Beat induction refers to the phenomenon whereby a metrical rhythm gives rise to a sense of isochronous pulse or beat. ERP experiment (64-channel EEG) investigated brain processes underlying beat induction. Results show that the evoked N2 is enhanced in regular as compared with random stimuli. Random omission of the third click on 50% of trials suppresses the N2 in the third interval but induces a later N3 when the 3 rd click is present
Beat induction refers to the phenomenon whereby a metrical rhythm gives rise to a sense of isochronous pulse or beat. ERP experiment (64-channel EEG) investigated brain processes underlying beat induction. Results show that the evoked N2 is enhanced in regular as compared with random stimuli. Random omission of the third click on 50% of trials suppresses the N2 in the third interval but induces a later N3 when the 3 rd click is present
Beat induction refers to the phenomenon whereby a metrical rhythm gives rise to a sense of isochronous pulse or beat. ERP experiment (64-channel EEG) investigated brain processes underlying beat induction. Results show that the evoked N2 is enhanced in regular as compared with random stimuli. Random omission of the third click on 50% of trials suppresses the N2 in the third interval but induces a later N3 when the 3 rd click is present
Beat induction refers to the phenomenon whereby a metrical rhythm gives rise to a sense of isochronous pulse or beat. ERP experiment (64-channel EEG) investigated brain processes underlying beat induction. Results show that the evoked N2 is enhanced in regular as compared with random stimuli. Random omission of the third click on 50% of trials suppresses the N2 in the third interval but induces a later N3 when the 3 rd click is present
ELECTROPHYSIOLOGICAL CORRELATES OF BEAT INDUCTION AS
INTERNALLY- AND EXTERNALLY-GUIDED ACTION
1 Neil P. Todd and 2 Ellen Seiss 1 Department of Psychology, University of Manchester, M13 9PL, UK. 2 Behavioural and Brain Sciences Centre, University of Birmingham, B15 2TT, UK. ABSTRACT In this paper we report the results of an ERP experiment (64- channel EEG) to investigate brain processes underlying beat induction. Musically trained subjects were required to listen to an anapest rhythm, consisting of three clicks, with inter-onset intervals of 500 ms, and a gap of 1000 ms, under active and passive conditions, preceded by a condition in which the stimulus was unpredictable. A further condition was added which introduced uncertainty into the presence or absence of the third click. The results show that the evoked N2 is enhanced in regular as compared with random stimuli. Random omission of the third click on 50% of trials suppresses the N2 in the third interval but induces a later N3 when the 3 rd click is present. Comparing active and passive conditions indicates that pre-movement negativities (PMN) overlap spatiotemporally with the regular N2 consistent with the N2 including a component which may be a correlate of movement preparation. We interpret the potentials as being consistent with the operation of two distinct sensorimotor circuits: (1) an automatic, internally driven circuit involving SMA for which the N2 may be a manifestation, and (2) an attention dependent, externally driven circuit involving PPC for which the N3 may be a manifestation. Likely both circuits are co-activated during the presentation of a rhythmic stimulus but the relative role of the two will depend on the predictability of the rhythm. 1. INTRODUCTION Beat induction or pulse sensation refers to the phenomenon whereby a metrical rhythm gives rise to a sense of isochronous pulse or beat (also known as the tactus). Research on beat induction overlaps with the field of sensorimotor synchronisation in which the principal experimental paradigm is synchronisation of simple repetitive movements with metronomic stimuli. Experimental studies of beat induction thus have similar methodologies, e.g. tapping, except that they involve non- isochronic, metrical rhythms as stimuli with a particular focus in the induction of a regular down beat, including in intervals for which there is no explicit stimulus present. Thus beat induction may be thought of as a kind of "mental continuation". Of particular relevance to this study is the established fact that beat induction has an existence region, i.e. the sense of pulse only occurs for metrical rhythms with a range of beat interval between about 300 and 900 ms with median about 450 ms (Moelents and van Noorden, 1999) which las led some authors to draw a parallel with natural rates of locomotion. The sense of pulse is also enhanced by syncopation, i.e. when rhythmic events do not fall on a predicted down beat. Whilst there are now numerous computational models and algorithms which attempt to simulate beat finding (for review see Todd , OBoyle and Lee, 2002) there are essentially only two theories which give an account of beat induction which have a bearing on possible underlying brain mechanisms in humans: (1) a beat is an attentional oscillation (AO) (Large and Jones, 1999) and (2) a beat is a movement intention (MI) (Todd et al., 2002). The AO theory supposes that there exists in the brain a network of coupled oscillators which synchronise with a periodic stimulus. The MI theory in contrast supposes that beat induction is a form of sensoriguided action in which the well-established motion planning network, involving pre-motor cortex (PMC), posterior parietal cortex (PPC) and cerebellum, is activated in conjunction with temporal coding of stimulus properties in auditory cortex by means of the modulation power spectrum. In this study we seek to investigate brain processes underlying beat induction by means of high resolution (64-channel) EEG. In particular we seek to find neurophysiological correlates of motion planning in a passive listening task in order to provide evidence for the MI theory. ERPs are an excellent tool for investigating sensorimotor synchronisation, given the fine temporal resolution, and produce well-defined stimulus and movement related potentials. Whilst MEG has advantages in some circumstances and has been used to identify stimulus and movement related brain events (Muller et al., 2000) for sensorimotor synchronisation MEG is less suitable for longer latency N2/P3 waves which are of interest in cognitive planning. A recent study of error correction in sensorimotor synchronisation by Praamstra et al. (2003) using EEG gives examples of stimulus and movement related potentials which are of relevance here and which we use as a basis for the identification of the various components in our study. Essentially there are three morphologies: (a) stimulus related potentials (SRPs), consisting of a P1, N1, P2 and N2, (b) movement related potentials (MRPs), consisting of a PMN, a re-afferance negativity (RAN) and a post-motion positivity (PMP) and (c) a hybrid SRP/MRPmorphology, obtained by locking the averaging to the stimulus during active synchronisation. The aim of the study reported here was to use a modification of the standard procedure for sensorimotor synchronization by stimulating with an anapest rhythm, rather than an isochronous sequence, consisting of three clicks and a gap in which a target beat is induced. The idea being that the gap would allow investigation of brain activity, either passive or active, induced before, at and after the target without being immediately masked by the standard SRPs. We hypothesized that potentials in the gap should show evidence of movement relatedness even in a passive condition. ISBN 1-876346-50-7 2004 ICMPC 212
2. METHOD Subjects 10 musically trained subjects were selected from staff and students at the University of Birmingham. Subjects were all right handed and had no hearing or cognitive impairments. Stimuli Stimuli were 1 ms, compressive square pulses, producing a robust click, presented at a loud but not uncomfortable intensity. Apparatus EEG, EMG and EOG were recorded continuously with Ag/AgCl electrodes located in scalp, muscle and periocular positions. Finger movements were recorded using a force-plate transducer. EEG locations were given by the 10-5 extension of the 10-20 system. EMG was recorded differentially in positions above the flexor and extensor indicies muscles. All signals were amplified by BioSemi Active-One Amplifiers and sampled at 512 Hz. Stimuli generated using Eprime software and delivered by detachable headphones placed under the cap. Procedure After preparation subjects were seated in a comfortable chair and were given instructions. They were told that they would be presented with a number of rhythms in sessions lasting about 4 mins. At the start of each session they would be given a warning tone. During the sessions they were to be as relaxed as possible and avoid unnecessary movements with gaze fixated but to maintain attention on the rhythm. Recordings were made under four conditions consisting each of two blocs of about 4 mins. The first three conditions were passive. The first was a random condition (300 repetitions), in which the inter-onset interval of the clicks varied randomly between 1000 and 2000 ms with a rectangular probability distribution. The second was a regular condition (200 repetitions), in which an anapest rhythm was presented consisting of three clicks with inter-onset interval of 500 ms (close to the optimum interval) followed by a gap of 1000 ms. The third was a regular-uncertain condition (400 repetitions), in which the anapest rhythm was randomly alternated with a rhythm consisting of two clicks with inter-onset interval of 500 ms followed by a gap of 1500 ms. The fourth condition was active (200 repetitions) in which the subjects were to synchronise extension of their right index finger to the beat of the regular anapest rhythm. They were given explicit instructions on what was meant by the beat, including the missing beat in the gap. They were also instructed that the required movement was a flick without any using any force on the return and without movement of the other fingers. This was to avoid significant activity in the flexor muscle. The movement target was a tap on the force-plate which was adjusted to be about 2 cm from the index finger which at rest hung loosely over the end of the arm rest. EEG/EMG analysis EEG was segmented into epochs off-line according to triggers which were time-locked to the stimuli. Epochs containing significant artifacts were rejected from the analysis. Averages were obtained for each of the four conditions for each subject and grand means obtained for each condition. For the regular uncertain condition the averaging was carried out separately for those trials with the 3rd beat present and those with the 3rd beat missing. EMG was rectified and low-pass filtered and averaged in the same way. Due to excessive artifact, data from two of the subjects were not included in the final analysis. 3. RESULTS
Figure 1: Grand mean from 8 subjects for the random condition. Potentials are centred at FCz with potentials measured at surrounding electrodes also shown. Considering first the random condition (Figure 1) N1, P2 and N2 potentials are identifiable with latencies of about 95, 190 and 280 ms and with amplitudes of about -5.0, +6.5 and 1.5 V respectively at FCz. Source analyses of the N1 supported a bilateral tangential source from within the sylvian fissure (Talairach-Tournoux coordinates, X=-48.7, 43.6, Y=-30.9, -30.4, Z=19.2, 23.1, RV = 2.8%). The N2, which was highly focussed over FCz, could be explained by a single source in medial frontal gyrus (BA 6) corresponding to the supplementary motor area (SMA) (Talairach-Tournoux coordinates, X=+5, Y=-21, Z=62, RV = 4.8%). A bilateral source model placed the generators near to the central sulcus and was therefore considered to be implausible.
Figure 2: Grand mean (centred at FCz) from 8 subjects for the regular anapest condition. In the regular anapest condition N1, P2 and N2 potentials can be also identified for each of the three clicks within the anapest. For the third interval prior to the gap these have peak latencies of about 90, 150 and 270 ms with amplitudes of about -0.8, +1.8 and -2.0 V at FCz. Thus all three waves peak earlier than the random condition. For amplitude the N1/P2 is significantly reduced in ICMPC8, Evanston, IL, USA August 3-7, 2004 213
amplitude compared with the random condition and appears to show adaptation over course of the anapest rhythm so that they are largest following the gap. The N2 in contrast appears to show integration over the anapest rhythm so that it is largest prior to the gap. Inspection of scalp distribution showed that the later aspect of the N1, up to about 120 ms, prior to the gap developed a more lateral distribution. Consistent with previous analyses (Scherg et al. 1989) the distribution could be explained by an additional bilateral radial source, the N150, superposed with the bilateral tangential source. The position of the N150 was located in BA 42 near the lateral surface of the superior temporal gyrus (Talairach-Tournoux coordinates, X=-65.9,+65.9, Y=-30.2, Z=8.4, RV = 5.8%). The tangential source was fixed at the position and orientation of the N1 from the random condition.
Figure 3: Scalp distribution of the N2 for the random case (left) compared with the N2 prior to the gap in regular anapest (right). The N2 prior to the gap in contrast to the random N2 was more widespread (Figure 3), as well as being larger in magnitude, and hence a bilateral source model seemed more appropriate. However, this gave coordinates, X=-20,+20, Y=-14.3, Z=43.8, (RV = 5.3%), which would place the sources at the juncture between anterior cingulate cortex (BA 24) and medial frontal gyrus (BA 6), and was therefore ambiguous. In addition to the potentials which could be compared with the random condition there was a P3 which followed the N2 prior to the gap and a series of emitted waves in the gap. The P3 had with a latency of about 370 ms and could be explained by a bilateral source within the sylvian fissure in BA 41 but more caudal than the N1 generators (Talairach-Tournoux coordinates, X=- 39.5,+39.5, Y=-41, Z=10.7, RV = 4.5%). The series of emitted waves in the gap appeared to be an echo of the evoked potentials in the 3 rd interval and hence were labelled eN1, eP2, eN2 and eP3. The eN1/P2 obtained just after the target beat had latencies of about 90 and 140 ms with small amplitudes. The eN2/P3 obtained from the average before the 1st click had latencies of about 300 and 350 ms with amplitudes of about -1.0 and +0.3 V. Reliable source analyses were not obtainable for the emitted potentials nor were they readily identifiable in the individual averages. Considering now the regular uncertain condition (Figure 4) in the first two intervals prior to the 3rd beat, the morphology of the three stimuli are very similar but with two subtle differences. In both variants of the uncertain condition the size of the P2s were larger, but the N1s not so, and there was a general enhanced negativity prior to 3rd beat.
Figure 4: Grand mean from 8 subjects for the regular anapest condition with grand means for the two variants of the uncertain condition superimposed. Case with 3rd beat present in red and case with 3rd beat missing in blue. In the third interval, however, there was a radical change in the morphology of the ERP. In the 50% uncertain case with 3rd beat present the regular N2 was suppressed and an early P3 appeared (labelled P3a) at about 300 ms with amplitude about 1 V followed by a later negativity (labeled N3) at about 400 ms with an amplitude of about 2.5 V. For the P3a a three dipole solution was attempted with a bilateral posterior source and a single anterior source (Talairach-Tournoux coordinates, X=-38.6,+38.6, Y=-57.9, Z=3.8, and X=-9.4, Y=36.3, Z= 17.7, RV = 3.8%). However, this was not considered satisfactory as the location of the bilateral sources in medial temporal lobe were not within 11mm of gray matter. The single anterior source located at the juncture between medial frontal (BA9) and anterior cingulate cortex (BA32) could alone account for the P3a with RV = 13.8%. The N3 could be accounted for by a bilateral source in temporoparietal cortex (Talairach-Tournoux coordinates, X=- 39.5,+48.3, Y=-37,-30.2, Z=31.8,23.5, RV = 5.5%). In this case the symmetry condition was relaxed and the right source was more lateral, anterior and deeper than the left. However both sources were located in BA40 within the inferior parietal lobule. For the 50% uncertain condition with 3 rd beat missing there was an eN1 at about 170 ms which had a more lateralised distribution similar to that of the N150 but a reliable source analysis was not obtainable. The P2/N2 was also supressed in this condition and was followed by a later P3 (labelled P3b) which has a similar latency as the P3 in the regular anapest, but with enhanced amplitude. For the fourth interval the morphology of all three conditions is again similar except that the amplitude of the eN1 was enhanced for the 50% uncertain 3rd beat present case. The peak latency of this eN1 was about 150 ms. For the active condition only four of the eight subjects were able to carry out the task to a sufficient degree of skill, i.e. were able to maintain good synchrony without significant flexor activity, thus only these subjects were included in the analysis of the active ICMPC8, Evanston, IL, USA August 3-7, 2004 214
condition. However, comparison of the grand mean of these four with the overall grand mean shows them to be representative.
Figure 5: Grand mean (centred at FCz) for the active condition for 4 skilled subjects compared with averaged EMG and force responses. The morphology of the first two intervals (Figure 5) are comparable to that of the hybrid SRP/MRPs from Praamstra et al (2003) as would be expected. For the third and fourth interval however the morphology of the potentials more closely resembles the pure MRP such that a PMN, RAN and PMP can be identified, even though the averages were stimulus locked. This result is consistent with the good performance of the four subjects in maintaining accurate synchrony as can be seen in the well- localised EMG and force responses. In the fourth interval the PMN showed a double peak indicative of a more complex internal structure. This may be related to the fact that the 4th interval is a pure cycle of continuation, and hence they were labeled cPMNa,b (respectively with latencies of about 270 ms and 310 ms) to distinguish them from the synchronisation PMN. Similarly the RAN in the gap just prior to the 1st click, which was more sharply focussed in time, is likely more of a pure MRP hence it was labeled cRAN.
Figure 6: Grand mean for the active condition for 4 skilled subjects compared with their corresponding mean for the passive condition. Examination of the scalp potential distribution showed that the latter part of the premovement negativity was focussed over left central electrodes (approximately C1) consistent with it being a correlate of activity in primary motor cortex. EMG started during the PMN about 200 ms before target time, followed by the force onset at about 150 ms before target and onset of RAN between about 80 to 60 ms before target. EMG reached a peak at about 90 ms before target followed by the peak force almost exactly on target. In the 4th beat the RAN peaked about 10 ms after the target. The cRAN appeared to peak just before the target by about 25 ms but it may be the case that its continuation past the target was masked by the early auditory waves. In order to directly compare active and passive conditions waveforms are superimposed in Figure 6. In the 3 rd interval before the missing beat there was a considerable temporal and spatial overlap between the N2 and PMN. In the 4th interval the eN2 was temporally aligned with the cPMNb, but the eN2 of Figure 2 (8 subjects) also shows a double peak. 4. DISCUSSION That the N1 was reduced in magnitude from the random to the regular condition and shows adaptation within the anapest is consistent with previously established literature indicating that the amplitude is inversely related to stimulus repetition rate, most likely due to the refractoriness of the N1. For the random condition the mean rate was between 1 and 0.5 Hz whereas for the anapest it was between 2 and 1 Hz. Of more interest was the observation that of the tangential and radial components of the N1, the radial component (N150) was more prominent in the anapest condition. This could be simply interpreted as the N150 being less susceptible to adaptation or it could be interpreted as it representing a more active process associated with regularity. The fact that the emitted N1s peaked in the uncertain conditions at about 150 ms and that they had a more lateral distribution similar to the N150, supports an interpretation whereby the generator of the N150 is active and that it is the primary source for the eN1. However a reliable source analysis was not obtainable from the emitted potentials and so a possible association must remain purely conjectural at this stage. The spatio-temporal overlap of the N2 with the PMN when comparing active and passive conditions is suggestive that the auditory N2 in the conditions presented here may be movement related, or at least include a component which is movement related. Such an interpretation of the N2 being movement related would be consistent with a location of the N2 generator(s) in the frontal lobe. It would also be controversial though as previous studies have located generators of the various negativities following the N1/P2 complex, including the mis-match negativity
(MMN) and slow wave (SW), in the temporal lobe (Sherg et al. 1989). However, frontal lobe generators have also been postulated for the late auditory potentials (Naatanen and Picton, 1987) and so a frontal source for the N2 in this case is not implausible. However, the source analysis for the regular N2s was ambiguous between SMA and ACC. In retrospect the simple bilateral model may have been inappropriate and a more complex model including both bilateral temporal sources and a source near the midline in SMA (as for the random N2) would give a more plausible account of the N2. This would be consistent with imaging studies which show activation of SMA in continuation, and the anapest rhythm does involve one cycle of continuation. Conversely an ACC source would seem less plausible as the task did not involved error detection or monitoring (Praamstra et al., 2003). ICMPC8, Evanston, IL, USA August 3-7, 2004 215
If the above interpretation is correct, that activation of SMA is centrally involved in beat induction, then it needs to be explained what is different with the random case which also indicated SMA activation. One possibility could be that both random and regular rhythms evoke the same basic auditory-motor mechanism, possibly related to a primitive orienting response (Naatanen and Gaillard, 1983). The basic response however could be suplemented with a rapid learning mechanism which forms a representation of the stimulus interval. Thus the apparent integration in the magnitude of the N2 over the anapest may be the result of long term potentiation (LTP) in the learning mechanism, possibly involving the hippocampus and basal ganglia. Such a learning mechanism would be capable of generating an expectation or intention on the basis of the interval. An active component to the regular evoked N2 thus provides an account of the emitted N2 in the gap. If the mechanism is generating a "pulse" of activity then it should be visible even without an explicit stimulus. The wave which we have identified as eN2 has a latency of 300 ms which is within 20 or 30 ms of the evoked N2. The amplitude of the eN2 is about 1.0 V which added to the random N2 gives a magnitude comparable to the regular N2. Thus the eN2 could plausibly represent a continuation of the expectation. This view is strengthened by the temporal alignment of the eN2 with cPMNb. We have already discussed the possible association of the eN1 with the N150 whereby the N150 could be a correlate of an active process in auditory cortex. Irrespective of whether this association holds up or not there are two possible mechanisms which could underly an active N1 component. The first is that the active component is self-generated in auditory cortex. As is reviewed elsewhere (Todd et al 2002) there is now a considerable body of evidence that many neurones in auditory cortex have receptive fields (RFs) which are tuned for modulation frequency between about 0.5 and 20 Hz and that collectively they compute a modulation power spectrum. The RFs come is a variety of species with both low-pass and band-pass types being present. Of particular relevance to our account is the fact that band-pass RFs are associated with "ringing" or damped oscillations. Thus the eN1 may be a correlate of ringing of band-pass RFs in auditory cortex. Given the fairly broad tuning (or heavy damping) of the RFs one would expect a spread of activation in the population and hence a dispersion in the temporal response, which is what is observed in the eN1. The location of the eN1/N150 generators in the lateral belt area of auditory cortex would also be consistent with the documented modulation sensitivity of this region (Rauscheker and Tian, 2000). This account is not precluded though by the second possible mechanism, which is that the active component is a response to the pulse generator via reciprocal connections between SMA and auditory cortex. Indeed it is likely that the two would work together. The modulation power spectrum in the lateral belt area provides candidate beat intervals, which for most metrical rhythms there are a number possible (unlike simple metronome tracking), and the pulse generator, informed by the natural frequencies of the body, reinforces one beat interval in particular. Turning now to the interpretation of the case when there is uncertainty in the presence of the third click of the anapest rhythm. We have noted in the results the dramatic change in morphology of the potentials in the third interval even though, it should be stressed, that when the 3rd click is present it is physically identical to the regular anapest. This change in morphology was accompanied by two more subtle changes, i.e. an enhancement of the P2 for the first two intervals and a general increase in negativity after the second N2 and prior to the 3rd click. Considering the subtle changes first, although we have not provided an account here for the origin of the P2 it is established in the literature that it is modulated by attention (Picton et al., 1974). Hence one simple account is that the enhanced P2 reflects a higher level of attention in the subjects for the uncertain condition. The negativity has similarities with a contingent negative variation (CNV) and therefore could be interpreted as a correlate of a kind of "readiness for action" process consistent with a higher level of attention. The change is morphology is principally characterised by a suppression or at least a masking of the N2 by the appearance of an early P3, which could be accounted for by a prefrontal source, followed by a later N3 localised in bilateral Tpt. Thus far we have placed the N2 and SMA at the centre of our account. Its disruption by the P3/N3 complex we take therefore to indicate that a quite distinct mechanism is being activated under this condition. The fact that the early P3 appears only after the detection of the stimulus present may be taken to indicate that it is a correlate of distributed process in prefrontal cortex involved in the detection of uncertain stimuli in the context of heightened attention and a prior expectation or readiness for action. As well as detection, the distributed process associated with the P3 may facilitate an action (in the general sense of the word) contingent on the detection. The PPC is thought to play a role, in conjunction with the cerebellum and premotor cortex, in the transformation of sensory to motor coordinates for sensory-guided action (Stein, 1995). At least two areas within PPC have been identified as auditory. The first, located in the interparietal sulcus (IPS), is associated with auditory spatial processing. The second auditory area is located in the temporoparietal region (Tpt), which overlaps with the inferior parietal lobule (IPL), and is associated more with temporal processing. The source analysis located the N3 in the temporoparietal area and hence we surmise that it is a correlate of audio-motor processing, including a circuit involving the cerebellum. In addition to the change in morphology in the third interval the uncertain condition gives rise to an enhanced eN1 in the fourth interval, and this could be explained by invigoration of the pulse generator through a boost by the IPL input. The alternative simpler explanation that the eN1 is enhanced due to greater level of attention is not consistent with the fact the evoked N1s are not enhanced. Both the SMA and IPL circuits we have proposed have influence upon primary motor cortex as well as indirectly through premotor, and prefrontal cortex and cerebellum. Motor cortex controls the musculoskeletal system through motor neurons in the spinal chord. In the potentials for the active case there was evidence that the later part of the PMN was consistent with it being a correlate of activity in primary motor cortex that occurred during the onset of EMG. Any contraction of the target muscles has propriceptive and somatosensory (tactile) consequences that are conveyed back ICMPC8, Evanston, IL, USA August 3-7, 2004 216
to the brain to primary and secondary somatosensory cortex. The RAN in the active case started just after the peak of EMG and reached a peak just before or just after the target beat. This suggests that the subjects here were using a somewhat different synchronisation strategy to that described by Muller et al (2000) who found a peak in the reafferance somewhat later after the target. 5. CONCLUSION In this paper we have presented data from an experiment to investigate brain processes underlying beat induction. The data support an interpretation whereby two distinct processes are active during listening to metrical rhythms. The first process, associated with the N2, seems to be completely automatic and may be instantiated in a brain circuit including the SMA. The second process, associated with the N3, is more attention dependent and is only activated for rhythms in which the arrival of events is uncertain and may be instantiated in a brain circuit involving the IPL and Tpt. In the introduction two alternative theories were outlined, the attentional oscillator theory and the movement intention theory. The interpretation we have outlined above is not consistent with either of these in their simplest of forms. Advocates of the AO theory, on the one hand, suppose that an attentional mechanism, usually implemented as a network of oscillators, "resonates" to regular rhythms and synchronizes its own period and phase with a particular metrical level in the rhythm. Thus a beat corresponds to a window of "attentional energy" that is maximal at the target. This interpretation is difficult to reconcile with the notion of attention as we have used it. According the AO theory the more regular the rhythm, the sharper the "attentional resonance" yet our data would suggest the opposite, i.e. the more regular the rhythm the less the attentional processing. More attention is only brought into play to deal with uncertain or unexpected events. The MI theory, on the other hand, as implemented in Todd et al. (2002) attempted to account for temporal tracking and beat induction without reference to an internal pulse generation mechanism and by means of externally guided action only, in conjunction with positive feedback from reafferance. This position is difficult to reconcile with the evidence that the N2/PMN includes generators in SMA/ACC and not lateral premotor cortex as would be expected if the beat were purely externally referenced. Given the association of the SMA with internally cued actions (Passingham, 1993) then the more regular rhythms most likely make use of an internal guidance mechanism. The truth of the matter therefore lies somewhere in between the two positions. The pulse generator we have proposed could be interpreted as a kind of attentional mechanism and the emitted N1 could be viewed as a reflection of this mechanism if it were the case that it was generated by via reciprocal connections between lateral belt of auditory cortex and SMA. It is however primarily a mechanism for internal generation of action through its influence on the SMA. The current proposal is therefore a more general version of the MI theory as it incorporates both major subcortical motor control structures. When a rhythm is isochronous or highly regular little attention is required to maintain an internal pulse or active synchronization and automated circuits likely involving basal ganglia and cerebellum can deal with the required processing. For more complex rhythms involving syncopated or unexpected events undoubtedly an attention dependent externally guided mechanism comes into play. During active synchronization both corollary discharge and reafference are available, which in principle could provide the positive feedback to both subcortical circuits required to sustain continuation without need for the internal feedback. Whilst the account we have developed is consistent with the data presented it must be regarded as tentative at this stage until further experiments are conducted. Investigations are needed of the effect of variation in the parameters of beat rate, intensity, frequency and rhythm complexity (not least syncopation) in order to substantiate the proposals we have made. Additionally other imaging methods are necessary to more accurately localise the hypothesised brain areas involved. 6. REFERENCES 1. Large and Jones (1999). The dynamics of attending, Psych. Rev. 106, 119-159. 2. Moelents, D. and Van Noorden, L. (1999). Resonance in the perception of musical pulse, J. New Mus. Res. 28, 43-66. 3. Muller, K., Schmitz, F., Schnitzler, A., Freund, H., Aschersleben, G. and Prinz, W. (2000). Neuromangnetic correlates of sensorimotor synchronisation, J. Cog. Neurosci. 12(4), 546-555. 4. Naatanen, R. and Picton, T. (1987). The N1 wave of the human electric and magnetic response to sound: A review and an analysis of component structure, Psychophysiol. 24, 375- 425. 5. Naatanen, R. and Gaillard, A.W. (1983). The orienting reflex and the N2 deflection of the ERP, In A Gaillard & W Ritter (eds), Tutorials in Event-Related Potential Research. North- Holland: Amsterdam, pp 119-141. 6. Passingham, R. (1993). The frontal lobes and voluntary action. Oxford University Press, Oxford. 7. Picton, T., Hillyard, S.A., Krausz, H.I. and Galambos, R. (1974) Human auditory evoked potentials II: Effects of attention, Electroencephal. Clin. Neurophysiol. 36, 191-199. 8. Praamstra, P., Turgeon, M., Hesse, C.W., Wing, A.M. and Perrya, L. (2003). Neurophysiological correlates of error correction in sensorimotor-synchronisation, Neuroimage 20, 1283-1297. 9. Rauschecker, J. and Tian, B. (2000). Mechanisms and streams for processing of 'what' and 'where' in auditory cortex, PNAS 97(22), 11800-11806. 10. Scherg, M., Vajsar, J. and Picton, T. (1989). A source analysis of the late auditory evoked potentials,. J. Cog. Neurosci. 1(4), 336-355. ICMPC8, Evanston, IL, USA August 3-7, 2004 217
11. Stein, J. (1995) "The PPC, the cerebellum and the visual guidance of movement". In F.W. Cody (Ed.) Neural Control of Skilled Human Movement. Portland Press: London. pp 31- 49 12. Todd, N.P.M, O'Boyle, DJ and Lee, CS (2002) A sensorimotor theory of beat induction and temporal tracking, Psych. Res. 36: 11191134. ICMPC8, Evanston, IL, USA August 3-7, 2004 218