Signbase, a Collection of Geometric Signs on Mobile Objects in The

www.nature.com/scientificdata

OPEN SignBase, a collection of geometric Data Descriptor signs on mobile objects in the Paleolithic Ewa Dutkiewicz1,3,4 ✉ , Gabriele Russo1,3, Saetbyul Lee2,3 & Christian Bentz 2,3 ✉

In the Paleolithic, geometric signs are abundant. They appear in rock art as well as on mobile objects like artworks, tools, or personal ornaments. These signs are often interpreted as a refection of symbolic thought and associated with the origin of cognitively modern behavior. SignBase is a project collecting the wealth of geometric signs on mobile objects in the European Upper Paleolithic, African Middle Stone Age (MSA), as well as selected sites from the Near East and South East Asia. Currently, more than 500 objects of the Aurignacian techno-complex (ca. 43,000 to 30,000 years BP) are registered in SignBase. They are linked to information about geographic and archaeological provenience, the type of object and material, size and preservation, and respective literature references. We identify around 30 diferent sign types found on these objects across Europe in the Aurignacian and illustrate how SignBase can be used to analyze geographical clusters. Ultimately, we aim to enable quantitative analyses of abstract graphical expression before the emergence of writing.

Background & Summary In the Paleolithic, geometric signs are found in parietal art as well as on mobile objects. Most of these signs appear in the period between 100,000 and 10,000 BP, but some examples are known from earlier periods1,2. Te Paleolithic is further subdivided into so-called techno-complexes. For instance, the Aurignacian is an Early Upper Paleolithic techno-complex dating to around 43,000 to 30,000 BP3–9. It roughly corresponds to the time when anatomically modern humans migrated into the Near East and Europe and encountered and lived alongside Neanderthals for several millenia. One of the characteristics of the Aurignacian is the abundant use of osseous material for the production of tools, weapons, ornaments, and artworks (see Fig. 1)10–22. Many of these objects are decorated with geometric signs. Geometric signs are sometimes referred to as abstract motifs, patterns, marks, or jottings in the literature. We use the term “sign” here in line with Peirce’s semiotics23,24, i.e. in the broad sense of a representation of some kind, which is then further subdivided into index, icon, and symbol10,25,26. In this context, the term “abstract” is ofen used to further underline that the signs are not obviously iconic, that is, they cannot be recognized by modern viewers as fgurative depictions. Simple geometric forms such as dots, lines, and crosses, as well as more complex patterns such as grids or overlapping crosses, are interpreted in this sense. However, in some cases, even seem- ingly abstract signs might bear some iconicity. For instance, when applied to animal fgurines, simple dots might refect patterns on fur which were discernible for the Paleolithic viewer. With this caveat in mind, we here choose to use the attribute “geometric” rather than “abstract”. We thus merely refer to the visual property of signs, without further delving into the issue of whether particular signs are to be seen as indices, icons, or symbols. Tere are several studies investigating geometric signs in parietal art27–30. However, studies scrutinizing signs on mobile objects, such as fgurines, tools, or personal ornaments, are rare and mostly limited to either single objects, or to particular assemblages10,31–36. SignBase aims to provide extensive data on mobile objects and the geometric signs found on these. A large body of data from the Swabian Aurignacian is available from previous collection eforts10,31,32, involving frst-hand analyses by the frst author. We complement this data with other Aurignacian assemblages available via the literature3,22,37–55. We thereby enable quantitative analyses of this rich material.

1Department of Early Prehistory and Quaternary Ecology, University of Tübingen, Tübingen, Germany. 2Department of General Linguistics, University of Tübingen, Tübingen, Germany. 3DFG Center for Advanced Studies “Words, Bones, Genes, Tools”, University of Tübingen, Tübingen, Germany. 4Museum für Vor- und Frühgeschichte, Staatliche Museen zu Berlin – Stiftung Preußischer Kulturbesitz, Berlin, Germany. ✉e-mail: [email protected]; [email protected]

Scientific Data | (2020)7:364 | https://doi.org/10.1038/s41597-020-00704-x 1 www.nature.com/scientificdata/ www.nature.com/scientificdata

Fig. 1 Examples of mobile objects with geometric signs from the Aurignacian. 1. Vogelherd, mammoth fgurine, ivory, length 5 cm (Lipták © University of Tübingen); 2. Hohlenstein-Stadel, deer-tooth, personal ornament, size 2.9 cm (Dutkiewicz © Landesamt für Denkmalpfege im Regierungspräsidium Stuttgart); 3. Vogelherd, lissoir, bone, size 21 cm (Lipták © University of Tübingen).

Early graphical expressions, such as found on objects from the MSA site Blombos in South Africa, have sometimes been interpreted as symbolism and associated with the cognitive modernity of humans, in some cases even with the presence of “fully syntactic language”56,57. However, a recent study performing experiments with modern-day participants comes to the conclusion that these engravings – while refecting “socially transmitted cultural traditions” – bear no clear indication of symbolism36. Systematic studies of geometric signs beyond the earliest Paleolithic fnds in South Africa will further help to establish their semiotic status. Tis will also give us a better understanding of their relation to later symbolic behavior such as early writing systems. Apart from insights into the evolution of human cognition, the geographic distribution of geometric signs – and the associated archaeological cultures – is a second major line of research. Geographic analyses promise to shed light on cultural developments and population turn-overs across the Late Pleistocene, as inferred by the archaeological and human fossil record58–61. As an example of practical applications in this direction, we give a preliminary clustering analysis of the sign types found across Europe in the Aurignacian. Methods Objects. Decorated mobile objects are mostly (though not exclusively) made from osseous material, like ivory, bone, or antler, and usually come from stratifed archaeological contexts. In contrast to the chronological difculties of dating parietal art62–64, mobile objects are usually well-dated, at least with reference to the given techno-complex. Te data of SignBase is structured according to these archaeological objects. Every artifact that carries geometric signs is assigned an object identifer consisting of a three-letter abbreviation of the excavation site (e.g. Vogelherd Cave: vhc) and a running, four-digit number. Te identifer is linked to detailed information about the object. Tis is frstly the techno-complex, geographic information, stratigraphical unit (layer), and the dating method(s). Te year of excavation is also indicated, since very old excavations may lack the relevant information, or it might be inadequate. Secondly, we give information about the object itself: the material, the type of object (object type), the dimensions (length, width, depth) as well as the state of preservation (complete, almost complete, and fragmented). Tis is followed by a short description of the object, as well as the relevant literature, such as excavation reports. Te data fle containing all objects and respective information is described in more detail in the section on Data Records. A website displaying this information per object is available online at www. signbase.org. So far, 531 mobile objects carrying geometric signs from 65 Aurignacian sites in Europe and the Near East are registered in SignBase. Te locations of excavation sites and individual artifacts can be seen in Fig. 2. Te majority of sites that yield artifacts carrying geometric signs of the Aurignacian derive from four main areas: South-West France (in particular the Dordogne), the Swabian Jura in southern Germany, as well as a series of sites in modern-day Belgium, and the Czech Republic. Tere are further isolated sites in Southern Spain (El Salitre),

Scientific Data | (2020)7:364 | https://doi.org/10.1038/s41597-020-00704-x 2 www.nature.com/scientificdata/ www.nature.com/scientificdata

60 52.5 Belgian Sites aa)) b)b) 50.0

50 Dordogne 47.5 lat lat Swabian Jura 45.0 40

42.5

30 -10 01020304050 40.0 -5 0510 15 20 lon lon

Fig. 2 Maps of the Aurignacian sites across Europe and the Near East. (a) Each triangle indicates an archaeological site where artifacts carrying geometric signs were found. A density plot is overlaid with high (red) and low (yellow) densities of sites. (b) Zoom into the areas yielding most objects with geometric signs. Individual artifacts are plotted as black dots (with some jitter added to avoid overplotting if many artifacts come from the same site). A density plot is overlaid with high (red) and low (yellow) densities of artifacts (for script see Supplementary File 1).

Material Number Total Osseous Animal tooth 4 Antler 32 Bone 202 Ivory 210 Shell 2 450 Rocks Calcite 1 Limestone 51 Flint cortex 10 Fossil 1 Sandstone 1 Slate 2 Steatite 1 Ochre 3 Stone indet. 2 72 Undetermined 9 Total 531

Table 1. Overview of the raw materials used for objects with geometric signs from the Aurignacian (n = 531).

Sicily (Riparo di Fontana Nuova), Israel (Hayonim cave), Iraq (Shanidar cave), and by the Black Sea (Muralovka) (Fig. 2a). While the density of sites is highest in the Dordogne, the density of artifacts is highest in the Swabian Jura (in particular the Vogelherd cave with overall around 170 artifacts) (Fig. 2b). Of course, both the distribution of sites and the distribution of objects are infuenced by historical factors such as excavation and publication eforts of particular universities and researchers. Note, however, that many Aurignacian sites across Europe have not yielded artifacts with geometric signs22,37. Hence, while the picture can still change as new sites are discovered and new artifacts published, we expect to have uncovered the main tendencies of the Aurignacian. Te majority (n = 450) of decorated mobile objects from the Aurignacian are made from osseous material (see Table 1). Rock material, like limestone, fint, or other rock types were used in 72 cases. Te objects carrying geometric signs are in 430 cases so-called symbolic artifacts (e.g. fgurines), tools are present in 72 cases (see Table 2).

Sign types. For each object registered in SignBase we then identify the types of signs represented on it. We defne mutually exclusive types, for instance, straight line, oblique line, radial line, notch, dot, cross, and more complex forms like grid, hashtag, and zigzag (see Figs. 3 and 4). Reduced pictograms like vulvae, or animal paws, are included in this collection as well since they constitute borderline cases between iconic and geometric. For instance, a paw might be a reduced geometric substitute for the entire animal. Overall, we identify 30 diferent sign types, while uncertain cases are subsumed under the category “other”. Each of the identifed sign types is marked as present/absent by one or zero. In Fig. 4, several examples of Aurignacian objects with diferent sign types are shown, such as line (aur0001, cas0014, lar0002, gdr0007, gpp0004, gdg0008, bla0014), oblique line (lar0002), notch (aur0001, cas0014, lar0002, gpp0005, msc0005, cel0005), oblique notch (gpp0004), circumferential

Scientific Data | (2020)7:364 | https://doi.org/10.1038/s41597-020-00704-x 3 www.nature.com/scientificdata/ www.nature.com/scientificdata

Object Type Number Total Symbolic Statuette 40 Possible statuette 25 Pendant 23 Personal ornament 21 Flute 9 Possible fute 16 Tube/fute 24 Tube 16 Antler object 3 Plaquette 4 Bloc 47 228 Tools Awl 21 Bâton percé/perforated baton 7 Blade 10 Compresseur 2 Crayon 1 Flake 10 Lissoir/spatula 43 Point 21 Rod 32 Other 18 165 Undetermined 138 Total 531

Table 2. Overview of the types of Aurignacian objects bearing geometric signs (n = 531).

spiral (cat0002), dot (gdg0003, bla0018, lar0002, bla0014), cross (gdg0008), hatching (gdg0008), and vulva (cal0002). Sometimes several sign types appear on a single object. In future versions of the database, numbers of sign occurrences per object, and coding of sign sequences will be included as well. Note that disagreements between codings by diferent researchers are unavoidable. Te description and typology of geometric signs must be based on visual impressions since we cannot understand the contextual relationships and meanings of such characters from today’s perspective. Any sign typology is hence to some extent subjective and leaves room for discussion. Diferent sign types can resemble each other, and it is sometimes difcult to determine them undoubtedly. We openly publish our coding decisions and hope for further input and discussion with other researchers. Furthermore, to estimate the degree of subjectivity in our choices, we have submitted the character types defned by us to several peers and calculated agreement scores (see the section on Technical Validation).

Frequencies of occurrence. Given our coding decisions, we can assess how ofen particular sign types occur. Some are frequently found across diferent sites and objects, while others are rarer or even restricted to a particular object (see Fig. 5). Te most frequent sign type is the simple notch (a short incision deeper than a line and in most cases applied on the edge of an object), occurring on almost half of the objects (48%, i.e. 254 of 531), followed by the line (33%), and cross (10%). Dots and cupules (7% and 2% respectively) are less frequent but still well-attested, whereas more elaborate signs such as hashtags, stars, or zigzags are exceedingly rare, and ofen associated with particular objects and sites. For example, clear instances of star-shaped signs (i.e. more than two lines crossing in a center point) are currently only attested on a fgurine from the Vogelherd cave (vhc0159), and an engraved ivory blade from the Grotte de La Princesse Pauline in Belgium (gpp0003).

Geographic clusters. Apart from diferences in frequencies of occurrence, sign types also difer regarding their geographic spread in the Aurignacian. While the most frequent sign types are spread widely across Europe and the Near East, others are geographically more confned (see Fig. 6). Te notch, for instance, is ubiquitous. It is found in some of the most westward (e.g. La Viña in Spain) and eastward sites (e.g. Hayonim Cave in Israel), as well as in the most southern (Sicily) and northern sites (Belgium) of central Europe. Others, such as crosses, hatchings, and dots, center around areas of high artifact density such as southwestern France, southern Germany, Belgium, and the Czech Republic. As an extreme example, abstract depictions of vulvae, while being attested a considerable amount of times (i.e. in 20 of 531 objects), are strictly limited to caves in the Dordogne, potentially indicating a local practice of graphic expression3,37,42,45,50,53,55.

Automated analyses. Beyond visual inspection of geographic maps, we here propose a more systematic way of scrutinizing clusters of artifacts and sign types. Tis also helps to better understand the type of data represented in SignBase. Take the following examples of binary sign type vectors for two objects from the Hohle Fels cave in the Swabian Jura (hfc0015, hfc0006), and Spy in Belgium (spy0023).

Scientific Data | (2020)7:364 | https://doi.org/10.1038/s41597-020-00704-x 4 www.nature.com/scientificdata/ www.nature.com/scientificdata

1 12 21

2 22 14

3 15 23

4 16

24, 25 5

26 6, 7

18 27 8, 9

10 19

11 20

Fig. 3 Schematic drawings of sign types as identifed for the Aurignacian (in brackets like shown in the data base): 1. Line (line); 2. Oblique line (obline); 3. Concentric lines (concenline); 4. Dashed line (dashline); 5. Radial line (radline); 6. Circumferential line (circumline); 7. Circumferential spiral (circumspiral); 8. Notch (notch); 9. Oblique notch (obnotch); 10. Radial notch (radnotch); 11. Circumferential notch (circumnotch); 12. Dot (dot); 13. Cupule (cupule); 14. Cross (cross); 15. Rhombus (rhombus); 16. Hashtag (hashtag); 17. Grid (grid); 18. Hatching (hatching); 19. Zigzag (zigzag); 20. Zigzag-row (zigzagrow); 21. Rectangle (rectangle); 22. Maccaroni (maccaroni); 23. V-Sign (v); 24. Pin to the lef side (pinlef); 25. Pin to the right side (pinright); 26. Star (star); 27. Vulva-Sign (vulva); 28. Paw-Sign (paw). Not shown in the table: anthropomorph, zoomorph, other.

hfc0006: 110100000101000000001000000000 Shfc0006 = {line, oblique line, dashed line, dot, cross, v-shape} hfc0015: 111000000100001000000000001000 Shfc0015 = {line, oblique line, radial line, dot, hatching, concentric line} spy0023: 100000000001000010001000000000 Shfc0015 = {line, cross, zigzag row, v-shape}

Tese vectors have 30 binary values – the value for “other” is discarded here, which leaves us with 516 objects (i.e. 15 objects have only sign type “other”). Te values refect whether a particular sign type is present (1) or absent (0). An equivalent representation is to give the set of sign types present on an object. Tese sets are displayed below the binary vectors. Te Jaccard distance65 between any given two sets A and B is then calculated as

Scientific Data | (2020)7:364 | https://doi.org/10.1038/s41597-020-00704-x 5 www.nature.com/scientificdata/ www.nature.com/scientificdata

2 4 5 1

7 8

10 9 11

12 13

Fig. 4 Examples of some Aurignacian mobile objects registered in SignBase and the identifed sign types (not in scale, for details see database): 1. aur0001: lines, notches; 2. cas0014: lines, notches; 3. cat0002: circumspiral; 4. gdg0003: dot; 5. bla0018: dots; 6. lar0002: line, obline, notch, dot; 7. gdr0007: line; 8. gpp0004: line, notch, obnotch; 9. msc0005: notch; 10. gdg0008: line, cross, hatching; 11. cel0005: notches; 12. cel0002: vulvae; 13. bla0014: dots, lines.

∣∣AB∩ dJaccard =−1, ∣∣AB∪ where the numerator is the cardinality of the intersection of the two sets, i.e. the number of shared sign types. Whereas the denominator is the cardinality of the union of two sets, i.e. the overall number of diferent sign types occurring on both objects together. Tus, the Jaccard distance between hfc0006 and hfc0015 is 3 dhfch0006, fc0015 =−1 ≈.067. Jaccard 9 While the Jaccard distance between hfc0006 and spy0023 is 3 dhfcs0006, py0023 =−1 ≈.057, Jaccard 7

Scientific Data | (2020)7:364 | https://doi.org/10.1038/s41597-020-00704-x 6 www.nature.com/scientificdata/ www.nature.com/scientificdata

notch 254 line 174 cross 53 obline 51 other 38 hatching 37 dot 35 vulva 20 grid 16 zoomorph 15 obnotch 15 cupule 13 radline 12 v 10 circumnotch

ype 10 zigzag 9 concenline 5

Sign T radnotch 5 maccaroni 4 rhombus 4 dashline 4 hashtag 3 star 2 anthropomorph 2 circumspiral 2 zigzagrow 2 pinright 1 pinleft 1 paw 1 rectangle 1 circumline 1 Frequency

Fig. 5 Frequencies of 31 diferent sign types (including “other”) across the 531 Aurignacian objects (for script see Supplementary File 2).

and for hfc0015 and spy0023 we have 1 dhfcs0015, py0023 =−1 ≈.089. Jaccard 9 Note that while hfc0006 shares three sign types with both hfc0015 and spy0023, the Jaccard distance measure “penalizes” the fact that there are overall more sign types occurring in hfc0006 and hfc0015 together (nine), compared to hfc0006 and spy0023 together (seven). Tus, the Jaccard distance is higher for the former. Te rationale behind this is that if there are many diferent types occurring in two vectors, then it is more likely that the same types occur in both by chance. Based on pairwise Jaccard distances we create a distance matrix for these three objects as below  ..  0067 057 D = 06..70089. Jaccard   05..7089 0  We can then use this distance matrix for cluster analysis.

Building a UPGMA tree. We here choose the so-called Unweighted Pair Group Method with Arithmetic mean (UPGMA) to create a clustering tree. UPGMA is an agglomerative bottom-up clustering method66. In the beginning, each “leaf” (object in our case) constitutes its own cluster. Given a distance matrix between clusters, the two clusters with the smallest distance are merged to yield a new cluster. Te average distance of this cluster to all other clusters is computed and compared to distances between the other clusters to decide the next merger. Te UPGMA algorithm thus successively merges clusters, until only one overall cluster (i.e. the fnal tree) is formed. Importantly, we here merely use this method to visualize the clustering of objects based on the Jaccard distances of their sign type presences. We do not claim that this clustering refects actual evolutionary relationships between objects and their sign types. For further details on calculating Jaccard distances and building UPGMA trees see the R code in the fles Supplementary File 4 as well as Supplementary File 5.

Scientific Data | (2020)7:364 | https://doi.org/10.1038/s41597-020-00704-x 7 www.nature.com/scientificdata/ www.nature.com/scientificdata

Notch Line 55 55 a)a) b)b) 50 50

45 45 lat lat

40 40

35 35

30 30 -10 010203040 -100 10 20 30 40 lon lon

Cross Obline 55 55 c)c) d)d) 50 50

45 45 lat lat

40 40

35 35

30 30 -10 010203040 -100 10 20 30 40 lon lon

Hatching Dot 55 55 e)e) f)f) 50 50

45 45 lat lat

40 40

35 35

30 30 -10 010203040 -100 10 20 30 40 lon lon

Vulva Grid 55 55 g)g) h)h) 50 50

45 45 lat lat

40 40

35 35

30 30 -10 010203040 -100 10 20 30 40 lon lon

Fig. 6 Maps for the presence/absence of particular sign types. Black triangles indicate archaeological sites where artifacts carrying geometric signs were found. Red triangles indicate the presence of a particular sign type (for script see Supplementary File 3).

Given the Jaccard distance matrix of our three example objects, the UPGMA method yields the tree in Fig. 7. In this simple example, the object from Belgium (spy0023) is frst merged with one of the objects from Hohle Fels in Germany (hfc0006), since they have the lowest distance in terms of sign type presence (0.57). Te second object from Hohle Fels (hfc0015) is then joined with them in the second step, yielding the overall tree. Te same method is applied to all 516 objects (excluding objects which are coded as displaying only the sign type “other”) and their Jaccard distances to generate the UPGMA tree in Fig. 8. Tis gives a general impression of how objects from sites in diferent countries cluster together based on the presence/absence of particular sign types. Te notch, for instance, is not only widespread (as was pointed out with reference to Fig. 6), but it is also ofen the only sign type present on objects. Tis is refected in the large cluster spanning the lower right quarter of the

Scientific Data | (2020)7:364 | https://doi.org/10.1038/s41597-020-00704-x 8 www.nature.com/scientificdata/ www.nature.com/scientificdata

0.29 spy0023

0.1 absent

0.29 present hfc0006 0

a Belgium 0.39 hfc0015 a Germany

t line oblineradlinedashlineconcenlinedo cross v hatchingzigzagro

Fig. 7 Example of a UPGMA tree for three objects. Te object identifers are given on the tips of branches. Colors indicate the country of provenience. A heatmap is given on the right of the tree with black indicating the presence of a sign type, and white indicating its absence for a given object. Only the sign types which occur in at

least one of the three objects are considered here (for script see Supplementary File 4). 8

4 2

35 2

1 2

y0025

f y0016

a s vhc0065 vhc0088

vhc0053

tdr0002 vhc0063

l sol0007

c cel0014 ztv0003

vhc00

cel0019 olv0001

cel0003 laf0022

laf0010 sp hfc003

cel0027 y0022

cel0025

0026 py vhc0099

cel0021 gkl0001 vhc0020

cel0029

cel0020

men0002 vhc0069

0023

py vhc0080 y0019 hfc0043

s vhc0027

vhc001 hfc0006

s ztv0002

s vhc0052

trm0004

3 sal0001hfc0048

hfc0005 sp

hfc0015 y0011

003

9 vhc015

y0027 laf002gdr00053

gkl0011 vhc0039

gkl0012 vhc0040

trm0012 vhc0042

sp vhc0170

sp vhc0068

hfc0009 vhc0083

gkl0028 vhc0025 trm0011

ndd0 vhc0171

vhc0095 hfc003

sh0001 vhc0024

sol0006

gpp0003 vhc0134

w vhc0001

1 psc000 vhc0108

vhc0011 btb0001

01 vhc0017

vhc0152 vhc0016 0004 4 vhc014

19 gfh0001

vdj0001 vhc0051 ndd0004 0010

sdn0001 sou0001

gdg0003 s

py sou0002

ecs00 cas0002

s cas0005

cas00 bla0018 2 9 vhc005

vhc0015 bla0013

y0009 bla0002w0001 vhc0008

gdg0008 vhc0041r

p b

s lus0003

stn0001 bla0014

hfc004 hfc0003

gpp00 tdc0001 brs0002

3 vhc000 vhc0147

vhc0087

cel0006 absent

men0003 cel0017

vhc014 cel0022

vhc015 cel0016

3 vhc015

004 cel0013

01 r0002

3 vhc014 cel0015

39 bla0019

r0 ve bla0015

laf0018 r00

ve bla0016

vhc01 laf0013

trm0007 laf0009 8

vhc012 laf0015 4

vhc011 cas0006 trm0009 present

cas0007

trm0008 bla0010

ilk0002 bla0009

kvs0001 bla0011 vhc0032

vhc0056 cel0002

cel0018 4

vhc004 poi0001 vhc0093

8 vhc009 lus0002

lus0001

gar0005 2

gar0008 lus0004

bla0008 6

gkl000 laf0011 laf001

4 hfc0050

gkl0026 hfc0035

gkl0005 hfc0002

bla000

gdr0007 08 hfc0049

hfc0034

gdr0001 hfc0022

gdr00 hfc0014

gdr0009 hfc0020 6

cas001 hfc0013

cas0011 hfc0023

ecs0003 hfc0037

cas0017 spy0005

cas0022 gdr0010 3

cas001 cas0012 a

cas0021 gdg0014 3

lak000 vhc0172 Belgium

lak0002 vhc0174

laf0006 vhc0058

laf0020

oi0001 ztv0001 r

y0021 ndd0001

sp vhc0142 laf0024

y0001

1 p vhc0034

s hfc0036 hfc001

27 hfc0017 a

laf00 msc0001

gdg0010 hdp0001 Croatia

gdg0006 vhc0090

gdg0004 bst0003

hfc0026 ltp0001

hfc0016 t0002

cross co notch

hfc0031 hfc0030 vulv

hfc0040 hfc0010

hfc0039 hfc0041 a

hfc0038 hfc0021 6

vhc011 hfc0012 Czech Republic

vhc0162 brs0001 line

gkl0025 gdr0002 6

vhc016 gkl0022

hfc0025 cot0001 5

vhc016 vhc0048

gpp0004 hfc0019

gdg0005 bla0012 a 5

vhc013 ecs0002 8

vhc016 taw0001 France

vhc0013 msc0006

trm0013 lus0006 vhc0036

vhc0089 hss0001

t0002 laf0007

vhc0091 hfc0001

hfc0004 hfc0045

vhc0005 gdg0011 a

s p

y0018 hfc0007 Germany

vhc0163 hfc0024

hfc0029 sol0001

vhc0120 sol0002

vhc0106 ras0001

vhc0028 pat0003

vhc0097 pat0001

vhc0074 pat0002

gdg0002 spy0024 a

gpp0002 spy0015

vhc0155 spy0017 Hungary

vhc0145 spy0003

vhc0158 spy0014

vhc0156 laf0019

vhc0002 laf0008

5 bla000 gkl0030

vhc0022 laf0025

vhc0107 laf0004 a

vhc0019 mla0002

lar0002 mla0001 Iraq

sp lar0001 y0012

mur0001 vin0001

hfc0018 cel0031

sp cas0010 y0006

hfc0027 cas0018

tdr0001 cel0004

bla0007 cel0012

bla0001 brs0003 a

gzr0003 bs 4

vhc016 bla0006t0001 Israel

mur0002 cas0015

mla0004 bla0003

vhc0102 gkl0017

6 vhc004 cel0030

vhc0160 gkl0024

trm0002 gkl0014

2 tdc000 gkl001

vhc0132 gkl0021 a

vhc0157 gkl0020 8

trm0006 gkl0019 Italy

vhc0026 gkl0027

vhc0029 gkl0029

1 fnr000 gkl0004

gzr0001 gkl0007

gzr0002 gkl0013

vhc0084 gkl0016

egh0001 gkl0003

vhc0060 gkl0015

gar0011 gkl0008 a

gar0012 gkl0009

gdg0013 gkl0010

cel0009 Poland

gar0004

aur0001 gkl0002

cas0003 gar0006

cas0008 cel0011

cas0020 cel0010

cas0001 cel0008

cas0014 cel0007

cas0009 cel0005

mla0003 g

laf0026 gar0003ar0010 a

gdg0012 gar0007

laf0005 g

hfc0028 g ar0009 Russia

hfc0044 fum0003ar001

hfc0046 vhc0092

hss0002 vhc0086

hfc0047 vhc0101

vhc0043 h 3

vhc0049 vhc0100ym0001

vhc0150 msc0005

vhc0146 vhc0105

vhc0112 vhc0103

vhc0081 vhc0062

vhc0096 vhc0061

vhc0094 vhc0054 a

vhc0131 vhc0045

0001

g vhc0064

cel0001 vhc0038

laf0021 Spain vhc0071

sgt0001 vhc0070

gdg0009 vhc0050

vhc0117 vhc0067

msc0002 h

msc0007 vhc0033

vhc0127 vhc005ym00095

vhc0126 vhc0057

vhc0004 vhc0066

vhc0076 rbb0003

hfc0008 vhc0082

gkl0023 vhc0078

vhc0010 pdh0001

vhc0014 rbb0001

vhc0018 vhc0085

vhc0021 vhc0030

vhc0006 vhc0031

vhc0007 vhc0079

vhc0009 sda0001

vhc0167 vhc0175

vhc0075 vhc0023

vhc0077 h

vhc0173 rbb0002

vhc0047 ltk0001

s h ym0003

vhc0118 hym0004

vhc0151 h vhc0125

py h vhc0148 ym0002 vhc0161 h

vhc0169 ilk0001 vhc0037

vhc0133 ym0005 vhc0136 ilk0003

vhc0140 ym0006

vhc0138

vhc0130 trm0001 ym0007

vhc0121 vhc0115 ym0008

trm0003 vhc0113

ve vhc0110

vhc0119 vhc0123

0020

vhc0129 vhc0109

vhc0104 vhc0124 vhc0073 vhc0137 trm0010 vhc0122

sp vhc0141

p r0003

y0007 y0008

Fig. 8 UPGMA tree for 516 Aurignacian objects and sign types. Tis tree is based on Jaccard distances of sign type presences/absences between pairs of objects. Only some sign types are represented in the presence/absence heatmap around the tree tips (line, notch, vulva, cross). With all sign types included, the plot would be too crowded (for code see Supplementary File 5).

circular tree in Fig. 8. Te objects in this cluster come from a wide range of sites – refected by the diferent colors representing modern-day countries. A similar picture emerges for the simple line (mainly upper and lower right quarter of the plot). It is also widely represented across sites, and ofen the only sign type present on an object. For geometric vulva representations, we fnd the opposite. Tey are exclusively found in France (Dordogne), and the objects carrying them mainly cluster together (upper right corner of the plot). Tere are only three objects carrying vulvae that diverge from this cluster since they carry other sign types (e.g. lines and notches) as well. Te cross, on the other hand, is represented in various clusters. Tere is the main cluster of crosses containing mainly objects from Vogelherd Cave (vhc, upper right corner), but there are also other smaller clusters involving objects from France and Belgium.

Scientific Data | (2020)7:364 | https://doi.org/10.1038/s41597-020-00704-x 9 www.nature.com/scientificdata/ www.nature.com/scientificdata

However, we do here not further delve into the issues of statistical analyses, hypotheses testing, and interpre- tation of clusters. Tese are topics for future studies using SignBase. Data Records Te data sets used for analyses in this article are available at fgshare67: Dutkiewicz, Russo, Lee, & Bentz. SignBase: collection and analysis of geometric signs on mobile objects in the Paleolithic. fgshare https://doi.org/10.6084/ m9.fgshare.c.4898643 (2020). Te names of data set fles are: - signBase_exampleObjects.csv (fgshare title: “Example Objects”) - signBase_Version1.0.csv (fgshare title: “SignBase Main Data File”) - TestCoding_000.csv (fgshare title: “Test Coding (Coder 0)”) - TestCoding_001.csv (fgshare title: “Test Coding (Coder 1)”) - TestCoding_002.csv (fgshare title: “Test Coding (Coder 2)”) - TestCoding_003.csv (fgshare title: “Test Coding (Coder 3)”) - TestCoding_004.csv (fgshare title: “Test Coding (Coder 4)”) Te main data of the current version of SignBase is given in signBase_Version1.0.csv. As the data set is going to grow, newer versions will be available via www.signbase.org. Te column names are given in parentheses and described in the following: • Microsof Access ID (access_id): Tis is an internal ID used by the Microsof Access database that the online version of SignBase is based on. For the general public, this ID is not important. • Object identifer (object_id): Te object ID is created by a three-letter (lower case) abbreviation of the site name, followed by a four-digit running number, e.g. for objects from La Ferrassie: laf0001, laf0002, etc. • Techno-complex (techno_complex): Entities in prehistory are based on material cultures. An archaeological culture is a recurring assemblage of artifacts from a specifc time and place that may constitute the material culture remains of a past human society. For the Paleolithic, there are mostly artifacts made from lithic or organic material. Characteristic artifact types or assemblages of artifacts defne the entity, here referred to as techno-complex. Tis might be, for instance, the Aurignacian, the Gravettian, or the Magdalenian in Europe, or the Middle Stone Age or the Later Stone Age in Africa. • Site and location (site_name, location, country, longitude, latitude): All the objects in SignBase are fnds from archaeological excavations. Tese may be caves or open-air sites. Te archaeological site (site_name) is indicated, as well as the closest community or town (location) and the country. For an exact location, the latitude and longitude are also given. • Layer (layer): In archaeological excavations, objects are found in units of sediments, which are called stratigraphical units or layers. Tese units are usually defned by the archaeologists during excavation and give information about the provenience of the object and its assignment to a particular techno-complex. • Dating method (dating_method): Tis describes the method that has been used for dating the fnd. For absolute dating in Prehistory, mainly radiocarbon dating is used (C14). Other ofen-used methods are Accelerator Mass Spectrometry (AMS), Termoluminescence (TL), optically stimulated luminescence (OSL), or Ura- nium–thorium dating. • Date (date_max-min): Te absolute dating (uncalibrated before present/BP) of the object or layer it derives from as given in the literature. • Excavation year (excavation_year): Te year the object was excavated. • Material (material): Te raw material of the object. For the Paleolithic, mostly osseous material, like ivory, bone, or antler is used. Other organic materials, like shells of mollusks, eggshells, or teeth, but also inorganic materials like rocks, pigments, or ceramics might appear. If the material has not been undoubtedly determined, this feature takes the value undetermined. • Type of object (object_type): Typological determination of the object, usually as indicated in the literature, or revised, if needed. Might be tools, personal ornaments, art fgurines, or any other type of archaeological object. If the type of object has not been undoubtedly determined, this feature takes the value undetermined. • Length, width, depth (length_mm, width_mm, depth_mm): Gives the dimensions of the object in millime- ters. Usually as indicated in the literature. • Preservation (preservation): Te preservation state of the object. Complete – the whole object is preserved; almost complete – the whole form of the object is preserved, with some damage; fragmented – only partly preserved, the original dimension and shape of the object is not preserved. If the preservation has not been undoubtedly determined, this feature takes the value undetermined. • Short description (short_description): Tis is a very brief (mostly one sentence) impressionistic description of the type of object and – in some cases – the respective signs represented on it. Tis mostly follows the description authors use in the original articles publishing the object. • General literature (general_literature): Literature references about the object, its provenience, excavation report, or dating of the object/stratigraphical unit. Technical Validation Establishing a sign type and its presence/absence on a particular object is a non-trivial task. Decisions are based on subjective judgment. We hence expect some disagreement between our “expert” coding and the potential coding of other researchers in the feld. To get a frst impression of the degree of subjectivity in our coding decisions, we have submitted 30 randomly chosen objects (of the 531 overall objects) to four colleagues at the University of

Scientific Data | (2020)7:364 | https://doi.org/10.1038/s41597-020-00704-x 10 www.nature.com/scientificdata/ www.nature.com/scientificdata

SignBase Coding Agreement with Coder 1 Agreement with Coder 2 Agreement with Coder 3 Agreement with Coder 4 vhc0163 vhc0158 vhc0144 vhc0139 vhc0134 vhc0095 vhc0076 vhc0015 trm0013 trm0003 spy0027 spy0007 pat0003 msc0007 lar0002 laf0012 hfc0017 object_id gkl0025 gkl0008 gdr0005 gdg0012 gdg0003 cel0030 cas0019 cas0009 cas0001 Agree.: 93 % Agree.: 94 % Agree.: 91 % Agree.: 93 % bla0019 bla0015 bla0005 Kappa: 0.44 Kappa: 0.44 Kappa: 0.29 Kappa: 0.35 aur0001 l l l l l t t t t t t t t t t s s s s s v v v v v e e e e e a a e e e a e e a a e e e e e w w w w w tch tch tch tch tch tch tch tch tch tch do do do do do lin e lin e lin e lin e lin e star star grid grid star grid grid star grid star pa pa pa pa pa other other other other other vulv vulv cros cros vulv cros cros vulv cros vulv notch notch notch notch notch pinlef pinlef pinlef oblin e oblin e oblin e oblin e oblin e pinlef pinlef zigza g cupule zigza g zigza g cupule zigza g cupule cupule cupule zigza g radlin radlin radlin radlin radlin pinright pinright pinright pinright pinright hashta g hashta g obno hashtag obno hashta g obno obno obno hashtag dashlin e dashlin e dashlin e dashlin e dashlin e hatchin g hatchin g hatchin g hatchin g hatchin g radnotch radnotch rhombus rhombus rhombus radnotch radnotch rhombus rhombus radnotch rectangl rectangl rectangl rectangl rectangl zoomorph zoomorph zigzagrow zigzagrow zoomorph zigzagrow zigzagrow circumlin circumlin zoomorph zigzagrow circumlin circumlin zoomorph circumlin maccaroni maccaroni maccaroni maccaroni maccaroni concenlin e concenlin e concenlin e concenlin e concenlin e circumspira circumspira circumspira circumspira circumspira circumno circumno circumno circumno circumno anthropomorph anthropomorph anthropomorph anthropomorph anthropomorph

Fig. 9 Evaluation of coding. 30 randomly selected objects with object identifers given on the x-axis. In the lef panel, the presence of a given type (y-axis) is given in black, grey tiles indicate the absence of the type according to the original coding by the SignBase team. Te plots to the right illustrate the agreement with test coders. Green indicates coding agreement, red indicates disagreement. Percentages of agreeing tiles, as well as Cohen’s Kappas, are given in the lower right corner (for code see Supplementary File 6).

Tübingen, who are familiar with the archaeological material. Familiarity with the objects and the respective literature is necessary since judging surface patterns is not possible without an understanding of their characteristic texture. Tis is particularly important given that coding decisions have to be taken based on pictures (photo- graphs and/or drawings, in some cases of low resolution). Disturbances of the material due to natural processes (cracks, wholes, bite marks, etc.), as well as functionally motivated adaptations, can be easily misinterpreted as intentional geometric signs by a layperson. Having said this, we did not explicitly teach the four test coders how to determine sign type presence. We merely provided them with the same interface used by the SignBase team to make coding decisions. In this interface, a picture of the respective object, alongside the meta-information described above, is provided. Furthermore, there are 31 predefned sign types to choose from, including a category “other”, in case none seem to ft. Test coders were instructed to choose the sign types they identify on any given object. While several diferent sign types can occur on a single object, each sign (i.e. pattern on the object’s surface) should only be associated with one sign type. For example, if a line is visible, the coder must decide whether it is a regular (i.e. straight) line or an oblique line. Te same pattern cannot be coded as both. Given the original expert coding by the SignBase team (Test Coding (Coder 0)), and the coding by four test coders (Test Coding (Coder 1), Test Coding (Coder 2), Test Coding (Coder 3), Test Coding (Coder 4)), we frstly calculate the so-called joint-probability agreement, namely, simply the percentage-wise overlap in coding decisions. Secondly, we calculate Cohen’s Kappa68. Tis is a more conservative metric of agreement devised to consider that for any given number of coding decisions, coders might agree just by chance. For example, if coders just choose uniformly at random between presence and absence for all the sign types, the expected agreement of coding decisions between two coders is already 50%. Te Kappa metric takes this chance agreement as a baseline. A Kappa value of 0 indicates that there is no coder agreement beyond that predicted by chance, while a Kappa of 1 indicates that there is perfect coder agreement. Additionally, the R package we use to calculate Kappa also provides a p-value. If this p-value is <0.05 for a given Kappa value, then it is signifcantly bigger than 0. In Fig. 9, the joint-probability agreement (Agree.), as well as Kappa values for pairwise comparisons between the original SignBase coding and the four test coders, is illustrated. Te joint-probability agreement ranges from 91% to 94% between our original coding and the coding by test coders. Cohen’s Kappa ranges from 0.29 to 0.44. While there is hence high coding agreement according to the joint-probability measure, the Kappa values indicate that the agreement between coders is rather moderate. However, given that coders were not explicitly instructed on how to make their decisions, even medium range Kappas are encouraging. Te p-values for all Kappas are <0.05 (in fact, they are output as 0), meaning that there is clearly more agreement between coders than expected by chance. Te R code with further explanations is given in Supplementary File 6. Code availability Te R code to replicate the analyses of this article can be found in a series of R markdown fles and at fgshare67: Dutkiewicz, Russo, Lee, & Bentz. SignBase: collection and analysis of geometric signs on mobile objects in the Paleolithic. fgshare https://doi.org/10.6084/m9.fgshare.c.4898643 (2020).

Figure 2: Supplementary File 1 and fgshare File 1 Figure 5: Supplementary File 2 and fgshare File 2 Figure 6: Supplementary File 3 and fgshare File 3 Figure 7: Supplementary File 4 and fgshare File 4 Figure 8: Supplementary File 5 and fgshare File 5 Figure 9: Supplementary File 6 and fgshare File 6

Received: 5 May 2020; Accepted: 29 September 2020; Published: xx xx xxxx

Scientific Data | (2020)7:364 | https://doi.org/10.1038/s41597-020-00704-x 11 www.nature.com/scientificdata/ www.nature.com/scientificdata

References 1. d’Errico, F. et al. Archaeological Evidence for the Emergence of Language, Symbolism, and Music - An Alternative Multidisciplinary Perspective. Journal of World Prehistory 17, 1–70 (2003). 2. Joordens, J. C. A. et al. Homo erectus at Trinil on Java used shells for tool production and engraving. Nature 518, 228–231 (2015). 3. White, R. et al. Context and dating of Aurignacian vulvar representations from Abri Castanet, France. Proceedings of the National Academy of Sciences 109, 8450–8455 (2012). 4. Higham, T. et al. Τesting models for the beginnings of the Aurignacian and the advent of fgurative art and music: Te radiocarbon chronology of Geißenklösterle. Journal of Human Evolution 62, 664–676 (2012). 5. Conard, N. J. & Bolus, M. Radiocarbon dating the appearance of modern humans and timing of cultural innovations in Europe: new results and new challenges. Journal of Human Evolution 44, 331–371 (2003). 6. Conard, N. J. & Bolus, M. Radiocarbon dating the late Middle Paleolithic and the Aurignacian of the Swabian Jura. Journal of Human Evolution 55, 886–897 (2008). 7. Jöris, O., Neugebauer-Maresch, C., Weninger, B. & Street, M. Te Radiocarbon Chronology of the Aurignacian to Mid-Upper Palaeolithic Transition along the Upper and Middle Danube. In New Aspects of the Central and Eastern European Upper Palaeolithic – Methods, Chronology, Technology and Subsistence. Symposium by the Prehistoric Commission of the Austrian Academy of Sciences; Vienna, November 9-11, 200 (eds. Neugebauer-Maresch, C. & Owen, L.) vol. 72 (Mitteilungen der Prähistorischen Kommission, 2010). 8. Jöris, O. & Street, M. At the end of the 14C time scale - the Middle to Upper Palaeolithic record of western Eurasia. Journal of Human Evolution 55, 782–802 (2008). 9. Nigst, P. R. et al. Early modern human settlement of Europe north of the Alps occurred 43,500 years ago in a cold steppe-type environment. Proceedings of the National Academy of Sciences 111, 14394–14399 (2014). 10. Dutkiewicz, E. & Conard, N. J. The symbolic language of the Swabian Aurignacian as reflected in the material culture from Vogelherd Cave (South-West Germany). In L’art au quotidien. Objets ornées du Paléolithique supérieur. Actes du colloque international Les Eyzies-de-Tayac, 16-20 juin 2014 (eds. Cleyet-Merle, J.-J., Geneste, J.-M. & Man-Estier, E.) vol. numéro spécial 149–164 (2016). 11. Tartar, E. Origin and Development of Aurignacian Osseous Technology in Western Europe: a Review of Current Knowledge. In Aurignacian Genius: Art, Technology and Society of the First Modern Humans in Europe. Proceedings of the International Symposium, April 08-10 2013, New York University (eds. White, R., Bourrillion, R. & Bon, F.) vol. 7 33–55 (2015). 12. Tartar, E., Teyssandier, N., Bon, F. & Liolios, D. Équipment de chasse, équipment domestique: une distinction efcace? Réfexion sur la notion d’investissement technique dans les industries aurignaciennes. In Normes techniques et practiques sociales. De la simplicité des outillages pré- et protohistoriques. XXVIe rencontres internationales d’archéologie et histoire d’Antibes (eds. Astruc, L., Bon, F., Léa, V., Milcent, P.-Y. & Philibert, S.) 107–117 (Éditions APDCA, 2006). 13. Wolf, S. Schmuckstücke – Die Elfenbeinbearbeitung im Schwäbischen Aurignacien. (Kerns Verlag, 2015). 14. Wolf, S., Münzel, S., Dotzel, K., Barth, M. & Conard, N. J. Osseous Projectiles from the Aurignacian and the Gravettian of the Swabian Jura (Southwest Germany) refect changing patterns of raw material, technology and typology. In Osseous Projectile Weaponry: Towards an Understanding of Pleistocene Cultural Variability (ed. Langley, M. C.) vol. VERT Series 71–87 (Springer, 2016). 15. Bolus, M. & Conard, N. J. Zur Zeitstellung von Geschossspitzen aus organischen Materialien im späten Mittelpaläolithikum und Aurignacien. Archäologisches Korrespondenzblatt 36, 1–15 (2006). 16. Dutkiewicz, E., Wolf, S., Floss, H. & Conard, N. J. Les objets en ivoire du Jura souabe. L’Anthropologie 122, 447–468 (2018). 17. Tejero, J.-M. & Grimaldi, S. Assessing bone and antler exploitation at Riparo Mochi (Balzi Rossi, Italy): implications for the characterization of the Aurignacian in South-western Europe. Journal of Archaeological Science 61, 59–77 (9). 18. Julien, M., Bafer, D. & Liolios, D. L’outillage en matières dures animales. In L’Aurignacien de la Grotte du Renne. Les fouilles d’André Leroi-Gourhan à Arcy-sur-Cure (Yonne) (ed. Schmider, B.) vol. XXXIVe supplément à Gallia Préhistoire 217–250 (2002). 19. Liolios, D. Variabilité et caractéristiques du travail des matières osseuses au début de l’Aurignacien: approche technologique et économique. (Université de Paris X-Nanterre, 1999). 20. Liolios, D. Refexions on the role of bone tools in the defnition of the Early Aurignacian. In Towards a defnition of the Aurignacian. Actes du Symposium de Lisbonne, 25-30 juin 2002 (eds. Bar-Yosef, O. & Zilhão, J.) vol. 45 (Instituto Portugues de Arqueologia, 2006). 21. Liolios, D. Les instruments osseux. In Les Aurignaciens (ed. Otte, M.) 137–151 (Éditions Errance, 2010). 22. Hahn, J. Aurignacien, das ältere Jungpaläolithikum in Mittel- und Osteuropa. (Böhlau, Verlag, 1977). 23. Peirce, C. S. Collected Papers of Charles Sanders Peirce. (Harvard University Press, 1931). 24. Peirce, C. S. Te Essential Peirce: Selected Philosophical Writings/edited by the Peirce Edition Project. vol. 2 (1893-1913) (Indiana University Press, 1998). 25. Sauvet, G. Les signes dans l’art mobilier. In Les objets au Paléolithique supérieur. Tome 2: les voies de la recherche. Colloque de Foix-Le Mas d’Azil (1987) (ed. Clottes, J.) 83–99 (Dir. Patrimoine, 1990). 26. Sauvet, G. Les signes pariétaux. In L’art pariétal paléolithique. Techniques et méthodes d’étude (ed. GRAPP) 219–234 (CTHS, 1993). 27. Breuil, A. H. Quatre cent siècles d’art pariétal. Les cavernes ornées de l’age du renne. (Fernand Windels, 1952). 28. Leroi-Gourhan, A. Préhistoire de l’art occidental. vol. 1 (Ed. d’Art Lucien Mazenod, 1965). 29. Clottes, J. & Lewis-Williams, D. Te Shamans of Prehistory. Trance and Magic in the Painted Caves. (Harry N. Abrams, 1998). 30. von Petzinger, G. Te First Signs. Unlocking the Mysteries of the World’s oldest Symbols. (Atria Paperback, 2016). 31. Hahn, J. Kraf und Aggression. Die Botschaf der Eiszeitkunst im Aurignacien Süddeutschlands? (Verlag Archaeologia Venatoria, 1986). 32. Dutkiewicz, E., Wolf, S. & Conard, N. J. Early symbolism in the Ach and the Lone valleys of southwestern Germany. Quaternary International 491, 30–45 (2018). 33. Texier, P.-J. et al. Te context, form and signifcance of the MSA engraved ostrich eggshell collection from Diepkloof Rock Shelter, Western Cape, South Africa. Journal of Archaeological Science 40, 3412–3431 (2013). 34. Henshilwood, C. S., d’Errico, F. & Watts, I. Engraved ochres from the Middle Stone Age levels at Blombos Cave, South Africa. Journal of Human Evolution 57, 27–47 (2009). 35. Conkey, M. W. et al. Te Identifcation of Prehistoric Hunter-Gatherer Aggregation Sites: Te Case of Altamira [and Comments and Reply]. Current Anthropology 21, 609–630 (1980). 36. Tylén, K. et al. Te evolution of early symbolic behavior in Homo sapiens. Proc Natl Acad Sci USA 117, 4578 (2020). 37. Otte, M. Le Paléolithique Supérieur Ancien en Belgique. (Musées Royaux d’art et d’histoire, 1979). 38. Chollot-Varagnac, M. Les origines du graphisme symbolique. Essai d’analyse des écritures primitives en Préhistoire. (Éditions de la Fondation Singer-Polignac, 1980). 39. Les chemins de l’art Aurignacien en Europe. Das Aurignacien und die Anfänge der Kunst in Europa. Colloque international, internationale Fachtagung, Aurignac, 16-18 septembre 2005. (2007). 40. Bourrillon, R. et al. A new Aurignacian engraving from Abri Blanchard, France: Implications for understanding Aurignacian graphic expression in Western and Central Europe. Quaternary International https://doi.org/10.1016/j.quaint.2016.09.063 (2017). 41. White, R. Production complexity and standardisation in early Aurignacian bead and pendant manufacture: evolutionary implications. In Te human Revolution. Behavioural and Biological Perspectives in the Origins of Modern Humans (eds. Mellars, P. & Stringer, C.) 332–390 (Princeton University Press, 1989).

Scientific Data | (2020)7:364 | https://doi.org/10.1038/s41597-020-00704-x 12 www.nature.com/scientificdata/ www.nature.com/scientificdata

42. Delluc, B. & Delluc, G. Les manifestations graphiques aurignaciennes sur support rocheux des environs des Exzies (Dordogne). vol. 21 (Éditions du CNRS, 1978). 43. Knecht, H. & White, R. Te Abri Cellier (or La Ruth), Commune de Tursac (Dordogne, France). Results of the 1927 Beloit College Excavations. in French Paleolithic Collections in the Logan Museum of Anthropology (eds. Breitborde, L. B. & White, R.) vol. 1 (1992). 44. San Juan-Foucher, C., Vercoutère, C. & Foucher, P. Parures et objets décorés auriganciens de la Grotte de Gargas (Hautes Pyrénées, France). Schmuck und verzierte Objekte aus dem Aurignacien der Höhle Gargas (Hautes Pyrénées, Frankeich). In Les chemins de l’art Aurignacien en Europe. Das Aurignacien und die Anfänge der Kunst in Europa. Colloque international, internationale Fachtagung, Aurignac, 16-18 septembre 2005 (eds. Floss, H. & Rouquerol, N.) vol. Cahier 4 89–104 (2007). 45. Mussi, M., Gioia, P. & Negrino, F. Ten small sites: the diversity of the Italian Aurignacian. In Towards a Defnition of the Aurignacian (eds. Bar-Yosef, O. & Zilhão, J.) vol. Towards a Defnition of the Aurignacian. Trabalhos de Arqueologia 45 (American School of Prehistoric Research/Instituto Português de Arqueologia, 2006). 46. Bartolomei, G. et al. La Grotte de Fumane. Un site aurignacien au pied des Alpes. Preistoria Alpina - MuseoTridentino di Scienze Naturali 28 (1992), 131–179 (1994). 47. d’Errico, F., Julien, M., Liolios, D., Vanhaeren, M. & Bafer, D. Many awls in our argument. Bone tool manufacture and use in the Châtelperronian and Aurignacian levels of the Grotte du Renne at Arcy-sur-Cure. In Te Chronology of the Aurigancian and the Transitional Technocomplexes. Dating, Stratigraphies, Cultural Implications. Proceedings of Symposium 6.I of the XIVth Congress of the UISPP (University of Liège, Belgium, September 2-8, 2001) (eds. Zilhão, J. & d’Errico, F.) vol. 33 247–270 (Instituto Português de Arqueologia, 2003). 48. L’Aurignacien de la Grotte du Renne. Les fouilles d’André Leroi-Gourhan à Arcy-sur-Cure (Yonne). vol. XXXIVe supplément à Gallia Préhistoire (CNRS Éditions, 2002). 49. Zilhão, J. Te Emergence of Ornaments and Art: An Archaeological Perspective on the Origins of “Behavioral Modernity”. Journal of Archaeological Research 15, 1–54 (2007). 50. Valoch, K. Paläolithische Archäologie in der ehemaligen Tschechoslovakei und ihr Beitrag zur mitteleuropäischen Forschung. Mitteilungen der Gesellschaf für Urgeschichte 19, 71–115 (2010). 51. Lejeune, M. Le Trou Magrite et l’art mobilier aurignacien en Belgique: synthèse et problems. Das Trou Magrite und die aurignacienzeitliche Kleinkunst in Belgien: Synthese und Probleme. In Les chemins de l’art Aurignacien en Europe. Das Aurignacien und die Anfänge der Kunst in Europa. Colloque international, internationale Fachtagung, Aurignac, 16-18 septembre 2005 (eds. Floss, H. & Rouquerol, N.) vol. Cahier 4 131–144 (2007). 52. Pirson, S. et al. Te stratigraphy of Spy cave. A review of the available lithostratigraphic and archaeostratigraphic information. Anthropologica et Praehistorica 123, 91–131 (2013). 53. Tejero, J.-M., Belfer-Cohen, A., Bar-Yosef, O., Gutkin, V. & Rabinovich, R. Symbolic emblems of the Levantine Aurignacians as a regional entity identifer (Hayonim Cave, Lower Galilee, Israel). Proceedings of the National Academy of Sciences 115, 5145 (2018). 54. Vanhaeren, M. & d’Errico, F. Aurignacian ethno-linguistic geography of Europe revealed by personal ornaments. Journal of Archaeological Science 33, 1105–1128 (2006). 55. Barandiarán, I. & García Diez, M. Les débuts du graphisme paléolithique dans le Nord de la péninsule Ibérique. Die Anfänge graphischer Gestaltung im Paläolithikum des Nordens der Iberischen Halbinsel. In Les chemins de l’art Aurignacien en Europe. Das Aurignacien und die Anfänge der Kunst in Europa. Colloque international, internationale Fachtagung, Aurignac, 16-18 septembre 2005 (eds. Floss, H. & Rouquerol, N.) vol. Cahier 4 (2007). 56. Henshilwood, C. S. et al. Emergence of Modern Human Behavior: Middle Stone Age Engravings from South Africa. Science 295, 1278–1280 (2002). 57. Henshilwood, C. S. & Dubreuil, B. Reading the artefacts: gleaning language skills from the Middle Stone Age in southern Africa. In Te Cradle of Language (eds. Botha, R. & Knight, C.) vol. 12 41–61 (Oxford University Press, 2009). 58. Posth, C. et al. Pleistocene Mitochondrial Genomes Suggest a Single Major Dispersal of Non-Africans and a Late Glacial Population Turnover in Europe. Current Biology 26, 827–833 (2016). 59. Fu, Q. et al. Te genetic history of Ice Age Europe. Nature 534, 200–205 (2016). 60. Rigaud, S., d’Errico, F. & Vanhaeren, M. Ornaments Reveal Resistance of North European Cultures to the Spread of Farming. PLOS ONE 10, e0121166 (2015). 61. Rigaud, S., Manen, C. & García-Martínez de Lagrán, I. Symbols in motion: Flexible cultural boundaries and the fast spread of the Neolithic in the western Mediterranean. PLOS ONE 13, e0196488 (2018). 62. González-Sainz, C., Ruiz-Redondo, A., Garate-Maidagan, D. & Iriarte-Avilés, E. Not only Chauvet: Dating Aurignacian rock art in Altxerri B Cave (northern Spain). Journal of Human Evolution 65, 457–464 (2013). 63. Hofmann, D. L. et al. U-T dating of carbonate crusts reveals Neandertal origin of Iberian cave art. Science 359, 912–915 (2018). 64. Pettitt, P. & Pike, A. W. G. Dating European Palaeolithic Cave Art: Progress, Prospects, Problems. Journal of Archaeological Method and Teory 14, 27–47 (2007). 65. van der Loo, M. P. J. Te stringdist Package for Approximate String Matching. Te R Journal 6, 111–122 (2014). 66. Huson, D. H., Rupp, R. & Scornavacca, C. Phylogenetic Networks: Concepts, Algorithms and Applications. (Cambridge University Press, 2010). 67. Dutkiewicz, E., Russo, G., Lee, S. & Bentz, C. SignBase: collection and analysis of geometric signs on mobile objects in the Paleolithic. fgshare https://doi.org/10.6084/m9.fgshare.c.4898643 (2020). 68. Cohen, J. A Coefcient of Agreement for Nominal Scales. Educational and Psychological Measurement 20, 34–46 (1960). Acknowledgements Tis project was funded by the DFG Center for Advanced Studies “Words, Bones, Genes, Tools” (DFG-KFG 2237). Open Access funding enabled and organized by Projekt DEAL. Author contributions E.D. was responsible for project inception, conceptualization, writing, and data collection. G.R. was responsible for data collection and drawing. S.L. helped with data and website management. C.B. contributed to project inception, conceptualization, statistical analyses, and writing. Competing interests Te authors declare no competing interests. Additional information Supplementary information is available for this paper at https://doi.org/10.1038/s41597-020-00704-x. Correspondence and requests for materials should be addressed to E.D. or C.B. Reprints and permissions information is available at www.nature.com/reprints.

Scientific Data | (2020)7:364 | https://doi.org/10.1038/s41597-020-00704-x 13 www.nature.com/scientificdata/ www.nature.com/scientificdata

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional afliations. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Cre- ative Commons license, and indicate if changes were made. Te images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.

Te Creative Commons Public Domain Dedication waiver http://creativecommons.org/publicdomain/zero/1.0/ applies to the metadata fles associated with this article.

Scientific Data | (2020)7:364 | https://doi.org/10.1038/s41597-020-00704-x 14