Intelligence 30 (2002) 289 – 302

EEG alpha power and intelligence

M. Doppelmayr*, W. Klimesch, W. Stadler, D. Pöllhuber, C. Heine
Department of Physiological Psychology, Institute of Psychology, University of Salzburg,
Hellbrunnerstr. 34, A-5020 Salzburg, Austria
Received 14 July 2000; received in revised form 16 January 2001; accepted 1 March 2001

Abstract

The hypothesis was tested whether alpha power in different subbands (lower-1, lower-2, and upper
alpha) is selectively related to intelligence. For 74 subjects, the EEG was recorded during a resting
session. Two different intelligence tests (LGT-3 and IST-70) were performed. We found a strong
positive correlation between intelligence and alpha power. Large differences between the two
intelligence tests and alpha power subbands were obtained. Whereas the upper alpha band shows the
strongest correlation with the IST-70, power in the two lower alpha bands tend to show a more
consistent relationship with the LGT-3. Considering the fact that the IST-70 has a strong emphasis on
semantic memory demands, whereas the LGT-3 focuses more on the ability to learn new material, we
conclude that upper alpha is related to the ability to process semantic information, whereas the two
lower alpha bands are associated with attentional demands that dominate during the encoding of new
information. D 2002 Elsevier Science Inc. All rights reserved.

Keywords: Alpha; Upper alpha; EEG; ERD; Intelligence; Memory; IST-70; LGT-3

1. Introduction

The alpha power and cognitive performance are related to each other in a rather complex
way. Two factors appear to play an important role: (1) the distinction between tonic and
phasic (or event-related) alpha power, and (2) as far as a relationship with intelligence is

* Corresponding author. Tel.: +43-662-8044-5120/5135; fax: +43-662-8044-5126.

E-mail address: michael.doppelmayr@sbg.ac.at (M. Doppelmayr).

0160-2896/02/$ – see front matter D 2002 Elsevier Science Inc. All rights reserved.
PII: S 0 1 6 0 - 2 8 9 6 ( 0 1 ) 0 0 1 0 1 - 5
290 M. Doppelmayr et al. / Intelligence 30 (2002) 289–302

concerned, the ‘‘neural efficiency hypothesis’’ must be taken into account (cf. Neubauer,
Freudenthaler, & Pfurtscheller, 1995). In a recent review focusing on alpha oscillations and
memory performance, Klimesch (1999) has emphasized that a consistent interpretation of
the relationship between alpha power and memory performance is possible only if a
dissociation between tonic changes in absolute and phasic or event-related changes in
relative alpha power is taken into account. Phasic (or event-related) changes are task and/or
stimulus related and occur at a rapid rate, whereas tonic changes are not (or less) under
volitional control and occur at a much slower rate (e.g., over the life cycle and in response
to circadian rhythms, fatigue, distress, neurological disorders, etc.). With respect to memory
performance, it was found that good performance is reflected by an increase in tonic alpha
power but a decrease in phasic (event-related) alpha power. As an example, in a similar
way as cognitive performance does, tonic alpha power increases from early childhood to
adulthood but decreases during the late part of the lifespan. When measuring phasic
changes in alpha power by using a method proposed by Pfurtscheller and Aranibar (1977),
Klimesch, Doppelmayr, Pachinger, and Russegger (1997) and Klimesch, Doppelmayr,
Russegger, Pachinger, and Schwaiger (1998) have found that the extent of event-related
desynchronization (ERD) in the upper alpha band is correlated with semantic memory
performance. The interesting point here is that tonic power (which can be measured only
during a ‘‘resting state’’ when subjects do not perform any task) and phasic or event-related
power (which is measured as the percentage of a change in power with respect to a
‘‘reference’’ or ‘‘resting’’ state) are related to cognitive performance but in different ways.
A possible explanation for this effect comes from a study by Doppelmayr, Klimesch,
Pachinger, and Ripper (1998) who found that the extent of ERD in a test interval (where
subjects actually perform a task) and alpha power in a reference interval (that precedes the
test interval and where subjects are not performing a task) are related to each other: The
larger the alpha power in the reference interval, the larger is the ERD in the test interval.
Thus, the extent of desynchronization (decrease in alpha power in the test with respect to
the reference interval) depends (at least in part) on the power of the reference (or ‘‘resting’’)
interval. In a pure statistical sense, the extent of an event-related band power change may
very well be independent from absolute power as measured during a resting or reference
condition. It was suggested (see Klimesch, 1999 for an extensive review) that the de-
pendency of event related on absolute power indicates a special physiological mechanism,
which possibly operates to increase the signal/noise ratio.
On the basis of this interpretation, it can easily be understood why studies measuring tonic
(or absolute) power find a positive and studies measuring phasic (or event-related) alpha
power find a negative correlation (or association) with intelligence. As an example, in an
interesting study, Neubauer et al. (1995) have found that the extent of a phasic change in
alpha power (as measured by ERD) is smaller for intelligent as compared to less intelligent
subjects. They have interpreted their results in terms of the ‘‘neural efficiency hypothesis,’’
which assumes that intelligent subjects are more efficient or skilled in performing a task and
hence show a smaller phasic response or ERD. On the other hand, studies focusing on tonic
power report a positive correlation with intelligence and alpha power as measured in a resting
condition (e.g., Jausovec, 1996). This finding is consistent with the assumption that a high
 M. Doppelmayr et al. / Intelligence 30 (2002) 289–302 291

tonic level in alpha power is related to good performance, but it does not tell us anything
about the validity of the ‘‘neural efficiency hypothesis’’ because tonic power is not measured
during actual task performance.
If we proceed from the assumption that a large extent of desynchronization reflects
cognitive performance (Klimesch, 1996, 1999) and that the extent of alpha desynchronization
depends on (tonic) absolute alpha power, the prediction is that even in a resting state absolute
alpha power (reflecting the tonic power level) is larger for intelligent than less intelligent
subjects. Because previous studies have shown that the relationship between (tonic and
phasic) alpha power and memory performance holds true primarily for the upper alpha band
[a band of 2-Hz width lying above the individually determined alpha frequency (AF)], the
question of interest is whether the expected relationship with intelligence is also confined to
the upper AF range as well.
It is known that EEG power is influenced by a variety of unspecific factors such as the
thickness of the skull or the volume of cerebrospinal fluid. Thus, the measurement of alpha
power will be influenced by nonspecific factors that increase the error variance of the
expected relationship with intelligence. On the other hand, the measurement of intelligence
may also be biased by unspecific factors such as motivation, attention, or preknowledge of
subtests. Because we had no opportunity to perform MRI scans (to measure the thickness of
the skull and the volume of cerebrospinal fluid), we tried to reduce the error variance for the
measurement of intelligence by performing two different tests, the IST-70 of Amthauer
(1970) and the LGT-3 of Bäumler (1974), and by additionally excluding subjects with
inconsistent results (see Section 2 below).

2. Method

2.1. Subjects and selection procedure

For a sample of 89 subjects, all psychology students (26 males and 63 females), two
intelligence tests (IST-70 and LGT-3) were performed. All subjects were right handed. They
participated in the present study as part of introductory course requirements. Handedness was
controlled by asking the subjects about the hand they use in different tasks such as
handwriting, throwing a ball, etc. A subject was considered right handed if he/she indicated
to use the right hand for all of these different tasks.
The EEG was analyzed only for those subjects who had consistent results in both
intelligence tests. A result was considered consistent if the difference between the
z-transformed summary scores of the two tests did not exceed a z-value of 1.5. This selection
procedure led to the exclusion of 15 subjects (4 males and 11 females). It was based on the
assumption that inconsistent results are due to either an artificially bad performance in one
test because of a lack of interest, fatigue, or to a much better performance in one test because
of preknowledge of this respective test.
A final sample of 74 subjects (22 males and 52 females) remained for data analysis. Their
mean age was 22.7 years with a range of 18–34 years.
292 M. Doppelmayr et al. / Intelligence 30 (2002) 289–302

It should be noted that data analyses were also performed on the original sample of 89
subjects. As compared to the final sample, similar results were obtained, but the number of
significant correlations was smaller.

2.2. Material

Each subject had to perform two different intelligence tests, the IST-70 and the LGT-3. The
IST-70 comprises the following nine subtests: sentence completion (SENTENCE), detection
of incorrect words (WORD), finding analogies (ANALOGY), finding common superordinate
categories (COMMON), solving calculation tasks (CALCULATION), finding rules in
sequences of numbers (NUMBER), solving mental rotation tasks (ROTATION), selection
of figures according to geometric rules (FIGURE), and performing a cued recall memory task
(MEMORY). The LGT-3 requires subjects to learn a city map (MAP), Turkish words
(TURKISH), phone numbers (PHONE), facts about a construction plan (FACTS) and to
memorize objects (OBJECTS) and symbols (SYMBOLS). A more detailed overview of the
different tasks is given in Table 1.

Table 1
IST-70
SENTENCE: Which is the correct completion of the sentence? A rabbit is most similar to a. . .(cat, squirrel, bunny,
fox, hedgehog).
WORD: Detect the incorrect word: table, chair, bird, locker, bed.
ANALOGY: Find the correct analogy: page: book = word: (sentence, chapter, letter, contents, title).
COMMON: Find common superordinate: knife, butter, newspaper, bread, cigar, collar.
CALCULATION: More or less complex calculation tasks: one pencil costs 25 Pfg, how much do you have to pay
for three pencils?
NUMBER: Find rule in number sequence: 2, 4, 6, 8, 10, 12.
FIGURES: A figure has been cut apart and the question is to determine which out of five possible templates this
figure corresponds to.
ROTATION: A cube with different symbols on each side is depicted. The subject has to determine which out of
five possible cubes is the same as the original one.

LGT-3
MAP: A specific route has to be learned on an symbolic map. All in all, about 30 crossroads and directions have to
be learned. The correct route has to be drawn in a new map.
TURKISH: 20 word pairs have to be learned. In a later recognition task, the correct words have to be identified out
of five possible.
OBJECTS: 20 different objects (ball, hammer, scissors. . .) have to be learned and recalled in a later free recall task.
PHONE: 13 different three-digit telephone numbers and the corresponding persons have to be learned.
FACTS: A text, including different facts about a building (names of the architects, number of rooms, size, etc.),
has to be learned and some questions about these facts have to be answered later.
SYMBOLS: 20 different symbols are depicted, each within a specific frame. In a later recall task, the subject has
to choose the correct frame out of five possible.
 M. Doppelmayr et al. / Intelligence 30 (2002) 289–302 293

2.3. Design and procedure

Subjects were told that they participate in a study, which aims to find electrophysiological
correlates of cognitive performance. First, their resting EEG was recorded for a period of
3 min with eyes closed after they were told to completely relax. After the EEG recording, we
asked subjects to perform the IST-70 and the LGT-3. The sequence of administering the two
tests was counterbalanced between subjects.

2.4. Apparatus

EEG signals were amplified by a 32-channel biosignal amplifier system (frequency

response: 0.16–30 Hz), subjected to an anti-aliasing filterbank (cut-off frequency: 30 Hz,
110 dB/octave), and were then converted to a digital format via a 32-channel A/D converter.
Sampling rate was 128 Hz. During data acquisition, EEG signals were displayed online on a
high-resolution monitor and stored on disk.

2.5. Acquisition and processing of EEG data

A set of 11 silver electrodes attached with a glue paste (Nihon Kohden Elefix EEG paste)
to the scalp was used to record EEG signals. Electrode diameter was 1 cm. The electrodes

Table 2
Behavioral results: IQ (LGT-3 and IST-70) and standard scores for subtests
Female Male
Minimum Maximum S.D. Mean mean mean t (df = 72)
IQ LGT-3 67 138 17.93 107.58 109.01 104.93
IQ IST-70 90 136 11.54 110.63 108.48 116.59
MAP 28 74 9.73 52.76 51.75 55.93
TURKISH 20 76 10.41 53.27 54.70 49.85
FACTS 27 74 10.80 52.29 53.56 49.85
PHONE 27 76 12.33 57.36 58.43 55.19
OBJECTS 24 72 8.88 50.45 51.54 47.96
SYMBOLS 29 78 10.90 55.67 55.57 56.48
SENTENCE 81 122 7.94 103.65 102.92 106.00
WORD 65 121 14.17 91.93 92.59 91.04
ANALOGY 93 127 6.96 107.88 107.33 109.52
COMMON 86 131 9.04 104.48 105.16 103.67
MEMORY 86 124 7.88 110.57 110.29 111.67
CALCULATION 85 131 10.11 105.11 102.97 110.33 2.48 *
NUMBER 87 127 10.39 108.71 106.98 112.96
FIGURE 80 123 9.58 105.13 103.95 108.41
ROTATION 81 129 12.37 106.02 103.90 111.59 2.29 *
Significant differences at or beyond the 5% level of significance between male and female results are indicated
by an asterisk.
294 M. Doppelmayr et al. / Intelligence 30 (2002) 289–302

were placed according to the International Electrode (10–20) Placement system at F3, F4, Fz,
C3, C4, Cz, P3, P4, Pz, O1, and O2. In addition to these 11 electrodes, two earlobe electrodes
(termed A1 and A2) were attached to the left and right ear. Furthermore, the electrooculogram
(EOG) was recorded from two pairs of leads in order to register horizontal and vertical eye
movements. Data were recorded against a common reference placed on the nose and off-line
algebraically re-referenced to linked earlobes.
The EEG recording of 3 min was arbitrarily segmented in 45 epochs with a length of 4 s
each. All of the 45 epochs were carefully checked individually for artifacts (eye blinks,
horizontal and vertical eye movements, muscle artifacts, etc.) by visual inspection. Because
low EEG frequencies may very easily get contaminated by eye movements, a very strict
criterion for rejecting eye movement artifacts was applied: Epochs that were associated even
with small changes in the horizontal or vertical EOG channel were rejected. After rejecting
artifacts, an average of 33 epochs remained.
For artifact free epochs, power spectra (within a frequency range of 0.5–35 Hz) were
calculated and then averaged for each subject and lead. Power estimates were obtained for
frequency steps of 0.25 Hz. These power estimates were averaged within the frequency limits
of three alpha subbands, which were obtained by applying a procedure used in a similar way
in a series of earlier studies (cf. Klimesch (1996)). First, mean AF was calculated for all leads
and subjects. Averaged over all subjects, it was 9.91 Hz. Then, by using AF as cut-off point,
we distinguished between an upper alpha band (AF–AF + 2 Hz) and two lower alpha bands,
the lower-2 alpha band (AF–AF 2 Hz), and the lower-1 alpha band (AF 2–AF 4 Hz).
AF was rounded to 10 Hz.
To obtain Gaussian distributed data, the log power values for each lead, subject, and the
three alpha bands were calculated. These log-transformed power values were correlated with
behavioral data (cf. Table 2 for an overview).

3. Results

3.1. Behavioral results

The results of the two intelligence tests are summarized in Table 2, displaying the values
for the whole sample as well as separately for men and women. In both tests, the IST-70 and
the LGT-3, subjects reached IQs of comparable magnitude. t tests have been performed to
compare male and female results, showing significant differences with higher IQ values for
males only for CALCULATION (t = 2.48, df = 72, P < .05) and ROTATION (t = 2.29,
df = 72, P < .05).

3.2. EEG results

Examples of power spectra are shown in Fig. 1. They reveal large differences between the
group of high vs. the group of lower test performance. Less intelligent subjects have smaller
alpha power than more intelligent subjects.
 M. Doppelmayr et al. / Intelligence 30 (2002) 289–302 295

Fig. 1. Absolute alpha power (during eyes closed) for subjects falling above (high IQ) and below (low IQ) the
grand average IQ of the sample of 74 subjects. The data are shown for two recording sites and reveal large
differences between the two subgroups. Within the (extended) AF range of 6 – 12 Hz, the high IQ group shows
larger absolute power than the low IQ group.

Fig. 2. Correlations between IQ (LGT-3 and IST-70) and log alpha power in three different subbands. The three
columns in each circle (symbolizing the head, seen from above) represent correlation coefficients with the lower-1,
lower-2, and upper alpha band (grey column). The level of significance is indicated by one asterisk (1%) or two
asterisks (0.1%). From upper left to lower right, the following recording sites are depicted: F3, Fz, F4, C3, Cz, C4,
P3, Pz, P4, O1, and O2. The left ‘‘head’’ shows correlations with the LGT-3 and the right those with the IST-70.
Note that the only significant correlations between alpha and the IST-70 can be observed in the high frequency
range of the upper alpha (10 – 12 Hz).
296 M. Doppelmayr et al. / Intelligence 30 (2002) 289–302
 M. Doppelmayr et al. / Intelligence 30 (2002) 289–302 297

The correlations of the mean standard scores (IQ) of the two intelligence tests with the log
absolute alpha power in the three subbands are shown in Fig. 2. Significant correlations at or
beyond the 1% level were found in 41 of the 66 possible cases (2 tests  3 subbands 
11 leads). If we consider correlations with the upper alpha band, a highly selective pattern of
results emerges. Now, a striking difference between the two intelligence tests can be
observed. Whereas the LGT-3 shows significant correlations with all three subbands (the
only exception is the lower-1 alpha at occipital leads), only the upper alpha band exhibits
correlations with the IST-70.
Significant correlations between the three alpha bands and the 15 subtests (6 for the LGT-3
and 9 for the IST-70) are shown in Fig. 3. At the 1% level, a large number of significant
correlations with subtests can be observed. Frontal, central, and parietal sites exhibit the
largest number of significant correlations.
For all but occipital sites, the most consistent correlations with respect to all three subbands
were found for TURKISH (of the LGT-3), CALCULATION, and FIGURE (of the IST-70).
Those subtests with an inconsistent pattern of correlations with subbands are MAP, PHONE,
and OBJECTS (of the LGT-3) and MEMORY and ROTATION (of the IST-70). An
intermediate position takes subtest COMMON. Two subtests of the LGT-3 (FACTS and
SYMBOLS) and four of the IST-70 (SENTENCE, WORD, ANALOGY, and NUMBER) do
not show significant correlations (at or beyond the 1% level) at all.
The most selective finding is the association of subtest MEMORY and ROTATION with
the upper alpha band. Without a single exception, these subtests correlate with the upper
alpha band only. For MEMORY, these correlations are significant at or beyond the 1% level
for all sites. For ROTATION, significant correlations were found for left frontal, all midline,
and three of the four right hemispheric sites.
As pointed out for example by Vogel (2000), males and females show differences in the
amount of high or low amplitude EEG in several frequency bands. Therefore, additionally,
analyses have been performed separately for men and women. The results show striking
differences with respect to the correlation of log power in the different frequency bands and
test scores. The number of significant correlations at or beyond the 1% are shown in Table 3
for the total sample of 74, for the female sample including 52, and for the male sample with
22 subjects. Additionally, the difference between male and female results is displayed.
While for women in the lower-1 alpha (including all 11 electrode sites, subtests and the
total scores) there are 25, in the lower-2 alpha 25, and in the upper alpha 69 significant
results, men show only 9, 3, and 5 significant correlations, respectively. In the lower-1 alpha,
the strongest sex-related differences can be seen for MAP, TURKISH, and CALCULATION,
in the lower-2 alpha for TURKISH, and in the upper alpha for the LGT-3 total score,

Fig. 3. Correlations between subtests of the LGT-3 and IST-70 and log alpha power in three different subbands.
The three columns in each circle (symbolizing the head, seen from above) represent correlation coefficients with
the lower-1, lower-2, and upper alpha band (grey column). The level of significance is indicated by one asterisk
(1%) or two asterisks (0.1%). From upper left to lower right, the following recording sites are depicted: F3, Fz, F4,
C3, Cz, C4, P3, Pz, P4, O1, and O2. Note that IST-70 subtests show significant correlations primarily for the upper
alpha band, whereas the opposite holds true for the LGT-3 subtests.
298 M. Doppelmayr et al. / Intelligence 30 (2002) 289–302

Table 3
Comparing the number of significant (at or beyond 1%) correlations between males and females

SENTENCE
IST-70 total
LGT-3 total

SYMBOLS
TURKISH

OBJECTS
PHONE

WORD
FACTS
MAP
All subjects (n = 74)
Alpha total 11 9 8 11 1 8 2
Lower-1 Alpha 9 8 11
Lower-2 Alpha 11 7 11 3
Upper Alpha 11 1 1 9 6

Females (n = 52)
Alpha total 8 9 11 1
Lower-1 Alpha 1 1
Lower-2 Alpha 1 1 1
Upper Alpha 9 1 2 11 9 6

Males (n = 22)
Alpha total 1 1
Lower-1 Alpha 9
Lower-2 Alpha 2 1
Upper Alpha

Difference (males females)

Alpha total 7 9 1 11 1
Lower-1 Alpha 1 9 1
Lower-2 Alpha 2 1 1 1 1
Upper Alpha 9 1 2 11 9 6

TURKISH, PHONE, and COMMON. In all cases, except the lower-1 alpha MAP, females
show more significant correlations. These big differences might be largely due to the
difference in the number of subjects. While there are 52 female participants, there are only
22 male subjects. Using a group of arbitrarily selected 22 females leads to similar results as
the male group, especially with respect to the number of significant correlations. The exact
results depend on the respective selection of subjects.
Considering the fact that for the IST total score there are very high correlation coefficients,
it seems somewhat puzzling that most of the single subtests revealed only weaker
correlations. This could possibly be due to the differential g-loadings of the subtests. In a
further analysis, we tested whether those subtests with higher g-loadings generally display a
higher correlation with intelligence (as measured with the IST or the LGT-3, respectively).
First, we computed the g-loadings of the IST and the LGT-3 subtests (loadings on the first
unrotated factor). Then, we correlated these g-values with the correlations of the respective
subtest with the total score of the IST-70 or the LGT-3. While these correlations showed a
 M. Doppelmayr et al. / Intelligence 30 (2002) 289–302 299

Subtest+total scores
Sum of all subtests
CALCULATION

LGT-3 subtests

IST-70 subtest
ROTATION
ANALOGY

NUMBERS
COMMON

MEMORY

FIGURE
1 5 1 9 7 30 23 53 73
3 9 4 19 16 35 44
2 9 6 21 17 38 49
5 11 7 4 6 25 33 58 79

4 1 4 12 18 30 47
11 3 1 14 24 25
1 1 8 1 1 1 1 15 25 25
9 2 6 4 1 28 22 50 69

1 1 2
9 9 9
1 1 3
5 5 5 5

4 1 4 11 18 29 45
11 3 1 14 15 16
1 1 8 1 1 1 9 15 24 22
9 3 6 4 1 28 17 45 64

highly significant result for the IST (r=.799, P < .01, df = 8), the result for the LGT-3 failed to
reach significance. This indicates that only for the IST those subtests exhibiting a higher
g-loading correlate higher with the total score than those with smaller g-loading. Similar
analyses have been performed to test if there is a relation between task complexity of a subtest
(defined by the number of correctly solved items) and its correlation with intelligence. The
results again show a significant correlation for the IST-70 (r=.732, P < .05, df = 8) and no
significant effects for the LGT-3.

4. Discussion

Although all subjects were psychology students, the broad range of IQ scores, as depicted
in Table 2, shows that these results are not restricted to a sample of very high intelligent
subjects but to a very broad population.
300 M. Doppelmayr et al. / Intelligence 30 (2002) 289–302

The findings of the present study show a strong positive correlation between intelligence
and alpha power. Most interestingly, we found consistent differences between the type of
intelligence test and alpha power in the three subbands. Whereas power in the upper alpha
band shows the most consistent and strongest correlation with the IST-70, power in the two
lower alpha bands tend to show a more consistent relationship with the LGT-3.
These findings support the results from a series of earlier performed experiments, which
revealed that the upper alpha band is related to semantic memory, whereas the two lower
alpha bands are associated with attentional demands (Klimesch, Doppelmayr, Pachinger, &
Ripper, 1997; Klimesch, Doppelmayr, Pachinger, et al., 1997; Klimesch, Doppelmayr, et al.,
1998; Klimesch, Russegger, Doppelmayr, & Pachinger, 1998; Klimesch, Schimke, &
Schwaiger, 1994; Klimesch, Vogt, & Doppelmayr, 2000; Vogt, Klimesch, & Doppelmayr,
1998). Comparing the subtests of the IST-70 and LGT-3 reveals that the former test has a
strong emphasis on semantic memory demands, whereas the latter focuses more on the ability
to learn new material. The emphasis of the IST-70 on semantic memory becomes clear when
considering the subtests, which are sentence completion (SENTENCE), detection of incorrect
words (WORD), finding analogies (ANALOGY), finding common superordinate categories
(COMMON), solving calculation tasks (CALCULATION), finding rules in sequences of
numbers (NUMBER), solving mental rotation tasks (ROTATION), selection of figures
according to geometric rules (FIGURE), and performing a cued recall memory task
(MEMORY). All of these subtests represent either verbal-semantic abilities (SENTENCE,
WORD, ANALOGY, and COMMON), verbal-semantic memory (MEMORY), or knowledge
about more formal and spatial semantic relationships (CALCULATION, NUMBER, ROTA-
TION, and FIGURE). Looking at the subtests of the LGT-3, on the other hand, demonstrates
that in all subtest new information has to be acquired during the test: learning a city map
(MAP), Turkish words (TURKISH), phone numbers (PHONE), facts about a construction
plan (FACTS), objects (OBJECTS), and symbols (SYMBOLS).
It has to be pointed out that some of the findings, especially concerning the subtests of the
IST-70, might be influenced by either differential g-loadings or task complexity (for a detailed
analysis of the IST-70, see Brocke, Beauducel, & Tasche, 1998). However, the importance of
those influences on the relationship between EEG alpha power and intelligence has to be
investigated in future research in more detail.
Given the long history of studies trying to find reliable EEG indicators of intelligence, the
question arises why so many studies report negative or inconsistent findings (cf. the reviews
by Gasser, Von Lucadou-Müller, Verleger, & Bächer, 1983; Mundy-Castle, 1958; Vogel &
Broverman, 1964, 1966 and the critical comments by Ellingson, 1966). We assume that the
following two reasons play a critical role. (1) None of the studies reported in the literature use
narrow frequency bands, which are adjusted to the actual AF of the studied subjects (cf.
Marosi et al., 1999 for a recent example). (2) Most studies did not take into account the
opposite effects of tonic and phasic alpha power (see Section 1). Studies, which use measures
that are less prone to biases from these two factors, report results that are similar to our
findings. The interesting studies by Anokhin, Lutzenberger, and Birbaumer (1999) and
Lutzenberger, Birbaumer, Flor, Rockstroh, and Elbert (1992) provide good examples to
highlight this point. These authors use a measure for the dimensional complexity of the EEG
 M. Doppelmayr et al. / Intelligence 30 (2002) 289–302 301

(DCx), which is thought to reflect the number of independently oscillating competitive neural
networks. If we assume that this measure is associated (at least to some degree) with the
extent of desynchronization, we may expect that strong synchronization (large alpha power)
should be positively related with low DCx. Because we found that large synchronization
(large alpha) is positively related to intelligence, we would assume that intelligence is
negatively related to DCx (more intelligent subjects show less dimensional complexity). The
findings of Anokhin et al. (1999) are in support with this interpretation. However, this
negative relationship might change under very difficult task conditions, where a more
distributed pattern of neuronal activation is necessary to obtain a high level of performance
(cf. Lutzenberger et al., 1992). Thus, if we assume that low DCx can be observed under
conditions, which Klimesch (1999) describes as ‘‘tonic’’ conditions, whereas high DCx can
be observed under conditions, which are related to ‘‘phasic or event-related’’ conditions, we
would expect that the relationship with intelligence might change as is the case for the
relationship between intelligence, tonic, and event-related alpha power. Large absolute but
small event-related alpha power is related to intelligence.
The finding that the faster frequency component of alpha is more closely related to
intelligence (as measured by the IST-70 as compared to the ‘‘learning test,’’ i.e., the LGT3)
than are slower frequency components is in good agreement with results from other research
groups. As an example, in an interesting study by Anokhin and Vogel (1996), it was found
that intelligence scores are positively related with AF. These findings indicate that mean and/
or peak AF of more intelligent subjects tend to fall within the frequency range of the upper
alpha band. Similar results have been reviewed by Oswald and Roth (1974) and have also
been obtained by Ellingson (1973) and Giannitrapani (1985).
Finally, it should be noted that there is good evidence that individual variations in the EEG
of healthy subjects (including power in different alpha bands) is determined almost exclusively
by genetic differences (for an extensive review, see Vogel, 2000). This evidence—together
with our findings—suggests that interindividual variations in intelligence are also significantly
determined by genetic factors.

Acknowledgments

This research was supported by the Austrian ‘‘Fonds zur Förderung der wissenschaftlichen
Forschung,’’ P-13047.

References

Amthauer, R. (1970). Intelligenz-Struktur-Test (I-S-T-70). Göttingen: Hogrefe.

Anokhin, A., & Vogel, F. (1996). EEG alpha rhythm frequency and intelligence in normal adults. Intelligence, 23,
1 – 14.
Anokhin, A. P., Lutzenberger, W., & Birbaumer, N. (1999). Spatiotemporal organization of brain dynamics and
intelligence: an EEG study in adolescents. International Journal of Psychophysiology, 33, 259 – 273.
Bäumler, G. (1974). Lern-und Gedächtnistest—LGT-3. Göttingen: Hogrefe.
302 M. Doppelmayr et al. / Intelligence 30 (2002) 289–302

Brocke, B., Beauducel, A., & Tasche, K. (1998). Der Intelligenz-Struktur-Test: analysen zur theoretischen Grund-
lage und technischen Güte. Diagnostica, 44 (2), 84 – 99.
Doppelmayr, M., Klimesch, W., Pachinger, Th., & Ripper, B. (1998). The functional significance of absolute
power with respect to event-related desynchronization. Brain Topography, 11 (2), 133 – 140.
Ellingson, R. J. (1966). Relationship between EEG and Test Intelligence: a commentary. Psychological Bulletin,
65 (2), 91 – 98.
Ellingson, R. J. (1973). Intelligence and frequency of the alpha rhythm. American Journal of Mental Deficiency,
78 (3), 334 – 338.
Gasser, Th., Von Lucadou-Müller, I., Verleger, R., & Bächer, P. (1983). Correlating EEG and IQ: a new look at an
old problem using computerized EEG parameters. Electroencephalography and Clinical Neurophysiology, 55,
493 – 504.
Giannitrapani, D. (1985). The electrophysiology of intellectual functions. New York: Karger.
Jausovec, N. (1996). Differences in EEG alpha activity related to giftedness. Intelligence, 23, 159 – 173.
Klimesch, W. (1996). Memory processes, brain oscillations and EEG synchronization. International Journal of
Psychophysiology, 24, 61 – 100.
Klimesch, W. (1999). EEG alpha and theta oscillations reflect cognitive and memory performance: a review and
analysis. Brain Research Reviews, 29, 169 – 195.
Klimesch, W., Doppelmayr, M., Pachinger, Th., & Ripper, B. (1997). Brain oscillations and human memory
performance: EEG correlates in the upper alpha and theta bands. Neuroscience Letters, 238, 9 – 12.
Klimesch, W., Doppelmayr, M., Pachinger, Th., & Russegger, H. (1997). Event-related desynchronization in the
alpha band and the processing of semantic information. Cognitive Brain Research, 6 (2), 83 – 94.
Klimesch, W., Doppelmayr, M., Russegger, H., Pachinger, Th., & Schwaiger, J. (1998). Induced alpha band power
changes in the human EEG and attention. Neuroscience Letters, 244, 73 – 76.
Klimesch, W., Russegger, H., Doppelmayr, M., & Pachinger, Th. (1998). Induced and evoked band power
changes in an oddball task. Electroencephalography and Clinical Neurophysiology, 108 (2), 123 – 130.
Klimesch, W., Schimke, H., & Schwaiger, J. (1994). Episodic and semantic memory: an analysis in the EEG-theta
and alpha band. Electroencephalography and Clinical Neurophysiology, 91, 428 – 441.
Klimesch, W., Vogt, F., & Doppelmayr, M. (2000). Interindividual differences in alpha and theta power reflect
memory performance. Intelligence, 27 (4), 347 – 362.
Lutzenberger, W., Birbaumer, N., Flor, H., Rockstroh, B., & Elbert, Th. (1992). Dimensional analysis of the
human EEG and intelligence. Neuroscience Letters, 143, 10 – 14.
Marosi, E., Rodriguez, H., Harmony, Th., Yanez, G., Rodriguez, M., Bernal, J., Fernandez, Th., Silva, J., Reyes,
A., & Guerrero, V. (1999). Broad band spectral EEG parameters correlated with different IQ measurements.
International Journal of Neuroscience, 97, 17 – 27.
Mundy-Castle, A. C. (1958). Electrophysiological correlates of intelligence. Journal of Personality, 26, 184 – 199.
Neubauer, A., Freudenthaler, H., & Pfurtscheller, G. (1995). Intelligence and spatiotemporal patterns of event-
related desynchronization (ERD). Intelligence, 20, 249 – 266.
Oswald, W. F., & Roth, E. (1974). Zusammenhänge zwischen EEG und Intelligenzvariablen. Psychologische
Beiträge, 16, 1 – 47.
Pfurtscheller, G., & Aranibar, A. (1977). Event-related cortical synchronization detected by power measurements
of scalp EEG. Electroencephalography and Clinical Neurophysiology, 42, 817 – 826.
Vogel, F. (2000). Genetics and the electroencephalogram. Berlin: Springer.
Vogel, W., & Broverman, R. M. (1964). Relationship between EEG and Test Intelligence: a critical review.
Psychological Bulletin, 62 (2), 132 – 144.
Vogel, W., & Broverman, R. M. (1966). A reply to ‘‘relationship’’ between EEG and Test Intelligence: a
commentary. Psychological Bulletin, 65 (2), 99 – 109.
Vogt, V., Klimesch, W., & Doppelmayr, M. (1998). High frequency components in the alpha band and memory
performance. Journal of Clinical Neurophysiology, 15 (2), 167 – 172.