Arthur R. Jensen (2000) A Nihilistic Philosophy of Science for a Scientific Psychology?. Psycoloquy: 11(088) Intelligence g Factor (49)

Volume: 11 (next, prev) Issue: 088 (next, prev) Article: 49 (next prev first) Alternate versions: ASCII Summary
Topic:
Article:
PSYCOLOQUY (ISSN 1055-0143) is sponsored by the American Psychological Association (APA).
Psycoloquy 11(088): A Nihilistic Philosophy of Science for a Scientific Psychology?

A NIHILISTIC PHILOSOPHY OF SCIENCE FOR A SCIENTIFIC PSYCHOLOGY?
Reply to Barrett on Jensen on Intelligence-g-Factor

Arthur R. Jensen
Educational Psychology
School of Education
University of California
Berkeley, CA 94720-1670

nesnejanda@aol.com

Abstract

Barrett's (2000) nihilistic notion of the philosophy of science is, I believe, very mistaken in its assumptions and limiting constraints, which would be largely rejected by most philosophers of science past or present. The conditions imposed by Barrett would greatly hinder or even prevent progress in understanding phenomena at the frontier of any relatively undeveloped science, not only for psychology, but in any other area of natural phenomena. The fallacy is in his assumption that we cannot measure anything until we already know everything about what we are attempting to measure. The history of every field of science amply contradicts this Draconian restriction.

Keywords

behavior genetics, cognitive modelling, evoked potentials, evolutionary psychology, factor analysis, g factor, heritability, individual differences, intelligence, IQ, neurometrics, psychometrics, psychophyiology, skills, Spearman, statistics
1. The keystone of Barrett's (2000) philosophy of science seems to be that all research, in order to be 'scientific' must be based on ratio scale measurement, which measures an attribute in equal and additive units on a scale with an absolute zero point that is perfectly isometric with the attribute being measured. Ordinal or rank-order scales, which constitute most of the measuring instruments used in differential psychology, including IQ and most other tests of cognitive abilities, are supposedly anathema to the advancement of science. Yet, as far as I know, the stricture imposed by Barrett has not been required for any other science. Many variables in the physical sciences have been studied with well-defined but nonratio or nonadditive scales - the hardness of gems, viscosity, and the refractive index. And many 'pure' or dimensionless numbers (as the correlation coefficient is) are seen in the physical sciences, such as the ratio of electrostatic to gravitational force.

2. In the history of measurement of certain properties, such as heat and electricity, virtually nothing is known initially about the nature of the property being measured. The early scales devised to measure these phenomena were neither ratio scales nor isomorphic with the property being measured. Temperature, for instance, was originally measured by thermometers that were at best ordinal scales, and even these had imperfect reliability in the psychometric sense. The kinetic theory of heat was still in the distant future when the first thermometers were invented. Does it make any sense to insist that because scientists do not yet have a ratio scale for measuring variation in some variable phenomenon X (in order to determine whether it has any systematic relationship to a variable phenomenon Y), they should be forbidden to determine whether there is a rank-order correlation between the ordinal measures of variables X and Y? Would such a correlation, if obtained, be scientifically meaningless? Experimental manipulations of X or Y could further establish that they have some common causal properties. Prohibitions of this kind of procedure are certainly not recognized by Charles Darwin's statement that "Science consists of grouping facts so that general laws and conclusions may be drawn from them".

3. I grant that ratio scales can prove useful, indeed essential, for answering certain questions in psychology, such as determining the precise form of mental growth curves for various abilities. We can do this, for instance, with reaction time (RT), which is a ratio scale, but not with, say, spelling ability, arithmetic reasoning, or artistic aptitude, for which there are no ratio scales, only ordinal scales. One of the attractive features of RT with respect to g theory is that ordinal measures of g are correlated with measures of RT, which may afford the possibility of measuring g on a ratio scale. Because a measure of RT based on any single elementary cognitive task (ECT) has only a relatively low correlation with psychometric g, developing a reliable and valid ratio scale of g based on RT would require an optimal combination of various ECTs. We already know that RTs derived from a number of diverse ECTs have multiple correlations with g factor scores that approach R = +.70, even though no special effort has yet been made to develop a battery of RT measures that might show even higher correlations with psychometric g. Developing a ratio scale of g based on time measurements may involve some mathematical tricks to maximize the correlation of these measurements with psychometric g, such as transformation of the scale, but this is a quite legitimate procedure and is commonly applied in the physical sciences. Psychologists are generally too reluctant to use transformations of scale to permit their data to show a better or simpler fit to their theoretical models. Normalized IQ scores are an example, based on the theoretical conception that general mental ability approximates a normal distribution in the population -- as do physical stature, brain weight, and many other anthropometric variables.

4. In the course of the scientific study of a phenomenon, concepts or theoretical constructs (such as g) typically 'grow' in definition, complexity, and precision. These attributes are not all there at the beginning, before a scientist is allowed to begin investigation, as Barrett apparently thinks is necessary. The science of genetics affords an example. Mendel, in 1865, explained the results of his simple breeding experiments with peas in terms of 'factors' (the word gene hadn't yet come into existence), about which nothing was known, except that they served in Mendel's model as the cause of the observed hereditary variation in a few specific characteristics of peas.

5. During the subsequent 88 years, Mendel's nondescript construct was elaborated both theoretically and empirically. T. H. Morgan hypothesized that genes are arranged on the chromosomes and, on the basis of linkage analyses from breeding experiments with fruit flies, measured genes' relative positions on the chromosome. (His ordinal scale [in 'decimorgans' or 'centimorgans'] at the time would probably have been unacceptable to Barrett.) Then H. Muller found that specific functions of single genes could be knocked out or mutated by X-rays, thereby further establishing their identity as molecular entities. Finally, in 1953, Crick and Watson discovered the specific molecular structure of the genetic code. (All of these scientists except Mendel won the Nobel Prize for their contributions.) This chain of development scarcely resembles Barrett's requirement that scientists must already know the essence of what they are investigating and measuring before they begin their empirical research.

6. Barrett is a recognized professional in the field of electro- physiological research on psychological variables, particularly relating certain brain variables, such as evoked potentials and nerve conduction velocity, to mental test scores that strongly reflect psychometric g. His empirical contributions in this field (some cited in Jensen, 1998) depend on ordinal scales for measuring mental abilities and, more generally, they do not in the least contradict the psychometric methodology or the investigative perspective of the research program illustrated in 'The g Factor' (Jensen, 1998, 1999). Presumably Barrett now has in mind a new and improved scientific methodology for the study of individual differences in behavioral traits, particularly g, the earliest and now most solidly established latent trait in the behavioral sciences. It will be interesting to learn just how it differs from the investigative paths that have led to our present knowledge about g.

7. As I am not an expert in the contemporary philosophy of science, I asked two university professors, who are recognized experts in this field and teach courses in it, for their opinions on Barrett's position. I will only quote a few of their key points:

    "Barrett's core error is not understanding the consequences of two
    metatheoretic truths that all contemporary philosophers of science
    accept, as far as I know: (1) Most theoretical concepts are defined
    by their role in the conjectured theoretical network. (A subset are
    'operationally' defined by a fairly direct tie to observations.)
    (2) The theoretical network is incomplete.  (3) It follows that
    theoretical concepts are 'open', or what logicians call 'partially
    interpreted'. Research continues precisely because they are open;
    the research task is to 'close' them, although never completely.
    Barrett ignores all this and seems uninformed as to contemporary
    metatheory, especially the theory of implicit definition and open
    concepts."

    "As you surmise, Barrett's animadversions against psychometrics
    apply across the board to all science."

REFERENCES

Barrett, P. (2000). Intelligence, psychometrics, IQ, g, and mental abilities: Quantitative methodology dressed as science. PSYCOLOQUY 11(046) ftp://ftp.princeton.edu/pub/harnad/Psycoloquy/2000.volume.11/ psyc.00.11.046.intelligence-g-factor.45.barrett http://www.cogsci.soton.ac.uk/cgi/psyc/newpsy?11.046

Jensen, A.R. (1998). The g factor: The science of mental ability. Westport, CT: Praeger.

Jensen, A.R. (1999). Precis of: "The g Factor: The Science of Mental Ability" PSYCOLOQUY 10(023) ftp://ftp.princeton.edu/pub/harnad/Psycoloquy/1999.volume.10/ psyc.99.10.023.intelligence-g-factor.1.jensen http://www.cogsci.soton.ac.uk/cgi/psyc/newpsy?10.023


Volume: 11 (next, prev) Issue: 088 (next, prev) Article: 49 (next prev first) Alternate versions: ASCII Summary
Topic:
Article: