Neuroscience Letters 269 (1999) 137±140

Differences in the ability to process a visuo-spatial task are re¯ected in event-related slow cortical potentials of human subjects

Claus Lamm*, Herbert Bauer, Oliver Vitouch, Reinhard GstaÈ ttner

Brain Research Laboratory, Department of Psychology, University of Vienna, Liebiggasse 5, A-1010 Vienna, Austria Received 24 March 1999; received in revised form 10 May 1999; accepted 10 May 1999

Abstract Recent positron emission tomography (PET) and electroencephalographic (EEG) studies suggest that higher ability in a cognitive task is associated with a more ef®cient neuronal processing of this task. However, the validity and general- izability of these studies is limited for several reasons. We investigated 20 male and 18 female human subjects with good vs. poor spatial ability performing a visuo-spatial task (cube test). Processing-related slow event-related potentials were recorded via 22 electrodes, evenly distributed over the scalp. Signi®cant differences between good and poor performers were found in both sexes: poor subjects showed higher activity in the parietal region, and their topography was more extended into fronto-central regions. Since the amount and topography of brain activity may vary considerably depend- ing on subjects' ability, we conclude that careful (experimental) control of task-speci®c ability of subjects is mandatory for cognitive neuroscience studies. q 1999 Elsevier Science Ireland Ltd. All rights reserved.

Keywords: Neuronal ef®ciency; Ability; Spatial cognition; Individual differences; Event-related potentials; Slow potentials; Human cognition

Having already been introduced in 1969 [4], the concept between good and poor performers was accomplished only of neuronal ef®ciency assumes that higher ability in a cogni- by means of a post-hoc median split. Also, group assign- tive task is associated with more ef®cient neuronal proces- ment was based on ability measures derived from tasks sing of this task. Several functional neuroimaging studies which were only correlated with, but not identical to those have led to revitalized speculations about this assumption. used during electrophysiological recordings. Furthermore, Positron emission tomography (PET) studies report positive of all studies cited so far, only one [11] investigated male correlations between ability and the decrease in glucose and female subjects, but no sex-speci®c analyses were metabolic rate (GMR) following training of a visuo- reported. Another PET study [6] which also investigated spatial/motor task [7], and negative correlations with both sexes shows no correlation between ability and GMR GMR during the processing of abstract reasoning [8] and in females during mathematical reasoning, but an inconsis- verbal ¯uency items [11]. Topographical recordings of slow tent positive one in males. However, the per-group sample potential shifts (SPSs) in good and poor spatial test perfor- size in this study was again rather low, the `low' ability mers, revealed less slow potential negativity in the good group was de®ned via an average mathematical ability, group over the left parieto-occipito-temporal region [14]. and groups also differed in their verbal ability. In subjects with higher reasoning scores (Raven test), less Using the method of slow potential topography [2], we event-related desynchronization, following presentation of a aimed to investigate whether the assumption of more ef®- sentence veri®cation task, was reported [10]. However, the cient processing of a visuo-spatial task in high ability generalizability of these studies is restricted for several subjects holds for both sexes. Spatial ability was chosen reasons. In the PET studies, sample sizes were rather low, since sex differences in this intellectual dimension are and no control group was investigated in the training study. well documented on in literature [17]. In two consecutive In the electroencephalographic (EEG) studies, the division studies, males [16] and females with different levels of abil- ity were investigated. In both experiments, a more sensitive separation of ability groups was achieved by pre-experi- * Corresponding author. Tel.: 143-1-427-747-829; fax: 143-1- mental selection of subjects whose test performance fell 427-747-859. either in the ®rst (poor performers) or in the fourth quartile E-mail address: [email protected] (C. Lamm)

0304-3940/99/$ - see front matter q 1999 Elsevier Science Ireland Ltd. All rights reserved. PII: S0304-3940(99)00441-3 138 C. Lamm et al. / Neuroscience Letters 269 (1999) 137±140 item presentation time was unrestricted. Starting with the button push until the response of the item, EEG was recorded via 22 Ag/AgCl electrodes evenly distributed over the scalp. Electrodes were mounted on small adapters, and the skin was scratched at each location using a sterile needle to minimize skin potential artifacts and to keep elec- trode impedance #1V. Adapters had been secured on the scalp via collodion with the help of an application cap [2], Fig. 1. Black/white sample of a dichotomized 3DC item. Correct yielding a 2 1 4 £ 5 equidistant matrix montage. While the answer: `yes' (i.e. the two cubes could be identical). positions on the median sagittal line corresponded exactly to those of the international 10±20 system, the other ones only (good performers) of the respective test calibration sample. approximated them due to the equidistance constraint In order to keep single trials short, the visuo-spatial task that (recording locations: Fp10,Fp20/F70,F30, Fz, F40,F80 / subjects had to process during EEG recordings was dichot- T30,C30,Cz,C40,T40 /T50,P30, Pz, P40,T60 /O70,O30, omized (Fig. 1), but in all other aspects identical to the one Oz, O40,O80; all referenced to linked mastoids). Electrodes version used for pre-selection (3 Dimensional Cube test, and adapters were ®lled with degassed electrolyte and, via a 3DC [5]). Since the 3DC ful®lls the criteria of the Rasch preampli®er, connected to a high input impedance ($100 model [13], it has the property of unidimensionality and GV) DC ampli®er with excellent baseline stability. EEG therefore measures the same latent cognitive dimension was sampled at 4 kHz (downsampling at 250 Hz for digital (ability) in different groups of subjects. This property was storage), and the frequency range was set from DC to 100 essential for our study since we wanted to ensure that beha- Hz (notch at 50 Hz). In order to control eye movement vioral differences between groups and sexes were based on artifacts, vertical (electrodes above and below the right `true' differences in spatial processing, and not on differ- eye) and horizontal electrooculogram (electrodes on the ences in processing strategies or other factors not attributa- outer canthi) were recorded bipolarly. After EOG correction ble to spatial cognition. using a linear regression algorithm [16], all trials (n ˆ 76 Altogether, 38 healthy right-handed [1] students aged 20± per subject on the average) were visually inspected, and 30 years participated in the two experiments (Male: good- those assessed to contain artifacts were excluded from poor ˆ 12/8; female: good/poor ˆ 10/8). Extreme groups further analyses. For all trials solved correctly, stimulus and sexes neither differed in age nor in their general verbal onset-related averages referenced to a 200 ms pre-stimulus ability and basic intelligence (tested using a brief word- baseline were computed for each subject, and topographical power test [15]; see Table 1). During the experiment, SPS amplitudes and their current source density transforms subjects were seated in a dimmed and sound-attenuated were mapped by means of a spherical-spline interpolation room. They had to process visuo-spatial items for a period algorithm [12]. For statistical analyses, seven measures of 30 min (in a separate experimental condition, they solved representing SPSs at different latencies were calculated. verbal reasoning tasks; only the results of the visuo-spatial These measures consisted of the mean DC amplitude in a condition are reported in this paper). Items appeared 2 s 200-ms interval centered around 1.9, 2.4, 2.9, 3.4, 3.9, 4.4 after pressing of a start button. Subjects had to decide and 4.9 s after stimulus onset. Since the same statistical whether two blue-painted cubes with white patterns on results (in terms of signi®cance level) were observed for their faces presented in the middle of a 1700 color computer all SPSs measures, only the results of the analyses at 4.9 s screen placed about 70 cm in front of them could be iden- will be reported on here. Greenhouse±Geisser corrected tical or not (answers `yes' or `no'). This comparison repeated-measures ANOVAs with factors sex (two levels), required the mental rotation and/or transformation of one group (two) and location (22) were computed for raw ampli- of the cubes. Since a speeded (computer-paced) item tude and z-normalized [9] values. Global ANOVA results presentation mode is known to affect unidimensionality, were evaluated in detail by six Bonferroni-corrected linear contrasts (adjusted a-level a ˆ 0.009) using speci®c error Table 1 variances [3]. Behavioral dataa Behavioral data are reported in Table 1. As for the SPS data, the ANOVA with raw amplitudes revealed signi®cant Age 3DC WPT % main effects: sex (F 1; 34†ˆ11:26, P ˆ 0:002), group Female good 22.5 ^ 2.0 16.3 ^ 1.0 33.3 ^ 3.1 90.3 (F(1,34) ˆ 10.39, P ˆ 0:003) and location F 21; 714†ˆ Female poor 23.6 ^ 1.8 4.1 ^ 2.0 35.7 ^ 1.8 68.2 91.16, e ˆ 0.24, P , 0:001), signi®cant interactions sex £ Male good 23.9 ^ 2.1 16.1 ^ 0.8 36.2 ^ 1.6 93.0 location (F 21; 714†ˆ2:91, e ˆ 0.24, P ˆ 0:015) and Male poor 25.8 ^ 2.8 6.1 ^ 2.2 35.8 ^ 1.8 77.0 group £ location (F 21; 714†ˆ4:57, e ˆ 0.24, P , a Mean ^ STD. 3DC: 3DC score (k ˆ 17 items), WPT, word- 0.001), and a three-way interaction that was not signi®cant power test score (42 items); %, percentage of correctly solved (F 21; 714†ˆ0:73, e ˆ 0.24, P ˆ 0:606). Linear contrasts spatial items during EEG recordings. (sexes pooled) revealed signi®cantly higher negative ampli- C. Lamm et al. / Neuroscience Letters 269 (1999) 137±140 139

Fig. 2. Grand mean waveforms (left) and topographical mappings (right) of SPSs during processing of 3DC items. From top to bottom: female-good/poor, male-good/poor. Waveforms: eight out of 24 recorded channels are displayed, stimulus onset at 0 s; mappings: only the value centered around 4.9 s post stimulus, and only negative shifts are displayed; top: potential mapping; below: current source density transforms. Note that the mV-range displayed is different between groups since the scale has been adjusted to maximal amplitudes (20 vs. 30 mV). tudes in the poor group for the fronto-medial (F30, Fz, F40; Using normalized values, the main effect location F 1; 34†ˆ14:78, P , 0:001, mean amplitude difference (F 21; 714†ˆ121:34, e ˆ 0.26, P , 0:001) and the inter- D ˆ 27:68 mV), the frontal (Fp10 to F80; F 1; 34†ˆ action group £ location were signi®cant (F 21; 714†ˆ5:56, 12:84, P ˆ 0:001, D ˆ 210:42mV), the central (C30, Cz, e ˆ 0.26, P , 0:001), while sex £ location (F 21; 714†ˆ C40; F 1; 34†ˆ15:16, P , 0:001, D ˆ 28:39 mV) and 2.08, e ˆ 0.26, P ˆ 0:062) and group £ sex £ location the parietal region (P30, Pz, P40; F 1; 34†ˆ11:25, (F 21; 714†ˆ0:97, e ˆ 0.26, P ˆ 0:445) did not reach P ˆ 0:002, D ˆ 27:15 mV). No signi®cance was observed signi®cance (a ˆ 0.05). for temporal (T30,T40,T50,T60) and occipital (O70 to O80) The corresponding SPSs waveforms, along with single- sites. Similar results were obtained if contrasts were calcu- latency topographical mappings of the value centered lated separately for males and females (signi®cant differ- around 4.9 s, are displayed in Fig. 2. In both groups and ences in parietal, frontal and central regions and no sexes, activity patterns are dominated parieto-occipitally, signi®cant temporal or occipital contrasts in either sexes). i.e. by a region which is typically attributed to the proces- 140 C. Lamm et al. / Neuroscience Letters 269 (1999) 137±140 sing of visuo-spatial information. However, the poor group cortical resources may not or only partly become engaged. shows substantially more slow potential negativity in this Thus, an important caveat for functional studies of cognitive area, and their topography is also more extended into the processing has to be raised, namely to (experimentally) fronto-central region. control the task-speci®c ability of subjects. The signi®cant interaction group £ location indicates that the ability differences in the amount of cortical activity are not restricted to either of the sexes. This interpretation is not [1] Annett, M., Left, Right, Hand, and Brain: The Right Shift contradicted by the signi®cance of the sex £ location inter- Theory, Erlbaum, London, 1985. action as the three-way interaction is not signi®cant: Thus, [2] Bauer, H., Slow potential topography. Behav. Res. Methods the two-way interaction only indicates a location-speci®c Instr. Comput., 30 (1998) 20±33. sex difference (which cannot be handled in detail here) in [3] Boik, R.J., A priori tests in repeated measures designs: spatial processing which is independent of group member- effects of non-sphericity. Psychometrika, 46 (1981) 241±255. [4] Ertl, J. and Schafer, E., Brain response correlates of psycho- ship. The generalizability of the group differences is addi- metric intelligence. Nature, 223 (1969) 421±422. tionally validated by the signi®cance of the linear contrasts [5] Gittler, G., Dreidimensionaler WuÈ rfeltest (3DW): Ein Rasch- when computed for separated samples. Our ®ndings do not skalierter Test zur Messung des raÈ umlichen Vorstellungs- only indicate that males and females of the poor group show vermoÈ gens, Theoretische Grundlagen und Manual, Beltz enhanced activity in item-processing regions, but also that Test, Weinheim, 1990. [6] Haier, R.J. and Benbow, C.P., Sex differences and lateraliza- wider and more distributed cortical activities may be tion in temporal lobe glucose metabolism during mathe- observed during task processing in these subjects. This is matical reasoning. Dev. Neuropsychol., 11 (1995) 405±414. shown by the analyses using normalized values, revealing [7] Haier, R.J., Siegel, B., MacLachlan, A., Soderling, E., Lotten- differences in topographic activity between the two groups, berg, S. and Buchsbaum, M.S., Regional glucose metabolic and by the linear contrasts, indicating activity differences in changes after learning a complex visuospatial/motor task: a positron emission tomographic study. Brain Res., 570 the frontal and central area. Also, only poor subjects showed (1992) 134±143. negative potentials at fronto-medial sites, while good ones [8] Haier, R.J., Siegel, B., Nuechterlein, K.H., Hazlett, E., Wu, did not. This suggests that this cortical region is involved in J.C., Paek, J., Browning, H.L. and Buchsbaum, M.S., Corti- task processing only in the poor group. cal glucose metabolic rate correlates of abstract reasoning The time windows used for analyses were identical for and attention studied with positron emission tomography. Intelligence, 12 (1988) 199±217. both groups. Thus, it might be argued that different stages of [9] Haig, A.R., Gordon, E. and Hook, S., To scale or not to scale: cognitive processing were compared between groups. As McCarthy and Wood revisited. Electroenceph. clin. Neuro- already noted, the same statistical results were observed physiol., 103 (1997) 323±325. for all nine time windows, re¯ecting the time course of [10] Neubauer, A., Freudenthaler, H.H. and Pfurtscheller, G., the SPSs patterns: In both groups, starting from 2 s after Intelligence and spatiotemporal patterns of event related desynchronization. Intelligence, 20 (1995) 249±266. stimulus onset, gradually increasing amplitudes, but almost [11] Parks, R.W., Loewenstein, D.A., Dodrill, K.L., Barker, W.W., no changes in topographies were observed. Also, at no point Joshii, F., Chang, J.Y., Emran, A., Apicella, A., Sheramata, in time did we observe a topography identical to the one of W.A. and Duara, R., Cerebral metabolic effects of a verbal the other group at a different point in time. Hence, it seems ¯uency test: a PET scan study. J. Clin. Exp. Neuropsychol., plausible that cognitive processing is different between 10 (1988) 565±575. [12] Perrin, F., Pernier, J., Bertrand, O. and Echallier, J.F., Sphe- groups over the whole interval of task processing. rical splines for scalp potential and current source density As we analyzed SPS amplitudes relative to a stimulus mapping. Electroenceph. clin. Neurophysiol., 72 (1989) onset-baseline, the smaller amplitudes in good performers 184±187. may alternatively be explained by a speci®c pre-activation [13] Rasch, G., Probabilistic models for some intelligence and of those cortical regions which are crucial for task-speci®c attainment tests (originally published 1960 in Danish), The University of Chicago Press, Chicago, 1980. processing. In order to exclude this interpretation, we also [14] RoÈ sler, F., Heil, M., Bajric, J., Pauls, A.C. and Hennighausen, analyzed pre-stimulus baseline amplitudes (mean interval E., Patterns of cerebral activation while mental images are 200 ms before stimulus onset). No signi®cant group effect, rotated and changed in size. Psychophysiology, 32 (1995) nor any signi®cant interaction including the factor group 135±149. could be observed, nor was any signi®cance obtained [15] Schmidt, K.H. and Metzler, P., Wortschatztest (WST), Beltz Test, Weinheim, 1992. using normalized data. Although our study supports the [16] Vitouch, O., Bauer, H., Gittler, G., Leodolter, M. and Leodol- neuronal ef®ciency assumption, further, and especially ter, U., Cortical activity of good and poor spatial test perfor- longitudinal, studies will be required to analyze the causal mers during spatial and verbal processing studied with directionality of the relationship between ability and brain slow potential topography. Int. J. Psychophysiol., 27 activity. For the moment, based on the evidence that the (1997) 183±199. [17] Voyer, D., Voyer, S. and Bryden, M.P., Magnitude of sex amount and even the localization of cortical activity varies differences in spatial abilities: a meta-analysis and consid- with subjects' performance levels, we conclude that, with eration of critical variables. Psychol. Bull., 117 (1995) 250± subjects too quali®ed or too well-trained, task-relevant 270.