A Failure to Discriminate: Querying Oxidizable Carbon Ratio (OCR) Dating
D. J. Killick, A. J. T. Jull, and G. S Burr
Oxidizable Carbon Ratio (OCR) dating is claimed by its inventor to be "an independent dating procedure. . .[that]. . .offers accuracy and precision in results, significant cost savings per sample, and meaningful age estimates for both archaeological features and the landforms which are the context of the archaeological site" (Frink 1997: 5). The stated basis of the method is that buried carbon in most archaeological soils undergoes (unspecified) changes of chemical state through time. The extent of these changes is measured by the Oxidizable Carbon Ratio, which is defined as the ratio of organic carbon determined by a wet oxidation procedure to total organic carbon, as measured by a low-temperature loss on ignition procedure (Frink 1992, 1994, 1995, 1997). OCR is claimed to have been calibrated against radiocarbon and historically dated features, from which an equation relating OCR to calendar age (see below) has been derived. The method is said to be applicable from the present to at least 10,000 years ago, and its precision is stated to a constant 3 percent of the age (Frink 1997: 3).
Since its introduction in 1992, use of the technique has spread rapidly. Some 4,000 OCR samples have been processed by Archaeology Consulting Team, of Essex, VT, the sole providers of OCR dating (D. S. Frink, personal communication 1999). The OCR Web site (members.aol.com/dsfrink/ocr/ocrpage.htm) lists domestic clients in 33 states as well as clients working in Argentina, Australia, Bermuda, Canada, the Czech Republic, Somalia, and South Korea. Domestic clients include federal and state agencies, museums, cultural resource management companies, and academics. The great majority of the clients are in the southeastern United States.
The rapid spread of acceptance of OCR dating by archaeologists concerns us for three reasons:
The remainder of this article elaborates on the second and third of these concerns.
Dating Equation and Assumptions
Frink (1992, 1994, 1995) has reported a novel method of dating archaeological strata by studying the "oxidizable carbon ratio" (OCR) of soils. In this technique, dried soils are measured for "easily oxidized carbon" using a wet dichromate oxidation (Walkley and Black 1935) and "total carbon" using a loss-on-ignition method (Ball 1964). The ratio of the amount of carbon in these two measurements is the OCR. By obtaining OCR ratios for contexts dated independently by radiocarbon and historical dates, Frink (1994) developed an empirical relationship between OCR and calibrated radiocarbon dates, based on an earlier different empirical formula of Frink (1992). He then further developed an empirical equation which he asserts can be used for dating many types of soils and archaeological deposits
where OCRdate is the estimated "age," Depth is the mean depth of the soil sample in cm, Tmean is the mean annual temperature (in °F), Pmean is the mean annual precipitation in cm/yr, Texture is the "mean texture" of the soil as defined by Frink (1994), pH is the pH of a 1:1 mixture of soil-water paste, and percent C is the total carbon in the sample (as measured by loss on ignition). The value 14.8888 is a constant, presumably derived from regression of OCR against historic and radiocarbon dates (Frink 1994).
Frink (1994) further asserts that the errors in OCRdate are small, 3 percent of the age, but without explanation. More recently, Pertulla (1997) published a list of some of Frink's OCRdate information with errors in age quoted of less than 10 years, also without explanation of the error estimates. Frink (1994) further asserts that "The OCR procedure may be used as an independent test of both the radiocarbon date and the archaeological assumption, verifying one or the other, or casting doubt on both." The claim that these relatively simple measurements and other parameters have small errors is not credible, and therefore warrants further discussion.
Issues of Accuracy and Precision
How can potential users of OCR decide whether the method is sufficiently accurate and precise for their needs? The only evidence provided for the accuracy of the method in Frink's published papers (Frink 1992, 1994, 1995, 1997) or on the OCR Web site is a single graphical plot of "calculated OCR date" against "calibrated 14C radiometric or documented age for cultural features". This is stated to be a plot of OCR dates obtained from features independently dated by radiocarbon, or else associated with features of known historical age. A regression line has been fitted to the 93 points plotted , and a correlation coefficient of .99 is claimed for the data. Since calibrated radiocarbon ages are age ranges, not points, it is inappropriate to use linear regression analysis on this data, but more importantly, the data underlying this plot are not published.We take issue with this proposed relationship and its assumptions. As mentioned, Frink (1994) presented a "calibration" of OCRdate with a correlation coefficient of 99 percent. For a parameter with 7 independent variables plus the fitting factor (constant), all of which have at least some errors, this assertion seems far-fetched. Surely, mean annual temperature and precipitation have significant errors. Surely the constant f derive from a fit of OCR to radiocarbon data must have an error yet this is specified to four decimal places! In addition, even if the 7 independently measured variables were totally without error, Frink would have to propagate the real error from the calibrated 14C measurements. We simply cannot see any way that these 7 independent variables plus the radiocarbon ages can give a total error in age of 3 percent.
Second, these 7 independent variables are in themselves questionable. We assume that Frink (1994) derived this equation from a purely statistical parametrization of a number of different possible variables. However, in order to determine a useful age equation, as eqn. 1 purports to be, we need to understand how these variables relate to the object of this exercise, the age of the sample. Frink's calibration curve on 93 soils from North American sites extends to over 10,000 years before present. We are asked to agree that mean current annual temperature (in F!) and mean current annual rainfall have any bearing on historical conditions for the sample. The texture parameter (as described by Frink, 1994) is ill-defined, and the inclusion of the square root of the pH and %C is simply inexplicable to us. We are unable to find any physical parameters which would explain the inclusion of these variables in this form. How can the square root of pH, itself the negative logarithm of the H+ ion concentration in a soil-water paste (in this case, the 1:1 paste mentioned) have any inverse linear relationship to age? How is the temperature in degrees F linearly related to any known physico-chemical process? The answer is: They are not.
If this were not enough, even the parameters actually measured in the OCR method leave us with many questions. Frink (1994, 1995) measured total carbon by loss on ignition (Ball 1964). We quote from the introduction of Ball (1964): "Loss on ignition has been widely dismissed as crude and inadequate as an estimate of organic matter or organic C content of soils . . ." This method does not measure total C. The weight loss on ignition includes loss of other volatiles and weight gain from oxidation of ferrous iron. Estimation of total carbon is best done by combustion and volumetric measurement of the carbon dioxide produced. Further, the identify of the "oxidizable" carbon measured by the dichromate titration method of Walkley and Black (1935) is not clear. In any case, there are many more precise ways to determine total C in samples, none of which are reported or discussed in Frink's publications.
Assumptions of the OCR Method
As has been stated by Frink (1994, 1995), the underlying idea of OCR dating is that older samples have less "oxidizable carbon" and that the organic material is converted over time into more refractory components. Although this idea is not without merit, we believe that it cannot easily be applied to soils. The residence time of carbon in different soil horizons and soil types is highly variable. (Indeed, Frink  points this out himself.) For example, Scharpenseel et al. (1996) showed that a wide range of radiocarbon ages can be obtained from different fractions of soils. For example, soluble fulvic acids in a soil from Sîllingen, Germany, were post-1950 A.D., but the base-soluble humic acids were ca.1560 B.P. and the insoluble humins were 2275 B.P. (standard errors omitted).
Chemical processes which must underly the conversion of carbon between the more and less "oxidizable" compounds do not obey linear relationships with temperature and concentration as in implied by eqn. 1. Chemical transformations are controlled by thermodynamics and kinetics. If a process is thermodynamically allowed, its actual rate of reaction will depend on the temperature and concentrations of the species present (e.g., Castellan 1964). Frink (1992, 1994) alludes to the exponential dependence of reaction rates on temperature. The simplest chemical reactions are first-order differential equations (Castellan 1964: 602), that is, ones where the reaction rate is controlled on the concentration of only one soluble species, or one species in the gas phase.
where N is the concentration (or partial pressure if a gas) of component N, which is a reagent in the process. Also, we know that the rate constant k has the form:
where ko is a constant, E is the activation energy for the process, R is the gas constant in the same units, and T is the absolute temperature in degrees Kelvin (Castellan 1964: 607) not, as in the equation of Frink, the temperature in degrees Fahrenheit!
These equations are provided to counter an important misrepresentation in Frink (1994). In this paper (subtitled "a proposed solution to some of the problems enountered with radiocarbon data"), Frink suggests that radiocarbon dating is limited in its accuracy because it is a random process. The article implies that OCR is a better dating method than radiocarbon because the chemical transformations underlying OCR are nonrandom. This is not so; as the equations above demonstrate, chemical processes are similarly random.
Certainly, there can be problems with radiocarbon dating in some situations. However, the parameters affecting radiocarbon are well known and to a large extent, well understood. The assumptions for radiocarbon dating are quite simple, and all revolve around the question of the initial value N0 in the radioactive decay equation:
where Nt is the number of 14C atoms present at time t, and N0 is the initial value. The amount of initial 14C, N0 , is known not to be constant with time, and this has been shown by the painstaking work over the last 30 years on the calibration of the radiocarbon time scale, with both dendrochronologically dated tree rings and, more recently, by comparison of 14C ages with U-Th ages in corals (Stuiver et al. 1998a,b). We know, for example, that these fluctuations limit precision in the range 1700 1950 A.D. Other differences in N0 due to a different initial reservoir of 14C are well known to occur in marine samples, lakes in calcareous deposits, where dead carbon has been added to the system ("the hard-water effect"), and occasionally in other systems. These effects are generally well understood. There are certainly other problems, and if we take an archaeological sample which consists of wood grown some years before the archaeological "event" in question, obviously radiocarbon dates the time of wood growth, not the event. We also know that N0 is higher for samples collected from material grown after 1950 A.D. due to addition of bomb 14C to the atmosphere. This, again, is well characterized and indeed, one can often date recent events very precisely. The point here is that the physical reality underlying the radiocative decay equation is well understood, and the validity of the equation is therefore universally accepted. For OCR dating, in contrast, we are asked to accept as valid a technique for which the underlying chemical mechanisms are unknown, and for which an equation is presented that departs in significant respects from long-established empirical laws governing all chemical reactions.
Summary and Conclusions
We have argued that that OCR cannot yet be accepted as a reliable method of dating soils and features of archaeological interest because: (1) the only available plot of calculated OCR dates against independently obtained ages (radiocarbon or historic) is statistically flawed and is based upon unpublished data; (2) the stated precision of the method (3 percent of age) is not credible; (3) no chemical basis for change in OCR with age has been demonstrated; (4) the equation that purports to derive the OCRdate from the OCR, depth of burial, soil texture, pH, total carbon, and present mean annual precipitation and air temperature cannot be correct; and (5) the methods used to measure the OCR (Ball 1964; Walkley and Black 1935) are crude and potentially inaccurate.
These flaws do not necessarily invalidate the concept of dating by oxidizable carbon ratio, and we are prepared to entertain the possibility that a more credible dating method might be developed around it. We do, however, doubt that so simple a method will ever be truly useful to the archaeologist. Several decades of work on soils has shown that they may contain several carbon-containing fractions of very different ages (Johnson and Johnson 1995; Sharpenseel and Becker-Heidmann 1992). Any dating procedure such as OCR that fails to discriminate between these various fractions is not likely to be able to distinguish events arising from human activity from those attributable to natural soil-forming processes.
In conclusion, we cannot at present recommend the addition of OCR to the archaeologist's battery of dating techniques. Beyond this, we are deeply disturbed by the willingness of so many archaeologists to adopt a technique that has not undergone the time-honored process of peer review at an appropriate scientific journal in this case, a journal of archaeological science or a journal of geochronology. One of us has argued elsewhere (Killick and Young 1997) that the education of archaeologists should include some mandatory training in the use of archaeometric techniques. In light of the rapid acceptance of OCR by archaeologists, we can only reemphasize this recommendation. ·D. J. Killick is associate professor in the Department of Anthropology, University of Arizona, Tucson. A. J. T. Jull and G. S. Burr work at the NSF Arizona Accelerator Mass Spectrometry Laboratory, University of Arizona.
References Cited, Killick et al.
Ball, D. F.
Castellan, G. W.
Frink, D. S.
1994 The Oxidizable Carbon Ratio (OCR): A Proposed Solution to Some of the Problems Encountered with Radiocarbon Data. North American Archaeologist 15: 1729.
1995 Application of the Oxidizable Carbon Ratio (OCR) Dating Procedure and its Implications for Pedogenic Research. Pedological Perspectives in Archaeological Research. SSSA Special Publication 44. Soil Science Society of America, Madison.
1997 Application of the Newly Developed OCR Dating Procedure in Pedo-Archaeological Studies. In Proceedings of the Second International Conference on Pedo-Archaeology. Anthropological Studies 10. South Carolina Institute of Archaeology and Anthropology, University of South Carolina, Columbia.
Johnson, C. W., and W. C. Johnson
Killick, D. J., and S. M. M. Young
Saunders, J. W., R. D. Mandel, R. T. Saucier, E. Thurman
Allen, C. T. Hallmark, J. K. Johnson, E. H. Jackson, C. M. Allen, G. L.
Stringer, D. S. Frink, J. K. Feathers, S. Williams, K.J. Gremillion, M.
F. Vidrine, and R. Jones
Scharpenseel, H. W., and P. W. Becker-Heidemann
Scharpenseel, H. W., F. Pietig, H. Schiffman, and P.
Stuiver, M., P. J. Reimer, E. Bard, J. W. Beck, G. S.
Burr, K. A. Hughen, B. Kromer, G. McCormac, J. van der Plicht, and M.
Stuiver, M., P. J. Reimer, and T. F. Braziunas
Walkley, A., and I. A. Black
The Scientific Basis of Oxidizable Carbon Ratio (OCR) Dating
Douglas S. Frink
Killick et al. have led the reader through a comparison of the OCR Carbon-Dating procedure to the laws of thermodynamics, linear statistical analysis, and established scientific practices, concluding with the "doubt [that] so simple a method will ever be truly useful to archaeologists." Beginning with a brief history of the OCR procedure as developed by this researcher (not inventor), I address each of these issues in turn. I will demonstrate that the development of the OCR procedure follows acceptable scientific practices and that the preceding article is more an example of human behavior than scientific critique.
Development of the OCR Carbon-Dating Procedure
In graduate school, I analyzed archaeological soil features using various soil analytic techniques including the Ball, Loss-on-Ignition, and Walkley-Black wet oxidation procedures. These two procedures are commonly used in soil analysis to characterize the organic carbon content in agricultural soils. Although these procedures do not yield the same values for organic carbon, the results obtained from analyses of agricultural topsoils form a constant ratio (Hesse 1971).
Results from the Ball and Walkley-Black procedures produced abnormal ratios during my analyses of hearth feature samples (predominantly containing charcoal) from local Native American sites. The variability in the ratio appeared linear when plotted against the 14C dates obtained from these features. Additional archaeological features evidencing more potential variables were then sought to test the preliminary observation (Frink 1992). Variability in the ratio was found to be related to soil texture and depth (factors that affect the rate of oxygen diffusion), temperature and moisture regimes (factors that control the rate of biological processes), and soil pH and concentration of organic carbon (factors that contribute to the viability and speciation of biological populations in the soil). The degree of effect for both pH and percent organic carbon was found to be nonlinear, and best modeled using the square-root of these values.
The seven variables (time, texture, depth, temperature, moisture, reactivity, and total organic carbon), along with a deduced constant (k) to account for other unidentified variables, form the OCR Carbon-Dating formula (Frink 1992, 1994). The equation used to calculate the OCRdate age estimate describes a dynamic system similar to Dokuchaev's and Jenny's models of soil genesis: Soils are the result of the interdependent dynamics of climate, biota, parent material, relief, and time (Buol et al. 1980; Jenny 1941). The variables deduced for the OCR formula reflect the variables of climate (temperature and rainfall), biota (soil depth, percent organic carbon, pH), parent material (texture, pH) and time (determined from documented events and 14C age estimates). Relief has not been found to be a major variable, as the samples are from flat-to-gently sloped terrain (Frink 1995).
The OCR formula, presented as a hypothesis to explain the variability in results found between the Ball and Walkley-Black procedures, continues to be tested using archaeological and pedological samples from throughout the world.
Comparison to the Laws of Thermodynamics
The OCR Carbon-Dating procedure differs epistemologically from radiometric carbon-dating procedures. Radiometric carbon-dating procedures measure the decay of unstable carbon isotopes, following a classical physics model of entropy. The OCR procedure does not directly measure an intrinsic characteristic of the soil organic carbon. Rather, it models the dynamic and nonlinear soil system and the relative reactivity of the soil's organic carbon within that system. Dynamic systems resist entropy by organizing and maintaining themselves at a distance far from equilibrium. The OCR procedure describes an evolving pedogenic system.
This difference in epistemology should not be misconstrued as describing two separate and contradictory sciences, nor as a false dichotomy. They are alternative ways of viewing different aspects of a common world. Such a world can be legitimately viewed through the reductionist approach, describing the behavior of individual particles, and through a holistic approach, describing the collective behavior of the interdependent parts composing a system. However, one approach cannot be used as the criterion for judging the validity of the other.
The ideas that form the basis of the OCR procedure are not fundamentally new or unique. The contention that the OCR Carbon-Dating procedure "departs in significant aspects from long established empirical laws governing all chemical reaction" (Killick et al.), is based on a narrow, and biased, concept of science that is founded exclusively on systems at or near equilibrium and governed only by entropic processes. The variables used in the OCR Carbon-Dating procedure equation, purported to be incorrect by Killick et al., directly translate into measurable aspects of the five factors of soil formation, the dominant model in pedogenics (Buol et al. 1980; Jenny 1941). Why these variables "cannot be correct" has not been proven: it is simply a stated and unsubstantiated belief.
The argument presented by Killick et al. serves to demonstrate Thomas Kuhn's (1996) thesis on the mechanisms by which the dominant paradigm within an established discipline will resist change resulting from and caused by new ideas. Among the practitioners, it is universally assumed that the dominant paradigm is unquestionable and correct. A crisis occurs when a new idea appears, and it becomes necessary to declare that the new idea is obviously false. The "scientific gaffe" of daring to compare the new OCR Carbon-Dating procedure to the established 14C Carbon-Dating procedure is such a case. Through quotations out of context, the reader is led to the inevitable conclusion that the OCR procedure must be false. Instead, the referenced article (Frink 1994) discusses the limitations of both the 14C and OCR procedures. The article concludes that the combined use of both procedures to obtain corroborative data from independent analytic processes may be scientifically prudent.
Comparison to Linear Statistical Analysis
Complex dynamic systems are synergistic in behavior and physiology. The sum of the parts do not describe the whole. Prigogine and Stengers (1984: 171) note that "One of the most interesting aspects of dissipative structures is their coherence. The system behaves as a whole, as if it were the site of long-range forces. In spite of the fact that interactions among molecules do not exceed a range of some 10-8 cm, the system is structured as though each molecule were 'informed' about the overall state of the system." In this informed condition, inherent errors of the parts are either accentuated through positive feedback or mollified through negative feedback. For example, as anthropologists, we may expect that any given human will have the physiological and behavioral characteristics that are considered normal with variability described by a small margin of error, even though the sum of the inherent errors of the billions of individual cells in that person would express an extremely large margin of error.
The soil system, as modeled by the OCR Carbon-Dating procedure, is a complex dynamic system in which the errors of the parts are mollified by their interactions. Fractionation of the organic carbons into humins and fluvic and humic acids prior to analysis, as suggested by Killick et al., is unwarranted. The analysis portrays the behavior of the humic material, as a whole, within the dynamic soil system. While I agree that more work, based on a larger database, is still needed to establish a statistical proof of the margin of error for the OCR procedure, the ± 3 percent error is validated by existing results.
Comparison to Established Scientific Practices
The OCR Carbon-Dating procedure is a new, and still experimental, procedure that has the potential to provide reasonably accurate and precise age estimates from organic carbon within an aerobic soil context at an affordable cost. The OCR formula, the specific changes in carbon that are being recorded, and the causal agent(s) of these changes are hypotheses. The phenomena of biodegradation and recycling of organic carbon within an aerobic soil context exist independently from the hypotheses posed by this researcher. The present research entails testing these hypotheses in a number of unique pedological and anthropological contexts in different soil series throughout the world.
Field trials are being conducted in conjunction with federal and state agencies, museums, Cultural Resource Management companies and academicians (not to mention other archaeologists, pedologists and geomorphologists) alluded to in the preceding paper. These people have the unique pedological and anthropological contexts necessary to test these hypotheses, and the need for an analytic method that may provide valuable information about the samples' contexts. The conclusion by Killick et al.that archaeologists are not capable of participating in such trials because they are untrained in archaeometric techniquesis patronizing and self-serving. Their stated concern with the "rapid spread of acceptance of the OCR Carbon-Dating procedure by archaeologists" underscores this point.
In critiquing the methods used, Killick et al. state that "the methods used to measure the OCR are crude and potentially inaccurate." This conclusion is based on an unconscionable abuse of the literature. They quote Ball's statement of the problem, that "Loss on ignition has been widely dismissed as crude and inadequate . . ." (Ball 1964: 85), without revealing that Ball then provides his new procedure addressing these shortcomings. The procedures used in the OCR analysis are not the most complex available, nor do they purport to recover all data. The eloquence of the OCR Carbon-Dating procedure is in its ability to accurately describe the behavior of complex soil system through the use of simple, accessible, and affordable analytical methods.
Furthermore, the aspersion that the OCR Carbon-Dating procedure has not yet appeared in a peer-reviewed journal is false. Although referenced several times by Killick et al., the peer-reviewed and published paper, Application of the Oxidizable Carbon Ratio Dating Procedure and Its Implications for Pedogenic Research (Frink 1995), is ignored in their discussion of peer-review of the OCR procedure. Publications by the Soil Science Society of America are conspicuously missing from their list of "appropriate scientific journals"; however, I suggest that a soil procedure should be critiqued by a peer group capable of judging its merits within the discipline of soil science rather than radiometrics.
I do not operate under the illusion that the initial hypotheses defining the OCR Carbon-Dating procedure will remain unchanged by future data and by the research of others. Were the procedure to remain unchanged, it would truly be an event unparalleled in the history of science. The present direction that must be taken for scientific inquiry and critique is clear. More data must continue to be gathered, and other, independent researchers must become involved. Rather than arguing the semantics of different disciplines (a course of action that leads to censorship), scientific trials demonstrating whether the OCR procedure can, or cannot, be duplicated would be more productive: in other words, Science. I renew my offer made to Dr. Killick in the personal communication referenced in the preceding paper. "Beyond the concern that this new method is becoming widely used before it has been critically reviewed, should be the concern that the results of this procedure still need to be independently duplicated. Toward this eventual trial, I maintain a collection of subsamples of the nearly 4,000 samples processed to date. I would welcome you, and/or others, to undertake this study" (D. S. Frink to D. J. Killick, letter, July 8, 1999).
Finally, I recognize that it is normal for any new idea to undergo a process of ridicule, critique, and, eventually, testing, prior to its ultimate acceptance or rejection. The history of the 14C Carbon-Dating procedure, as briefly enumerated in Killick and Young (1997), underwent a similar process during the first decades of its use. While this process may be normal and well illustrated in the history of science, it is important to distinguish between those aspects of this process that are strictly in the realm of human behavior and those that are scientific in nature. The dominant paradigm and its supporters will protect themselves from the influence of new ideas and resulting change. This resistance is commonly accomplished through control of funding and communications, but such resistance is clearly not scienceit is human behavior. ·Douglas S. Frink is with OCR Carbon Dating, Inc., in Essex Junction, Vermont.
References Cited, Frink
Ball, D. F.
Buol, S. W., F. D. Hole, and R. J. McCracken
Frink, D. S.
1994 The Oxidizable Carbon Ratio (OCR): A Proposed Solution to Some of the Problems Encountered with Radiocarbon Data. North American Archaeologist 15: 1729.
1992 The Chemical Variability of Carbonized Organic Matter through Time. Archaeology of Eastern North America 20: 6779.
Hesse, P. R.
Killick, D. J., and S. M. M. Young
Prigogine, I., and I. Stengers
back to top of page