Testing three alpha measures against two reasoning scores
The paper asks whether resting-state alpha activity tracks two separable reasoning abilities rather than a single composite IQ. It used the Cattell Verbal IIIB for verbal reasoning and the Cattell Culture Fair, Scale 3 for perceptual reasoning, converting all scores to the modern SD15 scale so the two tests could be compared.
Three EEG variables were tested: individual alpha peak frequency (IAPF), the frequency within the alpha band at which a person's alpha activity peaks; relative alpha power, defined as the percentage of 8–12 Hz activity relative to the full EEG spectrum averaged across electrode sites; and alpha coherence in the high alpha band (10–12 Hz) between Fp1–P3 and Fp2–P4. Coherence here measures how synchronized the signal is between two electrode sites, which the study treats as a candidate index of communication between frontal and parietal regions.
The design was correlational and regression-based, asking both whether the EEG variables co-vary with test scores and whether they predict them. Recordings were 19-channel quantitative EEG taken with eyes closed and eyes open, five minutes each.
Why a high-IQ sample changes the question
The study deliberately recruited from low-average to exceptionally gifted ability, but recruitment in the average range proved difficult. The result was a sample of 26 participants aged 21 to 82, more than half with IQs at or above the 98th percentile, and the paper states that findings cannot be generalised to the wider population.
That skew matters for how the null result should be read. The paper cites prior work suggesting psychometric variables correlate less well in high-ability populations and that highly speeded tests can introduce variance unrelated to the construct being measured. The Cattell tests are tightly timed, and on the separate visual continuous performance task most participants made few or zero errors and responded faster than age norms—a ceiling effect the author reads as evidence that speed carries little information at this ability level.
So the study is not a clean test of alpha–reasoning links across the full ability range. It is a test conducted mostly where the psychometric and possibly neural structure of intelligence may be most differentiated.
No significant alpha–reasoning associations emerged
No significant relationship was found between IAPF and either Cattell test. All IAPF values fell within the expected 7.5–12.5 Hz range, although two participants showed a low-amplitude fast pattern with no discernible alpha peak. The paper notes this fails to replicate earlier reports of a positive IAPF–IQ relationship.
Relative alpha power was likewise unrelated to either test. The analysis combined the alpha1 (8–10 Hz) and alpha2 (10–12 Hz) bands, and the author flags that a finer-grained band split might have behaved differently.
Alpha coherence between Fp1–P3 and Fp2–P4 also showed no significant association with verbal or perceptual reasoning. The study examined only these two electrode pairings out of 171 possible ones, a restriction the author identifies as a limitation rather than a tested equivalence.
Because these are null results from a small, skewed sample, they do not establish that the measures are unrelated. They establish that this study did not detect the relationships earlier work reported.
Arousal control did not rescue the correlations
One plausible confound is physiological arousal, which could shift alpha activity independently of reasoning ability. The study recorded heart rate variability and used it as a proxy for trait anxiety, then partialled it out of the IAPF–IQ correlations.
The effect was negligible, and heart rate variability scores sat within expected ranges for fit, healthy adults. The author interprets this as evidence that arousal levels were fairly similar across the group, so arousal is unlikely to explain the absence of an IAPF–reasoning link here.
This check strengthens the internal reading of the null result without changing its scope: it addresses one alternative explanation within this sample, not the sample's ability skew.
Recorded data left outside the analysis
The EEG session also produced event-related potentials during a modified Go-NoGo visual continuous performance task, in two ten-minute blocks. Those ERP data were not analysed in this study; only the task's behavioural data—omission and commission errors, reaction times, and reaction time variability—were examined.
The author suggests the unanalysed ERPs may relate more strongly to IQ scores than resting EEG alone, and that other recorded variables were too numerous and complex to include. This is a statement about untested data, not a finding, and it should not be read as evidence that ERPs would show the missing correlations.
The coherence analysis was similarly narrow: two frontal-parietal pairings in the high alpha band. The paper raises the possibility that short-range and long-range networks show different activation patterns, which a two-pair design cannot address.
The author's specialisation hypothesis remains untested
The paper offers an interpretation rather than a demonstrated mechanism: just as psychometric g weakens in high-ability populations and performance comes to rely on more specialised abilities, a hypothetical neural counterpart to g might also correlate less well with intelligence at the high end, with high-performing brains becoming more neurologically specialised.
This is presented as a proposal to explain why null findings in a high-IQ-heavy sample need not contradict earlier EEG and imaging work. The study did not test it directly—there is no comparison of correlation strength across ability levels in the reported results.
The author also argues that intelligence may map better onto two dimensions, such as fluid and crystallised ability, than onto one, which is the rationale for testing verbal and perceptual reasoning separately rather than a composite IQ. That rationale motivates the design but is not itself a result of this study.
What a decisive follow-up would need
The paper's own priority is methodological: a larger sample, a design able to compare ability levels rather than correlate within a skewed group, and a cognitive battery with a low speed loading and norms extending into the high range. The author notes that group-comparison designs may find significant effects more easily with small numbers than the correlational and regression approach used here.
A second priority is analytic breadth. The unanalysed ERP data, finer alpha band divisions, and the remaining electrode pairings are all identified as open questions rather than settled negatives. The paper also argues that first- and second-order CHC factors have received little neuroscientific attention relative to composite g.
The practical upshot for readers is a threshold for interpretation: this study's null results should not be treated as contradicting prior EEG–IQ findings, because its sample sits mostly at the high-ability extreme where the paper argues correlations may genuinely weaken. The most informative next step is not another small high-IQ correlation but a design that can compare correlation strength across the ability range while measuring reasoning factors separately.
Sources used in this answer
EEG Correlates of Verbal and Perceptual Reasoning
No significant relationships were found between individual alpha peak frequency, relative alpha power, or frontal-parietal high-alpha coherence and verbal or perceptual reasoning scores in a small sample skewed toward high IQ.