P300 and anticipated task difficulty

Internatmml Elsevier

Journal of Psychophysiology,

145

5 (1987) 145-149

PSP 00162

P300 and anticipated Peter Ullsperger

task difficulty

*, Hans-Gustav

I, Ulrich Neumann

Gille 1 and Manr’red Pietschmann

’

’ Department of Psychophysiology, and 2 Central Institute

Central Institute of Occupational Medicine of the G. D.R., Berlin (G. D.R.) of Cybernetics and Information Processing of the Academy of Sciences, Berlin (G. D. R.)

Key words: Event-related

(Accepted

6 March

1987)

potential;

P300; Task difficulty;

Reference

system

Recently, it has been shown that P300 measures depend on stimulus evaluation processes. This paper explores the effect of messages informing about the degree of difficulty of subsequent tasks on the P300 component. Ten healthy volunteers were asked to solve mental arithmetic tasks of graduated difficulty. The degree of difficulty was announced by messages which preceded the tasks and which were used as P300 eliciting stimuli. Measurements of the amplitude and latency of P300 components from 3 brain regions (Fz, Cz, Pa) were collected selectively for each category of task difficulty announced. Subjective ratings of difficulty and the frequency of errors confirmed the monotonic increase of task difficulty from categories I to VI. The mean P300 amplitude in categories I-VI showed a U-shaped trend, i.e. the highest amplitude occurred in the extreme categories ‘extremely easy’ and ‘very difficult’. No significant effect of stimulus categories on P300 peak latencies was found. It is concluded that the P300 amplitude reflects an evaluation in the sense of distance judgements on an internal scale which in this study concerned the task difficulty.

INTRODUCTION Although the functional significance of the endogenous-determined P300 component of the event-related brain potential (ERP) is not yet clear, the P300 is of growing interest as an index for assessing variations in cognitive activity (e.g. Donchin et al., 1978; Pritchard, 1981; Riisler, 1983; Kostandov, 1983; Ullsperger and Gille, 1985). The relevance of an eliciting event and its probability are broadly accepted as factors reliably influencing the P300 (Duncan-Johnson and Donchin, 1977; Donchin, 1979). In several investigations it was furthermore demonstrated that P300 amplitude varied with evaluation and categorization of stimuli according to different dimensions or sets of dimensions, e.g. physical features (Nash

Correspondence: P. Ullsperger, Zentralinstitut fur Arbeitsmedizin der DDR, Noldnerstrasse 40/42, DDR-1134 Berlin, G.D.R. 0167-8760/87/$03.50

0 1987 Elsevier Science Publishers

and Jasiukaitis, 1983), the letter position in the alphabet (Michalewski et al., 1984), interest value (Hornberg et al., 1984), and processing difficulty (Neumann et al., 1986). Regarding the latter result, the effect on P300 amplitude peaking at 300-600 ms after task presentation could not be explained by mental operations necessary for task performance per se, because the task solution took several seconds. Furthermore, a distinct effect of the difficulty of frequently presented reference tasks was shown (Ullsperger et al., 1986): a decrease of P300 amplitudes with increasing task difficulty was found when presenting reference tasks which were classified to the upper pole of the difficulty dimension. With reference tasks of low difficulty, however, the P300 amplitudes increased with increasing difficulty. Therefore, it was assumed that P300 amplitude changes might reflect a task evaluation in the sense of a first distance judgement comparing the actual task with a reference on the internal difficulty scale. To investigate this problem, in the

B.V. (Biomedical

Division)

146

present study the information about task difficulty was given explicitly and isolated from the tasks. Visually displayed verbal information about the difficulty of subsequent tasks were used as P300eliciting stimuli. We examined whether P300 amplitudes change systematically as a function of the degree of announced task difficulty.

MATERIALS

AND

METHODS

Subjects Ten healthy volunteers (3 women, 7 men) aged from 22 to 28 years (mean 25.7 years) participated in the experiment. They were paid for their participation. Procedure and stimuli The subjects sat in an armchair in front of a computer display at a distance of approximately 70 cm. The stimuli were always exposed on the same part of the screen. One of 6 different messages (categories I-VI, cf. Table I) informing the subject explicitly about the difficulty of the subsequent task or about the no-task (letter strings, category 0) appeared after he or she had asked for the message by pressing the return key. The time interval between pressing the key and presentation on the screen was 400 ms. Then, after pressing the

TABLE

I

Word stimuli and examples Category

Word stmulus

0

ABCDE

I

kssert leicht (extremely easy) Sehr leicht

II III IV V VI

(very easy) Leicht (easy) Mittel (medium) Schwer (difficult) Sehr schwer (very difficult)

of subsequent mental arithmetic tasks Task (examples)

Probability of occurrence 0.8

2=

0.033

2+5=

0.033

22+41=

0.033

9x3+91=

0.033

14x6+25 14x19x4-63=

=

0.033 0.033

same key again, a mental arithmetic task appeared whose difficulty agreed with the preceding message. In category 0 cases, no task was presented. After solving the task, the result was to be entered via the key board and the task vanished from the screen. The messages with subsequent tasks were presented equiprobably with the total probability of P = 0.2 embedded in a sequence of category 0 strings (P = 0.8). Altogether, 900 stimuli of the 7 categories were presented in an unpredictable sequence. The stimulus and task presentation and registration of errors were controlled by a microcomputer (K 1520, Robotron). The frequency of errors was calculated selectively for each category of task difficulty. The subjective difficulty of the tasks in categories I-VI was estimated by the subjects by means of a 7-point rating scale after the experiment. Recording The EEG signal was recorded from Fz, Cz and Pz with a linked-earlobe reference using Beckman Ag/AgCl electrodes. The vertical electro-oculogram (EOG) was recorded by electrodes fixed above and below the left eye. After amplification (time constant 2.5 s, high frequency cut-off 30 Hz; Mingograf, Elema-Schbnander) the signals were tape-recorded (EAM 500, Tesla) together with 7 different trigger impulses indicating the start and category of messages presented. The artifact-free EEG signals were selectively averaged, starting 200 ms before and ending 1000 ms after the presentation of the messages. The sampling interval was 5 ms; the number of averaged EEG epochs per category was between 25 and 30, depending on artifact frequency. From the 720 presentations of category-0 stimuli, 30 timelocked EEG epochs were randomly selected throughout the experiment and averaged. A baseline for computing the peak amplitudes of the averaged ERPs was calculated across the 200-ms prestimulus epoch. Peak amplitudes and latencies were measured by means of a VT 20 A computer (Videoton). All measures were obtained separately for the 3 EEG derivations and the 7 stimulus categories of each subject. The ERP parameters (P300 peak amplitude and latency)

147

were tested by ANOVAs (10 subjects, 7 categories, separately for derivations). Furthermore, the statistical analyses included multiple comparisons (Newman-Keuls test), monotonic and non-monotonic trend tests (Page, Ofenheimer; see Lienert, 1973). The level of significance was set to 0.05.

..ABCDE’

,

EXTREMELY

RESULTS Behavioural data The ratings of task difficulty showed a significant monotonic increase (Page-test, P I 5%) from category I to VI. The coefficient of concordance was 0.94, indicating a high degree of correspondence of difficulty estimation between subjects. The influence of graduated task difficulty was a significant factor (ANOVA) for the frequency of errors. The mean values increased systematically from 0.1 in I to 8.1 in VI. ERP data The ERP grand averages for the 7 stimulus categories are shown in Fig. 1. The highly probable ‘ABCDE’ stimuli evoked a small, sometimes negligible positive deflection at about 460 ms. The messages informing the subjects about the difficulty of the subsequent tasks, however, elicited clearly expressed P300 components. The highest amplitudes were found in both categories ‘extremely easy’ and ‘very difficult’. The results of ANOVAs computed separately for each of the derivations are given in Table II. A significant influence of stimulus categories was found in all derivations. The multiple comparisons of mean values indicated significant differences between P300 amplitudes of categories 0 on the one hand

TABLE Results

,,MEDIUM’

EASY

“VERY

EASY ~

,, EASY

”

1

0

200

.VEAV

400

500

800

Ins

0

DIFFICULT

zw

400

for P300 amplitudes

at the three derivations

Fz, Cz and Pz.

Source of variance

4

FZ

CZ

PZ

F5%

Subjects Category

9 6

8.84 * 17.65 *

10.56 * 16.54 *

15.99 * 20.13 *

2.06 2.27

*P 5 0.05

Ins

and I, II, III, V and VI on the other. Furthermore the mean values in category IV were significantly lower than in I, V and VI which was again true for the 3 derivations. The mean values of P300 amplitudes are shown in Fig. 2. A U-shaped trend of the mean amplitudes elicited by the equiprobable stimuli of categories I-VI, with its minimum at IV, is discernible. This bitone trend was statistically confirmed (Ofenheimer; cf. Lienert, 1973). No significant effect of stimulus categories was proved regarding the P300 peak latencies. The mean values were 463 ms in Fz, 461 ms in Cz and 466 ms in Pz.

of analyses of variance calculated

800

Fig. 1. Grand averaged waveforms across all subjects. ERPs were elicited by the messages indicated which preceded mental arithmetic tasks of corresponding difficulty. - - -, Fz; p, Cz; , Pz. Arrow indicates the start of presentation.

II

F-values

600

148

r 1 AMPLITUDE 25 !J” 20

1 F

. ..CDT~

,, ::~s”:

y-v .,::“,; ,,

,619””

.MtD,“M. p4rrlc”Lr’

;;py&,,”

Fig. 2. Mean values and standard errors of P300 amplitudes in Fz, Cz and Pz arranged as a function of increasing task difficulty.

DISCUSSION The messages which informed the subjects about the difficulty of the subsequent mental arithmetic tasks elicited ERPs with P300 amplitudes depending on the difficulty announced. The senseless ‘ABCDE’ stimulus which was presented in 80% of all cases and which was not followed by a task evoked a small positive deflection, only. This was expected because of its high probability of occurrence and its possible function as a reference. The P300 amplitudes in ERPs to the equiprobably presented 6 different word messages, however, varied in a U-shaped trend as a function of the announced task difficulty. The smallest values were found for the message ‘medium’. This result seems to be contradictory to those of earlier investigations (Neumann et al., 1986; Ullsperger et al., 1986) where linear relationships were found between P300 amplitude and task difficulty. Two major differences between the present and earlier results are to be considered. Firstly, in this study the information about difficulty was given explicitly before the task occurred. The message had the character of a warning stimulus, which may be used for an anticipatory adjustment of the information-processing system (Rosier, 1985; Ullsperger and Gille, 1980). However, it can hardly be assumed that the presentation of the 6 different messages themselves provoke “different extents of controlled information processing” which are sup-

posed to affect the P300 amplitude (Riisler, 1983). Secondly, the experiments were designed similar to a reference-task paradigm (Ullsperger et al., 1986). In this study, however, the reference tasks were replaced by senseless stimuli without subsequent tasks. In the earlier investigation the reference tasks matched the other tasks to the extent that all tasks could be ranged on a common dimension. The position of the reference on the internal difficulty scale was shown to be adjustable by the class of reference tasks chosen. By way of contrast in this study one can assume that the ‘ABCDE’ stimuli, despite their high probability, did not influence the position of the reference point of the internal difficulty scale. In other words, the senseless letter string was not included in the process of difficulty evaluation (cf. Sarris, 1971). The message ‘medium’ elicited the smallest P300 amplitudes in comparison to the other equiprobable presented messages. The amplitudes augmented relative to ‘medium’ with increasing differences on the difficulty scale. This was true in both directions, i.e. the highest amplitudes were found in the extreme categories I and VI. These results lead one to assume that the reference point on the internal difficulty scale was adjusted to ‘medium’. The fact that the minimum amplitude was found in category IV also speaks against an effect of physical length of messages alone: the number of letters per message in categories III, IV and V were identical. According to psychological findings obtained in research in the frame of adaptation-level theory (Krbner, 1976; for overview see Lauterbach and Sarris, 1980) one can expect that the reference point (or adaptation level) is arranged between the extreme poles of a given internal continuum. In this study the difficulty of subsequent tasks in category IV was rated as medium and probably the labelling ‘medium’ reinforced this classification by the subjects. Summing up, it can be stated that the amplitude of the P300 component reflects distance judgements on an internal continuum relative to the actual position of the reference point. Considering this suggestion, the results give reason to assume that the degree of anticipated, cognitive effort affected the P300 amplitude. To what extent

149

the first distance judgement on the basis of perceptual differences of stimuli may help the subject to adjust the CNS in a preparatory fashon appropriate for solving the task should be examined in future investigations. Nevertheless, the non-obtrusive determination of the adaptation level and and its changes within a certain reference system due to different influences will be of major importance in many psychophysiological investigations.

REFERENCES Donchin, E., Ritter, W. and McCallum, W.C. (1978) Cognitive psychophysiology: the endogenous components of the ERP. In E. Callaway, P. Tueting and S. Koslow (Eds.), Evenf-Related Brarn Potentials in Man, Academic, New York, pp. 349-441. Donchin, E. (1979) Event-related brain potentials: a tool in the study of human information processing. In H. Begleiter (Ed.), Evoked Brain Potentials and Behavior, Plenum, New York, London, pp. 13-88. Duncan-Johnson, C.C. and Donchin, E. (1977) On quantifying surprise: the variation of event-related potentials with subjective probability. Psychophysiology, 14: 456-467. Hiimberg, V., Griinewald, G. and Netz, J. (1984) Category scaling of the interest value of complex visual stimuli and late positive components of the evoked potential. In R. Karrer, J. Cohen and P. Tueting (Eds.), Brain and Information: Event-Related Potentials. Annals of the New York Academy of Saences, Vol. 425, The New York Academy of Sciences, New York, pp. 216-222. Kostandov, E.A. (1983) Funkcional’naja Asimmetrija PoluSarij Mozga i Neososnavaemoe Vosprijatie, Nauka, Moskva. KrGner, B. (1976) Selektive Prozesse in der Reizwahmehmung und psychophysiologische Aktivierung. Z. Exp. Angew. Psychol., 23: 623-665.

Lauterbach, W. and Sarris, V. (1980) Beitrlge zur psychologischen Bezugssystemforschung. Huber, Bern. Lienert. G.A. (1973) Vertedungsfreie Methoden m der Biostatistlk. Vol. I, Anton Hain, Meisenheim, pp. 358, 680. Michalewski, H.J., Kamel, A.M., Collier, J. and Starr, A. (1984) Mental distance judgement and P300. Psychophysrology, 21: 588. Nash, A.J. and Jasiukaitis, P.A. (1983) Effects of magnitude and direction of stimulus change on the human event-related potentials. Psychophysiology, 20: 460. Neumann, U., Ullsperger, P. and Gille, H.-G. (1986) Effects of graduated difficulty on P300 component of the event-related brain potential. Z. Psychol., 194: 25-37. Pritchard, W.S. (1981) Psychophysiology of P300. Psychol. Bull., 89: 506-540. Rasler, F. (1983) Endogenous ‘ ERPs’ and cognition: probes, prospects, and pitfalls in matching pieces of the mind-body puzzle. In A.W.K. Gaillard and W. Ritter (Eds.) Tutorials in ERP Research: Endogenous Components, Elsevier, Amsterdam, pp. 9-35. Riisler, F. (1985) Erelgniskorrelierte Hirnrindenpotentiale als Indikatoren kognitiver Prozesse bei der Informationsverarbeitung. Report DFG Project RO 529, Kiel. Sarris, V. (1971) Wahrnehmung und Urteil. Verlag fir Psychologie, GGttingen. Ullsperger, P. and Gille, H.-G. (1980) Contingent negative variation and mental arithmetic of tasks presented for limited periods. Wiss. Z. Karl-Marx-Univ. Leipzig, Math.Naturwiss. R., 29: 221-224. Ullsperger, P. and Gille, H.-G. (1985) Die P300-Komponente des ereignisbezogenen Himpotentials als lndikator bei der Untersuchung von Informationsverarbeitungsprozessen. Psychiat. Neural. Med. Psychol., 37: 582-588. Ullsperger, P., Peikert, D., Tessin, S., Gille, H.-G. and Metz, A.-M. (1986) Die P300 Komponente des ereignisbezogenen Himpotentials als Indikator in der Psychophysik kognitiver Prozesse. Z. Psychologie, 194: 365-375.