PreprintPDF Available

Evolutionary dynamics in the dispersal of sign languages

  • Trotters Independent Traders
Preprints and early-stage research may not have been peer reviewed yet.


While the evolution of spoken languages is well understood and has been studied using traditional historical comparative methods as well as newer computational phylogenetic methods, evolutionary processes resulting in the diversity of contemporary sign languages are poorly understood, and scholars have been largely unsuccessful in grouping sign languages into monophyletic language families. To date, no published studies have attempted to use language data to infer relationships amongst sign languages on a large scale. Here, we report the results of a phylogenetic analysis of 40 contemporary and 36 historical sign language manual alphabets coded for morphological similarity. Our results support grouping sign languages in the sample into six main European lineages, with three larger groups of Austrian, British, and French origin, as well as three smaller groups centering around Russian, Spanish, and Swedish. The British and Swedish lineages support current knowledge of relationships amongst sign languages based on extra-linguistic historical sources. With respect to other lineages, our results diverge from current hypotheses by indicating (i) independent evolution of Austrian, French, and Spanish from Spanish sources; (ii) an internal Danish subgroup within the Austrian lineage; and (iii) evolution of Russian from Austrian sources.
Evolutionary dynamics in the dispersal of sign
Justin M. Power1, Guido W. Grimm2and Johann-Mattis List3
1Department of Linguistics, University of Texas at Austin, USA
2Independent researcher, Orléans, France
3DLCE, Max Planck Institute for the Science of Human History, Jena, Germany
*corresponding author:
June 2019
While the evolution of spoken languages is well understood and has been studied
using traditional historical comparative methods as well as newer computational phy-
logenetic methods, evolutionary processes resulting in the diversity of contemporary
sign languages are poorly understood, and scholars have been largely unsuccessful
in grouping sign languages into monophyletic language families. To date, no pub-
lished studies have attempted to use language data to infer relationships amongst
sign languages on a large scale. Here, we report the results of a phylogenetic analy-
sis of 40 contemporary and 36 historical sign language manual alphabets coded for
morphological similarity. Our results support grouping sign languages in the sam-
ple into six main European lineages, with three larger groups of Austrian, British,
and French origin, as well as three smaller groups centring around Russian, Spanish,
and Swedish. The British and Swedish lineages support current knowledge of rela-
tionships amongst sign languages based on extra-linguistic historical sources. With
respect to other lineages, our results diverge from current hypotheses by indicating
(i) independent evolution of Austrian, French, and Spanish from Spanish sources;
(ii) an internal Danish subgroup within the Austrian lineage; and (iii) evolution of
Russian from Austrian sources.
Keywords: sign language, language phylogeny, language evolution, phylogenetic net-
1 Introduction
Linguistic analyses of the world’s sign languages (SLs) over the past 60 years have shown that the
human capacity for language is not limited to the oral-aural modality. Instead, homo symbolicus
has developed complex natural language in the gestural-visual modality as well, particularly in
deaf signing communities throughout the world. The development of educational institutions
for the deaf, which began during the Enlightenment in Europe, especially in the late 18th and
early 19th centuries, contributed to the formation of these deaf signing communities and the
emergence of widespread, conventional SLs [1, 2]. The languages that emerged in these newly
formed communities were soon dispersed to other parts of Europe and beyond. The success of
the rst public school for the deaf, the Institut National de Jeunes Sourds de Paris, founded
Power et al. Sign Language Dispersal June 2019
between 1759–71 [3], attracted educators from across Europe and the Americas, who came to
the Paris Institute to learn pedagogical methods, with the goal of establishing schools for the
deaf in their countries of origin [4]. Deaf students from other countries also came to Paris,
graduated, and returned home to work as teachers or to found new schools [5, 6]. Thus, within
a century of the founding of the Paris Institute, Langue des signes française (French SL) had
reached the Netherlands in 1790 [7], the US in 1817 [8], Brazil in 1857 [9], as well as other
countries, creating linguistic connections between geographically distinct signing communities.
Similarly, other European SLs spread along political and colonial pathways [10–12].
Although the history of European deaf educational institutions has been well-documented
[13, 14], much less is known about how SLs themselves relate to one another. Whereas the world’s
spoken languages have been classied in families and subfamilies based on their evolutionary
histories, few attempts have been made to form large-scale genetic classications of the world’s
SLs [15, 16], which are typically missing from overviews of the world’s language families (e.g.,
[17]; see [18, 19] for overviews including SLs). This lacuna in our knowledge of the history of an
entire class of human languages is due in part to challenges in understanding the evolutionary
processes that have shaped the diversity of contemporary SLs. In particular, using traditional
historical comparative methods, sign researchers have been unable to distinguish the results of
tree-compatible evolutionary processes—that is, patterns of similarity reecting inheritance in a
vertical ancestor-descendant relationship—from tree-incompatible processes, such as borrowing
and convergence. As a consequence of these methodological challenges, comparative studies of
SLs at times conate vertical and horizontal relationships in forming SL families [20, 41] or forgo
historical interpretations of their results [21].
While relationships amongst spoken languages have been studied using both traditional meth-
ods and, more recently, computational phylogenetic methods [22, 23], to date no published stud-
ies have attempted to use phylogenetic methods to infer relationships amongst SLs on a large
scale. Here, we use network-based exploratory data analysis (EDA) [24] with a sample of 76
SL manual alphabets (40 extant, 36 historical) as a rst step in investigating the evolutionary
histories of SLs and the processes that have shaped them. Our approach makes use of data-
display networks, which represent both tree-compatible and tree-incompatible patterns within a
data set and are therefore a useful starting point for understanding evolutionary processes and
formulating phylogenetic hypotheses.
2 Materials and Methods
2.1 Data
We created a data set of 76 manual alphabets, comprising 2,124 total entries from contempo-
rary (40, SI 2.1) and historical (36, SI 2.2) sources, both print and online, the geographical
distribution of which is illustrated in Figure 1. We transcribed manual alphabet handshape
forms using HamNoSys [25], a transcription system for SLs, and coded each for morphologi-
cal similarity using the EDICTOR tool [26], which was originally developed for the curation
and analysis of historical comparative spoken language data. The data can be accessed and
inspected through the EDICTOR tool (SI 1). In addition, we share the data in the TSV format
required for processing by the LingPy software package [27], and in CLDF format, following the
recommendations of the Cross-Linguistic Data Formats Initiative [28].
A manual alphabet (MA) is a set of forms representing an alphabet, in which one form
corresponds to one letter. An individual form is invariably comprised of a handshape with a
particular spatial orientation, and may include a characteristic location and/or movement. His-
torical comparative studies of languages typically use basic vocabulary as comparanda, but there
Power et al. Sign Language Dispersal June 2019
1 Yebra 1593 20 New Zealand Sign Language 39 Danish 1967 58 Jordanian Sign Language
2 Bonet 1620 21 Brazilian 1875 40 Danish Sign Language 59 Latvian Sign Language
3 French 1800 22 Indian Sign Languaage 41 Estonian 1988 60 Lithuainian Sign Language
4 British Sign Language 23 Afghan Sign Language 42 Estonian Sign Language 61 Mexican Sign Language
5 Austrian 1786 24 Albanian Sign Language 43 Finnish Sign Language 62 Norwegian 1900
6 German 1820 25 American 1886 44 Flemish Sign Language 63 Norwegian 1955
7 Italian 1897 26 American 1918 45 French 1803 64 Norwegian Sign Language
8 Dutch 1790 27 American Sign Language 46 French 1815 65 Pakistan Sign Language
9 Hungarian 1827 28 Austrian 1823 47 French 1856 66 Polish Sign Language
10 Spanish 1815 29 Austrian 1839 48 French Sign Language 67 Quebec Sign Language
11 Russian 1835 30 Austrian Sign Language 49 French-Belgian Sign Language 68 Russian Sign Language
12 Danish 1808 31 Brazilian Sign Language 50 German 1909 69 Dutch 1820
13 Swedish 1866 32 Bulgarian Sign Language 51 German 1916 70 Dutch Sign Language
14 American 1821 33 Catalan Sign Language 52 German Sign Language 71 Spanish 1845
15 Polish 1879 34 Croatian Sign Language 53 Greek Sign Language 72 Spanish 1859
16 Portuguese Sign Language 35 Czech Sign Language 54 Icelandic Sign Language 73 Spanish Sign Language
17 Norwegian 1893 36 Danish 1871 55 International Sign 74 Swedish Sign Language
18 Australian Sign Language 37 Danish 1907 56 Irish Sign Language 75 Turkish Sign Language
19 Icelandic 1857 38 Danish 1926 57 Italian Sign Language 76 Ukrainian Sign Language
39 40
41 42
62 63
Figure 1: Contemporary and historical sign languages in our sample, with locations being de-
rived from Glottolog [19] for contemporary languages and from city of publication for
historical manual alphabets.
are good reasons to begin with a comparison of MAs instead. First, because the creation and
transmission of many MAs are well-documented, the data provide a relatively well-understood
test case for studying evolutionary processes in SLs. There exist far fewer historical dictionaries
of the world’s SLs compared to historical examples of MAs, though some do exist [29–31]. Sec-
ond, while computer-readable transcription systems have been developed for SLs—HamNoSys
[25] and SignWriting [32] are the most commonly used—these are still not widely used in SL
lexicography (see [33] for a corpus-based dictionary including transcriptions in HamNoSys; see
[34] for a dictionary with transcriptions in SignWriting), and transcriptions in the two systems
are not straightforwardly comparable in all respects. Open, computer-readable, cross-linguistic
comparative data sets of SL vocabulary do not yet exist, due in part to the lack of consensus on
transcription system, but also to the time-consuming nature of SL transcription. Transcribing
handshapes in MAs instead of lexical signs—which typically include specications for orienta-
tion, location, and movement in addition to handshape—signicantly reduces the time necessary
to create a large cross-linguistic comparative data set.
An MA should not be understood as equivalent to a language’s phonological inventory.
Though features of an MA overlap with phonological features found in the lexicon, the two sets
of features are not co-extensive [35]. Further, the linguistic status of MAs and their relationships
to signed and written language have been the subject of debate. In some SLs ngerspelling—the
representation of a written word with a sequence of MA handshapes—is used frequently in every-
day discourse [36], and young children acquire MA handshapes and ngerspelling patterns before
they are able to read [37, 38]. However, usage of ngerspelling varies both cross-linguistically
Power et al. Sign Language Dispersal June 2019
[39] and across signers within the same signing community [40], which may have consequences
for evolutionary processes and mutation rates. Notwithstanding this variation in MA usage,
we show here that many of the historically-attested extra-linguistic connections amongst SLs
are represented clearly in the network analyses in Section 3. We take this as conrmation that
coherent, historically-relevant information is recoverable from a data set consisting of MAs.
2.2 Character Coding
We considered only handshapes and movements for determining similarity of MA forms, as
these are represented most consistently in both contemporary and historical sources. In coding
similarity across MAs representing dierent types of alphabets—for example, Latin and Cyrillic
alphabets—it is possible to consider both the form of the grapheme and the sound represented
by the grapheme as bases for organising the comparison. For example, the graphemic forms of
Latin B and Cyrillic Вare similar, but they represent dierent sounds: the voiced bilabial stop
(IPA [b]) for Latin and the voiced labiodental fricative (IPA [v]) for Cyrillic. Thus, a decision
must be made about whether to compare handshapes representing Cyrillic Вwith handshapes
for Latin B or V. In many such cross-alphabet comparisons, the documentary record is suggestive
about how the comparisons should be organised. In the example just mentioned, we compared
handshapes representing Cyrillic Вwith Latin B and not V because historical records provide
clues about how the Russian SL MA was adapted from other European MAs in the early 19th
century (see Section 4). When graphemic forms across two alphabet types are similar, as in
the case of Latin and Cyrillic just mentioned, the form of the grapheme typically becomes the
basis of comparison. However, there are many graphemes that dier in form across Latin-based,
Cyrillic-based, Greek, and Arabic-based alphabets, all of which are represented by MAs in our
sample. In such cases, the sound represented by the grapheme is the only basis for comparison.
Character mapping on selected networks can help to indicate how the coding of individual
concepts contributes to the overall dierentiation pattern (see SI 4.2).
In Figure 2, we exemplify our coding approach and the resulting binary matrices for use in
phylogenetic methods. For reasons of space we limit the number of taxa in the example to
those necessary for detailing our methods, and provide a second, more complex example in the
electronic supplementary material (SI 3.2). The left side of Figure 2 depicts handshape forms
representing four dierent graphemes, three of which represent the voiced velar stop (IPA [ɡ]):
Latin g, Cyrillic г, and Persian/Urdu ; and one for the voiced velar fricative (IPA [ɣ]): Greek
γ. Starting at the top of the gure, Afghan SL represents the grapheme by extending the
index and middle ngers, a form that is unique in this comparison. To represent the grapheme
g, historical Brazilian SL (Brazilian 1875) and French SL use forms with an extended index
nger and the thumb orientated in a similar direction. Pakistan SL uses a similar handshape
for the grapheme . Finally, Greek SL, Russian SL, and historical Russian SL (Russian 1835)
represent γand гwith extended index nger and thumb extended outward.
In the upper right side of Figure 2, we show how these handshapes were coded in the EDIC-
TOR tool [26]. The value in the “Concept” column enables comparisons across alphabets. Mor-
phological similarity was coded by assigning the same arbitrary numerical value in the “Cogid”
column for languages with similar forms. The column “Narrow concept” tracks the grapheme
represented in each MA, and “Year” indicates when historical sources were published, an empty
cell in the “Year” column reecting that the source is contemporary. Afghan SL, not being
similar with any other taxon, was assigned cogid 328. We coded the handshapes in Brazilian
1875, French SL, and Pakistan SL as similar and assigned them cogid 60. Handshapes in Greek
SL, Russian SL, and Russian 1835 were coded as similar and assigned cogid 4. The bottom of
Figure 2 shows how our coding translates to the binary matrix for use in phylogenetic methods.
Taking the character ID from the “Cogid” column, Brazilian 1875, French SL, and Pakistan SL
Power et al. Sign Language Dispersal June 2019
Figure 2: Simplied coding example for handshapes representing Latin g, Cyrillic г, Greek γ,
and Persian/Urdu (spatial orientations similar to those in the sources).
were scored as 1 for ID 60, while the other four taxa were scored as 0. Greek SL, Russian 1835,
and Russian SL were scored as 1 for ID 4, and the other taxa 0. Afghan SL was scored as 1 for
ID 328, while the other taxa were scored as 0.
2.3 Phylogenetic analysis
The matrix includes characters supporting language cliques that are compatible (inherited pat-
terns) and incompatible (horizontally-propagated patterns) with the splits in the unknown true
tree. The true tree is, in this case, not necessarily a sequence of dichotomous splits because the
tree can be anastomising: (i) one ancestor may have more than two direct descendants; and (ii) a
descendant may have more than one direct ancestor in the case that, for example, an MA is the
product of combining two (or more) dierent sources. Thus, any simple, dichotomising tree that
we infer or select using tree-reconstruction methods and commonly-used optimality criteria will
be incomprehensive, and any signal in the matrix reecting aspects not covered by the tree’s se-
lected or inferred topology will add to data incompatibility. Another source of tree-incompatible
signal is the inclusion of putative “ancestors” in the form of historical MAs, as well as their di-
rect or distant “descendants”. Spencer et al. [41] showed that the distance-based Neighbor-Nets
(NNets; [42]), which were designed to counter the problem of signal incompatibility, outperform
tree inferences when it comes to correctly depicting ancestor-descendant relationships.
With respect to the complex signal in the underlying matrices, we thus relied exclusively on
network-based EDA [24, 43] using planar (2-dimensional), distance-based (NNets), and multidi-
mensional, tree-sample-based splits graphs (Support Consensus Networks [44, 45], CNets). We
Power et al. Sign Language Dispersal June 2019
used PAUP* [46] to compute simple (Hamming) pairwise distances and establish non-parametric
bootstrapping (BS) branch support under the Least-Squares (LS) and Maximum Parsimony
(MP) optimality criteria. BS analysis used 10,000 pseudo-replicates; replicate trees were in-
ferred using the BioNJ algorithm [47] for LS and quick-and-dirty BS for MP as outlined by
Müller [48] (‘MulTrees’ option deactivated; only one tree saved per replicate). For Maximum
Likelihood (ML) BS support, we used 10,000 replicate trees generated with RAxML v.8.0.20
[49] and the standard model for binary data allowing for site variation modeled via the Gamma
function, and corrected for ascertainment bias (recommended setting for binary data without
invariable sites; the eect has only been tested for phylogenomic binary data; hence, we also ran
the same analysis without correcting for ascertainment bias). Splits graphs were inferred with
SplitsTree v.4.13.1 [50]. All analysis les are included in the electronic supplementary material.
Post-analysis character mapping was done by hand-and-eye following the logical framework of
median networks and guidelines provided by Bandelt et al. [51] for their manual reconstruction.
In contrast to a dichotomised and/or anastomised tree, a median network considers taxa to be
either tips or medians, representing ancestral variants connecting the tips. A full median network
includes all possible most-parsimonious solutions for the mutation of a character, character
complex, or data matrix. For this study, we establish the minimum amount of necessary changes
in each set of binary sequences representing concepts found in the standard Latin alphabet
(letters ato z) along time-ltered networks (SI 4.2).
3 Results
The NNet in Figure 3 allows dening eight main groups of diering coherence and uniqueness.
Each group forms a neighbourhood in the graph dened by a single, more or less prominent,
edge-bundle. Three of the groups collect SLs of (i) Austrian-, (ii) French-, and (iii) British-
origin; the oldest SLs in the rst two of these groups (Austrian 1786, French 1800) may reect
the older common bases from which the SLs in these groups are derived. The largely extinct
Austrian-origin group includes a single surviving contemporary SL, Icelandic SL. Most other
contemporary SLs (e.g., Austrian, Danish, and German SLs) of the Austrian-origin group are
now found in the French-origin group, which includes the International Sign MA. In addition, we
recognise (iv) an Afghan-Jordanian group, with lowest overall dissimilarity to the British-origin
group; (v) a Polish group that is connected with (vi) the Russian group via Latvian SL; (vii) a
distinct Spanish group including the oldest MAs in our data set (Yebra 1593, Bonet 1620); and
(viii) the very unique Swedish group which includes Portuguese SL. The spiderweb structure
of the centre of the NNet graph indicates the data cannot resolve the principal relationships
between the distinguished eight main groups.
The robustness of discriminating signal in the underlying matrix for each main group, es-
timated using non-parametric BS support, is shown in Table 1. In general, highly coherent,
distinct groups (Afghan-Jordanian, British-origin, Polish, Spanish, Swedish groups) received
moderate to high support (BS >48; usually >90 for at least two optimality criteria) irre-
spective of the optimality criterion used. Less coherent groups (Austrian-origin, French-origin,
Russian groups) received low support (BS <42). This demonstrates that the distance matrix
well reects the overall diversity patterns. Inter-group relationships are essentially unresolved:
best-supported alternatives have a BS 23. Ambiguous BS support (i.e., BS << 100) can result
from a lack of discriminating signal or internal signal conict, which can be explored using CNets
(see SI 4.3; electronic supplementary material). In the case of the low-supported Austrian- and
French-origin groups, no alternative nds a BS 15; these groups are poorly supported but
lack alternatives. The same holds for the much higher BS support of the Spanish group. In the
case of the Russian group, the low support relates to competing alternatives: the data prefer
Power et al. Sign Language Dispersal June 2019
Pairwise distance = 0.01
Albanian SL
Greek SL
Irish SL
French 1800
French 1803
Italian 1897
French 1856
Dutch 1790
Dutch 1820
French 1815
Austrian 1839
Pakistan SL
Quebec SL
Italian SL
Finnish SL
Danish SL
International Sign
Brazilian SL
French-Belgian SL
French SL
Mexican SL
Flemish SL
Dutch SL
Norwegian SL
Icelandic SL
Norwegian 1955
Polish 1879
Icelandic 1857
German 1916
German 1909
Austrian 1823
German 1820
Afghan SL
Jordanian SL
New Zealand SL
British SL
Australian SL
Indian SL
Turkish SL
Czech SL
Croatian SL
Polish SL
Lithuanian SL
Latvian SL
Estonian SL
Estonian 1988
Ukrainian SL
Bulgarian SL
Russian SL
Russian 1835
Bonet 1620
Yebra 1593
Catalan SL
Spanish SL
Portuguese SL
Swedish SL
Swedish 1866
Austrian 1786
Brazilian 1875
Hungary 1827
American SL
Austrian SL
German SL
before 1800
after 1800
after 1950
current SL
Used/ invented
Figure 3: Neighbour-net based on simple (Hamming) pairwise distances calculated from the
standard-coded Cogid binary matrix. Colours highlight the main groups and the Dan-
ish subgroup (cf. Figure 4; SI 4.2) within the Austrian-origin lineage. Neighbourhood-
dening edge-bundles are also highlighted.
Power et al. Sign Language Dispersal June 2019
Table 1: Non-parametric bootstrapping (BS) support for neighbourhood-dening splits of main
groups (highlighted in Figure 3). ML = maximum likelihood; ASC = corrected for as-
certainment bias; UNC = uncorrected for ascertainment bias; NJ = neighbour-joining;
P = parsimony.
SL group ASC UNC
Austrian-origin <15 <15 24 21
British-origin 99 99.5 99.8 91
French-origin <15 <15 42 16
Afghan-Jordanian 99.9 99.6 66 98
Russiana18 17 40 39
Polisha93 90 99 98
Spanish 84 85 71 49
Swedish 99 99 97 98
No alternative found with BS15
aNot including Latvian SL (see SI 4.3)
and would support partly incongruent tree-topologies (SI 4.3). The two major sources of signal
conict are (i) Latvian SL, which is substantially less dissimilar to the Polish group than all
SLs of the Russian group (BSNJ = 37, but BSML,MP < 15), hence, its intermediate placement
in the NNet (Figure 3); and (ii) the Russian 1835 MA. In this case, the BS support values can
vary substantially between optimality criteria: the distance-based NJ vs. the character-based,
mutation-probability naive MP vs. the character- and model-based ML. Within the Russian
group, Russian 1835 is most-closely related to contemporary Cyrillic-representing MAs, while
diering from Estonian SL, Estonian 1988, and Latvian SL. In the planar NNet, this conict is
resolved by placing Russian 1835 on the opposite side of Estonian and Latvian SLs, while the
n-dimensional CNets show according 3-dimensional boxes (when using a cut-o of BS 15).
Figure 4 shows stacked NNets including SLs from specic time periods: the lowest NNet in
the gure with SLs up to 1840; the middle from the mid-19th to the mid-20th century; and the
uppermost NNet including contemporary SLs. The bottom graph demonstrates the substantial
diversity amongst MAs by the early 19th century, with three main clusters: Austrian-origin,
French-origin, and Spanish. The Austrian-origin group diversied further during the second half
of the 19th century (middle graph), while the French SL MA was dispersed largely unmodied to
the Americas. The Russian group is closest to the Austrian-origin group (the early Austrian SL
MAs from 1786 and 1823, as well as their close relatives, the early German 1820 and Hungarian
1827 MAs) and most distant to the French-origin group. The third main distinct cluster in the
bottom graph is the Spanish group. The Swedish group, appearing rst in the middle graph, is
already unique in the second half of the 19th century. In the topmost graph, the overall picture
remains the same, with a few exceptions. First, Polish SL, which is found in the Austrian-origin
group in the middle graph, is positioned between the Austrian-origin and Russian groups and
forms a cluster with Lithuanian SL. Second, contemporary Norwegian SL separates from the
middle of the graph and is no longer grouped closely with Danish or Icelandic SLs. Third,
Austrian, Danish, and German SLs, earlier examples of which were found in the Austrian-origin
group, are grouped closely with the International Sign MA and with American SL.
The results of the EDA indicate that each major group goes back to an independent founding
event; hence, no to little support for the deepest splits, the spiderweb-like centre of the overall
NNet in Figure 3, and the three distinct clusters earliest SLs in Figure 4. Contemporary Spanish
Power et al. Sign Language Dispersal June 2019
Norwegian 1955
Danish 1967
Danish 1967
Icelandic 1857
Old French SG
Old French SG
Old French SG
French SG
American SG
American/Intl Sign SG
American/Intl Sign SG
Estonian 1988Estonian 1988
Brazilian 1875
1871 1808
1926 1967
Polish 1879
Italian 1897
Swedish 1866
French 1803
French 1800
Spanish 1815
Spanish 1815
Bonet 1620
Yebra 1593
Danish 1808 Austrian 1823 Hungary1827
Austrian 1839
Austrian 1786
Norwegian SL
Polish SL
Lithuanian SL
Latvian SL
Estonian SL
Ukranian SL
Bulgarian SL
Russian SL Albanian SL
Swedish SL
Portuguese SL
Greek SL
Irish SL
Dutch SL
Brazilian SL
French-Belgian SL
French SL
Mexican SL
Flemish SL
Pakistan SL
(3) Austrian SL
(4) German SL
(1) Quebec SL
(2) American SL
Italian SL
Finnish SL
Danish SL
International MA
Catalan SL
Spanish SL
Icelandic SL
German 1820
German 1820
French 1815
American 1821
French 1815
American 1821
Pairwise distance = 0.01
Figure 4: Time-/taxon-ltered stacked Neighbour-nets, based on the same distance matrix used
for Figure 3. Bottom NNet: SLs up to ca.1840; middle NNet: 1815–late 20th century;
top NNet: mid-/late 20th century–present. Abbr.: SG = potential subgroups within
the French-origin group.
Power et al. Sign Language Dispersal June 2019
and Catalan SLs are direct derivates of the oldest MAs in our dataset, Yebra 1593 and Bonet
1620, while the French- and Austrian-origin groups constitute the two main independent tradi-
tions in continental Europe. Slightly modied versions of the French SL MA were dispersed into
the Americas, with the American SL MA later forming the basis for the International Sign MA,
which had a homogenising eect on several European MAs. International Sign possibly aected
Norwegian SL and fully replaced the Austrian-origin handshapes in Danish, German, and Aus-
trian SLs. Standardisation also inuenced internal relationships within the French-origin group:
we observe a “taxonomic turnover” with the original SLs in Europe and the New World (closest
to French 1800) being replaced by versions very similar (“International Sign subgroup”) or more
similar (“contemporary French subgroup”) to American SL and International Sign than to the
French original, with the exception of contemporary Dutch (unique development), as well as
Greek and Irish SLs (still closest to original 18th/19th century French).
Contemporary Icelandic SL is a direct derivate from the Danish subgroup within the Austrian-
origin group; the same holds for Norwegian SL, which started to strongly deviate from the closely
related Icelandic SL in the second half of the 20th century. The Russian group can be linked
historically to the Austrian-origin group (see the bottom-most NNet in Figure 4) and underwent
substantial restructuring in the adaptation from representing a Latin-based alphabet to Cyrillic.
The Estonian and Latvian SL MAs are, to a lesser degree, 20th century derivates, with more links
to the contemporary Russian group than to the Russian MA from 1835 (see also SI 4.3). Swedish
constitutes an isolated, mainly unique tradition and is the basis of contemporary Swedish and
Portuguese SLs. Although we have not included any historical examples of either, the British-
origin and Afghan-Jordanian groups constitute isolated traditions that evolved independently
(British-origin) or largely independently (Afghan-Jordanian) of the European groups.
4 Discussion
Figure 5 shows ve European lineages and their hypothesised dispersal from the late 16th to the
late 19th century. While the earliest MAs in our sample, Yebra 1593 and Bonet 1620, are typically
identied as the ancestors of most one-handed MAs in the world today [36], we argue that
communities in Austria, France, and Spain independently formed MAs using the early sources.
We suggest that this independent formation supports identifying three separate SL lineages.
The Yebran and Bonetian MAs, together with other Spanish MAs in the sample (Spanish 1815,
1845, 1859, contemporary Spanish and Catalan SLs), constitute a Spanish lineage. The NNet
in Figure 3 gathers these MAs in a highly coherent group with moderate to high BS support
(Table 1) without alternatives (see CNets in the electronic supplementary material). Because
there existed no large-scale deaf educational institutions in Spain until the early 19th century,
it seems unlikely that the two early MAs were used widely prior to that period, particularly in
signing communities. Preserved as examples in books, the early MAs would not have evolved
due to processes implicated in their usage in a community, explaining the relative lack of change
in this lineage.
While the French MA was clearly formed using Spanish sources, we argue here that it did not
evolve directly from a shared Spanish-French origin. Independent formation of the French MA is
suggested by the clear separation of the French and Spanish groups in Figure 3 and of the earliest
French and Spanish MAs in Figure 4. The Paris Institute had been in existence for decades
before the school in Madrid was founded, making it unlikely that there could have been a common
Spanish-French basis from which the French MA evolved. Similarities between the earliest French
and Spanish MAs are explained by the fact that educators throughout Europe, including the
founder of deaf education in France, de l’Épée in Paris, had become aware of Bonet’s (1620)
Reduction de las Letras and its MA by the late 18th century [3, 52]. While the Spanish MA from
Power et al. Sign Language Dispersal June 2019
Figure 5: Hypothesised dispersal of European sign languages from late 16th to late 19th cen-
tury, based on results in section 3. Colour-coding reects ve hypothesised lineages:
Spanish, French-origin, Austrian-origin, British-origin, Swedish. Timeline reects ap-
proximate years of rst transmission of SLs, typically coinciding with establishment of
schools for the deaf or migrations of signers (except in the cases of 1 & 2, which track
publication of earliest manual alphabets in sample).
Power et al. Sign Language Dispersal June 2019
1815 shows little innovation compared to Yebra and Bonet, many dierences are observable
between the Yebran and Bonetian MAs and the earliest French MA in our sample (French
1800), in which just eleven forms remained unmodied and forms were added representing new
letters k,v, and w. In adapting the early Spanish sources, the signing community in Paris
changed them substantially, both consciously by innovating new forms and through usage in a
community, which resulted in only minimal modications to existing forms. It is clear that the
French MA drew on the original Spanish sources directly: where the original and later Spanish
forms diered, the French MA either kept the original form unmodied (e.g., in the handshapes
representing m,n) or modied the original and not the later Spanish form (e.g., in the handshape
representing q).
The other main continental European lineage is the Austrian-origin lineage. Founders of the
rst school for the deaf in Vienna, Joseph May and Friedrich Storch, visited de l’Épée and
the Paris Institute in 1777 to learn pedagogical methods and to subsequently establish deaf
education in the Habsburg Empire [10]. Perhaps because of these historical connections, it has
been thought that Austrian SL is related to French SL in an ancestor-descendant relationship
[15]. A large number of uniquely shared innovations (synapomorphies in biology) would support
such a conclusion, indicating a unied basis from which both MAs later diverged. Shared
innovations would be, in this case, those forms that diered from Spanish sources but that were
uniquely shared in French and Austrian MAs. In fact, we nd relatively few such potentially
derived forms between the earliest Austrian and French MAs, including those representing d,e,
l,r, and w. Forms for fand vwere shared by Austrian 1786 and French 1800, but also by early
unrelated SLs, such as Swedish 1866.
That these possible synapomorphies across Austrian and French MAs are too few is reected
in the lack of a French-Austrian neighborhood ‘trunk’ in the NNet of the earliest MAs in the
sample (Figure 4). If Austrian had indeed evolved from a French basis, the network should show
a prominent fan-like structure including both MAs, with the oldest MAs in the middle and the
most derived within each group as wings towards either side of the fan. The potential shared
innovations may be better characterised as early borrowings due to minimal contact between
French and Austrian signing communities; or as convergent evolution, in which handshape forms
independently evolved in similar ways either to iconically match the forms of similar graphemes,
or because forms were selected for that conferred articulatory and perceptual advantages. Fi-
nally, compare the positions of the Austrian, French, and Spanish groups in Figure 3 to the
positions of other languages thought to have evolved from French SL. For example, we know
that deaf French educators helped establish deaf education in the US [53], Brazil [54], and Mex-
ico [55]; and those MAs remain topologically close to contemporary French SL. In Figure 4,
the earliest American SL MA in the sample (American 1821) has diverged minimally from the
chronologically closest French MA from 1815. In contrast, early Austrian, French, and Spanish
MAs are found in diering neighbourhoods.
The results support classifying Danish as a sublineage of the Austrian-origin lineage. Peter
Atke Castberg, who founded the rst school for the deaf in Copenhagen in 1807, visited deaf ed-
ucational institutions in Germany, France, and Austria between 1802–5 [56]. Forms representing
c,g,h,o,p, and qin the earliest Danish MA from 1808 are shared with Austrian but not French.
In contrast, none of the forms in the earliest Danish MA from 1808 are unambiguously French
in origin. These patterns are reected in the bottom graph in the NNet in Figure 4; Danish
1808 is topologically closest to the early Austrian-origin group languages. The early Danish MA
shows several innovated handshapes, such as those representing d,f,k,s,u,v,w,y, and z, as
well as new forms for the Danish letters æand ø. Thus, there is some support for classifying
Danish as a separate lineage based on similar argumentation to that used above. In contrast
to the Austrian, French, and Spanish cases above, however, early Danish and early Austrian
Power et al. Sign Language Dispersal June 2019
MAs are consistently found in close topological proximity in the network analyses under various
methods and optimality criteria.
While the results support classifying Russian as a separate lineage created using Austrian
sources, any interpretation of the results with respect to Russian SL is complicated by substantial
adaptation and restructuring of an existing MA to represent the Cyrillic alphabet. Historical
connections between Austrian and Russian communities likely developed when an Austrian-
trained educator, Father Sigmund, and a Frenchman, Jean-Baptiste Jauret, helped establish
the rst school for the deaf in St. Petersburg during the period from 1806–10 [57, 58]. Similarities
between Russian SL and early Austrian MAs are reected in the NNet of early MAs in Figure
4, which places Russian 1835 closest to the Austrian-origin group languages compared to other
groups. Because of mismatches between the two alphabets, however, the Austrian MA underwent
substantial restructuring to represent graphemes in Cyrillic. New forms were invented for some
Cyrillic letters not found in the Latin alphabet; other forms were dropped; and some forms
were used to represent Cyrillic letters that appear similar to Latin letters, but which occupy
dierent positions in the alphabet. One consequence of this mismatch between alphabets is that
there are fewer cross-alphabet comparanda between Latin- and Cyrillic-representing MAs. In
addition, as discussed in Section 2, there are serious methodological challenges in deciding which
forms should be compared, with the resulting possibility that some connections between sources
and adaptations cannot be recovered. Thus, dierences between Latin and Cyrillic alphabets
may cause phylogenetic methods to overestimate the distance between the Austrian-origin and
Russian groups.
The British-origin group forms an independent lineage, with links to the early 2-handed MA in
Digiti lingua published in 1698 [59]. The NNet in Figure 3 shows a clear split of this group from
the centre of the graph. All of the MAs in this lineage are predominantly 2-handed, and some
characters are shared throughout the group (e.g., handshapes representing dand x). While the
links between British SL and SLs in Commonwealth countries are relatively well-documented
[12], less is known about how the British-origin MA came to other SLs in the group, namely,
Czech, Croatian, and Turkish SLs. Zeshan [60, 46] reports the presence of British educators in
Turkey in the early 1950s. Both Kuhn et al. [61, 55] for Croatian SL and Hudáková [62, 30] for
Czech SL report that 1- and 2-handed MAs are in use in Croatia and the Czech Republic: in
Croatia, the 2-handed alphabet is thought to be older, while the opposite may be true in the
Czech Republic. Future research could uncover historical examples of MAs from these SLs that
can help to clarify their connections to the British-origin group. Similarly, lexical investigations
are likely to shed more light.
Finally, the Swedish group, which consists of only historical and contemporary Swedish SL,
as well as Portuguese SL, forms a separate lineage. Bergman and Engberg-Pedersen [56] suggest
that Per Aron Borg, the founder of the rst school for the deaf in Stockholm in 1809, though
aware of de l’Épée’s work in the Paris Institute, may not have been familiar with the MA in use
in France when he created the Swedish SL MA. There are some similarities between an early
example of the Swedish SL MA from 1866 and MAs used in France and other parts of Europe,
such as in handshapes representing c,f,k,l,m,n,o,u, and v. Thus, while the new Swedish SL
MA was created mainly in isolation from other lineages, Borg may have borrowed handshapes
known widely in Europe. The connection between Swedish SL and Portuguese SL can also be
traced to Borg, who helped establish deaf education in Lisbon in 1823 [56].
5 Conclusion
Despite their relevance to our understanding of human linguistic diversity and to theories of
language change, the evolutionary histories of the world’s SLs have not, until now, been studied
Power et al. Sign Language Dispersal June 2019
using state-of-the-art methods. We have shown that computational phylogenetic methods can be
applied to SL data to uncover new insights into the evolutionary histories of SLs, to generate new
hypotheses about their relationships, and to better understand the evolutionary processes that
have shaped the diversity of contemporary SLs. Our analysis supports some aspects of existing
SL classications, but adds complexity to the overall picture, in particular to our understanding
of the evolution of SLs from early sources. Our discussion of the independent establishment of
SL lineages points to a characterisation of similarities across lineages as primarily horizontal, and
not due to descent from a common ancestor, while within-lineage diversication does appear to
be characteristically vertical in many cases. We anticipate that future studies of lexical data may
contradict our phylogeny based on MAs, in particular for SLs that adopted the International
Sign MA, because this adoption did not likely aect a language’s lexicon to any great extent.
Notwithstanding these limitations, we suggest that our analysis be taken in future research as
the best available phylogenetic classication of these SLs.
JML was supported by the ERC Starting Grant 715618 Computer-Assisted Language Compar-
ison ( We thank Tiago Tresoldi for sharing initial ideas on this
project, Harald Hammarström for providing information on proposed sign language classica-
tions, and Russell D. Gray and Richard P. Meier for comments on an earlier version of this
Author Contributions
JMP and JML initiated the study. JMP and JML designed the database and developed methods
for data coding. JML programmed the database system. JMP coded the data. GWG carried
out the phylogenetic analysis and designed graphics. JMP and GWG interpreted the results.
JMP and GWG wrote the rst draft. All authors read the last draft and agree on its contents.
Supplementary Material
The supplementary material accompanying this paper provides additional information in form
of an appendix, as well as all data, code, and analyses needed to replicate or test our studies.
It can be downloaded from
[1] Margret A. Winzer. Education, urbanization, and the deaf community: A case study of Toronto,
1870-1900. In John Vickrey Van Cleve, editor, Deaf history unveiled: Interpretations from the new
scholarship, chapter 9, pages 127–145. Gallaudet University Press, Washington, D.C., 1993.
[2] Jordan Fenlon and Erin Wilkinson. Sign languages in the world. In Ceil Lucas and Adam C.
Schembri, editors, Sociolinguistics and deaf communities, pages 5–28. Cambridge University Press,
Cambridge, United Kingdom, 2015.
[3] Susan Plann. A silent minority: Deaf education in Spain, 1550-1835. University of California Press,
Berkeley, 1997.
[4] Bencie Woll, Rachel Sutton-Spence, and Frances Elton. Multilingualism: The global approach to
sign languages. In Ceil Lucas, editor, The sociolinguistics of sign languages. Cambridge University
Press, Cambridge, United Kingdom, 2001.
Power et al. Sign Language Dispersal June 2019
[5] Harlan Lane. The wild boy of Aveyron. Harvard University Press, Cambridge, Massachusetts, 1976.
[6] Goedele De Clerck. The Flemish deaf community and the challenges of breaking through barriers in
the educational system. In Donald F. Moores and Margery S. Miller, editors, Deaf people around the
world: Educational and social perspectives, pages 157–177. Gallaudet University Press, Washington,
D.C., 2009.
[7] Penny Boyes Braem and Christian Rathmann. Transmission of sign languages in Northern Europe.
In Diane Brentari, editor, Sign languages, pages 19–45. Cambridge University Press, New York,
[8] John Tabak. Signicant gestures: A history of American Sign Language. Praeger Publishers,
Westport, Connecticut, 2006.
[9] Carlo Geraci. Italian Sign Language. In Julie Bakken Jepsen, Goedele De Clerck, Sam Lutalo-Kiingi,
and William B McGregor, editors, Sign languages of the world: A comparative handbook, chapter
III,I,19, pages 473–510. Walter de Gruyter, Inc. and Ishara Press, Berlin, Germany and Preston,
UK, 2015.
[10] Katarina Schalber. Austrian Sign Language. In Julie Bakken Jepsen, Goedele De Clerck, Sam Lutalo-
Kiingi, and William B McGregor, editors, Sign languages of the world: A comparative handbook,
chapter III,I,3, pages 105–128. Walter de Gruyter, Inc. and Ishara Press, Berlin, Germany and
Preston, UK, 2015.
[11] William B. McGregor, Janne Boye Niemelä, and Julie Bakken Jepsen. Danish Sign Language. In
Julie Bakken Jepsen, Goedele De Clerck, Sam Lutalo-Kiingi, and William B McGregor, editors, Sign
languages of the world: A comparative handbook, chapter III,I,7, pages 195–233. Walter de Gruyter,
Inc. and Ishara Press, Berlin, Germany and Preston, UK, 2015.
[12] Adam Schembri, Kearsy Cormier, Trevor Johnston, David McKee, Rachel McKee, and Bencie Woll.
Sociolinguistic variation in British, Australian and New Zealand Sign Languages. In Diane Brentari,
editor, Sign languages, pages 476–498. Cambridge University Press, New York, 2010.
[13] Harlan Lane. When the mind hears: A history of the deaf. Random House, New York, 1984.
[14] Per Eriksson. The history of deaf people: A source book. Daufr., Örebro, Sweden, 1998.
[15] Lloyd Anderson. A comparison of some American, British, Australian and Swedish signs: Evidence
on historical changes in signs and some family relationships of sign languages. First International
Symposium on Sign Language, 1979.
[16] Shi Yu, Carlo Geraci, and Natasha Abner. Sign languages and the online world online dictionaries &
lexicostatistics. In Nicoletta Calzolari (Conference chair), Khalid Choukri, Christopher Cieri, Thierry
Declerck, Sara Goggi, Koiti Hasida, Hitoshi Isahara, Bente Maegaard, Joseph Mariani, Hélène Mazo,
Asuncion Moreno, Jan Odijk, Stelios Piperidis, and Takenobu Tokunaga, editors, Proceedings of the
Eleventh International Conference on Language Resources and Evaluation (LREC 2018), Miyazaki,
Japan, May 7-12, 2018 2018. European Language Resources Association (ELRA).
[17] Lyle Campbell. Historical linguistics. Edinburgh University Press, Edinburgh, 3 edition, 2013.
[18] Ethnologue: Languages of the world, 2019.
[19] Harald Hammarström, Robert Forkel, and Martin Haspelmath. Glottolog, 2018.
[20] James Woodward. Some observations on research methodology in lexicostatistical studies of sign
languages. In Gaurav Mathur and Donna Jo Napoli, editors, Deaf around the world: The impact of
language, chapter 1, pages 38–53. Oxford University Press, Oxford, 2011.
[21] Russel R. Aldersson and Lisa J. McEntee-Atalianis. A lexical comparison of signs from Icelandic
and Danish Sign Languages. Sign Language Studies, 9(1):45–87, 2008.
Power et al. Sign Language Dispersal June 2019
[22] Vishnupriya Kolipakam, Fiona M. Jordan, Michael Dunn, Simon J. Greenhill, Remco Bouckaert,
Russell D. Gray, and Annemarie Verkerk. A Bayesian phylogenetic study of the Dravidian language
family. Royal Society open science, 5(3), 2018.
[23] Laurent Sagart, Guillaume Jacques, Yunfan Lai, Robin Ryder, Valentin Thouzeau, Simon J. Green-
hill, and Johann-Mattis List. Dated language phylogenies shed light on the ancestry of Sino-Tibetan.
Proceedings of the National Academy of Science of the United States of America, 116:1–6, 2019.
[24] David A. Morrison. Using data-display networks for exploratory data analysis in phylogenetic studies.
Molecular Biology and Evolution, 27:1044–1057, 2010.
[25] Thomas Hanke. Hamnosys-representing sign language data in language resources and language
processing contexts. In LREC, volume 4, pages 1–6, 2004.
[26] Johann-Mattis List. A web-based interactive tool for creating, inspecting, editing, and publishing
etymological datasets. In Proceedings of the 15th Conference of the European Chapter of the Associa-
tion for Computational Linguistics. System Demonstrations, pages 9–12, Valencia, 2017. Association
for Computational Linguistics.
[27] Johann-Mattis List, Simon Greenhill, Tiago Tresoldi, and Robert Forkel. LingPy. A Python library
for quantitative tasks in historical linguistics. Max Planck Institute for the Science of Human History,
Jena, 2018.
[28] Robert Forkel, Johann-Mattis List, Simon J. Greenhill, Christoph Rzymski, Sebastian Bank, Michael
Cysouw, Harald Hammarström, Martin Haspelmath, Gereon A. Kaiping, and Russell D. Gray. Cross-
linguistic data formats, advancing data sharing and re-use in comparative linguistics. Scientic Data,
5(180205):1–10, 2018.
[29] Don Francisco Fernandez Villabrille. Diccionario usual de mímica y dactilología: útil á los maestros
de sordo-mudos, á sus padres y á todas las personas que tengan que entrar en comunicación con
ellos. Imprenta del Colegio de Sordo-Mudos y Ciegos, Madrid, 1851.
[30] Pierre Pelissier. Iconographie des signes. Paul Dupont, Paris, 1856.
[31] Andreas C. Nyegaard. De Døvstummes Haandalphabet samt et Udvalg af deres lettere Tegn sam-
menstillet, tegnet, graveret og udgivet af en Forening af Dövstumme. Kongelige Dövstumme-Institut,
Copenhagen, 1871.
[32] Valerie Sutton. The SignWriting alphabet: The International SignWriting Alphabet 2010, ISWA
2010. The SignWriting Press, La Jolla, 2011. Online; accessed 12-April-2019.
[33] Joanna Łacheta, Małgorzata Czajkowska-Kisil, Jadwiga Linde-Usiekniewicz, and Paweł Rutkowski,
editors. Korpusowy słownik Polskiego Języka Migowego. Faculty of Polish Studies, University of
Warsaw, Warsaw, 2016. Online; accessed 28-March-2019.
[34] Bruno Sonnemans. LSFB asbl: An online dictionary and newspaper in LSFB free access of videos,
2016. Online; accessed 28-March-2019.
[35] Diane Brentari and Carol A. Padden. Native and foreign vocabulary in American Sign Language:
A lexicon with multiple origins. In Diane Brentari, editor, Foreign vocabulary in sign languages:
A cross-linguistic investigation of word formation, chapter 3, pages 87–119. Lawrence Erlbaum
Associates, Publishers, Mahwah, New Jersey, 2001.
[36] Carol A. Padden and Darline Clark Gunsauls. How the alphabet came to be used in a sign language.
Sign Language Studies, pages 10–33, 2003.
[37] Carol Tane Akamatsu. The acquisition of ngerspelling in pre-school children. PhD thesis, University
of Rochester, Rochester, New York, 1983.
Power et al. Sign Language Dispersal June 2019
[38] Carol A. Padden. Learning to ngerspell twice: Young signing children’s acquisition of ngerspelling.
In Brenda Schick, Marc Marschark, and Patricia Elizabeth Spencer, editors, Advances in the sign
language development of deaf children, chapter 8, pages 189–201. Oxford University Press, Oxford,
[39] Rachel Sutton-Spence, Bencie Woll, and Lorna Allsop. Variation and recent change in ngerspelling
in British Sign Language. Language Variation and Change, 2(3):313–330, 1990.
[40] Carol A. Padden. The acquisition of ngerspelling by deaf children. In Patricia Siple and Susan D.
Fischer, editors, Theoretical issues in sign language research, volume 2 of Psychology, chapter 10,
pages 191–210. University of Chicago Press, Chicago, 1991.
[41] Matthew Spencer, E. A. Davidson, A. C. Barbrook, and C. J. Howe. Phylogenetics of articial
manuscripts. Journal of Theoretical Biology, 227:503–511, 2004.
[42] Daniel Bryant and Vincent Moulton. Neighbor-net: An agglomerative method for the construction
of phylogenetic networks. Molecular Biology and Evolution, 21:255–265, 2004.
[43] John W. Tukey. Exploratory data analysis. Addison-Wesley, Reading, MA, 1977.
[44] Barbera Holland and Vincent Moulton. Consensus networks: A method for visualising incompati-
bilities in collections of trees, page 165–176. Springer Verlag, Berlin, Heidelberg, Stuttgart, 2003.
[45] Klaus Schliep, Alastair J. Potts, David A. Morrison, and Guido W. Grimm. Intertwining phylogenetic
trees and networks. Methods in Ecology and Evolution, 2017.
[46] David L. Swoord. Paup*: Phylogenetic analysis using parsimony (and other methods), 2002.
[47] O. Gascuel. BIONJ: An improved version of the NJ algorithm based on a simple model of sequence
data. Molecular Biology and Evolution, 14:685–695, 1997.
[48] Kai F. Müller. The eciency of dierent search strategies for estimating parsimony, jackknife,
bootstrap, and Bremer support. BMC Evolutionary Biology, 5:58, 2005.
[49] Alexandros Stamatakis. RAxML version 8: A tool for phylogenetic analysis and post-analysis of
large phylogenies. Bioinformatics, 30:1312–1313, 2014.
[50] Daniel H. Huson and Daniel Bryant. Application of phylogenetic networks in evolutionary studies.
Molecular Biology and Evolution, 23:254–267, 2006.
[51] Hans-Jürgen Bandelt, Vincent Macaulay, and Martin Richards. Median networks: Speedy construc-
tion and greedy reduction, one simulation, and two case studies from human mtDNA. Molecular
Phylogenetics and Evolution, 16:8–28, 2000.
[52] Charles Michel de L’Epée. La véritable maniere d’instruire les sourds et muets: Conrmée par une
longue expérience. Nyon l’aîné, Paris, 1 edition, 1784.
[53] Carol Neidle and Joan Cottle Poole Nash. American Sign Language. In Julie Bakken Jepsen, Goedele
De Clerck, Sam Lutalo-Kiingi, and William B McGregor, editors, Sign languages of the world: A
comparative handbook, chapter III,I,1, pages 31–70. Walter de Gruyter, Inc. and Ishara Press, Berlin,
Germany and Preston, UK, 2015.
[54] André Nogueira Xavier and Regiane Pinheiro Agrella. Brazilian Sign Language (Libras). In
Julie Bakken Jepsen, Goedele De Clerck, Sam Lutalo-Kiingi, and William B McGregor, editors,
Sign languages of the world: A comparative handbook, chapter III,I,4, pages 129–158. Walter de
Gruyter, Inc. and Ishara Press, Berlin, Germany and Preston, UK, 2015.
[55] David Quinto-Pozos. Sign language contact and interference: ASL and LSM. Language in Society,
37(2):161–189, 2008.
Power et al. Sign Language Dispersal June 2019
[56] Brita Bergman and Elisabeth Engberg-Pedersen. Transmission of sign languages in the Nordic
countries. In Diane Brentari, editor, Sign languages, Cambridge language surveys, chapter 1.4,
pages 74–94. Cambridge University Press, Cambridge, 2010.
[57] Igor A. Abramov. History of the deaf in Russia. In Renate Fischer and Harlan Lane, editors,
Looking back: A reader on the history of Deaf communities and their sign languages, volume 20,
pages 199–205. Signum-Verlag, Hamburg, 1993.
[58] Howard G. Williams and Polina Fyodorova. The origins of the St. Petersburg institute for the
deaf. In Renate Fischer and Harlan Lane, editors, Looking back: A reader on the history of Deaf
communities and their sign languages, volume 20, pages 295–305. Signum-Verlag, Hamburg, 1993.
[59] Author unknown. Digiti-lingua, or, The most compendious, copious, facil, and secret way of silent
converse ever yet discovered: Shewing, how any two persons may be capable, in half an hours time,
to discourse together by their ngers only, and as well in the dark as the light / by a person who has
conversed no otherwise in above nine years. P. Buck, London, 1698.
[60] Ulrike Zeshan. Aspects of Türk Isaret Dili (Turkish Sign Language). Sign Language & Linguistics,
6(1):43–75, 2003.
[61] Ninoslava Šarac Kuhn, Tamara Alibašić Ciciliani, and Ronnie B. Wilbur. Phonological parameters
in Croatian Sign Language. Sign Language & Linguistics, 9(1):33–70, 2006.
[62] Andrea Hudáková. Prstová abeceda pro tlumočníky. Česká komora tlumočníkŭ znakového jazyka,
Supplementary Information for the Paper
“Evolutionary Dynamics in the Dispersal of
Sign Languages”
Justin M. Power, Guido W. Grimm, and Johann-Mattis List
June 2019
1 Organisation of the supplementary material 2
2 Language Selection 2
2.1 ExtantManualAlphabets .................................. 2
2.2 Historical Manual Alphabets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
3 Data Preparation and Curation 4
3.1 TechnicalAspects ...................................... 4
3.2 Coding............................................ 4
4 Phylogenetic Analysis 5
4.1 Reasons for exploratory data analysis and e󰋳ects of incompatible signal . . . . . . . . . . 5
4.2 CharacterMapping ..................................... 6
4.2.1 Comprehensive character map for sign languages up to 1840 . . . . . . . . . . . . 8
4.2.2 Character map for sign languages from mid-19th to mid-/late 20th century . . . . . 9
4.2.3 Character map for contemporary sign languages . . . . . . . . . . . . . . . . . . 10
4.3 SplitSupport......................................... 10
Power et al. Sign Language Dispersal June 2019
Power et al. Dispersal of Sign Languages (Supplement) 2
1 Organisation of the supplementary material
The data was curated with help of the EDICTOR (List 2017). We used a server-based version to ease collab-
oration. A link to the database can be found at
Since the database curation process was in 󰋵ux for some time, and may change in the future, we pro-
vide a 󰋴nal stable dump of this database. The data itself is curated on GitHub (https://github.
com/lexibank/powerma), while the versions underlying this draft along with the results and the code
needed to convert the data into the formats required by the software packages we used can be found on
GitHub (
2 Language Selection
2.1 Extant Manual Alphabets
The table below shows the extant manual alphabets and sources in our sample, with IDs taken from Figures
1 and 5 in main text.
ID Manual Alphabet Abbrev Glottolog Source
23 Afghan Sign Language ZEA afgh1239 Afghan Sign Language 2001
24 Albanian Sign Language AlbSL alba1271
27 American Sign Language ASL amer1248 Tennant and Brown 1998, Lydell 2018
18 Australian Sign Language Auslan aust1271 Johnston 2014
30 Austrian Sign Language ÖGS aust1252 Lydell 2018
31 Brazilian Sign Language LSB braz1236 Lydell 2018
4 British Sign Language BSL brit1235 Brien 1992, Lydell 2018
32 Bulgarian Sign Language BZhE bulg1240 Lydell 2018
33 Catalan Sign Language LSC cata1287 Perelló and Masclans 1998
34 Croatian Sign Language HZJ croa1242 Kuhn et al 2006, Lydell 2018
35 Czech Sign Language CzSL czec1253 Hudáková 2008
40 Danish Sign Language DTS dani1246
70 Dutch Sign Language NGT dutc1253 Zwitserlood 2010
42 Estonian Sign Language EVK esto1238
43 Finnish Sign Language FinSL 󰋴nn1310 Kuurojen Liitto 1998
44 Flemish Sign Language VGT vlaa1235 Vertaal 2012
48 French Sign Language LSF fren1243 Lydell 2018
49 French-Belgian Sign Language LSFB lang1248,
52 German Sign Language DGS germ1281 Lydel 2018
53 Greek Sign Language GSL gree1271 Hatzopoulou 2008, Lydell 2018
54 Icelandic Sign Language ÍTM icel1236 Lydell 2018
22 Indian Sign Language IPSL indi1237 Lydell 2018
55 International Sign IS inte1259 Rubino et al 1975, Lydell 2018
56 Irish Sign Language ISL iris1235 Learn Irish Sign Language
57 Italian Sign Language LIS ital1275 Magarotto 1996, Lydell 2018
58 Jordanian Sign Language LIU jord1238 Hendriks 2008
59 Latvian Sign Language LSL latv1245, Lydell
60 Lithuanian Sign Language LGK lith1236, Lydell 2018
61 Mexican Sign Language LSM mexi1237 Lydell 2018
20 New Zealand Sign Language NZSL newz1236 McKee et al 2011
Power et al. Sign Language Dispersal June 2019
Power et al. Dispersal of Sign Languages (Supplement) 3
64 Norwegian Sign Language NTS norw1261 Tegnordbook 2016
65 Pakistan Sign Language PSL paki1242 Sabir Khan et al 2014, Lydell 2018
66 Polish Sign Language PJM poli1259 Łacheta et al 2016, Lydell 2018
16 Portuguese Sign Language LGP port1277 Lydell 2018
67 Quebec Sign Language LSQ queb1245
68 Russian Sign Language RSL russ1255 Lydell 2018
73 Spanish Sign Language LSE span1263 Blanco 2009
74 Swedish Sign Language STS swed1236 Svenskt teckenspråkslexikon 2014
75 Turkish Sign Language TİD turk1288 Zeshan 2003
76 Ukrainian Sign Language USL ukra1235 Lydell 2018
2.2 Historical Manual Alphabets
The following table shows the historical manual alphabets and sources in our sample, with IDs taken from
Figures 1 and 5 in main text, and abbreviations taken from the table in in Section 2.1.
ID Abbrev Year Location Source
14 ASL 1821 New York, USA Akerly 1821
25 ASL 1886 Washington D.C., USA Gordon 1886
26 ASL 1918 Iowa, USA Long 1918
12 DTS 1808 Copenhagen, Denmark Castberg 1818
36 DTS 1871 Copenhagen, Denmark Nyegaard 1871
37 DTS 1907 Copenhagen, Denmark Jorgensen 1907
38 DTS 1926 Copenhagen, Denmark Døvstumme 1926
39 DTS 1967 Copenhagen, Denmark Plum et al 1976
6 DGS 1821 Gmünd, Germany Alle 1821
50 DGS 1909 Leipzig, Germany Reuschert 1909
51 DGS 1916 Berlin, Germany Riemann 1916
41 EVK 1988 Tallinn, Estonia Toom 1988
19 ÍTM 1857 Akureyri, Iceland Sigurðsson 1857
3 LSF 1799-1800 Paris, France Unknown 1800
45 LSF 1803 Paris, France Sicard 1803
46 LSF 1815 Paris, France? de Ladebat 1815
47 LSF 1856 Paris, France Pelissier 1856
10 LSE 1815 Madrid, Spain Martí 1815
71 LSE 1845 Madrid, Spain Ballesteros and Villabrille 1845
72 LSE 1859 Madrid, Spain Carderera 1859
21 LSB 1875 Rio de Janeiro, Brazil da Gama 1875
7 LIS 1897 Milan, Italy Fornari 1897
9 MJ 1827 Vác, Hungary Schwarzer 1827
8 NGT 1790 Groningen, Netherlands Mörser and Guyot 1790
69 NGT 1820 Groningen, Netherlands van Heijningen Bosch 1820
17 NTS 1893 Oslo, Norway Svendsen 1893
62 NTS ca. 1900 Trondheim, Norway Bruun 1900
63 NTS 1955 Trondheim, Norway? Norske Døves Landsforbund 1955
5 ÖGS 1786 Vienna, Austria May 1789
Power et al. Sign Language Dispersal June 2019
Power et al. Dispersal of Sign Languages (Supplement) 4
28 ÖGS 1823 Vienna, Austria Venus 1823
29 ÖGS 1839 Vienna, Austria Czech 1839
15 PJM 1879 Warsaw, Poland Hollaka and Jagodzińskiego 1879
13 STS 1866 Stockholm, Sweden Paulsson 1866
11 RSL 1835 St. Petersburg, Russia Fleri 1835
1 Yebra 1593 Madrid, Spain de Yebra 1593
2 Bonet 1620 Madrid, Spain Bonet 1620
3 Data Preparation and Curation
3.1 Technical Aspects
To edit the data in a machine- and human-readable way, we used the EDICTOR application (List 2017),
which was originally designed for spoken languages and phonetic transcriptions. To annotate presumably
cognate handshapes, we used the “full cognate” annotation schema which essentially assumes that cognacy
is a transitive relation applying to a full form in a binary fashion (two forms are either cognate or not). To
export the data to the Nexus format, we made use of LingPy (List et al. 2018). The accompanying repository
with data-dump and source code shows how LingPy can be used for data conversion.
3.2 Coding
Simpli󰋴ed coding example for handshapes representing Latin h, n, and p, and Cyrillic нand п.
Coding is more complex when there is a mismatch across alphabet types of graphemic forms and sounds
represented. The 󰋴gure above exempli󰋴es our coding for ten MAs of handshapes representing Latin h,
n, and p, as well as Cyrillic нand п. Consider, 󰋴rst, the overlap in sounds represented across the two
alphabets. Latin h represents the voiceless glottal fricative (IPA [h]), a sound for which Cyrillic has no
corresponding letter; Latin n and Cyrillic нrepresent the voiced alveolar nasal (IPA [n]); and Latin p and
Cyrillic пrepresent the voiceless bilabial stop (IPA [p]). However, when considering the forms of the
graphemes, the two alphabets overlap in di󰋳erent ways: there are similarities between the forms of Latin h
and Cyrillic н, as well as Latin n and Cyrillic п.
The left side of the 󰋴gure shows our coding of handshapes representing the letters described above. We
coded the handshapes representing Latin h in contemporary Austrian SL, German SL, and American SL, as
Power et al. Sign Language Dispersal June 2019
Power et al. Dispersal of Sign Languages (Supplement) 5
well as historical American SL (American 1821), as similar and assigned them character ID 145 (in the col-
umn “Cogid”). Handshapes for Latin h in historical Austrian SL (Austrian 1823) and German SL (German
1820) were coded as similar and assigned ID 118. We chose to compare handshapes for Cyrillic нin his-
torical (Russian 1835) and contemporary Russian SL with the handshapes representing Latin h mentioned
above. Due to historical connections between Austrian-trained educators and the establishment of deaf edu-
cation in Russia in the early 19th century (Abramov 1993, Williams and Fyodorova 1993), we reasoned that
the similarity in handshape and graphemic forms for Latin h and Cyrillic нin Austrian 1823 and Russian
1835 are due to the adaptation of the Austrian h handshape in the Russian SL MA to represent Cyrillic н.
The alternative comparison, in which handshapes for Cyrillic нare compared with non-homologous hand-
shapes representing Latin n, is a possible but, in our view, incorrect approach based on the historical record.
Therefore, we coded the historical and contemporary Russian handshapes in the h-comparison, which can
be seen by observing the “Concept” and “Narrow concept” columns. The handshapes representing Cyrillic
нin Russian 1835 and contemporary Russian SL were coded as similar and assigned ID 299. Similarly,
we compared the handshape forms representing Latin n in Estonian SL and h in Latvian SL to the forms
in the h-comparison. That the handshapes in Estonian SL and Latvian SL are homologous to the Russian
SL forms for Cyrillic нis likely due to the history of deaf education in the former Soviet Union and the use
of Russian SL in Estonia. In addition, as we have described, the graphemic form of Latin h is similar to
the form of Cyrillic н, and both Latin n and Cyrillic нrepresent the alveolar nasal (IPA [n]). We coded the
forms for n in Estonian SL and h in Latvian SL as similar to the Russian SL form and assigned them the
character ID 299.
Next, historical examples of American SL (American 1821), Austrian SL (Austrian 1823), and German
SL (German 1820), as well as contemporary Latvian SL, use similar handshapes for Latin n and were
assigned ID 255. Handshapes for Latin n in contemporary American SL, Austrian SL, and German SL
were coded as similar and assigned ID 137. The cases of contemporary and historical Russian SL, as well as
Estonian SL, were more complex. Using similar reasoning to that described above for comparing handshapes
for Latin h and Cyrillic н, we compared the Austrian 1823 handshape representing Latin n with the Russian
1835 handshape for Cyrillic п, the latter grapheme representing the voiceless bilabial stop (IPA [p]). We
included the Russian 1835 handshape for пin the n-comparison and coded the handshape with ID 255.
The contemporary Russian SL handshape for пdi󰋳ers slightly from the Russian 1835 example, with greater
bending of the index and third 󰋴ngers, and was assigned ID 137. Estonian SL reversed the process just
described, taking the Russian SL handshape for Cyrillic пto represent Latin p. Thus, we included the
Estonian SL handshape for p in the n-comparison, assigning ID 255. In addition, because the Estonian SL
handshape for n was coded in the h-comparison, and because Estonian SL also has a handshape representing
Latin h, Estonian forms were coded twice in the set of character IDs for the h-comparisons. The Estonian
SL handshape for Latin h is unlike any other form and was assigned ID 113.
4 Phylogenetic Analysis
4.1 Reasons for exploratory data analysis and eects of incompatible signal
We assume that, in the case of language, evolutionary history does not strictly follow a tree model, but
includes reticulation as a result of borrowing and, in the case of MAs, partial or complete replacement by
standardisation or due to complex socio-political constraints. In addition, we expect a substantial amount of
positive selection due to articulatory and perceptual pressures within the framework of a limited number of
possible morphologies (handshapes), resulting in a high level of homoplasy, as similar handshapes evolved or
were conceived independently from each other. Homoplasy will provide support for topological alternatives,
outcompeting those best re󰋵ecting the evolutionary history of SLs. Finally, our data includes historical and
contemporary MAs (i.e., potential ancestors and their descendants), a situation poorly handled by trees and
best by NNets (Spencer et al. 2004). Each aspect can add an additional evolutionary/historical dimension,
Power et al. Sign Language Dispersal June 2019
Power et al. Dispersal of Sign Languages (Supplement) 6
and a tree is, per se, a 1-dimensional graph.
For instance, if A, B, and C have a common origin only expressed by similarity of B to both A and C,
it’s impossible to 󰋴nd a correct tree. The NNet can handle incompatible signal to some degree (B will be
placed between A and C) but is restricted to distance matrices, which can be strongly biased by missing
data artefacts (here, the number of Cogids applicable only for a subset of the SLs); and they are planar, 2-
dimensional graphs. If a taxon or lineage shares traits with more than two distinct, unrelated lineages (e.g.,
if B borrowed from the not directly related D), one of these relationships will get lost in the graph. The
limitation to two dimensions is the reason for the di󰋳erences between the all-inclusive NNet in Figure 3 and
the time-󰋴ltered NNets in Figure 4, both in the main text (e.g., regarding the placement of Russian 1835).
CNets are n-dimensional, that is, if the data re󰋵ect ndi󰋳erent topologies, they will all be represented in the
CNet. For the A-B-C-D example, we may 󰋴nd support for three partly incompatible splits: A + B, B + C, B
+ D. If the common origin of A, B, and C is well re󰋵ected in the character-matrix, we will get high support
for a fourth split, A + B + C, which competes with B + D, and both will be seen in the CNet. The NNet will,
in contrast, place B between A + C and D (aspect-wise correct), and the tree will place either (i) B as sister to
D and both as sister to A + C, or (ii) D as sister to A + B + C, which is equally incorrect. However, CNets are
based on a tree sample, hence, all limitations that apply for using tree-inference to reconstruct the history of
(here) SLs, apply also for each inferred pseudoreplicate tree during BS (here, 10,000 BS pseudoreplicates).
In addition to general branching artefacts, we will have more or less random branching patterns (NJ-
and ML-BS replicates are always fully-resolved trees, no branch has zero length, no polytomies). The use
of di󰋳erent optimality criteria allows for testing the stability of potential relationships seen in the BS tree
samples on which the BS-support CNets are based under di󰋳erent assumptions: NJ-BS CNets show the
robustness of signal based on overall similarity (or dissimilarity), minimising the e󰋳ect of single, potentially
misleading characters, but also in󰋵icting relationships based on dissimilarity to everything else (including
long-branching artifacts, LBA); P-BS CNets, also a󰋳ected by LBA, provide the most-conservative, but often
also least-discriminating result under the assumption that all character changes (mutations) have exactly the
same probability, an assumption that must be wrong for our data; and ML-BS CNets optimise a model that
allows for between-character variation but which runs the risk of over-weighting certain character splits. If
the CNets converge and support a neighbourhood seen in the NNet, it can be considered a data-unbiased
result; if they deviate from or contrast with each other, the resulting competing alternatives may be biased
by unrepresentative pairwise distance pro󰋴les due to missing data, or by inferred character mutations that
strongly depend on the assumed model. Given the complexity of character similarity and historical pathways
of SLs, it is impossible to judge, in such cases, which optimality criterion gives a better re󰋵ection of the true
situation. However, it is safe to assume that the true situation is one of the preferred alternatives, if it can
be explained by a single tree; or that the di󰋳erent preferred alternatives show di󰋳erent aspects of the true
situation, if too complex for a single tree.
4.2 Character Mapping
The 󰋴gures in this section show character maps based on the binary sequences encoding for the basic con-
cepts shared among all Latin alphabets, i.e. the standard set of 26 letters from “a” to “z”. In principle, for
each concept a (full) median network is reconstructed for the binaries encoding for a character and then
mapped visually on the time-taxon-󰋴ltered neighbour-nets. The (inferred or deduced) mutation is indicated
for each concept by arrows. When deduced, we consider not only the character cliques mapped on the
networks but also the age and country.
For instance, for the concept “k” in the oldest MAs set, we have three binary sequences (addressed and
abbreviated as “Dutch” = “Du.”, after the earliest MA showing this handshape, Dutch 1790 vs. Fr.” vs.
At.”; see annotation de󰋴nitions below). The full median network for handshapes of the concept “k” is a
triangle: to change from one binary sequence to another requires two steps (loss of the Cogid de󰋴ning the
one handshape, and gain of the Cogid de󰋴ning the other handshape). “k” is not de󰋴ned for the Spanish-
origin group including the oldest MAs (Bonet, Yebra) that in󰋵uenced the 󰋴rst MAs in all groups. Hence,
Power et al. Sign Language Dispersal June 2019
Power et al. Dispersal of Sign Languages (Supplement) 7
the Dutch version can be inferred as one primitive/commonly shared handshape evolved/changed from un-
known source (“Du.”), possibly competing with the extinct French handshape “→ †Fr. in the nearly
as old French 1800/1803 MAs. The original Austrian-origin group handshape (“At.”) is inferred to have
replaced the Dutch version (“At. Du.”): the neighbourhood linked to the Austrian handshape is embed-
ded in a much larger neighbourhood including earliest Danish and Russian MAs as well as Austrian 1823
with a “Dutch”-type handshape. Note that this is an inference using the logical framework for median net-
works and should not be viewed as conclusive evidence that the “Austrian”-type replaced the “Dutch”-type.
With respect to the age of the involved MAs and overall reconstructed history of SLs (see main text), it is
equally probable that younger MAs of the Austrian-origin group took over the more widespread “Dutch”-
type replacing the original “Austrian”-type typical for this lineage. Hence, there may be con󰋵icts in the
reconstructions shown in all three 󰋴gures, each one using a di󰋳erent taxon set and potentially a di󰋳erent set
of binaries. For example, in the mid-time network, we have an additional “Danish”-type and the “Dutch”-
type is inferred as the original handshape for “k” in all MAs including this concept. On the other hand, the
“Russian”-type handshape used in contemporary Norwegian SL can be deduced to represent a borrowing
or convergent development since older Norwegian MAs showed the “Danish”-type typical for the Danish
subgroup within the Austrian-origin group (“Da. Ru.”).
All annotations in the graphs relate to the labels used for binary sequences in the 󰋴le lists.xlsx in the
electronic supplementary material. We de󰋴ne the annotations below.
Annotation Denition
B Handshape that can be traced back to Bonet 1620, di󰋳erent from the handshape in
Yebra’s MA.
Y Handshape that can be traced back to Yebra 1593, di󰋳erent from the handshape in
Bonet’s MA.
O Original handshape: the handshape inferred as the original form of all covered European
lineages. When SLs of di󰋳erent groups share a handshape with Bonet and Yebra, this
handshape is labelled as “O”. Per de󰋴nition, “B”, “Y” and “O” are mutually exclusive.
C Cosmopolitan: the label indicates that this handshape is not only shared among di󰋳erent
groups covered in the networks but also in the British-origin and/or Afghan-Lebanese
groups, for which our data set includes no historical MAs (hence, not part of any graph
in Figures 4.2.1–4.2.3).
D Derived handshape: a handshape shared by SLs of di󰋳erent lineages. Per de󰋴nition, the
label “D” is mutually exclusive with all other labels, we only used it, when none of the
other labels apply.
“unique” Unique binaries: handshapes restricted to a single SL.
Highlights that this particular handshape is only found in historical MAs.
Two letters Handshapes (mostly) restricted to, exclusively found in a single main group.
At. Austrian/Austrian-origin group
Da. Danish/Danish subgroup within the Austrian-origin group
Du. Dutch: used for handshapes of the French-origin group not shared by French SLs but
found in the historical Dutch MAs (especially Dutch 1790, the oldest MA of the French-
origin group in our data set).
Fr. French/French-origin group
Ge. German: used for handshapes of the Austrian-origin group not diagnostic for the Danish
subgroup and not found in the oldest Austrian MAs but (historical) German MAs
Po. Polish/Polish group
Ru. Russian/Russian group
Power et al. Sign Language Dispersal June 2019
Power et al. Dispersal of Sign Languages (Supplement) 8
Sp. Spanish/Spanish-origin group: this can be a handshape exclusively found in Bonet,
Yebra and Spanish SLs as well as a handshape found only in younger Spanish MAs
but not Bonet or Yebra.
Sw. Swedish/Swedish group
Coloured mutations refer to the accordingly coloured edge-bundle (taxon split) in the respective graph:
mutations in light grey (and smaller) font unique mutations restricted to the respective OTU (“leaf”) of the
network; dark grey indicates the (inferred) handshape set of a hypothetical common “ancestor” (median
networks can place taxa/concepts in explicit ancestor-descendant relationships, with the inferred “medians”
representing hypothetical ancestors).
4.2.1 Comprehensive character map for sign languages up to 1840
The list highlights the handshapes shared (labelled subsequently as original/“O”, or Spanish/“Sp.”) and
di󰋳ering between the two oldest MAs (labelled as “B” or “Y”) in our data set (Yebra 1693, Bonet 1620).
The handshapes for the concept “d” are unique in both the Yebra and Bonet MA, all other MAs show
di󰋳erent handshapes (most show a handshape characteristic of the oldest MAs of the Austrian- and French-
origin groups).
Power et al. Sign Language Dispersal June 2019
Power et al. Dispersal of Sign Languages (Supplement) 9
4.2.2 Character map for sign languages from mid-19th to mid-/late 20th century
Similar to the character map above, each lineage can be characterised by group-speci󰋴c handshapes (in
coloured font). Note the striking di󰋳erence between the diversi󰋴cation patterns in the French- vs. Austrian-
origin group. The in󰋵ated fan-like structure of the neighbour-net for the Austrian cluster relates to a gradual
accumulation of lineage-speci󰋴c handshapes and their subsequent modi󰋴cation (and/or) replacement in sub-
groups. Potential older sources (Spanish 1815, Russian 1835, German 1820, Danish 1808, French 1815
and American 1821 MAs) are included to link this taxon set with the one shown in Figure 4.2.1.
Power et al. Sign Language Dispersal June 2019
Power et al. Dispersal of Sign Languages (Supplement) 10
4.2.3 Character map for contemporary sign languages
The taxon set also includes also historical but post-war MAs (second half of the 19th century). In addi-
tion to lineage-speci󰋴c handshapes, we 󰋴nd concepts directly linking the Polish and Austrian-origin group
(coloured font) by shared handshapes. Note the high level of inferred single-SL (tip) mutations, especially
in (phylogenetically) isolated SLs such as contemporary Norwegian or Albanian SL.
4.3 Split Support
Non-parametric bootstrapping (BS) support for competing relationships in the Polish and Russian groups
with respect to Latvian SL. ML = maximum likelihood; ASC = corrected for ascertainment bias; UNC =
uncorrected for ascertainment bias; NJ = neighbour-joining; P = parsimony.
Alternative type Group ASC UNC
Terminal Unchallenged Contemp. Cyrillic: Bulgarian SL, Russian SL, Ukranian SL 84 85 71 49
CNet & NNet Russian gr. incl. Russian 1835 37 37 27 54
CNet Contemp. Russian gr. + Latvian SL 35 37 18 <15
CNet & NNet Contemp. Russian gr. + Estonian SL <15 <15 46 21
Preferred Russian gr. incl. Russian 1835, excl. Latvian SL 18 17 40 39
CNet Russian gr. incl. Latvian SL excl. Russian 1835 18 <15 30 <15
Deep split
NNet-alternative Latvian SL part of Russian gr. 41 43 33 39
NNet-alternative Latvian SL part of Polish gr. <15 <15 37 <15
Power et al. Sign Language Dispersal June 2019
Power et al. Dispersal of Sign Languages (Supplement) 11
Abramov, Igor A. (1993). “History of the deaf in Russia”. In: Looking back: A reader on the history of Deaf communities and their sign languages.
Ed. by Renate Fischer and Harlan Lane. Vol. 20. Hamburg: Signum-Verlag, 199–205.
Akerly, Samuel (1823). Elementary exercises for the deaf and dumb. New York: E. Conrad.
Alle, Johann L. (1821). Anleitung taubstumme Kinder in Schreiben, Lesen, Rechnen zu unterrichten und sie moralisch-gut und bürgerlich-brauchbar
zu bilden. Gmünd: Verlag d. K. Taubstummen-Lehranstalt.
Ballesteros, Juan Manuel and Francisco Fernández Villabrille (1845). Curso elemental de instruccion de sordo-mudos. Parte segunda: Práctica de
la enseñanza. Madrid: Imprenta del Collegio de Sordo-mudos y Ciegos.
Blanco, Ángel Herrero (2009). Gramática didáctica de la Lengua de Signos Española (LSE). Spanish. Akademisk Forlag. Madrid: Ediciones SM.
Bonet, Juan Pablo (1620). Reduction de las letras y arte para enseñar a ablar los mudos. Madrid: Francisco Abarca de Angulo.
Brien, David, ed. (1992). Dictionary of British Sign Language: English. London: Faber & Faber.
Bruun, Hans Rener (1900). Det Norske Haandalphabet. Oslo Museum, Byhistorisk samling. Online; accessed 3-April-2019. Oslo.
Carderera, Mariano (1859). Apuntes sobre la educación elemental del sordomudo. Madrid: Imprenta de Ramón Campuzano.
Castberg, Peter Atke (1818). Forelæsninger over Døvstumme-Undervisningens Methode. Danish. Copenhagen: L. Beetens Forlag.
Czech, Franz Herrmann (1839). Nothwendigkeit der allgemein einzuführenden Elementar-Bildung der Taubstummen, aus den Verhältnissen derselben
zum Staate und zur Kirche hergeleitet und dargestellt. 2nd ed. Vienna: P. P. Mechitaristen.
Døvstumme-Raadet (1926). Ordbog i de døvstummes Tegnsprog. Copenhagen: Herm. Rolsteds Bogtr.
Fleri, Viktor I. (1835). Глухонемые, рассматриваемые в отношении к их состоянию и к способам образования, самым свойственным
их при. St. Petersburg: Типография А. Плюшара.
Fornari, Pasquale (1897). Il sordomuto e la sua istruzione: Manuale per gli allievi e le allieve delle R. scuole normali, maestri, genitori e lantropi.
Milan: Ulrico Hoepli.
Gama, Flausino José da (1875). Iconographia dos signaes dos surdos-mudos. Rio da Janeiro: Typographia Universal de E. & H. Laemmert.
Gordon, Joseph C. (1886). The American manual alphabet. Washington, D.C.: Bretano Bros.
Hatzopoulou, Marianna (2008). “Acquisition of reference to self and others in Greek Sign Language: From pointing gesture to pronominal pointing
signs”. PhD thesis. Institutionen för lingvistik.
Heijningen Bosch, M. van (1823). Berigt, houdende eenige wenken over het eerste onderwijs aan doofstommen: Benevens het hand-alphabet, in het
koper gegraveerd. 2nd ed. Groningen.
Hendriks, Bernadet (2008). Jordanian Sign Language: Aspects of grammar from a cross-linguistic perspective. Utrecht, The Netherlands: Landelijke
Onderzoekschool Taalwetenschap.
Hollaka, Józefa and Teo󰋴la Jagodzińskiego (2011). Słownik mimiczny dla głuchoniemych i osób z nimi styczność mających. Warszawa: Instytucie
Głuchoniemych i Ociemniałych.
Hudáková, Andrea (2008). Prstová abeceda pro tlumočníky. Česká komora tlumočníkŭ znakového jazyka.
Johnston, Trevor (2014). Auslan signbank. Online; accessed 27-March-2019.
Jørgensen, Johannes (1907). De døvstummes Haandalfabet og 280 af de almindeligste Tegn. Copenhagen: Alfred Jacobsens gra󰋴ske Etablissement.
Ladébat, André-Daniel La󰋳on de (1815). Recueil des dénitions et réponses les plus remarquables de Massieu et Clerc, sourds-muets, aux diverses
questions quileur ontété faites dans les séances publiques de M. l’Abbé Sicard, à Londres: Auquel on a joint l’alphabet manuel des sourds-muets,
le discours d’ouverture de M. l’Abbé Sicard, et une lettre explicative de sa méthode. Online; accessed 28-March-2019. Cox et Baylis.
Landsforbund, Norske Døves (1955). Håndalfabet for Døve. Sverresborg Trøndelag Folkemuseum, Trøndelag Folkemus. Online; accessed 3-April-
2019. Trøndelag.
Learn Irish Sign Language (2012). Online; accessed 28-March-2019.
List, Johann-Mattis (2017). “A web-based interactive tool for creating, inspecting, editing, and publishing etymological datasets”. In: Proceedings
of the 15th Conference of the European Chapter of the Association for Computational Linguistics. System Demonstrations. Valencia: Association
for Computational Linguistics, 9–12.
List, Johann-Mattis, Simon Greenhill, Tiago Tresoldi, and Robert Forkel (2018). LingPy. A Python library for quantitative tasks in historical lin-
guistics. Jena: Max Planck Institute for the Science of Human History.
Long, Joseph Schuyler (1918). The sign language: A manual of signs, illustrated. 2nd ed. Des Moines, IA: Robert Henderson.
Lydell, Thomas, ed. (2018). Spreadthesign. URL:
Magarotto, Cesare (1996). Vocabolario della Lingua Gestuale Italiana dei sordi. Armando Editore.
Martí Mora,Francisco de Paula (1815). Alfabeto manual para la instrucciónde los sordo-mudos del Real Colegio de Madrid. Biblioteca de Catalunya.
Online; accessed 3-April-2019. Barcelona.
May, Joseph (1789). Anleitung zum Unterrichte der Taubstummen nach der Lehrart des Herrn Abbe de l’Epée zu Paris, nebst einer Nachricht von
dem kaiserl. königl. Taubstummeninstitute in Wien. Vienna: k. k. Taubstummeninstitutsbuchdruckerey.
McKee, D., R. McKee, S. Pivac Alexander, L. Pivac, and M. Vale (2011). Online dictionary of New Zealand Sign Language. Online; accessed
28-March-2019. Wellington.
Mörser, I.L. and H.D. Guyot (1790). Het alphabet manuel of het vingerspraak voor doven en stommen. Groningen.
Nyegaard, Andreas C. (1871). De Døvstummes Haandalphabet samt et Udvalg af deres lettere Tegn sammenstillet, tegnet, graveret og udgivet af en
Forening af Dövstumme. Danish. Copenhagen: Kongelige Dövstumme-Institut.
Paulsson, Per (1866). Historik öfver folkunderwisningen i Sverige från äldsta till närvarande tid. Stockhom: Adolf Bonnier.
Pelissier, Pierre (1856). Iconographie des signes. Paris: Paul Dupont.
Perelló, Jorge and Juan Frigola Masclans (1998). Lenguaje de signos manuales. Madrid: CIE Inversiones Editoriales Dossat.
Plum, Ole Munk, Sofus Kjær, Asger Holm, and Orla A. Odde (1967). Håndbog i tegnsprog. Copenhagen: Danske Døves Landsforbund.
Reuschert, E. (1909). Die Gebärdensprache der Taubstummen und die Ausdrucksbewegungen der Vollsinnigen. Leipzig: Kommissions-Verlag von
H. Dude.
Riemann, G. (1916). Taubstumm und blind zugleich: Pädagogische und psychologische Darbietungen. 2nd ed. Berlin: Schriftenvertriebsanstalt
Rubino, F., A. Hayhurst, and J. Guejlman (1975). Gestuno: International sign language of the deaf . Carlisle: British Deaf Association.
Sabir Khan, Nabeel, Adnan Shahzada, Saleem Ata, Adnan Abid, Yaser Khan, Shoaib Farooq, Tahir Mushtaq, and Inayatullah Khan (2014). “A
Vision Based Approach for Pakistan Sign Language alphabets Recognition”. La Pensée 76.03, 274–285.
Power et al. Sign Language Dispersal June 2019
Power et al. Dispersal of Sign Languages (Supplement) 12
Šarac Kuhn, Ninoslava, Tamara Alibašić Ciciliani, and Ronnie B. Wilbur (2006). “Phonological parameters in Croatian Sign Language”. Sign
Language & Linguistics 9.1, 33–70.
Schwarzer, Anton (1827). Lehrmethode zum Unterrichte der Taubstummen in der Tonsprache für Lehrer. Buda: Universitäts-Schriften.
Sicard, Roch Ambroise Cucurron (1803). Cours d’instruction d’un sourd-muet de naissance, pour servir à l’éducation des sourds-muets, et qui peut
être utile à celle de ceux qui entendent et qui parlent. 2nd ed. Paris: Le Clere.
Sigurðsson, Sigfús (1857). Fingramálsstafróf. Akureyri: Jósef Grímsson.
Sonnemans, Bruno (2016). LSFB asbl: An online dictionary and newspaper in LSFB free access of videos. Online; accessed 28-March-2019.
Spencer, Matthew, E. A. Davidson, A. C. Barbrook, and C. J. Howe (2004). “Phylogenetics of arti󰋴cial manuscripts.” Journal of Theoretical Biology
227, 503–511.
Svendsen, Conrad (1893). De Døvstumme, deres Opdragelse i Hjem og Skole. Kristiania: H. Aschehoug & Co.s Forlag.
Svenskt teckenspråkslexikon (2014). Online; accessed 28-March-2019. Stockholm.
Tegnordbok (2016). Online; accessed 28-March-2019.
Tennant, Richard A. and Marianne Gluszak Brown (1998). The American Sign Language handshape dictionary. Gallaudet University Press.
Toom, Koostaja R (1988). Kõnelevad käed: Eesti Viipekeele sõnastik. Estonian. Tallinn: Tartu riiklik ülikool eesti NSV kurtide ühing.
tutkimuskeskus, Kuurojen Liitto ry ja Kotimaistenkielten, ed. (1998). Suomalaisenviittomakielen perussanakirja. Kotimaisten kielten tutkimuskeskuk-
sen julkaisuja 104. Helsinki: KL Support Oy.
unknown, Author (1799-1800). Alphabet, manuel-gure des sourds-muets de naissance. Online; accessed 28-March-2019. Paris.
(2001). Afghan Sign Language. Kabul, Afghanistan: Comprehensive Disabled Afghans Program (UN).
(2012a). Lietuvių Gestų Kalbos žodynas. Online; accessed 28-March-2019.
(2012b). Vertaalwoordenboek. Online; accessed 28-March-2019.
(2014). Eesti Viipekeele eesti keele sõnastiku. Online; accessed 28-March-2019.
(2016). Alfabēts. Online; accessed 28-March-2019.
(2019a). Dictionairre. Online; accessed 28-March-2019.
(2019b). Gjuha e Shenjave Shqipe. Online; accessed 28-March-2019.
(2019c). Ordbog over Dansk Tegnsprog. Online; accessed 28-March-2019.
(2019d). Sourdlang. Online; accessed 28-March-2019.
Venus, Michael (1823). Das kais. kön. Taubstummen-Institut in Wien, dessen Entstehung, Erweiterung und gegenwärtiger Zustand. Vienna: Anton
Williams, Howard G. and Polina Fyodorova (1993). “The origins of the St. Petersburg institute for the deaf”. In: Looking back: A reader on the
history of Deaf communities and their sign languages. Ed. by Renate Fischer and Harlan Lane. Vol. 20. Hamburg: Signum-Verlag, 295–305.
Yebra, Melchor de (1593). Libro llamado Refugium Inrmorum: Muy util y prouechoso para todo genero de gente : En el qual se contienen muchos
auisos espirituales para socorro de los aigidos enfermos, y para ayudar à bien morir a los que estan en lo ultimo de su vida ; con vn Alfabeto de
S. Buenauentura para hablar por la mano. Madrid: Luys Sa[n]chez.
Zeshan, Ulrike (2003). “Aspects of Türk Isaret Dili (Turkish Sign Language)”. Sign Language & Linguistics 6.1, 43–75.
Zwitserlood, I.E.P. (2010). “Laat je vingers spreken. NGT en vingerspelling”. Levende Talen Magazine 97.2, 46–47.
Łacheta, Joanna, Małgorzata Czajkowska-Kisil, Jadwiga Linde-Usiekniewicz, and Paweł Rutkowski, eds. (2016). Korpusowy słownik Polskiego
zyka Migowego. Online; accessed 28-March-2019. Warsaw: Faculty of Polish Studies, University of Warsaw.
Power et al. Sign Language Dispersal June 2019
ResearchGate has not been able to resolve any citations for this publication.
Full-text available
The amount of available digital data for the languages of the world is constantly increasing. Unfortunately, most of the digital data are provided in a large variety of formats and therefore not amenable for comparison and re-use. The Cross-Linguistic Data Formats initiative proposes new standards for two basic types of data in historical and typological language comparison (word lists, structural datasets) and a framework to incorporate more data types (e.g. parallel texts, and dictionaries). The new specification for cross-linguistic data formats comes along with a software package for validation and manipulation, a basic ontology which links to more general frameworks, and usage examples of best practices.
Full-text available
The Dravidian language family consists of about 80 varieties (Hammarström H. 2016 Glottolog 2.7) spoken by 220 million people across southern and central India and surrounding countries (Steever SB. 1998 In The Dravidian languages (ed. SB Steever), pp. 1textendash39: 1). Neither the geographical origin of the Dravidian language homeland nor its exact dispersal through time are known. The history of these languages is crucial for understanding prehistory in Eurasia, because despite their current restricted range, these languages played a significant role in influencing other language groups including Indo-Aryan (Indo-European) and Munda (Austroasiatic) speakers. Here, we report the results of a Bayesian phylogenetic analysis of cognate-coded lexical data, elicited first hand from native speakers, to investigate the subgrouping of the Dravidian language family, and provide dates for the major points of diversification. Our results indicate that the Dravidian language family is approximately 4500 years old, a finding that corresponds well with earlier linguistic and archaeological studies. The main branches of the Dravidian language family (North, Central, South I, South II) are recovered, although the placement of languages within these main branches diverges from previous classifications. We find considerable uncertainty with regard to the relationships between the main branches.
Full-text available
The fields of phylogenetic tree and network inference have dramatically advanced in the past decade, but independently with few attempts to bridge them. Here we provide a framework, implemented in the phangorn library in R, to transfer information between trees and networks. This includes: 1) identifying and labelling equivalent tree branches and network edges, 2) transferring tree branch support to network edges, and 3) mapping bipartition support from a sample of trees (e.g. from bootstrapping or Bayesian inference) onto network edges. The ability to readily combine tree and network information should lead to more comprehensive evolutionary comparisons and inferences. This article is protected by copyright. All rights reserved.
This is an accessible introduction to the major areas of sociolinguistics as they relate to sign languages and deaf communities. Clearly organised, it brings together a team of leading experts in sign linguistics to survey the field, and covers a wide range of topics including variation, multilingualism, bilingualism, language attitudes, discourse analysis, language policy and planning. The book examines how sign languages are distributed around the world; what occurs when they come in contact with spoken and written languages; and how signers use them in a variety of situations. Each chapter introduces the key issues in each area of inquiry and provides a comprehensive review of the literature. The book also includes suggestions for further reading and helpful exercises. The Sociolinguistics of Sign Languages will be welcomed by students in deaf studies, linguistics and interpreter training, as well as spoken language researchers, and researchers and teachers of sign language.
We present an initial description of the sign parameters in Croatian Sign Language. We show that HZJ has a comparable phonological structure to other known sign languages, including basic sign parts, such as location, handshape, movement, orientation, and nonmanual characteristics. Our discussion follows the Prosodic Model (Brentari 1998), in which sign structure is separated into those characteristics which do not change during sign formation ( inherent features ) and those that do ( prosodic features ). We present the model, along with discussion of the notion of constraints on sign formation, and apply it to HZJ to the extent that we are able to do so. We identify an inventory of the relevant handshapes, orientations, locations, and movements in HZJ, and a partial inventory of nonmanuals. One interesting feature of the HZJ environment is the existence of two fingerspelling alphabets, a one-handed and a two-handed system. We also provide additional analytical steps that can be taken after the initial inventory has been constructed. Both minimal pairs and constraints on sign formation are especially useful for demonstrating the linguistic systematicity of sign languages and separating them from gesture and mime.