6

In a comment to this question about age related IQ decline in professors and scientists, the following comment was left:

I noticed that lot of IQ tests (RPM or California) have different norms for age groups. So how do they measure decline of IQ. Is it decline of IQ as absolute mental strength? Or drop in numbers measured by specific test? – @ICanFeelIt

I thought I'd phrase it as a specific question here:

If IQ involves age specific norms, how do researchers study age related changes in IQ?

See also this earlier question on age related change in IQ.

Community
  • 1
Jeromy Anglim
  • 30,741
  • 11
  • 93
  • 221

2 Answers2

4

When studying adult IQ, general adult norms are often used. So for example, even if 70 year olds have lower IQ than 20 year olds on average, for research comparison purposes the same adult norms might be used to study age related cognitive differences. Thus, raw-scores and IQ scores will be almost perfectly correlated (except for small adjustments to the distribution resulting form the conversion from raw scores to IQ scores). When studying change in intelligence across childhood, test raw scores can often be used. Of course, things get more challenging studying young children who are unable to complete tests suited to older children and adult.

Updated comments

  • Cohort and related effects: I agree with @what's point that assessing whether differences between age groups based on cross-sectional data is problematic. Such differences can be due to cohort effects, death, differential participation rates, and so on. It's much better where possible where possible to study changes in IQ using a longitudinal design where the same participants are studied over extended periods of time. If a cross-sectional design is all that is available then studying the effect of covariates, sampling participants in a thoughtful way, and reasoning about cohort effects are all important.
  • Scaling raw scores to IQ: @what suggests that you need to relate measurements to a standardisation sample in order to compare research with other studies. First, a distinction can be made between "intelligence" and IQ. IQ is standardised by age. Intelligence is not standardised by age. I am assuming that any studies interested in age related declines in IQ are implicitly actually asking about age related declines in intelligence. That is because by definition at the group level, IQ never changes with age, because age is by definition controlled for. Second, there is an issue of what is a useful metric for describing changes in intelligence with age. I agree that IQ is a familiar metric with a clear mean and standard deviation. For that reason, it is often useful to use this metric with a single adult norm sample to study age related changes. That said, using raw scores would tell the same story regarding age related change; it's just that the metric would be a little less intuitive.
Jeromy Anglim
  • 30,741
  • 11
  • 93
  • 221
  • If you don't relate your measurements to the standardization sample, as you seem to suggest, you can only compare the different age groups with other age groups from the same sample, but you won't have any IQs and won't be able to compare your findings with other research. Also, you don't correct for changes in education and other cohort effects, so if you find that younger people perform better, this might be due to older people not having had the same education so they would have performed worse than the contemporary young in your sample even when they were young. –  Jul 28 '13 at 20:49
  • So your results won't tell you if what you see is an effect of the old having gotten less intelligent, or a general increase in intelligence. –  Jul 28 '13 at 20:51
  • @what thanks for the comments. I've tried to clarify a few points. – Jeromy Anglim Jul 29 '13 at 02:24
  • Then why do we normalize the IQ within age groups at all? That only makes it difficult to understand the absolute intelligence of an individual. For example, without having access to the norms and being able to perform the necessary calculations, I will not be able to draw any conclusions about my development from having my IQs at age 20 and at age 40. –  Jul 29 '13 at 10:56
2

Basically, what you do is you compare each age group IQ to the standardization sample.

According to Kaufman (2005, p. 172), "Parker (1986) had the clever idea of examining the comparative performance of year-of-birth cohorts by equating the standardization samples of the Wechsler-Bellevue I, WAIS, and WAIS-R."

Kaufman describes his procedure for a longitudinal study as follows (p. 172f.):

The WAIS and WAIS-R standardization samples are quite similar to each other, each matching relevant Census data on numerous key variables. They differ in that the data were collected 25 years apart, in approcimately 1953 and 1978. Thus, several cohorts in the WAIS sample are also represented in the WAIS-R sample. For example, adults born in the 1909-1913 cohort were tested at ages 40-44 in 1953 (on the WAIS), and again at ages 65-69 in 1978 (on the WAIS-R). To the degree that the two samples are comparable, a comparision of the test performance of 40- to 44-year-olds on the WAIS with that of the 65- to 69-year-olds 25 years later on the WAIS-R represents a longitudinal comparison of adults from the same cohort.

Before making the comparisons, he [Kaufman] verified empirically that the independent samples were extremely well matched and comparable within each of the four cohorts on the important variables of gender, race ..., geographic region, and educational attainment. Then, he had to convert sums of scaled scores on the WAIS and WAIS-R to a common yardstick to permit age-by-age comparisons, and chose to use the norms for ages 25-34 for all adults in the study. Next, he had to control for the fact that different tests (WAIS vs. WAIS-R) were administered at the two points in time. Conceptually, these two adult scales are interchangeable ..., but because of the "Flynn effect" the WAIS-R yields lower IQs. Kaufman added 6 to 6½ points to each WAIS-R IQ (the median differences from 20 studies totaling over 1.300 subjects) (Kaufman, 1990, Table 3.13) to convert these IQs to WAIS IQs. These "corrections" to the WAIS-R IQs helped answer the crucial question, "How many IQ points higher would adults have scored had they been administered the WAIS instead of the WAIS-R in 1978?" Finally, he applied a time-lag correction to control for cultural change during the 25-year span, just as Owens (1966) did in his Iowa State study. Adjustment for cultural change requires a comparison of the IQs earned by adults of the same age in 1978. The 1909-1913 cohort, for example, was 40 to 44 years old n 1953. This group was compared to adults aged 40-44 in 1978 to determine how cultural changes have affected test scores for this age group. Similar time-lag comparisons were conducted for each of the other three cohorts who, in 1953, were ages 20-29, 30-39, and 45-49. The analyses showed that cultural change affected each of the four cohorts about equally, producing about a 3-point IQ gain on the Verbal and Full Scales and about a 5½-point gain on the Performance Scale, presumable due to some type of culture-related change between 1953 and 1978 that affected all adults who were between the ages of 20 and 49 in 1953. Kaufman (1990) adjusted the estimated WAIS IQs earned by each cohort in 1978 for these time-lag effects to remove the influence of cultural change.

Sources:

  • Kaufman, A. S. (2005). Assessing adolescent and adult intelligence (3rd ed.). Wiley.
  • Parker, K. C. H. (1986). Changes with age, year-of-birth cohort, age by year-of-birth interaction, and standardization of the Wechsler adult intelligence tests. Human Development, 29, 209-222. doi:10.1159/000273048