Activation of brain mechanisms of attention switching as a function of auditory frequency change

COGNITIVE NEUROSCIENCE Activation of brain mechanisms of attention switching as a function of auditory frequency change Elena Yago, MarõÂa Jose Corral and Carles Escera CA Neurodynamics Laboratory, Department of Psychiatry and Clinical Psychobiology, University of Barcelona, P. Vall d'hebron 171, 08035-Barcelona, Catalonia, Spain CA Corresponding Author Received 3 October 2001; accepted 30 October 2001 The activation of the cerebral network underlying involuntary attention switching was studied as a function of the magnitude of auditory change. Event-related brain potentials (ERPs) were recorded during the performance of a visual discrimination task in which task-irrelevant auditory frequency changes of six different levels (5%, 10%, 15%, 20%, 40% and 80%) occurred randomly within the same stimulus sequence. All the frequency changes elicited a typical ERP waveform, characterized by MMN, P3a and RON, their respective amplitudes increasing linearly as a function of the magnitude of change. The results indicate that attentional processes in the brain may follow a linear function of activation, contrasting with the well-established logarithmic functions underlying perceptual and psychophysical processes. NeuroReport 12:4093±4097 & 2001 Lippincott Williams & Wilkins. Key words: Attention shifting; Distraction; Event-related brain potentials; Mismatch negativity; Orienting response; P3a; RON INTRODUCTION Event-related brain potentials (ERPs) can be used to track from the human scalp the activation of the cerebral network engaged by the unexpected occurrence of auditory stimulus change. At early stages of processing (150 ms) after the occurrence of such changes, the mismatch negativity (MMN) [1] is generated in the auditory cortex to trigger a switch of attention towards that potentially relevant stimulus. Successful attention switches lead to an effective orienting response, which is apparently signaled over the scalp at 300 ms by the subsequent P3a waveform (for reviews see [2,3]). The P3a has been proposed to re ect the activation of a large-scale cerebral network involved in the orienting response towards novelty and change in the acoustic environment [4±8]. P3a is larger in amplitude to more attention-catching events, such as novel sounds, which supports its role in the orienting of attention. Cerebral regions involved in P3a generation include the supratemporal auditory cortex, the anterior and posterior association cortices, and the posterior hypocampus (for reviews see [2,3]). In agreement with the involvement of this large-scale cerebral network in P3a generation, recent studies have suggested that, rather than a unitary process, the P3a is composed by two consecutive phases with different functional roles and different brain areas involved in their generation [8,9]. More recently, a late ERP response involved in the sequence of brain events underlying auditory change and novelty, processing has been described. This response, called the reorienting negativity (RON) [10], is generated over frontal areas [11] when irrelevant attention-catching events occur during the performance of a given task. In the original study [10], RON was elicited when subjects had to reorient attention to relevant aspects of the auditory stimulation (i.e. duration) after distraction caused by unexpected frequency changes in the same auditory modality. Subsequent studies have also shown that RON can be elicited during visual distraction of visual performance [12] and during auditory distraction of visual performance [9]. It has been proposed that RON re ects the reorienting of attention towards original task performance after momentary distraction, as supported by its larger amplitude to more task-disrupting events [9,11,12] and by the fact of being time-locked to the events of the primary task [9]. There is evidence suggesting that the sequence of brain events underlying involuntary attention may activate increasingly as a function of increasing the magnitude of auditory change. Indeed, Tiitinen et al. [13] have shown that the peak amplitude of MMN increases linearly as a function of the logarithm of the frequency deviance, and Amenedo and Escera [14] have reported a similar logarithmic relationship between MMN amplitude and deviance by using changes in stimulus duration (increments or decrements). According to this parametric increase of MMN amplitude to increasing the magnitude of deviance, one should expect a similar behavior of the subsequent processing stages in the cerebral network of involuntary 0959-4965 & Lippincott Williams & Wilkins Vol 12 No 18 21 December 2001 4093

E. YAGO, M. J. CORRAL AND C ESCERA attention, as indexed by P3a and RON. None of the above mentioned studies, however, analyzed these ERP responses, and therefore the present study aimed at investigating their activation as a function of increasing frequency change and their relationship to behavioral distraction. MATERIALS AND METHODS Eleven healthy human subjects (18±35 years; four males; two left-handed), with normal hearing and vision, participated in the study after informed written consent. Subjects were presented with 12 blocks of 375 stimulus pairs each delivered at a rate of one pair every 1.2 s. Each pair consisted of an auditory stimulus followed after 300 ms (onset-to-onset) by a visual stimulus. Auditory stimuli were a 600 Hz standard tone (p ˆ 0.82) and six different deviant tones (p ˆ 0.03 each), which, differed in frequency from the standard tone at 5% (630 Hz), 10% (660 Hz), 15% (690 Hz), 20% (720 Hz), 40% (840 Hz) and 80% (1080 Hz), occurring randomly within the stimulus blocks. All auditory stimuli were delivered binaurally through headphones with a duration of 150 ms (including 10 ms rise and fall times) and an intensity of 75 db SPL. Pairs with a deviant tone were preceded by a pair in which the tone was standard. Visual stimuli were either an even (2, 4, 6, 8) or an odd (3, 5, 7, 9) digit, presented equiprobably on a computer screen for 200 ms with a vertical visual angle of 1.78 and a horizontal angle of 1.18 (150 cm from the subject's eyes). Subjects were instructed to use their dominant hand to press one response button with the index nger to even numbers, and another response button with the middle nger to odd numbers, and to ignore the auditory stimulation. Both speed and accuracy were emphasized. Before the experimental session, subjects received one practice block in which the sounds were turned off, all subjects reaching a hit rate level of > 85%. A correct button press within 1100 ms after visual-stimulus onset was regarded as a hit, the mean reaction time being computed only for the hit trials. Hit rate and reaction time in standard pairs was compared by means of two-tailed t-tests with performance in each of the deviant pairs. The EEG (bandpass 0±100 Hz) was continuously recorded (sampling rate 500 Hz) from the 19 scalp electrodes of the 10-20 system (Fp1, Fp2, F7, F3, Fz, F4, F8, T3, C3, Cz, C4, T4, T5, P3, Pz, P4, T6, O1, O2), and from the left (LM) and right (RM) mastoids. Horizontal and vertical EOG were recorded with electrodes attached to the canthus and below the right eye, respectively. The common reference electrode was placed on the tip of the nose. ERPs were averaged off-line for each auditory stimulus type, for an epoch of 1200 ms including a preauditory stimulus period of 100 ms. Epochs in which the EEG or EOG exceeded 100 ìv, as well as the ve rst epochs of each block, were automatically excluded from averaging. The standard-tone pairs immediately following a deviant-tone pair were also excluded from the averages. Frequencies. 30 Hz were digitally ltered out from individual ERPs. The mean amplitude of the relevant ERP components was calculated in the difference waves obtained by subtracting the ERPs to the standard pairs from those to each of the six different types of deviant pairs. The mean amplitude of MMN was calculated in the 100±200 ms time window after stimulus onset. P3a was identi ed as the largest positive peak following MMN, and its mean amplitude was calculated in the 225±350 ms time window. Following the P3a, a late negativity was identi ed in the difference waves as the RON response, its mean amplitude being computed in a 100 ms time window around its maximum peak (450±550 ms). The mean amplitude of each ERP component at Fz for each deviant condition was compared against the zero level by means of t-test comparisons in order to determine whether each component was statistically signi cant. The scalp distribution of the different ERP responses was analyzed by ANOVA for repeated measures, with deviance (six levels), frontality (three levels from anterior to posterior) and laterality ( ve levels from left to right) performed on normalized mean amplitudes at the electrodes F7, F3, Fz, F4, F8, T3, C3, Cz, C4, T4, T5, P3, Pz, P4 and T6. When appropriate, Greenhouse-Geisser correction of the degrees of freedom was applied, the uncorrected degrees of freedom, the corrected p values and epsilon values being reported. A priori contrasts were run in the ANOVAs in order to compare the scalp distributions of the ERPs between adjacent deviant conditions. Regression analyses were applied on the data in order to estimate whether the mean amplitude of each of the studied ERP components could be predicted as a function of increasing the magnitude of deviance. RESULTS Standard deviant difference waves revealed an early negative response, peaking between 100 and 200 ms, which was identi ed as the MMN (Fig. 1a). MMN was statistically different from zero for all deviant types (Table 1), its Table 1. Mean amplitude and values of the t-test comparisons of the mean amplitude at Fz of MMN (100±200 ms), P3a (225±350 ms) and RON (450±550 ms) against the zero level. Deviance MMN P3a RON ìv t values ìv t values ìv t values 5% 0.69 2.68 0.13 0.34 1.36 5.73 10% 0.74 2.83 0.67 2.24 0.98 2.06 15% 0.93 2.91 0.58 1.47 1.23 3.47 20% 1.21 5.55 0.83 2.54 1.32 4.81 40% 0.72 2.91 1.02 2.76 1.62 4.27 80% 1.37 3.61 1.69 2.66 2242 5.8 p,0.1; p,0.05; p,0.01 4094 Vol 12 No 18 21 December 2001

BRAIN IMAGING AS A FUNCTION OF AUDITORY CHANGE (a) 2 µv Fz 2 µv N1/ MMN P3a RON 800 ms increased linearly as a function of deviance (slope 0.07 ìv every 10% change), although the regression model was only marginally signi cant (F(1,4) ˆ 3.25; p ˆ 0.14) probably due to a drop of MMN amplitude in the 40% deviant condition (Fig. 2). Following the MMN, a positive response was identi ed in the difference waves as the P3a (Fig. 1a). The mean amplitude of P3a in the 225±350 ms time window was signi cantly different from zero for the 10%, 20%, 40% and 80% deviancies, and tended to be signi cant in the 15% deviant condition (t(10) ˆ 1.47; p ˆ 0.17; Table 1). Difference waves revealed a different shape of the P3a in the 80% deviant condition, with an evident later phase as compared with the other conditions [8] (Fig. 1a). ANOVA showed a signi cant interaction between the factors deviance and frontality (F(10,100) ˆ 3.58; p, 0.02; å ˆ 0.37), indicating different scalp distributions of P3a across deviancies. A (b) 5% 10% MMN 2 1.8 1.6 1.4 1.2 1 0.8 0.6 0.4 0.2 0 Y ( 0.75) ( 0.007)*X p 0.14 15% 20% 40% P3a 2 1.8 1.6 1.4 1.2 1 0.8 0.6 0.4 0.2 0 Y 0.32 0.017*X p 0.04 80% / 1.3 µv / 2 µv / 2.5 µv Fig. 1. (a) Standard deviant difference waveforms at Fz for the six types of deviant conditions. (b) Scalp distribution maps of MMN, P3a and RON for the six types of deviant conditions. latency decreasing progressively with increasing the magnitude of the frequency change, as can be seen in Fig. 1a. ANOVA showed no signi cant interaction of the factor deviance with the factors frontality and laterality, suggesting no statistical differences in the scalp distribution of MMN elicited to the different deviant tones, which had a frontal-central maximum (Fig. 1b). The amplitude of MMN 0 Mean amplitude (µv) RON 2.5 2.3 2.1 1.9 1.7 1.5 1.3 1.1 0.9 0.7 0.5 Y ( 1.04) ( 0.015)*X p 0.005 5 10 15 20 40 80 Deviance in Hz (%) Fig. 2. Mean amplitudes (ìv) of the MMN (100±200 ms), P3a (225± 325 ms) and RON (450±550 ms) responses at the Fz electrode as a function of frequency change. Regression equations and their statistical signi cance are shown below each regression line. Vol 12 No 18 21 December 2001 4095

E. YAGO, M. J. CORRAL AND C ESCERA priori contrasts revealed no signi cant differences in scalp distribution for adjacent deviant conditions, but a signi cant different scalp topography between the 40% and the 80% deviant conditions (F(1,10) ˆ 26.02; p, 0.001), which was due to the more posterior distribution of P3a in the largest deviant condition, as can be seen in the distribution maps (Fig. 1b). Regression analyses showed that the amplitude of P3a increased linearly as a function of the frequency deviance (F(1,4) ˆ 36.2; p, 0.05), with a slope of 0.17 ìv every 10% change (Fig. 2). Subsequent to P3a, a late negative response was also observed in the difference waves (Fig. 1), peaking at about 500 ms, which was identi ed as the reorienting negativity or RON. The mean amplitude of RON was signi cantly different from the zero level for all types of deviant pairs (Table 1). RON had a frontal distribution which did not vary across deviancies (i.e., no signi cant interactions between the factors deviance, frontality and laterality were found in ANOVA; Fig. 1b). The mean amplitude of RON increased linearly as a function of deviance (F(1,4) ˆ 30.87; p, 0.01), with a slope of 0.15 ìv every 10% change (Fig. 2). t-test comparisons revealed signi cant distraction effects in the 10% deviant condition, as reaction time increased signi cantly by 7.7 ms (t 10 ˆ 3.24; p, 0.01) when visual stimuli were preceded by a 10% deviant tone in comparison to those visual stimuli preceded by a standard tone. In contrast, reaction time decreased by 11.3 ms (t 10 ˆ 5.21; p, 0.001) in the 5% deviant condition, suggesting a possible facilitatory effect in this condition. No signi cant behavioral effects were found in any of the other deviant conditions. DISCUSSION The occurrence of changes in sound frequency elicited a three-phase ERP waveform, in agreement with previous studies using similar distraction paradigms [8±10,12]. At early latencies ( 150 ms), the MMN was elicited to all deviance levels, its mean amplitude tending to increase linearly as a function of the magnitude of change. The similar scalp distribution of the MMN in all deviant conditions suggests larger activation of the same brain areas involved in MMN generation, rather than the contribution of additional brain regions to large deviancies. The pattern of growth of MMN amplitude to increasing deviance found here contrasts with the logarithmic functions described in previous studies [13,14]. These logarithmic functions were obtained in passive conditions, in the context of pre-attentive perceptual discrimination, and were similar to those described in animals at neural level, underlying visual [15] or somatosensory [16] perception. In the present study, although irrelevant to task performance, the MMN-eliciting sounds could have been attended as they warned the occurrence of the subsequent visual imperative stimulus [8]. In such a condition, some N1 enhancement, shown to be involved in frequency change detection [17] and contributing to the deviant standard difference wave to large deviancies [18], may have been modulated by attentional demands [19,20]. A previous study has also shown a linear increase of the neural activity, measured by fmri, in primary visual cortex to increasing behavioral accuracy in a visual-cued attention task [21]. Together with the linear functions obtained for the P3a and RON responses, these results suggest that attentional processes may follow different parametric laws than the well-established logarithmic functions underlying perceptual and psychophysical processes. Following the MMN, the P3a and RON responses were elicited, their amplitude increasing linearly as a function of deviance, with a slope of growth about double than that of the MMN. The scalp distribution of the P3a was more posteriorly distributed in the 80% deviant condition than in the smaller deviant conditions, suggesting the engagement of different brain regions for the largest frequency change. This parallels the different shape of the P3a waveform in this condition, in which a second phase of the response appeared evident. In previous studies, two stages of the P3a elicited to novel sounds have been described, with different functional roles and separate brain areas contributing to their generation as suggested by different scalp distributions [8,9]. The present results point to the involvement of the second phase of the P3a in the processing of stimulus salience, as it was observed only for the largest deviant condition. At a third stage of activation of the cerebral network of involuntary attention, the RON response was elicited in all deviant conditions. The scalp distribution analyses revealed that, similarly to MMN, the same cerebral regions were engaged during its generation for all levels of change, these being activated more strongly to larger deviancies according to a linear function. Interestingly, the present experiment yielded a lack of parallel behavior between the electrophysiological data and task performance, which needs clari cation in future studies. Indeed, the larger activation of the brain network underlying involuntary attention with increasing the magnitude of change was not paralleled by an attentional engagement re ected in behavioral distraction, i.e. subjects were only distracted in the 10% deviant condition, whereas a facilitatory effect was observed in the 5% deviant condition. At least two possible explanations may account for this lack of electrophysiological±behavioral parallel. First, it is plausible that behavioral distraction can only be observed when an optimal range of cerebral activation is reached, i.e, that obtained for a deviance around 10% [8,9]. Once the critical interval of distracting deviance is surpassed, the larger activation of the brain network underlying involuntary control of attention may re ect a compensatory effect that would prevent behavioral distraction from being manifested. In this context of interpretation, the larger RON response to increasing magnitude of change may therefore arise from a compensatory effort to reorient attention effectively towards the original task. The present results also suggest that RON re ects the reorienting to the originary task not necessarily after actual behavioral distraction [9,10,12], but after processing temporarily the preceding distracting stimulus, a processing which is in turn re ected in the preceding P3a response. The facilitatory effect observed in the 5% deviant condition can also be explained in this context. A small deviant change may suf ce to activate change detection, as indexed by MMN, but may fail to trigger successive orienting of attention, as indexed by the lack of a signi cant P3a. However, these small changes succeed in causing facilitatory effects on performance, probably via an enhancement of the processing of the subsequent visual stimuli. Recent 4096 Vol 12 No 18 21 December 2001

BRAIN IMAGING AS A FUNCTION OF AUDITORY CHANGE evidences have suggested that sounds can facilitate performance of a visual task. For example, cross-modal facilitation has been reported by McDonald et al. [22], who found better performance to visual targets preceded by warning tones informing on their location. Similar results have been reported by Hackley and Valle-InclaÂn [23] who found electrophysiological facilitation (i.e. earlier latencies of the lateralized readiness potential) at early stages of perceptual processing to visual targets presented simultaneously with a task-irrelevant sound. The present results suggest that, when a visual target is shortly preceded by an auditory change, facilitatory effects can be obtained for those auditory changes near the discrimination level. Alternatively, the context of stimulation may have had a determinant in uence in the behavioral results, as in previous studies distraction effects were obtained with different contextual stimulation. In these studies, subjects were presented with a similar auditory±visual paradigm in which the auditory stimulation intermixed deviant tones and novel sounds among the repeating standard tones [8,9], whereas in the present study no novel sounds were presented. It is plausible that the occurrence of novel sounds make the system controlling the orienting of attention more vulnerable to distraction due to a lack of brain resources to cope with novelty, whereas the occurrence of different deviant tones may not overload the brain resources to cope with distraction. Indeed, recent evidence [24] has shown that higher memory load results in greater interference effects on behavioral performance from distractor stimuli, as compared to a low memory load condition. These results may also explain why other authors [12] have found distraction effects on performance to small deviancies, as the distractors were presented in the same sensory modality than the main task, probably interfering much more than in the present study, in which distractors were presented in a different sensory modality than the stimuli of the primary task. Future studies should therefore consider the context of stimulation used to study the cerebral basis of involuntary attention and distraction, as it may affect the load of memory resources and, consequently, the interference of distractors on executive performance. CONCLUSION We found that the cerebral network underlying auditory change processing, as indexed by MMN, P3a and RON, activates linearly as a function of increasing the magnitude of change in sound frequency. Contrasting with the wellknown logarithmic functions describing perceptual and psychophysical processes, the present results suggest that attentional processes are best characterized by linear functions. Scalp distribution analyses revealed that the same brain regions contributed to the generation of MMN and RON responses in all deviant conditions, whereas additional brain regions were engaged in salience processing, as indicated by the different scalp distribution and an additional later phase of the P3a elicited to the largest deviance. Surprisingly, the linear increase of activation in this cerebral network was not paralleled by an attentional engagement re ected in behavioral distraction, as subjects were only distracted in the 10% deviant condition, whereas a facilitatory effect was observed in the 5% deviant condition. This lack of electrophysiological±behavioral agreement re ecting either low working memory load or the existence of an optimal range of activation in this cerebral network for interfering executive performance, remains to be clari ed in future studies. REFERENCES 1. NaÈaÈtaÈnen R. Attention and Brain Function. Hillsdale: Erlbaum; 1992. 2. Escera C, Alho K, SchroÈger E et al. Audiol Neurootol 5, 151±166 (2000). 3. Friedman D, Cycowicz YM and Gaeta H. Neurosci Behav Rev 25, 355±373 (2001). 4. Squires NK, Squires KC and Hillyard SA. Electroencephalogr Clin Neurophysiol 38, 387±401 (1975). 5. Knight RT. Electroencephalogr Clin Neurophysiol 59, 9±20 (1984). 6. Grillon C, Courchesne E, Ameli R et al. Arch Gen Psychiatry 47, 171±179 (1990). 7. Woods DL. Electroencephalogr Clin Neurophysiol 84, 456±468 (1992). 8. Escera C, Alho K, Winkler I et al. J Cogn Neurosci 10, 590±604 (1998). 9. Escera C, Yago E and Alho K. Eur J Neurosci 14, 877±883 (2001). 10. SchroÈger E and Wolff C. Neuroreport 9, 3355±3358 (1998). 11. SchroÈger E, Giard MH, Wolff C. Clin Neurophysiol 111,1450±1460 (2000). 12. Berti S and SchroÈger E. Cogn Brain Res 10, 265±273 (2001). 13. Tiitinen H, May P, Reinikainen K et al. Nature 370, 90±92 (1994). 14. Amenedo E and Escera C. Eur J Neurosci 12, 2570±2574 (2000). 15. Seidemann E, Zohary E and Newsome WT. Nature 394, 72±75 (1998). 16. Romo R, HernaÂndez A, Zainos A et al. Neuron 26, 273±278 (2000). 17. Yago E, Escera C, Alho K et al. Neuroreport 12, 2583±2587 (2001). 18. Scherg M, Vajsar J and Picton TW. J Cogn Neurosci 1, 336±355 (1989). 19. Hillyard SA, Hink RF, Schwent VL et al. Science 182, 177±180 (1973). 20. Teder W, Alho K, Reinikainen K et al. Psychophysiology 30, 71±81 (1993). 21. Rees D, Backus BT and Heeger DJ. Nature Neurosci 3, 940±945 (2000). 22. McDonald JJ, Teder-SaÈlejaÈrvi WA and Hillyard SA. Nature 407, 906±908 (2000). 23. Hackley SA and Valle-InclaÂn F. Nature 391, 786±788 (1998). 24. de Fockert JW, Rees G, Frith CD et al. Science 291, 1803±1806 (2001). Acknowledgements: This research was supported by grant PM99-0167 of the Spanish Ministry of Science and Technology and the Generalitat de Catalunya (1999FI-003 82-PGUB). We thank Vanessa Carral for her valuable help in the preliminary phases of the study. Vol 12 No 18 21 December 2001 4097