Department of Philosophy
It has sometimes seemed that those on the fringes of established science cry “paradigm bias” to explain why their work doesn’t get any attention when it is in fact the work itself that is to blame. Presumably, some evidence that conflicts with received views is ignored for good reason, and some without good reason. When an apparent anomaly is dismissed for no good reason, then the scientists in question are behaving badly.
But are they behaving “unscientifically”?
In particular, we will argue that, in periods of instability in science (“revolution,” if you like), it is in the very nature of science to treat anomalous evidence with hostility and suspicion, even when there is little evidential reason to suspect it.
For various genetic reasons, it seems that all aboriginal Americans are more closely related to one another than they are to any other populations, and are more closely related to the peoples of Asia than those of other parts of the world. The reasonable conclusion to draw from this evidence is that the first Americans migrated from Asia, either across the Bering Strait or across a land bridge.
Large-scale migration by boat is unlikely, even across so narrow a body of water as the Bering Strait, so a hypothesized Bering Land Bridge is the best hypothesis for a migratory route.
So confidently was this view held that in 1962, writing for Scientific American, William Haag could say,
There is an impressive array of evidence for the recent-migration view, and comparatively little for any earlier human presence in the Americas. What seemed to be evidence of earlier occupation has usually turned out to be misleading.
David Meltzer (1993) describes the situation this way:
There are at least three impressive kinds of evidence for a Late Pleistocene migration (or set of migrations):
All three kinds of evidence point to three waves of migration, the
earliest in the Late Pleistocene, as hypothesized. The earliest
clearly datable sites so far are those at Clovis and Folsom, and
they are no earlier than 11,500 BP.
The Amerind languages show the most variety, and are geographically the most widespread, being spoken in areas from Canada to Tierra del Fuego. These two facts argue for the relative antiquity of the common language from which they all derive.
The Eskimo-Aleut languages are fewer in number and more similar to one another. They are also spoken in a smaller area, around the northern coastal regions.
The Na-Dene group is intermediate in variety and extent. Those languages also are spoken in areas to the south of the greatest southern extent of the Eskimo-Aleut languages, but not so far south as the Amerind languages. Moreover, the language groups can be arranged in order of similarity to Old World languages, with Eskimo-Aleut being most like, and Amerind least like, the languages spoken in Asia.
This arrangement of languages points to three separate waves of migration, with the ancestors of Amerind speakers arriving first. However, this relative ranking gives us little in the way of absolute dating for the migrations.2
On the basis of this characteristic, a particular shovel-like shape to the incisors called sinodonty, he concluded that Native Americans divided into three genetically distinct groups: Eskimos, Athabaskans, and South Americans.
Christy Turner (1986) made a statistical analysis of American teeth to check this classification. Looking at other, similarly heritable characteristics of teeth, and cataloging similarities and differences from nine thousand different prehistoric Americans, he also concluded that Native Americans divided into three genetically distinct groups, but he identified the three groups more directly with Greenberg’s three linguistic groups.
When a gene expressed itself in a visible and easily-preserved part of an animal, such as a tooth, then we can use the variations in that part to date the genetic history of the animal. In the case of humans in North America, we can tell by distributions of kinds of sinodonty that the North American population split from the North Asian population about twelve thousand years ago - which confirms the late-Pleistocene migration view.
There are parts of the genetic code, however, that do not get expressed at all, or are expressed only in neutral characteristics. In those genes, the regular rate of mutation is not affected by environmental pressures. In particular, mitochondrial DNA (mtDNA) is not subjected to the mixing forces of fertilization, as all a creature’s mtDNA comes from its mother.
So given a reasonable estimate of how quickly and how regularly mutations occur in mitochondrial DNA, we can fairly accurately date when populations diverged. By that measure, Americans split from North Asians about 20,000 years ago.
This is earlier than the other methods gave us for a first
migration, but may be accounted for by the estimate of the rate of
Even though there are occasional finds that seem to be datable to much earlier, it is more reasonable to think there must be something wrong with the dates for those sites than to accept them at the cost of overturning so well-grounded a theory.
The inability to explain why a site seems to be earlier than the late Pleistocene is no bar to accepting the late migration theory, especially if the alternative is accepting an earlier migration while being unable to explain the linguistic, dental, and genetic evidence.
Meltzer (1993, p. 21) characterizes the archaeologist’s position this way:
Most of the archaeologists who give this understandable response are considerably less conciliatory than Meltzer. In fact, Haag’s response cited earlier, which dismisses claims of extreme antiquity for human presence in the Americas as irrational, is the norm rather than the exception.
The oldest sites that have stood up to careful scrutiny, and whose evidence is completely unambiguous, are Clovis and Folsom, both datable to after 12,000 BCE, and so completely consistent with the Late Pleistocene migration.
The specific archaeological project that is central to this work was located at Hueyatlaco, Valsequillo, which is a few kilometers south of Puebla, Mexico. The area had become very well known among archaeologists due to the varied extinct animal forms.
The initial excavation began in 1962. During the continued process of excavation five sites were discovered and stratigraphic sections sequenced (Irwin-Williams 1967a).
The final excavation at Hueyatlaco was concluded in 1973.
Field work continued throughout the excavational process by the members of the team, including Dr. Cynthia Irwin-Williams and Dr. Virginia Steen-McIntyre.3 Later consultants affiliated with the project were Ronald Fryxell, B. J. Szabo, and C. W. Naeser in continued efforts to resolve the dating controversy surrounding the evidence accumulated during the excavation process at Valsequillo, Mexico (Malde and Steen-McIntyre 1981).
There were no irregularities in the team’s methods, and the site was guarded to prevent tampering or accidental destruction of evidence (Irwin-Williams 1967a).
There were locations in the area suitable for camping, and nearby were sites suitable for slaughtering activities and sites that were appropriate for butchering procedures because of the close proximity of small streams. Irwin-Williams acknowledges that modern estimates regarding the presence of man in this locale ranges from 11,000 years to more than 30,000 years.
Controversy began in 1967, before the digs were completed. Despite the thorough efforts and the competence of the archaeological team members at Hueyatlaco, Jose L. Lorenzo, Director of Prehistory at the Instituto Nacional de Antropologia e Historia, launched several allegations regarding the integrity of the project at Hueyatlaco, El Horno, and Tecacaxco (commonly referred to as Valsequillo).
The most significant allegation was directed to the authenticity of the artifacts retrieved from the Hueyatlaco site.
Lorenzo (1967) alleged that some of the artifacts had been planted by laborers working at the site and then commingled with other artifacts in a way that made it impossible to separate and identify the planted artifacts.
The intentional commingling of evidence, if it occurred, would raise substantial doubts about the age of the site, as well as the integrity of the principal investigator and other members of the archaeological team.
The allegations were addressed by Cynthia Irwin-Williams (1967b) in the Paleo-Indian Institute Miscellaneous Publications stating that the,
During 1969, Cynthia Irwin-Williams further refuted Lorenzo’s allegations with written statements from three reputable professionals in the field of anthropology and archaeology (Irwin-Williams 1969).
Radiocarbon dating on molluscan fossils (shellfish) which showed an age greater than 35,000 years.
The uranium method gave a result of 260,000 ± 60,000 years. A mastodon tooth retrieved from El Horno was dated using the uranium method and was calculated to be older than 280,000 years. Likewise, a camel pelvis recovered from the Hueyatlaco site was dated using the uranium method closed system at greater than 180,000 years, and using the open system as 245,000 ± 40,000.
A horse metapodial recovered from the Atepitzingo site in the Valsequillo area was dated using the uranium method open system date at 260,000 ± 60,000 years.
In the concluding remarks of the article (Szabo, Malde, and Irwin-Williams 1969, p. 243) the authors noted, rather mildly, that some of these were perhaps too old stating that,
In the same article, Malde commented that one of the difficulties in evaluating the samples was possibly due to a lack of stratigraphic markers from the field for correlation with the various sample localities.
Later (the results were published in 1981), he and Virginia Steen-McIntyre would collect samples of the stratigraphic layers including samples of pumice and ash to resolve just this point.4 Additional stratigraphic information would help determine whether the artifacts were located in an erosional trough such as a stream channel, which would indicate that the beds bearing artifacts were of a younger age.
This possibility raised doubts that could not be ignored. Drs. Steen-McIntyre, Malde, and Roald Fryxell, a specialist in mapping sediment layers at archaeological sites, returned to Hueyatlaco for the additional excavation. The work to determine the stratigraphic sequence was undertaken in 1973. This final excavation established a sequence of age for the first time, showing that the artifacts did not lie within a stream channel and thus, were not younger than the ash deposits that covered them.
As the volcanic glass weathers, moisture moves through the exposed surfaces. In temperate climates this process may be completed in 20,000 years. As the pumiceous glass becomes hydrated, the vesicles also begin to collect water.
The total filling of the vesicles may require ten million years or so. Thus, evaluating the fill within the vesicles assists in age determination.
The examination process requires approximately eight hours of microscope time for each sample.
During the microscopic examination of the phenocrysts, Dr. Steen- McIntyre detected a phenomenon she described as resembling a picket fence. The samples, instead of having fresh-looking crystal surfaces, looked rather shaggy, having a “picket fence” appearance. The volcanic glass fragments were also weathered and had absorbed water from the soil in which they lay until excavated.
The minimum age ranged from 170,000 years to 260,000 years BP (Steen-McIntyre, personal communication with Suzanne Clark).
Szabo’s results, using the uranium-series method, ranged in age from 180,000 to 260,000 years BP. Naeser’s zircon fission-track method showed ages ranging from 170,000 years to 260,000 years BP. Both sets of dates agreed with Dr. Steen-McIntyre’s observations of 251,000 years. Three separate methods, calculated by three separate geologists, yielded similar results, yet the results met with skepticism and hostility.
It is not improbable that Irwin-Williams feared her career was in jeopardy in light of such dates.
She certainly feared (or at least was wary of) what might happen if she was associated with fringe elements. When, at a meeting of the Geological Society of America, Malde and Fryxell announced their early dates for the Valsequillo site - which dates were established by three independent dating methods - the announcement was reported on the UPI wire for November 14, 1973. Irwin-Williams reacted with anger.
In a letter dated 3 November 1974 to Alan L. Bryan, a colleague in Alberta, she said:
This sounds eminently reasonable. If two of the dating methods are experimental, and one gives an essentially worthless result, then the dates are surely suspect.
Compare the charge of irresponsibility with the text of their announcement, as reported in Quaternary Research (Steen-McIntyre, Fryxell, and Malde 1981):
In our view, the results reported here widen the window of time in which serious investigation of the age of Man in the New World would be warranted. We continue to cast a critical eye on all the data, including our own.
This declaration of ignorance hardly sounds rash and irresponsible. Moreover, Irwin-Williams seems to be getting the fission track date wrong.
Steen-McIntyre, in a letter to J.L. Bada, cites the date given by that method as 370,000 ± 200,000; a wide range of error, but hardly meaningless. The experimental methods (Tephrahydration and Uranium series) have since been found to be reasonably reliable.
Eventually, Steen-McIntyre was forced to choose a less controversial dissertation subject, how to examine volcanic ash samples. Steen-McIntyre finally obtained her degree in 1977. Between 1969 and 1973, frictions within the Valsequillo archeological team with regard to the date of the site were building. Malde was enthusiastically promoting an early date (ca. 200,000 years BP), while Irwin-Williams was promoting a more conservative, but still controversially early, date (ca. 20,000 years BP).5
Steen-McIntyre’s allegiance was with Malde, but her subordinate position on the team and in the profession of archeology led her to be more cautious. Her caution, along with her thorough scholarship, made it possible for her to continue to find employment.
In early 1973, Virginia Steen-McIntyre had achieved international recognition from several organizations, including the National Academy of Science, from whom she also received funding for foreign meetings and speaking engagements. She worked part-time in her area of expertise for a government laboratory, and even became an adjunct professor of Archaeology at Colorado State University.
The reviewing scientist recommended the grant be denied on the basis of Szabo’s involvement with the Valsequillo project. Szabo had been labeled an incompetent scientist and lacked credibility (Steen-McIntyre, personal communication with Suzanne Clark).
E. James Dixon (1993, p. 128) reports similar responses to his writings when he merely suggested a mechanism for migration other than the Bering land bridge:
It was not just Dixon’s colleagues that found his views dangerous; editors of journals criticized his professional writings, not because they failed to meet the journal’s scholarly standards, but because they argued against the received view.
Dixon had done a series of studies in which he and a colleague had grown hemoglobin crystals from material recovered from spear points. They matched the hemoglobin from those points with that found in living species, and also with specimens recovered from extinct species.
The result was that some of those points could be dated to well before the Clovis and Folsom barrier, as the animals whose blood was on them were extinct before 12,000 BP. So either humans were in America before the late Pleistocene, or these animals survived longer than is currently supposed.
Dixon sent these results to Science, with the following result (Dixon 1993, pp. 111-112):
In other words, there was no complaint about the article on either stylistic or technical grounds, but only about the conclusions for which he argued.
And yet it is a common response to anomalies. Not only in archaeology, but in every other science as well, challenges to the received view are treated with exaggerated suspicion. It is entirely reasonable to treat anomalies with suspicion. After all, if a piece of evidence comes to light that is inconsistent with a well-grounded theory, it is not always clear which of the two has to yield.
Frequently, apparent anomalies evaporate on further examination. There is some incentive for scientists to try to overturn received theories, and so they may overstate what their evidence shows. If a received theory is backed by lots of evidence, it would be irrational to abandon it at the first anomalous finding, even if there is no alternate explanation available for the anomaly.
Why these extreme reactions?
In every science, anomalies are met with this same hostility. It seems to be standard practice in science, and yet it sounds paradigmatically unscientific. The reason this is hard to square with our notions of science is that we are failing to see science as the socially embedded practice that it is.
These claims cannot both be true, and yet both seem plausible. It does seem that the scientific method (insofar as there is a single method) is designed precisely to root out error and tend toward truer and truer pictures of the world. On the other hand, scientists are people, and scientific investigation is done by people in societies, and it would be amazing if they didn’t bring their biases into the laboratory with them.
We have three choices:
In fact, the two view are indeed compatible.
When the proponents of the self-correcting nature of science say “Science is unbiased” and the proponents of science as an ideologically driven enterprise say “Science is biased,” they are not disagreeing, because they are talking at cross-purposes; they mean different things by the word ‘science’.
The former are talking about a method employed in theory choice, abstractly conceived; the latter are talking about a socially instantiated practice that has theory choice as a component. Consequently, it is possible for the abstractly characterized method of theory selection to be self-correcting, and yet be embedded in a larger practice which to some extent undermines, or even defeats, self-correction.
For example, many scientists would admit that particular scientists may have let bias creep into their work, but that when they were doing so, they were doing bad science. In other words, it is ideal science, or good science, that corrects itself. But both sides to the debate can agree that there is good and bad science.
The believers in science may admit that some scientists are biased, but they want to assert that it is not merely in the ideal that science corrects itself, but also in real practice. They want to claim that science as we actually do it has a tendency toward truth, which would be unwarranted if it were only science as ideally practiced that has that feature. Also, many of the political critics of science want to claim that even when science approaches the ideal of objectivity, it still serves political power. So the distinction between real and ideal science does not illuminate the problem.
The scientific method, as described in innumerable science textbooks, is something like this:
If the consequences are correctly deduced, and the experiments are well-designed and well-performed, then the original hypothesis is refuted, even if it was the pet hypothesis of a well-beloved and authoritative scientist.
Feynman goes on to say that this picture is a bit oversimplified, but his further remarks only serve to add details to the three-part structure: hypothesis, deduction, experiment.
The results of the experiment then have an effect on what hypotheses get proposed, so the process is a self-correcting spiral, homing in on accurate representation of the world. It is easy to see how this understanding of science would lead one to think that it could not possibly be biased. If a biased scientist presents a faulty hypothesis, it will not be borne out by experiment; and so bias is rooted out, at least in the long run, by the structure of science itself.
The classical “scientific method” is a component of science, but it is not the whole thing. They are thinking of science as a social practice that starts well before hypothesis with background information, distribution of resources and opportunities, and ends with publication and discussion of theories.
What theories are accepted, published, and discussed forms the new background information out of which new hypotheses arise, so on this picture, too, science spirals, but the spiral is guided by more than just observation and experiment. It is because of these additional forces on scientific inquiry that science (in the “practice” sense) can be biased, even if science (in the “method” sense) is immune to bias.
Theory choice has been the focus of much discussion of science, and so has become science itself for so many people, because it is amenable to abstract treatment. In particular, it is amenable to a normative understanding; understanding science as theory selection allows us to develop logics of science, and interpret particular cases of theory selection in terms of how well they achieve the goals of science, including an accurate picture of the world.
But obviously there is more to how science gets done, and more to what scientific theories we accept, than the logic of theory choice alone. The scientific practice, as actually undertaken by real, working scientists, is better represented as a three-stage structure, with theory choice taking place in a context of hypothesis selection and public uptake.
Scientists are partly hired, promoted, and otherwise evaluated on the strength of how interesting the problems are that they are pursuing, so what we find out about the world is in part a function of what presently employed scientists find interesting. Proponents of theories that postulate a pre-Clovis human presence in the Americas will (as Steen-McIntyre’s case shows) have trouble finding employment.
Hypotheses that no one respects will have trouble finding funding and support; hypotheses that are very radical will be difficult even to formulate, for lack of a history. So, what theories we accept is constrained by what hypotheses get tested. At the theory-uptake stage there are similar constraints. If no scientific society or journal finds your work important or interesting, it won’t get published, and so other scientists will not try to replicate the results, and the general public will never find out about it.
A lot of evidence against the standard view gets weeded out at this stage (as Dixon’s case shows). Evolutionary biology had to wait decades for Gregor Mendel’s groundbreaking work because it languished in a second-rate journal that nobody was reading.
Even if a paper on a problem considered marginal by the majority makes it to publication, if the scientific community doesn’t pick up on it, discuss it, and expand on it, it vanishes into obscurity. So while we confine ourselves to consideration of the scientific method, it is true that any hypothesis, no matter what it is or who brings it up, is treated equally, when we turn to the social practice of science, we see that only hypotheses that can attract enough interest to get resources, publication, and discussion really have a chance to be accepted.
Scientists who undertake projects outside the well-structured set of alternatives (like flat-earthers or creation scientists) are dismissed as crackpots. Scientific work that is within the pale of respectable work is then evaluated solely on the grounds of how well it meets the canons of science in the “method” sense.
Anything respectable as determined by the received view will be accepted as worth doing, and will have a chance at publication and funding.
The middle stage of theory choice looms large, and the forces that operate on problem selection and theory uptake have little work to do. In a time when evidence is turning up that calls a received view into question, the line between crackpottery and respectable science is temporarily blurred.
As a result, the first and third stages of the scientific enterprise take on a larger role. If it is no longer clear (except in extreme cases) who the crackpots are and who the good scientists are, the question of who gets hired, who gets funded, and who gets published will have a correspondingly larger effect on the resulting science. Also, without clear criteria for distinguishing between good science and bad, the criteria actually applied will be more prone to subjective bias.
Unfounded charges of incompetence or fraud will be much more common, and more injustices will be done.6