Survey, Excavation, and Geophysics at Songjiaheba—A Small Bronze Age Site in the Chengdu Plain

Archaeological survey in the Chengdu Plain of Sichuan Province has revealed settlement patterns surrounding Late Neolithic walled sites, including large numbers of small settlements from the Neolithic, Bronze Age, and Han Dynasty eras. Here geophysical survey and excavation at one of these small-scale sites dating to the Middle Bronze Age are reported, showing for the first time the value of high-resolution geophysics for evaluating site size and integrity in the Chengdu region.


INTRODUCTION
The Chengdu 成都 Plain, in the northwest corner of the Sichuan 四川 Basin (Figure 1), was the setting for the emergence of a complex civilization in the second millennium BC. This civilization is most notably associated with the site of Sanxingdui 三星堆, in Guanghan 廣漢, where two sacrificial pits discovered in 1986 revealed a rich and unexpected collection of jade objects, ceramics, elephant tusks, and elaborate bronze and gold objects (Bagley 2001;Flad 2012;Sichuan 1999). The discovery of the Sanxingdui pits was followed by research at other sites in the Chengdu Plain, including several loci in the city of Chengdu that post-date Sanxingdui, such as an elite residential location called Shi'erqiao 十二橋 (Sichuan et al. 1987; Jiang 1998), a zone of ritual deposits, cemeteries, and settlement areas named Jinsha 金沙 (Chengdu Institute 2006;Chengdu & Beijing 2002;Zhu et al. 2003), and Shangyejie 商業街, a Late Bronze Age elite burial site with large log coffins filled with lacquers and other elite objects (Chengdu 2002). Based on the excavations of these sites, we now know that bronze-producing communities that commanded multi-community networks of resource acquisition existed in the Chengdu Plain starting at least as early as the middle of the second millennium BC (Flad and Chen 2013).

<FIGURE 1>
Elsewhere in the Chengdu Plain, research in the last two decades has also located nine walled sites that predate Sanxingdui and which collectively comprise most evidence for the Late Neolithic Baodun 寶墩 culture ; Wang 2003). The Baodun culture is characterized by certain fine-ware ceramics, groundstone tools, and wattle-and-daub architectural features that have been discovered at these walled sites. There is no evidence that these walled sites were integrated into a single large regional polity, but large building foundations at several walled sites suggest that they may have acted as central places for more localized multi-site corporate groups, perhaps as loci of ritual activity, elite compounds, or refugia for residents at scattered house compounds or small hamlets. The walls at the sites required communal labor investment for their construction, thereby supporting the notion that these sites were occupied and maintained by fairly sizable populations. The walls also fit into a pattern of increasing social complexity seen across China during the late Neolithic (Demattè 1999;Liu 1996;Ren 1998;Underhill 1994;Underhill et. al. , 2002Xu 2000;Yan 1999;Yang 2004). The preliminary work done at these Neolithic walled sites, together with the research on Sanxingdui, Jinsha, Shanyejie, and other loci around Chengdu, has produced a basic archaeological chronology for the region (Table 1).

<TABLE 1>
At present, our basic understanding of the social developments in this region is based mainly on the investigation of these large, politically central sites. This research has allowed researchers to posit a transition of political power, and possibly population migration, from Sanxingdui to Jinsha and other loci in Chengdu during the middle part of the Bronze Age. This transition may have been violent, as we have iconographic evidence for war captives in the form of kneeling stone human figurines with hands bound behind their backs at both sites (Chengdu Wenwu 2006). Furthermore, non-functional weapons made of jade from both of these sites reflect the symbolic importance of artifacts of war in the associated society. 1 It is estimated that the Sanxingdui-Shi'erqiao transition took place sometime around 1200-1150 BC (Xu 2003).
Research to date has not, however, provided a systematic understanding of settlement patterns in the region, nor has much attention been paid to small sites, which were distributed across the plain and which probably housed a considerable portion of the population. Small settlements are poorly known from all parts of China throughout the Bronze Age because archaeologists have tended to focus on locations that are perceived to have been political centers.
In the Central Plains of North China, for example, attention has been focused almost entirely on cemeteries or large, walled settlements, many of which are often identified by researchers as possible former capitals of the dynastic houses that ruled the region according to traditional histories. Recent survey work and increased attention to smaller settlements has begun to mitigate this bias. 2 In the Sichuan basin, research on small sites has occurred occasionally in the Three Gorges and in areas around Chengdu where, respectively, a large dam project and urban development have created ad hoc opportunities for small settlements to be investigated (discussed in Flad & Chen 2013), but understanding of such sites has remained underdeveloped because of lack of systematic survey work and a predisposition to concentrate attention on larger sites with walls or relatively abundant, stratified deposits. A basic understanding of smaller settlements is vital, however, if we wish to begin to examine the broader network of communities that sustained the emergence of complexity. This is true in Sichuan as well, where very little attention has been paid to small settlements and their distribution. How widely spread and densely packed were settlements across the plain during various periods in the cultural chronology in this region? How diverse were these sites and what was their basic composition? University in St. Louis, McGill University, and National Taiwan University to answer some of these questions (Chengdu 2010). This project, the Chengdu Plain Archaeological Survey (CPAS), is part of a trend in Chinese archaeology to develop an understanding of settlement patterns through systematic regional survey. 3 Many of the surveys cited here have added crucial new data to our understanding of how early complex polities operated at a regional scale by outlining the spatial relationships among contemporary sites throughout several phases of prehistory. Similarly, CPAS aims to produce data that are representative of settlement patterns during the Baodun, Sanxingdui, Shi'erqiao, and later periods, and discuss the relationships among settlements during these different chronological phases.
As explained extensively elsewhere (Chengdu 2010), the survey procedures adopted by the CPAS project have been tailored to the particular features of the Chengdu Plain landscape, which is dominated by rice paddies, low degrees of surface visibility, and dispersed house compounds. Over the course of five seasons of fieldwork, we systematically examined approximately 347.5 square km using teams of archaeologists walking at approximately 50 meter intervals throughout a highly fragmented landscape of agricultural fields and house compounds.
The survey team collected all ceramics that were thought to be of Han period or earlier and marked these "find-spots" with GPS coordinates. Artifacts discovered in terrace walls, stream banks or other exposed profiles were similarly collected, and finds within 20-30 meters were collected together. In cases where decorated bricks or other identifiable architectural elements were identified, these were photographed and marked, but not always collected. Find-spots for surface collections are marked on Figure 2.

<FIGURE 2>
Due to poor ground visibility in many fields and the fragmented landscape being surveyed, we combined traditional field-walking survey techniques with systematic augering along transects to identify site densities and general patterns of settlement. Our procedure involved coring of four auger holes at "primary points" located at approximately 200 meter intervals along transects that cross the survey area. A total of 1,988 primary points were investigated over the course of the project. In those places where one or more of these primary holes recovered Han or earlier remains, additional "secondary points" were cored on transects extending to the cardinal directions from the primary point. Secondary points were first cored 10 meters from the original point, and when additional archaeology was discovered, the coring continued at successive 20 meter intervals along the four transects. In Figure 2, those locations where "secondary auger points" are marked near primary auger points are places where subsurface remains indicated concentrated loci of archaeological remains. These "activity areas" are the focus of the various research projects of the CPAS project.
Compared to previous work in the region, the CPAS project has been much more extensive. Most previous work has opportunistically focused on sites that were under threat from construction projects, or surveys of areas to be affected by road or railroad routes. Little effort has previously been focused on ensuring the representativeness of the data being collected whereas CPAS is aimed at creating a systematically collected picture of the area. The survey has focused primarily on a roughly circular area with a 10 km radius surrounding the walled Baodunculture site of Gucheng 古城 in Pi 郫 County (see Figures 1 and 2), as well as a secondary zone around the site of Yufucun 魚鳧 in Wenjiang 溫江 County, and has been able to identify many small sites in this region. The general results of the survey project are beyond the scope of this paper, but will be the subject of a project report currently in preparation.
In addition to establishing a new perspective on spatial patterns of settlements across the Chengdu Plain, CPAS has focused on evaluating the nature of small-scale settlement areas. As the CPAS survey area sites have been discovered, the project team has evaluated their integrity, size, and composition through augering, geophysics, and small-scale excavations.
Archaeological geophysics has not been used extensively or effectively in Chinese archaeology.
This project marks one of the first attempts to employ high-resolution geophysics for any archaeological purpose in China. As such, it remains a procedure that requires extensive ground-truthing, but the CPAS project has shown that it can be an effective approach for a variety of purposes in Chinese archaeology, including the evaluation of small-scale settlements.
Among the first of the CPAS sites to receive extensive attention was Songjiaheba 宋家河 壩 , which was threatened by local villagers extracting natural sand and gravel deposits from the site. Research at Songjiaheba demonstrated that it was a small residential area during the second half of the second millennium BC, roughly contemporary with the transition between the Sanxingdui and Shi'erqiao cultures. The research at Songjiaheba represents the first study of a small settlement from this period. It was also the first use of high-resolution geophysics in combination with survey and excavation data to understand a site of this nature in China.

SITE DISCOVERY
The site of Songjiaheba (centered on 0396996E 3414501N UTM Zone 48R, see

<FIGURE 3>
Subsequently, the auger survey transect crossed through the same location and two consecutive "primary points" were drilled within anthropogenic deposits (Plate 1 and Figure 4).
These primary points showed that anthropogenic soils were found on the terrace to the east of the sand-borrow pit. Subsequently, secondary points augered along transects in cardinal directions extending outward from the primary points allowed us to estimate that the preserved portion of the site extended approximately 220 meters from north to south, and reached east of the sandborrow pit approximately 100 meters (see Plate 1 for locations of surface survey finds, primary points and secondary points at the site, along with the relative abundance of datable archaeological material from the various locations). Whereas the ceramics recovered from these auger holes were similar to those collected during the surface survey, the auger holes were more effective at helping delineate the extent of the site. Surface finds, while ubiquitous on one side of the borrow pit, did not extend to the east where intact sub-surface remains were found. They date primarily to the Bronze Age, a period from which few small-scale settlement sites have been identified or investigated in the Chengdu Plain.

<FIGURE 4>
The auger testing at the site suggested that there were two cultural strata present. This was confirmed in the profile exposed by the sand-borrow pit. Samples of carbonized materials were collected during the site assessment from the edge of the borrow pit at location 06PXC0005. One (FCN 2261) was collected from about 75 cm below ground surface, while the other (FCN 2262) was collected from 1.3 meters deep. The radiocarbon dates for these samples suggest that the lower sample came from a stratum dated to the middle of the second millennium BC, while the later sample dated from the period of overlap between the Sanxingdui and Shi'erqiao / Jinsha cultural periods ( Table 2). Although one must be very cautious when using only two radiocarbon dates recovered from exposed strata, these dates accord well with the excavation results from the site discussed below. A range of geophysical techniques are available for archaeological prospection, including electrical methods (e.g. earth resistance), magnetic methods (e.g. magnetometry), and electromagnetic or EM methods (such as ground-penetrating radar or GPR). Based on the known ground conditions and the nature of the expected archaeological features, it was decided that magnetometry would be the most appropriate method for work in the Chengdu Plain. Earth resistance survey may be less effective due to the expectation that some archaeological remains would be below the detection limits of these techniques, but also because of the high water content of the soils resulting from irrigation. Measurement of soil resistance provides a measurement of the moisture content and porosity of buried materials. Saturated ground reduces the moisture contrast between archaeological features and natural sediments. EM techniques, including GPR, are particularly sensitive to the conductivity, hence water content, of the ground and therefore have the same limitations as earth resistance. Furthermore, the dense network of rice paddy walls that divide up the landscape in this region would make a technique such as GPR extremely difficult.
In contrast, magnetometers have several advantages that make them attractive for use in this region. Magnetometry is the most rapid prospection method, can positively locate a wide range of anthropogenic features, and is well established in archaeological prospection (Aspinall et al. 2008;Clark 1990;Gaffney & Gater 2003;Scollar et al. 1990). This technique relies on the magnetic contrast between archaeological features and surrounding natural sediments. It is therefore highly dependent on the geological and pedological environment at a site. When successful, it is possible to detect the slightly more magnetic fills of buried pits and ditches and features such as hearths where the soils and clays have been burnt, thereby enhancing their magnetization. The magnetometer used for this project, the Bartington Grad601-2 dual gradiometer, can detect features up to 3 meters deep, although this is dependent on the size of the feature and the degree of magnetic contrast with the surrounding material.
Magnetometer survey involves carrying an (usually handheld) instrument along transects within a gridded survey area ( Figure 5). Since vegetation and ground conditions effect the ease of data collection, in the Chengdu Plain, magnetometer surveys are restricted to those times of year when rice paddies are drained, but before or after they have a dense vegetable crop. In the case of these investigations at Songjiheba, the paddies within the survey area had been abandoned and were either stripped of topsoil or completely excavated to access the sand and gravel. The magnetometer survey discussed here was undertaken within 20 meter squares at a resolution of 0.5 m x 0.125 m, making it possible to identify the majority of wall, ditch, hearth, and large pit features.

<FIGURE 5>
Following the geophysical survey at Songjiaheba, a selection of soil and sediment samples were collected in order to allow their magnetic susceptibility (MS) to be determined in the laboratory. The MS value of a material is dependent on how readily it becomes magnetized when placed in a magnetic field. Therefore, a contrast between natural and anthropogenic deposits can be indicative of the likely success of magnetometer surveys (Table 3). A sample of anthropogenic material taken from the exposed section has a susceptibility of 23.5 x 10 -8 m 3 kg -1 , compared to a sample of adjacent 'natural' soil (5.8 x 10 -8 m 3 kg -1 ). These limited MS measurements indicate that a sufficient magnetic contrast exists between anthropogenic and natural deposits for archaeological features to be detectable with this technique.

<TABLE 3>
The magnetometer results from Songjiaheba, shown in Plate 2a, reveal a complicated picture of anomalies. These are evident across the entire survey, including the northwest area of the survey where the soil has been completely stripped down to the underlying sands and gravels. While some of these anomalies are probably due to topographic variations resulting from the quarrying of sand and gravel that left the surface uneven, the results indicate that the natural deposits are magnetically heterogeneous and produce clear responses. This is backed up by a considerable range in the MS values of a number of rocks collected from gravel in this area (Table 1). This range reflects the variety of rock types that make up the gravel. While individual stones with high MS will produce intense magnetic anomalies in their vicinity, at a distance (i.e., when buried under a meter of soil), individual rocks become virtually undetectable. However, larger natural features, such as sand lenses within the gravel, will produce measurable anomalies at the ground surface if an MS contrast exists between the sand and the surrounding gravel.
Despite the magnetic noise, it is possible to identify archaeological responses within the gravel. The more intense anomalies (in excess of ±12nT) at [1] in Plate 2b coincide with concentrations of brick, associated with Han tombs that had been sunk into the gravel but are now exposed. Whereas the bricks are exposed, they are no longer in situ, and no other components of the features remain. The attribution to the Han period is based on the brick composition, although no clear brick decoration was observed. A similar anomaly is seen at [2], which may also be due to a buried brick (possibly Han period) tomb.
In contrast to the quarried section of the survey area, the southeastern region is characterized by much subtler magnetic anomalies, generally less than ±2nT in strength. These are more difficult to interpret, as a natural origin cannot be ruled out. As discussed above, it is possible that variations within the sand and gravel deposits could produce such detectable anomalies. Without further investigation (i.e., excavation), it is not possible to state their origin with any certainty. However, there is evidence to suggest that they may have archaeological sources.
Firstly, in the southwestern part of the survey area there is a clear boundary between the subtle anomalies and a magnetically quiet area at [3]. It is possible that this represents a natural change in magnetic properties, (e.g., marking the edge of a palaeochannel); alternatively it might indicate the extent of buried archaeological features. It is interesting to note that the point where this boundary meets the edge of the quarry coincides with the location where anthropogenic deposits were visible in section. We consequently interpret this line as being the southwestern edge of the site area as reconstructed on Plate 1.
The lack of magnetic anomalies in area [3] also indicates that the ca.1.5 m overburden of soil/subsoil is sufficient to reduce the effect of natural responses to a level where they are essentially invisible. Unless there is a dramatic variation in the height of the gravel deposits, it is unlikely that the subtle anomalies found within the site area are geological in origin.
Numerous weakly positive responses have been identified (Plate 2b). The stronger of these may be due to discrete areas of magnetic enhancement associated with burning. They could therefore indicate the locations of hearths or pits containing burnt material (e.g. anomalies [4] to (07CPST1) and three trenches (07CPST2-4) were opened at the site (Plate 3). As a one-meter baulk was left around T1, a 9 x 9 m section of this unit was excavated. The other excavations comprised one 9 x 2.5 meter trench (T2) and two 6 x 1.5 meter trenches (T3-4). The excavations exposed 10 pits of various sizes, two burials, and one partial house foundation. An additional two 1 x 1 m units were excavated in July 2009 with the aim of examining the two round negative anomalies in the hope of extracting archaeobotanical material from the site (see below).
Excavations exposed five stratigraphic levels in this region ( Figure 6)

<FIGURE 6>
The first small 1 x 1m excavation unit, TU1, was placed at the location of magnetic anomaly [8] (see Plate 3). This unit was excavated to sterile soil at 60 cm. Three stratigraphic layers were visible in this unit. Directly underneath the topsoil, an artifact and ash-rich layer was unearthed. Although the boundaries of this feature were not found within this 1 x 1 m unit, the large quantities of ash, pottery, and grinding stone contained in this matrix are consistent with the fill of a trash pit. A large soil sample (53L) was taken from this layer. TU2, a second small unit placed over anomaly 12, had more complex stratigraphy and, like other areas of the site, contained 5 stratigraphic layers. Underneath topsoil, the second stratigraphic layer contained "Shang / Zhou" period (ca. 1500-300 BC) pottery and some ashy inclusions. 4 An 11L soil sample was taken from this layer. A thin stratum (Layer 3) containing no pottery or ashy inclusions was not sampled. Layer 4 contained a few ashy inclusions and an additional 13L soil sample was collected. Situated underneath this layer and cutting into what appeared to be sterile soil were the clear boundaries of a pit. An additional large sample of 31 L was taken from this feature.
The features uncovered in these six excavation units are consistent with those expected at a small settlement. The pits were all relatively shallow, between 30-60 cm in depth. They varied widely in shape, from circular to rectangular, and contained fragments of broken pottery and stone. Several artifacts were found near the partial house structure, a rectangular building constructed with wooden posts set into a shallow wall trench. The structure was approximately 4.3 meters along the one preserved side, but had no distinguishable features other than a wall trench and a single post-hole (Figure 7).

<FIGURE 7>
Ceramics found in house feature F1 and in contemporary pits that are covered by Level 5 suggest that this stratum was similar in date with Level 6 at the Meiyuan Locus of the Jinsha site (Chengdu 2004) and levels 12-13 at the site of Shi'erqiao (Sichuan et al. 1987). As pointed out by Xu (2003), these strata are contemporary with the latest Bronze Age levels at Sanxingdui, a period that somewhat post-dates our earlier radiocarbon date from the augering mentioned above. Of course, the radiocarbon sample does not come from an excavation context and merely provides a starting point for discussing chronological questions at the site. The features below level 5 may date to somewhere around 1200-1150 BC.
The Songjiaheba burials were more or less contemporaneous with each other, although M1 cut into M2 and was therefore slightly later ( Figure 8). They post-date the house foundation and Level 5. M1 was 2.95 X 0.87 meters and 0.50 m deep oriented at about 315º. M2 was almost the same size (2.95 X 0.90 X 0.50 m) and orientation. Human remains in both graves were poorly preserved, although teeth fragments remained in the northwest portion of both graves. The occlusial wear on the teeth discovered in M1 are consistent with those expected from an adult, perhaps near 30 years in age. In M2, the individual should be around 9 years old based on the recent eruption of the lower canines.

<FIGURE 8>
Both burials contained burial goods including ceramic vessels and stone "sticks" placed in the midsection of the deceased that were similar to objects found at other sites such as Shuiguanyin, in Xinfan county, not too distant from the Songjiaheba site (Sichuan 1959). The function of these stone sticks is unclear and more work is necessary to clarify how widespread this burial practice was in the region. The sticks were quite fragmentary as was apparently also true of those found at Shuiguanyin. M1 at Songjiaheba also contained a total of 10 ceramic vessels ( Figure 9A Figure 9B). These objects suggest that the burials were relatively close to one another in date. They may have been members of a single family, with a younger individual buried earlier than an older one. The stone "sticks" in both burials suggest a chronology contemporaneous with Shuiguanyin, the pointed-bottom saucer in M2 has parallels in level 12 at Shi'erqiao, and the pointed-bottom cup from the same burial is similar to ceramics found in Shi'eraio level 11. Materials from level 4, superimposed on top of the graves, are similar to objects from Levels 11-10 from the site of Shi'erqiao, perhaps contemporary with the early to middle parts of the Western Zhou (ca. 1150-771 BC; see Jiang 1998). All of these comparisons, coupled with the latter of the radiocarbon dates listed in Table 2, suggest that the graves date to sometime around 1100 BC.

COMPARISON OF EXCAVATION AND GEOPHYSICS RESULTS
These excavation results have allowed the magnetometer data from Songjiaheba to be assessed. It is clear that features containing burnt deposits and high concentrations of ceramic material produce detectable positive magnetic anomalies, the shape of which can be related to the source feature. However, perhaps in part due to the underlying magnetic responses to geological variations, other more subtle features have not been detected. The excavated structure, for example, does not appear to have been identified by this survey.
Pits filled with wood ash, despite their association with burning, do not usually produce a positive magnetic anomaly because of the absence of iron minerals in charcoal. Conversely, high quantities of such ash will produce a negative response because it contrasts negatively with natural soils and sediments. This has been shown to be the case at Songjiaheba as one negative anomaly was revealed to be associated with a wood ash-rich deposit. Unsurprisingly, since they often do not exhibit a sufficient magnetic contrast and are therefore difficult to identify in many environments, the graves did not produce any magnetic responses that would have aided in their location.
While somewhat problematic for the identification of subtle anomalies of archaeological origin, the ability of the magnetometer to respond to geological features has the benefit of providing information regarding buried geomorphological features. Results from other sites in the region (e.g. Horsley 2010) suggest that the pattern of magnetic responses in the northern portion of the survey indicate a palaeochannel that appears to divide the site in two. It is not possible to determine the age of this former river channel from the geophysical data alone. It is therefore unclear whether it was contemporaneous with the settlement or it represents a later event that washed away this area of the site.

ARCHAEOBOTANICAL RESULTS
Sediment samples were processed using simple bucket flotation similar to that described in Pearsall (2000). The light fraction was poured onto a 0.25mm nylon mesh and the heavy fraction was processed using a 0.5mm mesh sieve. To ensure efficiency and maximum data collection, different methods were applied to the sorting of fractions. The 2mm fraction was sorted in its entirety. Seeds and plant parts were then identified while charcoal was reserved for subsequent analysis. The 1mm, 0.5mm and 0.25mm fractions were sorted in their entirety for both carbonized and modern seeds, however charcoal and other parts were not pulled from these fractions. Identifiable plant parts, such as rice spikelet bases, were pulled from all fractions.
The Songjiaheba sediment samples yielded low densities of charred plant remains (see Table 4). Despite the fact that these remains were too sparse to aid in the interpretation of the site, they provide us with the first picture of the plants associated with subsistence at Sanxingduiculture settlements. Only recently has any archaeobotanical work been done in Sichuan; none of it has yet been published for the Chengdu Plain. 5 Extensive archaeobotanical analysis currently being undertaken by one of this paper's authors (D'alpoim Guedes) will soon provide a deeper understanding of agricultural activity and other plant exploitation in the region.

<TABLE 4>
Subsistence at Songjiaheba appears to have focused on rice (Oryza sativa). Rice remains constitute the most numerous taxa, with a total of 90 specimens in this assemblage. Examination of the rice spikelet bases confirms the presence of domesticated rice (Fuller et al. 2009). Most of these specimens were very fragmentary with only 10 measurable rice grains (see Table 5). In addition to rice, small amounts of foxtail millet (Setaria italica), peach pit (Amygdalus (Prunus) persica), and wild grape (Vitis sp.) were recovered, as were seeds of members of the Poaceae family and Amaranthaceae family. While one cannot rule out their role as food, these are also common arable weeds. It is likely that these weeds grew alongside the rice or millet plants and were introduced to the site when the harvest was brought in.

DISCUSSION AND CONCLUSION
The site of Songjiaheba was a small settlement occupied in two separate phases during the second half of the second millennium BC. In the earlier phase we found evidence of pits used to dispose of trash in the vicinity of residences. The second phase included two burials, most likely interred within a small settlement area. The site probably comprised a small hamlet, around 1.5 ha in size, near the banks of a stream. The general position and size of the site has been established using a combination of augering and magnetometer survey. Numerous magnetic anomalies consistent with buried archaeological features have been identified at the site and, although ground-truthing remains somewhat limited, is appears that many of these may have archaeological origins. Excavations have shown that magnetometry is particularly useful for identifying pit features. These features have provided the first data available on plant exploitation during the early Bronze Age in the Chengdu Plain.
These data contribute to the recent boom in survey archaeobotany in East Asia. The primary goal of extracting archaeobotanical data through survey has been to collect material from sites either not currently under excavation or to resample sites which were previously excavated. Much of this previous work has been carried out in regions where archaeological features are visible in agricultural terraces. As most sites in the Chengdu Plain are not visible on the surface, new methods of deciding where to target features are necessary. Magnetometry has been able to successfully locate such features.
Modern agricultural practices present limitations for geophysical survey in this environment. The extant paddy walls produce distinct anomalies in the survey data. While these are readily identifiable and can be ignored, they may obscure anomalies due to archaeological features buried below. Interestingly, no anomalies due to ancient paddy walls have been recognized, although it seems likely that the fields shifted over time.
Of greater concern for actually undertaking the surveys and collecting data are the modern crops. The magnetometer is a handheld instrument that does not need to make contact with the ground; however, the operator is required to walk along parallel lines spaced either 2m or 1m apart. It is not possible to survey areas where valuable crops are closely spaced without damaging them. Not only does this slow the rate of survey, it results in gaps in areas where agricultural fields are present.
The area around the site at Songjiaheba was being excavated for sand and gravel, not cultivated, so this problem was mitigated. Elsewhere in the CPAS survey area, ground conditions were more disruptive. One solution to this would be to conduct geophysical surveys at a different time of year, for example after the paddies are drained and the ground firm enough to walk on, but before crops would be damaged. Alternatively, if the crops are rotated, it should be possible to build up a complete picture by undertaking surveys over consecutive years.
Despite these issues, the potential of geophysics as a method to complement traditional survey techniques is demonstrated by the work at Songjiaheba. Furthermore, the work here shows that systematic survey even in difficult environments has the potential to significantly add to our knowledge about ancient settlement patterns. Prior to the initiation of the CPAS survey, sites such as Songjiaheba were not known from across the Chengdu Plain. It has been difficult, therefore, to determine whether the highly dispersed settlement pattern that existed through the historical period was also typical for the Neolithic and Bronze Ages. Songjiaheba was a small settlement location. It was perhaps not terribly consequential as a location of social or political influence during the period when it was occupied. However, the CPAS survery area and surrounding regions were quite possibly filled with small household groups and hamlets just like Songjiaheba, occupied by small families who maintained networks of relationships across the Chengdu Plain. Songjiaheba was one of hundreds of nodes in a web of human settlement that collectively made up the prehistoric settlement pattern of the region. We need to understand such small nodes if we hope to create a more nuanced interpretation of early occupation of this area.

ACKNOWLEDGEMENTS
The project depended on support from members of the Chengdu City Institute of Archaeology and funding from all collaborating institutions as well as the Wenner-Gren Foundation and the Luce Foundation. Two anonymous reviewers commented on a previous version of this manuscript and helped us clarify the significance of the project, and two reviewers for Asian Perspectives helped strengthen and clarify the arguments we present.         COLOR PLATESfor 2 sided insert Plate 1. Map of Songjiaheba showing the locations of cultural material found during the surface collection and auger surveys, the geophysical survey grid, the probable site extent, and the area of the sand-borrow pit. Location 06PXC0005 is the initial findspot, one of two surface findspots in the area (marked with open triangles). All other points on map are primary or secondary auger points. The two primary auger points are marked with a "P" adjacent to their location. For a detailed view of the magnetometer survey grid, see Plate 2.
Plate 2. Results of magnetometry survey at Songjiaheba. Left image, a: shows magnetometer data combined with auger survey results (white squares with dots, at locations with artifacts and without dot when no artifacts were present). Locations of subsequent excavation trenches are shown with dashed lines. Open triangle with green dot shows one of two surface findspots. Results are superimposed on satellite imagery from GoogleEarth. Right imageb: shows interpretation of signifi cant magnetic anomalies. Plate 3. Magnetometry data (left) and anomalies of probable cultural origin (right) for central area of cultural anomalies showing locations of excavation trenches T1-T4 and exploratory units TU1 and TU2. The grey angled feature in T1 is the foundation for house structure F1. Numbers refer to magnetic anomalies discussed in the text. B.