Signs on Stone Age objects: Precursor to written language dates back 40,000 years

(Press-News.org) EMBARGO February 23, 2026 at 3:00 PM U.S. Eastern time
Over 40,000 years ago, our early ancestors were already carving signs into tools and sculptures. According to a new analysis by linguist Christian Bentz at Saarland University and archaeologist Ewa Dutkiewicz at the Museum für Vor- und Frühgeschichte (Museum of Prehistory and Early History) in Berlin, these sign sequences have the same level of complexity and information density as the earliest proto-cuneiform script that emerged tens of thousands of years later, around 3,000 B.C.E. Using a computational approach, the team examined over 3,000 signs found on 260 objects to reveal insights on the origins of writing. Their findings, which will be published in the journal PNAS, were clear – and surprised even the researchers.

Palaeolithic objects dating back between 34,000 and 45,000 years bear mysterious sign sequences – often repeated lines, notches, dots and crosses. Many of these artefacts were discovered in caves in the Swabian Jura, such as a small mammoth found in the Vogelherd Cave in Lone Valley in south-western Germany. A Stone Age human carved the mammoth figurine out of a mammoth tusk and carefully engraved it with rows of crosses and dots. Other artefacts found in the Swabian Jura are also etched with signs. One of these objects is the 'Adorant', a mammoth ivory plate uncovered in the Geißenklösterle cave in the Ach Valley that depicts a hybrid lion-human creature. The object is likewise adorned with rows of dots and notches. Upon close inspection, another mythical depiction of a human-lion hybrid, the Lion Human from the Hohlenstein-Stadel Cave in the Lone Valley, reveals notches placed at regular intervals along the arm.

New findings show that these marks are there for a reason – Stone Age humans used them to convey information and to record their thoughts. 'Our research is helping us uncover the unique statistical properties – or statistical fingerprint – of these sign systems, which are an early predecessor to writing,' explains Professor Christian Bentz of Saarland University.

'The Swabian Jura is one of the regions where objects with this type of sign have been found most frequently, but there are, of course, other important regions. Countless tools and sculptures from the Palaeolithic, or the Old Stone Age, bear intentional sign sequences,' elaborates PhD archaeologist Ewa Dutkiewicz. The researchers travel together throughout Europe, visiting museums and archaeological sites to find new Stone Age signs. 'There are many sign sequences to be found on artefacts. We've only just scratched the surface,' says Dutkiewicz, who is a research associate and curator of the Stone Age department at the Staatliche Museen zu Berlin.

'The artefacts date back to tens of thousands of years before the first writing systems, to the time when Homo sapiens left Africa, settled in Europe and encountered Neanderthals,' explains the archaeologist. In a project funded by the European Research Council, the two researchers and their team are examining how Stone Age humans encoded information in sign sequences.

Information density similar to that of proto-cuneiform

The researchers analysed more than 3,000 geometric signs found on around 260 objects using computational approaches. Their aim was not to uncover the concrete meaning of the signs, which have not been deciphered. 'There are plenty of theories, but until now there has been very little empirical work carried out on the basic, measurable characteristics of the signs,' explains Bentz. His research deals with frequency trends and tangible, measurable aspects of the signs. This allows him to see what the sign systems have in common with later systems – and how they differ. The linguist aims to leverage statistics to uncover insights on the origins of information encoding.

'Our analyses demonstrate that these sign sequences have nothing to do with the writing systems of today, which represent spoken languages and are characterized by high information density. In contrast, the signs on the archaeological objects are frequently repeated – cross, cross, cross, line, line, line. This type of repetition is not a feature found in spoken language,' explains Christian Bentz. 'However, our findings also show that Palaeolithic hunter-gatherers developed a system of symbols that has an information density that is statistically comparable to the earliest proto-cuneiform tablets from ancient Mesopotamia, which came 40,000 years later. Sign sequences in proto-cuneiform script are also repetitive and the individual signs are repeated at a similar rate. In terms of complexity, the sign sequences are comparable,' says Bentz. 'Figurines exhibit higher informational density than do tools,' reports archaeologist Dutkiewicz, who also used to curate the Vogelherd archaeological park in the Swabian Jura.

The researchers were particularly surprised by how the sign systems measured up to proto-cuneiform. 'We hypothesized that the early proto-cuneiform script would be more similar to the writing systems of today, especially due to their relative proximity in time. Yet the more we studied them, the clearer it became that the early proto-cuneiform script is very similar to the much older Palaeolithic sign sequences.' That also means that little changed between the Old Stone Age and the emergence of the first proto-cuneiform scripts. 'Then, about 5,000 years ago, a new system emerged relatively suddenly that represents spoken language. The new system therefore has completely different statistical characteristics,' explains Bentz.

Computational analysis of Palaeolithic signs

For their research, the team digitalizes the sign sequences on archaeological objects in a database, which they then use to assess statistical properties in the Stone Age sign inventories. Using computer-assisted methods, Bentz looked into the potential to express information using the signs and compared this to the potential allowed by early cuneiform sequences and by modern writing. In their analysis, the researchers applied approaches from quantitative linguistics such as statistical modelling and machine learning classification algorithms.

'Because of the high rate of repetitions and the high predictability of the next sign, we were able to show that the entropy – a measure of information density – is comparable to that of proto-cuneiform, which came much later,' explains Christian Bentz. 'The human ability to encode information in signs and symbols was developed over many thousands of years. Writing is only one specific form in a long series of sign systems,' Bentz elaborates. 'We continue to develop new systems for encoding information. Encoding is also the basis of computer systems.' Large language models, which are currently one of the most visible forms of AI, rely on the fact that language sequences are predictable, meaning the model is able to determine which part of a word is likely to come next.

Uncovering the encoding from the Stone Age

The study does not reveal what the Stone Age humans were trying to record with the signs. 'But the findings can help us to narrow down potential interpretations,' explains Ewa Dutkiewicz. While the humans of today can access thousands of years of information and knowledge transfer that the humans of then could not, anatomically speaking, Stone Age humans had already reached a similar stage of development as modern humans. This means they likely had similar cognitive abilities as we do. The ability to record and convey information to others was extremely important for Palaeolithic humans. It may have allowed them to coordinate groups or even helped them survive. 'They were highly skilled craftspeople. You are able to see that they carried the objects with them. A lot of the objects fit right in the palm of your hand. That is another way in which the objects are similar to proto-cuneiform tablets,' explains Ewa Dutkiewicz.

The research is a part of the project 'The Evolution of Visual Information Encoding' (EVINE), which is being funded by an ERC Starting Grant from the European Research Council. Professor Christian Bentz joined Saarland University in 2025 and leads the EVINE research project. The project first began at the University of Tübingen and was then continued at the University of Passau at the Chair of Multilingual Computational Linguistics. Professor Bentz and Dr. Dutkiewicz examine how visual information encoding developed from the earliest signs to the writing of today. (ERC, EVINE, 101117111).
https://www.erc-evine.de/

Join the hunt for Stone Age signs on YouTube
Christian Bentz and Ewa Dutkiewicz travel across Europe, visiting museums and archaeological sites while searching for Stone Age signs. The duo documents their hunt on their YouTube channel. https://www.youtube.com/@StoneAgeSigns

END