I study cultural transmission, the way traditions are passed on from person to person. This key mechanism in human history and evolution bridges three distinct levels: individual human cognition, interactions between individuals, and population-level processes. To span all these levels, my work combines a broad range of methods, from lab experiment to quantitative cultural history. I developed a theory of cultural transmission which argues that it rests primarily on ostensive communication, as opposed to imitation. I also defended a view of cultural evolutionary dynamics where continuous, non-random transformations matter more to cultural change than selectionist dynamics, leading to debates on the nature and implications of cultural-evolutionary models. Since 2016, I have focused most of my research on one particular tool for cultural transmission: graphic codes.

Writing systems, pictographs, emblems, brands, seal markings are all graphic codes: they carry information by means of enduring images with standardised meanings. Humans are exceptionally good at inventing and learning communicative codes, especially spoken languages; but powerful graphic codes emerged rarely in human evolution, in contrast to the universality of language. I try to understand why.

Experimental methods help me explore the basic conditions that allow graphic codes to emerge. Me and my co-authors hypothesise that this is due to the peculiar conditions of asynchronous communication, where a messenger imparts information to a recipient across time frames. Synchronous (or face-to-face) communication is grounded in interactional mechanisms such as turn-taking or repair, as well as pragmatic factors like common ground. In contrast, messengers and senders engaged in asynchronous communication have little common ground to share, and no opportunities to repair misunderstandings.

Understanding the evolution of writing requires us to explore other graphic codes that cannot be classified as writing: mnemonic codes, flags and other heraldic emblems, symbols found on coins, etc. These codes, known as "semasiographies", do not encode components of spoken languages. They are also drastically limited, compared to writing, in the quantity and variety of information that they can encode. These limitations can teach us about the obstacles that blocked the development of writing systems in most cultures before Neolithic times. But their strengths tell us how far the human capacity for asynchronous communication may extend, in the absence of writing. Large, standardised datasets allow me to use information theoretic-tools to measure the capacity of two graphic codes, heraldic emblems and ancient coinage, to contain and impart information. Standard entropy measurements can be applied to the organisation of components inside a graphic symbol, telling us, for instance, to what extent the colours and figures that make up coats of arms occur in a non-random fashion or how predictable the occurrence of symbols on a coin can be.

Researchers know a great deal about what makes shapes easy or difficult for human vision to process. Complex shapes, with a high ratio of contour to surface, take more time to process and tend to have less appeal than simple ones. (Letters like O or C are simpler than W or Q in this sense.) Cardinal straight lines (horizontals and verticals, as in H or E) are processed more fluently than oblique ones (as in W or X). The scripts used by the writing systems of the world must weigh these demands of visual simplicity against the need to encode large quantities of information (which overly simple or uniform shapes cannot do). Finding an optimal balance between informativeness and simplicity is a difficult task, for which a gradual and protracted process of cultural evolution might be needed. I study the ways in which various literate cultures have solved this problem, and we ask whether the last three millennia have seen global improvements in this regard.

This research yields two kinds of results. First, it validates laboratory results for the first time outside the lab: we show that the psychophysical biases documented experimentally have real cultural consequences. For instance, I showed that the vast majority of the world’s scripts obeyed complex constraints related to the organisation of cardinal and oblique straight lines in letters. Second, I contribute to the budding field of cultural evolution by putting its hypotheses to the test: my work on the evolution of cardinal lines within letters addressed the question whether cognitively appealing cultural forms need a protracted evolutionary process to arise.

This research is part of a broader attempt to integrate the methods of cultural history with those of experimental psychology. My work in collaboration with Piers Kelly, James Winters, and Helena Miton, on the evolution of the Vai syllabary of Liberia, exemplifies this. Of all the invented writing systems that appeared in colonial or post-colonial contexts in the last two centuries, the Vai script is one of the best documented. Our Vai project started with the construction of a dataset documenting the shape of Vai letters in the various historical stages that the syllabary went through. The team then applied a variety of compression algorithms, verifying our prediction that Vai letters became simpler in the course of their 170-years evolution. Another paper, by Helena Miton and me, asks the same question on a larger scale. It compares the visual complexity of letters in more than 100 systems, which allows us to consider phylogenetic relationships between writing systems, and to test for evolutionary trends. Recently, we used the same dataset to show, with Alexey Koshevoy, that frequently used letters are simpler in a broad range of scripts — a phenomenon that we argue is analogous to what linguists know as Zipf's Law of Abbreviation.

Laboratory experiments are limited in their demographic range (typically, a few Western undergraduates) and in the scale over which they unfold (two persons for a few hours). Gamified apps or applets allow me to move beyond these constraints. I have supervised two gamified "citizen science" projects.

Glyph is a participatory science project where players are challenged to find the best ways to classify the shapes of the world's writing systems. 

The Color Game. Players of the “Color Game” smartphone app were asked to organise black and white symbols into graphic codes that they used to communicate about colours. The players could be from anywhere on the planet (the app was translated into seven languages and available worldwide). They could play as often as they wanted with whomever they wanted. The app’s sophisticated back-end interface guaranteed that players remain entirely anonymous, and could not contact their friends. More than 4000 players interacted through the app, creating a miniature language whose evolution is completely documented. We used it to test hypotheses regarding the evolution of graphic codes.

Quantitative cultural history applies quantitative tools to historical changes in cultural forms. I have done quantitative cultural history of letter shapes, coins, heraldry, or portraits, but I also on topics beyond graphic codes or visual culture, such as literary fiction. In a paper with Alberto Acerbi, we show that the strong decline in emotional expression in 19th and 20th century literature was driven by a collapse of positive emotions. From a methodological point of view, we argue that quantitative cultural history benefits from combining results from big-data corpus such as Google Ngram with smaller and better-understood corpora. I also did studies asking why people die in novels or why monsters are dangerous.