Although the Hendricksons have proposed that the string length measure
of human auditory evoked potential complexity is positively related t
o psychometric intelligence, this relationship is variable and may eve
n in some circumstances be reversed. Previous studies relating string
length to IQ are reviewed and it is proposed that inconsistencies betw
een these reports reflect procedural differences in task-attention req
uirements or the absence thereof. In order to test this hypothesis, 21
subjects participated in two string length conditions: one in which a
ttention was not required and one in which subjects were required to c
ount oddball stimuli. String length was shown to be the product of an
interaction between intelligence and attention such that high-IQ subje
cts recorded lower brain evoked potential string lengths in the attend
condition, whereas low-IQ subjects showed maximal string lengths unde
r attended conditions. This result is compatible with both the D.E. He
ndrickson and Hendrickson (1980) and Bates and Eysenck (1993) findings
of positive and negative IQ-string correlations, respectively, and su
pports the hypothesis that string length indexes efficiency and capaci
ty under attended and unattended conditions, respectively. The string
length correlation with IQ is thus suggested to result from opposite a
ttention-induced changes in string length in high- and low-IQ subjects
. This was supported by a correlational analysis of the difference in
string length between high and low attention conditions revealing a wi
de spread high correlation maximal at frontal sites, where the differe
nce measure accounted for over half the variance in intelligence test
scores (uncorrected R(2) = 0.53). It is suggested that the string diff
erence measure is a more reliable estimate of intelligence than raw st
ring length in either attended or unattended conditions. The neural-ef
ficiency model of intelligence is discussed in light of these experime
ntal findings.