Another kind of calibration seems to be necessary for radiocarbon dates taken on human bones, if the humans ate a lot of fish. It has long been recognized that in salt-water seas, organic substances like shell or fish bones absorb old carbon that is in solution in the water, which makes radiocarbon dates on shell and fish come out too old. This is called the “reservoir effect” because seas act as a reservoir of old carbon. Recent studies have indicated that the same problem can affect organisms that lived in fresh water, and most important among these were fish. Fish absorb old carbon in solution in fresh water, and people who eat a lot of fish will digest that old carbon and use it to build their bones. Radiocarbon dates on their bones will come out too old. Dates measured on charcoal or the bones of horses and sheep are not affected, because wood and grazing animals do not absorb carbon directly from water like fish do, and they do not eat fish. Dates on human bone can come out centuries older than dates measured on animal bone or charcoal taken from the same grave (this is how the problem was recognized) if the human ate a lot of fish. The size of the error depends on how much fish the human ate and how much old carbon was in solution in the groundwater where he or she went fishing. Old carbon content in groundwater seems to vary from region to region, although the amount of regional variation is not at all well understood at this time. The amount of fish in the diet can be estimated on the basis of 15N levels in bone. Fish have much higher percentages of 15N in their tissues than does any other animal, so humans with high 15N in their bones probably ate a lot of fish. High 15N in human bones is a signal that radiocarbon dates from those bones probably will yield ages that are too old.
Figure A1. A proposed linear correlation between the % of 15N in dared human bone (bottom) and the number of radiocarbon years that should be sub-tracked from radiocarbon dares (top) before they are calibrated.
Research to correct for reservoir effects in the steppes is just beginning as I write this, so I cannot solve the problem. But many of the radiocarbon dates from steppe archaeology are from cemeteries, and the dated material often is human bone. Widespread tests of the 15N in human bone from many different steppe cemeteries, from Kazakhstan to Ukraine, indicate that fish was a very important part of most ancient steppe diets, often accounting for 50% of the meat consumed. Because I did not want to introduce dates that were probably wrong, I used an approach discussed by Bonsall, Cook, and others, and described by them as preliminary and speculative. They studied five graves in the lower Danube valley where human bone and animal bone in the same grave yielded different ages (see chapter 7 for references). Data from these graves suggested a correction method. The average level of 15N in the human skeletons (15.1%) was equated with an average radiocarbon error (425 ± 55) that should be subtracted prior to calibrating those dates. These averages could be placed on a scale between the known minimum and maximum levels of 15N found in human bone, and, speculatively, a given level of 15N could be equated with an average error in radiocarbon years. The scale shown in figure A.1 was constructed in this way. It seems to yield results that solve some long-problematic dating offsets in steppe chronology (see ch. 9, notes 4, 16, and 22; and ch. 12, note 30). When I use it—when dates are based principally on human bone—I warn readers in the text. Whatever errors it introduces probably are smaller than those caused by ignoring the problem. All the radiocarbon dates listed in the tables in this book are regular BP and calibrated BCE dates, without any correction for the reservoir effect.
Figure A.1 shows the correction scale I used to revise dates that were measured from human bone in regions where I knew the average 15N levels in human bone. The top number is the number of years that should be subtracted from the BP radiocarbon date; the bottom number is the 15N level associated with specific subtraction numbers.
TABLE A.1
The average 13C and 15N% in human bone from seventy-two individuals excavated from graves in the Samara oblast, by time period.
Table A.1, based on our own studies in the Samara oblast, shows the average 15N content in human bone for different periods, taken from measurements on seventy-two individuals.
NOTES
CHAPTER 1. THE PROMISE AND POLITICS OF THE MOTHER TONGUE
1. Bloch 1998:109.
2. See Sapir 1912:228.
3. Cannon 1995:28–29.
4. Poliakov 1974:188–214.
5. Veit 1989:38.
6. Grant 1916.
7. For “external origin” passages in the Rig Veda, see Witzel 1995. For “indigenous origin” arguments, see N. Kazanas’s discussions in the Journal of Indo-European Studies 30, nos. 3–4 (2002); and 31, nos. 1–2 (2003).
8. For the Nazi pursuit of Aryan archaeology, see Arnold 1990.
9. For goddesses and Indo-Europeans, see Anthony 1995b; Eisler 1987, 1990; and Gimbutas 1989, 1991. For Aryan-identity politics in Russia, see Shnirelman 1998, 1999.
10. Heidegger 1959:37–51, contrasted to Boaz 1911. For the non-Aryan element in the Rig Veda, see Kuiper 1948, 1991.
11. Harding and Sokal 1988.
12. The American Heritage Dictionary has thirteen hundred unique Proto-Indo-European roots listed in its appendix. But multiple reconstructed words are derived from the same root morphemes. The number of reconstructed words with distinct meanings is much greater than the number of unique roots.
13. For doubts about proto-languages and tree diagrams, see Lincoln 1991; and Hall 1997. For a more nuanced view of tree diagrams, see Stewart 1976. For “creolization” and convergence creating Proto-Indo-European, see Renfrew 1987:78–86; Robb 1991; and Sherratt and Sherratt 1988.
14. For framing, see Lakoff 1987:328–37.
CHAPTER 2. HOW TO RECONSTRUCT A DEAD LANGUAGE
1. Here is the text of the tale:
A sheep, shorn of its wool, saw some horses, one moving a heavy cart, another carrying a big load, a third carrying a human speedily. The sheep said to the horses: “It pains me [literally, “the heart narrows itself for me”] to see human driving horses.” The horses said: “Listen sheep, it pains us to see that human, the master, makes the wool of the sheep into a warm garment for himself and the sheep no longer has any wool!” On hearing that the sheep ran off into the fields.
It is impossible to construct whole sentences like this with confidence in a language known only in fragments. Proto-Indo-European tense markers in the verbs are debated, the form of the relative pronoun is uncertain, and the exact construction of a Proto-Indo-European complement (sheep saw horse carrying load) is unknown. Linguists still see it as a classic challenge. See Bynon 1977:73–74; and Mallory 1989:16–17.
2. This chapter is generally based on four basic textbooks (Bynon 1977; Beekes 1995; Hock and Joseph 1996; and Fortson 2004), and on various encyclopedia entries in Mallory and Adams 1997.
3. Embleton 1991.
4. Pinker 1994.
5. An example of a change in phonology, or pronunciation, that caused shifts in morphology, or grammar, can be seen in English. German has a complex system of noun and pronoun case endings to identify subjects, objects, and other agents, and verb endings that English lacks. English has lost these features because a particular dialect of Middle English, Old Northumbrian, lost them, and people who spoke the Old Northumbrian dialect, probably rich wool merchants, had a powerful effect on the speech of Medieval London, which happened to give us Modern English. The speakers of Old Northumbrian dropped the Germanic word-final n and m in most suffixes (esse’, not essen, for “to eat”). In late Old English the pronunciation of many short vowels (like the final -e that resulted here) was already merging into one vowel (the [uh] in sofa, called schwa by linguists). These two shifts in pronunciation meant that many nouns no longer had distinctive endings, and neither the infinitive nor the subjunctive plural verb had a distinct ending. Later, between 1250 and 1300, the word-final schwa began to be dropped from most English speech, which wiped out the distinction between two more grammatical categories. Word order became fixed, as few other guides indicated the differe
nce between subject and object, and auxiliary particles like to, of or by were employed to distinguish infinitives and other forms. Three shifts in pronunciation were responsible for much of the grammatical simplification of modern English. See Thomason and Kaufman 1988:265–275.
6. For Grimm’s Law, see Fortson 2004:300–304.
7. Some linguists argue that the Proto-Indo-European root did not begin with k but rather with a palato-velar, a kh-type sound, which would require that the first consonant was moved back in the centum languages rather than forward in the satem languages. See 1994:251–252. Thanks to Bill Darden for pointing this out.
8. Hock and Joseph 1996:38.
9. For pessimistic views on the “reality” of reconstructed Proto-Indo-European, see Bynon 1977; and Zimmer 1990. For optimistic views, see Hock and Joseph 1996:532–534; and Fortson 2004:12–14.
10. Hall 1950, 1976.
11. Bynon 1977:72. Mycenaean was in a transitional state in 1350 BC, when it was recorded. Some Proto-Indo-European words with kw had already shifted to k in Mycenaean. The alternation between *kw and *p probably was already present in some dialects of Proto-Indo-European.
12. For doubts on reconstructed meanings, see Renfrew 1987:80, 82, 260. For the argument that comparing cognates requires that the meanings of the compared terms are subjected to fairly strict limits, see Nichols 1997b.
CHAPTER 3. LANGUAGE AND TIME 1
1. See Swadesh 1952, 1955; and Lees 1953.
2. The replacement rate cited here compares the core vocabulary in Modern English to the core vocabulary in Old English, or Anglo-Saxon. Much of the Old English core vocabulary was replaced by Norse, but, since Norse was another Germanic language, most of the core vocabulary remains Germanic. That is why we can say that 96% of the core vocabulary remains Germanic, and at the same time say that the replacement rate in the core vocabulary was a high 26%.
3. Much of the information in this section came from Embleton 1991, 1986. See also McMahon and McMahon 2003; and Dyen, Kruskal, and Black 1992. Many linguists are hostile to any claim that a cross-cultural core vocabulary can be identified. The Australian aboriginal languages, for example, do not seem to have a core vocabulary—all vocabulary items are equally vulnerable to replacement. We do not understand why. Both sides of the debate are represented in Renfrew, McMahon, and Trask 2000.
4. Meid 1975; Winfred 1989; and Gamkrelidze and Ivanov 1984:267–319.
5. Ivanov derived Hittite (Northern Anatolian) and Luwian (Southern Anatolian) separately and directly from Proto-Indo-European, without an intervening proto-language, making them as different as Celtic and Greek. Most other linguists derive all the Anatolian languages from a common source, Proto-Anatolian; see Melchert 2001 and Diakonoff 1985. Lydian, spoken on the western coast of Anatolia in the Classical era, might have descended from the same dialect group as Hittite. Lycian, spoken on the southwestern coast, probably descended from the same dialects as Luwian. Both became extinct in the Classical era. For all these topics, see Drews 2001.
6. For the Anatolian languages, see Fortson 2004:154–179; Houwink Ten Cate 1995; Veenhof 1995; and Puhvel 1991, 1994. For the glottalic perspective, see Gamkrelidze and Ivanov 1995.
7. Wiluša was a city west of the Hittite realm. It is very possible that Wilusa was Troy and that the Trojans spoke Luwian. See Watkins 1995:145–150; and Latacz 2004.
8. The non-Indo-European substrate effect on Luwian was described by Jaan Puhvel (1994:261–262) as “agglutinative creolization … What has happened to Anatolian here is reminiscent of what became of French in places like Haiti.” Hittite showed similar non-Indo-European substrate effects and had few speakers, causing Zimmer (1990:325) to note that, “on the whole, the Indo-Europeanization of Anatolia failed.”
9. Melchert 2001.
10. Forster 2004; Baldi 1983:156–159.
11. Lehrman 2001. The ten innovations that Lehrman identified as distinctive of Proto-Indo-European included two phonological traits (e.g., loss of the laryngeals), three morphological traits in nouns (e.g., addition of the feminine gender), and five morphological traits in verbs.
12. See Sturtevant 1962 for the Indo-Hittite hypothesis. For Anatolian as a daughter of very early Proto-Indo-European, see Puhvel 1991. Lehrman (2001) pointed out that Anatolian had a different word from Proto-Indo-European for man, usually considered part of the core vocabulary. The Anatolian term (*pāsna-) used a root that also meant “penis,” and the Proto-Indo-European term (*wtro-) used a root that also meant “strength.” Proto-Anatolian and Proto-Indo-European did, however, share cognate terms for grandfather and daughter, so their kinship vocabularies overlapped. Classic Proto-Indo-European and Anatolian probably emerged from different places and different times in the Pre-Proto-Indo-European dialect chain.
13. For Pre-Greek language(s) of Greece, see Hainsworth 1972; and Francis 1992.
14. For the oldest language in the Indic branch I use the term Old Indic instead of Indo-Aryan. The standard nomenclature today is Indo-Iranian for the parent, Avestan Iranian for the oldest Iranian daughter, and Indo-Aryan for the oldest Indic daughter. But the designation Aryan for Indic is unnecessary; they were all Aryan. For the language and history of the Rig-Veda, see Erdosy 1995.
15. For Old Indic terms among the Mitanni, see Thieme 1960; Burrow 1973; and Wilhelm 1995. I thank Michael Witzel for his comments on Mitanni names. Any errors are my own.
16. For a date for Zarathustra before 1000 BCE, see Boyce 1975; and Skjærvø 1995. For the “traditional” date promulgated by ancient Greek sources, five hundred years later, see Malandra 1983.
17. Clackson (1994) and Hamp (1998) argued that Pre-Armenian was linked to the Greek-Indo-Iranian block. See also the isogloss map in Antilla 1972, figure 15.2. Many of the shared lexical items are discussed and described in Mallory and Adams 1997. I am grateful to Richard Diebold for his analysis of Greek/Indo-Iranian relations in a long letter of October 1994, where he pointed out that the shared innovations link Greek and Iranian closely, and Greek and Indic somewhat less.
18. See Rijksbaron 1988 and Drinka 1995 for the shared poetic functions of the imperfect. Poetics, shared phrases, and weapon terms are reviewed in Watkins 1995, chap. 2, 435–436.
19. See Ringe et al. 1998; and also Ringe, Warnow, and Taylor 2002. Similar cladistic methods were applied to a purely lexical data set in Rexová, Frynta, and Zrzavý 2003.
CHAPTER 4. LANGUAGE AND TIME 2
1. See Darden 2001, esp. 201–204, for the etymology of the term wool. For the actual textiles, see Barber 2001, 1991; and Good 1998.
2. The “unspinnable” quotation is from Barber 2001:2. The mitochondrial DNA in modern domesticated sheep indicates that all are descended from two ancient episodes of domestication. One cluster (B), including all European and Near Eastern sheep, is descended from the wild Ovis orientalis of eastern Anatolia or western Iran. The other cluster (A) is descended from another Ovis orientalis population, probably in north-central Iran. Other wild Old World ovicaprids, Ovis ammon and Ovis vignei, did not contribute to the genes of domesticated sheep. See Hiendleder et al. 2002. For a general discussion of sheep domestication, see Davis 1987; and Harris 1996.
3. In the Ianna temple of Uruk Iv (3400–3100 BCE) artists depicted women making textiles. The later Sumerian names for some months incorporated the term for plucking sheep. The zoological evidence suggests that the months were named this way during the Late Uruk period or afterward, not before.
4. Zoological evidence for wool production in the Near East is reviewed by Pollack (1999:140–147). For Arslantepe, see Bökönyi 1983. An earlier date for wool sheep could be indicated by a couple of isolated pieces of evidence. The phase A occupation at Hacinebi on the Euphrates, dated 4100–3800 BCE, had spindle-whorls that seemed the right weight for spinning wool, which requires a light spindle; see Keith 1998. A clay sheep figurine from Tepe Sarab in western Iran (Kermanshah) seems to show a wooly fleece, from a level dated about 5000 BCE. For a broader discussion, see Good 200
1.
5. For the caprids (sheep and/or goats) at Khvalynsk, see Petrenko 1984. Petrenko did not report the age at death for all the caprids in the Khvalynsk graves, but six of the twelve with reported ages were adults. Sacrificial deposit #11 contained 139 bones of caprids representing four adults and five sub-adults, and the average adult withers height was 78 cm, almost 15 cm taller than other European Neolithic caprids. For Svobodnoe sheep, see Nekhaev 1992:81. For sheep in Hungary, see Bökönyi 1979:101–116. For sheep in Poland, see Milisauskas 2002:202.
6. For wool at Novosvobodnaya, see Shishlina, Orfinskaya, and Golikov 2003. For evidence of Catacomb-period wool (dated ca. 2800–2200 BCE) in the North Caucasian steppes, see Shishlina 1999. Sherratt’s updated comments on wool are included in the revised text of an older article in Sherratt 1997a.
7. The term for hub or nave, which is often included in other lists, also meant “navel” in Proto-Indo-European, so its exact meaning is unclear. For the wheel-wagon vocabulary, see Specht 1944. Three influential updates were Gamkrelidze and Ivanov 1984:718–738; Meid 1994; and Häusler 1994. I first published on the topic in Anthony and Wailes 1988; and also in Anthony 1991a, 1995a. As with most of the topics covered in this book, there is an excellent review of the Indo-European wheel vocabulary in Mallory and Adams 1997.
The Horse, the Wheel, and Language: How Bronze-Age Riders From the Eurasian Steppes Shaped the Modern World Page 50