Why counting is hard to find in the ground
Counting is an activity, not an object. A person who tallies sheep on their fingers leaves nothing behind, and a great deal of early counting probably worked that way: with fingers, pebbles, sticks or the body. What archaeologists can actually recover are durable objects that carry deliberate marks, and from those they have to infer what, if anything, the marks stood for. That inference is the heart of the subject, and it is why careful writers use words like “may” and “could” about almost every item below.
It helps to separate three things that are easy to blur. The first is number sense: the ability to tell one from two or three at a glance, which human infants and some other animals appear to share. The second is tallying, matching marks one-to-one with things, which needs no number words at all. The third is counting proper, with an ordered sequence of words or signs that can in principle be extended without limit. A marked bone could belong to any of the three, and the object alone rarely says which.
The oldest marked objects
A few finds come up again and again. A baboon fibula from the Lebombo Mountains, on the border between Eswatini and South Africa, carries a row of notches, commonly reported as 29, and is usually dated to more than 40,000 years ago. Some researchers have compared that figure with the length of a lunar month in days. A single object with a roughly lunar number of cuts cannot show that the comparison was intended, though, and the count itself depends on how worn or broken marks are read.
A wolf bone from Dolni Vestonice in the Czech Republic, dated to roughly 30,000 years ago, is usually described as having about fifty-five cuts, some arranged in groups of five. Grouping would be interesting because it suggests organising marks rather than simply adding them. The Ishango bone from the Democratic Republic of Congo, often placed around 20,000 years ago, is the most discussed of all, and its notch groups have prompted many competing readings.
Older engraved pieces of ochre from Blombos Cave in South Africa, more than 70,000 years old, carry cross-hatched patterns. These are generally discussed as evidence of symbolic behaviour, not as tallies, and most researchers would not call them counting at all.
What the marks can and cannot show
A row of notches could record animals killed, days passed, phases of the moon, goods owed, or nothing numerical at all. Cuts are also made when butchering meat, gripping a tool or decorating a surface. Several features make a tally more plausible: regular spacing, repeated grouping, and signs that marks were added in separate sessions, for example changes in the width or angle of strokes. The researcher Alexander Marshack, who examined many such objects under the microscope in the 1960s and 1970s, argued that this kind of variation pointed to marks accumulated over time. Others have been more cautious, noting that any set of marks can be made to look patterned if one is free to choose how to group and read them.
Even a convincing tally has limits. One-to-one matching shows that someone kept track of a quantity. It does not show that they had number words, understood addition, or thought of numbers as abstract things. A notched stick is better evidence of memory aid than of mathematics.
Evidence beyond bones
Language offers indirect clues. Studies of some Amazonian groups, such as the Munduruku and the Pirahã, have been used to ask whether exact counting depends on having number words, and how people manage with limited number words. The interpretations are debated, and these are present-day speakers, not windows into the Stone Age. They do show that quantity can be handled in several ways without a long counting sequence.
The body is another source. In many languages the word for five is related to the word for hand, and the way people tally with fingers, joints and other body parts shapes the base they count in. The most elaborate recorded examples are the body-based systems of New Guinea, which treat parts of the arm, shoulder and face as a numbered sequence. This is why body counting is one of the best ways to imagine how a system could work with no tools whatsoever.
Objects other than bones matter too. Small clay tokens found at Near Eastern sites from about 8000 BCE onward have been interpreted by Denise Schmandt-Besserat as an early accounting method that eventually led to writing and numerals. Other scholars dispute how directly the tokens connect to later script, and what each shape meant is not settled. Knotted cords are a further possibility. The best-documented example is the Inca quipu, but such cords rot easily, so their absence from earlier periods says little.
Reading the evidence responsibly
Two habits help when reading claims about prehistoric counting. The first is to ask what a simpler explanation would be. If a row of cuts could be a decoration or a butchery mark, a claim that it is a lunar calendar needs further support, such as repeated grouping or matching marks on several objects. The second is to remember survivorship: bone, antler and stone last, while wood, cord, hide and sand do not, so the surviving tallies are probably a small and unrepresentative sample of what people did.
Dating adds another layer of uncertainty. Many of the best-known pieces were found in early excavations, and their ages have been revised as methods improved. Figures such as “40,000 years” or “20,000 years” are best read as rough ranges, not exact dates.
From tallies to number systems
Counting beyond a handful of items needs grouping, and the natural groupings come from the body: five fingers on a hand, ten on two hands, twenty with the toes, or twelve if the finger joints are counted. Which of these a community chose left traces in its number words long after the practice faded.
Written numerals appear only after thousands of years of such practice, in Mesopotamia and Egypt by about 3000 BCE. Their descendants are traced in the study of numeral systems, which picks up where the evidence for unwritten counting runs out. Before that point, what survives is a handful of scratched objects, some language data and later descriptions of living practices. Treating each with caution, and not stretching one find into a story about the whole of prehistory, is the most reliable way to read it.