ARCHAEOLOGICAL
SURVEY
|
ILLINOIS ARCHAEOLOGICAL SURVEY |
|
The Case for the 15 Meter Transect Interval
The 15m transect interval is a viable, tried and true technique for finding sites during Phase I archeological surveys. Problems and criticisms of the 15m interval derive from academia. This is because academic archeologists have maintained a distorted view of survey work that persists to this day. This distorted view is historic and endemic, but more recent theoretical developments have produced new and different distortions.
Historically, academics have focused their efforts on individual sites above all else. This is convenient for them for two reasons. It facilitates the logistics of field school investigations. It helps them to think about sites as laboratories for studying human behavior. This is fine, but the mindset breeds problems which can result in a disconnect between academic archeology and the real-world concerns of cultural resource management (CRM).
Many academics view walkover survey only as a method for getting archeological data from a site. By focusing on large, known sites, many academic archeologists forget that survey is a necessary way to discover sites in the first place. For example, some of the earliest published work on surface survey (Redman and Watson 1970) examined the problems involved in linking surface manifestations with buried deposits. Do the types and density of surface artifacts relate directly to subsurface deposits? What kind of techniques should be used to accurately or systematically garner information on surface finds, so that a viable picture of subsurface deposits can be generated? These are important questions, but they are questions of surface artifact collection procedures, not site discovery procedures.
Others confuse survey and sampling at the site level with survey and sampling at the landscape level. Seminal work by Redman and Watson (1970: 281) introduced the 10% simple random sample as technique for reliably gathering representative data from the surface of a site. Others uncritically apply this approach to site discovery surveying. This can produce the profound result of missing the few, rare sites on the landscape and thereby a poor approximation of archeological reality (Whalen 1990: 224).
The important point to remember is that any survey is a sampling survey. Kintigh (1990: 238) sums this up nicely: "We are all aware that in doing survey we miss detail. Thus, we must consider the efficiency of a survey with regard to questions being addressed and the survey conditions (for example, the degree of ground cover)."
In CRM work, the objective of the Phase I Survey is to locate and record sites and make a preliminary assessment of their potential for listing on the National Register of Historic Places. Thus, we are looking for sites that contain enough scientific potential for National Register listing. In practical terms and in plowzone contexts, this excludes isolated finds and very small sites. Where topographic conditions do not otherwise suggest the potential for deeply-buried deposits, these types of sites are never recommended for additional evaluation, much less NR listing.
If discovering isolates and extremely small sites in the plowzone is not the objective of CRM survey efforts, then we need not employ techniques necessary to ensure their discovery. Our surveys can be designed to locate sites that we consider important - those that have a chance for NR eligibility and which will consist of more than two or three artifacts.
Some archeologists still worry that standard pedestrian survey will not locate 100% of the artifacts on the ground surface. This is of especial concern to the advocates of "site-less" archeology. This concept was developed by archeologists working in places like the American Southwest, where large portions of the landscape are covered by continuous artifact scatters of variable density and where defining the edges of a site are truly problematic. In Arizona, for example, scatters of fewer than 10 artifacts have been considered "small sites," and there is considerable debate as to whether they should even be recorded. This debate has solidified the position of "site-less" archeology proponents, who claim that only they can construct valid land use histories because only they stand firmly on the bedrock of archeological theory (e.g., Wandsnider and Camilli 1992: 183).
This orientation epitomizes the disconnect between academia and real-world CRM practices. A major proponent of "site-less" archeology expresses dismay that so-called full-coverage surveys are only sampling surveys. He complains that:
". . . contemporary field methods generally mirror those of traditional field walking. In the search for surface artifacts, linear transects typically are examined across the area under study, albeit with much tighter intervals between members of the survey team (15-m, 10-m, and even 5-m intervals have been reported; Ebert 1992). As a consequence of these field procedures, many "full-coverage" surveys have yielded, in reality, only samples of the surface archaeology, and probably very biased ones at that (Kvamme 1998: 128)."
Kvamme and other proponents of "site-less" archeology have lost touch of the fact that survey is a sampling procedure which acknowledges that complete coverage is an ideal that can never be attained, but which must instead be approximated given real-world considerations for time, effort, and monetary cost.
While opening new interpretive horizons and freeing themselves of the need to define site boundaries, the proponents of site-less archeology burden themselves with even greater methodological issues. Because they acknowledge no interstitial space between sites, they cannot concentrate their efforts, but must instead examine every part of the landscape with equal intensity so as to locate every artifact visible on the ground.
Kvamme (1998: 130) ignores these problems in his case study from Colorado. His study area is a 6 ha (14.8 a) parcel, which he gridded out in 1m squares and intensively surface-collected over the course of seven field seasons. Of course, academic archeologists have the prerogative to conduct such measured and intensive research. How many real-world CRM projects could justify spending seven years to survey a 15 acre parcel? Having mislaid the distinction between surveying to discover sites and surveying to discover every artifact in a given area, Kvamme (1998: 129) proposes rigorous new field methods that illustrate his disconnect with real-world archeological concerns. He recommends laying out surveyed grids of squares, 1-5m on a side, and using "saturation survey" - intensive, multi-directional walkover coverage within each square.
Archeologists surveying in the Sand Hills of western Nebraska adopted elements of the "site-less" archeology approach while incorporating vegetation sampling techniques borrowed from plant biologists (Burger, et al 2002). They were convinced that "traditional pedestrian surveys result in shaky inferences about landscape patterns due to a lack of understanding of the properties of the sample;" the sample being the sites and isolates typically recorded during traditional walking surveys (Burger, et al 2002: 413). Soon, however, they found themselves walking 70cm transects (shoulders touching) and then crawling across the ground on hands and knees, with shoulders touching, as dictated by the new survey methodology (Burger et al 2002: 414). The time and effort required for this sort of survey was not reported.
Another example of the problems inherent in "site-less" archeology is a sampling survey reported by (Ebert 1992) on the Green River in southwest Wyoming. Thirty random 500 by 500m squares, comprising a 1% sample of the study area (750ha or 1853.1 acres) were laid out. The site discovery method consisted of a five-man crew who walked at 5m transect intervals, marking artifact locations with pin flags. They were followed by a separate recording crew. The discovery crew alone covered about 25ha (61.8a) per day. This works out to coverage of 5ha (12.3a) per person per day, with no recording. The work of the discovery crew was considered to be "relatively fast (Ebert 1992: 163)." The overall speed of this survey effort was such that only 25 of the original 30 survey quadrats were surveyed before time and money ran out (Ebert 1992: 161). Overall, 369 man-days were expended to survey and record surface data on 625ha (1544.4a) of land. This works out to individual coverage of about 1.7ha (4.2a) per person per day for survey and recording combined.
In Illinois, we need not concern ourselves with the methodological contortions some archeologists have devised in order to accommodate the "site-less" landscape concept. Except in some restricted topographic settings (around Cahokia, and along some sections of the lower Illinois River bluff line, for example), this problem of continuous sites is a non-issue.
Unlike Arizona, our "small sites" really are small. Some archeologists in Illinois define "isolate" sites as scatters of three or fewer artifacts. But this author's own rule of thumb is to label every plowzone scatter with three or more artifacts as a "site." Given the quantity of naturally-occurring till chert in the soil and the high potential for plow-manufactured flakes, I do not typically record occurrences of one or two flakes. Single stone tools are recorded as isolated finds. Widely-scattered whiteware, ironstone, and stoneware sherds are noted but not recorded as sites, due to the potential for inclusion via manure spreading and roadside trash discard.
Very little experimental work has been undertaken to compare and test the utility of various walkover methods. What little there is of surface survey theory development and experimental work has been undertaken by academics. Most recent work has been done to support the "site-less" archeology concept.
Baker (1978) looked the simple mechanics of artifact shape and size, and concluded that larger artifacts will necessarily be over-represented in surface collections. A variety of natural and cultural processes work to keep a disproportionate number of large artifacts exposed on the ground surface. This is not bad news for surveyors, except that the effect also assists private artifact collectors in depleting sites of stone tools and other larger items.
An experiment in artifact recognition was conducted by Ebert (1992: 163-165) as part of the random sampling survey on the Green River. One quadrat was seeded with painted washers and nails, and the surveyors covered it with their standard 5m interval transects. Only two-thirds of the seeded items were recovered by the crew.
The results are predictably out-of-touch with the realities of CRM work. Ebert (1992: 165) considers reducing the transect interval down from 5m, surveying each area twice, and even paying crew members per artifact rather than hourly. Working from Ebert's data, Wandsnider and Camilli (1992: 184) conclude that "an acceptable transect interval of 15 m" will at most discover only 6-13% of artifacts in low-density surface distributions. Wandsnider and Camilli express regret that land managers and field archaeologists alike would reject the 2m transect interval that they consider is really needed.
Further, Wandsnider and Camilli (1992: 174) argue that artifact discovery is a two-step process, which begins with the discovery of all artifacts within 1-2m immediately surrounding the surveyor's feet. In the next step, the surveyor supposedly expands his search beyond the immediate area - but only if artifacts have been discovered in the initial 1-2m radius! Thus, Wandsnider and Camilli (1992: 182) demonstrate that a surveyor's gaze won't drift more than 2m off the centerline of his transect unless he has practically stepped on an artifact first, woken up, and started looking around in earnest. My own experience is that surveyors can effectively scan a swath 16m wide (8m ahead and to each side), simply by turning their heads from side-to-side while walking.
The literature is rife with fretting over the skill level and zeal of individual crew members, and how this can affect their ability to find artifacts on the ground surface. Wandsnider and Camilli (1992: 185) callously describe the archeological surveyor as "a sentient instrument capable of learning and of boredom." Because academics are necessarily surrounded by untrained workers, they assume the lowest common denominator for the skill level of individual crew members. Things are different in the CRM world, where experienced, dedicated, and highly-trained career archeologists comprise the field crews.
Proponents of "site-less" archeology focus their criticism on the supposed ineffectiveness of standard walkover survey for finding individual artifacts. They do not address the effectiveness of finding sites, because they do not acknowledge the concept of defining sites per se. Logic and experiences shows that it is possible to miss some artifacts along a transect and still find the others that constitute a site. The "non-site survey" approach makes this sort of sampling anathema. The approach requires that all artifacts must be found, in order to make complete sense of artifact loss/discard patterns across the landscape. In real-world archeology, however, we are happy to discover sites, even if a few isolates go undetected. In the realm of cultural resource management, this is considered an acceptable trade-off between data precision and data-collection cost.
A few highly experienced field archeologists (including some in academia) have looked at the costs and mechanics involved in walkover survey. When using the 15m transect interval in Illinois, the author has found that, as a rule of thumb, one person could cover 40 acres per day. This included recording time for two or three sites each with fewer than a couple of dozen artifacts. Recording included close-interval walkover (at 3m transects) in and within 15m of each artifact find spot, and individual piece-plotting of artifact locations using hand compass and 50m tape. Sites with more artifacts required extra recording time, but in general, the 40 acre/day rule worked nicely to estimate time and labor for 15m transect interval walkover surveys in northern Illinois.
In one of the first published accounts of a systematic site discovery survey, Lovis (1976) employed a 91m (100 yd) transect interval for locating sites in the wooded landscape near Little Traverse Bay in lower Michigan. He concluded that a 23m (25 yd) interval "would probably prove more reliable (Lovis 1976: 371)."
Schiffer and Wells (1982) examined ten variables which define the level of effort in archeological surveys. They assert that transect spacing "is probably the single most important factor affecting level of effort (Schiffer and Wells 1982: 347)." They recognize that the obtrusiveness of each archeological manifestation -- ranging in scale from an isolated flake to a temple mound -- is an important second factor that affects site discovery and, therefore, the level of effort involved.
In reviewing general problems with survey techniques, Schiffer and Wells (1982: 370) state that transect intervals of 20-50m are not adequate for discovering unobtrusive archeological sites. They point out that other researchers have demonstrated that higher precision requires greater survey intensity. They conclude that "a crew spacing of 15 m can be used on units of intermediate size to yield good discovery of smaller sites. Because unobtrusive remains and smaller sites are often more abundant, this procedure should provide a cost-effective compromise survey strategy (Schiffer and Wells 1982: 370)."
George Cowgill (1990) plays with numbers and rough estimates for the chances of site discovery to critique the 30m survey transect interval commonly employed across the American West. He assumes that a surveyor has a 100% chance of detecting a site if he passes 1m or more inside the periphery of site, a 95% chance if he passes between 1m inside and 1m outside its perimeter, a 75% chance if he passes 1-3m outside the site, a 50% chance if he passes within 3-5m outside, and no chance if he passes more 5m away. Cowgill (1990: 254) states that these figures are sheer guesses, but believes -- based upon his own experience -- that "they are plausible for light sherd and artifact scatters in places where vegetation and . . . other obscuring phenomena are fairly sparse but not absent." Given these probabilities and using a 30m transect interval, there would be only a 56% chance of detecting sites with a diameter of 10m or less. A 16m interval would be required to produce a 95% probability of detecting a 10m diameter site. Cowgill's basic assumptions are quite conservative, since they imply that a surveyor cannot see artifacts on the ground more than 5m distant. Certainly this is not the case for trained archeologists who possess adequate eyesight and are walking farm fields with good to excellent ground surface visibility.
A mathematical approach toward estimating the effectiveness of site discovery has been proposed. Sundstrom (1993: 92) discards intuition and examines "the mathematical relationships between the transect pattern and the probability of finding sites of a given size." If one assumes that any line transect will result in discovery if the transect intersects the site, it is a relatively straightforward matter to calculate the spacing required in order to insure discovery of sites of any particular dimension. Her formula (dealing with circular sites) is as follows, where P is the probability of discovery, r is the site radius, and d is the transect interval:
P = 2r ÷ d
So, for example, to get a 90% probability of finding all sites larger than 10m in diameter, it would be necessary to employ an 11.11m transect interval (Sundstrom 1993: 93). Using the Sundstrom formula, a 15m interval would insure 90% probability of finding 13.5m diameter sites, while 5m spacing would have a 90% chance of finding all 4.5m diameter sites. Of course, the mathematical formula assumes that a surveyor will always recognize a site if his transect intersects one. Further, and more importantly, the formula does not take account for the fact that surveyors can see artifacts and recognize sites "for several meters beyond the actual transect lines (Sundstrom 1993: 94)."
Thus, in settings where ground surface visibility is not an issue, Sundstrom's discovery estimates must be understood to be absolute minimum figures. Her formula can be modified to account for the swath of discovery that a survey transect actually represents. This can be done by adjusting the radius of the target site by whatever distance at which we consider a surveyor can see artifacts and recognize sites. The site radius figure is increased by whatever distance we assume the surveyor can see beyond his line transect. The adjusted formula is:
P = 2(r + v) ÷ d
where v is the distance at which the surveyor can recognize a site.
New estimates for the effectiveness of the 15m transect interval can thus be generated given various assumptions of the ability of field archeologists to recognize sites at a distance. Wandsnider and Camilla (1992) barely trust surveyors to find artifacts more than 2m off-transect. Assuming this meager distance, a 15m transect interval would have a 100% probability of finding all sites 11m in diameter or larger. In Illinois, the current 5m interval assumes that each surveyor can see at least 2.5m off-transect. Plugging this 2.5m visual acuity figure into the formula, we find the resulting 15m transect interval would insure 100% discovery of all sites 10m in diameter or larger. States surrounding Illinois recommend a 10m transect interval, a figure which assumes that surveyors can see up to 5m distance. If we plug that number into the modified formula, we find that a 15m interval insures a 100% chance of discovering all sites 5m in diameter or larger. Finally, if we employ the author's own (wildly optimistic?) visual acuity estimate of 8m (for open farm fields with excellent ground surface visibility), the resulting 15m interval insures 100% discovery of all sites.
References Cited
| Burger, Oskar, Lawrence C. Todd, Paul Burnett, Tomas J. Stohlgren, and Doug Stephens | ||||
| 2002 | Multi-scale and nested-intensity sampling techniques for archaeological survey. Journal of Field Archaeology 29(3-4): 409-423. | |||
| Cowgill, George L. | ||||
| 1990 | Toward refining concepts of full-coverage survey. In: S. K. Fish and S. A. Kowalewski (eds.), The Archaeology of Regions: The Case for Full-Coverage Survey, pp. 249-259. Smithsonian Institution Press, Washington, D.C. | |||
| Ebert, James I. | ||||
| 1992 | Distributional Archaeology. University of New Mexico Press, Albuquerque. | |||
| Kintigh, Keith W. | ||||
| 1990 | Comments on the case for full-coverage survey. In: S. K. Fish and S. A. Kowalewski (eds.), The Archaeology of Regions: The Case for Full-Coverage Survey, pp. 237-242. Smithsonian Institution Press, Washington, D.C. | |||
| Kvamme, Kenneth L. | ||||
| 1998 | Spatial structure in mass debitage scatters. In: A. P. Sullivan III (ed.), Surface Archaeology, pp. 127-141. University of New Mexico Press, Albuquerque. | |||
| Lovis, William A., Jr. | ||||
| 1976 | Quarter sections and forests: an example of probability sampling in the northeastern woodlands. American Antiquity 41(3): 364-372. | |||
| Redman, C. L., and P. J. Watson | ||||
| 1970 | Systematic, intensive surface collection. American Antiquity 35(3): 279-291. | |||
| Schiffer, M. B., and S. J. Wells | ||||
| 1982 | Archaeological surveys past and future. In: R. H. McGuire and M. B. Schiffer (eds.), Hohokam and Patayan: Prehistory of Southwestern Arizona, pp. 345-383. Academic Press, New York. | |||
| Sundstrom, Linea | ||||
| 1993 | A simple mathematical procedure for estimating the adequacy of site survey strategies. Journal of Field Archaeology 20(1): 91-96. | |||
| Wandsnider, L., and E. Camilli | ||||
| 1992 | The character of surface archaeological deposits and its influence in survey accuracy. Journal of Field Archaeology 19: 169-188. | |||
| Whalen, Michael E. | ||||
| 1990 | Sampling versus full-coverage survey: an example from western Texas. In: S. K. Fish and S. A. Kowalewski (eds.), The Archaeology of Regions: The Case for Full-Coverage Survey, pp. 219-236. Smithsonian Institution Press, Washington, D.C. | |||