- Email: [email protected]
Interference between two concurrent tasks is associated with activation of overlapping fields in the cortex
Cognitive Brain Research 6 Ž1997. 1–8
Research report
Interference between two concurrent tasks is associated with activation of overlapping fields in the cortex Torkel Klingberg ) , Per E. Roland DiÕision of Human Brain Research, Department of Neuroscience, Karolinska Institute, Doktorsringen 12, S-171 77 Stockholm, Sweden Accepted 18 February 1997
Abstract Interference between two concurrent tasks can be measured as an increased reaction time during simultaneous performance compared to when each task is performed alone. We tested the hypothesis that two tasks interfere because they require activation of overlapping areas of the cerebral cortex. With positron emission tomography we measured cortical activation as fields with significant increase in regional cerebral blood flow during single task performance of an auditory and a visual gorno-go task and an auditory and a visual short-term memory ŽSTM. task. In a separate experiment we measured the degree of interference between the two gorno-go tasks and between the two STM tasks during dual task performance. Both the two gorno-go tasks and the two STM tasks activated overlapping parts of the cortex and interfered significantly during dual task performance. The two STM tasks had a larger volume of overlap and also significantly larger increase in reaction time during dual task performance, compared to the gorno-go tasks. The results thus indicate that two concurrent tasks interfere, with a resulting increase in reaction time, if they require activation of overlapping parts of the cortex. q 1997 Elsevier Science B.V. Keywords: Short-term memory; Working memory; Reaction time; Regional cerebral blood flow ŽrCBF.; Positron emission tomography ŽPET.
1. Introduction When two tasks are performed concurrently, there is often an impairment in performance, which can be measured as an increase in error rate or reaction time w2,5x. Such dual task interference has been described as a competition for attentional resources w9,14x, or competition for information processing mechanisms w16,32x. However, the neural mechanisms underlying dual task interference are not known. There has been no systematic comparison of the pattern of cerebral activation, as seen with positron emission tomography ŽPET. or MRI, with dual task interference. With PET one can measure cortical activation as volumes of cortex which increase their metabolism when the brain is activated. What appears to be continuous areas of activations in the PET images, are probably columns w21x or patches w12x of active synapses. If two tasks significantly increase the regional cerebral blood flow ŽrCBF. in an identical volume of cortical tissue, they are likely to
)
Corresponding author. Fax: q46 Ž8. 309045; E-mail: [email protected] 0926-6410r97r$17.00 q 1997 Elsevier Science B.V. All rights reserved. PII S 0 9 2 6 - 6 4 1 0 Ž 9 7 . 0 0 0 1 0 - 4
interfere with each other, either because the two tasks depend on activation of the very same population of neurones, or because the two tasks activate neuronal populations within the volume, which are non-identical but inhibit each other by local mechanisms when activated simultaneously w13,28,31x. Both mechanisms would disturb the neuronal processing and decrease the probability of a correct answer. Alternatively, the cortical field or the neuronal population which is required for two tasks could first be dedicated to one task and then to the other, which would increase the mean reaction time. The hypothesis in this study was then that if the areas of cortex that are activated by task A overlaps with the areas of cortex activated by task B, then the two tasks will interfere. We measured significant increases in rCBF during a task relative to a passive control state as clusters of voxels having high Student’s t-values w25x. The degree to which two tasks activate identical areas of cortex can then be determined as the overlaps between these clusters of significant blood flow increase in the two tasks. We analysed the volume and location of the overlaps in PET data from two gorno-go tasks and two short-term memory ŽSTM. tasks w17x. The rCBF increase was measured only during single task performance, not during dual task performance.
2
T. Klingberg, P.E. Rolandr CognitiÕe Brain Research 6 (1997) 1–8
In a separate psychophysical experiment, with a different group of subjects, the interference between the tasks during dual task performance was evaluated and compared to the overlaps in the PET data.
2. Materials and methods 2.1. Psychophysical experiment 2.1.1. Subjects Fifteen paid, healthy volunteers, 12 males and 3 females, mean age 23.4 years, participated. 2.1.2. Tasks In the visual gorno-go task ŽFig. 1, task 1. the luminance of a circular field on a computer screen changed at random intervals Ž2.00–5.00 s.. The subjects were asked to respond when they detected a decrease in luminance, but not when there was an increase in luminance. In the auditory gorno-go task ŽFig. 1, task 2. a tone changed in frequency at random intervals Ž2.00–5.00 s.; the subjects were asked to respond when the tone changed from a higher frequency to a lower one, no response was required when the tone changed from a lower frequency to a higher one. In the simultaneous gorno-go task ŽFig. 1, task 3., the visual and the auditory gorno-go tasks were performed simultaneously. The dual task conditions always consisted of performing one visual and one auditory task to avoid perceptual interference between the tasks w20x. In the dual tasks, stimuli varied randomly and independently of each other, with the only restriction that a change in one sensory channel was always separated by at least 1.00 s from a change in the other sensory channel. This was to avoid what has been termed the ‘‘psychological refractory period’’, i.e., an increased reaction time when the two stimuli which call for a response are separated by less than 300 ms w24x. In the visual STM task ŽFig. 1, task 4., a circular field on a computer screen changed in luminance between one of six high luminance levels – target levels – and a baseline luminance level. The 1.00 s stimulation at target luminance levels were separated by 2.00–5.00 s delays with baseline luminance. The subjects were asked to compare each target luminance level with the previous one, and to respond if the last luminance level was lower than the previous one; no response was required when there was a change from a lower to a higher luminance. In the auditory STM task ŽFig. 1, task 5., a series of high pitched tones – target tones – of different frequencies were presented. Target tones were separated by temporal intervals of 2.00–5.00 s, during which a low pitched tone – baseline tone – was presented. The duration of the target tones was 1.00 s. The subjects were asked to compare each target tone with the previous target tone, and to respond only if the last target tone was lower in frequency than the
Fig. 1. Experimental design of the gorno-go and the STM tasks. The lines indicate how the stimuli changed in frequency Žauditory tasks. or luminance Žvisual tasks. over time. The arrows point at the two frequency levels Žor luminance levels. which were compared by the subjects. The subjects were asked to respond if the second level Žbelow the second arrowhead. was lower than the previous one Žbelow the first arrowhead..
previous one ŽFig. 1, task 5.. In the simultaneous STM condition ŽFig. 1, task 6., the auditory and the visual stimuli varied independently of each other, with the only restriction that a change in one sensory channel was always separated by at least 1.00 s from a change in the other sensory channel. The gorno-go and the STM tasks were designed to be similar with respect to sensory stimulation, selective attention, response selection and motor output. The difference was the delay between the stimuli in the STM tasks, a delay which required stimulus information to be kept ‘‘on-line’’ for a brief period. 2.1.3. Stimuli The visual stimulus consisted of a circular field of 15 cm in diameter, which was presented against a black background on a computer screen, 100 cm from the subjects in a completely dark room. The minimum and maximum luminance values were within 18–36 cdrm2 in the detection tasks, and 4–18 cdrm2 in the STM tasks. In the STM tasks there was also an inter-stimulus ‘‘baseline’’ of luminance that was fixed at 3 cdrm2 .
T. Klingberg, P.E. Rolandr CognitiÕe Brain Research 6 (1997) 1–8
The auditory stimuli were tones presented through a pair of earphones. The minimum and maximum frequencies were within 650–780 Hz. In the STM tasks there was an inter-stimulus ‘‘baseline’’ frequency that was fixed at 550 Hz. During a pre-test session the stimulus levels were adjusted so that the subjects would achieve an aROC value of approximately 0.85 in the non-simultaneous tasks. Reaction times were recorded by the same computer that controlled the stimuli and were defined as the time from the onset of the stimulus to the pressing of the button. Correct responses should occur within 2.00 s in order to be counted as a ‘‘hit’’. In both the gorno-go tasks and the STM tasks the subjects pressed the left button on a computer mouse with their right index finger when they responded to visual stimuli and pressed the right button with their right third finger when they responded to auditory stimuli. Failure to respond within 2.00 s after stimulus onset was taken as a ‘‘miss’’. Responses on non-target trials Žwhen the stimulus was higher in luminance or frequency than the previous one. were registered as false alarms. 2.1.4. Procedure The tasks were performed in blocks of six Žfor example task numbers 4, 6, 5, 2, 1, 3. Žsee Fig. 1.. Each task was thus performed once before any task was repeated. The order of the tasks within a block was balanced over the whole test session. A total session consisted of four blocks. Each task was performed for 90 s. After a 10 s break a text appeared on the screen, describing the next task, e.g. ‘‘simultaneous performance of gorno-go’’. 2.1.5. Interference measurement During dual task performance, there is a possibility that subjects give priority to one task over the other, which could increase the reaction time and error rate more in the task that is not given priority. We were not interested how such priorities were given. The increases in reaction time were therefore computed by subtracting the sum of the reaction times for the two tasks in the single condition ŽRT,aud,single and RT,vis,single. from the sum of the reaction times during simultaneous performance:
D RT s Ž RT ,aud, sim q RT ,Õis, sim . y Ž RT ,aud, single q RT ,Õis, single . . W here aud s auditory, vis s visual, sim s simultaneous. The gorno-go D RT values were then compared to the STM D RT values. We calculated the area under receiver-operating characteristic curve ŽaROC. values to assess the performance w10x. When we talk of increase in error rate, this refers to a decrease in aROC values. Data were sampled for each person giving only one point on the ROC. In this case the aROCs Ž P Žhit. y P Žfalse alarm..r2 q 0.5; where P s probability. This calculation gives an aROC of 0.5 at
3
chance level performance and a value of 1.0 at perfect performance. The change in error rate was calculated as the sum of aROC during the simultaneous performance minus the sum of the aROC during the two single presentations:
D aROCs Ž aROC,aud, sim q aROC,Õis, sim . y Ž aROC,aud, single q aROC,Õis, single . For each subject, D RT and DaROC was calculated for the gorno-go tasks and for the STM tasks. 2.1.6. Test for normality All data of D RT and DaROC were tested to see if the distribution differed significantly from a normal distribution. The DRT values were found not to be normally distributed ŽShapiro–Wilk W-test, P - 0.01 for both the gorno-go and the STM group.. By visual inspection we detected three values of DRT being outliers which were found to be 5.0 S.D. or more from the mean of the rest of the values within each group and could thus not be considered to belong to the same distribution as the rest of the points within each group w11x. A possible explanation for this could be that the subjects had not followed the rule of answering as quickly as possible on all tasks. The three outlier points of D RT were excluded from further analysis. After exclusion the distribution of D RT was normal for both the STM and the gorno-go group ŽShapiro–Wilk W-test, P ) 0.10.. The median value for D RT was not changed by the exclusion and was 0.06 s for the gorno-go group and 0.18 s for the STM group, both before and after exclusion of the three outliers. 2.2. PET experiment Ten subjects had their rCBF measured during a visual gorno-go task Žtask 1 in Fig. 1., an auditory gorno-go task Žtask 2 in Fig. 1. and during a control state. In the control state the subjects only received passive auditory and visual stimulation without any task to perform. Another group of 10 subjects had their rCBF measured during an visual STM task Žtask 4 in Fig. 1., during an auditory STM task Žtask 5 in Fig. 1. and during a control state which was identical to the control state in the gorno-go experiment. Control rCBF was subtracted from task rCBF and significant activations were determined as clusters of voxels with rCBF increase above a certain threshold, these images are called cluster images. The cluster images ŽSTM visualycontrol . and ŽSTM auditoryycontrol . were then compared with each other, voxel by voxel. Corresponding voxels that were significant in both subtraction images are called overlapping voxels. This is illustrated in Fig. 2. Clusters from the auditory subtraction image are shown in green. Clusters from the visual subtraction image are shown in blue, and the overlapping voxels are shown in red.
4
T. Klingberg, P.E. Rolandr CognitiÕe Brain Research 6 (1997) 1–8
Exactly the same procedure was carried out to create an overlap image from the two subtraction images Žauditory gorno-go – control. and Žvisual gorno-go – control.. 2.2.1. Scanning procedure and statistics Subjects received an intravenous bolus injection of 70 mCi of 15 O-butanol before each PET scan. Local tissue radioactivity was measured with an 8-ring Ž15-slice. PC2048-15B Scanditronix PET scanner. Images were reconstructed with a 4 mm Hanning filter, giving a final image resolution of approximately 4.2 mm. Values of rCBF were calculated using a dynamic algorithm w18x. Global blood flow was normalized so that white matter blood flow was 20 mlr100 grmin in all images. The rCBF images were transformed to a standard format using
the MRI of each subject and a computerised brain atlas w25x. Statistical analysis was performed as described by Roland et al. w27x. Task minus control subtraction images were made for each subject through voxel by voxel subtraction of the rCBF of the control from each of the task images. A Student’s t-value was calculated for each voxel. Voxels above a certain t-threshold and occurring in clusters above a certain size were considered significant, where the t-threshold and cluster size were set so that there was a probability of 0.6 for having one or more false positive clusters in the whole brain w27x. The non-significant voxels were set to zero. This image is called a cluster image. This article only concerns clusters that were found in the same place in two different subtraction images. The probability
Fig. 2. Illustration of the overlap in cortical activation between the two STM tasks. Clusters from the auditory minus control subtraction image are shown in green. Clusters from the visual minus control subtraction image are shown in blue, and the overlapping voxels are shown in red. The clusters are superimposed on a horizontal section of the MR image used for anatomical standardisation.
T. Klingberg, P.E. Rolandr CognitiÕe Brain Research 6 (1997) 1–8
of two clusters, e.g. 0.5 cm3 large, occurring at the same place within the brain, and both clusters being false positives, is of course smaller than 0.6. An apparent overlap could theoretically arise if two adjacent but non-overlapping areas of rCBF increase are measured with a method with low spatial resolution. However, all the reported areas of overlap extended at least 5 mm from border to border along an axis connecting the centres of gravity of the two clusters. This is more than the spatial resolution of the PET camera and ensures that the overlaps are real and not a mere artefact due to low spatial resolution. It should also be kept in mind that a significant cluster, based on measured rCBF changes from 10 subjects, probably is smaller than the neuronal volume activated in each of the 10 individuals, since the cortical activation is not located in exactly the same place in each of the individuals, and the volume of increase that reaches significance level across subjects is located over the area where the majority of subjects has increased metabolism. The volume of each of the overlaps is for the same reason probably underestimated. The volumes of overlap reported here can therefore not be taken as absolute and quantitative measurements, but only permit relative comparisons.
3. Results
5
Table 2 Location and volume of the cortical overlaps between the auditory and the visual STM task Area
Coordinates
Sup. frontal g ŽR. Inf. frontal g ŽL. Frontal operculum ŽL. Cingulate cortex ŽR. Inf. parietal cortex ŽR. Ventromedial frontal ŽR. Orbitofrontal cortex ŽR.
y20, 47, 41 33, 55, 0 41, 10, 16 y2, 16, 45 y40, y48, 49 y7, 24, y7 y11, 42, y15
mm3 90 230 300 260 240 190 280
S 1590 Sup.ssuperior; inf.s inferior; g s gyrus; L s left hemisphere; R s right hemisphere.
percentage overlap varied between 19 and 48% in the gorno-go tasks and between 18 and 74% in the STM tasks. The mean percentage overlap was 33% for the gorno-go tasks and 50% for the STM tasks. The percentage overlap was also calculated as O divided by the mean volume of A and B, which resulted in a mean percent overlap for the gorno-go tasks of 27% and for the STM tasks of 40%. The overlaps were thus larger between the two STM tasks, also relative to the volume of the cortical activations that formed the overlaps. 3.2. Task performance during PET scanning
3.1. OÕerlap in cortical actiÕation Location and volume of the cortical overlaps are shown in Tables 1 and 2. The gorno-go tasks overlapped in six locations with a total volume of 820 mm3. The STM tasks overlapped in seven locations with a total volume of 1590 mm3 ŽFig. 3.. To assess if the volume of overlap was related to the size of the clusters forming the overlap, we also analysed the percentage of overlap for each cluster. If a cluster A overlaps with a cluster B Žwhere A is smaller than B. and thus creates an overlap O, then the percentage overlap for O is the volume of O divided by the volume of A. The
Performance was evaluated as aROC values and was 0.93 Ž"0.03. and 0.84 Ž"0.03. in the auditory and visual gorno-go tasks respectively, and 0.96 Ž"0.01. and 0.95
Table 1 Location and volume of the cortical overlaps between the auditory and the visual gorno-go task Area
Coordinates
Sup. frontal g. ŽR. Inf. frontal g. ŽR. Sup. frontal g mesial part ŽR. Frontal operculum ŽR. Sup. temporal g. ŽR. Lingual g. ŽL.
y7, 13, 55 y32, 48, y1 y4, 63, 12 y43, 20, 1 y53, y24,y7 28, y82, y23
mm3 120 160 180 80 180 100
S 820 Sup.ssuperior; inf.s inferior; g s gyrus; L s left hemisphere; R s right hemisphere.
Fig. 3. The total volume of overlap between the two gorno-go tasks Žleft bar. and between the STM task Žright bar.. The superimposed line-graph indicates the increase in reaction time Ž D RT. during dual compared to single performance of the gorno-go and STM tasks. The D RT was significant for both tasks Ž ) P - 0.05; ) ) P - 0.01. and was significantly larger for the STM tasks compared to the gorno-go tasks Ž P 0.05..
T. Klingberg, P.E. Rolandr CognitiÕe Brain Research 6 (1997) 1–8
6
Table 3 Error rates and reaction times in the gorno-go tasks
RT Žs. aROC
4. Discussion
Single performance
Simultaneous performance
Visual
Auditory
Visual
Auditory
0.80 Ž"0.07. 0.82 Ž"0.03.
0.94 Ž"0.06. 0.85 Ž"0.03.
0.79 Ž"0.04. 0.79 Ž"0.03.
1.00 Ž"0.06. 0.83 Ž"0.03.
RT s reaction time; aROC sarea under receiver-operating characteristic curve; Ž"S.E.M...
Ž"0.03. for the auditory and visual STM tasks respectively. 3.3. Psychophysical experiment Each of the 15 subjects performed all three gorno-go tasks Žtasks 1, 2 and 3 in Fig. 1. and all STM tasks Žtasks 4, 5 and 6 in Fig. 1.. After controlling that the values were normally distributed, means and standard error of means ŽS.E.M.. were calculated for D RT and DaROC. The increase in reaction time Ž D RT computed as described in Section 2.1.5. during gorno-go was 0.05 Ž"0.02. s, and during STM 0.19 Ž"0.05. s ŽFig. 3.. This increase was significant for both tasks Ž P - 0.05 for gorno-go tasks, and P - 0.01 for the STM tasks, one-tailed Student’s t-test.. Furthermore, the increase was significantly larger in the STM tasks than in the gorno-go tasks Ž P - 0.05.. Accuracy and reaction times in the STM and gorno-go tasks are shown in Tables 3 and 4. The decrease in aROC Ž DaROC computed as described in Section 2.1.5. during the simultaneous performance was significant for both tasks: y0.055 Ž"0.027. and y0.052 Ž"0.026., for the gorno-go and STM tasks respectively Ž P - 0.05, one-tailed Student’s t-test.. To assess the effect of task difficulty in the single task on the increase in error rate during dual performance, we computed the correlation between the mean aROC value for the two single tasks ŽŽ aROC,aud,single q aROC,vis,single.r2. and the DaROC for each individual. The correlation between these two variables was not significant Ž r s y0.25, P ) 0.18.. Task difficulty in the single tasks did thus not significantly effect the relative increase in error rate Ž DaROC. during simultaneous performance. Neither was the mean aROC value for the two single tasks correlated with the relative increase in reaction time, D RT Ž r s 0.01, P ) 0.94.. Table 4 Error rates and reaction times in the STM tasks
RT Žs. aROC
Single performance
Simultaneous performance
Visual
Auditory
Visual
Auditory
0.90 Ž"0.05. 0.85 Ž"0.02.
0.83 Ž"0.04. 0.86 Ž"0.02.
0.96 Ž"0.05. 0.82 Ž"0.02.
0.96 Ž"0.05. 0.84 Ž"0.02.
RT s reaction time; aROC sarea under receiver-operating characteristic curve; Ž"S.E.M...
The cortical fields activated by the auditory gorno-go task overlapped with the fields activated by the visual gorno-go task in six locations. The total overlap was 820 mm3. The two STM tasks had overlapping activations in seven locations with a total volume of 1590 mm3. At an anatomical and macroscopic level, the cortical activation in the areas of overlap was obviously not specific to the sensory modality of the stimuli and was located outside the primary sensory areas. No cortical overlaps were located in primary motor cortex, premotor cortex or the supplementary motor area ŽTables 1 and 2.. When the two gorno-go tasks were performed simultaneously there was a significant interference between the tasks, detected as an increase in error rate as well as in reaction time, compared to when the gorno-go tasks were performed one at a time. Simultaneous performance of the two STM tasks was similarly associated with significant increases in both reaction time and error rate. In these two independent PET studies, overlapping cortical activity between single tasks was associated with interference during simultaneous performance, in accordance with the hypothesis. Furthermore, the larger volume of overlap in the STM tasks was accompanied by a significantly larger increase in reaction time, compared to the gorno-go tasks ŽFig. 3.. This is in accordance with the hypothesis, but the data are not sufficient to specify the relationship between overlap and increase in reaction time, for example we cannot say that there is a linear relation between the two variables. It should be pointed out that the size of the overlaps is not absolutely quantitative, but depends on the sensitivity of the PET camera, the data treatment, the statistical method for evaluating the data and the level of statistical significance chosen. The comparison of two pairs of tasks is therefore always relative and requires that the methods are identical for both pairs of tasks, which was the case in the present study. Kinomura et al. w15x showed with PET that when the brain goes from a relaxed state into a state of increased arousal, as happens when one performs a reaction time task, this increases the blood flow in the midbrain reticular formation and the thalamus. This finding suggests that there will always be an overlap between two tasks which involve an increase in arousal compared to a passive sensory stimulation during control. If such subcortical overlaps result in interference is not known. We did not perform any PET scanning during dual task performance. Even if such scanning had been performed, the limited time resolution of the PET method makes it impossible to decide whether the two tasks actually activate the very same neuronal population at the same time, or if there is a time-sharing activity, as discussed in Section 1. The STM tasks had about twice as large volumes of overlap as had the two gorno-go tasks, but the increase in error rate was equally large in both pairs of tasks. This
T. Klingberg, P.E. Rolandr CognitiÕe Brain Research 6 (1997) 1–8
difference in total volume of cortical overlap, but lack of difference in error rate, could be taken as evidence that error rate is not related to cortical overlaps at all. It is possible that the physiological basis for errors is not the same as the one for reaction time increases. It is also possible that one needs to take into account the location of the cortical overlaps to explain the error rates, while only reaction time increases are related to the total volume of overlap, independent of the location of the overlaps. None of the overlaps between the two gorno-go tasks was located at exactly the same site as the overlaps between the two STM tasks ŽTables 1 and 2., which demonstrates that the interference was not a result of any particular and specifically located overlap. It is not possible from the present data to determine whether the interference was caused by particular overlaps or due to the fact that overlaps were present at many sites. The increased reaction time during the simultaneous performance of the STM tasks is in agreement with findings in previous studies of dual task performance, where reaction time increases with increasing short-term memory load w2,19,30x. Task difficulty is sometimes suspected to be a confounding factor when evaluating the amount of interference between two tasks. However, there was no significant correlation between task difficulty Žmean aROC. in the single tasks and the increase in reaction time during simultaneous performance. Roland w26x, p. 424, suggested that the synapses in cortex always are activated in large ensembles called functional cortical fields, and that two tasks which activate an identical functional field cannot be performed simultaneously. The hypothesis in this article is similar but more generally formulated, since all overlap in cortical activity between two tasks could cause interference, as discussed in Section 1. Our data show that interference is associated with overlaps of less than 50% of the overlapping activated fields, at least when multiple such overlaps exist, as is shown by the interference between the two gorno-go tasks ŽTable 1.. The hypothesis of this article also bears resemblance to ideas formulated by Kinsbourne and Hicks w16x and Wickens w32x. Nestor et al. w22x reported that in Alzheimer’s patients, decreased metabolism in the right parietal lobe is correlated with the relative increases in reaction time during dual task performance. Other studies show that patients with Alzheimer’s disease are impaired with respect to both reaction time and errors w3,4x. Increases in reaction time during dual task performance, compared to single task performance, have also been shown in patients with close head injuries w29x. In the present article we only dealt with PET data from single task performance and tried to make predictions about dual task performance. It has been suggested that performance of dual tasks requires some additional factor, such as a time-sharing ability w1,6,8x, ‘‘supervisory function’’ w23x or ‘‘central executive function’’ w2x not required
7
in single tasks, and that the deterioration in performance in dual tasks is caused by the malfunction of this central executive w3,7x. Dual task interference due to requirement of activating the same part of cortex is a fundamentally different source of interference, but it does in no way exclude additional sources of interference.
5. Conclusion In conclusion, the present results show that if two tasks activate overlapping parts of the cortex, there will be significant interference with increased reaction time when the tasks are performed simultaneously. This is a physiological basis for dual task interference which has not been demonstrated previously.
Acknowledgements We would like to thank Professor Lars-Goran ¨ Nilsson ¨ for helpful suggestions on the manuscript, Fredrik Ohrstrom ¨ for computer programming and Erik Eliasson for help with testing the subjects. This research was supported by the EJBL Foundation.
References w1x P.L. Ackerman, W. Schneider, C.D. Wickens, Deciding the existence of a time-sharing ability: a combined methodological and theoretical approach, Hum. Factors 26 Ž1984. 71–82. w2x A. Baddeley, Working Memory, Oxford University Press, Oxford, 1986. w3x A. Baddeley, R. Logie, S. Bressi, S.D. Sala, H. Spinnler, Dementia and working memory, Quart. J. Exp. Psychol. 38A Ž1986. 603–618. w4x A.D. Baddeley, S. Bressi, S.D. Sala, R. Logie, H. Spinnler, The decline of working memory in Alzheimer’s disease, Brain 114 Ž1991. 2521–2542. w5x D.E. Broadbent, Task combination and selective intake of information, Acta Psychol. ŽAmst.. 50 Ž1982. 253–290. w6x J.B. Brookings, D.L. Damos, Individual differences in multiple-task performance, in: D.L. Damos ŽEd.., Multiple-Task Performance, Taylor and Francis, London, 1991, pp. 363–386. w7x M. D’Esposito, J.A. Detre, D.C. Alsop, R.K. Shin, S. Atlas, M. Grossman, The neural basis of the central executive system of working memory, Nature 378 Ž1995. 279–281. w8x G. Fogarty, Timesharing in relation to broad ability domains, Intelligence 11 Ž1987. 207–231. w9x A. Friedman, M.C. Polson, D.G. Dafoe, S. Gaskill, Dividing attention within and between hemispheres: testing a multiple resources approach to limited-capacity information processing, J. Exp. Psychol.: Hum. Percept. Perform. 8 Ž1982. 625–650. w10x D.M. Green, J.A. Swets, Signal Detection Theory and Psychophysics, Wiley and Sons, New York, 1966. w11x F.E. Grubbs, Sample criteria for testing outlying observations, Ann. Math. Stat. 21 Ž1950. 27–58. w12x S.L. Juliano, P.J. Hand, B.L. Whitsel, Patterns of increased metabolic activity in somatosenory cortex of monkey Macaca fascicularis, subjected to control cutaneous stimulation: a 2-deoxyglucose study, J. Neurophysiol. 46 Ž1981. 1260–1284.
8
T. Klingberg, P.E. Rolandr CognitiÕe Brain Research 6 (1997) 1–8
w13x S.L. Juliano, B.L. Whitsel, M. Tommerdahl, S.S. Cheema, Determinants of patchy metabolic labeling in the somatosensory cortex of cats: a possible role for intrinsic inhibitory circuitry, J. Neurosci. 9 Ž1989. 1–12. w14x D. Kahneman, Attention and Effort, Prentice Hall, Englewood Cliffs, NJ, 1973. w15x S. Kinomura, J. Larsson, B. Gulyas, ´ P.E. Roland, Activation by attention of the human reticular formation and thalamic intralaminar nuclei, Science 271 Ž1996. 512–515. w16x M. Kinsbourne, R.E. Hicks, Functional cerebral space: a model for overflow, transfer and interface effects in human performance: a tutorial review, in: J. Requin ŽEd.., Attention and Performance VII, Erlbaum, Hillsdale, NJ, 1978, pp. 345-362. w17x T. Klingberg, P.E. Roland, R. Kawashima, Activation of multi-modal cortical areas underlies short-term memory, Eur. J. Neurosci. 8 Ž1996. 1965–1971. w18x R.A. Koeppe, J.E. Holden, W.R. Ip, Performance comparison of parameter estimation techniques for the quantitation of local cerebral blood flow by dynamic positron computed tomography, J. Cereb. Blood Flow Metab. 5 Ž1985. 224–234. w19x G.D. Logan, Short-term memory demands of reaction-time tasks that differ in complexity, J. Exp. Psychol. 6 Ž1980. 375–389. w20x R.H. Logie, Visuo-spatial processing in working memory, Quart. J. Exp. Psychol. wAx 38 Ž1986. 229–247. w21x V.B. Mountcastle, An organizing principle for cerebral function. The unit module and the distributed system, in: G.M. Edelman, V.B. Mountcastle ŽEds.., The Mindful Brain, Plenum, New York, 1978, pp. 7-50. w22x P.G. Nestor, R. Parasuraman, J.V. Haxby, C.L. Grady, Divided
w23x
w24x w25x
w26x w27x
w28x w29x
w30x w31x
w32x
attention and metabolic brain dysfunction in mild dementia of the Alzheimer’s type, Neuropsychologia 29 Ž1991. 379–387. D.A. Norman, T. Shallice, Attention to Action: Willed and Automatic Control of Behavior, Center for Human Information Processing, Technical Report No. 99, 1980. H. Pashler, Dual-task interference in simple tasks: data and theory, Psychol. Bull. 116 Ž1994. 220–244. P.E. Roland, C. Graufelds, J. Wahlin, L. Ingelman, M. Andersson, ˚ ˚ A. Ledberg, J. Pedersen, S. Akerman, A. Dabringhaus, K. Zilles, Human brain atlas: for high-resolution functional and anatomical mapping, Hum. Brain Mapping 2 Ž1994. 1–12. P.E. Roland, Brain Activation, Wiley-Liss, New York, 1993. ˚ P.E. Roland, B. Levin, R. Kawashima, R. Akerman, Three-dimensional analysis of clustered voxels in 15 O-butanol brain activation images, Hum. Brain Mapping 1 Ž1993. 3–19. D.J. Simons, Temporal and spatial integration in the rat S1 vibrissa cortex, J. Neurophysiol. 54 Ž1985. 615–621. F. Stablum, G. Leonardi, M. Mazzoldi, C. Umilta, ` S. Morra, Attention and control deficits following closed head injury, Cortex 30 Ž1994. 603–618. S. Sternberg, High-speed scanning in human memory, Science 153 Ž1966. 652–654. E. Welker, M. Armstrong-James, H.V.d. Loos, R. Kraftsik, The mode of activation of a barrel column: response properties of single units in the somatosensory cortex of the mouse upon whisker deflection, Eur. J. Neurosci. 5 Ž1993. 691–712. C.D. Wickens, The structure of attentional resources, in: R. Nickerson ŽEd.., Attention and Performance VIII, Erlbaum, Hillsdale, NJ, 1980.
Research report
Interference between two concurrent tasks is associated with activation of overlapping fields in the cortex Torkel Klingberg ) , Per E. Roland DiÕision of Human Brain Research, Department of Neuroscience, Karolinska Institute, Doktorsringen 12, S-171 77 Stockholm, Sweden Accepted 18 February 1997
Abstract Interference between two concurrent tasks can be measured as an increased reaction time during simultaneous performance compared to when each task is performed alone. We tested the hypothesis that two tasks interfere because they require activation of overlapping areas of the cerebral cortex. With positron emission tomography we measured cortical activation as fields with significant increase in regional cerebral blood flow during single task performance of an auditory and a visual gorno-go task and an auditory and a visual short-term memory ŽSTM. task. In a separate experiment we measured the degree of interference between the two gorno-go tasks and between the two STM tasks during dual task performance. Both the two gorno-go tasks and the two STM tasks activated overlapping parts of the cortex and interfered significantly during dual task performance. The two STM tasks had a larger volume of overlap and also significantly larger increase in reaction time during dual task performance, compared to the gorno-go tasks. The results thus indicate that two concurrent tasks interfere, with a resulting increase in reaction time, if they require activation of overlapping parts of the cortex. q 1997 Elsevier Science B.V. Keywords: Short-term memory; Working memory; Reaction time; Regional cerebral blood flow ŽrCBF.; Positron emission tomography ŽPET.
1. Introduction When two tasks are performed concurrently, there is often an impairment in performance, which can be measured as an increase in error rate or reaction time w2,5x. Such dual task interference has been described as a competition for attentional resources w9,14x, or competition for information processing mechanisms w16,32x. However, the neural mechanisms underlying dual task interference are not known. There has been no systematic comparison of the pattern of cerebral activation, as seen with positron emission tomography ŽPET. or MRI, with dual task interference. With PET one can measure cortical activation as volumes of cortex which increase their metabolism when the brain is activated. What appears to be continuous areas of activations in the PET images, are probably columns w21x or patches w12x of active synapses. If two tasks significantly increase the regional cerebral blood flow ŽrCBF. in an identical volume of cortical tissue, they are likely to
)
Corresponding author. Fax: q46 Ž8. 309045; E-mail: [email protected] 0926-6410r97r$17.00 q 1997 Elsevier Science B.V. All rights reserved. PII S 0 9 2 6 - 6 4 1 0 Ž 9 7 . 0 0 0 1 0 - 4
interfere with each other, either because the two tasks depend on activation of the very same population of neurones, or because the two tasks activate neuronal populations within the volume, which are non-identical but inhibit each other by local mechanisms when activated simultaneously w13,28,31x. Both mechanisms would disturb the neuronal processing and decrease the probability of a correct answer. Alternatively, the cortical field or the neuronal population which is required for two tasks could first be dedicated to one task and then to the other, which would increase the mean reaction time. The hypothesis in this study was then that if the areas of cortex that are activated by task A overlaps with the areas of cortex activated by task B, then the two tasks will interfere. We measured significant increases in rCBF during a task relative to a passive control state as clusters of voxels having high Student’s t-values w25x. The degree to which two tasks activate identical areas of cortex can then be determined as the overlaps between these clusters of significant blood flow increase in the two tasks. We analysed the volume and location of the overlaps in PET data from two gorno-go tasks and two short-term memory ŽSTM. tasks w17x. The rCBF increase was measured only during single task performance, not during dual task performance.
2
T. Klingberg, P.E. Rolandr CognitiÕe Brain Research 6 (1997) 1–8
In a separate psychophysical experiment, with a different group of subjects, the interference between the tasks during dual task performance was evaluated and compared to the overlaps in the PET data.
2. Materials and methods 2.1. Psychophysical experiment 2.1.1. Subjects Fifteen paid, healthy volunteers, 12 males and 3 females, mean age 23.4 years, participated. 2.1.2. Tasks In the visual gorno-go task ŽFig. 1, task 1. the luminance of a circular field on a computer screen changed at random intervals Ž2.00–5.00 s.. The subjects were asked to respond when they detected a decrease in luminance, but not when there was an increase in luminance. In the auditory gorno-go task ŽFig. 1, task 2. a tone changed in frequency at random intervals Ž2.00–5.00 s.; the subjects were asked to respond when the tone changed from a higher frequency to a lower one, no response was required when the tone changed from a lower frequency to a higher one. In the simultaneous gorno-go task ŽFig. 1, task 3., the visual and the auditory gorno-go tasks were performed simultaneously. The dual task conditions always consisted of performing one visual and one auditory task to avoid perceptual interference between the tasks w20x. In the dual tasks, stimuli varied randomly and independently of each other, with the only restriction that a change in one sensory channel was always separated by at least 1.00 s from a change in the other sensory channel. This was to avoid what has been termed the ‘‘psychological refractory period’’, i.e., an increased reaction time when the two stimuli which call for a response are separated by less than 300 ms w24x. In the visual STM task ŽFig. 1, task 4., a circular field on a computer screen changed in luminance between one of six high luminance levels – target levels – and a baseline luminance level. The 1.00 s stimulation at target luminance levels were separated by 2.00–5.00 s delays with baseline luminance. The subjects were asked to compare each target luminance level with the previous one, and to respond if the last luminance level was lower than the previous one; no response was required when there was a change from a lower to a higher luminance. In the auditory STM task ŽFig. 1, task 5., a series of high pitched tones – target tones – of different frequencies were presented. Target tones were separated by temporal intervals of 2.00–5.00 s, during which a low pitched tone – baseline tone – was presented. The duration of the target tones was 1.00 s. The subjects were asked to compare each target tone with the previous target tone, and to respond only if the last target tone was lower in frequency than the
Fig. 1. Experimental design of the gorno-go and the STM tasks. The lines indicate how the stimuli changed in frequency Žauditory tasks. or luminance Žvisual tasks. over time. The arrows point at the two frequency levels Žor luminance levels. which were compared by the subjects. The subjects were asked to respond if the second level Žbelow the second arrowhead. was lower than the previous one Žbelow the first arrowhead..
previous one ŽFig. 1, task 5.. In the simultaneous STM condition ŽFig. 1, task 6., the auditory and the visual stimuli varied independently of each other, with the only restriction that a change in one sensory channel was always separated by at least 1.00 s from a change in the other sensory channel. The gorno-go and the STM tasks were designed to be similar with respect to sensory stimulation, selective attention, response selection and motor output. The difference was the delay between the stimuli in the STM tasks, a delay which required stimulus information to be kept ‘‘on-line’’ for a brief period. 2.1.3. Stimuli The visual stimulus consisted of a circular field of 15 cm in diameter, which was presented against a black background on a computer screen, 100 cm from the subjects in a completely dark room. The minimum and maximum luminance values were within 18–36 cdrm2 in the detection tasks, and 4–18 cdrm2 in the STM tasks. In the STM tasks there was also an inter-stimulus ‘‘baseline’’ of luminance that was fixed at 3 cdrm2 .
T. Klingberg, P.E. Rolandr CognitiÕe Brain Research 6 (1997) 1–8
The auditory stimuli were tones presented through a pair of earphones. The minimum and maximum frequencies were within 650–780 Hz. In the STM tasks there was an inter-stimulus ‘‘baseline’’ frequency that was fixed at 550 Hz. During a pre-test session the stimulus levels were adjusted so that the subjects would achieve an aROC value of approximately 0.85 in the non-simultaneous tasks. Reaction times were recorded by the same computer that controlled the stimuli and were defined as the time from the onset of the stimulus to the pressing of the button. Correct responses should occur within 2.00 s in order to be counted as a ‘‘hit’’. In both the gorno-go tasks and the STM tasks the subjects pressed the left button on a computer mouse with their right index finger when they responded to visual stimuli and pressed the right button with their right third finger when they responded to auditory stimuli. Failure to respond within 2.00 s after stimulus onset was taken as a ‘‘miss’’. Responses on non-target trials Žwhen the stimulus was higher in luminance or frequency than the previous one. were registered as false alarms. 2.1.4. Procedure The tasks were performed in blocks of six Žfor example task numbers 4, 6, 5, 2, 1, 3. Žsee Fig. 1.. Each task was thus performed once before any task was repeated. The order of the tasks within a block was balanced over the whole test session. A total session consisted of four blocks. Each task was performed for 90 s. After a 10 s break a text appeared on the screen, describing the next task, e.g. ‘‘simultaneous performance of gorno-go’’. 2.1.5. Interference measurement During dual task performance, there is a possibility that subjects give priority to one task over the other, which could increase the reaction time and error rate more in the task that is not given priority. We were not interested how such priorities were given. The increases in reaction time were therefore computed by subtracting the sum of the reaction times for the two tasks in the single condition ŽRT,aud,single and RT,vis,single. from the sum of the reaction times during simultaneous performance:
D RT s Ž RT ,aud, sim q RT ,Õis, sim . y Ž RT ,aud, single q RT ,Õis, single . . W here aud s auditory, vis s visual, sim s simultaneous. The gorno-go D RT values were then compared to the STM D RT values. We calculated the area under receiver-operating characteristic curve ŽaROC. values to assess the performance w10x. When we talk of increase in error rate, this refers to a decrease in aROC values. Data were sampled for each person giving only one point on the ROC. In this case the aROCs Ž P Žhit. y P Žfalse alarm..r2 q 0.5; where P s probability. This calculation gives an aROC of 0.5 at
3
chance level performance and a value of 1.0 at perfect performance. The change in error rate was calculated as the sum of aROC during the simultaneous performance minus the sum of the aROC during the two single presentations:
D aROCs Ž aROC,aud, sim q aROC,Õis, sim . y Ž aROC,aud, single q aROC,Õis, single . For each subject, D RT and DaROC was calculated for the gorno-go tasks and for the STM tasks. 2.1.6. Test for normality All data of D RT and DaROC were tested to see if the distribution differed significantly from a normal distribution. The DRT values were found not to be normally distributed ŽShapiro–Wilk W-test, P - 0.01 for both the gorno-go and the STM group.. By visual inspection we detected three values of DRT being outliers which were found to be 5.0 S.D. or more from the mean of the rest of the values within each group and could thus not be considered to belong to the same distribution as the rest of the points within each group w11x. A possible explanation for this could be that the subjects had not followed the rule of answering as quickly as possible on all tasks. The three outlier points of D RT were excluded from further analysis. After exclusion the distribution of D RT was normal for both the STM and the gorno-go group ŽShapiro–Wilk W-test, P ) 0.10.. The median value for D RT was not changed by the exclusion and was 0.06 s for the gorno-go group and 0.18 s for the STM group, both before and after exclusion of the three outliers. 2.2. PET experiment Ten subjects had their rCBF measured during a visual gorno-go task Žtask 1 in Fig. 1., an auditory gorno-go task Žtask 2 in Fig. 1. and during a control state. In the control state the subjects only received passive auditory and visual stimulation without any task to perform. Another group of 10 subjects had their rCBF measured during an visual STM task Žtask 4 in Fig. 1., during an auditory STM task Žtask 5 in Fig. 1. and during a control state which was identical to the control state in the gorno-go experiment. Control rCBF was subtracted from task rCBF and significant activations were determined as clusters of voxels with rCBF increase above a certain threshold, these images are called cluster images. The cluster images ŽSTM visualycontrol . and ŽSTM auditoryycontrol . were then compared with each other, voxel by voxel. Corresponding voxels that were significant in both subtraction images are called overlapping voxels. This is illustrated in Fig. 2. Clusters from the auditory subtraction image are shown in green. Clusters from the visual subtraction image are shown in blue, and the overlapping voxels are shown in red.
4
T. Klingberg, P.E. Rolandr CognitiÕe Brain Research 6 (1997) 1–8
Exactly the same procedure was carried out to create an overlap image from the two subtraction images Žauditory gorno-go – control. and Žvisual gorno-go – control.. 2.2.1. Scanning procedure and statistics Subjects received an intravenous bolus injection of 70 mCi of 15 O-butanol before each PET scan. Local tissue radioactivity was measured with an 8-ring Ž15-slice. PC2048-15B Scanditronix PET scanner. Images were reconstructed with a 4 mm Hanning filter, giving a final image resolution of approximately 4.2 mm. Values of rCBF were calculated using a dynamic algorithm w18x. Global blood flow was normalized so that white matter blood flow was 20 mlr100 grmin in all images. The rCBF images were transformed to a standard format using
the MRI of each subject and a computerised brain atlas w25x. Statistical analysis was performed as described by Roland et al. w27x. Task minus control subtraction images were made for each subject through voxel by voxel subtraction of the rCBF of the control from each of the task images. A Student’s t-value was calculated for each voxel. Voxels above a certain t-threshold and occurring in clusters above a certain size were considered significant, where the t-threshold and cluster size were set so that there was a probability of 0.6 for having one or more false positive clusters in the whole brain w27x. The non-significant voxels were set to zero. This image is called a cluster image. This article only concerns clusters that were found in the same place in two different subtraction images. The probability
Fig. 2. Illustration of the overlap in cortical activation between the two STM tasks. Clusters from the auditory minus control subtraction image are shown in green. Clusters from the visual minus control subtraction image are shown in blue, and the overlapping voxels are shown in red. The clusters are superimposed on a horizontal section of the MR image used for anatomical standardisation.
T. Klingberg, P.E. Rolandr CognitiÕe Brain Research 6 (1997) 1–8
of two clusters, e.g. 0.5 cm3 large, occurring at the same place within the brain, and both clusters being false positives, is of course smaller than 0.6. An apparent overlap could theoretically arise if two adjacent but non-overlapping areas of rCBF increase are measured with a method with low spatial resolution. However, all the reported areas of overlap extended at least 5 mm from border to border along an axis connecting the centres of gravity of the two clusters. This is more than the spatial resolution of the PET camera and ensures that the overlaps are real and not a mere artefact due to low spatial resolution. It should also be kept in mind that a significant cluster, based on measured rCBF changes from 10 subjects, probably is smaller than the neuronal volume activated in each of the 10 individuals, since the cortical activation is not located in exactly the same place in each of the individuals, and the volume of increase that reaches significance level across subjects is located over the area where the majority of subjects has increased metabolism. The volume of each of the overlaps is for the same reason probably underestimated. The volumes of overlap reported here can therefore not be taken as absolute and quantitative measurements, but only permit relative comparisons.
3. Results
5
Table 2 Location and volume of the cortical overlaps between the auditory and the visual STM task Area
Coordinates
Sup. frontal g ŽR. Inf. frontal g ŽL. Frontal operculum ŽL. Cingulate cortex ŽR. Inf. parietal cortex ŽR. Ventromedial frontal ŽR. Orbitofrontal cortex ŽR.
y20, 47, 41 33, 55, 0 41, 10, 16 y2, 16, 45 y40, y48, 49 y7, 24, y7 y11, 42, y15
mm3 90 230 300 260 240 190 280
S 1590 Sup.ssuperior; inf.s inferior; g s gyrus; L s left hemisphere; R s right hemisphere.
percentage overlap varied between 19 and 48% in the gorno-go tasks and between 18 and 74% in the STM tasks. The mean percentage overlap was 33% for the gorno-go tasks and 50% for the STM tasks. The percentage overlap was also calculated as O divided by the mean volume of A and B, which resulted in a mean percent overlap for the gorno-go tasks of 27% and for the STM tasks of 40%. The overlaps were thus larger between the two STM tasks, also relative to the volume of the cortical activations that formed the overlaps. 3.2. Task performance during PET scanning
3.1. OÕerlap in cortical actiÕation Location and volume of the cortical overlaps are shown in Tables 1 and 2. The gorno-go tasks overlapped in six locations with a total volume of 820 mm3. The STM tasks overlapped in seven locations with a total volume of 1590 mm3 ŽFig. 3.. To assess if the volume of overlap was related to the size of the clusters forming the overlap, we also analysed the percentage of overlap for each cluster. If a cluster A overlaps with a cluster B Žwhere A is smaller than B. and thus creates an overlap O, then the percentage overlap for O is the volume of O divided by the volume of A. The
Performance was evaluated as aROC values and was 0.93 Ž"0.03. and 0.84 Ž"0.03. in the auditory and visual gorno-go tasks respectively, and 0.96 Ž"0.01. and 0.95
Table 1 Location and volume of the cortical overlaps between the auditory and the visual gorno-go task Area
Coordinates
Sup. frontal g. ŽR. Inf. frontal g. ŽR. Sup. frontal g mesial part ŽR. Frontal operculum ŽR. Sup. temporal g. ŽR. Lingual g. ŽL.
y7, 13, 55 y32, 48, y1 y4, 63, 12 y43, 20, 1 y53, y24,y7 28, y82, y23
mm3 120 160 180 80 180 100
S 820 Sup.ssuperior; inf.s inferior; g s gyrus; L s left hemisphere; R s right hemisphere.
Fig. 3. The total volume of overlap between the two gorno-go tasks Žleft bar. and between the STM task Žright bar.. The superimposed line-graph indicates the increase in reaction time Ž D RT. during dual compared to single performance of the gorno-go and STM tasks. The D RT was significant for both tasks Ž ) P - 0.05; ) ) P - 0.01. and was significantly larger for the STM tasks compared to the gorno-go tasks Ž P 0.05..
T. Klingberg, P.E. Rolandr CognitiÕe Brain Research 6 (1997) 1–8
6
Table 3 Error rates and reaction times in the gorno-go tasks
RT Žs. aROC
4. Discussion
Single performance
Simultaneous performance
Visual
Auditory
Visual
Auditory
0.80 Ž"0.07. 0.82 Ž"0.03.
0.94 Ž"0.06. 0.85 Ž"0.03.
0.79 Ž"0.04. 0.79 Ž"0.03.
1.00 Ž"0.06. 0.83 Ž"0.03.
RT s reaction time; aROC sarea under receiver-operating characteristic curve; Ž"S.E.M...
Ž"0.03. for the auditory and visual STM tasks respectively. 3.3. Psychophysical experiment Each of the 15 subjects performed all three gorno-go tasks Žtasks 1, 2 and 3 in Fig. 1. and all STM tasks Žtasks 4, 5 and 6 in Fig. 1.. After controlling that the values were normally distributed, means and standard error of means ŽS.E.M.. were calculated for D RT and DaROC. The increase in reaction time Ž D RT computed as described in Section 2.1.5. during gorno-go was 0.05 Ž"0.02. s, and during STM 0.19 Ž"0.05. s ŽFig. 3.. This increase was significant for both tasks Ž P - 0.05 for gorno-go tasks, and P - 0.01 for the STM tasks, one-tailed Student’s t-test.. Furthermore, the increase was significantly larger in the STM tasks than in the gorno-go tasks Ž P - 0.05.. Accuracy and reaction times in the STM and gorno-go tasks are shown in Tables 3 and 4. The decrease in aROC Ž DaROC computed as described in Section 2.1.5. during the simultaneous performance was significant for both tasks: y0.055 Ž"0.027. and y0.052 Ž"0.026., for the gorno-go and STM tasks respectively Ž P - 0.05, one-tailed Student’s t-test.. To assess the effect of task difficulty in the single task on the increase in error rate during dual performance, we computed the correlation between the mean aROC value for the two single tasks ŽŽ aROC,aud,single q aROC,vis,single.r2. and the DaROC for each individual. The correlation between these two variables was not significant Ž r s y0.25, P ) 0.18.. Task difficulty in the single tasks did thus not significantly effect the relative increase in error rate Ž DaROC. during simultaneous performance. Neither was the mean aROC value for the two single tasks correlated with the relative increase in reaction time, D RT Ž r s 0.01, P ) 0.94.. Table 4 Error rates and reaction times in the STM tasks
RT Žs. aROC
Single performance
Simultaneous performance
Visual
Auditory
Visual
Auditory
0.90 Ž"0.05. 0.85 Ž"0.02.
0.83 Ž"0.04. 0.86 Ž"0.02.
0.96 Ž"0.05. 0.82 Ž"0.02.
0.96 Ž"0.05. 0.84 Ž"0.02.
RT s reaction time; aROC sarea under receiver-operating characteristic curve; Ž"S.E.M...
The cortical fields activated by the auditory gorno-go task overlapped with the fields activated by the visual gorno-go task in six locations. The total overlap was 820 mm3. The two STM tasks had overlapping activations in seven locations with a total volume of 1590 mm3. At an anatomical and macroscopic level, the cortical activation in the areas of overlap was obviously not specific to the sensory modality of the stimuli and was located outside the primary sensory areas. No cortical overlaps were located in primary motor cortex, premotor cortex or the supplementary motor area ŽTables 1 and 2.. When the two gorno-go tasks were performed simultaneously there was a significant interference between the tasks, detected as an increase in error rate as well as in reaction time, compared to when the gorno-go tasks were performed one at a time. Simultaneous performance of the two STM tasks was similarly associated with significant increases in both reaction time and error rate. In these two independent PET studies, overlapping cortical activity between single tasks was associated with interference during simultaneous performance, in accordance with the hypothesis. Furthermore, the larger volume of overlap in the STM tasks was accompanied by a significantly larger increase in reaction time, compared to the gorno-go tasks ŽFig. 3.. This is in accordance with the hypothesis, but the data are not sufficient to specify the relationship between overlap and increase in reaction time, for example we cannot say that there is a linear relation between the two variables. It should be pointed out that the size of the overlaps is not absolutely quantitative, but depends on the sensitivity of the PET camera, the data treatment, the statistical method for evaluating the data and the level of statistical significance chosen. The comparison of two pairs of tasks is therefore always relative and requires that the methods are identical for both pairs of tasks, which was the case in the present study. Kinomura et al. w15x showed with PET that when the brain goes from a relaxed state into a state of increased arousal, as happens when one performs a reaction time task, this increases the blood flow in the midbrain reticular formation and the thalamus. This finding suggests that there will always be an overlap between two tasks which involve an increase in arousal compared to a passive sensory stimulation during control. If such subcortical overlaps result in interference is not known. We did not perform any PET scanning during dual task performance. Even if such scanning had been performed, the limited time resolution of the PET method makes it impossible to decide whether the two tasks actually activate the very same neuronal population at the same time, or if there is a time-sharing activity, as discussed in Section 1. The STM tasks had about twice as large volumes of overlap as had the two gorno-go tasks, but the increase in error rate was equally large in both pairs of tasks. This
T. Klingberg, P.E. Rolandr CognitiÕe Brain Research 6 (1997) 1–8
difference in total volume of cortical overlap, but lack of difference in error rate, could be taken as evidence that error rate is not related to cortical overlaps at all. It is possible that the physiological basis for errors is not the same as the one for reaction time increases. It is also possible that one needs to take into account the location of the cortical overlaps to explain the error rates, while only reaction time increases are related to the total volume of overlap, independent of the location of the overlaps. None of the overlaps between the two gorno-go tasks was located at exactly the same site as the overlaps between the two STM tasks ŽTables 1 and 2., which demonstrates that the interference was not a result of any particular and specifically located overlap. It is not possible from the present data to determine whether the interference was caused by particular overlaps or due to the fact that overlaps were present at many sites. The increased reaction time during the simultaneous performance of the STM tasks is in agreement with findings in previous studies of dual task performance, where reaction time increases with increasing short-term memory load w2,19,30x. Task difficulty is sometimes suspected to be a confounding factor when evaluating the amount of interference between two tasks. However, there was no significant correlation between task difficulty Žmean aROC. in the single tasks and the increase in reaction time during simultaneous performance. Roland w26x, p. 424, suggested that the synapses in cortex always are activated in large ensembles called functional cortical fields, and that two tasks which activate an identical functional field cannot be performed simultaneously. The hypothesis in this article is similar but more generally formulated, since all overlap in cortical activity between two tasks could cause interference, as discussed in Section 1. Our data show that interference is associated with overlaps of less than 50% of the overlapping activated fields, at least when multiple such overlaps exist, as is shown by the interference between the two gorno-go tasks ŽTable 1.. The hypothesis of this article also bears resemblance to ideas formulated by Kinsbourne and Hicks w16x and Wickens w32x. Nestor et al. w22x reported that in Alzheimer’s patients, decreased metabolism in the right parietal lobe is correlated with the relative increases in reaction time during dual task performance. Other studies show that patients with Alzheimer’s disease are impaired with respect to both reaction time and errors w3,4x. Increases in reaction time during dual task performance, compared to single task performance, have also been shown in patients with close head injuries w29x. In the present article we only dealt with PET data from single task performance and tried to make predictions about dual task performance. It has been suggested that performance of dual tasks requires some additional factor, such as a time-sharing ability w1,6,8x, ‘‘supervisory function’’ w23x or ‘‘central executive function’’ w2x not required
7
in single tasks, and that the deterioration in performance in dual tasks is caused by the malfunction of this central executive w3,7x. Dual task interference due to requirement of activating the same part of cortex is a fundamentally different source of interference, but it does in no way exclude additional sources of interference.
5. Conclusion In conclusion, the present results show that if two tasks activate overlapping parts of the cortex, there will be significant interference with increased reaction time when the tasks are performed simultaneously. This is a physiological basis for dual task interference which has not been demonstrated previously.
Acknowledgements We would like to thank Professor Lars-Goran ¨ Nilsson ¨ for helpful suggestions on the manuscript, Fredrik Ohrstrom ¨ for computer programming and Erik Eliasson for help with testing the subjects. This research was supported by the EJBL Foundation.
References w1x P.L. Ackerman, W. Schneider, C.D. Wickens, Deciding the existence of a time-sharing ability: a combined methodological and theoretical approach, Hum. Factors 26 Ž1984. 71–82. w2x A. Baddeley, Working Memory, Oxford University Press, Oxford, 1986. w3x A. Baddeley, R. Logie, S. Bressi, S.D. Sala, H. Spinnler, Dementia and working memory, Quart. J. Exp. Psychol. 38A Ž1986. 603–618. w4x A.D. Baddeley, S. Bressi, S.D. Sala, R. Logie, H. Spinnler, The decline of working memory in Alzheimer’s disease, Brain 114 Ž1991. 2521–2542. w5x D.E. Broadbent, Task combination and selective intake of information, Acta Psychol. ŽAmst.. 50 Ž1982. 253–290. w6x J.B. Brookings, D.L. Damos, Individual differences in multiple-task performance, in: D.L. Damos ŽEd.., Multiple-Task Performance, Taylor and Francis, London, 1991, pp. 363–386. w7x M. D’Esposito, J.A. Detre, D.C. Alsop, R.K. Shin, S. Atlas, M. Grossman, The neural basis of the central executive system of working memory, Nature 378 Ž1995. 279–281. w8x G. Fogarty, Timesharing in relation to broad ability domains, Intelligence 11 Ž1987. 207–231. w9x A. Friedman, M.C. Polson, D.G. Dafoe, S. Gaskill, Dividing attention within and between hemispheres: testing a multiple resources approach to limited-capacity information processing, J. Exp. Psychol.: Hum. Percept. Perform. 8 Ž1982. 625–650. w10x D.M. Green, J.A. Swets, Signal Detection Theory and Psychophysics, Wiley and Sons, New York, 1966. w11x F.E. Grubbs, Sample criteria for testing outlying observations, Ann. Math. Stat. 21 Ž1950. 27–58. w12x S.L. Juliano, P.J. Hand, B.L. Whitsel, Patterns of increased metabolic activity in somatosenory cortex of monkey Macaca fascicularis, subjected to control cutaneous stimulation: a 2-deoxyglucose study, J. Neurophysiol. 46 Ž1981. 1260–1284.
8
T. Klingberg, P.E. Rolandr CognitiÕe Brain Research 6 (1997) 1–8
w13x S.L. Juliano, B.L. Whitsel, M. Tommerdahl, S.S. Cheema, Determinants of patchy metabolic labeling in the somatosensory cortex of cats: a possible role for intrinsic inhibitory circuitry, J. Neurosci. 9 Ž1989. 1–12. w14x D. Kahneman, Attention and Effort, Prentice Hall, Englewood Cliffs, NJ, 1973. w15x S. Kinomura, J. Larsson, B. Gulyas, ´ P.E. Roland, Activation by attention of the human reticular formation and thalamic intralaminar nuclei, Science 271 Ž1996. 512–515. w16x M. Kinsbourne, R.E. Hicks, Functional cerebral space: a model for overflow, transfer and interface effects in human performance: a tutorial review, in: J. Requin ŽEd.., Attention and Performance VII, Erlbaum, Hillsdale, NJ, 1978, pp. 345-362. w17x T. Klingberg, P.E. Roland, R. Kawashima, Activation of multi-modal cortical areas underlies short-term memory, Eur. J. Neurosci. 8 Ž1996. 1965–1971. w18x R.A. Koeppe, J.E. Holden, W.R. Ip, Performance comparison of parameter estimation techniques for the quantitation of local cerebral blood flow by dynamic positron computed tomography, J. Cereb. Blood Flow Metab. 5 Ž1985. 224–234. w19x G.D. Logan, Short-term memory demands of reaction-time tasks that differ in complexity, J. Exp. Psychol. 6 Ž1980. 375–389. w20x R.H. Logie, Visuo-spatial processing in working memory, Quart. J. Exp. Psychol. wAx 38 Ž1986. 229–247. w21x V.B. Mountcastle, An organizing principle for cerebral function. The unit module and the distributed system, in: G.M. Edelman, V.B. Mountcastle ŽEds.., The Mindful Brain, Plenum, New York, 1978, pp. 7-50. w22x P.G. Nestor, R. Parasuraman, J.V. Haxby, C.L. Grady, Divided
w23x
w24x w25x
w26x w27x
w28x w29x
w30x w31x
w32x
attention and metabolic brain dysfunction in mild dementia of the Alzheimer’s type, Neuropsychologia 29 Ž1991. 379–387. D.A. Norman, T. Shallice, Attention to Action: Willed and Automatic Control of Behavior, Center for Human Information Processing, Technical Report No. 99, 1980. H. Pashler, Dual-task interference in simple tasks: data and theory, Psychol. Bull. 116 Ž1994. 220–244. P.E. Roland, C. Graufelds, J. Wahlin, L. Ingelman, M. Andersson, ˚ ˚ A. Ledberg, J. Pedersen, S. Akerman, A. Dabringhaus, K. Zilles, Human brain atlas: for high-resolution functional and anatomical mapping, Hum. Brain Mapping 2 Ž1994. 1–12. P.E. Roland, Brain Activation, Wiley-Liss, New York, 1993. ˚ P.E. Roland, B. Levin, R. Kawashima, R. Akerman, Three-dimensional analysis of clustered voxels in 15 O-butanol brain activation images, Hum. Brain Mapping 1 Ž1993. 3–19. D.J. Simons, Temporal and spatial integration in the rat S1 vibrissa cortex, J. Neurophysiol. 54 Ž1985. 615–621. F. Stablum, G. Leonardi, M. Mazzoldi, C. Umilta, ` S. Morra, Attention and control deficits following closed head injury, Cortex 30 Ž1994. 603–618. S. Sternberg, High-speed scanning in human memory, Science 153 Ž1966. 652–654. E. Welker, M. Armstrong-James, H.V.d. Loos, R. Kraftsik, The mode of activation of a barrel column: response properties of single units in the somatosensory cortex of the mouse upon whisker deflection, Eur. J. Neurosci. 5 Ž1993. 691–712. C.D. Wickens, The structure of attentional resources, in: R. Nickerson ŽEd.., Attention and Performance VIII, Erlbaum, Hillsdale, NJ, 1980.