The Vinča signs

There is no single object. The material is several thousand incised marks on fired clay from more than 40 settlements of the Vinča culture, with the densest finds at Vinča-Belo Brdo, Turdaș, Banjica and Divostin⁠[21]⁠[9]. Andrej Starović's survey of 17 museum collections counted over 1,000 inscribed objects carrying over 1,500 signs; about 95 per cent of the well-recorded finds come from houses and their pits, and inscribed objects occurred in all 79 houses he examined⁠[21]. The carriers are mostly pots, with signs on the base, near the base or on the body; then spindle whorls, figurines, loom weights and a few lids, altars and tablets⁠[9]⁠[14]. Most signs were scratched into the clay before firing; from the Vinča B2 phase signs scratched after firing become common⁠[21].

There is no corpus edition. Winn's 1981 book is the standard sign list, built from Vinča, Turdaș, Banjica and Divostin and reducing 210 signs to about 30 core shapes with variants⁠[9]. Merlini's DatDas databank is far larger, 5,421 signs on 971 artifacts, but it has been published only as summaries and inventories in his 2009 monograph and on the prehistory.it site, not as data⁠[10]. The one open machine-readable corpus is Vinca.xml, encoded from Winn's catalogue by Wu, Solman, Linehan and Sproat in 2012 and released under the Apache 2.0 licence: 591 documents, 804 sign tokens and 185 sign types⁠[13]⁠[22].

A few objects carry more. The three Tărtăria tablets (two rectangular, one round and 6 cm across, two of them pierced) are in the National Museum of Transylvanian History in Cluj-Napoca⁠[23]. The Gradešnica plaque, a shallow clay tray 12.5 by 10.5 cm with signs in four registers inside and a figure with nine signs outside, is in the Vratsa museum⁠[18]. The Dispilio tablet is a wooden board from a Greek lake settlement, dated to 5260 ± 40 BC, still in conservation and never fully published⁠[19].

The dating is now secure for the culture, if not for every inscribed object. Bayesian models of radiocarbon dates from Vinča-Belo Brdo put the start of Vinča occupation in the 52nd century cal BC and the end in the second half of the 46th century, when two successive horizons of houses burned within about 25 years of each other⁠[1]⁠[2]. The culture as a whole runs from about 5400 to 4500 BC; Haarmann's "Danube script" extends the span to 3500 BC by including later Chalcolithic cultures⁠[12]. These dates make the Vinča signs older than proto-cuneiform, whose earliest tablets at Uruk date to about 3300 BC⁠[16].

The sign shapes are known and mostly simple. Winn's list runs from single strokes, crosses and chevrons through combs and meanders to compound signs; Haarmann counts 700 to 1,000 components if every variant is kept apart⁠[9]⁠[12]. Many recur across sites hundreds of kilometres apart. Comb and brush motifs, which make up about a sixth of all signs, are plausibly tallies⁠[3].

The statistics are known and they are unlike those of any script. In the open corpus 84 per cent of documents are a single sign, 9 per cent are two signs, and only four documents exceed eight signs; the most frequent sign (Winn 66) accounts for 6 per cent of tokens, and 72 of 185 types occur once⁠[13]. Mäder, working on the longer sequences, found 59 per cent of signs behaving as pottery marks (one sign, on the base, repeated within a site but not combined), 11.5 per cent as whorl ornament, and 29.5 per cent as candidates for a non-phonetic notation⁠[14]. Nothing is known of any sound values, any direction of writing or any numeral system beyond the tallies, because there is no text long enough or repeated enough to show them.

Three kinds of test have been applied. The first is comparison with known scripts. Vlassa and Hood matched Tărtăria signs with Uruk III and Jemdet Nasr signs of about 3000 BC⁠[6]⁠[7]. Calibrated radiocarbon then made the Vinča context too old for Mesopotamian influence, so either the match was coincidence or the tablets were intrusive; Neustupný set out the dilemma in 1968 and Renfrew made it the showcase of the radiocarbon revolution⁠[24]⁠[5]. Haarmann's matching of Vinča shapes with Linear A signs across a gap of two thousand years has not persuaded Aegean epigraphers; Owens, reviewing the Balkan material in Kadmos, found no link to the Aegean scripts and no text long enough to meet the usual criteria for writing⁠[25]⁠[20].

The second is the archaeological test of context. Lazarovici and Merlini re-examined the 1961 find and obtained radiocarbon dates of 6310 ± 65 BP for the human bones and 6210 ± 65 BP for animal bone from the adjacent pit house; they argued the tablets' fragility came from acid cleaning, not from firing in the museum, and that the find is a genuine early Vinča complex⁠[8]⁠[23]. The weakness is that Vlassa never published the stratigraphy, so the tablets are dated by association with bones he said were found with them⁠[26]. Erika Qasim went further and argued the tablets are modern copies of Sumerian signs from books available in Romania; the re-investigators reject this and it has no wider following⁠[27].

The third is statistical. Wu and colleagues encoded Winn's corpus precisely so that Vinča could serve as a known non-linguistic comparison for the Indus and Pictish entropy debates⁠[13]⁠[28]. Mäder's 2019 study is the only one to test the Vinča sequences themselves, by repetition, position on the object and comparison with pottery-mark systems; it is also the only study to find that Winn's list arbitrarily rotated and merged signs⁠[14]. Nobody has run the test the other way: a sign-sequence corpus checked against a null model of potters' marks of comparable size.

Five positions are live, and they are positions about what kind of thing the signs are, not readings of them. The first is that the signs are a writing system, the Old European or Danube script, able to record language or at least fixed meanings, developed independently in the Balkans before Mesopotamia⁠[11]⁠[12]⁠[10]⁠[21]. Its proponents point to the size of the inventory, the recurrence of shapes across sites, deliberate placement on objects and the domestic ubiquity Starović documented. No proponent has offered a reading, and the position's criteria (not decoration, not potters' marks, conventional, systematic) are negative ones that a mark system could also meet⁠[12]⁠[17].

The second is proto-writing: a conventional notation that conveyed meanings such as ownership, quantity or ritual category without encoding speech. This is Winn's own conclusion, built into his title, and it fits the tallies and the few repeated compounds⁠[9]. Peter Damerow's account of proto-cuneiform as a knowledge-recording system with little phonetic content shows what such a system looks like when it is large and administrative; the Vinča material has no equivalent of its numerals and totals⁠[29].

The third is that the marks are potters' and owners' marks, decoration and occasional tallies, made in a household economy that had no use for records; this is Chapman's and the Serbian excavators' view and Palavestra's⁠[15]⁠[17]. The fourth, Mäder's, is mixed: most signs are marks and ornament, but about a third of the longer sequences may be a notation of some kind, which is not writing⁠[14]. The fifth is that the Tărtăria tablets at least are something else, either intrusive objects of about 3000 BC showing Mesopotamian contact⁠[7] or modern⁠[27]. Pešić's 1987 reading of the signs as an alphabet ancestral to Etruscan and Cyrillic is not a scholarly position; it rests on shape resemblance alone⁠[17].

The question is answerable, because it is about the structure of a corpus rather than about an unknown language. What is needed first is an open sign-level corpus: every inscribed object with its site, phase, object type, position of the signs and sign sequence in a published order, with drawings, so that Winn's merging of variants and Merlini's much larger DatDas can be checked against each other⁠[9]⁠[10]⁠[14]. The Vinca.xml file is a start, but it covers four sites and reproduces Winn's rotations⁠[13].

With such a corpus, three tests follow. One is whether the multi-sign sequences repeat more often than chance across sites, as the names and commodity entries of a real administrative script do and as potters' marks do not⁠[16]. Another is whether sign position on the object predicts the sign, which would mark pottery marks, or whether sign order within a sequence predicts the next sign, which would mark a notation; Sproat's corpora of heraldry, kudurru emblems and barn stars give the comparison set for the latter⁠[28]. The third is whether any numeral system exists beyond strokes, since every genuine early script began with counting⁠[29].

For Tărtăria specifically, direct dating of the tablets' clay by thermoluminescence would settle whether they were fired in antiquity or in 1961, and a re-excavation of Vlassa's trench at Gura Luncii, which Lazarovici has begun, could find the pit's stratigraphy⁠[8]⁠[23]. For Dispilio, the full publication the excavators have promised since 1994 is still the one thing that would let anyone judge the tablet⁠[19]. A long text, of fifty signs or more in an excavated context, would change everything; none has been found in 150 years of digging, which is itself evidence⁠[20].