To the Editor: In his letter, Dr. Gold puts forward five points from which he concludes that the "appearance of stability" in IQ in our subjects is actually "likely evidence of intellectual decline." We discussed three of Dr. Gold’s points in our original paper; the remaining two are matters of test artifact.

• Dr. Gold points out that we included six to seven subjects with IQs less than 75 and raises the issue of comparing WISC and WAIS scores among subjects termed "intellectually limited." The cutoff for "intellectually limited" is usually a full-scale IQ lower than 70. We have, therefore, reexamined our data in subjects whose full-scale IQs were lower than 70. During the follow-up period, 50% of these subjects showed improved IQ, and 50% showed a decline. There were no significant differences between IQs at time 1 and time 2 in the subjects whose full-scale IQs were lower than 70, a finding similar to the findings for the group overall. Thus, there is no substance to this point.

• Dr. Gold and Dr. David raise the issue of the short form of the WAIS-R. There are considerable available data concerning the reliability and validity of the five-subtest short form of the WAIS-R; these are discussed in our original paper. Advantages of the short form are that it 1) omits several of the subtests (e.g., information and arithmetic) that assess educational attainment and 2) avoids tests (e.g., those requiring the subjects to accurately name five prime ministers of Great Britain since World War II) that would be affected by a long period of serious mental illness. In using this form, we also considered issues of test fatigue in patients who, in general, remain chronically unwell R5615511CHDBEJBA.

  Dr. Gold also refers to data he has collected on adults with schizophrenia. His results confirm that equivalent IQs are derived 1) when 10 subtests of the WAIS-R and 2) just five subtests are administered; however, the former produces a lower mean subtest score. Although the difference between the means of the 10 and the five subtest scores reaches statistical significance, in practice such a difference would never be considered evidence of cognitive decline. The IQ literature suggests that a scaled score should be at least 3 points above or below the mean before it can be considered noteworthy R5615511CHDCHAFF. Dr. Gold presents scores that are approximately one-third of a scaled score from the mean.

• Stevens et al. state that in "the majority of adults with schizophrenia . . . the premorbid intelligence and mental statuses are within normal limits" and cite a 1991 study by Frith et al. They highlight the fact that for our group, the mean premorbid IQ is 82.4 with at least 10 subjects having IQ measures lower than 80. Closer examination of the Frith et al. study demonstrates that premorbid IQ was "estimated" retrospectively, after illness onset, by employing a word reading test—the National Adult Reading Test. This test can provide both overestimates and underestimates of IQ R5615511CHDCDFEFR5615511CHDCCEEB. Furthermore, Nelson R5615511CHDCHBGD suggests that there are difficulties in using the National Adult Reading Test as a measure of IQ when IQ falls close to the test’s basal and ceiling levels. The basal IQ that can be estimated by the National Adult Reading Test is 84, i.e., to make 100% errors on that reading test would provide a full-scale IQ estimate of 84. It can go no lower. It is hardly surprising, then, that the conclusion of studies employing this type of methodology is that premorbid IQ falls within normal limits since the measures they employ mean that it is impossible to estimate IQs lower than 84. The literature indicates that the mean IQ in adult schizophrenic samples is generally lower than the mean of the general population R5615511CHDBHDBH; studies estimating premorbid IQ in schizophrenic patients should employ assessment techniques that encompass the full distribution.

• Stevens et al. also comment on the increases and declines in IQ across the group, a phenomenon that does appear to reflect regression toward the mean; as they point out, the majority of those with low IQs demonstrate an increase, and the majority with higher IQs demonstrate a decline. They suggest, however, that such individual variation requires a nonstatistical explanation. Certainly, there are many possibilities; as stated in our original paper, changes in another cognitive function such as memory, orientation, attention, or frontal function may account for this. The WAIS-R summary IQ is a rough index to the general functioning of an individual and is a composite score of a number of abilities and skills. It is also possible that interrupted education was at play in this group. IQ tests are heavily dependent on acquired knowledge and can often be viewed as tests of educational attainment. A proportion of our subjects were hospitalized early in life. One could hypothesize that their ability to benefit from their educational experiences was, at best, heavily compromised.

• We do agree with our colleague Dr. David that one must be cautious in interpreting longitudinal data, but we would point out that such a design does control for most of the age cohort effects that plague cross-sectional studies.

• Dr. David is, of course, correct that the group we studied is unusual in that the schizophrenic subjects had already been sufficiently deviant as children to be referred to a child psychiatry department. We did not claim that they were representative of all schizophrenic patients but pointed out that, as he knows, all four recent cohort studies that examined childhood IQ in representative samples of preschizophrenic patients found it to be significantly reduced.

In summary, the evidence is now conclusive that children who go on to develop schizophrenia have lower-than-average IQs. Dr. David believes that there then ensues further decline, although he acknowledges that "truly longitudinal data on this point are sparse." If he and our other correspondents wish to continue to believe in this "myth" of further decline, that is their right. If they wish the rest of us to share their belief, they must produce convincing longitudinal data.