ELSEVIER

INTERNATIONAL JOURNALOF PSYCHOPHYSIOLOGY

International Journal of Psychophysiology 17 (1994) 47-56

Mental comparison of visually presented two-digit numbers: a P300 study

Kerstin Grune a, Peter Ullsperger a,*, Matthias Miille b, Axe1 Mecklinger ’

u Federal Institute for Occupational Medicine, NtildnerstralJe 40-42, D-10317 Berlin, Germany, ’ Medical University, Haus 23 a, Ratzeburger Allee 160, D-23564 Liibeck, Germany,

’ Psychological Institute, Free Unictersity Berlin, Habelschwerdter Allee 45, D-14195 Berlin, Germany

(Accepted 20 January 1994)

Abstract

Comparison processes were investigated in a multiple-stimulus paradigm with a pseudo-random sequence of visually presented numbers 11 to 20. The subjects’ task was to compare each current number with the preceding one and to indicate whether it was larger or smaller. ERPs were selectively averaged for the 10 number stimuli, for three functional conditions according to the information the numbers provide about the ensuing response and for the differences between consecutive numbers. P300 amplitude averaged for each number stimulus showed a U-shaped trend with largest amplitudes for the numbers 11 and 20. It was found that P300 amplitudes change with the amount of information the current number is delivering for the ensuing response. This information delivery is related to the processing in the subsequent trial, as revealed by a negative correlation between P300 in the current trial and RT in the ensuing trial. Reaction times decreased significantly with increasing difference between the current number and the preceding one. This symbolic distance effect was not found for P300 parameters. The dissociation between RT and P300 data provides evidence for the assumption that under the present experimental conditions informational transaction leading to the distance effect occur after the elicitation of P300.

Key words: Mental comparison; Event-related brain potential; P300 component; Symbolic distance effect

1. Introduction

Mental comparison processes have long been an object of psychological interest (Moyer and

* Corresponding author. Tel.: (+49-30) 5513 8510; Fax: (+ 49-30) 5513 8171.

Landauer, 1967; Sekuler and Mierkiewicz, 1977; Link, 1990). A basic phenomenon of mental com- parison is the so-called symbolic distance effect (SDE, Moyer and Dumais, 1978). The greater the psychological difference between members of a pair of symbols stored in memory, the faster people can compare their magnitudes. Moyer and Landauer (1967) observed SDE also in digit com-

0167.8760/94/$07.00 0 1994 Elsevier Science B.V. All rights reserved

SD1 0167-8760(94)00003-W

parison tasks. They found that decisions (which digit of a given digit pair was larger) were faster when the difference between digits was large rather than small.

Distance effects were: shown to be a general property of memorial size comparison as ob- served in tasks with judgements of animal sizes (Moyer, 1973) and with those concerning every- day objects as well (Paivio, 1975). Interestingly, the SDE seems not to be restricted to human cognition. D’Amato and Colombo (1990) found a distance effect in monkeys tested with pairs of items with which they were familiar.

Although the SDE characterizes memory com- parisons quite generally, the cognitive mecha- nisms underlying this phenomenon remain unre- solved. There are two classes of explanations to account for the SDE. So-called analog models assume that subjects retrieve information from analog size representations in memory in order to make comparative judgments. For example Moyer and Landauer (1967) proposed that items are converted to analog representations or percept- like entities and than they are compared by the same processes that compare perceptual informa- tion.

The second class of models for the SDE as- sumes that the nature of the code used in a symbolic comparison is more abstract than an analog code and only includes information about the symbols’ relative position on a task relevant dimension. The semantic coding model advanced by Banks (1977) for example assumes three se- quential processing stages which combine addi- tively to determine reaction time CRT) in sym- bolic comparisons; encoding, choice and response selection/execution.

The encoding stage generates discrete seman- tic codes for each stimulus based on information represented in semantic memory. These discrete codes are the basis for operating in the subse- quent stages. In the choice stage the codes are processed until they distinguish between the two stimuli, and then match one of the resultant codes to the code for the instructions. In the final stage a response is selected and executed. This model accounts for the SDE because the proba- bility that two items receive identical codes in the

initial encoding stage increases as the distance between the two items decreases.

An additional phenomenon observed in sym- bolic comparison tasks is the so-called end-anchor effect (Moyer and Dumais, 1978; Henderson and Well, 1985). It has repeatedly been shown that different strategies can be used depending on whether or not a given pair of items contains an end-term (Banks et al., 1975; Potts, 1974; Meck- linger et al., 1993). For example Potts (1974) demonstrated that the comparison process can be by-passed completely if the first member of a pair of to-be-compared items is an end-term within the stimulus scale. Based on these results it can be assumed that when pairs of symbols are com- pared at least two forms of representation are held in working memory. First, a representation of the relative ordering of the symbols and sec- ond, information whether a symbol is an end term within the stimulus scale or not (cf. Meck- linger et al., 1993).

Most of the work investigating the cognitive processes underlying performance in comparison tasks has involved behavioral measures. A com- plementary approach is to record event-related potentials (ERP) elicited by the stimuli in a men- tal comparison task and to use components of the ERP to make inferences about the timing and nature of stimulus processing under different ex- perimental conditions.

Event-related brain potentials (ERPs) elicited by the numerical stimuli can be used as markers of encoding and comparison processes occurring in a particular task situation. In response to pre- sentation of discrete numbers the ERPs are recorded from the scalp. This approach is partic- ularly promising because the P300 component of the ERP has been shown to covary with stimulus evaluation processes (McCarthy and Donchin, 1981). P300 latency has been associated mainly with the duration of stimulus evaluation and less with response selection and execution processes (Magliero et al., 1984). Donchin and Coles (1988) argue that P300 indicates processes involved in strategic future-oriented information processing. In addition, Ullsperger and colleagues NJlls- perger and Gille, 19X8; Ullsperger and Baldeweg, 1992) showed that P300 amplitude reflects the

K. Grune et al. /International Journal of Psychophysiology 17 (1994) 47-56 49

distance between an actual stimulus and the cur- rent reference or Adaptation level (Helson, 1964) of the inferred internal model.

Ruchkin et al. (1990) used ERP measures in an Sl-S2 paradigm to investigate digit compari- son processes. The amount of outcome informa- tion delivered by Sl was manipulated by varying the amount of prediction uncertainty that Sl could resolve. The five odd numbers 1, 3, 5, 7, and 9 were used for Sl, the four even numbers 2, 4, 6, and 8 served as S2. The task was, prior to the start of each trial, to predict if Sl would be greater or lesser than S2. ERPs were recorded to Sl and to S2 stimuli. The P300 (P3b) was sensi- tive to delivery of information (complete, partial) by Sl. P300 amplitudes elicited by S2 stimuli were inversely related to the subject’s outcome expec- tations.

Recently, Mecklinger et al. (1993) used ERP measures to investigate processes underlying mental comparison. The authors used a serial paired comparison task with the spoken digits 1 to 5. The subjects’ task was to decide whether each digit presented was smaller or larger than the preceding one. Shortest RTs were obtained for the two end terms “one” and “five”. The amplitude of the P300 component was found to covary with the information the eliciting digit provided to the response to the ensuing digit: P300 was largest for the two end terms which provided full information with respect to the re- sponse to the ensuing digit; it was intermediate for the two digits “two” and “four” which deliv- ered partial information with respect to the ensu- ing response and smallest for the digit “three” which provided no information about the upcom- ing response.

Rather than auditory presentation of numbers from 1 to 5 (Me&linger et al., 1993) in the present experiment numbers from 11 to 20 were presented visually. This would allow us to exam- ine the stimulus-to-category assignment of ten digits and comparison processes for a larger num- ber of numerical differences as well. Again as in the Mecklinger et al. study three functional con- ditions were constructed by sorting the numbers with respect to the information each digit pro- vides about the subsequent response. The first

condition (FULL) comprised trials with the end- term numbers 11 and 20, conveying full informa- tion about the ensuing response. The second con- dition included numbers providing partial infor- mation about the ensuing response (PART). Fi- nally, the third condition (NONE) included trials with the middle-term numbers 15 and 16 which provided no information about the subsequent response.

An additional issue addressed in this study was to examine the extent to which the SDE is re- flected in P300 measures. The semantic coding model proposed by Banks (1977) assumes that the discrimination of stimulus codes within the choice stage is responsible for the SDE. If the P300 is related to the choice stage it should be affected when the comparison of two symbols takes longer.

2. Materials and methods

2.1. Subjects

Ten adult right-handed volunteers (7 female) participated in the experiment (age range: 22-39 years, mean: 29.6 years). The subjects had normal or corrected to normal vision. Eight subjects had experiences with former ERP studies.

2.2. Stimuli

The time course of stimulus presentation is presented in Fig. 1. The stimuli were the numbers from 11 to 20. Two-digit numbers were used because they allow an examination of comparison processes between 10 different numbers under

I, trial 1 trial 2 trial 3

” XX 11 XX 18 XX 13 XX

)I~ ~1 -! -~ II

500 ms 2s

’ Task: 18 )( 11 ? 13 >( 18 ? I-

Fig. 1. Time course of stimulus presentation in the serial

paired number comparison task.

similar visual conditions. They were presented in a random sequence in the centre of a 15” com- puter screen (size of numbers: 5 x 6 mm; distance between subject and screen ca. 70 cm). The ten numbers were presented equiprobably. They were randomized with the constraints that identical numbers did not occur consecutively and that the differences between consecutive numbers were equiprobable. The differences between consecu- tive numbers ranged from - 7 to 7.

The interval between two consecutive stimulus onsets (stimulus onset asynchrony, SOA) was 2 s. Each number stimulus was presented for 500 ms. The two digits were then replaced by two X which were again replaced by the two digits of the next trial. Stimulus presentation and collec- tion of behavioral data were controlled by an IBM-compatible 386 computer.

2.3. Procedure

The subjects’ task was to decide whether the current number was larger or smaller than the preceding one. If it was larger, they had to press the right mouse button with the middle finger of the right hand. If the number was smaller than the preceding one, they had to respond by press- ing the left mouse button with the index finger of the right hand. The subjects were instructed to respond as quickly and accurately as possible. Prior to the test session the subjects performed 100 trials of practice. The test session included 1120 trials divided into two blocks separated by a lo-min break. The duration of the entire session was about 60 min.

2.4. ERP recording and data acquisition

The EEG was recorded with Beckman Ag/AgCl electrodes at the electrode positions Fz, Cz and Pz according to the International lo-20 System. Linked earlobe electrodes served as reference. The ground electrode was placed on the mid forehead. Vertical eye movements were monitored by means of a left supraorbitally lo- cated electrode.

The scalp electrodes were fixed with Grass electrode cream. Electrode impedance was kept

below 5 k0. The EEG and EOG were amplified by Graphtek-amplifiers (time constant: 5 s; upper frequency: 30 Hz). All channels were recorded continuously and A-D converted at a frequency of 250 Hz. EEG and EOG recording was con- trolled by a 486 IBM-compatible computer con- nected to an A-D conversion interface (Cam- bridge Electronic Design 1401).

2.5. Data anulysis

2.5.1. Performance duta

Reaction time (RT) was defined as the interval between the onset of number presentation and the subject’s keypress. Only correct trials were selected for RT averaging. In the first stage RTs and number of errors were sorted according to the three functional conditions described above. RTs were averaged for the end terms “11” and “20” (maximum 224 trials), which provided full information about the following response (FULL). The second condition included the numbers “12”, “13”, “14” and “17”, “I?,“, “19” (PART; maxi- mum 672 trials), providing partial information about the outcome of the next trial. The mid terms “1.5” and “16” (maximum 224 trials) which provided no information about the response to the ensuing number were averaged in the third condition (NONE).

In stage two we sorted all trials into three groups according to the amount of information provided by the current number. RTs from trials following the FULL condition were averaged in condition FOLLOWING-FULL (FO-FULL). The same was done for PART (FO-PART) and NONE conditions (FO-NONE).

In stage 3 RT and the number of errors were averaged according to the numerical differences between consecutive numbers ( - 7 to 7).

2.5.2. ERP data ERPs time-locked to the onset of number

stimuli were averaged from 200 ms prior to the stimuli to 800 ms thereafter. A baseline averaged across the first 50 data points (200 ms) preceding the stimulus was subtracted from each data point in the waveforms. Trials containing ocular or EEG artifacts (criterion: +50 pV1 or incorrect

K. Grune et al. /International Journal of P.rychophysiology 17 (1994) 47-M 51

responses (errors and misses) were excluded from further analysis. Before the measurement of the P300 component EEG signals were digitally low- pass filtered at 3.5 Hz (24 dB).

The amplitude of the P300 was defined as the maximum positive deflection occurring between 300 and 700 ms post-stimulus. Its latency was defined as the time point of maximum positive amplitude in this time range.

P300 amplitude and latency were measured in ERPs which were selectively averaged for num- bers (11 to 201 and for the numerical differences between consecutive numbers (- 7 to 7).

The averaging procedure described for analy- sis of performance data was also employed to average the ERP-trials according to the three functional conditions described above (FULL, PART and NONE).

2.6. Statistical analyses

Repeated-measures ANOVAs were computed with the factor Condition (FULL, PART, NONE or FO-FULL, FO-PART, FO-NONE) for behav- ioral and with the factor Condition (FULL, PART, NONE) for P300 data. The data was also quantified in repeated-measures ANOVAs as a function of differences between consecutive num- bers with the factors absolute value of numerical Difference (7) and Response type (2 - larger vs. smaller) as well as the factor Electrode (3 - Fz, Cz, Pz) for the ERP data. In order to correct for violations of sphericity (Vasey and Thayer, 1989) all within-subjects main effects and interactions with two or more degrees of freedom in the numerator were adjusted using the Huyn-Feldt Epsilon. All post-hoc comparisons were com- puted with the Bonferroni test. The significance level of statistical tests was set to 0.05.

3. Results

3.1. Influence of information delivery of current numbers

3.1.1. Reaction time Table 1A presents mean RTs for the condi-

tions FULL, PART and NONE. There is no

Table 1

Mean values of reaction time with corresponding standard deviations averaged according to the three functional condi-

tions defined on the basis of the amount of information

provided by the numbers, (A) for the current and (B) for the

following trials

(A) Condition Reaction Time SD

(ms) (ms)

FULL

PART

NONE

58h x5

569 I IO

569 128

(B) Condition Reaction Time

(ms)

SD (ms)

FO-FULL

FO-PART

FO-NONE

475 154

5x9 IO4

616 102

Numbers per functional condition in current (A) or in follow-

ing trials (B): FULL: Il. 20; PART: 12, 13, 14, 17, 18, 19;

NONE: 15, 16.

significant influence of the factor Condition on RT (F(2,18) = 1.67; p > 0.05).

3.1.2. P300 Fig. 2 shows the grand mean waveforms of the

ERPs selectively averaged for the numbers 11 to 20. The ERPs show the following peaks Nl, P2, N2 and P3. Neither the amplitudes nor the peak- latencies of Nl, P2 and N2 revealed systematic changes as a function of the eliciting numbers. However, for P300 smallest amplitudes were found in the middle of the number scale with an increasing trend to both ends of the number set.

The corresponding mean values of P300 ampli- tude as a function of eliciting numbers are pre- sented in Fig. 3.

Fig. 4 shows the mean P300 amplitudes after collapsing the data for the functional conditions FULL, PART and NONE. Largest amplitudes were found when full information was provided and smallest in the NONE condition. Repeated- measures ANOVA confirms the influence of the factor Condition on P300 amplitude (F(2,18) = 77.13, p < 0.05, E = 1.0). Post-hoc comparisons indicate significant differences between the con- ditions FULL and NONE (p < 0.05).

As revealed by Figs. 2 and 4 P300 amplitude increases from frontal to parietal location (Elec- trode: F(2,18) = 20.6, p < 0.05, E = 1.0).

3.2. Influence of current numbers on RT in ensuing trial.7

The RT mean values given in Table 1B suggest that the more information provided in the cur- rent trial about the response probability for the following trial, the shorter the RT. In addition, a repeated-measures ANOVA also reveals a signif-

* Y” -PI ---.a - PZ

Fig. 2. Grand mean ERP waveforms selectively averaged for

the 10 number stimuli 1 I to 20.

icant influence of amount of information deliv- ered on RT (F(2,18) = 22.54, p < 0.05, E = 0.53). The post-hoc comparison FO-FULL vs FO- NONE (p < 0.05) was significant.

Moreover, there was a significant correlation between the P300 amplitude and the RTs in corresponding subsequent trials (correlation coef- ficient: -0.558, p < 0.05). The more information about subsequent response is provided by a stim- ulus the larger the P300 amplitude to this stimu- lus and the faster the RT in the following trial.

3.3. Effects of numerical differences between con-

secutil’e number stimuli

3.3.1. Performance data Mean RT and accuracy data for the numerical

differences ( - 7 to 7) are presented in Fig. 5. RTs and the number of errors decrease with increas- ing differences between two consecutive num- bers. For RT a significant main effect of factor Difference was found (F(6,54) = 49.55, p < 0.05, E = 0.53). There is a significant linear trend as a function of the absolute numerical differences from 1 to 7 (t = 9.84; p < 0.05), i.e. the greater the difference the faster the response.

RT is also significantly affected by Response type (F(1,9) = 5.72; p < 0.051, suggesting that “larger” responses were faster than “smaller” responses. This was particularly true at larger differences where a significant interaction be- tween numerical Difference and Response type (F(6,54) = 2.69; p < 0.05; E = 0.88) was found.

For the number of errors a significant main effect of the factor Difference was also found (F(6,48) = 13.47; p < 0.05; E = 0.38), in addition, the linear trend as a function of numerical differ- ence was significant (t = 4.3; p < 0.05).

3.3.2. ERP data In the ERPs selectively averaged for numerical

differences, peaks Nl, P2 and N2 did not display any systematic changes as a function of the nu- merical differences between consecutive number stimuli. In addition, a P300 component increasing in amplitude from Fz to Pz (Electrode: F(2,18) = 20.05; p < 0.05; E = 1) was observed. The P300

K. Grune et al. /International Journal of Psychophysiology 17 (1994) 47-56 53

__ Fz 1 0

11 12 13 14 15 16 17 18 19 20

Number stimulus

Fig. 3. Mean values of P300 amplitude as a function of the number stimuli 11 to 20 with indication of quadratic regression lines for

the three derivations Fz, Cz, and Pz.

FULL PART NONE

Fig. 4. Mean values and standard errors of P300 amplitudes

computed separately for the three functional conditions de-

fined according to the amount of information provided by the

eliciting number stimulus (FULL: 11, 20; PART: 12, 13, 14,

17, 18, 19; NONE: 1.5, 16).

latencies ranged from 460 ms to 510 ms. Signifi- cant P300 amplitude and latency changes or trends as a function of the numerical differences were not found.

4. Discussion

In the present experiment RT and frequency of errors decreased with increasing numerical dif-

Fig. 5. Mean reaction times (ms) and number of errors (n) as a

function of the numerical differences between subsequent numbers.

ference between numbers to be compared. The behavioral data are consistent with the classical and multiply replicated findings of Moyer and Landauer (1967) where subjects were required to decide which of a pair of single digits (l-9) was numerically larger. The time required for this numerical judgment was an inverse function of the numerical difference between the presented digits. Interestingly, under the present conditions of a serial paired number-comparison task the SDE is not reflected by P300 latency or ampli- tude data.

An interpretation for this dissociation between RT and P300 latency data can be derived from the semantic coding model advanced by Banks (1977). The model assumes that in the first stage tic. the encoding stage) semantic codes of the stimuli are generated while the process of com- parison which is responsible for the SDE oper- ates in the choice stage. Thus, the observation that P300 latencies do not increase with decreas- ing distances between the to-be-compared items implies that under the present conditions P300 is not related to the choice stage. It is noteworthy that there has been an extensive discussion on the relation between P300 latency and the re- sponse choice stage in recent years (Magliero et al., 1984; Ragout and Renault, 198%. Given that there are functional similarities between the re- sponse choice stage assumed in visual matching tasks and the choice stage proposed in the se- mantic coding model, the data suggest that P300 elicited in this serial comparison task is not re- lated to the response choice stage (cf. McCarthy and Donchin, 1981) ‘.

The P300 amplitude of visually elicited ERPs was found to vary with the three functional condi- tions FULL, PART and NONE. Based on the knowledge concerning the position of the incom- ing stimulus on the internal number scale subjects extract information about the probability of the response in the subsequent trial. Largest P300 amplitudes were found in the FULL condition, smallest amplitudes in the NONE condition. Our results corroborate the findings of Mecklinger et al. ( 1993) and suggest that information extraction is reflected by P300 of auditory and visual ERPs as well.

The results support the notion that during symbolic comparisons two different kinds of in- formation are held in working memory: a symbol’s

’ In a recent study (Grune et al., lY93) it was shown, that the

numerical distance between to-be-compared numbers can af-

fect the P300 component, i.e., under conditions of simultane- ous presentation of the numbers. This finding support the

assumption that under the present conditions of consecutive

presentation of the numbers to compare the choice stage might be outside of a time range in which the P300 is affected.

relative position within the stimulus scale and whether it is an end-item or not. The systematic changes in RT in trials following each of the conditions FULL, PART, NONE and the nega- tive correlation between RT and P300 amplitude measured in the current trials, confirms our as- sumption that this information was strategically processed to foresee the response probability in ensuing trials. The more information about re- sponse probability that subjects could extract from a particular number, the faster they responded in subsequent trials.

These findings are in accordance with results of Gratton et al. (1990). They investigated the processing of a priming stimulus varying in its information content. Subjects were shown se- quences of two letters Sl and S2. The imperative stimulus S2 had a 0.8, 0.5 or 0.2 probability of physically matching Sl. The different probability conditions were signalled by the position of a dot flanking the Sl letter. RT and accuracy data indicated priming of responses as a function of the information conveyed by Sl. Amplitudes of P300 were sensitive to the amount of stimulus information. The P300 elicited by S2 stimuli re- flects the utilization of information delivered by the Sl stimuli.

Moreover, in the Sl-S2 paradigm of Ruchkin et al. (1990) P300 appeared sensitive to the amount of prediction uncertainty that Sl could resolve. In the present study the subjects per- formed a serial number comparison, therefore a distinction between Sl and S2 stimuli was not possible. Here each stimulus carried information about a forthcoming response and may be under- stood as an Sl stimulus, however, it also simulta- neously represents the imperative stimulus 62) in that it completely resolves the uncertainty about the response in the very current trial. This “dou- ble function” of number stimuli should be disen- tangled in experiments with pairwise number pre- sentation (cf. Grune et al., 1993).

The systematic P300 amplitude changes in de- pendence on the eliciting numbers can also be seen as an effect of stimulus-to-number-category assignment. The results hint at an internal repre- sentation of the number scale in order to arrange the individual number stimuli. It is argued that

K. Grune et al. /International Journal of f%ychophysiology 17 (1994) 47-56 55

P300 reflects the distance of the current number from the adaptation level (Helson, 1964; Ullsperger and Baldeweg 1992; Mecklinger and Ullsperger, 1993) of the internal number scale. The adaptation level can be expected for the present task to be located near 15 and 16.

D’Amato, M.R. and Colombo, M. (1YYO) The symbolic dis-

tance effect in monkeys (Cebus aprlla). Anim. Learn. Be-

hav., 18: 133-140.

Donchin, E. and Coles, M.G.H. (1988) Is the P300 componenr

a manifestation of context updating? Behav. Brain Sci., 11:

357-373.

Similar findings were reported by Yanai and Nageishi (1993) who associated numbers from 0 to 9 with ten corresponding buttons (choice reac- tion task) and got a U-shaped function for P300 amplitude with smaller amplitudes in the middle of the number scale. These findings support the view that prior to numerical comparisons, individ- ual numbers are encoded and arranged into an internal number scale held in working memory. Consequently, in this experiment a similar U- shaped trend for P300 amplitude can be expected for both the information-extraction and the adap- tation-level hypotheses.

Gratton, G., Bosco, C.M., Kramer. A.F.. Coles. M.G.H.,

Wickens. C.D. and Donchin, E. (1990) Event-related brain

potentials as indices of information extraction and re-

sponse priming. Electroenceph. Clin. Neurophysiol., 75:

419-432.

Grune. K., Mecklinger, A. and Ullsperger, P. (lY93) Mental

comparison: P300 component of the ERP reflects the

symbolic distance effect. NeuroReport. 4: 1272-1274.

Helson, H. (1964) Adaptation-Level Theory. An experimental

and systematic approach to behavior. Harper and Row.

New York.

Henderson, J.M. and Well, A.D. (1985) Symbolic comparisons

with and without perceptual referents: is internal informa-

tion used‘? Memory Cognit., 13: 176-182.

Link. St. (1YYO) Modelling imageless thought: the relative judgment theory of numerical comparison. J. Math. Psy-

chol.. 34: 2-41. In conclusion it can be said that the perfor-

mance data showed the expected symbolic dis- tance effect, although this was not expressed in P300 data. P300 amplitude changes support infor- mation extraction from number stimuli or/and the presence of an internal number scale where the individual number stimuli are arranged dur- ing encoding. P300 data also confirmed that in- formation extraction is strategic so as to get as much information about the subsequent re- sponses as possible.

Magliero, A., Bashore. T.R., Coles, M.G.H. and Donchin. E.

(1984) On the dependence of P300 latency on stimulus

evaluation processes. Psychophysiology, 21: I7 I - 186. McCarthy. G. and Donchin, E. (1981) A metric for thought: a

comparison of P300 latency and reaction time. Science, 21 I: 77-79.

Mecklinger, A., Ullsperger, P., M611e. M. and Grune, K. (1994) Event-related potentials indicate information ex-

traction in a comparative judgment task. Psychophysiology,

31: 23-28.

Acknowledgements

Mecklinger, A. and Ullsperger. P. (1993) P3 varies with stimu-

lus categorization rather than probability, Electroenceph.

Clin. Neurophysiol., 86: 395-407.

Moyer, R.S. (1973) Comparing objects in memory: evidence

suggesting an internal psychophysics. Percept. Psychophys.. 13: 1X0-184.

This research was supported by grant UL 111 /l-l of the Deutsche Forschungsgemeinschaft (DFG). We wish to thank W. Dehoff and B. Minuth for their support during data acquisition.

References

Moyer, R.S. and Dumais, S.T. (lY78) Mental Comparison.

Psychol. Learn. Motiv., 12: 117-155.

Moyer, R.S. and Landauer, T.K. (1967) Time required for

judgmentsof numerical inequality. Nature, 215: lSlY-1520.

Paivio, A. (1975) Perceptual comparisons through the mind’s

eye. Memory Cognit., 3: 635-647.

Potts, G.R. (lY74) Storing and retrieving information about ordered re!ationships. J. Exp. Psychol.. 103: 431-439.

Ragout, R. and Renault, B. (1985) P300 and S-R compatibil- ity: A reply to Magliero et al. Psychophysiology. 22: 34Y-

352. Banks, W.P. (1977) Encoding and processing of symbolic Ruchkin, D.S., Johnson Jr., R., Canoune, H.L., Ritter, W. and

information in comparative judgements. In: G.H. Bower Hammer, M. (1990) Multiple sources of P3b associated (Ed.), The Psychology of Learning and Motivation, Vol. with different types of information. Psychophysiology, 27: 11, Academic Press, New York. 157-176.

Banks, W.P., Clark, H.H. and Lucy, P. (1975) The locus of the semantic congruity effect in comparative judgments. J.

Exp. Psychol. Hum. Percept. Perform., 104: 35-47.

Sekuler, R. and Mierkiewicz. D. (1977) Children’s judgments of numerical inequality. Child Dev., 48: 630-633.

Ullsperger, P. and Baldeweg, T. (1992) Interpreting P300

56 K. Gnrne Ed al. /International Journal of Psychophysiology I7 (1994) 47-56

amplitude changes with adaptation level theory. Behav. Brain Sci., 14: 733-734.

Ullsperger. P. and Gille, H.-G. (1988) The late positive com-

ponent of the ERP and adaptation level theory. Biol.

Psychol., 26: 299-306. Vasey, M.W. and Thayer, J.F. (1989) The continuing problem

of false positives in repeated measures ANOVA in psy-

chophysiology: a multivariate solution. Psychophysiology,

24: 479-486. Yanai, K. and Nageishi, Y. (1993) Examination of adaptation

level theory by the help of P300 measurement. Int. J.

Psychophysiol., 14: 1.56.