We usually reconsider a definition when we find something it cannot explain. A book without paper, a memory kept in a device, or a paragraph produced by software makes an old word harder to use. We can stretch the word, replace it, or examine what we expected it to mean in the first place.

Tools have been producing this difficulty for a long time. Fire is an early example. Cooking changes what people can eat and how food is prepared. Warmth changes the use of a dwelling. Light makes activities possible after sunset. The words food, shelter, and safety remain useful, although the conditions they describe have changed.

The history needs care. Burnt material in an archaeological site does not tell us everything about who lit a fire or how it was maintained. A 2026 study of burnt small-mammal bones at Wonderwerk Cave adds evidence of Early Pleistocene fire use while distinguishing that evidence from the ability to make and control fire.[^1] We cannot reconstruct the first hearth from these remains. We can investigate particular changes associated with fire without inventing that scene.

Some of those changes can be studied experimentally. Rachel Carmody and colleagues found that cooking increased the energy mice obtained from meat and tubers.[^2] The animal model helps explain a consequence of food processing; dating human cooking requires other evidence. Food still supplies nourishment, but preparation affects how much of that nourishment becomes available.

Other changes concern how people spend their time. Polly Wiessner studied conversations among contemporary Ju/’hoansi communities. Practical affairs were prominent during the day; storytelling and other shared cultural activities occupied more of the firelit evenings.[^3] These conversations belong to contemporary communities, with their own histories. They offer an example of a setting supporting different activities, rather than a record of prehistoric thought.

A definition can describe an enduring purpose and still leave out the practices that fulfil it. Calling shelter protection says little about a chimney, a smoke alarm, or the arrangement of rooms. Those details belong to the life of a dwelling. When enough of them change, the short definition may remain correct while becoming less informative.

This distinction matters especially when our tools handle words and ideas. Early Mesopotamian writing included administrative records; the system developed to represent language more extensively.[^4] A written record could then serve purposes beyond the transaction that first required it. It could carry an account to another place or preserve it after its maker died.

Plato’s Phaedrus already examines the difference between possessing a record and learning from it. In the story of Theuth and Thamus, writing promises help with memory and wisdom, but its critic worries that written signs will substitute reminders for the exercise of memory.[^5] The problem remains familiar. I can keep an explanation indefinitely without becoming able to explain it myself.

Keeping information outside the head is often useful. Research calls many such practices cognitive offloading: actions that reduce internal cognitive demands through the use of the environment.[^6] A reminder can spare me the effort of holding an appointment in mind. A calculation on paper can help me finish a problem that exceeds my unaided working memory. Each arrangement has to be judged by the work it helps me do.

The boundary of that work is disputed. Andy Clark and David Chalmers argue that an external resource may, under suitable conditions, participate in a cognitive system. Their notebook example depends on reliable access and habitual use, rather than the mere existence of stored information.[^7] That philosophical argument is one possible account of the relationship between a person and a tool. Ordinary acts of note-taking do not, by themselves, settle it.

Books present another difficulty. UNESCO’s 1964 definition specified a printed, non-periodical publication of at least forty-nine pages, excluding the covers, for the purpose of production statistics.[^8] The rule helped people count publications consistently. It was never an adequate account of everything a book can mean to its reader. An electronic book complicates the rule while retaining many familiar features of a book: a title, an organised text, and a work to which someone can return.

AI complicates the history of that text. Two volumes may look alike even though one writer composed every sentence and another accepted generated prose. The physical category book cannot tell us how either work was made. We need further questions about judgment, responsibility, and authorship.

The essays in this series examine these problems separately. Memory leads to knowledge, then to understanding and the tools used in thinking. From there the discussion reaches the boundaries of mind, the purpose of an essay, the persistence of books, and the authorship of words. The final essay considers how connections between these subjects can help a reader follow an idea.

The common concern is practical. Before changing a definition, we should establish what has changed in the thing, which distinction we need, and what the existing word still explains. Some definitions survive because they continue to do their work. Others need revision because our lives have given us a new question.