Actually, WMI and PSI are both more likely to be lower than VCI/VSI/FRI in high cognitive learners, rather than the reverse. It's not only PSI that tends to be lower.
However, there are other factors that might allow WMI to be higher than PSI. To your question about low VSI/PSI, one factor appears to be physical development. For example, if a very young child happens to have the spatial reasoning of a much older child, but the manual dexterity of their chronological age, having to mediate visual spatial thinking through a fine-motor-involved task may confound manual dexterity with visual spatial thinking, creating an artifactually lower estimate of their visual spatial ability than they actually have. This would affect at least one of the VSI subtests, and both of the PSI subtests, but not the WMI subtests nearly as much (hypothetically, maybe a little if the child is young enough that speech articulation could be a factor, or if they are sufficiently impaired that they cannot manage pointing).
The FRI, WMI and VCI subtests can all be completed entirely verbally (verbal responses can be accepted on the pointing items), so motor dexterity does not have to be a factor.
Another factor, separately from and in addition to manual dexterity, is speed. Both VSI and all PSI subtests are timed, which means that a slow, deep or extremely thorough thinker may have discrepancies between their absolute accuracy and their normative scores. For some, it may be that they really need more processing time, and for others, it may be stylistic. (I once had a student complete the FRI subtests by identifying the correct response within the first few seconds, but then systematically going through all of the multiple choice options testing and disproving them. It took 45 minutes to complete a subtest that is normally done in about 10 or less. But they got them all right.)
Only one FRI and one ancillary WMI subtest are timed, and all VCI subtests are untimed, so speed has less of an impact on these indices.
FRI would be one of the less likely indices to be a relative weakness in GT learners, since it's probably the best measure of higher-level thinking, and is amenable to both verbal and nonverbal strategies.
