General ability, defined as psychometric g, arises from the empirical fact that scores on various cognitive tests are positively correlated in the population. The g factor is highly stable across different factor analytic algorithms, across different test batteries and across different populations. Because all cognitive tests, from the simplest to the most complex, regardless of their informational content, are g-loaded to varying degrees, g cannot be described in terms of the tests' content, or even in psychological terms. It is actually a property of the brain. The loadings of various tests on g, from tests of sensory discrimination and reaction time to those of highly complex problem solving, predict those tests' degree of correlation with a number of non-psychometric variables: the test's heritability, inbreeding depression, coefficient of assortative mating, brain size, reaction time, brain nerve conduction velocity, brain glucose metabolic rate and features of brain evoked potentials. Although some of the brain's cognitive functions are modular, the g factor reflects the all-positive correlations among virtually all cognitive functions that show individual differences. I hypothesize that the brain contains no module for general problem solving. Correlations between individuals' performances in various cognitive tasks result from quantitative individual differences in physiological conditions that do not constitute the brain's modular and other neural design features but do influence their speed and efficiency of information processing.