Glozel: the inscribed tablets of the Fradin field

The Glozel collection is about 3,000 objects dug from one field and its surroundings between 1924 and 1936, almost all of them by Morlet and Fradin without a recorded stratigraphy⁠[1]⁠[22]. The inscribed pieces are the heart of it. About 100 are flat fired-clay tablets, most inscribed on one face with six or seven lines of signs cut before firing; some bricks and vases carry signs as well, and there are inscribed bone, pebbles and stone⁠[1]⁠[23]. Around them is a mixed assemblage that the sceptics found impossible as one culture: Upper Palaeolithic-looking engravings of reindeer, bone harpoons of Mesolithic type, Neolithic-looking flint and stone, clay idols with exaggerated sex, face urns, hand-printed tablets and glass⁠[22]⁠[13]. Dussaud noted that only about half a dozen tablets were found in front of witnesses other than Morlet and Fradin⁠[6].

The only edition is Morlet's Corpus des inscriptions of 1965, with drawings of the tablets on pages 69 to 90, the other inscribed objects on pages 25 to 68, and a sign table of 111 signs⁠[20]⁠[19]. Crawford found 44 tablets usable in it, two inscribed on both faces, and rejected several as too damaged to read⁠[17]. There is no photographic corpus, no database of sign occurrences and no agreed transliteration. The collection is held by the Fradin family in their private museum at Glozel; institutional archives are at the Musée d'Archéologie nationale, the Archives nationales and the Épernay museum, and more than 100 contemporary publications are digitised on the Maison de l'Orient's Reinach portal⁠[21]⁠[24].

The signs are linear, made of strokes, and most resemble letters of early Mediterranean alphabets. Morlet compared them to Phoenician and claimed near identity⁠[3]. Dussaud agreed that the 22 letter-shapes on the first tablets matched Phoenician, but the Phoenician of the Eshmunazar sarcophagus of the fifth century BC rather than the archaic script of Ahiram, which he had just published; a Neolithic ancestor should not look like a late descendant⁠[6]. Later tablets added Iberian-looking signs and five-point pebbles, which Dussaud and Vayson de Pradenne tied to publications that had reached Morlet in the meantime⁠[6]⁠[7]. Isserlin's 1976 comparison, commissioned alongside the TL work, found only partial resemblances to Greek, Etruscan and Latin and closer but still partial ties to the Iberian scripts⁠[25]. Camille Jullian read the signs in 1926 as late Latin cursive carrying magical formulas, a reading nobody else has reproduced⁠[19].

Of the physical dates the firm points are these. Thermoluminescence (TL) measures the radiation dose a fired clay has absorbed since it was last heated, so it dates the last firing, not the making. About 40 TL dates exist on Glozel ceramics; they fall into an Iron Age to Gallo-Roman group, a medieval group centred on the thirteenth century, and a modern group, with two inscribed vitrified tablets at AD 1910 and 1920⁠[15]⁠[14]. Only one date, a pooled bone sample of about 17,000 years that its own authors discount, is older than the Iron Age⁠[15]. Radiocarbon on bone, charcoal and ivory gives the third to twentieth centuries AD⁠[15]⁠[1]. The field contains glassworking debris in every trench within the site, and the Ministry's team judged medieval glassmaking the only undisputed activity there⁠[15]. Neutron activation at Toronto confirmed that tablets and an idol were made of local Glozel clay⁠[14].

Three kinds of test have been run. The first is excavation under control. The 1927 commission dug a trench for four days with the ground guarded at night and found objects only in loose, recently disturbed earth, together with fresh root fragments and clay still soft enough to mark; it concluded that the objects were not ancient⁠[8]⁠[9]. Morlet's supporters answered with their own Comité d'études in April 1928, which reached the opposite conclusion⁠[10]. The state's five trenches of December 1983 were the first with modern recording; the two outside the traditional limits were sterile and the three inside produced glassworking waste, a few sherds and no Neolithic or Palaeolithic layer⁠[15].

The second is laboratory dating. The 1974 TL programme at Edinburgh, Risø and the French atomic energy laboratories gave Iron Age dates and was presented as proof of authenticity⁠[11]⁠[4]. Zimmerman's zircon test excluded artificial irradiation as a way of faking age⁠[26]. Aitken and Huxtable pointed out that objects dug up in the 1920s have no measurable burial dose-rate and that any reheating resets the clock⁠[12]. Barbetti's archaeomagnetic study then found one tablet heated twice⁠[27], and the Oxford laboratory in 1983 to 1985 obtained medieval dates for three inscribed pieces and twentieth-century dates for two vitrified tablets⁠[15]⁠[14]. Risø and Oxford disagreed on the same tablet by over a thousand years⁠[15]. Gaston Bayle's 1929 forensic examination had already found modern dyed wool and cotton fibres and fresh plant matter inside the tablets⁠[1].

The third is analysis of the signs. Crawford's 1977 computer study compared the sign sequences of 34 tablets with Basque, Berber, Chinese, Finnish and Iberian corpora and found none of the distributional properties of a natural-language text; he suggested a non-linguistic function or a syllabary and said linguistics alone could not decide authenticity⁠[17]. No later statistical study has been published.

On authenticity four positions have been held. Morlet and Reinach held that the site was Neolithic and its alphabet the oldest known; no Neolithic deposit or date has ever been found, and the position is abandoned⁠[3]⁠[10]⁠[15]. The 1927 commission, Dussaud, Vayson de Pradenne and most prehistorians since have held that the whole collection was made in the 1920s⁠[8]⁠[6]⁠[22]. The TL team and the pro-authenticity colloquia at Vichy argue for a genuine Celtic and Gallo-Roman occupation of about 300 BC to AD 300, with the medieval and modern dates explained by later fires and kilns reheating old objects⁠[28]⁠[14]. The state's position since 1995 is a middle one: the field held a medieval glassworks, some Iron Age and medieval objects may be genuine, and the collection as a whole, the inscribed tablets included, was enlarged by forgery in or after the late nineteenth century⁠[15]⁠[16]. Sociologists of science have treated the affair as a case where no shared criterion of authenticity existed⁠[5].

On the script three readings exist. Hitz derives a 17-letter primary alphabet of about 300 BC from Lepontic and Etruscan models, expanded by about 200 BC to 26 letters and 40 to 50 ligatures, and reads Celtic dedications such as nemu Chlausei, which he renders as the sacred grove of Glozel⁠[18]⁠[29]⁠[19]. The reading has no support among Celtic epigraphers and depends on many signs assigned ad hoc. Jullian's Latin magical reading and the Semitic, Basque, Berber and Turkic proposals of the 1920s are not pursued⁠[19]. The alternative, held by Dussaud and by Jean-Pierre Adam, is that the signs were borrowed from published alphabet tables and arranged without linguistic order, which is what Crawford's statistics would predict⁠[6]⁠[17].

The authenticity question could be closed by dating the inscription rather than the clay. A TL or archaeomagnetic date fixes the last firing; a sign cut into wet clay before firing is as old as that firing, but a sign scratched after firing is not, and the two cases have not been separated systematically across the corpus. Microscopy of sign floors on the dated tablets, with the Risø and Oxford samples re-measured on the same surfaces, would show whether the Iron Age group carries inscriptions made before firing⁠[12]⁠[15]. The Ministry team recommended technological and typological study rather than more dates, and in particular a study of medieval and modern glassmaking in the Bourbonnais hills, which would explain the kilns and the vitrified tablets⁠[15].

The script question can be tested now. A machine-readable transcription of Morlet's Corpus, with a sign list keyed to his 111 signs and Hitz's 80, would allow Crawford's 1977 test to be repeated on all the tablets with modern methods: sign-frequency distributions, conditional entropy and repeated-sequence counts against natural-language controls and against random draws from a Phoenician or Iberian alphabet table⁠[17]⁠[25]. If the sequences behave like a language, Hitz's Celtic reading can be scored on held-out tablets. If they behave like random draws, the forgery account gains a quantitative footing. A find of a Glozel-type inscribed object in a dated context outside the Fradin collection would change everything; none has been reported in a century⁠[16].