1. Introduction

Meta-analyses of archaeozoological data provide explanations of fine-grained questions using large databases (Conolly et al. 2011; Cranbrook and Piper 2009; Emery 2007; Grayson and Meltzer 2002; Grayson 2005; Rick et al. 2011). A classic example of a meta-analysis is the recognition that we only have evidence for Clovis-era hunter-gatherers in North America exploiting two of the 35 large mammal genera that became extinct during the late Pleistocene (Grayson and Meltzer 2002). Meta-analyses of faunal data provide (i) quantifiable means to test hypotheses about regional or cultural adaptations; (ii) an understanding of the composition and state of a regionally-geographically specific eco/arti-factual database; and (iii) information on shifting patterns in the archaeological record through time. In Southeast Asia, broad syntheses of different eco/arti-fact classes have shed light on topics as diverse as prehistoric landscape modification and forest disturbance (Hunt and Rabett 2014), geographical distribution of rice production and consumption (Castillo 2011; Castillo and Fuller 2010), and historic trends in the methodological and conceptual analysis of flaked stone artifacts (Marwick 2007). Aside from qualitative syntheses of Southeast Asian faunas (Gorman 1971b), little archaeozoological meta-analysis has occurred in this region.

Identifying subsistence change at the Pleistocene-Holocene transition is a popular topic in archaeozoological endeavors (Conolly et al. 2011; Stiner and Munro 2002; Zeder 2012). This transition is significant during human prehistory because of worldwide climatic change that transpired, away from a glacial period into the climatic and environmental regime that we see today (Straus et al. 1996). In mainland Southeast Asia, this transition is significant but few meta-analyses attempt to understand the inherent subsistence changes that may have emerged (cf. Rabett 2012).

In this paper I conduct a compositional overview of the late Pleistocene and Holocene archaeozoological record in Thailand and Peninsular Malaysia to establish the methodological and quantitative patterns found within this large faunal dataset. I then use nestedness and squared chord distance metrics to explore shifting faunal exploitation in Thai-Malay assemblages spanning the Pleistocene-Holocene transition.

2. The Dataset

The primary goal of this analysis is to examine trends in forager subsistence activity in Thailand and Peninsular Malaysia, thus this sample is limited to a total of 28 sites within both regions that span the late Pleistocene to Holocene and have archaeozoological deposits that are primarily pre-agricultural in nature (Figure 1; Table 1). I use the presence of ceramics as a coarse proxy for agriculture. There are two important caveats to this.

Figure 1 

Location of sites discussed in text. (1) Spirit Cave, Banyan Valley Cave, Tham Phaa Can (2) Tham Lod Rockshelter, Ban Rai Rockshelter (3) Ban Chiang (4) Non Nok Tha (5) Sai-Yok Rockshelter (6) Khao Talu Cave, Heap Cave, Ment Cave (7) Lang Kamnan Rockshelter (8) Nong Nor (9) Pak Om Cave (10) Buang Beab Cave (11) Lang Rongrien Rockshelter, Khao Toh Chong Rockshelter, Moh Khiew Cave I, Moh Khiew Cave II (12) Sakai Cave (13) Thung Nong Nien Rockshelter, La Sawang Rockshelter (14) Gua Kepah (15) Gua Gunung Runtuh, Gua Harimau, Gua Ngaum, Gua Teluk Kelwar (16) Gua Peraling, Gua Chawas, Gua Cha (17) Seberang Perak (18) Gua Tenggek, Gua Sagu (19) Gua Kechil.

(North to South)
Location Recorded Faunal Data Available Radiocarbon Dates
Site References
(*Faunal Reference)

Spirit Cave Mae Hong Son Province, Thailand Presence/Absence 11690 ± 560 –
7,622 ± 300 14C a BP
Gorman 1969; Gorman 1971a; 1971b; 1972; *Higham 1977; Lampert et al. 2003; White 2004 SC
Banyan Valley Cave Mae Hong Son Province, Thailand Presence/Absence 5,360 ± 120 –
930 ± 80 14C a BP
*Higham 1977; *Reynolds 1992 BVC
Tham Phaa Can
(Steep Cliff Cave)
Mae Hong Son Province, Thailand Descriptive 7,500 ± 160 –
5,180 ± 110 14C a BP
*Higham 1989; White and Gorman 2004 TPC
Tham Lod Rockshelter Mae Hong Son Province, Thailand NISP 34,029 ± 598 –
12,100 ± 60 14C a BP
*Amphansri 2011; Marwick 2013; Marwick and Gagan 2011; Shoocongdej 2006 TL
Ban Rai Rockshelter Mae Hong Son Province, Thailand Descriptive 10,210 ± 50 –
7,250 ± 40 BP
*Treerayapiwat 2005 BR
Sai-Yok Rockshelter Kanchanburi Province, Thailand Presence/Absence N.D. (Hoabinhian-Holocene) van Heekeren 1961; van Heekeren 1962; *van Heekeren and Knuth 1967 SY
Khao Talu Cave Kanchanburi Province, Thailand MNI 9,530 ± 1050 –
2,800 ± 300 14C a BP
Pookajorn 1981; Pookajorn 1984; *Pookajorn 1988 KT
Ment Cave Kanchanburi Province, Thailand MNI 8,400 ± 640 –
3,200 ± 370 14C a BP
Pookajorn 1981; Pookajorn 1984; *Pookajorn 1988 MC
Lang Kamnan Cave Kanchanburi Province, Thailand NISP 27,110 ± 500 –
6,110 ± 60 14C a BP
*Mudar 1996; Shoocongdej 1996b Shoocongdej 2000; Shoocongdej 2010 LK
Long Rongrien Rockshelter Krabi Province, Thailand NISP 43,000 –
2,530 ± 45 14C a BP
Anderson 1990; Anderson 1997; Bulbeck 2014; *Kijngam 1990; *Mudar and Anderson 2007 LR
Khao Toh Chong Rockshelter Krabi Province, Thailand NISP 13,026 ± 45 –
149 ± 25 14C a BP
Conrad et al. 2013; Marwick et al. 2013; *Van Vlack 2014 KTC
Moh Khiew Cave I Krabi Province, Thailand MNI 25,800 ± 600 –
4,240 ± 150 BP
*Chaimanee 1994; Pookajorn 1991; Pookajorn 1996 MKI
Moh Khiew Cave II Krabi Province, Thailand NISP 12,480 ± 830 –
2,100 ± 690 14C a BP
Auetrakulvit *2004a; Auetrakulvit 2004b; Auetrakulvit 2005; Auetrakulvit et al. 2012 MKII
Sakai Cave Trang Province, Thailand MNI 9,280 ± 180 –
7,620 ± 160 14C a BP
*Chaimanee 1994; Pookajorn 1991; Pookajorn 1996 SK
Thung Nong Nien Rockshelter Trang Province, Thailand NISP 9,810 ± 420 –
8,370 ± 370 14C a BP
Auetrakulvit *2004a; Auetrakulvit 2004b; Auetrakulvit 2005 TNN
La Sawang Rockshelter Trang Province, Thailand Descriptive 14,000 – 10,000 years BP *Auetrakulvit 2005 LS
Guar Kepah Penang Province, Malaysia Descriptive 5700 ± 50 BP Bulbeck 2003; *van Stein Callenfels 1936 GK
Gua Gunung Runtuh Perak Province, Malaysia NISP 13,600 ± 120 –
2,620 ± 80 14C a BP
Bulbeck 2014; *Davidson 1994; *Zuraina et al. 1994 GGR
Gua Harimau Perak Province, Malaysia MNI 14,140 ± 795 –
1,760 ± 195 14C a BP
*Bulbeck 2003 (Original-Zolkurnian 1998); Bulbeck Pers. comm. 2014 GH
Gua Ngaum Perak Province, Malaysia MNI 6,370 ± 90 –
5,990 ± 80 14C a BP
*Bulbeck 2003 (Original-Zolkurnian 1998); Bulbeck Pers. comm. 2014 GN
Gua Teluk Kelawar Perak Province, Malaysia MNI 10,245 ± 80 –
6,100 ± 100 14C a BP
*Bulbeck 2003(Original-Zolkurnian 1998); Bulbeck Pers. comm. 2014 GTK
Gua Peraling Kelantan Province, Malaysia NISP 12,000 years BP – Present *Adi 2007 GP
Gua Chawas Kelantan Province, Malaysia Descriptive 12,550 years BP – Present *Adi 2007 GCH
Gua Cha Kelantan Province, Malaysia Presence/Absence 6280 ± 170 –
820 ± 80 14C a BP
Bulbeck 2003; *Groves 1985; Sieveking 1954 GC
Seberang Perak (Bukit Perang) Perak Province, Malaysia Descriptive 5,970 ± 50 BP *Adi 1983; Bulbeck 2014 SP
Gua Kechil Pahang Province, Malaysia NISP 4,800 ± 800 14C a BP *Dunn 1964, Dunn 1966; *Medway 1969 GKH
Gua Tenggek Pahang Province, Malaysia NISP 10,660 ± 110 –
10,545 ± 80 14C a BP
*Rabett 2012; Zuraina et al. 1998 GT
Gua Sagu Pahang Province, Malaysia NISP 14,410 ± 180 –
1,240 ± 100 14C a BP
*Rabett 2012; Zuraina et al. 1998 GS

Table 1

Archaeological sites, locations, methods of faunal recording, available date ranges and key references for Thai-Malay sites. This table uses published radiocarbon dates, including mollusk dates (Bulbeck 2014), as such; the dates reported here should not be regarded as an absolute chronological range. Instead these dates act as a coarse date range. For each site, see cited literature for exact chronological details.

* = Represents publications where faunal data is extracted.

First, many of the reported radiocarbon dates for these sites come from uncontrolled deposits (not in stratigraphic association), or from material (molluscs) that have uncorrected dates which do not account for reservoir effects or diagenetic processes (Bulbeck 2014). Second, while I am primarily interested in forager subsistence strategies, several of the sites span the Neolithic transition into the Protohistoric period. Commonly included in these later Holocene assemblages are increased ceramic accumulations and adzes, both suggesting a sedentary agricultural adaptation.

At Khao Toh Chong Rockshelter, these artifact classes increase rapidly in abundance after 7,000 years ago (Van Vlack 2014). Thus, it is difficult to tell when faunal remains were deposited due to mobile hunter-gatherer groups or agriculturalists in nearby settlements. I do not attempt to dichotomize faunal data to differentiate these groups (although this would be a worthwhile and interesting endeavor in the future). Instead, I simply use all available data during the span of these sites and assume that hunter-gatherer groups are the primary accumulating agents. For this reason, I specifically exclude sites and faunal data from known settlement agricultural sites (generally open sites in flood plain valleys). Some well-documented examples of these include Ban Chiang and Non Nok Tha (Higham 1977; Higham and Kijngam 1979).

Assemblages chosen for this analysis span the late Pleistocene through the Holocene, approximately 45,000 years ago to the present, in Thailand and Peninsular Malaysia (Table 1; Figure 1). With the exception of the Guar Kepah (GK) and Sebarang Perak (SP) shellmounds in Peninsular Malaysia (Bellwood 2007), all sites are either caves or rockshelters. This provides the ability to understand stratigraphic and chronological sequences of faunal exploitation and subsistence activities over long temporal spans (Anderson 1989; Anderson 1997; Anderson 2005; Straus 1990; Straus 1979), but also may indicate that site type bias is occurring in this region (Higham 2013).

Aside from similar chronologies and site types, these sites also share characteristics in their lithic techno-complex assemblages. Cave and rockshelter sites in Thailand and Peninsular Malaysia dating to the late Pleistocene and Holocene often contain deposits described as Hoabinhian (Bronson and White 1992; Dunn 1970; Forestier et al. 2015; Gorman 1972; Marwick 2007; Matthews 1966; Reynolds 1990; Shoocongdej 2000; Tan 1997; White and Gorman 2004), a term variously used to document a lithic technology, temporal period, subsistence economy and/or ethnicity (Marwick 2007). Recent research on the lithic assemblages from Tham Lod (TL) and Ban Rai (BR) in northwest Thailand show that paleoenvironmental processes drove considerable variability in the procurement strategies and technological adaptations of so-called Hoabinhian foragers (Marwick 2013). Some archaeologists thus avoid the term Hoabinhian to describe general archaeological periods, ethnicities or economies and instead explicitly use the term for conveying information on a distinct artifact typology (i.e., Sumatraliths) with formal characteristics (Shoocongdej 2000). All 28 sites reviewed for this work date to the late Pleistocene or Holocene and have some Hoabinhian type material within their deposits.

I also chose the 28 sites in this review based on availability of published or gray literature. Excluded sites include Buang Baeb and Pak Om Rockshelters, which both have large faunal assemblages, but are recorded in Thai language reports only (Reynolds 1990; Shoocongdej 1996a; Pers. comm. Cholawit Thongcharoenchaikit 2014). In addition, data from the Peninsular Malaysian sites, Gua Harimau (GH), Gua Ngaum (GN) and Gua Teluk Kelawar (GTK) are extracted from Bulbeck (2003). The 28 sites included here represent some of the best evidence for prehistoric hunter-gatherer occupations in mainland Southeast Asia (see previous syntheses by Adi 2007; Shoocongdej 1996a), and are historically used in other reviews of archaeological material from this region (Marwick 2007; Reynolds 1990; Shoocongdej 1996a; Shoocongdej 2000).

2.1 Potential Issues with the Dataset

There is a considerable literature on the benefits and potential bias derived from using NISP versus MNI quantification in vertebrate faunal studies (see Lyman 2008 for overview). Since this is a meta-analysis, I am limited by counts of taxonomic abundance provided in the published or gray literature. Here, I focus on summarizing NISP counts to limit any additional biases created by aggregating MNI values from sites. Separate quantification issues occur when counting the taxonomic abundance of invertebrate faunas (Claassen 1998). I do not quantify invertebrate faunas from these 28 sites, due to dissimilar recording techniques, and thus my investigation of these issues is limited.

Additionally, there is inconsistency on screen size use (or reporting sieving methods) in the archaeozoological literature from this region. While some excavations used small-sized screens (e.g., Gorman 1971b), others lack published screen size information. It is possible that several animal types, including fish, birds, small mammals and small reptiles, are absent from this faunal dataset due to screen size biases (see Nagaoka 1994; Nagaoka 2005; Quitmyer 2004 for examples).

Non-human accumulation of faunal material represents another potential bias in this dataset. I am unaware of any robust taphonomic analyses in mainland Southeast Asian archaeozoological contexts (see Forestier et al. 2015 for a recent application), but some discussion of these issues occurs in island Southeast Asia literature (Stimpson 2009; Stimpson 2012). Rodents, raptors, mammalian carnivores and some reptilian carnivores could potentially accumulate faunal remains in cave and rockshelter sites, independent of human activity (Blumenschine 1988; Faith 2013; Lyman 1994; Schmitt and Lupo 1995), but analyses focused on this topic in Southeast Asia are lacking. Furthermore, some fauna from these 28 sites likely accumulated through natural death events, specifically bat species (Higham 1972; Rabett et al. 2011).

There is a tendency in Southeast Asian archaeozoology to assume that all fauna in cave and rockshelter sites accumulated through human activity. Without re-investigation of faunal assemblages focused on taphonomic processes, it is difficult to determine which taxa from these sites truly reflect human subsistence strategies. I expect that the signal of human subsistence activity should be clearly represented in this analysis based on large abundances of specific prey types that were consumed throughout the late Pleistocene and Holocene in this region. As such, I do not discard any faunal data in this research, even though it was potentially accumulated through non-anthropogenic means.

Seasonal bias in human occupation of cave and rockshelter sites is an additional potential limitation in this analysis. Interestingly, this question remains relatively unanswered during this temporal period. Based upon the association of wet and dry season resources (freshwater gastropods and bivalves) in the same stratigraphic contexts at Spirit Cave (SC) in northwest Thailand, Gorman (1971a; 1971b; 1972) argued that this represented year-round site occupation by hunter-gatherer groups. Yet, more recent research at Lang Kamnan Cave (LK) in central Thailand suggests that only a wet season occupation is represented in the assemblage based upon the presence of botanical and land snail remains (Shoocongdej 1996b; Shoocongdej 2000). At Lang Rongrien Rockshelter (LR) in southern Thailand, faunal data suggests that the site was intermittently occupied during the late Pleistocene and Holocene, in both the wet and dry seasons (Anderson 1990; Anderson 2005; Mudar and Anderson 2007). Clearly, additional research is required before identifying site-specific seasonal occupations in mainland Southeast Asia.

Typically, only the qualitative association of wet and dry season resources in cave and rockshelter sites is used to determine seasonal occupation events. These resources are useful as paleoecological proxies for seasonality, but depositional resolution of cultural activity remains low. Archaeozoologists in mainland Southeast Asia must develop new techniques to identify what is likely low intensity, short occupation, seasonal events in sites. Micromorphology of sediments to examine micro-stratigraphic layers and geochemical analyses of fauna may be useful techniques in this regard.

Finally, some prehistoric sites in mainland Southeast Asia may represent specialist activities. Two notable possibilities are Tham Phaa Can (Steep Cliff Cave) and Nong Nor. Tham Phaa Can does not have an identified and published faunal assemblage, only a descriptive account remains (Higham 1989). Still, this account suggests that large artiodactyls and turtles or tortoises dominate the assemblage. Similarly, Nong Nor is dominated by approximately 5.5 million common orient clams (Meretrix lusoria), indicating that hunter-gatherer groups with a marine adaptation intensively exploited resources at this site (Higham and Thosarat 1998). There is little evidence for specialist subsistence strategies at other sites in Thailand and Peninsular Malaysia. Since the faunal assemblage is not quantified from Tham Phaa Can, any potential bias is limited from this site and it is included here. Nong Nor is excluded due to the considerable specialist activity represented in the assemblage.

Although some bias is present in Thailand and Peninsular Malaysian faunas, specifically non-human accumulation of remains and seasonal occupation events, the nuances of these processes remain unknown. Thus, I include all sites and faunal data (except for Nong Nor) to investigate shifts in methodological techniques, faunal composition, and changing subsistence strategies during the late Pleistocene and Holocene.

3. Methods

First, I aggregate all faunal publications by decade to explore how archaeozoological research emerged during the twentieth and twenty-first centuries. Decadal divisions of quantitative units employed by Southeast Asian archaeozoologists provide insight into the shifting methodological techniques used in this same period. Thai-Malay faunal research has its own discrete temporal and methodological qualities that are expressed by these analyses.

Next, I separate all identified Thai-Malay fauna into total NISP counts per taxonomic classification. By examining Thai sites and Peninsular Malay sites alone, differences and similarities between geographic regions, dominant faunas and total abundances appear. All numerical and taxonomic data employed in this research is available as supplemental information in the University of New Mexico digital electronic repository, including source code for analyses conducted in R (3.1.1) and RStudio (0.98.1028) (Conrad 2015; http://hdl.handle.net/1928/25699).

3.1 Nestedness

I use nestedness analysis to explore change in composition of these faunas through time. Nestedness uses a presence-absence matrix to establish whether faunas within a set are subsets of each other (Atmar and Patterson 1993; Jones 2004; Jones 2015; Lyman 2008; Patterson 1990; Staniczenko, Kopp and Allesina 2013). As an ecological metric, nestedness is important because it allows inference into how random or non-random a faunal assemblage is, and is thus useful in studies of culturally accumulated remains (Kougioumoutzis, Simaiakis and Tiniakou 2014; Simaiakis and Strona 2014). Originally created to deal with species level extinctions, nestedness provides a quantitative means to test if the diversity of species is broadening or narrowing through time, or across space, when coupled with analysis of changing species richness (NTAXA) (Atmar and Patterson 1993; Patterson 1990). By quantifying the degree to which species enter and leave a population (e.g., island biotas), nestedness provides a value to measure the degree to which local extinction and colonization occurs.

A benefit of nestedness analyses in archaeozoological research is that quantitative units in the original literature can easily be transformed into presence-absence matrices, even to the level of basic description. While methods of quantifying nestedness have changed over time (Atmar and Patterson 1993; Patterson 1990), the basic principles remain. A perfectly nested assemblage occurs when successive populations have a perfect subset number of species in relation to the population prior (Atmar and Patterson 1993; Jones 2004; Jones 2013; Jones 2015). Using the open-source software NeD (Strona et al. 2014), I calculate NODF (nested overlap and decreasing fill; Almeida-Neto and Ulrich 2011; Guimarães Jr. and Guimarães 2006) and T (nestedness temperature; Atmar and Patterson 1993; Guimarães Jr. and Guimarães 2006). When a perfectly nested order is calculated the NODF = 100 while T = 0. A perfectly random and non-nested assemblage exhibits a NODF = 0 and T = 100 (Almeida-Neto, Guimarães Jr. and Lewinsohn 2007; Almeida-Neto et al. 2008; Atmar and Patterson 1993; Jones 2015; Ulrich, Almeida-Neto and Gotelli 2009). NeD quantifies z-scores and gives a significance value based upon 500 simulated null matrices using the ‘CE’ (proportional row and column totals) algorithm (Bascompte et al. 2003; Jones 2015). SC, Banyan Valley (BVC), TL, Khao Talu (KT), Ment (MC), LK, Sai Yok (SY), LR, Khao Toh Chong (KTC), MKI, and Sakai (SK) are quantified with nestedness metrics.

3.2 Squared Chord Distance

Here, I use squared chord distance analysis to understand similarity in faunas through time. Squared chord distance analyses (otherwise known as dissimilarity coefficients) originated in the ecological literature describing the similarity between biological assemblages (Gavin et al. 2003; Overpeck, Webb III and Prentice 1985). Known as signal-to-noise matrices, squared chord distance uses the weighted percentage of variables (fauna) as a “signal” of similarity (Gavin et al. 2003; Overpeck, Webb III and Prentice 1985). By quantifying faunal classes as a weighted percentage, an emphasis on the relative proportion of each taxonomic class or species relative to all others (but not to their absolute values) gives squared chord distance its uniqueness as a broadly applicable measurement (Ludwig and Reynolds 1988). This metric produces a value between 0–2, with 0 representing a perfectly similar assemblage based on taxonomic composition between two successive pairs of assemblages or stratigraphic contexts (Faith and O’Connell 2011; Ludwig and Reynolds 1988). Regardless of how fauna is reported, given that the same reporting technique is used for the complete assemblage (or compared contexts), squared chord distance is applicable.

Archaeozoological application of squared chord distance is limited, but Faith and colleagues successfully express its quantitative value for faunal assemblages (2013; Faith and O’Connell 2011). By using squared chord distance analyses, rapid shifts in paleoenvironmental conditions negatively impact small mammals by increasing turnover and community fragmentation in southwestern Australia during the Pleistocene, prior to human mediated impacts (Faith and O’Connell 2011). Taphonomic and paleoecological investigations of Middle and Later Stone Age deposits in the Cape Floristic Region of South Africa show similar results (Faith 2013). Here, shifts in taxonomic composition are not correlated with taphonomic processes when quantified using chord distance (Faith 2013).

In Thai-Malay sites, fauna is recorded differentially as NISP or MNI counts, and squared chord distance provides a proxy measure to understand similarity in taxonomic composition through time between sites with contrastingly recorded quantitative units. Squared chord distance is quantified for the following sites – TL, KT, MC, LK, LR, KTC, MKI, SK – between successive stratigraphic contexts in each assemblage.

4. Results

The Thai-Malay archaeofaunal record is abundant and diverse. At least 163 distinct taxonomic classifications are represented in Thailand and Peninsular Malaysian archaeological sites with a total NISP of 29,842 (Tables 1 and 2; Appendix 1 Table A), but this number is only a minimum because of the number of publications (in Thai or Malay) that I am unable to access. As a whole, molluscs (bivalves and gastropods) appear most frequently in sites followed closely by sambar deer (C. unicolor), hard-shell and soft-shell turtles and tortoises (Testudines), Indian muntjac deer (M. muntjak), and wild boar (S. scrofa). Due to scattered recording techniques applied to molluscan assemblages, their analytical value in this study is low and they are not explored in greater detail (Table 2). Thailand and Peninsular Malaysia have a comparable division of faunal publications using different recording techniques (Table 3), but their use through time has shifted. Until the 1990s and later, MNI, presence/absence, and descriptive publication styles were popular, but during this decade and after NISP became the dominant technique employed (Figure 2).

Site Recorded Faunal Data

Spirit Cave Descriptive
Banyan Valley Cave Present
Steep Cliff Cave Present
Tham Lod Rockshelter Mass (grams)
Ban Rai Rockshelter Raw Count
Sai-Yok Rockshelter Present
Khao Talu Cave Descriptive
Ment Cave Descriptive
Lang Kamnan Cave Mass (grams)
Long Rongrien Rockshelter NISP
Khao Toh Chong Rockshelter NISP+MNI
Moh Khiew Cave I Mass (grams)
Moh Khiew Cave II Present
Sakai Cave Mass (grams)
Thung Nong Nien Rockshelter Present
La Sawang Rockshelter NISP
Gua Kepah Descriptive
Gua Gunung Runtuh NISP
Gua Harimau Unknown
Gua Ngaum Unknown
Gua Teluk Kelwar Unknown
Gua Peraling Unknown
Gua Chawas Unknown
Gua Cha Unknown
Seberang Perak Descriptive
Gua Kechil NISP
Gua Tenggek NISP
Gua Sagu NISP

Table 2

Methodological recording of molluscan fauna from Thai-Malay sites. Unknown indicates that mollusc data could not be identified in the site literature. Present indicates that a passing note of molluscan presence is provided for the site, but no absolute counts. Raw count is molluscan data that is aggregated into broad categories, for example shellfish or mollusc, and recorded in the literature.

Class Order Site Specific Taxonomic Classifications Common English Name ΣOccurrence ΣNISP

Mammalia Proboscidea Elephas sp. Elephants 1
Elephas maximas Asian Elephant 1 9
Dermoptera Cynocephalus variegatusI Sunda Flying Lemur 2 1
Primates Primate Primates 4 38
Nycticebus coucang Bengal Slow Loris 2
Cercopithecidae Old World Monkeys 3 17
Macaca sp. Macaque 8 58
Macaca assamensis Assam Macaque 1
Macaca fascicularis Crab-eating Macaque 2 28
Macaca nemestrina Southern Pig-tailed Macaque 3 1
Colobinae Colobinae Monkeys 1 3
Presbytis sp. Surilis 5 6
Presbytis obscuraII See T. obscurus below 1
Presbytis cristataIII Silvery Lutung 1
Presbytis melalophos Sumatran Surili 1
Presbytis sp./Macaca sp. Surils/Macaques 2 2175
Trachypithecus sp. Lutungs 5
Trachypithecus obscurus Ducky Leaf-monkey 1 2
Hylobatidae Gibbons 1 23
Hylobates sp. Gibbons 2
Hylobates sp. Old World Monkeys/Gibbons 3
Hylobates lar Lar Gibbon 1
Rodentia Rodentia Rodents 10 458
Sciuridae Squirrels 7 60
Ratufa sp. Giant Squirrels 1 25
Ratufa bicolor Black Giant Squirrel 2 6
Hylopetes sp. Flying Squirrels 1 1
Hylopetes phayrei Indochinese Flying Squirrel 2
Hylopetes platyurus Jentink’s Flying Squirrel 1 1
Petaurista sp. Nocturnal Flying Squirrels 1 2
Petaurista petaurista Red Giant Flying Squirrel 4 2
Callosciurus sp. Squirrels 3
Callosciurus prevostii Prevost’s Squirrel 1
Tamiops mcclellandii Himalayan Striped Squirrel 1
Rhizomyidae Bamboo Rats 4 83
Cannomys badius Lesser Bamboo Rat 2 2
Rhizomys sp. Bamboo Rats 3 3
Rhizomys pruinosus Hoary Bamboo Rat 2 1
Rhizomys sumatrensis Large Bamboo Rat 3 11
Porcupine/Bamboo Rat Porcupine/Bamboo Rat 1
Muridae Rats and Mice 6 81
Hapalomys longicaudatus Marmoset Rat 1 1
Leopoldamys sp. Giant Rat 1
Rattus sp. Rats 8 120
Rattus andamanensis Sikkim Rat 1 1
Hystricidae Old World Porcupines 7 100
Atherurus macrourus Asiatic Brush-tailed Porcupine 3 4
Hystrix sp. Porcupine 7 3
Hystrix brachyura Malayan Porcupine 2 2
Lagomorpha Lepus peguensis Burmese Hare 1 1
Erinaceomorpha Echinosorex gymnurus Moonrat 1
Chiroptera Chiroptera Bats 4 1
Microchiroptera Microbats 1
Pteropodidae Megabats 2 1
Pteropus sp. Flying Foxes 1
Rhinolophus sp. Horseshoe Bats 1
Hipposideros sp. Roundleaf Bats 3 10
Hipposideros lylei Shield-faced Roundleaf Bat 1
Megaderma sp. Vampire Bats 2
Miniopterus sp. Long-winged Bats 1
Myotis sp. Mouse-eared Bats 1
Pholidota Manis javanica Sunda Pangolin 2 32
Carnivora Carnivore Carnivores 8 373
Felidae Cats 2 9
Felis bengalensisIV Leopard Cat 1
Felis viverrinaV Fishing Cat 1
Neofelis nebulosa Clouded Leopard 2 1
Panthera sp. Leopard/Tiger 1 1
Panthera pardus/Neofelis nebulosa Leopard/Clouded Leopard 1 3
Panthera pardus Leopard 2 1
Panthera tigris Tiger 4 2
Viverridae Viverrids 4 16
Arctictis binturong Binturong 2 2
Arctogalidia trivirgata Small-toothed Palm Civet 1
Paradoxurus sp. Civets 1 7
Paradoxurus hermaphroditus Asian Palm Civet 2
Hemigalus derbyanus Banded Palm Civet 1
Viverra zibetha Large Indian Civet 1 1
Canidae Canids 1 1
Canis sp. Dogs 3 4
Cuon alpinus Dhole 2 1
Ursidae Bears 3 94
Helarctos malayanus Sun Bear 5 5
Ursus thibetanus Asian Black Bear 3
Mustelidae Otters/Badgers/Martens 2 2
Mustela sp. Weasels 1 1
Aonyx cinerea Asian Small-clawed Otter 2
Lutra sp. Otters 2
Arctonyx collaris Hog Badger 5 79
Martes flavigula Yellow-throated Marten 2
Ursidae/Felidae/Cuon Bears/Cats/Dhole 1
Perissodactyla Tapirus indicus Malayan Tapir 3 4
Rhinocerotidae Rhinoceros 3 18
Dicerorhinus sumatrensis Sumatran Rhinoceros 3
Rhinoceros sp. Rhinoceros 3 2
Rhinoceros sondaicus Sunda Rhinoceros 2
Artiodactyla Ungulate Ungulates 3 73
Artiodactyls Even-toed Ungulates 4 49
Sus sp. Pigs 9 227
Sus scrofa Wild Boar 11 868
Sus barbatus Bearded Pig 2 14
Tragulidae Mouse-Deer 3 14
Tragulus sp. Mouse-Deer 3
Tragulus javanicus Java Mouse-Deer 3 9
Tragulus napu Greater Mouse-Deer 3 2
Tragulus sumatrensisVI* 1
Cervidae Deer 8 445
Axis porcinus Hog Deer 6 2
Muntiacus sp. Barking Deer 1 1
Muntiacus muntjak Indian Muntjac 15 185
Cervus eldiVII Eld’s Deer 1 3
Cervus unicolorVIII Sambar Deer 17 2387
Axis/Cervus eldi Hog/Eld’s Deer 2 27
Axis/Muntiacus sp. Hog/Barking Deer 1 2
Cervus eldi/unicolor Eld’s/Sambar Deer 1 1
Bovidae Cattle/Buffalo/Goats 6 411
Bos sp. Cattle 10 19
Bos gaurusIX Gaur 3 1
Bos javanicus Banteng 2
Bubalus bubalis Water Buffalo 1
Bos/Bubalus Cattle/Water Buffalo 1 4
Capridae Goat/Sheep 1
Capricornis sumatraensis Sumatran Serow 2
Naemorhedus sp. Gorals 3 233
Capricornis sumatraenis/Naemorhedus goral Serow/Goral 3
Insectivora InsectivoresX Insectivores 2 3
Malacostraca Decapoda Brachyura Crabs 1 1
Potamidae Freshwater Crabs 10 261
Bivalvia/Gastropoda Bivalvia/Gastropoda Bivalvia/Gastropoda Bivalves and Gastropods 20
Actinopterygii Pisces Fish 9 577
Cyprinidae Carps and Minnows 1
Barbus sp. Barbels 1
Pangasius sutchiXI Shark Catfish 1 5
Nigripceps sp.XII Catfish? 1 1
Chondrichthyes Rajiformes Rajidae Skates 1 4
Reptilia Reptilia Reptilia Reptiles 3 29
Crocodilia Crocodylinae Crocodiles 2
Squamata Squamata Scaled Reptiles 3 24
Lacertilia Lizards 6 12
Agamidae Lizards/Dragon Lizards 2 6
Serpentes Snakes 7 341
Pythonidae Pythons 4 520
Python sp. Pythons 1 5
Python reticulatus Asiatic Reticulated Python 1 3
Varanidae Carnivorous/Frugivorous Lizards 1 1
Varanus sp. Monitor Lizards 8 3975
Varanus salvator Water Monitor 2 14
Testudines TestudinesXIII Turtles 16 14503
Cuora cf. ambionensis Southeast Asian Box Turtle 1
Cyclemys cf. dentata Asian Leaf Turtle 2 7
Cyclemys/Heosemys Freshwater Turtles 1 1
Siebenrockiella crassicollis Black Marsh Turtle 1 1
Geoemydidae Pond/Wood Turtles 2 148
Emydidae Pond/Marsh Turtles 2
Trionychidae Softshell Turtles 1
Dogania subplana Malayan Softshell Turtle 2 6
Amphibia Anura Anura Frogs 7 142
Hylidae Tree Frogs 1 2
Aves Aves Aves Birds 8 148
Galliformes Gallus sp. Red Junglefowl 4 118
Anseriformes Anatidae Duck 1 1
Galliformes/Anseriformes Fowl Fowl, sm. 1 2
Passeriformes Passerine, sm. Perching Birds 1 4

Table A

Taxonomic summary of Thailand and Peninsular Malaysian faunal data.

Method Thailand Peninsular Malaysia

NISP 6 5
MNI 4 3
Presence/Absence 3 1
Description 3 3

Table 3

Total count of Thailand and Peninsular Malaysia sites with differential faunal recording methodologies.

Figure 2 

Decadal division of faunal publications using differential methodological recording techniques during the 1930s–2010s.

Archaeofauna identified in this meta-analysis occupy a mosaic of environments, including savannah grasslands, open forests, dense jungles, riparian, lakes, marsh/estuaries, and marine habitats. A full range of size classifications are also represented in this collection, from large sized elephants to small sized mice, rats, and bats. Clearly, select species (and habitats) were exploited more intensely than others throughout this region. Large abundances of turtles, tortoises, deer and pigs indicate that both aquatic and terrestrial species were consumed.

When comparing NISP counts between Thailand faunal classifications, several trends appear (Figure 3). Testudines clearly dominate the fauna with over 14,000 identified specimens, with the next closest classes being monitor lizards (Varanidae) and deer (Cervidae), but this regional trend is primarily driven by the dominance of Testudines from Moh Khiew Cave II (MKII). When MKII is removed deer have the highest NISP values, with Testudines only behind by a small factor (Figure 4). Furthermore, when Testudines are removed primates, deer and monitor lizards are more evenly distributed and all share relatively high NISP values (Figure 5).

Figure 3 

Grouped classifications from Thailand sites expressing total NISP counts per site (left column), and total NISP counts for all sites (right column).

Figure 4 

Grouped classifications from Thailand sites expressing total NISP counts per site (left column), and total NISP counts for all sites (right column) without Moh Khiew Cave II.

Figure 5 

Grouped classifications from Thailand sites expressing total NISP counts per site (left column), and total NISP counts for all sites (right column) without Testudines.

Peninsular Malaysian sites show a slightly different trend. Here, NISP values are substantially lower than Thai sites and thus the spread between sites and classes is considerably less. Instead of deer displaying high NISP values, wild boar, and Testudines tie for the top NISP counts, both with 212 (Figure 6). A surprising transition in the Peninsular Malaysian data is the lower abundance of deer compared to other classes. Deer appear in less abundance than primates, birds and fish in this region. The significance of this change is discussed below.

Figure 6 

Grouped classifications from Peninsular Malaysian sites expressing total NISP counts per site (left column), and total NISP counts for all sites (right column).

Nestedness values suggest that during the late Pleistocene and Holocene, hunter-gatherer subsistence strategies were variable, but clearly influenced by archaeozoological investigator bias (Table 4). Only sites with NISP recording have significantly nested faunas. At SC a nested fauna appears without being identified as NISP values, but this is potentially due to the high degree of documentation provided by both Gorman (1971b) and Higham (1977). MKI is also nested, but less strongly, only the T metric is significant. These data suggest that differing levels of methodological recording and analysis are driving the nested or non-nested state of faunal assemblages.

Site Nestedness NTAXA

NODF z-score p-value T z-score p-value Holocene PHT Late Pleistocene Pleistocene

SC 70.142 2.051 <0.05* 15.836 –2.239 <0.05* 23 8 5
TL 66.775 1.795 <0.05* 1.422 –1.715 <0.05* 18 22
LK 59.635 1.986 <0.05* 18.379 –2.363 <0.01* 11 8 13
LR 58.883 1.387 >0.05 21.837 –2.502 <0.01* 23 19 13
KTC 51.277 2.954 <0.01* 21.07 –4.215 <0.001* 21 16 28
MKI 61.56 1.35 >0.05 20.253 –2.847 <0.01* 32 34 30
BVC 36.307 –1.591 >0.05 38.823 0.657 >0.05
KT 62.137 1.116 >0.05 17.385 –1.593 >0.05
MC 63.089 1.202 >0.05 16.657 –1.484 >0.05
SY 34.286 –0.54 >0.05 25.582 –1.202 >0.05
SK 57.429 0.128 >0.05 26.667 –0.571 >0.05

Table 4

Nestedness results for Thailand faunal assemblages, and NTAXA data for significantly nested faunas. NODF = 100 in a perfectly nested assemblage/NODF = 0 in a non-nested assemblage. T = 0 is a perfectly nested assemblage/T = 100 in a non-nested assemblage.

* = Significant at the 95% level or greater.

It is intriguing that the spread of nestedness values is much higher in northern Thailand than in peninsular Thailand. This may be a result of several factors, but comparable nestedness for LR, KTC, and MKI indicate that faunal accumulation occurred similarly at these sites throughout the late Pleistocene and Holocene. These patterns are potentially related to the limits of hunter-gatherer mobility on the Peninsula, prey type diversity or habitat fragmentation throughout the Pleistocene-Holocene transition. In general, when sites are quantified by NISP values, nested assemblages occur through time. This may be indicative of a large-scale pattern involving narrowing diets over the course of the Holocene, or site-level bias based on archaeozoological quantification techniques (i.e., identification of a limited number of skeletal elements is skewing the true species diversity).

Squared chord distance results provide a slightly more complex story of faunal similarity and dissimilarity through time (Figure 7). While several sites have a remarkably similar taxonomic composition in successive contexts (TL, KTC, MKI, SK), others show increased dissimilarity over time (KT, MC, LK, LR). Interestingly, the sites with similar taxonomic composition through the late Pleistocene and Holocene are located in peninsular Thailand (except for TL in both region and time). Lang Rongrien has a dramatic change in faunal composition, but this is potentially driven by the rigorous reanalysis of only the Pleistocene faunas at the site (Mudar and Anderson 2007).

Figure 7 

Squared chord distance values for Thailand and Peninsular Malaysian sites. S = Surface context.

5. Discussion

In general, the changes in recording methodology and quantitative units employed by mainland Southeast Asian archaeozoologists working in Thailand and Peninsular Malaysia throughout the twentieth and twenty-first centuries have followed similar patterns globally during this time (Lyman 2008; Lyman available from the author). The popularity of MNI as an archaeozoological recording technique reached its prime during the 1960s–1980s because faunal analysts felt that it provided a more accurate measure of taxonomic abundance, skeletal part frequencies, and meat weight estimation (Lyman available from the author). Based on the recording methodology employed by Thai-Malay archaeozoologists, MNI reached its peak during the 1980s-1990s before dropping off in relationship to NISP. This upward trend in analytical recording of NISP in Thai-Malay assemblages allows for continued, and future, statistical analyses of Southeast Asia faunal assemblages and a more rigorous understanding of the true nature of the archaeozoological record (diversity/richness) in this region (cf. Lyman 2008: 172–213). Furthermore, based upon the dissimilar recording of molluscan taxonomic abundance, Southeast Asian archaeozoologists should attempt to standardize quantification techniques. Documentation of NISP, MNI and weight values for all invertebrates will enhance the ability to quantitatively examine shifts in invertebrate composition and diversity through time (Claassen 1998; Giovas 2009; Mason, Peterson and Tiffany 1998).

Occurrence and NISP counts for Thai-Peninsular Malay Pleistocene-Holocene sites indicate that while large sized prey types are common, and were consumed by hunter-gatherers, small sized prey types, which are often slow moving or immobile, tend to occur most frequently in faunal assemblages. Mainland Southeast Asia supports an extremely diverse and rich array of mollusc, turtle and tortoise species (Dudgeon 2000; Köhler et al. 2012; Thirakhupt and van Dijk 1994; Turtle Taxonomy Working Group 2014). While previous work begins to examine the role that these animals played in prehistoric forager subsistence strategies in this region (Conrad in press; Mudar and Anderson 2007; Pritchard, Rabett and Piper 2009; Rabett 2012; Van Vlack 2014), this overview establishes their importance in the Pleistocene and Holocene diet of Thai-Malay foragers. These animals not only provide a wealth of information on prehistoric subsistence strategies, but they also have the potential to be used as proxy evidence for paleoecological and paleoenvironmental shifts.

Additionally, Testudines and Mollusca commonly occupy a niche habitat, stagnant and slow-moving freshwater environments, where wet-rice cultivation and horticulture is possible (Van Vlack 2014). By better identifying the prehistoric exploitation of these animals, it furthers our insights into exactly how, and when, the replacement of wild resources for domestic emerged in mainland Southeast Asia. Currently, archaeobotanical and genetic evidence suggests that rice and millet agriculture entered Southeast Asia through human migration from China during the mid-Holocene (Castillo 2011; Fuller 2011; Fuller et al. 2011; Gross & Zhao 2014; Higham 1996; Londo et al. 2006). Yet, understanding how subsistence adaptations shifted for indigenous hunter-gatherer groups in mainland Southeast Asia during the emergence of agriculture remains weak. Future analysis should attempt to standardize and gain robust taxonomic identifications for Testudines and Mollusca specimens to better understand shifts in their species richness, diversity and composition through time (e.g., Pritchard, Rabett and Piper 2009).

Originally, this research focused on using a meta-analysis of faunal data to understand historical methodologies and Pleistocene-Holocene subsistence strategies from mainland Southeast Asia. By exploring the true nature of these data it appears that broad-scale patterns may not necessarily characterize this region; instead, localized trends emerge. What results do suggest is that prehistoric hunter-gatherer groups consumed a large abundance of turtles, deer, monitor lizards, primates and wild boar, but to varying degrees of intensity. In Thailand, turtle exploitation is dominant, but MKII clearly saturates the record. Overall, Thailand assemblages appear to be driven by the extraordinary level of high precision identification, recording and quantification of the faunas from MKII and TNN (Auetrakulvit 2004a). Without these NISP counts, the faunal record is skewed towards deer as the dominant species exploited in the past. This suggests that with future archaeozoological research, the character of the faunal data from Thailand may change once additional assemblages are recorded, and previously excavated assemblages are reanalyzed with modern techniques.

The breadth of taxonomic classifications identified by archaeozoologists working in Thailand and Peninsular Malaysia also suggests that both human and non-human agents are accumulating fauna in cave and rockshelter sites. While naturally high levels of endemic faunal diversity suggests that hunter-gatherer groups had access to a wide range of potential prey types (Charoenwongsa 1987), this does not necessarily mean that hunter-gatherers always exploited a broad diversity of resources. This meta-analysis indicates that species-specific exploitation is common throughout the late Pleistocene and Holocene.

One striking aspect of this dataset is the difference in abundance between deer and wild boar in Thailand and Peninsular Malaysia. Although an unexpected result, this information is useful both for understanding paleoecology, and as a prediction of the types of prey that may be identified in Peninsular Malaysian versus Thailand sites. Further research is required to understand the exact processes of this difference, but using modern ecological knowledge as a predictive framework, the dominance of wild boar in Peninsular Malaysia may be a reflection of more suitable habitat availability for this species and thus higher population densities. Wild boar tends to enjoy dense tropical forest vegetation in contrast to larger sized cervids that occupy transitional and marginal zones at the edges of forest-grassland habitats (Francis 2008; Lekagul and McNeely 1977). If these predictions are an accurate reflection of these species’ paleoecology, then tracing the relationship of deer to wild boar through the Pleistocene and Holocene in mainland Southeast Asia could be used as a proxy for shifting environmental conditions and habitat type exploitation. This is the type of predictability and hypothesis construction that meta-analyses, like this one, bring to the archaeozoological record in Southeast Asia.

Nestedness and squared chord distance results suggest that there may be differences occurring in the accumulation and exploitation of species between northern Thailand and the Thai-Malay peninsula. Similarity in both faunal composition and NODF values in southern Thailand indicates that hunter-gatherer groups shared a common foraging strategy during the late Pleistocene and Holocene. Why these similarities exist are currently unknown, but the substantial increase in sea levels and loss of landmass at the end of the Pleistocene on the Thai-Malay Peninsula are likely to be significant drivers of these results (Anderson 1990; Dunn and Dunn 1977; Horton et al. 2005; Sathiamurthy and Voris 2006; Scheffers et al. 2012; Sinsakul 1992; Tjia 1996; Voris 2000). Lowered hunter-gatherer mobility or lower human population densities during this period may have allowed groups to exploit a similar and consistent prey base over time.

While the results of nestedness and squared chord distance analyses describe coarse trends in the Thai-Malay faunal record, the more direct utility of these methods is their application for assemblages recorded with diverse techniques (NISP/MNI/Presence-Absence). In Southeast Asia, a region with a rich history of archaeological investigation spanning the past 150 years or more, quantitative techniques like nestedness and squared chord distance provide the ability to compare sites with diverse and distinct assemblages that utilize older quantitative methodologies. Nestedness also suggests that methods of identification are potentially driving quantifiable patterns in these assemblages. Due to the nature of fragmentation and poor preservation in Southeast Asian faunal assemblages, archaeozoologists typically only identify teeth because of their robustness and preservation (Wattanapitaksakul 2006). Thus, faunal data likely lacks the complete suite of diversity present, providing a potential driver of these nestedness results.

Overall, the Thailand and Peninsular Malaysian archaeozoological record is robust and unique in its diversity and abundance. Continued faunal research in this region will only clarify and expand the results, questions, and themes presented in this analysis. It is also my hope that this overview and synthesis of faunal data provides a useful guide for future research and hypothesis testing in mainland Southeast Asia.

6. Conclusion

Meta-analysis of Thailand and Peninsular Malaysian faunal assemblages establishes the current state, and patterns present, in the archaeozoological records from 28 sites in this region. Small sized slowly moving fauna (mollucs, turtles and tortoises) dominate the occurrence and NISP counts, but further research is needed to establish firm taxonomic identifications and standardization of recording methodology. These results are important for future hypothesis based research investigations. For example, why do wild boar appear in greater abundance in Peninsular Malaysia than Thailand? Are similar foraging economies occurring in southern Thailand during the Pleistocene-Holocene transition, driving the similarity in nestedness and squared chord distance results? Finally, how does the exploitation and consumption of turtles and molluscs relate to the onset and transition to agriculture and domestication? These types of questions are extremely significant to the archaeological record in this region, and archaeozoological data increases our ability to answer them. Finally, while this meta-analysis focused on Thailand and Peninsular Malaysia, the records of Myanmar, southern China, Laos, Cambodia, Vietnam, Island Malaysia and beyond, have the potential to provide substantial insights into broader trends through greater Southeast Asia. The exploitation of turtles and molluscs clearly extends into the deep past and continued research should focus on understanding their exploitation, consumption, and ecological histories in Southeast Asia.

Competing Interests

The author declares that they have no competing interests.