Scott Morrison
MA Candidate, Educational Technology
Concordia University
Montreal, Canada
smorr@alcor.concordia.ca
There is a central warning that runs throughout Steven Jay Gould's Book, The Mismeasure of Man. Gould alerts the reader to his concerns about the objectivity of science and the role that the scientist has in the corroboration of empirical findings for his or her theory or research program. Gould argues that the science underlying the hereditarian school of intelligence is flawed in that the research and findings are result of a priori convictions of the researchers, not of objective inquir y into the nature of human intelligence. Accompanying the prime fallacy of a priori science are the additional errors of reification, which leads to circular reasoning, and mismatched logic that hereditarians employ while arguing their theories. Throughout the book Gould revisits these three errors as warnings for social science research. He admonishes the reader to become more critical of scientific conclusions and scientists to consider how their convictions may consciously, or unconsciously, influence their findings and interpretations.
In this review I summarize the thrusts of Gould's criticism and, to an extent, analyse their nature. The structure of this summary parallels the historical nature of Gould's arguments up to the time of psychometricians. I then summarize twentieth centur y biological determinists with an overview Gould's three main critiques of the psychometricians. Additionally I criticize Gould's approach to the debate on intelligence tests through a criticism of the rhetorical nature of his arguments. In doing so I co nsider the nature of rhetoric in science, how social science is particularly susceptible to rhetoric, and the resultant imperative for critical review of social science research.
The organization of Gould's arguments is historical and attacks specific figures and schools of science throughout the last 150 years. Starting primarily with polygenists such as Aggasiz in the mid-nineteenth century, Gould charts the use of quantitative methods to demonstrate the efforts of scientists to quantify intelligence and consequently rank individuals, genders, cultures, social classes, and races in terms of intellectual ability.
Gould's method of argumentation here is ruthless but valid. His first critical attack is launched against Morton. Using the raw data that Morton provided, Gould reveals mathematical and statistical errors that Morton committed (albeit unwittingly) in hi s comparison of the cranial measurements between different races and genders to illustrate the supremacy of European Males. These errors consisted of unequal representation of racial groups and gender and selective use of data. In addition Gould attacks Morton's subjective a priori belief that "directed his tabulations along preestablished lines." (p 69) These assumptions were that intelligence could be measure by brain size (cranial capacity) and that whites, on average, were more intelligent th an those individual of other races.
As the sophistication of social science research, statistics, and psychology increased, so did the sophistication of the arguments of biological determinists. Statistics led to added power to the arguments of craniometrists, such as Broca, that there are differences in physical brain size and structure between different races, classes, and genders. While Broca statistics were less open than Morton's to criticism, Gould still points out that Broca's facts were gathered selectively and then manipulated un consciously in the service of prior conclusions. (p. 85.)
The use of physical characteristics were not used by biological determinists for exclusively intellectual discrimination. Gould details the work of Lombroso and his colleagues whose theories purported that physical characteristics were indicators of mora lity -- that criminals can be identified through physical appearance. The target of Gould's arguments here is the false logic that ontogeny recapitulates phylogeny; that appearance is a reflection of reality. By this logic, if people or races appeared c riminal or inferior, it is because they are criminal or inferior. Gould does not rely on revealing quantification anomalies to refute Lombroso however (Lombroso himself used few statistics to support his theories) but on the use of the arguments of scien tists of Lombroso's time, his own rhetoric, and the appeal to late twentieth century commonsense to see through Lombroso's fallacies. A corollary of The Mismeasure of Man is to discuss how research affects social policy and that research is not a purely academic matter. Gould's first major illustration of this purpose occurs in his discussion on how Lombroso's support of biological determinism impacted on social policy of his time. Gould makes the point stronger and more relevant to the present by describing how Lombroso's ideas re-emerged in more recent theories such as the XYY issue of the 1960's- 70's, and how similar racist ideas still have the potential to influence politicians and decision makers whose primary concern is not the validity o f research but the promotion of specific policies.
Well over half of Gould's book is dedicated to a scientific discrimination of greater sophistication than the craniology and theories concerning avatism of the latter part of the nineteenth century. The principle target of Gould's book is the twentieth c entury science of psychometrics and the efforts by many psychologist to measure general intelligence. While Gould arguments loosely follow the historical development of psychometry, there are concerns that Gould addresses in his discussion of each develo pment. To provide a more cohesive overview of Gould's position towards psychometry, I shall address these concerns rather than parallel Gould's discussion. Gould's criticism of psychometricians has three dimensions: the criticism of their reasoning, the criticism of their technique, and the criticism of their intentions. Each criticism warrants more complete discussion.
Gould repeatly condemns the a priori assumptions of psychometricians such as Goddard, Terman, Yerkes, Spearman, and Burt. Their reasoning is circular in that they assume a single dimension to intelligence, they create instruments and use other res earch techniques to measure the single dimension, and consequently they validate their theory that general intelligence has a single dimension because they were able to obtain scores and rankings of that dimension. The "Factor Analytic School" is somewha t more sophisticated in that they search for factors by using mathematical correlations in their data and find that indeed factors of correlates exist. They then assume that the major factor is their presupposed factor of general intelligence. Gould arg ues that the statistical evidence that psychometricians use are artifacts of intelligence tests and analytic techniques that circularly support a priori assumptions of the nature of intelligence, not reflections of a true general intelligence that can be stated as a single number. The fallacy is that their techniques were specifically design or chosen to return a single dimension of intelligence. To use a cliché, if one goes looking for trouble, one is bound to find it.
Regardless of the question of whether a single factor or score reflecting true intelligence can be measured (in effect, the question of whether the statistics are what the psychometricians purport them to be), the psychometricians' assumption that such a score is innate or the result of heredity is an even more profound error of reasoning. Gould spends much time addressing this error. The following three points overviews Gould's summary of this poor reasoning: 1) Psychometricians hold an unfounded view of intelligence that specifies natural and trained intelligences. 2) They use the technique of factor analysis, which cleanly partitions intelligence measures into two parts, to understand the variance of scores in intelligence tests. 3) These partition s are interpreted to be a general intelligence (the general factor of covariance), and a specific intelligence (the residual variance due to individual score differences). In short, given an a priori view of intelligence which specifies two forms of intelligence (natural and trained), and the analytic technique returns two variances (correlated and uncorrelated) psychometricians did not have to think hard to support what they were looking for: a 'link' between correlated factors to natural intell igence and residual to trained intelligence. The link is the belief that correlations indicate a biological commonality.
Gould argues that such a link is unsubstantiated. The ability to distinguish variance and factors does not support a relation of a general factor to heredity, and certainly not the causal relationship in which the reified general intelligence is the sour ce of the correlation of the data that produces the factor. The psychometricians' ability to see such relationships still lie in their reification statistical results and conclusions that a factor on an intelligence test is a measure of general intellige nce.
From the initial steps of their research, the psychometricians enter biases into the results. Gould shows how the intelligence tests used are culturally and socially biased through discussing examples of test items. Generally stated, the intelligence te sts used presupposed that the subject either came from or had a basic familiarity with white middle to upper class North American culture. Gould used Yerkes' data (the largest body of data used to demonstrate statistical differences in mental age between races) to illustrate empirically the confound of social familiarity. Yerke's mental age correlated highly (r = .75) with years of schooling, years of residency in American culture, and diseases of poverty. While this immediately raises the ques tions surrounding the construct validity of the test, Yerkes interpreted such correlations to indicate that higher intelligence leads to greater education, wealth, and reflected certain immigration trends, rather than interpreting such correlations as in dications of bias in the instrument. Not only were the tests used biased but they were improperly administered. This stemmed from the construction of easy to widely administer tests such as Yerke's army intelligence tests and Terman's National Intelligence Test -- not the tests that Binet o riginally devised or envisioned. Gould again exemplified Yerke. While Yerke prescribed a specific protocol for test administration, the testers where "rarely able to carry out little more than a caricature of their own stated procedure." ( p. 201) Goul d cites many documented reasons for this, not the least being hostility towards the test and logistical difficulties. The result was that test data were confounded by varied administration of the tests. Regardless, Yerkes used the suspect data and conti nued with the analyses and interpretation.
Gould uses the example of Yerke again to demonstrate how unsatisfactory or unexplained results in the data were massages to achieve a better analysis. In order to compensate for unexpected scores on their instruments, Yerke's chief statistician, Boring, massaged raw data in such a way that rendered resultant analysis suspect. As Gould discusses, the unexpectedly high frequency of zero scores on certain instruments were not taken as indications of possible test bias, but were redistributed in so that a priori assumptions could be better supported. In essence, the undesirable scores were "calibrated" or grossly regressed against subjects scores on other instruments. The compounded error of this manipulation was that the zero scores affected were those collected from instruments administered to the less literate (most often blacks and immigrants) and the zero scores were unjustifiably lowered according to poor performance on other measures. In short Gould argues that any alternative interpretatio ns, were explained away whenever statistically possible by massaging the data and statistics or looking at selective results.
To summarize, Gould argues that the many testing and data irregularities as well as the presence of alternative hypotheses were discounted by the psychometricians seeking a measure of intelligence even though such problems would cause more cautious resear chers to question their results. Gould demonstrates that as more evidence contrary to the psychometrician's interpretation were revealed, they went to greater lengths to defend the results that supported their convictions, even when it made more sense to put their idee fixe to rest.
Gould criticizes the technique of factor analysis on other grounds. The scientists who support factor analysis use a more sophisticated tool and less sloppy analyses. Accordingly, Gould attacks the technique itself more than the methodological errors of the researchers. The specific target of Gould's criticism is the statistician's ability to use multiple methods of extracting factors and rotating them in a multivariate space to achieve different results. Gould plays the specific statisticians of the factor analytic school off of each other (eg Thurstone versus Spearman and Burt) when discussing factor rotation. Thurstone's methodology consisted of rotating explained factors in multivariate space to achieve the best loadings of variables on unique fa ctors, rather than using an averaged factor (principle components). The use of oblique rotation were later used to counter Thurstone's theory of primary vectors of intelligence through the introduction of second order factor analysis. Ironically, Gould uses the multiple factor analytic techniques of the psychometricians to support his central argument against the psychometricians; the argument that the results are more artifacts of the many ways to use the technique than of the data. In this light, the results cannot not dictate any certain interpretation, let alone a meaning that can be reified. Factor analysis has been misused by the factor analytic school of researchers.
While Gould exonerates Binet, who intended that intelligence testing be used to identify individuals so that they may be assisted in schooling, as well as Goddard and Brigham (a follower of Terman) who recanted many of their earlier positions, he still co ndemns their contributions to the school of biological discrimination that affected social policy in America and Britain for decades of this century.
Although much of The Mismeasure of Man centres on Gould's criticisms of scientific techniques, the them of biases and intentions in research is truly the term of the book. To be more precise, the use of science to justify and promote specific di scriminatory convictions is the main target of Gould's attacks whereas the critique of specific research and techniques is merely an avenue of attack. This thesis is further clarified in the analysis and critique of Gould arguments that follows.
As Gould's arguments develop and he addresses scientists who use more sophisticated techniques and arguments to promote their convictionsand accordingly, his arguments change. Towards the end of the book, Gould presents his case against researchers such as Terman and Burt with a second syllogism. This syllogism subtly shifts the order of premises and conclusions in the first syllogism:
Analysing and contrasting the two syllogisms is enlightening. The first premises of each argumentation style seem similar but with closer examination, I argue that Syllogism A holds at premise one but the syllogism B doesn't hold at its first premise. H aving pre-conceived notions of what to look for (eg theories and hypotheses) does not imply that a scientist will ensure that the notions are supported. It must be understood that there is a danger of such happening, but there is not direct equation of s uch happening. Otherwise, all but the most empirical and exploratory of studies (and chance discoveries) must be suspect of bias. This raises that question of whether or not such strict empiricism is the nature of social science. The scientific method necessitates hypotheses and in doing so, does it not allow itself to be criticized on the point of a priori convictions? Discussion of the nature of science, howeve,r is too involved to investigate here and my main concerns are the nature of Gould 's argumentation and establishing the weakness in Gould's second form of argumentation.
Contrasting the two styles of argumentations suggests that while the former syllogism is valid and logical, the second is weak and rhetorical. With the first syllogism, premise two can be supported empirically and premise one holds logically. Gould repl icates analyses to show empirically where analyses were manipulated regardless of the intentionally in these manipulations. The effectiveness of second syllogism, however, depends on Gould's ability to construct and present his argument rather than the e mpirical and logical strength of the argument itself. Gould constructs the second syllogism on his ability to convince the reader that the scientist in question had a priori assumptions. Gould supports his second syllogism with a shift from syl logism A to syllogism B and with rhetoric.
In addressing the first point, I argue that the second syllogism is superficially effective because the subtle shift in the order of premises and conclusions. The reader is prepared to agree with the first premise of syllogism B because it seems to follo w easily from the conclusion of syllogism A. The first syllogism prepares the reader to agree with the first premise of the second syllogism even though this syllogism does not necessarily hold. This shift in style not only demonstrates the sophisticat ion of Gould's rhetorical ability but also the rhetoric which underlies much of Gould's argument against scientific determinism.
Ironically, rhetoric is Gould's primary weakness and strength. In addition to a rhetorical argumentation style, Gould uses rhetorical language throughout to build an argument against the biological determinists. The usage of rhetoric is too numerous to cite fully but a few short examples represent such use (italics added):
"Terman relentlessly emphasized limits and their inevitability . He needed less than an hour to crush the hopes and belittle the efforts of struggling, 'well-educated' parents afflicted with a child of IQ 75." (p. 179.)
"One might suspect that Yerkes, in his blindness , didn't read his own information." (p 217.)
"Hearnshaw has convinced me that the very enormity and bizarreness of Burt's fakery forces us to view it not as the 'rational' program of a devious person trying to salvage his hereditarian dogma when he knew the game w as up, ... but as the actions of a sick and tortured man." (p. 236.)
The main reason for illustrating Gould's rhetorical language and argumentation style is not to discredit Gould. His conclusions are agreeable but they are not conclusive in that not all biological determinists and racists can be similarly countered. Rac ist intentions, as undesirable as they may be, will not always lead to massaged data and analyses. There exists a theoretical danger under Gould's weak argumentation and Meehl's paradox of social science that a racist's convictions of difference between groups will be statistically supported with methodologically sound research. The avenue of attack that Gould uses against the methods of biological determinists is ultimately shortsighted because it discredits only present methods, not any future attempt s by scientists with more sophisticated methods.
Gould's use of rhetoric, incidently, introduces questions concerning the role of rhetoric in science. Is the rhetorical argument a necessary evil in social science because of the inability for a researcher to accept the null hypothesis? As one cannot, u sing inferential statistics, support the equality of two groups and statistical support is merely corroboration of a theory, debate on many issues may need to be more rhetorical than scientific. In this light, I argue that Gould must use the rhetorical argument to counter scientific determinism because he cannot use statistics to support an argument of no difference between races and classes. Gould essentially is forced to used rhetoric to develop an ethical argument, not a statistical argument.
To examine this further, the prevalence of rhetoric in Gould's arguments suggests that he is using an ethical argument against the ontological arguments of the determinists. Gould currently attacks the determinists "is" with his own "ought not" which h e then uses to weakly argue "is not". But a firm "is not" is not truly argued, nor can it be. Gould cannot demonstrably support that there is no difference between individuals and groups of people as such corroboration is not possible in social science research and null hypothesis testing. Because of the inability in inferential statistics to conclusively prove no difference, support of the hypothesis of difference is always possible, and eventually likely, for scientists intent on using statistics to support discriminatory views.
Given this impasse, it may be more useful to clearly develop clear arguments against the ethical positions of the biological determinists. Racism is an ethical issue, no doubt, and the direction of Gould's rhetoric should be firmly against the ethics and rhetoric of the determinist's position, not against their statistics. To argue scientific racism or classism, one should primarily counter the racist's "ought" with an "ought not" and their "is" with a "so what". It is more important to, as a society, deal responsibly and ethically with differences between individuals and groups rather than trying to show that differences do not exist, an avenue that, given Meehl's paradox for social science, will surely fail.
This critique of Gould is short -- too short to fully analyze Gould's arguments and critique them in depth. An overview will have to suffice. To support Gould's intentions and his efforts I raise some philosophical questions. Does the null hypothesis m ethod inadvertently necessitate rhetorical arguments? Is it because of limitations in the methodology and Meehl's paradox that Gould needs to be a rhetorician as well as a statistician? Perhaps Gould had little choice to argue as he did.
Gould draws the reader's critical eye towards the nature of bias and a priori assumptions in social science research but I note that in doing so, he must use rhetorical arguments. As such arguments are not empirically based, Gould cannot escape th e biases his own rhetoric carries. The analysis of The Mismeasure of Man has ironically supported Gould's central theme of bias in social science by illustrating his own necessary use of rhetorical rather than empirical arguments in addition to hi s overt illustrations of how biases influence the research of biological determinists. To quote another's observation, "The result, ultimately, may be that applied social science will be able to do no better than affirm the prejudices of those who conduc t [and argue] research." (Shaw, 1991)