Our reference: AMPER 1 P-authorquery-v11 AUTHOR QUERY FORM Journal: Ampersand Article Number: 1 Please e-mail or fax your responses and any corrections to: E-mail: [email protected]Fax: +44 1392 285879 Dear Author, Please check your proof carefully and mark all corrections at the appropriate place in the proof (e.g., by using on-screen annotation in the PDF file) or compile them in a separate list. Note: if you opt to annotate the file with software other than Adobe Reader then please also highlight the appropriate place in the PDF file. To ensure fast publication of your paper please return your corrections within 48 hours. For correction or revision of any artwork, please consult http://www.elsevier.com/artworkinstructions . Location in article Query / Remark click on the Q link to go Please insert your reply or correction at the corresponding line in the proof Q1 Please confirm that given names and surnames have been identified correctly. Q2 As per the journal style, only a minimum of 1 and a maximum of 6 keywords are allowed. Please provide/retain any 1-6 keywords. Q3 Please check and confirm the terms ‘Wanderwörter’ and ‘Wanderwort’ given throughout the file. Q4 In the place of Haspelmath (2009) we have cited both Haspelmath (2009a) and Haspelmath (2009b). Please check, and correct if necessary. Q5 Mallory and Adams (1997), Sapir (1916), Appleyard (1999), Oswalt (1964), Sawyer (1991), Golla (2007), Akerman and McConvell (2002), Green (2009), Haspelmath and Tadmor (2009), Mixco (1978), Kroeber (1961) and McConvell and Laughren (2004) is/are cited in the text but not provided in the reference list. Please provide it/them in the reference list or delete these citations from the text. Q6 The citation Gamkrelidze and Ivanov (1984), Dolgopol0sky (1998) has been changed to match the author name/date in the reference list. Please check here and in subsequent occurrences, and correct if necessary. Q7 The citation Berndt (1964) has been changed to match the author name/date in the reference list. Please check here and in subsequent occurrences, and correct if necessary. Q8 Figs. 8–16 is/are not cited in the text. Please check that the citation(s) suggested by the copyeditor is/are in the appropriate place, and correct if necessary. Q9 An extra closing parenthesis is deleted. Please check, and correct if necessary. Q10 An extra opening parenthesis (2 cases) has been deleted. Please check, and correct if necessary. Q11 Shall we change ‘bighorn ‘sheep’‘’ and ‘”‘antelope’“’ to “bighorn ‘sheep”’ and “antelope” respectively in this paragraph? Q12 Uncited references: This section comprises references that occur in the reference list but not in the body of the text. Please cite each reference in the text or, alternatively, delete it. Any reference not dealt with will be retained in this section. Q13 Please update the status of publication for this reference. Q14 Please confirm the year for the following reference(s): Brown et al. (forthcoming) and Epps (forthcoming). Q15 Please update the status of publication for this reference. Please check this box or indicate your approval if you have no corrections to make to the PDF file Thank you for your assistance. Page 1 of ...1...
19
Embed
Wanderwoerter in languages of the Americas and Australia
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Our reference: AMPER 1 P-authorquery-v11
AUTHOR QUERY FORM
Journal:Ampersand
Article Number: 1
Please e-mail or fax your responses and any corrections to:
Please check your proof carefully and mark all corrections at the appropriate place in the proof (e.g., by using on-screen annotationin the PDF file) or compile them in a separate list. Note: if you opt to annotate the file with software other than Adobe Reader thenplease also highlight the appropriate place in the PDF file. To ensure fast publication of your paper please return your correctionswithin 48 hours.
For correction or revision of any artwork, please consult http://www.elsevier.com/artworkinstructions.
Locationin article
Query / Remark click on the Q link to goPlease insert your reply or correction at the corresponding line in the proof
Q1 Please confirm that given names and surnames have been identified correctly.
Q2 As per the journal style, only a minimum of 1 and a maximum of 6 keywords are allowed. Please provide/retainany 1-6 keywords.
Q3 Please check and confirm the terms ‘Wanderwörter’ and ‘Wanderwort’ given throughout the file.
Q4 In the place of Haspelmath (2009) we have cited both Haspelmath (2009a) and Haspelmath (2009b). Please check,and correct if necessary.
Q5 Mallory and Adams (1997), Sapir (1916), Appleyard (1999), Oswalt (1964), Sawyer (1991), Golla (2007),Akerman and McConvell (2002), Green (2009), Haspelmath and Tadmor (2009), Mixco (1978), Kroeber (1961)and McConvell and Laughren (2004) is/are cited in the text but not provided in the reference list. Please provideit/them in the reference list or delete these citations from the text.
Q6 The citation Gamkrelidze and Ivanov (1984), Dolgopol0sky (1998) has been changed to match the authorname/date in the reference list. Please check here and in subsequent occurrences, and correct if necessary.
Q7 The citation Berndt (1964) has been changed to match the author name/date in the reference list. Please checkhere and in subsequent occurrences, and correct if necessary.
Q8 Figs. 8–16 is/are not cited in the text. Please check that the citation(s) suggested by the copyeditor is/are in theappropriate place, and correct if necessary.
Q9 An extra closing parenthesis is deleted. Please check, and correct if necessary.
Q10 An extra opening parenthesis (2 cases) has been deleted. Please check, and correct if necessary.
Q11 Shall we change ‘bighorn ‘sheep’‘’ and ‘”‘antelope’“’ to “bighorn ‘sheep”’ and “antelope” respectively in thisparagraph?
Q12 Uncited references: This section comprises references that occur in the reference list but not in the body of thetext. Please cite each reference in the text or, alternatively, delete it. Any reference not dealt with will be retainedin this section.
Q13 Please update the status of publication for this reference.
Q14 Please confirm the year for the following reference(s): Brown et al. (forthcoming) and Epps (forthcoming).
Q15 Please update the status of publication for this reference.
Please check this box or indicateyour approval if you have nocorrections to make to the PDF file
Wanderwörter in languages of the Americas and Australia
Q1∧Hannah
∧Haynie a,
∧Claire
∧Bowern a,∗,
∧Patience
∧Epps b,
∧Jane
∧Hill c,
∧Patrick
∧McConvell d
a Yale University, United Statesb University of Texas, Austin, United Statesc University of Arizona, United Statesd Australian National University, Australia
h i g h l i g h t s
• We examine Wanderwörter in Australia and the Americas.• Wanderwörter exhibit higher levels of borrowing than most loanwords.• These items spread in both chain-like and starburst borrowing networks.• Wanderwörter are often acculturation terms, ritual objects, and technologies.• Diffusion of cultural or technological innovations plays an important role.
a r t i c l e i n f o
Article history:Received 29 May 2014Accepted 23 October 2014Available online xxxx
Wanderwörter are a problematic set of words in historical linguistics. They usually make up a smallproportion of the total vocabulary of individual languages, and only a minority of loanwords.
They are, however, found frequently in languages from across the world.There is, to our knowledge, no general synthesis of Wanderwörter patterns, causes of exceptionally
high borrowing rates for particular lexical items, or estimates of their frequency across language families.Claims about the causes of their spread exist, but have not been widely tested. Nor, despite researchers’intuitions that Wanderwörter form a distinct type of borrowing, is there a clear demonstration thatWanderwörter are, in fact, different from other loanwords in any concrete way.
In the present paper, we examine the phenomenon of Wanderwörter using a standard sample ofvocabulary in languages of Australia, North America and South America. The investigation presentedhere examines Wanderwörter in great enough detail to answer questions about the linguistic and socialprocesses by whichWanderwörter migrate as well as the shapes and densities of the resulting networks.We show that Wanderwörter can be categorically distinguished from other borrowing.
The study of Wanderwörter to date has focused on agricultural or industrialized societies; however,the phenomenon is well attested in networks of smaller languages. There are areal differences in types ofWanderwörter and the networks through which they spread. Specific categories of cultural association,including but not limited to agricultural cultivation, condition widespread borrowing.
Wanderwörter are outliers in the realm of loanwords, borrowed far more frequently than typicallexical items but still a subset of a more general phenomenon. We show that the link betweenWanderwörter and cultural diffusionmay be amore sound basis for defining this term than the traditionaldefinitions that invoke the loan frequency, areality, or untraceability of these terms.
small group of neighboring languages. They usually make up a1
small proportion of the total vocabulary of individual languages,2
and only a minority of loanwords. They are, however, found fre-3
quently in languages from across the world. While there is pre-4
vious research on individual Wanderwörter and linguistic areas,5
there is, to our knowledge, no general synthesis of Wanderwörter6
patterns, causes of exceptionally high borrowing rates for particu-7
lar lexical items, or estimates of their frequency across language8
families. Definitions of Wanderwörter vary, and while there are9
claims about the causes of their spread (such as links to trade),10
these claims have not beenwidely tested. Nor, despite researchers’11
intuitions that Wanderwörter form a distinct type of borrowing, is12
there a clear demonstration that Wanderwörter are, in fact, differ-13
ent from other loanwords in any concrete way.14
In the present paper, we examine the phenomenon of Wander-15
wörter using a standard sample of vocabulary in languages of Aus-16
tralia, North America and South America, coded for etymological17
status. This controlled sample provides an unprecedented oppor-18
tunity to examine the phenomenon ofWanderwörter in away that19
permits us tomove beyond isolated anecdotes, to clarifywhy some20
lexical items become involved in such complexes, while others do21
not, and to define a threshold for identifying the number of loan22
events which qualifies a word as being a Wanderwort. The inves-23
tigation presented here examines Wanderwörter in great enough24
detail to answer questions about the linguistic and social processes25
by which Wanderwörter migrate as well as the shapes and densi-26
ties of the resulting networks. The following issues are addressed:27
a. Can Wanderwörter be categorically distinguished from other28
borrowing?29
b. Are there areal differences in the types of Wanderwörter we30
find evidence for and the networks throughwhich they spread?31
c. Dowordswithin certain semantic fields or pragmatic categories32
becomeWanderwörter more readily than others?33
1.2. Defining a Wanderwort34
The clearest statements about the characteristics of Wander-35
wörter as a class come from Campbell and Mixco (2007), Hock and36
Joseph (1996), and Roberge (2010). Campbell and Mixco (2007:37
220) define a Wanderwort as follows (numbers and emphasis38
ours):39
1: a borrowed word diffused across numerous languages,40
2: usually with awide geographical distribution;41
3: typically it is impossible to determine the original donor lan-42
guage from which the loanword in other languages originated.43
Clearly the notions of ‘numerous languages’ in 1 and a ‘wide geo-44
graphical distribution’ are vague, as is the qualification ‘usually’ in45
2. Criterion 3 is even more problematic. Making the definition of a46
Wanderwort dependent on the impossibility of finding the original47
source is unworkable. One can say that the original source has not48
been found, or is not certain, but inmany cases ifmore data are con-49
sidered or more interpretations explored, then plausible hypothe-50
ses about the original source can be proposed. Further, if an original51
source is evident and the Wanderwort set meets the other crite-52
ria, it is odd to exclude it from the category. For example, words53
like ‘coffee’ and ‘tea’ would be excluded from category of Wander-54
wörter under criterion 3, even though they fit criteria 1 and 2.55
In another of the few treatments of Wanderwörter in the basic56
texts, Hock and Joseph (1996: 254) translate the termas ‘‘migrating57
words’’ and define these as words that ‘‘spread over vast territories58
through a chain of borrowings’’. That is, to the basic definition out-59
lined by Campbell andMixco (2007) they add the notion of a ‘chain’60
of borrowing, implying thatWanderwörter (typically) move in one
direction from source to target with this move being repeated as 61
the original target becomes a source for further borrowing. This is 62
one way in which one might want to distinguish Wanderwörter 63
borrowing from ordinary borrowing, which typically occurs be- 64
tween a single pair of languages (Haspelmath and Tadmor, 2009; 65
Bowern et al., 2011). This requirement that Wanderwörter exhibit 66
chain borrowing is in contrast with Campbell and Mixco’s require- 67
ment that the loans be ‘widespread’, which is not specific about the 68
trajectory of borrowing. To our knowledge a systematic study of 69
borrowing network shapes has not been undertaken, and it is thus 70
unclear whether trajectory of borrowing is a significant criterion 71
for distinguishing loan type (or is simply epiphenomenal). 72
Roberge’s (2010: 411) definition (following Polomé, 1992 and 73
others) includes ‘‘a special category of loanwords that spread 74
across languages, usually in connection with trade or the adoption 75
of external technological, economic, or cultural practices’’. Roberge 76
thus focuses on Wanderwörter as a special category by virtue 77
of the means by which they spread, and the circumstances of 78
their adoption, rather than the frequency with which they are 79
loaned or other areal associations. Roberge’s criteria also imply 80
that Wanderwörter in indigenous languages might be particularly 81
associated with the colonial period, given the widespread changes 82
in cultural practices and economies for those groups. 83
Conversely, Haspelmath (2009a, 2009b: 45) refers to Wander- 84
wörter aswordswhich ‘‘appear in languages from a number of lan- Q4 85
guage families in a particular area’’ (also known as ‘areal roots’), 86
and points out that if these are old roots, it may be difficult to es- 87
tablish that they are loanwords. This approachmay appear to adopt 88
the perspective of the ‘area’ or Sprachbund rather than the trajecto- 89
ries of individual words. Another interpretation would be that the 90
‘areas’ in question may not necessarily be already∧recognized as 91
exhibiting heavy lexical (or other) diffusion (cf. Nelson-Sathi et al., 92
2011). 93
In summary, definitions of Wanderwörter refer to four distinct 94
properties of borrowing. First is the number of loan events, and 95
whether the number of individual borrowing events between 96
languages is systematically greater for Wanderwörter than for 97
other types of words. Secondly is the structure of the loan 98
network. That is, do Wanderwörter show particular patterns in 99
how they are loaned across language communities that make 100
them distinct? Alternatively, such networks could simply reflect 101
the major conduits by which other cultural innovations diffuse 102
across communities. Thirdly is the type of word that is particularly 103
susceptible to becoming a Wanderwort, and whether (and for 104
what reason) certain items or semantic fields are particularly 105
associated with the first two criteria (see also Section 1.4 106
below). Finally, though we cannot explore this topic here, one 107
might ask whether there are linguistic properties which define 108
Wanderwörter as distinct from other loan categories, either by 109
a tendency to phonological readaptation (or non-adaptation) or 110
the speed of their adoption. The first set of criteria refer to 111
structural properties of the borrowing patterns, and can be used 112
to distinguish Wanderwörter from other types of loans, while the 113
second involves the composition of the class of lexical items that 114
might be defined as Wanderwörter. We discuss the first three 115
properties of potential Wanderwörter in this article; borrowing 116
amounts in Section 2, the shape of the network structures (and 117
potential correlations with other conduits for technology, such as 118
trade networks) in Section 3, and semantic fields in Section 4. 119
Due to data limitations we are unable to address the question 120
of whether putative Wanderwörter have particular phonological 121
features. In Section 5, we discuss the problem of distinguishing 122
Wanderwörter from lexical resemblances that may reflect ancient 123
genetic unity or substratal phenomena. Section 6 provides a brief 124
summary of the paper. 125
H. Haynie et al. / Ampersand xx (xxxx) xxx–xxx 3
We show that there is, indeed a category of widespread loans1
which can be distinguished from other loans by virtue of the2
number of times they have been borrowed. We find some limited3
support for the ‘chain’ idea, though we note that this is somewhat4
epiphenomenal. We also find limited support for the idea that5
Wanderwörter are associated with particular semantic fields, with6
the exception of acculturation terms, which consistently stand out7
in our survey.8
1.3. Hypotheses about Wanderwörter spread9
The diffusion of innovations has been a major focus in com-10
munication theory and sociology since the work of Rogers (1962);11
cultural diffusion has likewise been important as a focus in some12
schools of anthropology (e.g. Kroeber, 1940), with various theories13
proposed to explain the rate of uptake of different kinds of innova-14
tions in different groups. Investigation of the spread of associated15
terminology has not generally been linked to these fields of study.16
Linguists have mainly approached the issue from the angle of ety-17
mology, with the exception of ‘acculturation’, in which a powerful18
group has invaded, controlled or exerted major influence on an-19
other groupwith a distinct culture (cf. Brown, 1999). Studies of the20
linguistic effects of such profound cultural influence often include21
reference to Wanderwörter (for instance Brown, 1999) as well as22
alternative strategies for naming novel items, such as the coining of23
neologisms or the semantic extension of indigenous terms to new24
meanings (Basso, 1967; Campbell and Grondona, 2012).25
Previous literature has identified both potential triggers for26
Wanderwörter diffusion and semantic fields that seem to be par-27
ticularly associatedwith them.One of themost common categories28
of claimed Wanderwort consists of words referring to material29
culture items. Diffusion can occur through physical movement of30
these items through trade, learning of techniques formaking items,31
or copying of styles by people in a sequence of societies. Often these32
processes occur in combination, providing a motivation for both33
the trigger of spread and the associated linguistic semantic field.34
Domesticated plants and animals are also among themost com-35
monly cited examples of Wanderwörter in Eurasia and Africa.36
Some of them are thought to have diffused between proto-37
languages at dates coinciding with the earliest agriculture (thus38
linking flora/fauna to technical innovation). This may be the39
case with proto-Indo-European *bhars ‘barley’, (related to proto-Q540
Semitic *burr-/barr- ‘threshed grain’ (Gamkrelidze and Ivanov,Q641
1995: 836; Mallory and Adams, 1997: 51-2)). However, as with42
other etymologies at such a time-depth, this is debated. Other cul-43
tivars arrived in Europe later, after first cultivation in Asia around44
3500 years ago; this trend is exemplified by the term ‘rice’ and45
many related forms in theMiddle East deriving from Proto-Central46
South Dravidian *vari (itself perhaps borrowed from another lan-47
guage family, cf. Pejros, 1997: 97). Similar etymological patterns48
can also be associated with a diffusion of a new variety and/or49
function, as with the Wanderwort %kannabis1 ‘hemp’ in Europe.50
The word was first associated with a narcotic variety which spread51
around 3000 years ago from the East, but later named varieties52
1 We mark Wanderwort etyma sets with a % sign to denote that the word isnot a reconstruction, but rather a generalization across forms that have historiesboth of adaptation through loanhood and regular sound change, and that are oftennot reconstructible to a single form because of their complex history. An examplefrom North America is %palata ‘‘acorn woodpecker’’, which represents the likelysource of attested forms like Northern Sierra Miwok palat:ata, Yowlumne palakak,and Tübatulabal ta:la’gakt. In some cases, we can be fairly sure of the shape of thesource item, as with the words for ‘datura’, where the source language is knownwith some confidence, or for ‘pelican’ or ‘wildcat’ where the items have beenreconstructed for a protolanguage that is either the donor language or the parent ofthe donor language.
used only for textiles and rope (Barber, 1991; Sapir, 1916; Mc- 53
Convell and Smith, 2003).2 54
Ritual objects and decorations are other itemswhich have a pat- 55
tern of following paths of cultural influence and trade routes and 56
are candidates for widespread loans. For example, below in Sec- 57
tion 3.4 we discuss pearlshell, which is clearly a valued trade item 58
in Australia, and which moves far outside its origin area on the 59
coast into other regions where it is not naturally found. The exis- 60
tence of extensive trade routes for such items has been studied us- 61
ing historical and∧archeological evidence (McCarthy, 1939; Berndt Q7 62
and Berndt, 1964: 128-9; Akerman and Stanton, 1994) but not lin- 63
guistic evidence to any extent. Like ritual or decorative items, sub- 64
stances with medicinal or psychotropic properties are also traded 65
and passed from the areas where they grow to other areas. In the 66
Americas, psychotropic substances are highly valued and traded, 67
and in many cases have been obtained from specific source areas 68
and subsequently grown locally. Due to the relationship between 69
religion/ritual and healing, as well as the ritual use of psychotropic 70
substances, novel cultural usemight lead both ritual items andpsy- 71
chotropic substances be associated with widespread diffusion of 72
lexical items, which would therefore make them good candidates 73
for Wanderwörter. 74
Other hypotheses about Wanderwort spread relate to the de- 75
mographics of the donor and recipient populations. Such hypothe- 76
ses (such as the greater likelihood of spread of words from larger 77
populations to smaller ones) apply to general loan events too, 78
but may be particularly pronounced in the case of Wanderwörter, 79
where the number of borrowing events makes demographic cor- 80
relations easier to recover. Such correlations between population 81
and direction of borrowing could plausibly be related to the greater 82
number of tieswhich speakers of larger languagesmayhave and/or 83
the relative political power and prestige of larger languages, which 84
can foster borrowing by raising both the appeal and the accessibil- 85
ity (e.g. via lingua franca effects) of the larger language relative to 86
others. Geographical parameters may also affect the direction of 87
diffusion, such as from coastal languages into inland languages or 88
vice versa, as these parameters may shape the pathways by which 89
physical resources move (see further Section 3.4). It may also be 90
that long-distance Wanderwörter are less likely to show these ef- 91
fects statistically, since they cross languageswithmanydistinct de- 92
mographic and geographical profiles. 93
1.4. Hypotheses about types of words that become Wanderwörter 94
One of the reasons thatWanderwörter as a category are difficult 95
to pin down is that theremaybe several reasonswhy a root appears 96
to be widespread in an area. In cases where loans may be old, it 97
is often difficult to distinguish Wanderwörter from items that are 98
widespread for other reasons, such as substratal relics of an earlier 99
language, universal sound symbolic properties, or even inheritance 100
from a common (ancient) proto-language. For instance, some of 101
thosewho support the idea of aNostraticmacrofamily regard some 102
etyma which are shared between Indo-European and these other 103
families as belonging to ‘proto-Nostratic’, such as *woyn- ‘wine’ 104
in proto-Indo-European and *wayn in proto-Semitic. However 105
even some of those who support the Nostratic hypothesis regard 106
such items as Wanderwörter, moving with the diffusion of wine 107
and wine-making, for both linguistic and non-linguistic reasons 108
(Dolgopolsky, 1998; Appleyard, 1999; cf. also Sherratt, 1995). 109
2 Recent studies of dispersal of cultivars have tended to be interdisciplinary,involving
∧archeology and paleobotany together with linguistics, making it possible
to plot the paths of dispersal of the plants along with the words used to namethem, as in studies of banana species in south-east Asia and the Pacific (Denhamand Donohue, 2009), which then arrived and diffused in Africa (Blench, 2009).
4 H. Haynie et al. / Ampersand xx (xxxx) xxx–xxx
Without the requirement of chain borrowing or network1
characteristics, adoption of substratal terms may be difficult to2
distinguish from radially dispersed Wanderwörter, particularly if3
Wanderwörter are defined by their occurrence in a number of4
languages across a broad geographic region. Prior etymological5
analysis can permit us to sort out the most likely cases of long-6
distance borrowing from cases thatmay involvewidespread sound7
symbolism or nursery formations, and from other caseswhichmay8
resembleWanderwörter distributions, such as substratal adoption9
due to language shift (McConvell, 2009, 2011).10
1.5. Languages and data11
Our sample includes 53 languages of northwestern Australia,12
55 North American languages from California and the Great13
Basin, and 27 languages of the Amazon basin in northwestern14
South America.3 All of the Australian languages and the majority15
of the American languages are associated with hunter-gatherer16
societies. In theNorthAmerican and SouthAmerican samples some17
languages of cultivators are included (this is especially true for18
South America, where most groups rely on some combination of19
these subsistence patterns). The Australian sample includes both20
∧Pama–Nyungan and non-Pama–Nyungan languages in 11 different21
genetic groups. The North American sample includes languages22
from four unrelated families, and six isolates. Eight families and23
two isolates are included in the SouthAmerican sample. Fig. 1 gives24
a map of language locations.25
This language sample differs from those in most treatments of26
Wanderwörter, which typically include languages of metropolitan27
or agricultural communities, affected by large-scale or global flows28
of ideas and innovations and associated with the diffusion of29
major world religions, empires, trade, new crops and technologies30
(Haspelmath and Tadmor, 2009). Here we focus on small-scale31
foraging (or partially foraging) communities. While many of these32
are now part of global networks of ideas and technologies to33
some extent, many of the Wanderwörter we consider here do not34
relate to modern cultural contacts, but to contact unrelated to35
colonization or absorption into global trade networks.36
For each language, we collected a standard sample of 204words37
of basic vocabulary, along with approximately 120 words (the38
number varies slightly among the three samples) for flora and39
fauna, and approximately 100 cultural vocabulary items for each40
area. The basic vocabulary sample includes forms with relatively41
culture-free references, such as body parts and words for natural42
phenomena such as ‘water’ and ‘sky’ (see Bowern et al., 2011).43
The ethnobiological sample (Bowern et al., 2014) includes words44
that are unlikely to have a special cultural load (such as ‘small45
bird’), but also words for plants and animals of heightened cultural46
significance (e.g. ‘eagle’, of ritual and mythological significance in47
many parts of the Americas and Australia). The cultural sample48
includes items of indigenous culture (e.g. ‘spearthrower’, ‘arrow’)49
aswell as items introduced since the time of European colonization50
of each of the study areas (e.g. ‘gun’, ‘horse’). For each language, all51
items are coded for etymological status and assigned to cognate52
classes. Loan-family sets, as noted in footnote 1 above, are marked53
with %.54
3 While the South American sample focused on these 27 languages, wealso considered information from 76 additional languages (most also fromthe northwest Amazon), representing thirteen additional language families andfourteen isolates. As with the 27 focal languages, data from these languageswere systematically collected for the full word list under consideration. Thiscomparison provided important information about the distribution and history ofWanderwörter in the region. Where relevant, data from these additional languagesare included in the South American examples below.
Fig. 1. Case study areas.
The composition of this three-part word list creates a sample 55
with a variety of borrowing levels, ranging from little-loaned 56
words like pronouns to highly borrowed acculturation terms. This 57
approximates the overall range of borrowing levels in the sample 58
languageswithout explicitmanipulation of the∧word list to achieve 59
a particular distribution of loan frequencies. Items were assigned 60
to etyma sets according to the tools of the ComparativeMethod. All 61
itemswhich stem in someway from the same etymon are grouped 62
together. However, for the purposes of establishingWanderwörter 63
counts, only items which are demonstrably loans are included in 64
counts. The older the putativeWanderwort, the more different the 65
forms are from each other and the harder it is to be confident 66
in distinguishing ancient loans from chance resemblances. We 67
have not included forms where the resemblances involve only CV, 68
in order to reduce the likelihood of unwittingly including chance 69
resemblances. We have also been conservative in requiring a high 70
degree of semantic resemblance, or a solid justification in the 71
ethnographic record for a semantic change where this is involved, 72
as in the case of ‘moon’ ∼ ‘datura’ in Section 4.3. 73
2. Assessing Wanderwörter as a distinct category of loanwords 74
As discussed in Section 1, conventional definitions of Wander- 75
wörter do not adequately establish a categorical difference be- 76
tween Wanderwörter and other loans. Heuristic approaches to 77
identifyingWanderwörter, such as arbitrary threshold numbers of 78
loan events or network links, are useful for detecting likely loan 79
chains but tend to be sensitive to the specific configuration of the 80
loan network. Without an explicit definition or heuristic system 81
by which to classify items as Wanderwörter, it is difficult to as- 82
sess the notion that this phenomenon represents a different type 83
of borrowing from multiple individual lexical exchanges between 84
pairs of languages. We begin to chip away at this task by exam- 85
ining the hypothesis that Wanderwörter comprise a category of 86
loans separate from typical borrowing with regard to levels of 87
borrowing. While not all highly loaned items are necessarily Wan- 88
derwörter (under some definition), quantification of loan events 89
allows us to explore loan frequency patterns and the distribution 90
of different meanings and pragmatic categories along those loan 91
frequency clines. This allows us to compare our notions of Wan- 92
derwörter with empirical facts. 93
In previous work (Bowern et al., 2011), we established that 94
languages in our sample for the most part borrow little from their 95
∧neighbors, with the vast majority exhibiting borrowing levels in 96
basic vocabulary of under ten percent. The South American area 97
in particular exhibits very low borrowing levels. A small number 98
of languages in the North American and Australian samples are 99
outliers, with levels of borrowing in basic vocabulary ranging 100
H. Haynie et al. / Ampersand xx (xxxx) xxx–xxx 5
between 20% and 50%. Borrowing levels in the∧flora–fauna sample1
are on average higher, averaging 9.8% (Bowern et al., 2014).2
Material culture loan figures, however, are considerably higher,3
with an average of over 20% across case study areas.44
To test the notion that Wanderwörter form a separate class of5
borrowing, we use statistical methods. We apply the X-means sta-6
tistical clustering algorithm (Pelleg and Moore, 2000) to this data7
to partition it into awell-fitting number of discrete categories. This8
extension of K -means clustering (MacQueen, 1967) groups data9
points into clusters, inferring the number of clusters by selecting10
a model that minimizes both the within-cluster sum of squares11
error and the complexity of the model (assessed using the cor-12
rected Akaike Information Criteria (Cavanaugh, 1997)). This proce-13
dure strikes a balance between penalizing the creation of excessive14
numbers of partitions and finding a solution with tightly clustered15
categories and very littlewithin-group variation. By identifying the16
location of natural breaks between categories in this way, we pro-17
vide some statistical evidence for the categorization of lexical data18
into different loan classes. If Wanderwörter are in fact a distinct19
loan class, we should find a cluster of high-loan items within our20
sample. Clustering solutions with large numbers of inferred par-21
titions and those whose partition boundaries do not represent a22
clear division between highly loaned and less-highly loaned items23
would not provide evidence for the unique status of a class ofWan-24
derwörter. We first partition the data on statistical grounds, and25
then look for patterns within the data which might explain why a26
given set of words cluster together. We do this by temporarily set-27
ting aside questions about semantic field and loan network struc-28
tures, and concentrate purely on the number of loan events that29
each lexical item undergoes.30
2.1. Borrowing levels in the full lexical sample31
At the most basic level, then, the status of Wanderwörter as32
distinct from other loans depends on whether some words are33
borrowed more frequently than other words. While variation in34
the ‘borrowability’ of lexical items is a well-known and empiri-35
cally substantiated concept (Haspelmath and Tadmor, 2009), the36
distribution of lexical items across the range of borrowing lev-37
els has received little systematic study. A simple bar plot (Fig. 2)38
shows the number ofmeanings in our total sample at each attested39
level of borrowing. As wemight expect, the distribution is skewed,40
with the majority of items borrowed minimally or not at all. How-41
ever, with increasing borrowing levels we find a smooth curve, and42
a long tail that stretches into very high numbers of loans.5 The43
highest number of loan events in our sample for a given mean-44
ing category6 is 45. Note that this measurement oversimplifies in45
a number of ways, for example by not distinguishing etyma bor-46
rowed several times from ‘meaning categories’ borrowed multi-47
ple times, but potentially independently. We further refine this48
distinction in Section 2.2 but here treat the measurement as one49
which showswhichmeanings are particularly associatedwith high50
numbers of borrowing events, whether dependent or independent.51
4 We note that the set of material culture terms also includes postcolonialacculturation terms, associated with loans having their origin in Europeanlanguages. However, loans in this domain are not confined to acculturation terms,and not all acculturation terms are loaned.5 Loan counts in Fig. 2 include the total number of forms for a particular meaning
that are coded as loans, loans into protolanguage, or loans in an unknown direction,summed across all three study areas.6 A ‘meaning category’ is a translation equivalent. That is, the meaning category
‘horse’ includes all words, regardless of etymology, which can be used as translationequivalents for the word ‘horse’.
Fig. 2. Loans by meaning category across case study areas.
Although this unimodal distribution does not show evidence 52
for discrete classes of loans in the form of multiple distributional 53
peaks, the curve itself suggests that only a small number of words 54
are very highly loaned, while the vast majority of words do not 55
participate in widespread borrowing. 56
X-means clustering partitions this data into three categories. 57
The first includes words borrowed up to three times (n = 1110); 58
another includes items borrowed four to 14 times (n = 299); the 59
third category includes items borrowed 15–45 times (n = 23). 60
While within-cluster variance and model AICc point to a three- 61
category solution rather than a binary division between highly 62
loaned items and ‘regular’ borrowing, the bulk of the data fall into 63
just the first two of these three classes. 64
As seen from the full list of data in the supplementarymaterials 65
(see Appendix A), the division between the low- and moderate- 66
borrowing categories does not align precisely with clear seman- 67
tic or pragmatic categories. However, several generalizations can 68
be made about these categories and the items that fall in them. 69
First, the majority of basic vocabulary falls into the low-borrowing 70
category.7 Secondly, the extremely high-borrowing category is 71
comprised almost entirely of acculturation terms which were in- 72
troducedwith colonial contact. These include items such as ‘paper’, 73
‘gun’, ‘corn’, and ‘cat’. Non-acculturation terms in this third cate- 74
gory are the ethnobotanical terms for Datura wrightii and ‘grass’, 75
both from the North American sample area. The overall patterns in 76
this data suggest that acculturation terms, perhaps not surprisingly 77
(cf. Brown, 1999), tend to be very frequently loaned, while a fur- 78
ther division between moderately loaned words and infrequently 79
loanedwordsmay exist but is less clearly associatedwith semantic 80
or pragmatic differences among meanings or specific sociohistori- 81
cal settings. This suggests that there is, indeed, a category of ‘super- 82
borrowings’, which are particularly associated with acculturation 83
processes. However, a weakness of this approach is that it does 84
not distinguish general levels of meaning borrowing from borrow- 85
ing of specific etyma. A further disadvantage is that items that are 86
identified as probable loans, but with unknown direction, are po- 87
tentially overweighted.8 We refine this method in the following∧
88
section. 89
7 The supplementary information gives further details about the types of itemsin each category (see Appendix A).8 In our dataset, 961 words (out of 55,378 total data points) are described as
probable loans on the basis of their phonology, morphology, and/or distribution.However, the source of the loan is unknown. For example, pilthi ‘bird’ is a wordshared only by two adjacent languages with otherwise little vocabulary in common(Malyangapa and Paakintyi in the Australian case study area). This distributionstrongly implies that the word is a loan from one language into the other. However,the direction of loanhood is not ascertainable from current data. If, however, loanswith this coding are treated as equivalent to loans where the direction is known,they endupoverweighted in loan counts because bothdonor and recipient languageare coded as ‘‘loan direction unknown’’.
6 H. Haynie et al. / Ampersand xx (xxxx) xxx–xxx
Fig. 3. Loan events in Australia.
2.2. Borrowing levels of individual etyma in Australian languages1
To further investigate natural patterns in loanword frequency2
and the evidence they bring to bear on the nature of Wander-3
wörter, we look more closely at loans within the Australian case4
study area and examine them by etyma set. Fig. 2 above plots5
the number of meanings associated with various levels of over-6
all borrowing across all three study areas. However, it does not7
take into account the fact that the same ‘meaning category’ could8
be associated with loan events of very different etyma. For exam-9
ple, the forms %jirigi and %bandalmada are both associated with10
the meaning ‘‘bird’’, but because they are separate etyma, the loan11
events associated with them should be counted separately. While12
Fig. 2 provides information about the tendency for certain se-13
mantic categories to be loaned with varying frequency, the loan14
counts for form-meaning pairs plotted in Fig. 3 provide a better15
summary of the etymon loan networks typically associated with16
Wanderwörter. Focusing on this more detailed information for the17
Australian study area has the further advantage of allowing the18
implementation of a weighting scheme to prevent items coded as19
‘loan direction unknown’ from being overcounted.20
The overall picture of loan events in Australia resembles the21
skewed, unimodal curve present for meanings across all three22
study areas, though the curve is slightly steeper for this case study23
subset than for the overall data. X-means clustering for this data24
subset suggests that a two-cluster model best fits the data, rather25
than the three-category solution proposed for the global dataset.26
As we found for the global data, however, the break between cate-27
gories partitions items with three or fewer borrowings from those28
with four or more borrowings. The lower optimal number of clus-29
ters in the classification of Australian loan event data probably re-30
flects the fact that the curve for this study area shows a sharper31
drop-off as borrowing levels increase, with highly loaned items32
falling into a flatter tail. In other words, the ‘moderately loaned’33
category in Fig. 2 is inferred because a relatively substantial num-34
ber of meanings are neither associated with rare borrowing nor35
with extremely high borrowing. In Fig. 3 we see that the majority36
of etyma are borrowed a single time, with a dramatic drop in the37
number of etyma that are borrowed even twice. Etyma borrowed38
more than three times are distributed relatively evenly among39
higher borrowing rates, forming a shallow tail in the distribution40
and preventing a ‘moderately loaned’ category from emerging.41
Again, we find the majority of basic vocabulary items in the42
low-borrowing category and the majority of cultural terms, and43
especially postcolonial acculturation terms, in the high-borrowing44
category. However, the pattern is less distinct near the border be-45
tween the two categories returned by X-means clustering. The46
classification of basic words such as ‘dust’ and ‘small’ in the cat-47
egory of etyma most frequently borrowed is not easily explained48
by traditional accounts of Wanderwörter presented above.9 In all49
9 In these particular cases, ‘‘dust’’ is borrowed into Gooniyandi, Gajirrabeng,Gurindji, Kija, Miriwung, and Wunambal, all from
∧Ngumpin–Yapa languages.
likelihood, the boundary between the number of loan events as- 50
sociated with true Wanderwörter and the level of borrowing that 51
‘ordinary’ items typically undergo is a fuzzy one, with individual, 52
unrelated loan events occasionally adding up to numbers that ap- 53
proximate the borrowing levels associated with Wanderwörter. 54
Quantification of borrowing can only tell part of the story regard- 55
ing Wanderwörter, since high numbers of loan events for a given 56
meaning does not of itself entail the existence of a Wanderwort. 57
The levels of borrowing exhibited by Australian data and in the 58
global dataset provide mixed evidence regarding the distinction 59
betweenWanderwörter and other loans. On the one hand, the dis- 60
tribution of loan frequencies exhibits a continuous cline in loan 61
levels rather than the sort of multimodal peaks that would pro- 62
vide powerful evidence for a categorical distinction betweenWan- 63
derwörter and ‘ordinary’ borrowing. However, clustering analysis 64
identifies a statistical division between etyma loaned three times 65
or fewer and those more highly borrowed that may represent a di- 66
vide between ‘ordinary’ loans in the head of the distribution and 67
Wanderwörter in the long tail of the distribution. This result is con- 68
sistent with the conventional wisdom that ‘ordinary’ loanwords 69
are far more common than Wanderwörter, and suggests that the 70
much greater variation in borrowing levels among themost loaned 71
items reflects the importance of individual words’ histories for de- 72
termining how far and wide a word may wander. It also indicates 73
that while high loan status may be part of the definition of a Wan- 74
derwort, it is not the only component worth examining. 75
3. Patterns in network structure and directionality of borrow- 76
ing 77
In this section we discuss the structure of Wanderwörter bor- 78
rowing chains inmore detail. Because of the difficulty inmeasuring 79
network shapes mathematically with the data available, we con- 80
centrate instead on characterizing the general patterns evident in 81
our dataset of high and ultra-high loan items. While graph theory 82
(and particularly components of social network analysis) have a 83
number of tools to measure and compare the structure of graphs 84
such as those produced by putativeWanderwort patterns, our data 85
have too many holes to make such comparisons reliable. We pro- 86
vide some discussion of this in Section 3.2. 87
Previous work on Wanderwörter has∧emphasized a chain-like 88
pattern for loans, in which a word is borrowed in sequence and 89
thereby spreads linearly across an area (see Campbell and Mixco, 90
2007). 91
(1) A → B → C → D. 92
However, the mapping of widespread loans in our dataset shows 93
that there are, in fact, several different types of loan pattern. In 94
addition to the chains pattern, a ‘radial’ or ‘star’ pattern, where a 95
single language loans a term into several (or all) of its∧neighbors, 96
is also widespread. This is illustrated in (2), where the notation 97
means that Language A was the donor language for terms into 98
languages B, C, and D. 99
(2) A → B, C, D. 100
Combinations of these patterns are also found, where etyma 101
participate both in chains and radial borrowing patterns. 102
‘‘Small’’ is due to a number of loans from Nyulnyulan into surrounding languages.It seems that these loans from members of a family/subgroup into several otherlanguages account for much of this unexpected pattern. We might consider theseto be independent borrowing events (of nonetheless cognate items) rather thantrue Wanderwörter, however there is no principled way to say when this sort ofunderspecified source can be counted as part of a Wanderwort network and whenit is associated with individual loan events.
H. Haynie et al. / Ampersand xx (xxxx) xxx–xxx 7
One pattern that we do not find robustly attested in the dataset1
is multiple independent borrowings of the same lexical item from2
discontiguous languages. That is, theoretically it would be possible3
to achieve Wanderwort status (at least in numerical terms) by4
a loan structure such as that illustrated in (3) below, where the5
same etymon has been borrowed independently between pairs6
of languages. The scarcity of such a pattern in the dataset is7
not, perhaps, surprising, but it is worth mentioning given that it8
enforces the way in which Wanderwörter might be different from9
simple repeat (but adjacent) borrowings of the same lexical item.10
It further points to the importance of trade or other factors in11
spreading such loans.1012
(3) A → B, C → D, E → F13
where A, C, and E are closely related but not contiguous14
languages.15
3.1. Long chain networks16
We find relatively few long chains in the dataset, characterized17
by three or more consecutive borrowing events in which a18
borrowing language serves as the source for a subsequent loan. In19
North America we find long chains involved in complex borrowing20
networks, such as in the terms for datura and acorn woodpecker,21
discussed below in Sections 4.4 and 4.6, respectively. The items22
associated with these etyma are traded and used in ritual in this23
study area, which may explain the number of chain-like links in24
their borrowing networks as well as the occurrence of parallel25
chains and star-like branching in those same networks.26
Long chains appear to be relatively frequent in the Australian27
case study area compared to other areas. Study of these long28
chains is hampered to some extent by the difficult in ascertaining29
the direction of borrowing, however. For example, in the case30
of %tyimpila ‘speartip’, ten languages from five different families31
along the Fitzroy River (and east into the Victoria River district)32
share an almost identical form, strongly indicating that it is a recent33
loan.34
The only long chain identified to date in the South American35
sample involves the etymon %hipa ‘coca’ (see Section 3.4.3 below),36
but the scarcity of long chains in this region may be in part due to37
problems of detection; because common source languages belong38
to widespread and discontiguous families (such as Arawak and39
Tupí-Guaraní), multiple borrowings from closely related source40
languages are not easily distinguished from chains of borrowings.41
3.2. Star and short chain networks42
Widespread loan words may reflect star-shaped (radial)43
spreads from all edges of a source language, rather than, or in ad-44
dition to, the chaining of loan moves described in Hock and Joseph45
(1996). This star-shaped pattern of borrowing is perhaps the most46
common configuration of Wanderwörter networks in the North47
American case study, where languages all around the edges of the48
source language have borrowed some item. In this case, there is no49
borrowing chain; each attestation of the form represents a trans-50
fer of the etymon from the same source to one of many recipients.51
A good example is a word for ‘dog’ in the North American case52
study region. Here theword appears to be Proto-Miwokan, and has53
10 We note, however, that this pattern probably has occurred in Amazonia,particularly involving the Arawak and
∧Tupi–Guarani language groups, which are
widely discontiguous in the Amazon basin but are known to have been influential inmaintaining large-scale trading networks (see, e.g. Rydén’s 1962 discussion of Tupí-Guaraní terms for ‘salt’ in Amazonia). However, the pattern is difficult to identifywhere so many of the relevant daughter languages are currently unattested. Theword for ‘dog’ illustrated in (4) may be another example.
been borrowed into languages that surround the Miwokan family 54
(though probably from several different Miwokan daughters, indi- 55
cating that the source of the star may be a set of closely related 56
b. %mani- ‘datura’ Gabrielino> Cupan> Serran> Chumashan, 72
Tübatülabal 73
c. %wits-∼wich-‘bird, small bird’ Southern Numic > Serran, 74
Southeastern Yokuts > Ineseño 75
d. %?ui ‘roadrunner’∧Yok–Utian > Nisenan, Western Mono, 76
Tübatülabal, Kawaiisu > Chemehuevi, 77
e. %hus ‘buzzard’ Sierra Miwokan > Washo, Wintun, Maiduan, 78
Wintun> Lake Miwok 79
g. %chipik ‘beaver’ Yokutsan > Kitanemuk > Kawaiisu; Yokut- 80
san > Chumash 81
h. %molo- ‘black bear’ Yokutsan > Kitanemuk, Tübatülabal 82
> Shoshone; Kitanemuk > Kawaiisu 83
i. %hawuts ‘fox’ Yokutsan > Western Mono, Ventureño 84
Chumash; Numic > Northern Sierra Miwok > Washo 85
j. %saka ‘black willow’ Serran > Nim-Yokuts > Salinan, Sierra 86
Miwok 87
k. %lima ‘brown bear’ Pomoan > Wappo > Wintu 88
l. %hola ‘badger’ Sierra Miwok > Maidu > Washo 89
m. %?aq ‘eat pinole’ Chumashan > Californian > Southern 90
Numic 91
It is immediately obvious that there is some overlap between the 92
meanings that appear in this list of items and the itemswith chain- 93
like borrowing patterns more typical of the classical description 94
of Wanderwörter. For example, %pahmo is a word in the tobacco 95
complex that, like the more-widespread %sokon, originates in 96
∧Yok–Utian and spreads into Uto-Aztecan languages on the one 97
hand and intoMaidu andWasho on the other. %mani- ‘Datura’, like 98
*moji ∼ *moni ‘moon, datura’, is a word that probably originates 99
in Gabrielino and the Cupan languages, and spreads both within 100
California Uto-Aztecan and into one variety of Chumashan. These 101
cases involve culturally important plants and similar languages 102
to examples associated with longer borrowing chains, and the 103
difference in chain length is probably either accidental or a 104
reflection of imperfect data. 105
3.3. Supernova patterns 106
A third type of structure is where most or all languages in 107
a region show a particular word. These are most common in 108
our acculturation vocabulary sample (e.g. ‘horse’) but also occur 109
in a few other items, such as Banisteriopsis caapi (%kapi) in 110
languages of the Upper Rio Negro region of Amazonia. Logically, 111
such distributions could have their origins in several different 112
loan structures. They could be from long chains that have folded 113
back on themselves across a region, for example, or from star-like 114
networks, where loans have radiated into surrounding languages 115
until they have covered all languages in region. Finally, they could 116
spread ‘wave-like’, where every language in a region rapidly adopts 117
a word. We lack the data to be able to distinguish these scenarios 118
with any reliability at this point. 119
8 H. Haynie et al. / Ampersand xx (xxxx) xxx–xxx
Fig. 4. Major loan patterns: Australian case study area.
3.4. Trade networks and Wanderwörter networks: social network1
analysis2
Social network analysis tools allow us to plot Wanderwörter3
networks according to the number of loans linking individual4
languages, and to compare these networks to possible correlates of5
Wanderwort spread. Here, we compare Wanderwörter networks6
to known trade networks in the Australian and North American7
case study areas to assess the likelihood that Wanderwörter travel8
straightforwardly along trade networks. This raises the question9
of whether all (or most) Wanderwörter spread through trade, or10
whether alternative sources are possible.11
We test this notion by mapping loan patterns geographically,12
using graph visualization tools typically employed for Social13
Network Analysis.11 We present data for Australia and North14
America. Owing to the high degree of difficulty in establishing pre-15
contact geographic settlement patterns in our SouthAmerican case16
study, we do not attempt to do the same for Amazonia here.17
3.4.1. Australian case study area18
Fig. 4 shows the major links map for loans in the Australian19
case study area side by side with a map of the route of pearl-20
shell (after Akerman and Stanton, 1994) in Fig. 5, which proxies21
widespread trade of other items.1222
Figs. 4 and 5, which represent the network of loan transmission23
in our Australian study area and known trade routes in the24
region, respectively, provide some perspective on the notion that25
Wanderwörter are correlated with trade or diffuse along trade26
corridors. The geography of the borrowing network in Fig. 4 bears27
only limited resemblance to the network of trade routes in Fig. 5,28
suggesting thatwhile economic trademay contribute to the spread29
of Wanderwörter, it is not the sole mechanism, or perhaps even30
the primary mechanism ofWanderwort diffusion. If regional trade31
networks were the primary conduit for Wanderwörter spread, we32
should expect a closermatch between our network generated from33
highly borrowed lexical items and that generated from traded34
items. This conclusion is modified in Section 3.4.2, however, based35
on further data from California.36
In individual cases, however, highly traded items may in fact37
predispose the linguistic forms used to refer to them to be widely38
11 Here we use Gephi open graph visualization software to map borrowingnetworks (Bastian et al., 2009).12 The path of pearlshell trade is similar to a number of the other major pathsfor traded goods in the area, though few goods participated in such a widespreadnetwork as the pearlshell one. For local trade networks that are congruent with thecontinent-wide one, see McBryde (1984a,b), for example.
Fig. 5. Pearlshell trade networks (after Akerman and Stanton, 1994).
Fig. 6. Loan networks in California.
borrowed. For example, whereas aggregate loan networks map 39
poorly onto general trade routes, the specific network along which 40
the term for ‘pearlshell’ itself has spread is a very plausible route 41
for pearlshell trade. That is, trade itself is not sufficient to establish 42
patterns by which Wanderwörter are spread, but the widespread 43
exchange of an item may lead to loans for its name. This is 44
congruent with the findings in Bowern et al. (2014), which found 45
that trade itself did not predict high loan levels for flora/fauna 46
items, but some of the highest loaned items were nonetheless 47
highly traded. Thus trade is neither necessary nor sufficient for 48
Wanderwort status, at least in this case study area. 49
3.4.2. North American case study area 50
Figs. 6 and 7 below provide trade network and loan data which 51
parallel those in the Australian case study area from Figs. 4 and 5. 52
The loan networks in the North American study area (Fig. 6) 53
more closely resemble known trade networks in the region 54
(adapted here in Figure 7 from Davis, 1961), particularly in the 55
web of contact around the Clear Lake area, and in the central- 56
ity of Yokuts to trade networks involving its neighbors. The rel- 57
atively greater isomorphism between loanword networks and 58
economic trade networks in North America might suggest a
H. Haynie et al. / Ampersand xx (xxxx) xxx–xxx 9
Fig. 7. California trade networks.
stronger association between lexical borrowing and economic1
trade in this region than in Australia. It is important to note, how-2
ever, that the structure of these two social networks in the Cal-3
ifornia area may be constrained by physical geography, creating4
a limited number of possible pathways for both economic and5
linguistic contact.13 With the Pacific Ocean and Coastal Range to6
the west, the imposing Sierra Nevada Mountains on the east, and7
much travel outside of the Central Valley∧funneled through rugged8
river valleys and canyons, the physical location of the Yokuts lan-9
guages makes their communities likely hubs of contact among10
∧neighboring linguistic groups from all sides. In this case, the paral-11
lels between trade andWanderwörtermay indicate shared linguis-12
tic and economic history, but ultimately this may reflect not a nec-13
essary link between trade and linguistic borrowing but rather the14
more general conditioning of social contact in this region. In con-15
trast, some Australian trade networksmay be shapedmore by cus-16
tomary partnerships that evolve throughout history than by phys-17
cal trade andWanderwörter networks in Australia to be geograph-19
ically more diffuse and exhibit less obvious spatial parallelism.20
While general patterns are useful, it is also worth consider-21
ing the ways in which particular languages contribute to network22
structures. That is, the structure of a loan network is not inde-23
pendent of the languages that it comprises. Dominant languages24
within a region, for example, will tend to produce star-like bor-25
rowing patterns, as they loan words to their neighbors. We briefly26
investigate these patterns for North America. Example (6) shows27
the number of cases in which a language (or language family in the28
case of loans ofwords that reconstruct to protolanguages) has been29
the source or language of origin of a chain borrowing.30
13 However, McCarthy (1939) quotes contemporary ethnographic sources forQueensland (e.g. Roth) in which trade pathways are likewise strongly constrained,if not strictly by geography then by custom. However, McCarthy also describes theimportance of local barter systems, which have a less formal network structureand are less clearly defined than either the regional networks or the continent-wide highways, such as that characterized by the pearlshell network in Fig. 5 above.This may explain at least some of the variation between the Australian and NorthAmerican areas.
(6) Number of Wanderwörter with short and long chains, by 31
ultimate source language or group 32
3 links: Yokutsan source: 2 33
Miwokan source: 3 34
Chumashan source: 3 35
Gabrielino source: 1 36
Patwin (Wintun) source: 1 37
2 links: Yokutsan Source: 5 38
Gabrielino source: 1 39
Southern Numic source: 1 40
Sierra Miwok source: 2 41
Serran source: 1 42
Pomoan source: 1 43
Chumashan source: 1 44
Example (6) shows that different source languages dominate in 45
lexical complexes that have three-link chain borrowing, versus 46
lexical complexes that have two-link chain borrowing. Miwokan 47
and Chumashan, thought to be longer established in California, 48
dominate the three-link systems, with Yokutsan, which Golla 49
(2007) suggests is a late intrusion, moving into the southern 50
Central Valley from theGreat Basin about 1500 years ago, in second 51
position. This may reflect the greater age of three-link systems. 52
On the other hand, Yokutsan is clearly the most common source 53
in the perhaps more recent two-link borrowing chains, reflecting 54
an important impact of this intrusive radiating language group 55
on its∧neighbors. Yokutsan, along with Miwokan, is also the most 56
common source in star-shaped borrowing. There appears to be 57
slightly more variety in the source languages for two-link chains. 58
(7) Directionality of chain borrowing in short and long chains 59
3 links: West > East: 7 60
East > West: 0 61
North > South 1 in ‘mountain lion’ 62
South > North 2 63
2 links: West > East: 9 64
East > West: 1 65
South > North: 4 66
North > South: 0 67
As example (7) demonstrates, both three-link borrowing chains 68
and two-link borrowing chains exhibit strong, and similar, direc- 69
tional tendencies; most loans in chain borrowing in California go 70
from west to east. Where the dominant directionality is longitudi- 71
nal, chain borrowings are more likely to go from south to north. 72
3.4.3. South American case study region 73
In South America, it is in many cases very difficult to deter- 74
mine the source and pathway of Wanderwörter. This difficulty is a 75
result of various factors; in particular, the prevalence of small 76
language families with few or no attested sisters constrains the 77
comparative assessment of a source, and a lack of data for many 78
languages creates challenges for establishing both sources and 79
pathways. In many cases, we can identify a generalized local dis- 80
tribution of a shared form, but can do little more than guess be- 81
yond localized borrowing events. However, we can be reasonably 82
certain that Wanderwort diffusion occurred in chain-like fashion 83
in some cases. At least two links can be identified in chains in- 84
volving a probable Arawak source (‘spidermonkey’), a Carib source 85
(‘iguana’, ‘howler monkey’), and a Tupí-Guaraní source (e.g. ‘gourd 86
dipper’, ‘beans’). At least three links are probably involved in the 87
diffusion of terms for ‘coca’ (Erythroxylum coca), which apparently 88
originated with Boran or Witotoan languages and spread to the 89
northwest through Tukanoan, Arawak, Nadahup, and other lan- 90
guages of the region (see example (8) below; compare also the 91
complex spread of terms andmeanings associatedwith the etymon 92
%kumu, as described in Section 4.4 below, which cover roughly the 93
same area). 94
10 H. Haynie et al. / Ampersand xx (xxxx) xxx–xxx
Where borrowing chains can be identified in the northwest1
Amazon, loan pathways tend to originate with languages that are2
located along the larger rivers and move into the upland regions,3
with regional systems like the Vaupés area as local foci of diffusion.4
This directionality is probably associatedwith the role that Arawak5
and∧Tupi–Guarani speakers apparently played as brokers of an6
extensive systemof trade routes that followed the rivers (Hornborg7
and Eriksen, 2011).8
4. Cultural categories and the tendency of words to wander9
As discussed in Section 1 above, the distinction between Wan-10
derwörter and ordinary borrowing relies in part on an intuition11
that certain types of words are more likely than others to be12
borrowed widely across neighboring languages. That is, the13
definition of Wanderwörter is not based on the frequency of bor-14
rowing events alone. In this section we review arguments about15
different cultural categories (or lexical semantic fields) and the ev-16
idence within our dataset for items in those categories to be par-17
ticularly associated with Wanderwörter.18
4.1. Cultural items19
Hock and Joseph (1996: 242) claim that ‘‘words for cultural20
items or concepts are especially apt to become widely dispersed’’.21
In its most basic form, this hypothesis would entail that words22
for culturally significant items are more likely to becomeWander-23
wörter than are basic vocabulary items. The class of ‘cultural items’24
is itself a very broad category, of course, and the interaction be-25
tween cultural salience, etymological stability, and novelty of use26
must be considered in linking cultural significance to the likeli-27
hood of widespread diffusion. Whereas cultural centrality is asso-28
ciated in the ethnobiological literature with a resistance to lexical29
replacement for already-salient items (e.g. Berlin et al., 1973; Hunn30
and Brown, 2011), the existence of novel items and even novel uses31
for items creates situations where cultural importance increases32
the likelihood of borrowing (see further Bowern et al., 2014).33
The introduction and spread of new, culturally important items34
creates a natural opportunity for the spread of words. For example,35
the etymology of the word ‘coffee’ outlined by deVaan (2008)36
coincides with the drink’s spread from North Africa to first Turkey,37
then Italy, and then to other parts of Europe. Such a process38
was probably behind the spread of etymon %hipa ‘‘coca’’ in the39
Source/product polysemy is common in Australian languages (see10
Dixon, 1980: 103) and can lead to semantic change. For instance,11
a tree species name can become an artifact name, either retaining12
or eventually losing its tree meaning (Dixon, 2002: 56).14 In this13
case, the apparent loss of the original meaning assists in plotting14
the direction of the diffusion of the Wanderwort: where the15
original polysemy is attested (northern Jarragan) is most likely the16
origin of the word. The southward diffusion of the root %ngapale17
along the two aforementioned paths was associated with only the18
spearthrower meaning, not the tree species. (See Figs. 8–16.)Q819
4.3. Economic and cultural innovations20
Meanings associated with trade, economy, and cultural innova-21
tion are also prevalent in lists of Wanderwörter in all three case22
study areas. The natural correlation between economic systems,23
areal contact, and the diffusion of items associated with cultural24
change helps to explain the prevalence of economic and subsis-25
tence terms among the identifiedWanderwörter. This category in-26
cludes both items like ‘string of shell money’ in North America and27
domesticated species likemaize and beans in South America. To il-28
lustrate this category, we lookmore closely at the latter. Maize and29
14 Such patterns are common in our dataset. For example yirrikili is a widespreadgeneric term for boomerang in the Western Kimberley and Pilbara; it is also acommon word for Hakea arborescens, the tree from which boomerangs are verycommonly made.
Fig. 9. Languages showing reflexes of %bea ‘maize’.
Fig. 10. Languages showing reflexes of %kana ‘maize’.
beans, two of the most important food crops in the Americas, were 30
introduced to Amazonia before the arrival of Europeans. 31
All the flora Wanderwörter we identified in Amazonia are 32
domesticated, including post-Conquest introductions and a few 33
pre-Conquest items, such as maize and beans. Maize (Zea mays) 34
was domesticated from local grasses around 3500 years ago in 35
South-Eastern Mexico and spread into both North and South 36
12 H. Haynie et al. / Ampersand xx (xxxx) xxx–xxx
Fig. 11. Languages showing reflexes of %kumana ‘beans’.
Fig. 12. Languages showing reflexes of %purutu ‘beans’.
America over the next 2000 years. In the case of Uto-Aztecan this1
spread of maize cultivation may have been closely associated with2
spread of the languages (Hill, 2001). Beans (Phaseolus lunatus and3
Phaseolus vulgaris) were domesticated along the Peruvian coast4
of South America as early as 5600 years ago, with subsequent5
spread into the Andes and thence into the Amazon basin (Kaplan6
and Lynch, 1999; Kwak and Gepts, 2009). Maize is represented7
Fig. 13. Languages showing %tykuli ‘pearlshell’.
Fig. 14. Languages showing %moji ‘datura’.
by several etyma in Amazonia, notably the multiply borrowed 8
forms %bea and %kana in the northwest. A widespread term for 9
‘beans’, %kumana, probably originated in∧Tupi–Guarani, but was 10
likely disseminated in part via Arawak languages, which were also 11
a source of one ‘maize’Wanderwort. Quechua appears to have been 12
the source for the other widespread term for ‘beans’, %purutu; 13
however, since regional Spanish also adopted a variant of this 14
H. Haynie et al. / Ampersand xx (xxxx) xxx–xxx 13
Fig. 15. Languages showing %kumu ‘drum’.
Fig. 16. Wanderwort patterns for %palaka ∼ %panak ‘woodpecker’.
Wanderwort, it is possible that diffusion of some of the %purutu1
forms was mediated more directly by Spanish.152
15 For further discussion and reconstructions, see Brown et al. (2014, forthcom-ing), Chacon (2013), Mello (2000) and Ramirez (2001).
(10) %bea ‘maize’ 3
Chocoan *pe <- -> Tukanoan *we’a (Cubeo we’á, West 4
In the Americas, where agriculture, or at least plant cultivation, 20
played some role in the subsistence of many societies, the 21
spread of borrowed terms for domesticated species may well 22
have accompanied the spread of agricultural cultivation. Where 23
cultivation was practiced in the Americas, these items may 24
have been moved by humans into new areas (as seeds, root- 25
stock, etc.) or in some cases the idea of utilizing previously 26
existing wild varieties as crops may have diffused, creating an 27
opportunity for cultivar terminology to spread along with these 28
crops. While more focused work on a case-by-case basis is needed 29
to determine the extent to which this scenario has applied, it has 30
been reasonably well established in particular regions (e.g. the 31
partial adoption of agricultural technology by foragers in the 32
northwest Amazon; Epps, forthcoming) and for particular cultivars 33
(e.g. Brazil nut, Bertholletia excelsa; Shepard and Ramirez, 2011). 34
Pre-contact animal domestication only occurred in restricted areas 35
in the Americas and is not reflected in our sample; however, 36
domesticated animals introduced by Europeans did give rise 37
to borrowing events, including early post-Contact diffusion of 38
Wanderwörter (see further Section 4.6). 39
4.4. Ritual and spiritual associations 40
This category includes both items associatedwith ritual use and 41
those with more abstract links to spiritual beliefs. Wanderwörter 42
of this type appear to have accompanied the spread of ritual and 43
mythological complexes throughout the study regions, however 44
the exact paths of many of these spiritual elements are unclear. 45
Examples of this are ‘pearlshell’, a decorative item in Australia, 46
and ‘Datura’, a psychotropic substance in North America; see 47
also the example of ‘‘coca’’ (Erythroxylum coca) in South America, 48
mentioned above. 49
Shells are trade items in many areas of the world, often valued 50
for their beauty and for use as ritual objects and decorations. 51
In Australia the pearlshell (Pinctada maxima) is used this way 52
(Akerman and Stanton, 1994)—as a pendant and pubic covering 53
for men and boys in initiation ceremonies over a wide area of 54
northern Australia and beyond. The mother-of-pearl is also often 55
carved to create objects of value. The source of these shells is on the 56
Kimberley coast from where they were traded as far as the South 57
Australian Bight and Gulf of Carpentaria on the other side of the 58
continent. 59
One of the terms for this shell in a number of languages 60
is %tyakuli ‘pearl shell’.16 The origin of the term appears to be 61
16 Pearlshells from Nyulnyulan language areas were also traded, but theNyulnyulan term is %riityi, anotherWanderwortwith aWestern/Southern diffusionroute, shown as a zone on the Australian trade route map in Fig. 5.
14 H. Haynie et al. / Ampersand xx (xxxx) xxx–xxx
tyakoli in Worrorra (or a closely related language), the coastal1
language spoken in one of the regions where the shells were2
collected. The path of borrowing can be traced from Worrorran3
into Jarragan languages, and thence further east into the∧Non-4
Pama–Nyungan language Jaminjung and to the∧Ngumpin–Yapa5
subgroup of Pama–Nyungan, then into Anmatyerr, an Arandic6
Tümpisa Shoshone, Big Smokey Valley Shoshone mui-ppeh,50
Kawaiisu , Northern Sierra Miwok moo-tah 99K North-51
ern Paiutemoohoo’oo ‘‘opium’’52
Takic languages have two variants of a word for ‘moon’. The ex-53
pected regular form reflects Proto-Northern Uto-Aztecan .54
However, Cupeño has both ‘month’ and ‘moon’.55
Cahuilla has meni-ly ‘moon’. The Salinan forms suggest that this56
variation may be quite old (a similar /y/ ∼ /n/ alternation appears57
17 We suggest a possible Yokutsan intermediary between Salinan and SierraMiwok. However, the word for ‘‘Datura’’ in Yokuts varieties recorded by C.H.Merriam is universally %ta:nai (related to a verb for ‘‘to be drunk’’). The connectionto Sierra Miwok of Salinan may be direct trade, or it may be that the form onceappeared in Plains Miwok, another possible intermediary language.
in Takic *kwija- ∼ *kwini- ‘Black Oak, Quercus kelloggii’), and the 58
Mutsun and Miwokan items appear to have the Migueleño Sali- 59
nan word, and ultimately a Takic form with /n/, as a source. An 60
obvious question is whether a semantic change from Takic ‘moon’ 61
to ‘datura’, a hallucinogenic plant, in the borrowing languages can 62
be justified. In this case, the association is clear; Applegate (1975) 63
has clearly shown the association between themoon and the plant 64
in the Chumash datura cult, and the association is widely docu- 65
mented in the southern California languages.18 66
From South America, a particularly interesting example of a 67
Wanderwort associated with cultural and ritual practice involves 68
a set of items with strong cultural significance in the northwest 69
Amazon, all of which are assigned to the etymon %kumu, 70
with different patterns of polysemy in different languages. This 71
Wanderwort is found among languages of the East Tukanoan 72
family (Vaupés region) and of the area to the southwest occupied 73
by the ‘People of the Center’ (speakers of Bora,Witoto, Arawak, and 74
other languages); these regions share a number of other linguistic 75
and cultural features that indicate a history of contact. Variants 76
of the %kumu etymon mean ‘signal drum’ (a large drum made 77
from a whole section of tree trunk) for Bora and Arawak speakers, 78
while in Tukanoan languages %kumu exhibits polysemy among the 79
meanings ‘canoe’,19 ‘bench’, and ‘healer/shaman’ (for whom the 80
traditional seat is the carved, painted Tukanoan bench).20 Notably 81
similar terms also occur in both regions with the meaning ‘laurel 82
tree’ (fam. Lauraceae), which is a standard material from which 83
signal drums and canoes are made. Unfortunately, comparative 84
data are at this point too scanty to work out the full story of how 85
this Wanderwort spread. 86
(16) %kumu ‘signal drum’, ‘laurel tree (fam. Lauraceae)’, ‘canoe’, 87
‘bench’, ‘healer/shaman’ 88
‘signal drum’: Bora (Bora khuúmu, Muinane ), Arawak 89
(Resigaro koómó, Yucuna kumu) 90
‘laurel tree (fam. Lauraceae)’: East Tukanoan (Makuna kuma, 91
Tukano ), Arawak (Yucuna kumujlé), Bora (Bora 92
khuumúru-he) 93
‘canoe’: East Tukanoan (Barasano k u bu-ã, Carapana kubu-ã, 94
18 Klar (1977) considers the Chumashan form, with variants found in allChumashan languages, to be Proto-Chumashan. However, by chance all of thesounds in the word that Klar (1977) reconstructs as Proto-Chumashan *mom’oj (areduplication of *moj) have been stable in the history of Chumashan, so we cannotdistinguish in this case a genuine Proto-Chumashan form from a word that hasspread by borrowing. Therefore this reconstruction is not definitive. However, thesource of the loan probably pre-dates the loss of intervocalic /j/ in Gabrielino (themost likely source language, given its close geographic and cultural associationwith Chumash); the historically-attested form in Gabrielino is /mwa:r/ from Pre-Gabrielino *moja-la. The absence of the reduplication in Salinan suggests that thereduplication may be a relatively recent innovation in Chumashan. The WesternShoshone word with the root mui- has sometimes been suggested to come fromwords for ‘‘crazy’’. However, the Numic words are not cognate with one anotherand it is likely that the Shoshone word comes from Tübatulabal, possibly withsome phonological convergence with words for ‘‘crazy’’ given the well-knownhallucinogenic effects of consuming infusions of the plant.19 While polysemous uses of the %kumu etymon do not appear to include both‘‘signal drum’’ and ‘‘canoe’’ in any of the languages concerned, the distinct etymon
does share themeanings ‘‘canoe’’ and ‘‘signal drum’’ across the East Tukanoanfamily; the presence of an apparently cognate form meaning ‘‘canoe’’ in WestTukanoan suggests that the ‘‘canoe’’ meaning was prior (but cf. Chacon, 2013).20 An association between benches and canoes among Tukanoan peoples mayderive from a belief that shamanic activities recreate the voyage of the ancestralcanoe, a central theme of Tukanoan mythology.
H. Haynie et al. / Ampersand xx (xxxx) xxx–xxx 15
Thus in summary,words denoting items or plants involved in ritual1
are well attested in our∧dataset.2
4.5. Acculturation terms3
Widespread lexical items may reflect repeated instances of4
borrowing from a spreading language. Examples can be found in5
loans from colonial languages. For instance, in the North American6
sample, nearly all languages have a loan from Spanish caballo for7
‘horse’. However, in no case havewe been able to demonstrate that8
the loan was more than two borrowing steps away from Spanish9
itself, and in most cases the form of the loan suggests a borrowing10
directly from Spanish, as with Maidu kawáju ‘horse’, which is very11
close to the Spanish form even though Maidu, a language of the12
Sierra Nevada highlands in northeastern California, was spoken13
at a considerable distance from any early settlement of Spanish14
speakers.15
In contrast, in the Australian sample, words for ‘horse’ are not16
usually from English (the contact language of the colonial period),17
even though the terms are widespread. Rather, terms such as18
%timana, %yarraman, %nantu and %yawarta are in origin terms in19
Australian languages. %nantu, for example, is the Kaurna (Adelaide20
area) word for ‘ghost‘, while %yawarta is a species of kangaroo in21
∧Nyungan (Walsh, 1992).22
Acculturation terms such as ‘horse’ show up in our loan count23
data as extremely highly borrowed terms, borrowed so frequently,24
in fact, that they are assigned their own category by data parti-25
tioning algorithms. Whether the type of widespread loan patterns26
associated with colonialism and acculturation is to be considered27
a sub-type of Wanderwörter or an entirely different phenomenon28
depends ondefinitional criteria. As noted above in the case of Span-29
ish ‘horse’ in North America, the widespread presence of colo-30
nial languages often makes it particularly difficult to distinguish31
between multiple independent borrowing events involving the32
colonial source language and various recipient languages, versus33
chain- or network-like sets of borrowing events by which a word34
∧traveled through multiple languages.35
4.6. Wanderwörter less clearly linked to cultural diffusion36
The majority of the terms identified as Wanderwörter in our37
three case study areas can be linked to the diffusion of cultural38
items and practices through the types of associations illustrated39
above.However, a small number ofWanderwörter are foundwhich40
are not of exceptional economic importance, nor are they obviously41
related to ritual or other cultural significance in the recent period.42
Examples include a generic term for ‘fish’ in North America and43
%buthuru, a highly borrowed term for ‘ear’ in Australia. One44
hypothesis is that these items were associated with diffusions of45
cultural patterns in the distant past but evidence of these routes46
has become hard to gather or interpret. Particularly in the case47
of ritual or mythological associations, the pathways of cultural48
diffusion may be difficult to trace at deep historical distances. The49
Australian term is a recent loan from English ‘bottle’ and so may50
have been associated with necronym taboo replacement.51
This type of Wanderwort, less securely linked to cultural diffu-52
sion, is particularly common in the domain of flora and fauna. Such53
Wanderwörter are not likely to be associated with movement of54
the items named into new areas, as suggested for trade, cultivar,55
and ritual items above, since the species involved are widespread56
and stable. However, it is possible that some of these Wander-57
wörter are associated with very early spread of species or species58
use.59
An example of a Wanderwort displaying this sort of unclear60
link to cultural diffusion can be seen in the terms for ‘iguana’61
in northwestern Amazonia (probably themselves related), which62
have beenwidely borrowed among languages of the region (aswell 63
as into Spanish, probably fromArawak, and thence from Spanish to 64
22 However, this possibility is limited in our case study area because of the numberof languages from
∧non-Pama–Nyungan families. Ancient retentions are likely to
form only a very small proportion of our sample, butmight be a possibility for someof the items which are widespread in
∧Pama–Nyungan but only in a few, adjacent
∧Non-Pama–Nyungan languages. They would therefore be regular loans into the
∧Non-Pama–Nyungan languages, not Wanderwörter.
The Yuman languages, Salinan, and Seri/Comcaac are divided from 57
one another by Uto-Aztecan and Chumashan languages that do not 58
exhibit any similar item for big-game ungulates. It is, of course, 59
possible that this is an old loan, exchanged before the expansion 60
of Chumashan and Uto-Aztecan into their present locations. The 61
other possibility is that this is a vocabulary item inherited from a 62
common ancestor of Salinan, Yuman, and Seri/Comcaac. Yuman, 63
Salinan, and Seri/Comcaac have all been regarded as candidates for 64
a membership in a Hokan genetic unit. Most Americanist linguists 65
consider proposals for a Hokan family of languages to be unproven 66
and perhaps unprovable. However, not only are the words for Q1167
‘bighorn ‘sheep’‘’ and ‘’’‘antelope’‘‘’ similar across these languages, 68
but two Yuman languages (in different sub-branches), Salinan, 69
and Seri/Comcaac all have similar words for both ‘Bighorn sheep’ 70
and ‘antelope’, suggesting the possibility of some kind of ancient 71
derivational relationship between names for these two animals.23 72
The animals are quite different (although both are important game 73
animals), so this is a possible ‘‘shared anomaly’’ thatmay represent 74
an ancestral ‘Hokan’ process. The ‘‘’‘antelope’‘‘’ form appears to 75
have been loaned into Yokutsan and Wappo, neither of which 76
has ever been suggested as a ‘Hokan’ language. Note that no 77
similar form is attested in the so-called ‘Hokan isolates’ of northern 78
California: Pomoan, Shastan, Yana, and Washo. 79
A further way in which languages in an area may come to 80
share forms, but without a process of widespread borrowing, is if 81
a word is borrowed from one ancestor language into another and 82
the languages subsequently diversify. For example, a number of 83
Nyulnyulan and∧Pama–Nyungan languages share reflexes of the 84
form *mara ‘hand’; rather than being either a widespread loan 85
or an ancient retention from Proto-Australia, however, the form 86
was most likely borrowed from a precursor of one of the modern 87
∧Ngumpin–Yapa languages into Proto-Nyulnyulan, as evidence by 88
the shared reflex –marla (the sound change of r > rl is regular 89
in∧Ngumpin–Yapa; see McConvell and Laughren, 2004). We have 90
excluded such forms from our counts where we have the evidence 91
to do so (for example, where a lexical item follows regular sound 92
correspondences between related languageswhich suggests that it 93
is an inheritance rather than widespread loan). 94
5.2. Substratal elements 95
Another possibility that must be considered in the case of a ge- 96
ographically widespread lexical complex is the presence of a sub- 97
stratum language from which words have been borrowed.24 Such 98
languages may be ancient and unattested, compounding the prob- 99
lem of identifying them as sources. The option of∧analyzing appar- 100
ent Wanderwort-like networks as manifestations of substrates is 101
more or less confined to cases where we can independently find 102
evidence of a substrate and where there are attested languages 103
closely related to the substrate language to bear witness. 104
5.3. Sound symbolism 105
Widespread lexical items may reflect independent inventions 106
of rather similar lexical items, motivated by iconicity (as in the 107
case of sound symbolism and ideophones). For instance, bird 108
names are likely sites for sound-imitative formations that may be 109
23 The abstract suffixes -X and -Y in the labels for the lexical complexes aboverepresent this proposal.24 In this paper we are using ‘substratum’ to mean that language A which existedin an area prior to language shift to language B, and which provided elements to alanguage B+, changed from B by contact. There are no assumptions or implicationsthat the substratal language A was socially subordinate to B (as in some definitionsof the term).
H. Haynie et al. / Ampersand xx (xxxx) xxx–xxx 17
quite similar cross-linguistically because the designated bird has1
a distinctive cry. Hunn (1975) has discussed this possibility for2
names for owls in the Americas. However, sound imitative roots3
are sometimes deformed by regular sound changes, and so cannot4
be repeated independent inventions. They may be inherited or5
re-borrowed into other languages in the modified form following6
sound change. The potential for sound symbolism to obscure7
or lend uncertainty to the diagnosis of Wanderwort status is8
discussed above in Section 4.6 in relation to North American acorn9
woodpecker terms.10
6. Conclusions11
Wanderwörter have∧traveled between languages and peoples12
not only in the wake of imperial expansion but also between the13
smaller languages and societies of the world, giving voice to new14
ideas and names to new products and practices. The limited study15
of Wanderwörter to date has focused on agricultural or industri-16
alized societies; however, as we show here, the phenomenon is17
well attested in networks of smaller languages, including those of18
hunter-gatherers.19
There is evidence that agriculture plays a role in the pattern-20
ing of Wanderwörter, since words for cultivated crops are often21
widespread, even in regions like Amazonia where general borrow-22
ing is very low. Words for wild plants, in contrast, rarely travel as23
Wanderwörter, unless they are also∧artifact terms. Nevertheless,24
our exploration of the cultural, pragmatic, and lexical semantic25
facets of Wanderwort spread demonstrate that specific categories26
of cultural association, including but not limited to agricultural cul-27
tivation, seem to condition widespread borrowing. We find Wan-28
derwörter not only in agricultural and industrial societies, but also29
in situations of contact among exclusively non-agricultural groups30
that predate colonial influences.31
Trade in a broad sense is not confined to commercial markets32
but has played a significant role in all the small and forager33
societies we have looked at. Some prized items were exchanged34
along long chains and frequently this also involved the passing on35
of words for the items, whether these involved new technologies,36
ritual paraphernalia or drugs. However, Bowern et al. (2014)37
suggest that trade status alone is not enough to elevate the loan38
likelihood of lexical items, and we find that although certain39
traded items appear to spread along networks associated with40
trade pathways, the overall association between Wanderwörter41
networks and trade networks is inconsistent.42
While Wanderwörter continue to be a difficult linguistic43
phenomenon to describe and∧analyze, the results of this study44
point us toward a clearer notion of what these items are and why45
they are borrowed so widely. Our definition of Wanderwort might46
be edited in light of these findings to reflect the conclusion that47
Wanderwörter are essentially outliers in the realm of loanwords,48
borrowed far more frequently than typical lexical items but still49
a subset of this more general phenomenon. They are traditionally50
defined as any widely borrowed items that diffuse through areal51
or chain-like networks. However, we note that this diffusion is52
typically made possible by the spread of cultural items, customs,53
or ideas, and suggest that the link between Wanderwörter and54
cultural diffusionmay be amore sound basis for defining this term55
than the traditional definitions that rely on specific claims about56
the loan frequency, areality, or untraceability of these terms.57
Uncited references58
Q12
Akerman and Patrick, 2002, Hill, 2011 andWhitaker et al., 2007.59
Appendix A. Supplementary material 60
Supplementary material related to this article can be found 61
online at http://dx.doi.org/10.1016/j.amper.2014.10.001. 62
References 63
Akerman Kim, Patrick McConvell. ‘Wommera’ - The Technology and Terminology 64
of the Multipurpose Spearthrower in Australia. In: ARCLING II: conference 65
on the archaeology and linguistics of australia. Canberra: Australian National 66
University Centre for Research on Language Change; 2002. 67
Akerman Kim, Stanton John. Riji and Jakuli: Kimberley pearl shell in Aboriginal 68
Australia. Darwin: Northern Territory Museum of Arts and Sciences; 1994. 69
Applegate Richard B. The datura cult among the Chumash. J California Anthropol1975;2:7–17.
70
Balée William. Historical-ecological influences on the word for cacao in Ka’apor.Anthropol Linguist 2003;45:259–80.
71
Barber EJW. The archaeolinguistics of hemp. In: Prehistoric textiles: the develop-ment of cloth in the Neolithic and Bronze Ages. Princeton: Princeton UniversityPress; 1991. p. 36–8.
72
Basso Keith. Semantic aspects of linguistic acculturation. Amer Anthropol 1967;69(5):471–7.
73
Bastian Mathieu, Heymann Sebastien, Jacomy Mathieu. Gephi: An open source 74
software for exploring and manipulating social networks. Paper presented at 75
international AAAI conference on weblogs and social media, 2009. 76
Berlin Brent, Breedlove Dennis E, Laughlin Robert, Raven Peter H. Culturalsignificance and lexical retention in Tzeltal-Tzotzil ethnobotany. In: EdmonsonMunro S, editor. Meaning in Mayan languages. The Hague: Mouton; 1973.p. 143–64.
77
Berndt Ronald, Berndt Catherine. The world of the first Australians. Sydney: UreSmith; 1964.
78
Blench Roger. Bananas and plantains in Africa: Re-interpreting the linguisticevidence. Ethnobot Res Appl 2009;7:363–80.
79
Bowern C, Epps P, Gray R, Hill J, Hunley K, McConvell P, et al. Does lateral 80
transmission obscure inheritance in Hunter-Gatherer languages? PLoS ONE 81
The paleobiolinguistics of domesticated bean (Phaseolus sp.). Ethnobiol Lett2014 [forthcoming].
Q1486
Campbell Lyle, Grondona Verónica. Linguistic acculturation in Nivaclé and Chorote.Internat J Amer Linguist 2012;78(3):335–67.
87
Campbell Lyle, Mixco Mauricio J. A glossary of historical linguistics. Edinburgh:Edinburgh University Press; 2007.
88
Capell Arthur. Mythology in northern kimberley, North-West Australia. Oceania1939;9:382–404.
89
Cavanaugh Joseph E. Unifying the derivations of the Akaike and corrected Akaikeinformation criteria. Statist Probab Lett 1997;31:201–8.
90
Chacon Thiago. On proto-languages and archaeological Cultures: Pre-history andmaterial culture in the Tukanoan family. Rev Bras Linguíst Antropol 2013;5:217–45.
91
Davidson Daniel S. The spearthrower in Australia. Proc Amer Phil Soc 1936;76(4):445–83.
92
Denham Tim, Donohue Mark. Pre-Austronesian dispersal of banana cultivars westfrom New Guinea: Linguistic relics from eastern indonesia. Archaeol Oceania2009;44(1):18–28.
93
deVaan Michiel. On Wanderwörter and substrate words in etymological research.In:MooijaartMarijke, vanderWalMarijke, editors. Yesterday’swords. Contem-porary, current and future lexicography. Cambridge Scholars Publishing; 2008.p. 199–207.
94
Davis James T. Trade routes and economic exchange among the indians of california.Reports of the university of california archaeological survey no. 54. Berkeley(CA): University of California Berkeley Anthropology Department; 1961.
95
Dixon RMW. The languages of Australia. Cambridge: Cambridge University Press;1980.
96
Dixon RMW. Australian languages: their nature and development. Cambridge:Cambridge University Press; 2002.
97
Dolgopolsky Aharon. The nostratic macrofamily and linguistic palaeontology.Cambridge: McDonald Institute; 1998.
98
Q15Epps Patience. Language and subsistence patterns in the Amazonian Vaupés.
In: Güldemann Tom, Rhodes Richard, McConvell Patrick, editors. The lan-guages of hunter-gatherers: global and historicals perspectives. Cambridge:Cambridge University Press; 2014 [forthcoming].
99
Gamkrelidze Thomas V, Ivanov Vjacheslav V. Indo-European and the Indo-Europeans: a reconstruction and historical typological analysis of a protolan-guage and proto-culture. Parts I and II. Berlin: Walter De Gruyter; 1995.
Gayton AH(Anna Hyde). Yokuts and Western mono ethnography. II: Northernfoothill Yokuts and western mono. Anthropol Records 1948;10(2):143–301.
1
Haspelmath Martin. Lexical borrowing, concepts and issues. In: Haspelmath,Tadmor, editors. Loanwords in theworld’s languages: a comparative handbook.Berlin: Walter de Gruyter; 2009a. p. 35–54.
2
Haspelmath Martin, Tadmor Uri, editors. Loanwords in the world’s languages: acomparative handbook. Berlin: Walter de Gruyter; 2009b.
3
Hill Jane. Proto-Uto-Aztecan: A community of cultivators in central Mexico? AmerAnthropol 2001;103(4):913–34.
4
Hill Jane. Borrowed names and indexical function in the Northern Uto-Aztecanbotanical lexicon. In: Whallon Robert, Lovis William, Hitchcock RobertK,editors. Information and its role in Hunter-Gatherer bands. Los Angeles (CA):UCLA Cotsen Institute of Archaeology Press; 2011. p. 167–80.
5
Hock Hans Henrich, Joseph Brian D. Language history, language change, and lan-guage relationship: an introduction to historical and comparative linguistics.Berlin: Mouton; 1996.
6
Hornborg Alf, Eriksen Love. An attempt to understand Panoan ethnogenesis inrelation to long-term patterns and transformations of regional interaction inwestern Amazonia. In: Hornborg Alf, Hill Jonathan D, editors. Ethnicity inancient amazonia. Boulder: University Press of Colorado; 2011. p. 129–54.
7
Hunn Eugene. Words for owls in North American Indian languages. Internat J AmerLinguist 1975;41:237–9.
8
Hunn Eugene S, Brown Cecil H. Linguistic ethnobiology. In: Anderson EN, PearsallDeborah M, Hunn Eugene S, Turner Nancy J, editors. Ethnobiology. Hoboken(N.J.): Wiley-Blackwell; 2011. p. 319–34.
9
Kaplan Lawrence, Lynch Thomas F. Phaseolus (fabaceae) in archaeology: AMSradiocarbon dates and their significance for pre-Colombian agriculture. EconBot 1999;53:261–72.
10
Klar Kathryn Ann. Topics in historical Chumash grammar [Ph.D. dissertation]Berkeley: University of California; 1977.
11
Kofod Frances. The miriwung language (east Kimberley): a phonological andmorphological study [MA thesis]. University of New England; 1978.
MacQueen JB. Some methods for the classification and analysis of multivariate24
observations. In: Le Cam Lucien M, Neyman Jerzy, editors. Proceedings of the25
fifth berkeley symposium on mathematical statistics and probability. Berkeley,26
California: University of California Press; 1967; 281–97.27
McCarthy FD. ‘Trade’ in aboriginal Australia and ‘trade’ relationships with TorresStrait, New Guinea, and Malaya. Oceania 1939;9:405–38.
28
McConvell Patrick. Loanwords in Gurindji, a Pama–Nyungan language of Australia.In: Haspelmath Martin, Tadmor Uri, editors. Loanwords in the world’slanguages. Berlin: Walter de Gruyter; 2009. p. 790–822.
29
McConvell Patrick. Lexical contact phenomena in Australian linguistic prehistory: 30
substrates andWanderwoerter. Paper presented to the symposiummethodology 31
McConvell Patrick, Smith Mike. Millers and mullers: The archaeolinguistic stratig-raphy of seed-grinding in Central Australia. In: Andersen H, editor. Languagecontacts in prehistory: studies in stratigraphy. Amsterdam: Benjamins; 2003.p. 177–200.
34
Mello Antônio Augusto Souza. Estudo histórico da familia linguistica Tupi-Guaraní[Ph.D dissertation] Universidade Federal de Santa Catarina; 2000.
35
Nelson-Sathi Shijulal, List Johann-Mattis, Geisler Hans, Fangerau Heiner, GrayRussell D, Martin William, et al. Networks uncover hidden lexical borrowingin Indo-European language evolution. Proceedings of the Royal Society B:Biological Sciences 2011;278(1713):1794–803.
36
Pejros Ilya. Are correlations between archaeological and linguistic reconstructionspossible? In: Blench Roger M, Spriggs Matthew, editors. Archaeology andlanguage I. London: Routledge; 1997. p. 149–57.
37
Pelleg Dan,Moore Andrew. X-means: Extending K -meanswith efficient estimationof the number of clusters. In: Langley Pat, editor. Proceedings of theseventeenth international conference on machine learning. San Francisco:Morgan Kaufmann Publishers; 2000. p. 727–34.
38
Polomé E. Indo-European and substrate languages in thewest. Arch Glottol Italiano1992;67:66–85.
39
Ramirez Henri. Línguas Arawak da Amazônia setentrional: comparação e descrição.Manaus: Editora da Universidade; 2001.
40
Roberge Paul. Contact and the history of Germanic languages. In: Hickey Raymond,editor. Handbook of language contact. Chichester: Wiley-Blackwell; 2010.
41
Rogers Everett M. Diffusion of innovations. Glencoe: Free Press; 1962. 42
Rydén Stig. Salt trading in the Amazon basin: Conclusions suggested by thedistribution of Guarani terms for salt. Anthropos 1962;57:644–59.
43
Shepard Jr, Glenn H, Ramirez Henri. Made in Brazil: Human dispersal of the Brazilnut (Bertholletia excelsa, Lecythidaceae) in Ancient Amazonia. Econ Bot 2011;65(1):44–65.
44
Sherratt A. Introduction: Recupliar habits. In: Goodman J, Sherratt A, LovejoyP, editors. Consuming habits: drugs in history and anthropology. London:Routledge; 1995. p. 1–46.
45
Troike Rudolph C. Woodpeckers and linguistic evidence for trade in central 46
Californian languages. MS: University of Arizona; 2009. 47
Walsh Michael. A nagging problem in Australian lexical history. In: Dutton Tom,Ross Malcolm, Tryon Darrell, editors. The language game: papers in memory ofDonald C Laycock. Canberra: Pacific Linguistics; 1992.
48
WalshGL,MorwoodMJ. Spear and spearthrower evolution in the Kimberley region,N.W. Australia: Evidence from rock art. Archaeol Oceania 1999;34(2):45–59.
49
Whitaker Adrian R, Eerkens Jelmer W, Spurling Amy M, Smith Edward L, GrasMichelle A. Linguistic boundaries as barriers to exchange. J Archaeol Sci 2007;35:1104–13.