenT components of response selection to coexist simultaneously; our minimally constrained, event-related, functional magnetic resonance imaging (fMRI) experiment identified prefrontal regions that were activated at the moment of action execution (Hunter et al., 2003). In that experiment, our data were time-locked to the execution of an explicit action and revealed activated foci located at the frontal poles, in Brodmann area (BA) 10. We were initially surprised not to find activation of DLPFC. However, we reasoned that our approach differed from previous studies in a critical respect: the utilization of event-related fMRI to study motor action which is entirely self-initiated (c.f., Rowe et al., 2000). Previous studies had used dblockedT fMRI (series of individual events modeled as a single, prolonged, block event; Hyder et al., 1997) or positron emission tomography (PET; not involving a time-series model; Frith et al., 1991; Jenkins et al., 2000; Spence et al., 1998) to investigate the dwhichT or dwhenT aspects of response selection. They would have identified those brain regions activated within a block (fMRI) or scanning session (PET), regardless of the temporal relationship between such activation and individual, intrablock (session), motor-action events. Crucially, such studies could not reveal when DLPFC underwent activation relative to the moment of action execution. The apparent discrepancy between our event-related work (DLPFC activation absent) and the blocked fMRI/PET work of others (DLPFC activation present) suggested a new hypothesis: that dorsolateral prefrontal cortex is involved
M.D. Hunter et al. / NeuroImage 23 (2004) 1186–1191
during the production of spontaneous motor action, but its involvement is pivotally before the act, so that it undergoes a relative ddeactivationT at the moment of action execution (relative to an earlier, activated state). Materials and methods Subjects We used event-related fMRI to investigate healthy, right-handed (Oldfield, 1971), male volunteers. Six subjects participated in an initial experiment, and six different subjects participated in two further replication experiments. Subjects were aged 25 F 3 years (Experiment 1) and 29 F 5 years (Experiments 2 and 3). No subject had any clinically significant history of medical disorder. All subjects gave informed consent to participate in the study, which was approved by the North and South Sheffield Research Ethics Committees. Motor-action paradigms Inside the scanner, subjects (with their eyes closed) were required to generate sequences of spontaneous motor action by deciding when to perform button-pressing movements with the digits of their right hand. In Experiments 1 and 2, subjects were also free to choose, for each motor-action event, which of two adjacent buttons to press with their right index or middle fingers. Experiment 1 differed from Experiment 2, however, in that we requested that subjects not produce motor actions more frequently than every 10–12 s, while emphasizing that the decision regarding when to act was that of the individual subject (and that they should not estimate time by counting). Our rationale for the temporal dpart-constraintT was to maximize statistical power (on functional image analysis; below) by maintaining an implicit physiological baseline in the fMRI time series. Because Experiments 2 and 3 sought to replicate findings in those anatomical regions of interest identified by Experiment 1, their statistical power requirements were reduced. Hence, we were able to increase the ecological validity of these experiments by allowing temporally dunconstrainedT responding. Experiment 3 differed from Experiment 2 in that subjects were restricted to utilizing a single button and their index finger to produce sequences of motor action. Hence, Experiment 1 contained a dwhichT and dwhenT choice, subject to temporal dpart-constraint.T Experiment 2 also contained a dwhichT and dwhenT decision, but was temporally dunconstrained.T Experiment 3 contained a dwhenT choice only and was temporally dunconstrained.T The intrascanner response box was optically connected, via an interface (New Micros Inc., Dallas, TX), to an Apple Macintosh G3 computer. All responses were automatically recorded. In Experiments 1 and 2, for each subject, we calculated an information score based upon the frequency with which the four possible second-order diads (i.e., button A followed by button B; AA; BB; BA) appeared in the sequences that were generated (Baddeley, 1966). The information score describes the amount of information, in dbits,T that is transmitted by a sequence. This score was then transformed into percentage redundancy, where a truly random sequence is 0% redundant, and a stereotypic sequence is 100% redundant. In the temporal domain, interresponse interval data were also calculated.
1187
Functional image acquisition Event-related fMRI was performed on a 1.5-T system (Eclipse, Philips Medical Systems, Cleveland, OH) at the University of Sheffield. At each of 600 time points (Experiment 1) and 450 time points (Experiments 2 and 3), 10 noncontiguous transverse images were acquired, covering most of the frontal and parietal cortices (gradient-recalled echo-planar imaging [EPI]; repetition time = 1 s; echo time = 40 ms; slice thickness = 6 mm; interslice gap = 1 mm; field of view = 240 mm; in-plane matrix = 128 128). Experiment 1 lasted 10 min, Experiments 2 and 3 lasted 7 min 30 s; we utilized less functional time points in the latter experiments because of storage considerations resulting from the acquisition of consecutive, large data sets from the same subjects on the same day. Functional image analysis Group analyses were carried out using statistical parametric mapping in SPM99 (http://www.fil.ion.ucl.ac.uk/spm). Images were realigned, spatially normalized (Friston et al., 1995) and smoothed with a Gaussian kernel of 6-mm full width at halfmaximum. Motor-action events were modeled by a canonical hemodynamic response function (HRF) and its temporal derivative. High and low pass filters were applied to the blood oxygenation level dependent (BOLD) response data. SPM99 combines the General Linear Model and Gaussian Field Theory to draw statistical inferences from BOLD response data regarding deviations from the null hypothesis in threedimensional brain space. We used a fixed-effects model to assess the BOLD response associated with motor-action events. Such a model can allow qualitative inferences to be drawn regarding normal functional anatomy, typical of the population being studied (Friston et al., 1999). In this case, we were solely interested in a qualitative factor: the direction of an effect (activation or deactivation). The identification of brain areas where the hemodynamic series correlated with our model (baseline was implicit) produced a parametric brain map in the stereotactic space of the Montreal Neurological Institute (MNI; Evans et al., 1993). In Experiment 1, we used a conservative statistical threshold ( P b 0.05, corrected for multiple comparisons in the whole brain) that does not take into account prior anatomical hypotheses regarding the likely location of activation/ deactivation maxima. Experiments 2 and 3 explicitly tested hypotheses generated by Experiment 1 and were, therefore, appropriately thresholded at P b 0.001, uncorrected (the conceptual principle that allowed us to permit temporally unconstrained responding in those experiments, with consequent increased convolution of the dbaselineT and reduced statistical power). In view of such varying power between the experiments, we emphasize that these experiments were designed to stand alone as a series of replications and not to be contrasted across any particular experimental factor. In addition to group average analyses and to permit population level inference, we performed conjunction analyses across subjects, thresholded at conjoint P b 0.001, uncorrected, identifying those foci common to each subject in the three experiments (results marked with an asterisk in Table 1). According to Friston et al. (1999), the probability of a conjunction across n subjects given that the proportion (c) of the population exhibiting the effect is less than a certain proportion (c c) that defines typicality (i.e., the null hypothesis;
1188
M.D. Hunter et al. / NeuroImage 23 (2004) 1186–1191
Table 1 Similar neuroimaging and behavioral findings in separate, differentially constrained experiments Neuroimaging data Activation Left frontal pole (Experiment 1)* (Experiment 2) (Experiment 3)* Right frontal pole (Experiment 1)* (Experiment 2)* (Experiment 3) Supplementary motor area (Experiment 1)* (Experiment 2)* (Experiment 3)* Left primary motor cortex (Experiment 1)* (Experiment 2)* (Experiment 3)* Deactivation Left dorsolateral prefrontal cortex (Experiment 1)* (Experiment 2)* (Experiment 3)*
Talairach co-ordinates (x, y, z)
t Statistic
38, 55, 12 32, 50, 20 28, 50, 20
6.75 3.52 4.08
34, 55, 17 38, 50, 21 42, 53, 14
7.13 3.97 3.72
2, 12, 44 2, 6, 51 2, 6, 44
10.47 8.90 10.41
32, 14, 63 34, 13, 60 30, 18, 64
14.35 10.07 13.46
53, 28, 13 50, 24, 12 40, 32, 19
7.34 4.39 3.81
Behavioral data
Interresponse interval (s)
Redundancy (%)
(Experiment 1) (Experiment 2) (Experiment 3)
19.05 F 5.04 8.28 F 4.96 10.41 F 5.38
3.80 F 6.15 5.32 F 4.49
Group average data are presented. Two-tailed statistical thresholds are P b 0.05, corrected for multiple comparisons in the whole brain (Experiment 1) and P b 0.001, uncorrected (Experiments 2 and 3). * Foci also present in conjunction analyses, across all subjects, in each experiment (conjoint P b 0.001, uncorrected). The average dactivationT data and behavioral data (mean F standard deviation) in Experiment 1 are from Hunter et al., 2003, cited to demonstrate replication.
in our case that less than half of the population shows the effect and c c = 0.5) is
Fig. 1 demonstrates the clear functional dissociation between this area and the bilateral activation foci seen at the frontal poles (BA 10), which we observed in our original analysis (Hunter et al., 2003). Fig. 2 shows averaged hemodynamic series from the three main regions of interest (right and left frontal poles, and left DLPFC) in a time frame encompassing the period between 7-s preaction and 4-s postaction. The deactivation of left DLPFC is evident. Experiments 2 and 3 also demonstrated deactivation of left DLPFC (BA 46) and replicated our earlier findings (Hunter et al., 2003) of activation in the frontal poles (BA 10), supplementary motor area and primary motor cortex (Table 1). Thus, in separate, differentially constrained, experiments we were able to demonstrate a consistent functional dissociation at the moment of spontaneous motor-action execution between anterior (activated) and dorsolateral (deactivated) prefrontal regions. We also observed behavioral consistency (Table 1). In the spatial (dwhichT) domain of response selection, Experiment 2 yielded redundancy data comparable with our previous findings (Hunter et al., 2003); subjects utilized buttons A and B to generate sequences of high randomness content (Baddeley, 1966). In the temporal (dwhenT) domain, the mean interresponse interval and its variance were similar in Experiments 2 and 3 (both temporally dunconstrainedT). Modeling ddeactivationT as prior dactivationT In Experiment 1, we were able to further deconstruct the time course of left DLPFC because the mean interresponse interval approached 20 s. Fig. 2 shows that the mean hemodynamic response in left DLPFC peaked at 4 s before action execution. The canonical HRF used in SPM99 models peak hemodynamic response at 5 s following the deventT (with a degree of variation conferred by the first temporal derivative). Hence, we reasoned that if deactivation of left DLPFC was an expression of its prior activation (rather than a negative deviation from a flat dbaselineT), then the same area that was deactivated in our model (time-locked to the moment of action execution) should be activated in an alternative model: time-locked to an event defined as occurring 9 s before action execution.
½ad ð1 cc Þ þ bd cc n Where a is the specificity (i.e., [0.001]1/6 in the current example, as the conjoint P threshold in SPM99 is 0.001 and there are six subjects) and b is the sensitivity (paradoxically, the most conservative approach is to assume high sensitivity, because it is very unlikely that an insensitive test could detect an effect in n subjects). Hence, conjoint P = 0.001 in a group of six corresponds to P = 0.05 in the population for the effect being typical, that is, occurring more often than not, if sensitivity = 90%. For the purposes of reporting and neuroanatomical labeling, the coordinates of significant areas of activation were transformed from MNI space into the stereotactic space of Talairach and Tournoux (1988). Results In Experiment 1, we observed deactivation of left DLPFC (BA 46), at the moment of spontaneous motor-action execution.
Fig. 1. Functional dissociation at the moment of spontaneous motor-action execution. Group average data from Experiment 1 are projected on the SPM99 dsmooth brainT (www.fil.ion.ucl.ac.uk), viewed from the front and left. Significant clusters ( P b 0.05, corrected) are color-coded. Areas that demonstrated a significant positive correlation with the hemodynamic model of neural activation, at the moment of motor-action execution, are shown in red (activation); negatively correlated areas, the most extensive of which was left DLPFC, are shown in blue (deactivation).
M.D. Hunter et al. / NeuroImage 23 (2004) 1186–1191
1189
Fig. 2. Temporal response dynamics in the periaction period. Averaged hemodynamic time series from selected regions of interest in Experiment 1: right frontal pole (34, 55, 17), left frontal pole (38, 55, 12), and left dorsolateral prefrontal cortex (53, 28, 13). These filtered data represent the first eigenvariate from spheres of radius 5 mm centered upon the activation maxima.
Group average analysis revealed that this was the case (BA 46; activation coordinates: 51, 30, 14 [c.f., deactivation: 53, 28, 13]; peak t statistic = 5.98; 66 voxels exceeded threshold P b 0.05, corrected for multiple comparisons in the whole brain). We also identified those areas that were ddeactivatedT at 9 s before action execution. The largest involved primary motor cortex (BA 4; 32, 14, 63; t = 7.97; 117 voxels exceeded P b 0.05, corrected). This was the same region that activated at the point of action execution (Table 1). Thus, the left DLPFC and motor cortex were effectively dmirror imagesT of each other. At the moment of action execution, motor cortex activated (while DLPFC deactivated), and at 9 s before execution (the approximate midpoint of the mean interresponse interval), DLPFC activated (while motor cortex deactivated). Discussion We have used event-related fMRI to demonstrate functional dissociation between distinct prefrontal regions at the moment of spontaneous motor-action execution. Specifically, we have shown that, while frontal pole regions are activated at the point of action execution, left DLPFC deactivates, a finding crucially different from those of previous studies. The area of left DLPFC deactivation that we observed, in Experiment 1, was maximal at a relatively inferior location, close to premotor cortex (BA 45/46). In Experiment 2, the deactivation maximum was located at almost exactly the same coordinates (Table 1), perhaps reflecting the methodological similarity between Experiments 1 and 2 (both of which permitted dwhichT and dwhenT response selection). In Experiment 3, the deactivation maximum was more anteromedial, close to some of those coordinates for left DLPFC activation maxima which have been reported in the previous literature (Frith et al., 1991; Hyder et al., 1997; Spence et al., 1998). Nevertheless, we are cautious in comparing such coordinates in view of the many potential sources of variation between respective studies: individual subject anatomy, resolution
of imaging technique, and data preprocessing, not least image smoothing. However, the qualitative points remain: first, that previous work described DLPFC activation, whereas our eventrelated experiments demonstrate deactivation at the point of action, and second, that the provision of a dwhenT choice in the psychological paradigm was sufficient to elicit this effect. The existence of a dwhichT response selection parameter may modulate the deactivation process but is not a prerequisite. In this paper, the term ddeactivationT means reduced fMRI (BOLD) signal. For several reasons, the term ddeactivationT has been controversial in fMRI research. First, the BOLD signal is always a relative measure (Arthurs and Boniface, 2002). In the current study, we found left DLPFC to be deactivated at the moment of motor-action execution relative to activated areas at the frontal poles and relative to its own, prior activation. Secondly, fMRI utilizes an indirect (vascular) marker of neural activity. We must, therefore, have some understanding of the neurovascular coupling between neuron and BOLD signal to interpret both increases and decreases of the latter. Simultaneously acquired fMRI and in situ electrophysiological data have helped to elucidate the nature of this relationship (Logothetis et al., 2001). On the basis of such work, the BOLD signal is currently thought to represent a positive correlate of the neuronal local field potential (a measure of synaptic activity). In straightforward terms, increased BOLD signal corresponds to increased synaptic activity (which, itself, may correspond to increased neuronal firing). However, neural dactivityT at this level may represent excitation or inhibition (at the neurotransmitter level). It has been suggested that inhibition is associated with lower metabolic demand than excitation, because there are fewer cortical inhibitory neurons, which are, additionally, more efficient than their excitatory equivalents (Arthurs and Boniface, 2002). Supporting evidence for this hypothesis comes from a combined fMRI/ transcranial magnetic stimulation paradigm, which demonstrated that, unlike excitation, inhibition of the motor cortex did not evoke measurable change in the BOLD signal (Waldvogel et al., 2000). Although it has also been argued (Tagamets and Horwitz, 2001) that
1190
M.D. Hunter et al. / NeuroImage 23 (2004) 1186–1191
inhibition might raise the BOLD signal, in the context of minimal local excitation, one interpretation of our results could be that relative deactivation of left DLPFC, at the moment of action execution (relative to a previously activated state), reflects reduced excitation in that region at that moment. The bilateral activation at the frontal poles that we observed was almost identical to the result of a PET study examining prospective memory (Burgess et al., 2001). In that study, subjects were expecting to perform a cued motor task at infrequent intervals. Critically, by comparison with a baseline condition, it was the expectation of action execution (the online maintenance of dwhat to do,T whether or not execution followed) that was associated with such activation. Hence, it is possible that, at the moment of action execution, anterior prefrontal activation is related to the monitoring of motor intention (dpress a button at willT) and might serve to drefreshT prospective memory of the required motor behavior. In psychological terms, such activity might therefore correspond to an dimplementation intentionT (Sheeran, 2002). However, and more importantly, we found evidence for deactivation of left DLPFC at the moment of action execution, relative to its prior activated state. This suggests early involvement of DLPFC during the temporal evolution of action initiation and execution, an obligate effect congruent with the region’s functional anatomy: its role in the free selection of response behaviors (Petrides and Milner, 1982) as a component of the supervisory attentional system (Shallice and Burgess, 1996). Although DLPFC also supports working memory (Postle et al., 1999), its involvement during action selection is distinct and dissociable (Rowe et al., 2000). Unlike Rowe et al. (2000), we did not find evidence of DLPFC activation at the moment of action execution. However, that group’s paradigm involved cued dselection from memoryT (c.f., spontaneous motor-action execution), and their DLPFC activation foci were in the right hemisphere (c.f., our deactivation foci in the left, at a relatively inferior location, maximal adjacent to premotor cortex). These differences are likely to reflect functional specificities within regions of the prefrontal cortex, differentially probed by distinct neuroimaging paradigms. Hence, we have a hypothesis regarding the dcausalT direction of willed action, namely, that the supraordinate precedes the subordinate (in this case, that left DLPFC activates before premotor and motor cortices). This hypothesis is supported by electrophysiological and magnetoencephalographic data that indicate prefrontal and premotor activation early in the evolution of motor behavior (Goldman-Rakic et al., 1992; Nishitani and Hari, 2000). Yet, we are also demonstrating temporal dissociation within higher, executive regions, since left DLPFC is activating earlier than the frontal poles and supplementary motor area. Also, we have previously reported that, among those executive regions activated proximal to the moment of action execution, the left frontal pole (BA 10) and rostral supplementary motor area (BA 6) have significantly shorter BOLD response latencies than primary motor cortex (Hunter et al., 2003), a difference that could correspond to the fMRI equivalent of the electrophysiological readiness potential but which does not amount to the profound difference between deactivated DLPFC and other activated brain regions that we report here. Hence, using eventrelated fMRI, we have discerned a sequence from the activation and subsequent deactivation of left DLPFC through the activations of the frontal poles and supplementary motor area, to the activation of motor cortex—an evolution of volition: voluntary behavior. Regarding the biological significance of these changes in signal, Ball et al. (1999) have used high-resolution EEG, combined
with fMRI, to show that the expected readiness potential originating in the supplementary motor area during the emergence of voluntary action actually undergoes a sharp reduction immediately before its execution. Those authors argued that such a reduction in signal may trigger the motor act by releasing the primary motor cortex from putative inhibition. It is conceivable (though conjectural) that our observed reduction in DLPFC activation in the moments before action could similarly act as a release signal for premotor cortex. Regarding subjective phenomenology, we acknowledge that we could not know what our subjects were dthinkingT while they performed our experiments, nor do we know the latency between their decisions to act and the actions themselves. However, we need not invoke a direct correspondence between the physiology of motor execution and the phenomenology of intention (as our experiments are designed to address the biology of action). Nevertheless, it has been suggested that there may be a robust coupling of the experience (phenomenology) of will and specific, identified behaviors on the part of the agent even when they are temporally dissociated by as much as 5 s (Wegner, 2002). Also, crucially, conscious intention has been shown to follow electrophysiological evidence of action initiation (Libet et al., 1983). Hence, our data inform the physiology of action, although not necessarily its subjective phenomenology. We should not equate left DLPFC activation with the moment of perceived subjective choice (even if it is the moment of neural choice). While our results are entirely compatible with earlier PET and blocked fMRI studies that identified the involvement of DLPFC in response selection (Frith et al., 1991; Hyder et al., 1997; Jenkins et al., 2000; Spence et al., 1998), they significantly refine their interpretation. In those studies, DLPFC activation was revealed because of its temporal integration over the scanning window (i.e., it occurred at some point during the action tasks involved). Hence, a parsimonious (but incorrect) interpretation might have been that DLPFC was activated at the moment of action execution. However, crucially, our studies demonstrate that this is not the case and that, at the moment of spontaneous motor-action execution (relative to its prior state), DLPFC deactivates. Acknowledgments M.D.H. is supported by the Wellcome Trust. An investigatorled award to S.A.S. from Cephalon (UK) supported R.D.J.G. We thank our radiographer colleagues at the Academic Unit of Radiology, University of Sheffield, for assistance during image acquisition. References Arthurs, O.J., Boniface, S., 2002. How well do we understand the neural origins of the fMRI BOLD signal? Trends Neurosci. 25, 27 – 31. Baddeley, A.D., 1966. The capacity for generating information by randomisation. Q. J. Exp. Psychol. 18, 119 – 129. Ball, T., Schreiber, A., Feige, B., Wagner, M., Lucking, C.H., KristevaFeige, R., 1999. The role of higher-order motor areas in voluntary movement as revealed by high-resolution EEG and fMRI. NeuroImage 10, 682 – 694. Burgess, P.W., Quayle, A., Frith, C.D., 2001. Brain regions involved in prospective memory as determined by positron emission tomography. Neuropsychologia 39, 545 – 555.
M.D. Hunter et al. / NeuroImage 23 (2004) 1186–1191 de Zubicaray, G.I., Williams, S.C., Wilson, S.J., Rose, S.E., Brammer, M.J., Bullmore, E.T., et al., 1998. Prefrontal cortex involvement in selective letter generation: a functional magnetic resonance imaging study. Cortex 34, 389 – 401. Desmond, J.E., Gabrieli, J.D.E., Glover, G.H., 1998. Dissociation of frontal and cerebellar activity in a cognitive task: evidence for a distinction between selection and search. NeuroImage 7, 368 – 376. Dilman, I., 1999. Free Will: A Historical and Philosophical Introduction. Routledge, London. Evans, A.C., Collins, D.L., Mills, S.R., Brown, R.D., Kelly, R.L., Peters, T.M., 1993. 3D statistical neuroanatomical models from 305 MRI volumes. Proc.-IEEE Nucl. Sci. Symp. Med. Imaging Conf. 108, 1877 – 1878. Friston, K.J., Ashburner, J., Frith, C.D., Poline, J.B., Heather, J.D., Frackowiak, R.S.J., 1995. Spatial registration and normalization of images. Hum. Brain Mapp. 2, 165 – 169. Friston, K.J., Holmes, A.P., Worsley, K.J., 1999. How many subjects constitute a study? NeuroImage 10, 1 – 5. Frith, C.D., Friston, K., Liddle, P.F., Frackowiak, R.S.J., 1991. Willed action and the prefrontal cortex in man: a study with PET. Proc. R. Soc. London, B. 244, 241 – 246. Goldman-Rakic, P.S., Bates, J.F., Chafee, M.V., 1992. The prefrontal cortex and internally generated motor acts. Curr. Opin. Neurobiol. 2, 830 – 835. Hunter, M.D., Farrow, T.F.D., Papadakis, N.G., Wilkinson, I.D., Woodruff, P.W.R., Spence, S.A., 2003. Approaching an ecologically valid functional anatomy of spontaneous bwilledQ action. NeuroImage 20, 1264 – 1269. Hyder, F., Phelps, E.A., Wiggins, C.J., Labar, K.S., Blamire, A.M., Shulman, R.G., 1997. bWilled action:Q a functional MRI study of the human prefrontal cortex during a sensorimotor task. Proc. Natl. Acad. Sci. U. S. A. 94, 6989 – 6994. Jahanshahi, M., Dirnberger, G., Fuller, R., Frith, C.D., 2000. The role of the dorsolateral prefrontal cortex in random number generation: a study with positron emission tomography. NeuroImage 12, 713 – 725. Jenkins, I.H., Jahanshahi, M., Jueptner, M., Passingham, R.E., Brooks, D.J., 2000. Self-initiated versus externally triggered movements. II. The effect of movement predictability on regional cerebral blood flow. Brain 123, 1216 – 1228. Libet, B., Gleason, C.A., Wright, E.W., Pearl, D.K., 1983. Time of conscious intention to act in relation to cerebral activity (readinesspotential). The unconscious initiation of a freely voluntary act. Brain 106, 623 – 642. Logothetis, N.K., Pauls, J., Augath, M., Trinath, T., Oeltermann, A., 2001. Neurophysiological investigation of the basis of the fMRI signal. Nature 412, 150 – 157.
1191
Nishitani, N., Hari, R., 2000. Temporal dynamics of cortical representation for action. Proc. Natl. Acad. Sci. 97, 913 – 918. Oldfield, R.C., 1971. The assessment and analysis of handedness: the Edinburgh inventory. Neuropsychologia 9, 97 – 113. Petrides, M., Milner, B., 1982. Deficits on subject-ordered tasks after frontal- and temporal-lobe lesions in man. Neuropsychologia 20, 249 – 262. Phelps, E.A., Hyder, F., Blamire, A.M., Shulman, R.G., 1997. FMRI of the prefrontal cortex during overt verbal fluency. NeuroReport 8, 561 – 565. Postle, B.R., Berger, J.S., D’Esposito, M., 1999. Functional neuroanatomical double dissociation of mnemonic and executive control processes contributing to working memory performance. Proc. Natl. Acad. Sci. U. S. A. 96, 12959 – 12964. Rowe, J.B., Toni, I., Josephs, O., Frackowiak, R.S.J., Passingham, R.E., 2000. The prefrontal cortex: response selection or maintenance within working memory? Science 288, 1656 – 1660. Shallice, T., Burgess, P., 1996. The domain of supervisory processes and temporal organization of behaviour. Philos. Trans. R. Soc. London. Ser. B. 351, 1405 – 1412. Sheeran, P., 2002. Intention–behaviour relations: a conceptual and empirical review. In: Stroebe, W., Hewstone, M. (Eds.). Eur. Rev. Soc. Psychol. vol. 12. John Wiley and Sons Ltd, Chichester, UK, pp. 1 – 36. Spence, S.A., Hirsch, S.R., Brooks, D.J., Grasby, P.M., 1998. Prefrontal cortex activity in people with schizophrenia and control subjects. Evidence from positron emission tomography for remission of bhypofrontalityQ with recovery from acute schizophrenia. Br. J. Psychiatry 172, 316 – 323. Spence, S.A., Liddle, P.F., Stefan, M.D., Hellewell, J.S.E., Sharma, T., Friston, K.J., Hirsch, S.R., Frith, C.D., Murray, R.M., Deakin, J.F.W., Grasby, P.M., 2000. Functional anatomy of verbal fluency in people with schizophrenia and those at genetic risk. Focal dysfunction and distributed disconnectivity reappraised. Br. J. Psychiatry 176, 52 – 60. Spence, S.A., Hunter, M.D., Harpin, G., 2002. Neuroscience and the will. Curr. Opin. Psychiatry 15, 519 – 526. Tagamets, M.-A., Horwitz, B., 2001. Interpreting PET and fMRI measures of functional neural activity: the effects of synaptic inhibition on cortical activation in human imaging studies. Brain Res. Bull. 54, 267 – 273. Talairach, P., Tournoux, J.A., 1988. A Stereotactic Co-planar Atlas of the Human Brain. Thieme, Stuttgart. Waldvogel, D., van Gelderen, P., Muellbacher, W., Ziemann, U., Immisch, I., Hallett, M., 2000. The relative metabolic demand of inhibition and excitation. Nature 406, 995 – 998. Wegner, D.M., 2002. The Illusion of Conscious Will. MIT Press, Cambridge, MA.