Mariusz S. Ziółkowski, Mieczysław F. Pazdur †, Andrzej Krzanowski, Adam Michczyński

Some problems of the radiocarbon dating of the Central Andean Cultures

Mariusz S. Ziółkowski

It is obvious that the basis of all archaeological research is to establish the chronology (relative and absolute) of the analyzed phenomena. In Andean archaeology, absolute dating is mainly accomplished by the C-14 dating method, and the dates from the Andean sites appeared in the very first radiocarbon laboratory lists, in the early fifties, i.e. more than 40 years ago. The thermoluminescent, paleomagnetic and obsydian methods are considerably less frequently used. Unfortunately, in spite of very prominent advancement of the exploration of Prehispanic cultures, and the above mentioned fact of the very early application of the C-14 method in the Andean archaeology, the absolute chronology for this region is still the object of many controversies.

This situation issues, among others, from the fact that until now a complete catalogue of radiocarbon dates for this region did not exist. The data were dispersed in many publications, in various countries, some dates have never been published and were available only in laboratory archives or remained in the possession of individual researchers. This situation and the complications arising from it may be best illustrated by the fact that the fundamental work „Chronology in New World Archaeology” (Taylor, Meighan, ed.1978) does not contain a chapter on chronology of the Central-Andean cultures.

But the idea of a general catalogue of radiocarbon dates for the Andean cultures is not new. In 1951 Junius Bird published the first list, with about 20 dates. In the following years some efforts have been made to establish catalogues for particular cultures and/or regions (Engel 1966; Bischof 1972; Ponce Sangines 1972; Burger 1981; Watson 1986) or for the entire considered area. Later R. Ravines' catalogues were especially interesting (Ravines and Alvarez Sauri, 1967; Ravines 1982). The largest one (Ravines 1982) contains 752 C-14 dates. This one, according to our estimation represents about 25% of all the dates available at that moment, i.e. in 1994.

Of course the problem does not end in simply listing and publishing full information for example from Radiocarbon, among other reasons because his review publishes only a low percentage of all the datas produced by numerous laboratories now. Until the early sixties, it was still possible to publish an almost complete list of currently produced new datings in Radiocarbon (or earlier in Science). Rapidly though, it appeared that the growing number of dates produced every year (several tens of thousands, obviously not only for archaeology) could not be presented in one review any more. In the eighties appeared the idea of creating a general worldwide data base of all the radiocarbon datings, the so-called IRDB (International Radiocarbon Database). This project has been so far reported, principally for need of financial support. In this situation the alternative solution was to start establishing such a database limited to one country only or a well delimited cultural area, keeping in mind though the possibility of a future unification of such regional databases in more general system. As far as Andean region, as concerned a lot of new problems arise, with which our predecessors have been already confronted. The most obvious of those problems is the data selection and clear presentation (related to the datings) for archaeological purposes. And so 3 main problems connected with data selection criteria arise:

  1. What kind of information should a catalogue card contain?
  2. How to determine precise territorial and/or cultural and temporal limits for the enquiry?
  3. How to estimate reliability of dates, depending on the kind of sample material, the method of dating employed, the general reliability of the laboratory, etc.?

Before coming to the resumed presentation of the contents of our Catalogue, it is necessary to offer some information about our methods for collecting data.

As we already mentioned, we made good use of previous catalogues such as Engel 1966a, Ravines and Alvarez Sauri 1967 and others, as well as the lists of datings from Radiocarbon, Science and Archaeometry, but not uncritically, as we will explain further below. We also went to some 80 radiocarbon laboratories in the World, requesting them to send us their records related to the radiocarbon datings for Bolivia, Ecuador and Peru. We received positive answers from about 45 of them; in which most of the datings had already been published in the mentioned reviews. But in some cases, as for example in that of the radiocarbon laboratory of the U.S. Geological Survey (code W), the major part of the datings send to us were still unpublished. We received especially important assistance from the Centre des Faibles Radioactivite Laboratoire Mixte CNRS-CEA in Gif sur Yvette, France (codes Gif and Gif A), which offered us the access to all the records, information and publications collected in their archive.

We soon realized that it would be suitable and fruitful to confront two sources of informations: the archaeological reports concerning the context of the samples with the detailed laboratory reports about the method of datation employed and its final results. For that reason, when we finished collecting the data of our Catalogue, we addressed ourselves once more to the radiocarbon laboratories, which had produced the datings we collected, requesting from them a confirmation of the results we reported. It was especially important for datings which had not been previously published in Radiocarbon and were known only from the archaeological publications. The datings which have been confirmed by the laboratories in such a way carry in our Catalogue, in the field „References”, the code of the laboratory accompanied by the symbol „&” (for example W&, see W-310 in the Peru section of the Catalogue). This verification led us to a series of corrections; in the Index 4 we list all the datings in which appeared significant differences (> 10 yrs in the conventional age BP or the standard error) according to the sources they were collected from. Some especially striking differences are discussed below. We are well aware of the fact that, in spite of a keen cross-control we made, probably the present Catalogue is not exempt of such mistakes as well. Therefore we would be most thankful to the Readers to inform us about any error they might come upon. We plan to publish regularly eventual corrections and supplements to the Catalogue in a special Bulletin.

As it was said at the beginning, our aim was not merely to report the radiocarbon datings available in the literature but rather to offer the possibility of comparing the results from different laboratories, produced at different times. It is precisely on this point we came upon with the first problem that may be defined as

1. Corrections, improvements and calibrations (or what was indeed reported as a C-14 dating?)
  The detailed description of the present procedures in the field of the radiocarbon dating methods are discussed in the chapter of Mr and Mrs Pazdur (Andes, 1994:25-62). Therefore I will limit here myself with some consequences of the historical changes in these methods immediately perceptible in the data we collected.
  a) The problem of the exact values of the half-life of the C-14 isotope
  At the turn of the fourties and the fifties series of measurements of the half-life of the C-14 isotope was made, on the basis of which H. Libby established the mean value at 5568±30 years. This value is commonly known as so-called Libby Half-life. More exact measurements realized in the early sixties led to a re-evaluation to the improved estimation of 5730±40 years. But for the sake of keeping a common scale in the publicated results, it was agreed at the 5th International Conference on Radiocarbon Dating in 1962 to maintain the Libby Half-life as the conventional basis for the calculation of datings. It means practically that all the C-14 dates calculated conforming to this convention are reduced by approx. 3%. But this difference has been considered as negligeable compared to the degree of exactitude of the counting methods available at that time, and the especially to the magnitude of the average standard errors. Anyway, it is possible to correct this difference, by multiplying the conventional age BP (calculated according to the Libby Half-life in reference to 1950) by the coeficient 1.03 (wihout changing the value of the standard error). Let us underline that all the modern computer programs for calibration (among which the software elaborated in Gliwice and used for the calibration of the present Catalogue) automatically carry out this correction. In the literature on Andean archaeology it is still possible to find datings reported in both versions, i.e. counted with Libby Half-life together with Half-life=5730 years, or even in the second form only. One ought to remember this last example, especially when using computer calibration programmes such as those by Stuiver and Reimers (1993) in to avoided the second correction (by the 1.03 coeficient) of a date already produced with the Half-life=5730 yrs.
  b) The problem of the accuracy of the radioactivity measurement
  As we mentioned above, the first radiocarbon datings concerning the Andean region are almost as old as the radiocarbon method itself. Already at the end of the forties and the beginning of the fifties datings were made for sites of such importance as Huaca Prieta, Paracas Necropolis, Huaca Negra, Cahuachi and others. Leaving appart the mentioned problem of the exact duration of the Half-life of the C-14 isotope, it is necessary to recall the fact that those first datings were produced with the so called „solid carbon” method, rather imprecise and completely abandoned nowadays. This method, which will not be discussed in details here (as it is presented in the chapter by Mr and Mrs Pazdur, Andes 1994:55 and ss) led to significant imprecisions, leading to standard errors of 200 yrs or even more. Moreover the results of succesive countings of the same sample were sometimes very different. Let us take as an example the case of the datings L-116a, produceed on a single sample of charcoal from Huaca Prieta. The value reported in our Catalogue is 3780±100 BP according to the Laboratory list . But this sample was previously dated in the same laboratory (i.m. Lamont Observatory) by the black carbon method at 3650±400 BP and by the University of Chicago at 4298±230 BP (see the datings L-116a and C-598 in the Peru section).

At that time, as well as now, some authors used to present the mean value obtained from different countings, although others present only the result of one of the countings, which they suppose more acceptable. For example the dating C-271 from Paracas Necropolis which appears as 2257±200 in our Catalogue, conforming to the mean value given by Arnold and Libby from two different countings (2190±350 and 2336±300) is reported by Ravines from the first value only at 2190±350 (compare Arnold and Libby, 1951: 119 with Ravines 1982: 170). Even more significant is the case of the datings made from an ensemble of atlatls found in a grave at Cahuachi (Nasca). The dating C-521 is reported in the laboratory list as 2211±200, as an average value between two very significantly different countings: 1681±250 and 2477±200. Still for example H. Silverman recently published only the first counting, for the only reason that it fitted better with the ceramic Nasca 3 found in the same grave, and thus rejected the second as aberrant (Silverman 1993: 39). In addition to this question of differences in results of countings we must add the problem of

  c) the „Suess effect” and the standard of radioactivity
  The datations by the radiocarbon method were initially carried-out on the assumption that the percentage of the C-14 isotope in the atmosphere had not changed for millenia. The standard reference was thus established as the contemporary measure of the C-14 in the atmosphere. But it appeared rather quickly that such was not the case. Since the middle of the 19th century and the industrial age when a quantity of C-14 free CO2 was ejected into the atmosphere (resulting from the combustion of fossil coal and petroleum), the proportion of C-14 diminished compared to the global amount of carbon in the atmosphere. It was established that this diminution amounted to 2.5 % in 1950, compared to pre- industrial times (see Suess, 1955: pass., Pazdur and Pazdur 1982: 26), and for that reason the age of the datings made on the basis the „atmospheric standard” was underestimated by the same quotient of about 2.5 %. It is for that reason that in the laboratories lists from the fifties appear datings corrected by the so-called „Suess effect” (or „industrial effect”) ranging from +80 to +200 and even more years. A significant example appears in the case of L-122a from Huaca Negra, reported 3150±90 (as a mean value of two countings: 3200±100 and 3050±90) which, according to the „Suess correction” produces 3400±90 (see L/2: 410). For that reason in archaeological literature we sometimes find, under the same code, various datings, depending on whether the „Suess correction” was taken into account or not. A good example of this situation probably appears in the series of datings L-268a (1710±80), L-268e (900±70), L-268f (970±70), L-268g (1200±80), L-268h (1430±80), produced from samples collected by Strong from different sites in the Nasca Valley. Although in the original list of the Lamont Laboratory those dates appear as quoted above, in the later 6th list of the radiocarbon laboratory of the University of Pennsylvania in 1963 all those datings were reported as being 200 years older, i.e. 1910, 1100, 1170, 1400 i 1630 years, with the unchanged respective standard errors. It seems we are confronted here with the „Suess effect” taken indeed into account although not mentioned. Unfortunately we were not able to confirm this supposition for the reason that the archives of the Lamont Laboratory of those years are not available.

Of course the problems concerning the „Suess effect” vanished with the introduction of the NBS Oxalic Acid radioactivity standard, but anyhow, a lasting effect of this situation is perceptible in the datings from the fifties, which are still rather difficult to compare with those produced later, both because of the inaccuracy of the countings as well as that of the radioactivity standard. Therefore although we present a dendrochronological calibration of those first datings from the fifties, together with the later ones in the Catalogue, we must keep in mind that they are just mere approximations and should by no means serve as a confident base for absolute chronology.

The next cause for modifications of the datings leading to other problems with the comparison of results produced in different periods and which must be taken into account as well is:

  d) the „isotopic fractionation” commonly called the „dC-13 correction”
  This problem is discussed in detail in the chapter by Mr and Mrs Pazdur (loc. cit.) therefore I shall only content myself with mentioning that this question refers to the slightly different amounts of C-14 contained in various living organisms. This difference is estimated by the help of the similar proportional absorbtion by living organisms of C-12 and C-13 isotopes, presenting the advantage of being more easily measurable.

The correction base on the „isotopic fractionation”, sometimes unclearly called „calibration” (a term which may lead to confusion with the „dendrochronological calibration”) is peculiarly significant when applied to datings made on samples of sea shells or some kind of grains such as maize. Depending of the sort of shell analyzed this correction may reach 400 years and for maize around 245 years, etc. (see the exact values in table 2 of the Pazdur's paper, loc. cit.). As a simple illustration let us consider the abstact case of a similar dating results of 3150±50 BP, obtained from three different samples: shell, maize and humus. The corresponding correction to dC-13 would produce respectively 2700±50 BP, 2865±50 BP and 3130±50 BP. This situation renders the most difficult use of old datings especially when the material of the sample is not reported.

The next source of confusion results principally of the so called „human factor”, concerning

  e) the „BP, AD, BC” and the treacherous question of the year „0” (Radiocarbon Time).
  In the very beginning of the radiocarbon datings it was accepted that the conventional radiocarbon age of the sample is expressed in years , counted backward from the year when the dating was produced. This habit was necessaryly adding by itself a new parameter of corrections as results produced in successive years were obviously refering to different „0” base dates. The importance of this variation factor was negligeable, considering the broad margins of inexactitude in the datings produced in those years (see above), especially as soon (n 1962) a general convention agreed with establishing the „0” point of C-14 datings as the year 1950. For physicists the question was solved, but some problems remained with reports of the C-14 datings in archaeological publications. In archaeology the convention was to present chronological data in terms of years BC/AD. Theoretically, the problem of conversion from BP to BC/AD scale is very easy, consisiting simply to substract 1950 years from the BP dates, given by the laboratories. Still, this elementary operation prove to be sometimes distrustful, because complementary complications emerged from the traduction of the English terms BC, AD and BP to Spanish, where theoretically the English „BC” (before Christ) corresponds to „a.C.” (antes de Cristo), whereas „AD” (Anno Domini) corresponds to „d.C.” (después de Cristo). All this become even more intricated with the appearance of another, rather enigmatic abreviation „AC”, which in some publications seems to mean „AD” and in others is interpreted as „BP”. For the sake of the clarity of the present exposition, I will not discuss the additional differences in the use of „BC, AD and BP” abreviations in minuscule or capital letters here....

Unfortunatelly the mistakes in interpreting the datings, as the results of misunderstanding of the different abreviations mentioned above, clearly distort the Catalogue of Ravines and Alvarez Sauri from 1967 and find their consequences in the 1982 list by Ravines. As an illustration, the dating L-268e, the correct value of which reported in the laboratory list is 900±70 BP (see L/3: 163) is first erroneously reported as 900±70 AD (i.e. 1050 BP - Ravines, Alvarez Sauri 1967: 22) and later appears as 2850 BP (Ravines 1982) as a second confusion by the author, reusing his former erroneous interpretation with a new intervertion of 900 AD into 900 BC (i.e. exactly 2850 BP). To continue, let us consider the well known date W-310, produced on a sample closely associated with the giant ground drawings on the Nasca desert. Its correct value, confirmed to us by the Laboratory, is „< 200 BP” (namely: younger than 200 radiocarbon years Before Present) and not 1757 +\- 200 BP (i.e. AD 193±200), as it appears in the Ravines catalogue (Ravines, 1982: 174). The probable origins of this error are presented in the commentary to W-310 (Andes 1994:479-480).

This kind of misinterpretation touches sometimes a whole serie of datings, for example in the same catalogue by Ravines, all the dates from the Guitarrero cave appear in the wrong column, i.e. in BC instead of BP, in such a way that all of them agree in being older by 1950 years (see Ravines, 1982). In this case we prefer to believe in an error at the printers rather than in a mistake of inattention by the Author.

In our Index 4 we present all the datings in which we observed a significant difference in the reported values (> 10 years) of the age BP or the standard error according to various sources.

It is worth mentioning that the conversion of conventional radiocarbon datings BP (Before Present) into the BC/AD scale was acceptable at a time when it was believed that the radiocarbon age corresponds approximatively to the calendric age (expressed in solar years). At least from the early sixties it became evident that this was not the case since the assumption by Libby that the percentage of C-14 in the atmosphere was stable over a long period of time proved not to be true. In that situation it appears that the radiocarbon age of the sample rather reflects its radioactivity, and should serve only as a base for the calculation of its probable calendrical age. This point leads us to the next group of questions related to

  f) the dendrochronological calibration.
  The first reconsideration of the previously accepted idea of the stablity of the C-14 percentage in the atmosphere came with the recognition of the „Suess effect”. With the improvement of the C-14 measurements it appeared that the datings of some samples of well known age (e.g. of. from the Early Dynasties of the Ancient Egypt) gave results too short by at least a few centuries, even when taken into account all the corrections known then.

In consideration of that point, precise dendrochronological researches (let us mention here the important contribution of C.W. Ferguson) led to the establishment of the so-called „curves of calibration” based on the changes of content of C-14 in the atmosphere during the last millennia. Thanks to this, it is possible now to convert the conventional radiocarbon age of a sample into the most probable calendrical age. This question is discussed in detail in the chapter by Mr and Mrs Pazdur as well as that of Mr Michczynski (Andes 1994), therefore I shall only mention here that before the general acceptance as a standard, in 1986, of the Stuiver's and Reimers' calibration curve, other systems of calibration, based on preliminary and rough curves, had been in use since the early seventies. Some of those systems had been applied in Andean archaeology, especially the systems of Klein et al, 1974, Damon et al., 1979 and Suess et al. 1982. In the Catalogue ANDES those earlier calibrations are quoted by us in the Archaeological comment or Laboratory comment, in order to allow for the comparison with the results of more recent methods (see the commentary to the Catalogue). In the specialized literature we sometimes encounter, fortunately not so frequently, cases in which the author only reports the calibrated value, without presenting the basic conventional age BP. This renders the rather difficult use of such datings, especially when the calibration method used is not mentioned (see the datings SMU-1787 and SMU-1788 in the Peru section).

Another specific problem related to the dendrochronological calibration is that of the so called „correction for the Southern Hemisphere”. From the last years publications of the analysis of samples of well known age from South Africa, it was noticed that the content of C-14 in the atmosphere of the Southern Hemisphere is slightly different (higher by ca 3%) of that in the Northern Hemisphere, a situation leading to a constant difference of about 30 - 40 radiocarbon years. To make things easier let us consider that for any conventional dating made from a sample from Peru, for example, we should add 30 years (i.e. 2050±30 conv. BP must be converted to 2080±30 conv. BP, etc.) and it is only this second corrected value which should be taken as a base for the dendrochronological calibration. So far everything would be fine, but we have to take in account the fact that atmospheric and climatological limits of the sub-equatorial zone are far from being as strictly delimited as the geographical equator, moreover this zone is a region of significant atmospherical exchanges between the Northern and Southern Hemispheres. Therefore it is possible that the differences in the percentage of C-14 in the atmosphere when passing from one Hemisphere to the other, occur rather gradually than abruptly. So, practically, which sort of correction we should apply to a dating made on a sample coming from an archaeological site located at the Latitude 2ºS.? Should it be 40 years? or maybe only 15? Or even only 10? And consequently, what should be the correction for a sample collected at the Latitude 10ºS., etc? Precisely because of these doubts and in order to avoid a still greater confusion in the data, we have decided not to introduce this correction, until the uncertitudes of its real value for the region between Lat. 1 N and 20 S, can be solved. Still, it is necesary to remember that in some laboratories this correction of 30 or 40 years is introduced systematically. This may not have a too great importance for datings of the order of 6000±100 conv. BP, but becomes significant for high-resolution datings (with a standard error of ca 20 yrs) from later periods (i.e. Late Intermediate Period or Late Horizon). In the Catalogue ANDES there are a few datings with the „Southern Hemisphere correction”, it is the case of, for example series of five datings from the Yumes site in Ecuador (AA-1760, AA-1762, AA-1763, AA- 1764, AA-1765).

I should now dedicate a few words to the above mentioned question of

  g) the standard error or sigma and its place in the archaeological literature.
  It may seem somehow exagerrated to mention such a trivial point here, but I consider it appropiate to remember that a conventional C-14 dating should be reported in the following way (an example chosen arbitrarly): 3500±70 BP. This means in practice that the true value of the dating (still in radiocarbon and not solar years) is situated between 3430 and 3570 conv. BP at the probability level of 68%, whereas with the probability level of ca 95%, it would range between 3360 and 3640 conv. BP (amounting to value of 2SD on each side of the mean value, i.e. 3500 BP). It is only the range of values between those limits (and not the mean value alone!), that should be the object of dendrochronological calibration. Unfortunately, in some publications issued in the times when archaeologists (and not only them) were over-confident in the exactness of the C-14 method, we are given chronologies based only on the mean values (in conventional radiocarbon years) with the complete omission of the corresponding standard errors. It is according to this convention one of the earliest lists of datings was publilshed, namely that of F. Engel, 1963, in which indeed the author spoke about the standard error in his Introduction but in the list itself the datings are reported only as mean values. In the archaeological literature we find a quantity of datings for which there is simply no mention of the value of the standard error, in our Catalogue such is the case of the series of datings produced by the now inactive C-14 laboratory of the Science Museum of Victoria, Australia (see V-275 and ss.). Of course in similar situations of the complete lack of information about the standard error makes any kind of serious dendrochronological calibration impossible.

At the end of this paragraph in which we exposed some problems associated mainly with the archaeological interpretation (sometimes imprecise or even erroneous) of the data produced by the laboratories, we would like to continue with the delicate matter of

  h) corrections, inaccuracies and mistakes in the laboratories themselves.
  A fact which is commonly known although rather discretly kept is that among the C-14 laboratories there was (an still is) a significative difference not only in the accuracy of the measurements (depending on the sophistication of the instruments used) but simply in the viability of the respective laboratories, depending on other factors. Those factors precisely constitute the delicate matter I mentioned here above and without wishing enter into details I simply recall here that with the aim of establishing something like a general standard of viability, there are periodical intercomparisons between several laboratories, consisting of dating of the same samples of well known absolute age. The results of such comparisons are published in Radiocarbon. Of course the participation in such verificatory programmes is not compulsary and therefore not all the active laboratories take part in them, but all the same, there is an unformal but well accepted ranking of C-14 laboratories according to the exactness of their results. In the calibration program published by Stuiver and Reimers in 1993 we can even find a special coeficient named „k” (ranging from 1 to 4) which is used to mulptiply the standard errors (before the calibration), according to the estimated degree of accuracy of the results produced by respective laboratories, the ceoficient „1” being attributed to the most reliable ones, of course. It is probably useless to mention that such an idea of „formal ranking” (and in such a way) was rather coldly received, and we have never had heard about a laboratory which would use another coeficient than „1” for its own results...
Anyway it is very important to report, the corresponding code and laboratory number together with any C-14 dating. This is essential information, even putting aside the above mentioned question of viability, allowing the user to address directly the laboratory with any question concerning a particular dating for more details. Unfortunately there is rather too frequent custom even in the latest literature, to forget to give the complete laboratory reference or report them too imprecisely. In the present Catalogue to these datings are atribuited an arbitrary code „ZZZZ”. I should also remember that some laboratories during the long period of activity changed their identification codes, and, sometimes, even the numbering of their datings. For example the first laboratory codes of Gif (Gif-sur-Yvette) were Gsy and Sa. In the case of GrN (Groningen) it was previously Gro and in that of GaK (Gakushuin) it was G., etc. In the particular case of the last one, the change of the code concerns also the numbering of datings, for example the reference Gak-106 corresponds to the ancient G-606, etc.

Another problem is that sometimes preliminary results from laboratories reach the first archaeologist and are published without waiting for the final results, different in some cases.

A good example of such situation concerns the dating GX-1127 from Chavin de Huantar, reported by H. Amat O. as 3150 BP without the corresponding standard error (Amat, 1976: 544). When we addressed ourselves to Geochron Laboratories with a request for the missing value of the standard error, we received an answer that the analyzed sample had been to small to allow a dating at all (letter of Dr G. Wilcox, Geochron Laboratories, April 27th, 1994). So it seems to us that H. Amat O., when publishing the dating 3150 BP reliate on a preliminary laboratory estimation, without waiting or not receiving the final report. Independently and enigmatically, the same dating GX-1127 appears sometimes as 3077 BP or 1127 BC and this, as we suspect, is most probably the result of a confusion of the dating itself and its code number, namely 1127 (see the commentary to GX-1127, in Andes 1994:285). In addition I have to mention also that it sometimes happens that the laboratories are unable to confirm or submit more details about datings they produced, especially at the turn of the fifties and sixties, mainly because of some gaps in their archives. In some cases it is not even possible to identify the laboratory number of the dating, for example, among others, the whole series of datings from Cerro Sechin produced in the radiocarbon laboratory of the Pontificia Universidad Catolica del Peru in Lima (see PUCP - XX1 - XX16, in Andes 1994:388 and ss.), which by the way does not appear, to our knowledge, in any of the lists issued by this research center.

The procedure of the analysis of the samples from their submittion to the laboratory until the publication of the results is anyway so complicated, that even the most reknowned centers are not exempt of some errors. A well known case is that of a series of ca 400 datings produced by the British Museum C-14 Laboratory between 1980 and 1984, with a systematic error of about 200 yrs due to a technical deficiency. Immediately as the error was detected, the Laboratory informed about it in the columns of both Antiquity and Radiocarbon and subsequently published a new list of corrected datings. Unfortunately, it was too late to establish an exact correction for all of the ca 400 datings, a problem was even more complicated by the fact that some of those erroneous results had already gone their own way quoted in various archaeological publications. For the data collected in our Catalogue, this concerns four datings from Guarumal, Ecuador (BM-1682R, BM-1684, BM-1688, BM- 1689) and one from Cusichaca, Peru (BM-1633R). Taking into account the possibility of errors ( not frequent, it is true, but still probable) from the laboratories, it is recommended in the case of findings of particular importance for the chronology of the investigated site, to divide the sample in fragments in order to send it to several laboratories for independent datings.

This is precisely what we have done with the case of an extremely interesting finding of a preceramic burial in Cahuachi, Nasca, Peru: after five coincidental datings obtained from the Gliwice Radiocarbon Laboratory, the Laboratory itself sent parts of the samples to three other laboratories (Gif-sur-Yvette, Gif A - accelerator and Groningen). The perfect agreement of the results from those laboratories with that from Gliwice constitute a strong argument to support the preceramic dating of the burial and the associated wooden structure (compare section Peru, records Gd-2994, Gd-2996, Gd- 3441, Gd-4393, Gd-4394, Gif-8128, GrN-16593 and TAN- 89171). By the way, those agreeing results constitute, to some extent, a confirmation of the reliability of the laboratories that participated in the mentioned cross-dating.

2. The next important group of problems we met in our enterprise turns around the question of the original location of the samples or, more concretely, the name and geographical identification of the site, where it was collected. It would seem that such a question is too trivial to even be mentioned, because most evidently radiocarbon datings are primarily for the establishment of the chronology of archaeological sites, so that at least the name of the site should accompany every dating, together with the code and laboratory number. In practice the situation appears sometimes more complex. It is especially connected with the data collected from laboratory archives where information is based upon descriptions sent by archaeologists together with the samples. Those descriptions are sometimes so laconic and unprecise that it is absolutely impossible to make sure even of the name itself of the site and its localization, therefore it would be too much asking for more detailed information about the archaeological context. Therefore datings classificated as „Unidentified site name” or, in a better case, designated as, for example, „Quito region” or „Baños quad” or other similar kinds of approximate localizations appear in our catalogue. It even sometimes happens that datings identified with the same laboratory code and number (and, of course, the same age BP) are attributed to two absolutely different sites. It is precisely the case with the dating N-87 which in one of the sources is attributed to Paracas and in the other to La Florida in the Rimac Valley ( see the commentary to this dating in Andes 1994:361). Because it happens that the laboratory concerned is actually no longer active, we are unable to define the correct attribution of this dating. A related question is the precision of localization of the sites in terms of administrative units such as, for example, provinces and districts in the case of Peru, as well as their geographical coordinates. In some cases the differences between the coordinates given by several authors for the same site varied by some twenty arc minutes, and it happens even that according to the reported values the concerned pre- hispanic village should have been situated far out in the Ocean. The above mentioned problems with localization and identification of sites resulted in some important consequences in the presentation of data in the Catalogue as well as in their geographical indexation. First of all we have to resign from the presentation of the geographical distribution of the datings on maps for now. This is because the MapInfo software use needs exact geographical coordinates of the sites. The verification of all those data would delay considerably the presentation of the Catalogue to the readers. For this reason we decided to delay the geographical presentation for later, in one of the volumes of the projected specialized Bulletin.

But even with a well defined name and approximate localization appear sometimes rather relevant differences for example in the site number or its description. It is therefore difficult to establish if, e.g. two sites from Tumbaco valley (Ecuador), reported by different authors as Cumbaya Z3B3-021 (Santa Lucia) and Santa Lucia ED16, respectively, are really the same one or two neighbouring ones. A similar situation appears in particular for the area of great concentration of archaeological sites, in which the datings may be reported in association with a peculiar monument in the region or with the region or complex in general. Such is the case, among others, with the archaeological sites of the Quebrada de Chilca, Paracas, Cahuachi, Ancon, Batan Grande, Caballo Muerto, etc. Even in the situation when we had good reasons to suppose that some of the sites labelled under somewhat different, but similar, references by the authors in fact concern the same one (see e.g. different „Ancon, Tank site” or „Qaluyu”) we chose rather to present them in the Index as separate than to risk possible confusion. We hope to be able to introduce the relevant precisions to that part of the Database in the near future.

3. The last of general problems we met in the completion and analysis of data for our Catalogue was the question of attribution of datings to chronological classificatory periods of the Andean archaeology, a problem intimately related to the elaboration of a Chronological Index of datings. Initially we had been planning to attribute each peculiar dating in the classical chronological frame such as Late Horizon, Late Intermediate Period, Middle Horizon etc. for Peru, as a reference base for the index. Soon, although, very essential difficulties appeared: on which criteria should such a classification be established? On the basis of stratigraphy established by the archaeologist who collected the sample and submitted it to the laboratory, or on the results of the absolute dating by radiocarbon? In the first of those cases the period established according to the stratigraphical level in which the sample was collected might be significantly different from the radiocarbon dating because of, for example, the contamination of the sample itself. The result of such a choice would lead to the attribution to different periods of datings of similar absolute age, because of their different stratigraphical position. In the second case e.g. of an attribution of a dating to archaeological periods on the only base of its radiocarbon age, would lead to the rejection of all the stratigraphical and contextual evidences.

If we had taken into account only datings which results had been fully analyzed and accepted by archaeologists themselves, it would have considerably reduced the number of dates in our Catalogue. In that situation we decided to include in the "Archaeological comment" to the datings, only the stratigraphical attribution, reported in terms of regional period or phase, nevertheless we built the Index indendependently of this information, taking into account only the values of the conventional radiocarbon age of the datings, to dispatch them in the arbitrary radiocarbon time periods BP.

Those limits correspond more or less to the cultural chronology of the Andean region, for the above mentioned reasons we have not considered suitable to attribute them the names of the cultural periods. We should also explain the reasons why we have chosen to construct the Chronological Index using the conventional radiocarbon age BP and not the calibrated values, although the calibrated age is obviously of greater interest to the users. There are two main reasons for such a choice: first, most of the datings from the Andean region have been given until now only in conventional radiocarbon age BP. If we had produced a calibrated conversion of those dates in the Index, this could have led to confusion and rendered more difficult the consultation of the Catalogue. The second reason is essentially technical. As it is exposed in details in the chapters by A. Michczynski and A. and M. Pazdur, we have chosen a form of presentation of calibated dates which consists of offering time ranges of given probability level, and renounced from presenting the calibrated equivalent of the conventional radiocarbon mean value (with corresponding standard errors, etc). Moreover, in respect to the described dendrochronological calibration curves a single conventional date would have produced several time ranges of different probality levels. In this situation the indexing of calibrated datings would have been difficult to present clearly. The decision of establishing the Chronological Index on the base of the conventional radiocarbon age BP is a compromise between the theoretical basis we had impose on ourselves and the practical usefulness of the Catalogue.

After presenting the main problems we met during the collection and classification of data introduced in the catalogue, I would like to present a few general remarks concerning the datings presented here, in the Catalogue. Some general statistics are given in the following table:

 

 
Categories of sites according to the number of datings per site Nº of datings
  1% 2-4% 5-9% >10% Total % %
Bolivia 17 58.6 10 34.5 0 0 2 6.9 29 4.5 99 3.7
Ecuador 65 43.3 53 35.3 16 10.7 16 10.7 150 23.6 657 24.6
Peru 182 39.7 177 38.6 59 12.8 41 8.9 459 71.9 1917 71.7
Total 264 41.4 240 37.6 75 11.8 59 9.2 638   2673  
 

 

The Database includes 2672 datings from about 630 sites, among which the major part is represented by archaeological samples, but we have also considered the results from other fields of research, such as paleoclimatology or geology. We felt that information concerning for example the last glaciation in the Andes or the changes in the processes of fluvial deposits in Amazonia could be interested for paleoecoligical reconstructions in archaeology. We noticed that the distribution of datings is not equal for the three concerned countries and the number of dates from Bolivia is very low. It obviously may depend, in some degree, on the unsufficient consultation of relevant archives, anyway, when compared to the 657 datings from Ecuador, we can clearly see that the basis for analysis and discussion of the absolute chronology of, e.g. the archaeological cultures of the Bolivian Altiplano is rather scarce. Another interesting constatation concerns the question of the number of sites and the the average number of datings per site. The Catalogues includes 630 sites, among some may appear under slightly different names (for the reasons I presented previously). Fortunately, those cases are not numerous, therefore is it reasonable to consider that the total number of dated sites oscilates around 620. This produces a mean number of ca 4,3 datings per site, ranging from a record number of 118 datings for the same site (or better to say an archaeological complex) of Cahuachi , Nasca (Peru) to 273 sites with only one dating per site (see table 1). Evidently the concept of „site” is rather vage as it may designate archaeological complexes spred on several square kilometers as well as a small shell midden. Looking further to the statistical table, we observe that for the 30 Bolivian sites (the various sectors of Tiwanaku appearing as gathered together) the mean value is still 3,3 datings per site, but we find that 59 datings come from only 3 sites (mean value: 29,7 datings per site) whereas the remaining 27 sites had only from 1 to 3 datings each (in total 40, mean value 1.48 datings per site). This not only reflects the difference of importance of the sites concerned but also the attitude of many archaeologists, to often over-estimating the diagnostic importance of a single radiocarbon dating. We are well aware here of the impact of the economical factor, the radiocarbon datings being rather expensive, nevertheless I would like to insist on the fact that only a series of radiocarbon datings from a single context may offer a solid basis for chronological estimations. Even though, sometimes several coinciding datings made from a single context may prove misleading. Those cases are, fortunately, seldom and occur only in very specific conditions, but, us an example, let us remember the well known case of the „Otavalo Man” (see Birm-331a,b,c, Birm-332, Birm-360a,b, Q-1032 and Q-1033 in the Ecuador section).

Concluding remarks
The problems presented above concern the process of collecting radiocarbon dates and the various limitations concerning the possibilitiess of their use for forming chronological scales, might give the impression of denigrating the C14 method itself and the reliability of the results. Omitting the question of various corrections and associated question of the exactness of the measurement itself in conventional radiocarbon years, especially discouraging seems to be some of the results of calibration, where we obtain a range of perhaps 800 years and not only for the dating of layers from the early Preceramic or Palaeo- Indian phases, but from about 2500 BP (see e.g., the dating TK-344 in the Peru section). Have we in creating our database and critically examining the data collected here unwittingly provided arguments for those archaeologists who question the usefulness of the C14 absolute dating method? It seems to me that from a certain moment archaeologists began to expect too much of the C14 method, treating it as a „deus ex machina” (nomen omen) completely authoritative in questions of chronology. In such a situation successive corrections and qualifications in the interpretation of the results of the radiocarbon dating created disappointment which in turn created distrust. This was perhaps more an emotional than analytical approach to the problem. We all are aware for example that such fundamental techniques such as stratigraphy also hold many surprises, and we often have to alter some earlier interpretations. Nobody however has questioned the sense of using the stratigraphic method in fieldwork and the interpretation of the results of excavations. Our approach to C14 dating should be similar, especially now in a situation where we are acquiring successive improvements in the dendrochronological calibration and it is now beginning to regain the trust of investigators.

Both in this article and that of Mr and Mrs Pazdur (Andes 1994:25 and ss) we have discussed different methods of minimalising the possibilities of obtaining dates which are not in agreement with the stratigraphy, or simply undiagnostic from the point of view of absolute dating. Apart from careful selection of material for dating, of especial significance is the extension of the requirement of dating series of samples, and not just single samples from a particular context. This is especially the question if it possible to divide the samples between different laboratories. The next extremely important question is the matter of the choice of laboratory to which the sample is sent. For example in the investigation of sites from EIP or MH where we wish to conduct an analysis of events in a relatively short timespan we should aim to obtain dates with standard errors no greater than 40-50 years (some labs can achieve dates with standard errors of the range of 20-25 years). Only in the case of such a scale of accuracy we can count on receiving calibrated results which are accurate enough for our analyses. There remains the question of the different statistical processes for the analysis of sequences of radiocarbon dates. From the Andean area we should note the extremely interesting attempt to determine the absolute chronology for the Jama Valley (Ecuador) site carried-out by Zeidler et al. In any case we may regard as finished the period in which the use of single radiocarbon dates was acceptable for the dating of sites, and analyses of absolute chronology are carried-out on dates given in radiocarbon years BP. The C14 method has been developed for over 40 years and it has allowed us to improve the process of analysis itself as well as to define more precisely its various limitations. In connection with Andean archaeology, this method will long remain indispensible. Because of the practical impossibility of using dendrochronology in the mid- Andean area (lack of suitable trees) and the rather limited use of other techniques of dating (e.g., palaeomagnetic or thermoluminescence) the C14 method will remain, despite all its limitations, the most accurate we have at our disposal.

 

Selected bibliography

Amat Olazabal H., 1976. Estudios arqueológicos en la cuenca del Mosna y en el Alto Marañon. [in:] Actas del XLI Congreso Internacional de Americanistas, México, pp. 532-544.
Arnold J.R., Libby W.F., 1951. Radiocarbon Dates, Institute for Nuclear Studies, University of Chicago. Science, vol. 113, no 29271, pp. 111-120.
Bischof H., 1972. The origins of pottery in South America - Recent radiocarbon dates from Southwest Ecuador. [in:] Atti del XL Congresso Internazionale degli Americanistici, Roma-Genova 1972, vol. 1, Genova 1973, pp. 269-281.
Bowman S.G.E., Ambers J.C., Leese M.N., 1990. Re-evaluation of British Museum radiocarbon dates issued between 1980 and 1984. Radiocarbon, vol. 32, no 1, pp. 59-79.
Burger R., 1981. The Radiocarbon evidence for the temporal priority of Chavín de Huántar. American Antiquity, vol. 46, no 3.
Damon P.E., Ferguson C.W., Long A., Wallick E.I., 1974. Dendrochronologic Calibration of the Radiocarbon Time Scale, American Antiquity, t. 39, pp. 350-366.
Engel F., 1963. Datations à l'aide du Radio-Carbone 14 et problèmes de la préhistoire du Pérou. Journal de la Société des Américanistes, t. LIII, Paris, pp. 101-132.
Engel F., 1966. Geografía humana prehistórica y agricultura precolombina de la Quebrada de Chilca. Universidad Agraria, Lima.
Ferguson C.W., 1969. A 7014-Year Annual Tree-Ring Chronology for Bristelcone Pine, Pinus Aristrata, from White Mountain, California. Tree Ring Bulletin, t. 29, pp. 3-29.
Klein J., Lerman J.C., Damon P.E., Ralph E.K., 1982. Calibration of Radiocarbon dates: tables based on the concensus data of the Workshop on calibrating the radiocarbon timescale. Radiocarbon, vol. 24, pp. 103-150.
Krzanowski A., Michczynski A., Pazdur M.F., Ziólkowski M.S., 1994. Komputerowa baza danych datowan radioweglowych rejonu Andów Srodkowych. Zeszyty Naukowe Politechniki Slaskiej, Geochronometria 10, pp. 139-151.
Pazdur A., Pazdur M., 1982. Chronometria radioweglowa jako metoda badawcza w archeologii, mozliwosci, ograniczenia, perspektywy. Przeglad Archeologiczny, vol. 30, pp. 5-45.
Ponce Saginés C., 1972. Tiwanaku: espacio, tiempo y cultura. La Paz.
Ravines R., 1982. Panorama de la arqueología andina. Instituto de Estudios Peruanos, Lima.
Ravines R., Alvarez Sauri J.J., 1967. Fechas radiocarbónicas para el Perú. Arqueológicas 11, Museo Nacional de Antropología y Arqueología, Lima.
Rowe J.H., 1965. An interpretation of radiocarbon measurements on archaeological samples from Peru. [in:] Proceedings of the 6th International Conference on Radiocarbon and Tritium Dating, pp. 187-198.
Silverman H., 1993. Cahuachi in the Ancient Nasca World. Iowa University Press.
Suess H.E., 1955. Radiocarbon Concentration in Modern Wood, Science, vol. 120, pp. 415-417.
Suess H.E., 1979. A calibration table for conventional radiocarbon dates. [in:] R. Berger, H.E. Suess (eds.). Radiocarbon Dating. Proceedings of the 9th International Radiocarbon Conference, University of California Press, Berkeley, pp. 777-785.
Watson J., 1986. 14C and Cultural Chronology of the North Coast of Peru. [in:] R. Matos, S. Turpin, H. Eling (eds.). Andean Archaeology Monograph XXVII, Institute of Archaeology UCLA.
Taylor R.E., Meigham C.W., 1978. Chronologies in New World Archaeology. Academic Press.
Zeidler J.A., Buck C.E., Litton C.D., 1993. The integration of archaeological phase information and radiocarbon results from the Jama River Valley, Ecuador: a Bayesian approach. (ms).
Ziólkowski M.S., Pazdur M.F., Krzanowski A., Michczynski A., 1994. ANDES. Radiocarbon database for Bolivia, Ecuador and Peru, Warszawa-Gliwice.

We are very grateful to Prof. Christopher Bronk Ramsey for his support in the installation of the date calibration system, based on the OxCal program.