63 VISUAL SOURCING OF (CENRAL EASIrRN CALIFORNIA OBSIDIANS Robert L. Bettinger Michael G. Delacorte Robert J. Jackson As compared with their counterparts of barely a generation ago, modern archaeologists practice science with the help of a bewildering and ever increasing battery of technical aids. As the direct result of research achievements in physics, chemistry, biology, and electronics and vigorous pro- grams in government, industry, and academia to develop the practical applications of these achievements, a generation of sophisticated technology applicable to virtually every phase of archaeological work, from site survey to report prepar- ation, is now available. Unfortunately, many of these techniques are very expensive, a drawback all too apparent in an era of dwindling funds for research in archaeology that is likely to get worse before it gets better. It would seem prudent therefore, periodically to take stock of the technology considered "state of the art" in archaeology in terms of its performance relative to less expensive alternatives -- perhaps less technical - but potentially capable of achieving comparable or closely comparable results. In each such instance, the question is whether the differences in results justify the differences in cost. The specific problem dealt with below is that of obsidian source analysis. The techniques now available for this kind of work include neutron activation and several variations of x-ray fluorescence (cf. Harbottle 1982). It is generally accepted that these chemical methods of sourcing provided the first reliable means for establishing the parent geological source of obsidian artifacts. Prior to their advent, the attribution of archaeological obsidian to specific quarries was based on a combination of visual inspection and commrn sense, the results of which were rightly considered speculative and seldom put forth with any conviction, although they have proved right (e.g, Bennyhoff 1956; cf. Jack 1976) about as often as wrong (e.g., Goldschmidt and Driver 1940: 120; cf. Hughes 1978: 54). The reliability of chemical sourcing methods, despite their expense and inaccessibility to many, resulted almost imnediately in the development of an entirely new field of obsidian source analysis. The theoretical and methodological naievete evident in the early literature of the field, attributable to its origins in and subsequent emphasis on scientific technique, has evaporated as research has increasingly addressed itself to issues firmly grounded in anthropological theory. Despite this, the nascent field of obsidian source analysis still retains some dubious axioms. Among the most basic of these is the proposition that volcanic glass flows are so variable in their macroscopic physical properties that reliable source identifications can only be made by one of the accepted chemical sourcing methods. Curiously, this proposition has seldom been put to the test. This would present no problem if source analyses were inexpensive or required of only a few specimens in each instance, but neither condition applies. 64 The least expensive chanical sourcing methods costs approximately $15 per sample and with increased awareness of sampling as an archaeological problem it is apparent that for a broad range of problems source analysis must be performed on large suits of specimens often numbering into the thousands (cf. Hughes and Bettinger 1984). In more philosophical perspective, the claim that obsidian sources are never macroscopically distinguishable agrees with neither camnun sense nor experience; it is no more logical than its alternative, that they can always be distinguished megascopically. Mlany archaeologists, whatever they are willing to put into print, express confidence in their ability to recognize specific glass sources, though few have any idea as to the degree to which their impres- sions are correct. 1 In the following sections, we explore the viability of megascopic obsi- dian sourcing as an alternative to the more formidable chemical methods avail- able. The term megascopic here refers to unaided visual inspection, and to visual inspection aided only by low power magrnification (lOx - 30x). Background During the course of probabilistic surveys of a large archaeological transect centered near the modern town of Big Pine, California (Bettinger 1975; 1977), it became evident that obsidian exhibiting certain properties of color, texture, banding, inclusions, and iridescence varied in abundance according to location within the transect; it increased in frequency near the known location of the Fish Springs obsidian source (cf. Steward 1933: 262), and decreased with increased distance fram that point. By the close of the survey, familiarity with this distinctive obsidian and its pattern of distri- bution made it appear likely that Fish Springs was the source of this partic- ular glass. In turn, this suggested that Fish Springs produced in some frequency a variety of glass that could be distinguished visually from glass produced by other sources. Local amateurs, far more familiar than we with the welter of archaeological sites in Owens Valley, expressed similar opinions about glass they identified as deriving from the Queen source, in the Truman Meadows area north and east of Big Pine on the California-Nevada border. To explore further these possibilities, in 1977 two undergraduate students in the Department of Anthropology at New York University were famil- iarized with the properties of the glass believed to derive from Fish Springs. These two students then identified the frequency of such glass in samples of debitage given to them without provenience fram sites located during the probabilistic surveys of the Big Pine transect. The results showed the expected patterning, sites near the Fish Springs source yielding high fre- quencies of this variety of glass, while sites farther away yielded lower 1 Editor's note: workers at Sonoma State University report success rates between 80 - 95% for visual source identifications of Borax Lake, Mt. Konocti, Napa Glass Mountain and Annadel obsidian at archaeological sites in the North Coast Ranges of California (see Origer 1982: 194-197). 65 frequencies. Further, as previously observed in the field, the frequencies of this type of glass were extremely high near the source (often in excess of 90Z) which suggested that the attributes that had been employed to identify Fish Springs obsidian characterized the bulk of the material produced by this source. These findings indicated that for at least some obsidian sources, objective and explicit criteria could be developed to permit reliable maga- scopic identification of the majority of glass fram a single source. Building on this assumption, quarry specimens and chemically sourced samples of obsidian debitage were inspected and a series of criteria were developed as a basis for the visual identification of obsidian from four major obsidian sources, three located in eastern California and one in western Nevada. These sources were: Casa Diablo, Mount Hicks, Truman Meadows (Queen), and Fish Springs. Since the initial work, Bodie Hills has been added to this list. Between 1980 and the present, the identification criteria for these glasses has been refined and the reliability of visual identification has been tested in three different instances. Procedures In all cases the techniques for megascopic identification have been the same. Specimens, either debitage or tools, are first inspected without magni- fication and candled to ascertain color and translucency. Each is then examined under a lOx-30x variable power binocular dissecting microscope with movable high intensity light source held in various positions to reveal inclusions, color, texture, and other salient attributes. It has been our practice to have two individuals, each similarly equipped with a binocular microscope and light source, work side by side examining the same specimen in turn, identifying its source independently. The identifications are compared and in the case of discrepancies the specimen is reexamined to resolve the differences. As the technique has been perfected, there has been a conscious effort to produce a specific source identification for each specimen and to minimize the frequency of specimens left unattributed to source. The reasoning here is that rigorous adherence to objective identification criteria in these tests, rather than more subtle intuitive impressions, alone will result in the iso- lation of traits that permit replicable results with visual sourcing. If a particular piece exhibits traits that satisfy the criteria specified for a particular source, e.g., in opacity, it is always identified as belonging to that source, regardless of whether it resembles in other respects the balance of specimens visually identified as belonging to that source or whether the identifier genuinely believes it to be from that source. The point is that in visual sourcing, the gross megascopic variability observed in glass from a particular location is reduced to a few key traits which it is assumed to display invariably and which at the same time others are assumed never to display. The empirical utility of the traits selected in any given instance can only be ascertained if the logical consequences of this assumption for identification are adhered to strictly. In this scheme, the only specimens 66 that are left unidentified are those that exhibit none of the traits or a com- bination of traits that identify the sources encompassed in the research and are hence presumed to be from other sources. The distinguishing traits currently used to identify the five glass sources within the Inyo-Mono region now considered potentially susceptible to megascopic sourcing are described in some detail below. Casa Diablo. This is the largest source in eastern California and certainly the one with the widest distribution when frequency is considered (Ericson 1981). The trait taken to distinguish this source is its near-uni- form opacity. All except the thinnest pieces permit the passage of little or no light. The degree of light transmission may vary between or within pieces of this glass but it is never more than faintly translucent. Some of this obsidian shows a deep silky sheen that approaches chatoyance. On the basis of earlier tests neither texture, which varies from very waxy to coarse grained, nor color, which varied from black to red and brown, are considered reliable in identifying this source. Fish Springs. This small source is located near the modern town of Big Pine. As discussed elsewhere (Bettinger 1982), there are two distinguishing attributes of this source: 1) much of it is green; and 2) it displays feathery brown inclusions that show an iridescent play when struck by low- angle light. All except the smallest pieces exhibit some heterogeneity in the transmission of light, which may vary from near-transparent to near-opaque, the transition fram one to the other occurring in the form of sharply defined and typically contorted flow bands. The clearer material contains abundant black and white phenocrysts that are most obvious only upon microscopic exam- ination. Also characteristic are white-gray streaks or thin bands of what appear to be volcanic ash. Minor fractions are brilliant red or exhibit a silky silver sheen produced by elongated vesicles; neither of the latter two are considered definitive traits. Queen (Truman Meadows). The distinguishing trait of Queen obsidian is its translucency and lack of minor inclusions. Much of this glass is clear almrst to the point of transparence, usually with a distinct gray or golden brown cast. Minute phenocrysts are very sparse. Flow banding is conmorn and occurs as thin, dense planar bands that are almost perfectly parallel to each other. In identifying this source, care must be taken that these planar bands are not viewed from the perpendicular, in which case the glass will appear to be nearly opaque. The only eastern California source with which Queen may be commnly confused is Mt. Hicks. Mt. Hicks. This source resembles Queen in that a substantial fraction of it is clear and without magnification appears nearly transparent and free of inclusions. Segregation of the tw is as follows. First, Mt. Hicks seldom occurs in large pieces, greater than 2 cm. in diameter, without some inclusions in the form of banding or clouding. Second, Mt. Hicks glass exhibits substantial quantities of-small phenocrysts readily visible upon magnification. Third, Mt. Hicks does not exhibit the striking parallel planar banding found in Queen glass. The banding is invariably contorted and individual bands frequently intersect each other. The only eastern California-source with which Mt. Hicks is generally confused is Bodie Hills. 67 Bodie Hills. To date, this source has proved the most troublesaoe to identify megascopically, in part because it constitutes but a very small frac- tion of the collections with which we have worked. The traits used define two distinct forns. One exhibits contorted, but generally linear, gray-green bands and masses interspersed with much sller quantities of clear bands- and masses containing minute black and white phenocrysts visible under magnification. This form might be confused with Fish Springs except that it lacks the brown and white banding characteristic of that source. The second Bodie Hills form is a mixture of very dense black and, in smaller quantities, red-brown dentri- tic masses interspersed with much smaller quantities of clear material contain- ing minute black and white phenocrysts. Quantification of Reliability The reliability of megascopic source identification is most conveniently tested by comparing these identifications with identifications obtained on the same material by one of the standard chemical characterization techniques (either x-ray fluorescence or neutron activation). It is, of course, assumed that the megascopic identification is done without knowledge of the results of chemical characterization. Consequently, megascopic identification should precede chemical characterization. The matches and msmatches between chemical identifications, in this study by x-ray fluorescence, and visual identifications can be quantified in several ways. The simplest expresses the number of correct visual identifi- cations as a fraction of the correct and incorrect identifications, combined, for all sources. In general terms, this expresses the overall reliability of visual source identifications and is an appropriate measure of the utility of the procedure in establishing the gross composition of a large assortment of glass from a site or region. The range of archaeological questions to which such data are applicable include patterns of resource acquisition, either by trade or direct procurement, territoriality (e.g. Bettinger 1982), and social ranking (e.g. Hughes 1978) to cite a few examples. In some cases, two or more sources that are difficult to distinguish visually can be meaningfully combined. Thus, until recently in our research Queen and Mt. Hicks sources were merged into a single co-source, which seemed justified on the grounds that they are relatively close together geographically. More drastically, in much earlier central Owens Valley studies, Fish Springs, which is readily identified, was compared against all others, which were then less well known. The extent to which such merging of visually similar sources is useful will depend on the question at hand and the relative location of the sources potentially to be merged with respect to each other and to the remaining sources. The reliability of visual identification also can be couched in terms applicable to an individual source. Here, as when all sources are considered at once, the number of correct identifications is the total number of pieces from a source correctly identified as belonging to that source. The errors, however, can be of two kinds and hence admit of no simple quantification. In reference to a single source, there are first errors of conmmission, those in which other sources are wrongly confused with the source in question. In the tabulated test results, in which we adopt the convention of casting x-ray 68 fluorescence identifications as the columns and visual identifications as the rows, these equal the marginal row total less the number of correctly identi- fied pieces. Still speaking of a single source, there are also errors of omission, those in which specimens of that source are incorrectly attributed to another source. These equal the column total less the number of correct identifi- cations. It should be obvious that for any collection of pieces subjected to visual sourcing, the number of errors made overall is equal to the total number of errors of omission and, at the same time, the total number of errors of comnission. That is, a piece incorrectly attributed is at once an error of commission (with respect to the source to which it was wrongly attributed) and omission (with respect to the source to which it actually belonged). This is not true in the case of the individual source, where an error of omission does not imply an accompanying error of comnission -- nor an error of com- mission one of omission. Consequently, sources may vary according to the degree to which they are subject to errors of connission and, similarly, to errors of omission. Further, there are at least two senses in which it can be useful to dis- tinguish between errors of commission and errors of omission for individual sources. First, inspection of the errors of omission and comrission discloses which specific sources are most coamnly confused with each other and, if the distinguishing visual criteria are kept in mind, the basis for that confusion. Thus, errors of ccmmission occur when'-the traits taken to distinguish the source in question are found in glass from a different source; errors of omission occur when the traits taken to distinguish a source are not present in glass from that source. Careful inspection of such data provides a sound basis for improving identification criteria for future applications. Second, it is useful to distinguish between errors of commission and errors of omission because in some applications of source information, errors of comnission are relevant and errors of omission are not. In these circum- stances, visual source identification can provide reliable data for sources -infrequently subject to errors of commission regardless of the frequency of errors of omission for that source. The most recurrent situation of this kind is in relation to source specific obsidian hydration dating. For this purpose, source identification is relevant only for the sample submitted for rind measurement and not the entire population of specimens representing that source. Thus, it need only be certain, or relatively certain, that the specimens visually identified as belonging to a source are from that source, which is to say that the source is not greatly subject to errors of commission. The same sort of utility would apply to any archaeological question in- volving only one specimen rather than a suite of specimens. For instance, the ability to determine with, say, 95% reliability that a given specimen is Fish Springs obsidian might be useful in any number of contexts. 69 It is possible, of course, to minimize ers of ccnnission by being ex- ceptionally conservative in identifications and deferring all questionable iden- tifications to a residual category of unidentifiable specimens. This, however, effectively eliminates the application of the results to broader questions regarding the gross corposition of an obsidian collection, for it cannot be assumed that the number of specimens for which visual identifications are quite certain is proportional to the total number of specimens of that source in the collection. Some sources are likely to be more distinctive than others and so exhibit a higher frequency of certain identifications. For these reasons and for the reason that it tends to make the recognition of reliable source traits more ambiguous, we have eschewed this conservative strategy in our visual source identifications. Results Tables 1 through 3 report the results of three successive tests in which we have compared visual source identifications of archaeological materials from central Owens Valley (Table 1), Long Valley (Table 2), and western Nevada (Table 3) with chemical source identifications of the same materials. Table 4 sumarizes for each test the frequency of overall errors, errors of coirmission, and errors of omission. In each case it was presumed on the basis of both geographical location and previous source analysis that Fish Springs, Casa Diablo, Queen, Mt. Hicks, and Bodie Hills glass would account for a substantial fraction of the material present. In addition, these sources were considered susceptible to visual identification and at least provisional identification criteria were available for each one. The visual identification criteria for certain of these sources, however, differed to some degree between the three tests. The currently accepted iden- tifying traits described earlier were used in the most recent test (western Nevada). The identifying criteria for Fish Springs in the first (Owens Valley) and second (Long Valley) tests were the same as those currently accepted. The identification criteria for Casa Diablo were the same as currently accepted in the Long Valley test but not in the earlier Owens Valley test, in which it was defined by a combination of traits consisting of susceptibility to light transmission (opaque), color (grey/black), and texture (fine-grained). In the Owens Valley test, Queen was identified on the basis of transparency and Mlt. Hicks on the basis of contorted banding and cloudy inclusions. The results of this test showed these criteria to be invalid and in the following Long Valley test Queen and Mt. Hicks were treated as a combined co-source iden- tified by the presence of complete or fractional translucence. Bodie Hills was identified by the presence of abundant black dendritic masses in both the Owens Valley test and the Long Valley test. Central Owens Valley. The first test (Table 1) was carried out in 1980 on a sample of 39 projectile points and 15 pieces of debitage selected from sites located in the Big Pine archaeological transect mentioned earlier. In this first trial of visual sourcing, the principal question was whether Fish Springs 70 obsidian, which was defined in this and all subsequent tests in the manner described earlier, could be reliably segregated from other glass occurring on archaeological sites in Owens Valley. At the time we believed it also would be possible to segregate other glasses visually, but our notions about the defining criteria that would apply to these were less certain than for Fish Springs. That is, the identifying traits enployed for Casa Diablo, Mt. Hicks, and Queen were neither as well defined nor as well understood as they are at present. The effects of this differential familiarity are evident in the results from the test. Every piece of Fish Springs glass in the test sample was identified correctly as Fish Springs. However, two additional pieces, one from Casa Diablo and one from Mt. Hicks, were incorrectly identified as Fish Springs. Casa Diablo was correctly identified in 7 of 11 instances and it was not confused with any other source. On the other hand, 4 pieces of Casa Diablo glass were not correctly attributed to this source; 1 was identified as Fish Springs, and the other 3 were considered unidentifiable. Mt. Hicks and Queen were not successfully separated: six pieces of AMt. Hicks were incorrectly identified as Queen, a seventh as Fish Springs, and an eighth was unidenti- fiable. Queen was not mistaken for any other source, but (as just noted) 6 pieces of Mt. Hicks were wrongly identified as Queen. In addition to these, also unidentified was a single piece of Mono Craters/Mono Glass Mountain, two chemically indistinguishable glass sources, one located east of Long Valley, the other in the Mono Basin. Finally, 5 pieces from an as yet unknown source, probably in Nevada, were correctly identified as deriving from sources outside central eastern California. Long Valley. A second test of visual source identification was performed in early 1983 on a sample of 57 projectile points collected during surface surveys in Long Valley. At this time it was believed that the Queen and Mt. Hicks sources could not be visually separated with any degree of confidence, so these two sources were lumped together as a single co-source. The results of the test are indicative of our familiarity with different glass sources at the time and of the distinctive differences between these particular glass sources. Casa Diablo, Queen, and Mt. Hicks are among the most easily and reliably recognized glasses found in eastern California and we had frequently encountered them during previous research in both Owens Valley and Long Valley. Long term familiarity with these glasses, coupled with their dominance in the Long Valley collection, resulted in very accurate visual source identifications for this particular collection. Fifty-one of fifty-seven pieces, or 80%, were attributed to the correct source. Minimal representation of Fish Springs, Bodie Hills, and Mono Craters/Mono Glass Mountain glass precludes an accurate statistical assessment of success in identifying these sources. The mediocre performance with Fish Springs is probably an aberration of sample size; the poor performance with Bodie Hills and Mono Glass Mountain/Mono Craters, on the other hand, is reasonabily indi- cative of the limited degree to which they were susceptible to visual identifi- cation at the time. Visual identification criteria have yet to be determined for Mono Glass Mountain and Mono Craters. Fortunately, neither source 71 contributes significantly to any archaeological assemblage in central eastern California so far analyzed by x-ray fluorescence. Nevada. In the most recent test of visual source identifications, a sample of 60 projectile points from Hidden Cave, western Nevada, and Gatecliff Shelter, central Nevada, that had been previously sourced by x-ray fluorescence (Hughes 1983a, 1983b) were examined and attributed to sources on the basis of visual identification criteria described above. On the basis of careful inspection of previously sourced collections, the explicit criteria described earlier were used to distinguish Queen, Mt. Hicks, and Bodie Hills glass. The results indicate that Queen and Mt. Hicks glasses can be successfully segregated in the vast majority of cases by visual inspection alone: out of a combined total of 37 pieces of Queen and Mt. Hicks obsidian, 31 (89%) were correctly identified, and only 3 of the misidentifications were the result of confusion between these two particular sources; the other 3 errors resulted from AMt. Hicks glass being misattributed to Bodie Hills. That highly reliable criteria for segregating Bodie Hills from other sources remain to be discovered, if indeed they exist, is shown by the rela- tively low reliability of visual identifications. Only 7 of 10 specimens were correctly attributed from a group of 11 specimens visually identified as belonging to that source. Nbreover, the data in Table 3 clearly indicate that Bodie Hills is likely to resist accurate visual sourcing because it often is confused with two other sources, Casa Diablo and Mlt. Hicks, that are them- selves visually quite distinct and seldom confused with each other or with other sources. It is the case, therefore, that Bodie Hills must lie between these two sources in terms of its visual characteristics and be sufficiently variable that it is confused with them and they with it. The results of this particular trial indicate that Bodie Hills is confused with Mt. Hicks at roughly the same rate Mt. Hicks is confused with Bodie Hills, i.e. between 14% and 10% of the time. These data also indicate that Casa Diablo glass is never wrongly identified as Bodie Hills, but that much Bodie Hills glass, about 20%, is wrongly identified as Casa Diablo. When only eastern California sources are considered, the accuracy of visual source identifications in this particular test is 83%. X-ray fluor- escence, however, indicated that nearly 18% of the collection in question consisted of pieces attributable to the Majuba Mountain source in northwestern Nevada (Hughes 1983b). We were entirely unaware of the existence of this source since it only recently has been reported in the literature. In addition, Hughes' (1983b) work on this source suggests that it may be difficult to dis- tinguish from Casa Diablo even by x-ray fluorescence unless the proper elements are measured. That this source is likely to be a major problem in attempts at visual sourcing in areas where it is found is shown by the results reported in Table 3. Here, Majuba Mountain is confused with both Queen and Casa Diablo. In this respect it is rather similar to Bodie Hills, which, as just discussed, overlaps with both Mt. Hicks (a "clear' source) and Casa Diablo (an "opaque" source). Given this, it is curious that Majuba Mountain is not as frequently confused with Bodie Hills as might be expected. The obvious implication is that Majuba Mountain is not variable in appearance in the sense 72 that it might be described as grading between clear and opaque. Rather, these data suggest that Majuba Mountain takes on two distinct forms: a dark, opaque fonn practically identical to Casa Diablo, and a clear form identical or nearly identical to Queen. Reinspection of the pieces in question suggests some dimensions along which it may be possible to segregate Queen from Majuba Mountain, but not that differentiate Majuba Mountain from Casa Diablo. Conclusions and Implications As stated at the outset, at issue here is the long standing axiom that obsidian glass sources cannot be reliably identified without recourse to chemical characterization. In operational terms, the question is whether the substantial cost of chemical characterization relative to visual identification is justified by the observed differences in results obtained by these two techniques. Within the range of possible outcomes implied, at one extreme visual identification might be judged so unreliable or misleading that one would conclude that chemical characterization is always worth the cost re- gradless of the problem under scrutiny. On the other hand, at the other extreme, visual identification might be found to match results obtained by chemical characterition exactly and lead to the conclusion that the cost of the latter is never justified. As one might reasonably have expected, the results of the three tests reported above indicate that at least for the west- ern Great Basin, neither extreme applies and that the actual situation falls somewhere in the middle. Several more concrete conclusions drawn fran these data are discussed below. First, in the most general sense, it is quite clearly the case that many obsidian sources produce glass that can be successfully identified mega- scopically. More than three-quarters (78%) of all the specimens considered in the three tests were correctly identified using visual identification criteria. Second, not all eastern California sources can be visually identified with equal success. At present, Fish Springs produces the best results over the widest area. Although it did not appear in the samples from western and central Nevada there is no reason to believe that it would have been missed .had it been present as it does not in any respect resemble the other sources represented in those collections. In addition, we suggest that identifications attributed to this source are sufficiently reliable to be used as a basis for source-specific obsidian hydration dating. Mt. Hicks glass is likewise successfully identified over a large area extending as far east as central Nevada. Nevertheless, Bodie Hills glass and Queen glass are mistaken for Mt. Hicks glass often enough that megascopic iden- tifications of this source are probably unreliable when used as a control for obsidian hydration dating. The Queen source is so readily confused with the Majuba Mountain source that visual identification is likely to be of limited value in central and western Nevada until the Majuba Mountain source is better known. With respect to assemblages from eastern California, on the other hand, Queen can be readily 73 identified. Further, since other sources are seldom confused with it, with this geographic limitation, identifications obtained visually provide a sound basis for source specific obsidian hydration dating. Like Queen, Casa Diablo glass is too often confused with Majuba Mountain glass to be consistently identified in western and central Nevada. Further, in eastern California, Bodie Hills obsidian is taken for Casa Diablo obsidian with sufficient regularity (ca. 20%) that in assemblages with substantial quantities of Bodie Hills obsidian, visual identification of Casa Diablo obsidian will be subject to considerable error of comnission and therefore not suitable as a control for obsidian hydration dating. Where Bodie Hills is not present in any quantity, as was found to be the case in Long Valley and Owens Valley, for example, visual identification of Casa Diablo glass is quite reliable and p perfectly suitable for use in conjunction with source specific obsidian hydration dating. Bodie Hills would appear to be the least reliable source for visual iden- tification. It is so frequently taken for the two sources geographically closest to it, Mt. Hicks (ca. 10%) and Casa Diablo (20%), and, in turn, AMt. Hicks is so frequently taken for Bodie Hills (ca. 14%), that visual identifi- cation will be of only modest utility for assemblages that contain Bodie Hills glass in any appreciable quantity. Nevertheless, certain kinds of questions, for example territoriality, might be addressed with source information of the accuracy observed for Bodie Hills, though source specific obsidian hydration dat ing is not amnng them. From the specific conclusions above regarding the utility of visual iden- tification of obsidian in eastern California follow certain broader implications for the potential role of visual identification in archaeological investigation. For one thing, visual identification cannot replace chemical characteriza- tion. Chemical characterization is ordinarily the means by which the accuracy of visual identification is judged and therefore visual identification can never be more accurate than properly executed chemical characterization and ordinarily will be, in varying degrees, less accurate depending on the source in question. Since it invariably entails a sacrifice of accuracy, the only legitimate use of visual identification is in relation to analysis that would not otherwise be performed owing to the expense of chemical characterizations, either those per- taining to obsidian source analysis directly or those made possible by the availability of resources that would have otherwise been allocated to chemical characterization. Archaeology is, after all, subject to severe practical con- straints among the most important of which are time and money. Archaeologists routinely accept these constraints and have developed a host of sampling strate- gies that enable them to extrapolate about unknowably large populations from limited information within certain limits of error. Visual identification should be undertaken for the same reasons and is, likewise, subject to error. We suggest that rather than using either technique to the exclusion of the other, chemical characterization is most appropriately employed where visual identification is the most subject to error. Visual identification can be 74 undertaken where chemical characterization is simply too costly, for example in cases that entail extremly large numbers of specimens. The precise mixture of these techniques, of course, will vary with circum- stances, but past experience in eastern California suggests that a rough guide- line is provided by the observed tendency for certain artifact categories to exhibit greater variability in obsidian sources than others. In particular, within a given site highly curated tools specifically projectile points, generally exhibit a wider variety of sources and especially distant exotic sources than do non-curated categories, most notably debitage (Hughes and Bettinger 1984). Since the number of projectile points is generally low when compared to the balance of the chipped lithic assemblage present, and since visual identification is the least reliable where source diversity is great and likely to include exotics, the use of chemical characterization is clearly indicated. Taken as a first step, this establishes the likely range of sources represented in the balance of the lithic assemblage -- which will improve per- formance in any subsequent visual identifications -- and at the same time serves as a basis for the source specific obsidian hydration that is routinely per- formed on projectile points. Debitage, by contrast, normally occurs in far greater quantity and is generally composed of a few local sources; visual identification is clearly the more appropriate technique for establishing the source composition of this category. To conclude, we have stressed that visual identification is potentially capable of performing tasks that cannot be practically performed by chemical characterization. We believe that it is always worthwhile to examine the possible uses of this technique. If the utility of visual identification of obsidian is discounted out of hand without assessing the feasibility of its application a whole class of interesting archaeological problems are, at least for the present, going to go unexplored. 75 Table 1. COanparison between visual identification and chemical characterization of obsidian fran central Owens Valley. Visual Identif ication Fish Springs Casa Diablo Queen Mt. Hicks Other Chemical Characterization Fish Springs Casa Diablo Queen Mt. Hicks Other* 25 1 7 3 3 1 6 1 1 6 Total 27 7 9 1 10 Total 25 11 3 9 6 54 *Note: five specimens are from unknown sources and one is from Mono Craters/Mono Glass Mountain. Table 2. Comparison between visual identification and chemical characterization of obsidian from Long Valley. Visual Identification Fish Springs Casa Diablo Queen/Mt. Hicks Bodie Hills Other Chemical Characterization Queen/ Bodie Fish Springs Casa Diablo Mt. Hicks Hills Other* 1 1 39 1 10 1 1 1 1 1 Tbtal 2 40 11 1 3 *Note: all three specimens are from Mono Craters/Mono GJlass Mountain. Table 3. Comparison between visual identification and chemical characterization of obsidian from western and central Nevada. Visual Identification Casa Diablo Queen Mt. Hicks Bodie Hills Chemical Characterization Majuba Casa Diablo Queen Mt. Hicks Bodie Hills 'Mountain Total 3 12 3 2 19 3 1 7 3 6 8 18 23 11 1 10 10 60 Total 1 41 12 1 2 57 3 15 22 TIbtal Cg 1 o o CD 0 o S o a n 8 g r- .r.4 ~ ~ 44 * 1141-A W11 2 , o 1n a I 0 o .HILCIO , o0 tl C) v, 0 LO, O EH C4 8 00 cn 0 00 jn 0 I I T-- O 1 8I CD 0 II 0 '-4 0 m 0 LO 0 O) m 0 ~ I~~~~~~~~~ 0q .0 I I I 1 I I om 00 I. Cs- I I I I I I bQ 0 C: r-- r- 9 s d H . s (aW, 0 8 g g X *1 - .H~~~~~ Pk & m & 9~ 76 co, 00 CN b *,.4 0 CQ) aH 0 0 C) ad Ez I I 77 References Bettinger, 1975 R.L. The surface archaeology of Owens Valley, eastern California: pre- historic man-land relationships in the Great Basin. Ph.D. dissertation, Department of Anthropology, University of California, Riverside. 1977 Aboriginal human ecology in Owens Valley: prehistoric change in the Great Basin. American Antiquity 42(1): 3-17. Bennyhoff, 1956 J.A. An appraisal of the archaeological resources of Yosemite National Park. University of CaZifornia ArchaeoZogicaZ Survey Reports 34. Berkeley. Goldschnidt, W.R., and H.E. Driver 1940 The Hupa white deerskin dance. University of California Publica- tions in American ArchaeoZogy and EthnoZogy 35: 103-142. Harbottle, 1982 G. Chemical characterization in archaeology. historic Exchange, edited by J.E. Ericson 51. Academic Press, New York. In: Contexts for Pre- and T.K. Earle. Pp. 13- Hughes, R.E. 1978 Aspects of prehistoric Wiyot exchange and social ranking. JournaZ of CaZifornia Anthropology 5(1): 53-66. 1983a X-ray fluorescence characterization of obsidian. In: The Archaeology of Monitor Valley 2: The Archaeology of Gatecliff Shelter, by D.H. Thomas. Anthropological Pcapers of the American Museum of NaturaZ History 59: 401-408. 1983b Obsidian source use at Hiidden Cave, Churchill County, Nevada. In: Archaeology of Hidden Cave, Nevada, by D.H. Thomas. AnthropologicaZ Papers of the American Museum of NaturaZ History. In press. Hughes, R.E. and R.L. Bettinger 1984 Obsidian and prehistoric sociocultural systems in California. Exploring the Limits: Frontiers and Boundaries in Prehistory, edited by S.P. DeAtley and F. Findlow. British ArchaeologicaZ Reports, InternationaZ Series. In press. In: Jack, R.N. 1976 Prehistoric obsidian in California I: geochemical Advances in Obsidian GZass Studies, ArchaeoZogicaZ Perspectives, edited by R.E.. Taylor. Pp. 183-217. Park Ridge, New Jersey. aspects. In: and GeochemicaZ Noyes Press, 78 Origer, T. M. 1982 Temporal control in the southern North Coast Ranges of California: the application of obsidian hydration analysis. M.A. thesis, Department of Anthropology, San Francisco State University. Steward, J.H. 1933 Ethnography of the Owens Valley Paiute. Publications in American Archaeology and University of California Ethnology 33(3): 233-350.