of this study is that the G factor scores derived from the rat tests were correlated with brain weight (r = +48) to a degree not unlike that found between g and brain size in humans.
A third study, by Francis Crinella and Jen Yu in 1995, used as genetically inbred strain of 120 adult Sprague-Dawley laboratory rats, in which there are hardly any natural individual differences. Individual differences had to be minimal as the aim of the experiment was to discover how much each specific region of the brain was involved in each of the various measures. This was done surgically by creating small lesions in 48 specific brain sites selected on the basis of the findings on the rat brain-behavioral correlates found in previous research by Crinella and Yu.  In the lesioned group, only one of the 48 sites was lesioned in easch pair of rats. A control group of 24 rats was given the same surgical procedure, but without creating a lesion at any site. After full recovery from the operation, all 120 rats were tested in each of seven diverse laboratory tests of learning and problem solving (reasoning). The performance measures from this battery were all positively correlated and yielded a g factor that accounted for 34 per cent of the total variance in the seven tests. Probably the most important finding is the every high correlation between the various tasks' G loadings and the number of brain structure that are significantly involved in task performance --a rank-order correlation of +.9.1. For example, one of the most highly G-loaded (.81) tasks (a detour problem, which requires reasoning) is significantly influenced by each of 17 brain structures, whereas a relatively simple conditioned avoidance tasks, with a lowest G, loading of only .08, significantly involved only four brain structures. The unlesioned control group performed better on each of the tasks than did the lesioned group. The vector of standardized mean differences between the unlesioned and lesioned (UL) groups on each of the seven tasks had a significant rank-order correlation of +.75 with the vector of G loadings for each of the tasks. The G factor correlated -.45 with the presence of any brain lesion--a higher correlation than was found for any single test. In brief, there was a strong relation between a task's G loading and its degree of sensitivity to the effects of brain damage in general. The authors suggest that "where the investigator is interested in detecting presence of any type of neuropsychological deficit, as opposed to damage that only affects a particular cognitive/neural system, g would bet the most sensitive measure" (p. 243). The results of this study lend support to the theory that tests with higher g loadings involve proportionately more neural processes than tests with lower g loading, even when studied in nonhuman animals.