Skip to main content
NCBI home page
Springer logoLink to Springer
. 2026 Feb 13;37(1):23–39. doi: 10.1007/s12110-026-09512-5

Exploring the Emergence of Organized Colouration in Paintings Through Cultural Transmission

Mayuko Iriguchi 1,#, Sota Kikuchi 2,#, Takashi Morita 3,4,5, Hiroki Koda 2,6,✉
PMCID: PMC13079504  PMID: 41686391

Abstract

Painting and drawing are symbolic representations that visually transform natural scenes into line and colour. These visual expressions change through cultural transmission—the process by which artworks are reproduced and modified across individuals or communities. While structured patterns have been observed in transmitted language and music, it remains unclear whether similar structuring occurs in colour use for visual representations. This study investigates the emergence of structured colour expression in painting through cultural transmission. We conducted a transmission chain experiment using colouring books. Participants viewed and memorised a coloured page, then reproduced the colour patterns from memory. Their reproduction was used as the stimulus for the next participant. Two colouring books were used, each with five initial colourings. Each transmission chain involved ten iterations, with 100 participants total (five chains per book). In the final generation, participants used fewer and more consistent colours for recognizable objects compared to the initial random-like patterns. This indicates a structuring of colour use over time. However, familiar colour choices (e.g. green for leaves) emerged in only some chains, showing partial influence of prior object-colour knowledge. Our findings suggest that, like language and music, colour use in visual art becomes structured through cultural transmission. While not all chains developed familiar colouring, the overall simplification and regularity of colour patterns highlight the role of shared knowledge and perceptual expectations in shaping visual cultural evolution.

Supplementary Information

The online version contains supplementary material available at 10.1007/s12110-026-09512-5.

Keywords: Paintings, Cultural transmission, Cultural evolution

Introduction

Painting and drawing are products of higher cognitive functions that are unique to humans (Saito et al., 2014) and have been attested cross-culturally, just like language and music. Psychologically, painting and drawing represent humans’ cognitive properties regarding visual-information processing and their symbolization of the visual world (Makuuchi et al., 2003). That is, a painting is not a simple reproduction of the visually perceived world―recovering features such as colours, luminance, or structural patterns―but is rather influenced by symbolic/categorical perceptions of objects and/or backgrounds, based on our prior beliefs or memories. Artistic representations, as exemplified by painters such as Monet and Cezanne who emphasized some visual features of light and colours, reflect the organization of perceptual and conceptual processes in the brain, rather than the direct replication of retinal images (e.g., Zeki, 1999). Moreover, paintings have been utilized as a means of conveying information across generations and between cultures in a symbolic way, similarly to written languages. For example, the paintings left by early human, possibly including Neanderthals, are considered to be a trace of primitive symbolic communication in cognitive archaeology (Aubert et al., 2014; Hoffmann et al., 2018); they utilized colours and line patterns to symbolize observed information and communicate it among individuals.

Accordingly, unveiling how symbolic communication emerged in a particular culture and also finding cross-culturally general patterns in such emergence is of a major interest in the study of cultural evolution (Deacon, 1998; Mithen, 1996). In the past two decades, researchers have explored possible accounts for the emergence of such symbolic communication systems under the experimental paradigm of cultural transmission (Kirby et al., 2008, 2015; Perfors & Navarro, 2014; Silvey et al., 2015). In the experiment, participants play “the game of telephone”; they are ordered sequentially and instructed to memorize what their preceding participant say/show and convey it to the next participant as it is heard/seen (without being notified of the transmissive design of the experiment). For example, participants may be asked to pass on some random text words/sentences via its reading and writing (Kirby et al., 2008, 2015; Perfors & Navarro, 2014; Silvey et al., 2015), or some rhythmic sound pattern by hearing and tapping (Jacoby et al., 2024; Jacoby & McDermott, 2017; Nave et al., 2024; Ravignani et al., 2016). The key idea behind this experimental design is that the carried signal is iteratively transformed as it passes between participants, and this transformation can be modelled computationally to capture participants’ cognitive biases or prior beliefs about the signals. Specifically, when each transmission is modelled as a process of Bayesian inference, the entire chain converges to the prior distribution held by the participants, regardless of the initial stimuli (Griffiths & Kalish, 2007). Thus, the transmission of initially random texts and rhythms can give rise to organized artificial languages and musical patterns, respectively (Jacoby & McDermott, 2017; Kirby et al., 2008). Similarly, even a highly descriptive depiction of the world—such as a realistic painting—could eventually be reformatted into a symbolic representation through cultural transmission, guided by humans’ underlying cognitive bias towards categorical perception. This offers a possible account for the emergence of symbolic communication systems.

As already reviewed, paintings and drawings, like language and music, serve as a means of symbolic communication. Accordingly, we would expect that some structured patterns could emerge in the process of visual perception, reproduction, and transmission of paintings/drawings between individuals, reflecting their prior knowledge or bias. For instance, humans are supposed to share some belief about the canonical colours on the objects drawn in a painting―like “a rose is red” and “a leaf is green”―based on their previous experiences. Such prior knowledge about objects’ colour can override or complement our physical sensing and modify our recognition of the colour (Hansen et al., 2006; Hasegawa et al., 2024; Kimura et al., 2013; Olkkonen et al., 2008; Witzel et al., 2011). For instance, a banana drawn in greyscale is perceived slightly yellowish by an observer (called memory colour effect; Hansen et al., 2006; Olkkonen et al., 2008).

In the previous studies of cultural transmission, researchers have focused on the development of shapes in drawings. Indeed, the first experimental study dates back to 1932, when Frederick Bartlett (Bartlett, 1932) demonstrated an emergence of drawing. Just as in the modern paradigm, he initially prepared an “unrecognizable” instance of an Egyptian Hieroglyph, and then instructed the participants to memorize and reproduce it, serially serving the stimulus for the next generation. Consequently, the drawing by the participants “evolved” to an owl or a black cat. Moreover, Bartlett also shared the modern interpretation of this result; humans are strongly biased towards their own cultural expectations when reconstructing information from memory, and this memory or cognitive bias emerged in drawings due to the exaggerated consequences of serial reproduction. Consistent results were reported in recent studies as well (Tamariz & Kirby, 2015); transmission of drawings converged to simplified/structured patterns, presumably reflecting the prior knowledge or belief of participants. These findings support the argument that symbolic systems can emerge in transmission of paintings under a perceptual bias of humans.

Despite the rich history of investigation on the emergence of shape patterns in the literature, however, little exploration was made into cultural transmission of colours as a pictorial expression. Colour perception and categorization are believed to be shaped by cultural and linguistic backgrounds (e.g., Regier & Kay, 2009) and this hypothesis should be tested in the context of cultural transmission. Thus, we experimentally investigated whether the structured colouration could emerge in a cultural transmission of paintings. Specifically, we utilized a “colouring book”, containing line drawings with multiple colour-fillable areas, and instructed participants to memorize a sample colouring pattern and reproduce that pattern as precisely as possible in a limited time. The reproduced paintings were transmitted to other participants (in the next generation) following the paradigm of the previous cultural transmission experiments. We recorded and analysed the evolution of the colours used in the participants’ paintings.

Prior to the conduction of the experiment, we expected three main results. First, due to the memory colour effect, familiar colouring patterns, seen on real objects, were thought to emerge in the consequence of the transmission. Second, each categorically recognized segment of objects in the drawings―such as petals and leaves of flowers―would be painted with a single or few, consistent colour(s); consequently, we expected uneven use of colour types (i.e. dominance of a few colours in the entire colouring book). Finally, fewer transmission errors would occur in later iterations, due to the emergence of organized―thus more easily memorable―patterns as expected in the second.

Materials and methods

Participants

We recruited 117 university students (age range 18–31, mean age 20.64) from Kyoto and Keio Universities as participants in the experiment. Prior to the experiment, their colour vision was assessed using the Ishihara tests, a standard test for colour blindness in Japan, and none exhibited any colour vision deficiency. Seventeen of them did not follow the instruction of the experiment (specifically, they put more than one colour within a single colouring area), and accordingly, they were excluded from the analysis. This left the data from 100 participants, who were grouped and ordered into ten chains, each consisting of ten participants. The ten chains were further categorized into two groups and presented with different stimuli as will be explained in the next section (i.e. 2 groups × 5 chains × 10 participants).

Stimuli and Response Format

We used two types of colouring books as the experimental stimuli: roses are drawn in one and animals of four species in the other (Fig. 1 A and B). The rose colouring book included seven flowers consisting of petals, leaves, stems, and sepals; there were 213 line-distinguished areas. The animal painting included two rabbits, one parrot, one dog, and one cat, having 186 colourable areas. The drawings were approximately of size 15 × 15 cm, and were printed on a A4 paper with grey background. Participants were provided with pens in six colours: black, red, blue, green, yellow, and brown.

Fig. 1.

Fig. 1

Open in a new tab

Schematic diagram of the cultural transmission experiment using colouring books. A is a rose, and B is an animal colouring book. C Generation 0 sample colouring books of 10 chains. D Flow of the cultural transmission experiment using a colouring book. First, participants viewed a sample colouring book for one minute and memorized its colour pattern (memorization phase). Then, the participants reproduced the colour pattern in twenty minutes (reproduction phase). All colouring areas should be painted. The resulting colouring book was used as a sample for the next generation of colouring books. This was repeated and transmitted up to generation 10

Procedures

In the experiment, participants were presented with a sample painting (either of roses or of animals) and asked to memorize and reproduce it by filling a blank colouring book. The initial stimuli used for the first-generation participants (called the “generation 0” painting, and listed in Fig. 1 C) were constructed by assigning a random colour to each line-distinct area of the colouring book.1 These random colours were sampled uniformly and independently from the six options (black, red, blue, green, yellow, and brown; using the RANDOMBETWEEN(1,6) function in Microsoft Excel), and the resulting paintings were colour-printed on a A4-sized paper. It is noteworthy that the colours used in the initial stimuli did not perfectly match those of the pens used by the participants, due to the technical limitations.

Five different colourations were built to initiate the transmission chains of each type of the colouring books (“rose” and “animal”). Then, the reconstruction of these initial colouring patterns by the participants of generation 1 yielded the stimuli for those of generation 2, and so on (Fig. 1D). After ten iterations of this transmission process, we obtained the responses from 100 participants in total (2 stimulus types × 5 chains × 10 generations).

The detailed procedure for each participant’s trial is as follows. First, they were presented with a sample painting (either the initial stimuli or response from the previous generation, but without being notified that the transmissive design of the experiment) and asked to memorize its colouring pattern as precisely as possible within a minute. After this memorization phase, the participants were instructed to flip the sample painting, so that it was no longer visible to them, and they immediately started filling the blank book to recover the observed pattern. Specifically, the participants were instructed to paint all the white areas in twenty minutes by choosing one of the six colours―and not two or more―for each. This time limit was only intended to pace the participants, and the actual trials lasted until everyone completed the task.2

It is of note that the experimental instruction did not mention the treatment of the background. Consequently, one participant (chain A, generation 4; see Fig. 2) painted the background, contrary to our intension. Due to the limited availability of the participants, we continued the experiment on that chain and also included the results in the analysis.

Fig. 2.

Fig. 2

Open in a new tab

Example results of colour transmission process. (Top two rows) The most remarkable familiar colour emerged in the rose colouring book. (Bottom two lines) Transmission results of animal colouring book

Our experiment complied with the Guidelines for Research in Human Participants, and was approved by the Human Research Ethics Committee of Primate Research Institute, Kyoto University (Permit No. 2018-03).

Re-identification of Colours

The paintings collected from the participants were scanned into digital images (in the JPEG format), and the colour of each line-distinct area of the colouring books was re-identified systematically in the Hue- Saturation-Value (HSV) feature space. We first manually sampled three representative pixels within that area. Then, the RGB values of these pixels were extracted using the “cv2.imread” function of OpenCV (wrapped as a Python module; Bradski, 2000). We also used the OpenCV function, “cv2.cvtColor”, for the RGB-to-HSV conversion.

Each representative pixels of the paintings were then classified into one of the six colour categories, or identified as “NULL” (i.e. colourless), based on their HSV values in the following procedure.

  • The pixel was identified as “NULL” when its saturation was under 20 (S ≤ 20) whereas the value was greater than 200 (V ≥ 200).

  • The pixel was categorized into “black” when its saturation and value were both under 100 (S ≤ 100 and V ≤ 100).

  • Otherwise (S ≥ 20), the pixel was classified by their hue into the nearest neighbour of the following category means:

    • ◦ Red: 0 or 180 (0 ≤ H < 4 or 145 ≤ H < 180).
    • ◦ Brown: 8 (4 ≤ H < 19).
    • ◦ Yellow: 30 (19 ≤ H < 55).
    • ◦ Green: 80 (55 ≤ H < 95).
    • ◦ Blue: 110 (95 ≤ H < 145).

The colour category of the line-distinct area was then defined by the majority of the three-pixel colours. Although it was theoretically possible that all the three pixels were identified as having different colours―in such cases, we planned manually colour identification―there was no such complete disagreement in practice.

Analysis

The data collected from the transmission were analysed from three perspectives: (1) transmission accuracy, (2) randomness of colour use, and (3) local consistency of the colouration.

Transmission Accuracy

We first evaluated the transmission accuracy between successive generations, assessing whether each area of the colouring books was painted with the same colour as in the previous generation (coded as “1”) or not (coded as “0”). Specifically, we investigated if this accuracy was influenced by the number of transmissions performed on the paintings, estimating its statistical significance by the generalized linear model (GLMM) that predicts the binary accuracy values from the generation. We built distinct linear models for each type of the colouring book (rose or animals). The statistical test was implemented in R (R Core Team, 2023), using the “glmer” function in the “lme4” package (Bates et al., 2015). The generation effect was categorically represented in the reverse Helmert coding―which contrasts each generation with the mean of the previous generations―and treated as fixed effect terms. The transmission chains were used as a random effect, and the error distribution was binomial (formulated in R as glmer(correct ~ generation + (1|chain).

Randomness of Colour Use

Second, we assessed the degree of randomness in the colour assignments by utilizing the perplexity statistic. Formally, the perplexity Inline graphic was defined by two to the exponent of the colour-proportion entropy:

graphic file with name d33e527.gif

where the integer Inline graphic) indexes the six colours (black, red, blue, yellow, green, and brown), and Inline graphic denotes the relative frequency of each colour, normalized s.t. Inline graphic. The perplexity offers an intuitive interpretation of the colour proportions by associating their dispersion with the uniform distribution over Inline graphic (e.g. the uniform distribution over six categories has the perplexity of Inline graphic).

The perplexity estimated from the empirical frequencies was evaluated against the random baseline built as follows. First, random probabilities of occurrence (Inline graphic) over six colours were sampled from the Dirichlet distribution with a concentration parameter of 1 (Inline graphic; i.e. uniform distribution over the simplex of Inline graphic dimensions), and the numbers of occurrences of six colours in the 213 areas of the rose painting and 186 areas of the animal painting were sampled from the multinomial distribution parameterized by Inline graphic. Then, the perplexity of the random colouration was calculated from the relative frequencies of these colour samples. We collected 10,000 random perplexities in this procedure, and the statistical significance of the experimental result was assessed by the proportion of the random perplexities smaller than that of the experiment (i.e. the p-value was estimated in a Monte Carlo manner). The random sampling from the Dirichlet and multinomial distributions were implemented by the “rdirichlet” and “rmultinom” functions of R respectively (the former is available in the package “MCMCpack”; Martin et al., 2011).

Local Consistency of Colouration

In addition, we evaluated the spatial distributions of the colour use. Specifically, we examined if neighbouring blocks in the colouring book were painted in the same colour, resulting in a stronger local colour consistency compared to the global ratio. We formalized this test by counting the number of pairs of adjacent colouring areas that were assigned the same colour (the rose and animal colouring books included 504 and 339 pairs of neighbouring areas in total). Then, the statistical significance of these counts was estimated by comparing them with 10,000 random re-assignments of the colours used in each participant’s response (i.e. colour shuffling within the colouring books). Specifically, the p-value was defined by the proportion of the random re-assignments that resulted in a greater number of neighbour colour match than the experimental result.

Results

Qualitative Overview

Figure 2 presents examples of the transmission trajectories. Overall, we observed a tendency that distinct flower parts (e.g. leaves and petals) and individual animals were painted with a few consistent colours. Most remarkably, one of the transmission chains achieved a surprisingly familiar colour patterns (the top row of the figure).

Transmission Accuracy

Figure 3 illustrates the transmission accuracy between each successive generation. Overall, the accuracy increased along with the cumulative number of transmissions (i.e. fewer errors were made in the later generations). The GLMM analysis revealed the statistically significant improvements at 3rd and 5-10th generations of the “rose” chains (Inline graphic) and 2-10th generation of the “animal” chains (Inline graphic at 3rd generation and Inline graphic elsewhere; see also the Supplementary Table S1, S2).

Fig. 3.

Fig. 3

Open in a new tab

Line plots of the performances for each generation (Left: Rose type; Right: Animal type). Asterisks represent the significant improvements from the previous generations (Inline graphic). The dashed line represented the mean values of 5 chains

Randomness of Colour Use

Figure 4 shows the proportion of the six colours used in each generation. The use of certain colours became dominant in the later generations (e.g. red and brown in the rose paintings, and brown and green in the animal paintings), in the compensation for the fewer occurrences of others (e.g. yellow and blue in the rose paintings, and red and black in the animal paintings). We quantitatively evaluated these uneven frequencies by the perplexity statistic.

Fig. 4.

Fig. 4

Open in a new tab

Proportions of the six colours used in each generation (Left: Rose type; Right: Animal type). As the transmission progressed, the relative frequency of the colours became uneven

Figure 5 A presents the global colour perplexities―based on the colour proportions in the entire colouring books―over generations together with the top 95% range of the random perplexities estimated by the Monte Carlo method (shaded areas). We identified roughly two trends within the ten chains. On the one hand, three of the five “rose” chains (A, H, J) and one “animal” chain (E) achieved small perplexities of 1.75–3.21 at generation 10. On the other hand, the remaining chains maintained perplexities of 4.33–5.65, indicating the preservation of colour diversity. The overall statistical significance of these perplexities, aggregated across the five “rose” chains, was Inline graphic, and that of the “animal” chains was Inline graphic; thus, the global perplexities were not significantly smaller than those of random colour ratios.

Fig. 5.

Fig. 5

Open in a new tab

A Line plots of the global colour perplexities for each generation (Left: Rose type; Right: Animal type). Shaded regions represent the 95-percentile range of the random perplexities estimated by the Monte Carlo method. B Line plots of the geometric-average perplexity within each painting (Left: Rose type; Right: Animal type). Shaded regions represent the 95-percentile range of the random perplexities estimated by the Monte Carlo method

By contrast, when we segmented the rose paintings into eight parts―distinguishing seven flowers and grouping all the leaves and stems into one3―their geometric-average perplexity within the paintings ranged between 1.3066 and 3.1203 at generation 10 (Fig. 5B). Similarly, the geometric-average perplexity over the five individual animals was 1.4500–3.3388. The overall statistical significance of the “rose” chain was Inline graphic and that of the “animal” chains was Inline graphic. These results indicate that the distinct segments/objects were painted only with a few primary colours.

Local Consistency of Colouration

Figure 6 A presents the counts of colour match between neighbouring areas per generation and chain. In addition, Fig. 6B plots the proportion of the random colour re-assignments that resulted in greater match counts than the observed values. As expected from the analysis of perplexities in the previous section, we observed an increasing trend over generations, and the colour-matched neighbours at generation 10 amounted 41.07–80.16% in the “rose” chains and 50.44–84.96% in the “animal” chains. The overall statistical significance of the colour match rates was Inline graphic for the “rose” chains (aggregated across the five chains; not significant mainly due to the chain labelled J) and Inline graphic for the “animal” chains (no random colour permutation yielded greater counts than any of the observed counts).

Fig. 6.

Fig. 6

Open in a new tab

A Line plots of the counts of colour match between neighbouring areas for each generation (Left: Rose type; Right: Animal type). B Line plots of p-value (the proportion of the random colour re-assignments that resulted in greater match counts than the observed values) of each generation

Discussion

This study explored an emergence of organized and systematic painting patterns using the experimental paradigm of cultural transmission. Consequently, on the one hand, we found simple and consistent colouring patterns in the last few generations of the transmission chains, with each segment (rose petals vs. leaves) and object (individual animals) painted with only two or three colours. On the other hand, the familiar and plausible colours of petals, leaves or animals were not always observed, contrary to our pre-experimental predictions. We quantitatively confirmed this high-level generalization through the analysis of the colour perplexities and colour match between neighbouring areas. In addition, we also showed that fewer transmission errors were made in the later generations of the transmission, which indicates an emergence of easily communicable patterns.

Under the assumption of Bayesian inference, paintings of the last few generations of the transmission chains represent participants’ prior beliefs on the plausible painting patterns (Griffiths & Kalish, 2007). In this view, the present results would reflect a high-level bias towards simpler and more consistent colouring patterns, rather than a bias towards the familiar colour patterns recalled to us for specific objects (e.g. the belief that leaves are green). Intensive use of a few colours, quantitatively represented as the low colour perplexities, is likely to be strongly influenced by our way of looking at familiar scenes; we tend to perceptually identify or segment objects by their consistent colours amongst diverse visual information. In addition, preference for simpler patterns is consistent with the previous attempts employed in Bayesian models of various cognitive domains, including language learning (Blei et al., 2003; Feldman et al., 2013; Morita & O’Donnell, 2022; Teh et al., 2006), categorization of visually perceived objects (Salakhutdinov et al., 2012) and classification of natural numbers (Tenenbaum, 1999a, b).

Prior to conducting the experiment, we expected an emergence of more realistic colouring patterns, such as red on the rose petals and green on the leaves, in the outcomes of the transmissions. Indeed, we observed such realistic colouration in some of the transmission chains, but they were not the majority (2 or 3 out of 10). Instead, the transmission results are best characterized by the slight dominance of brown and the relatively little use of blue in both the “rose” and “animal” paintings. This finding is the principal contribution of the present study.

Retaining the assumption of Bayesian inference, one possible account for the result is that the participants were not strongly guided on the specific contents of the line drawings (e.g. petal, leaf, dog, parrot, etc.). Instead, their bias might have followed more abstract, higher-level colouring pattern shared across general objects or their segments, only avoiding universally rare patterns with too many colours in segments.

Alternatively, our results could also be interpreted as reflecting memory limitations; simple and consistent colouring patterns are more easily and precisely transmittable, and therefore are more likely to survive a long transmission chain (Miller, 1956). This interpretation aligns with Tamariz and Kirby (2015), who observed that simplified and structured patterns emerged during the transmission of line drawings. They argue that the compressibility of transmitted signals is a key determinant for their long-term survival under repeated transmission.4

In either case, the deviation from familiar colouration observed in this study may elucidate the emergence of more abstract, symbolic forms of artwork in the human history. Zeki (1999) argues that paintings are not literal replications of the visually perceived world, but are instead shaped by symbolic and categorical perceptual processes. Conversely, cultural transmission provides an algorithmic account of how such abstract artworks could arise from attempts at faithful reproduction; even when individuals strive to memorize and reproduce what they observe as precisely as possible, a chain of such replications can unveil their common, underlying bias towards simpler and more consistent patterns, which are consequently advantageous for communication purposes. Similar findings have been reported in the previous studies on artificial languages and rhythms, which demonstrate the emergence of morphologically systematic lexicons (Kirby et al., 2008, 2015; Perfors & Navarro, 2014; Silvey et al., 2015) and integer-rational rhythmic structures (Jacoby et al., 2024; Jacoby & McDermott, 2017; Nave et al., 2024; Ravignani et al., 2016), respectively. Thus, cultural transmission offers a universal hypothesis for the origin of various organized artifacts and/or behaviours of humans.

Finally, we acknowledge several limitations of the present study. First, existing artworks exhibit substantial diversity, as exemplified by the markedly different colouring styles observed in realism and abstract expressionism. Within the Bayesian framework, such complexity makes it difficult to exhaustively sample prior knowledge from a limited number of participants. Furthermore, recent studies have begun to explore extensions of the original serial transmission paradigm (e.g., by introducing communication within the same generation; Kirby et al., 2015). It remains an open question what kinds of patterns would emerge in paintings under such extended conditions, and a comprehensive comparison among different transmission designs—including a baseline condition without any transmission (e.g., Derex et al., 2024)—will be required necessary in future investigations.

Supplementary Information

Below is the link to the electronic supplementary material.

Supplementary Material 1 (2.9MB, pdf)

Acknowledgements

We are grateful to Naofumi Nakagawa, Ryotaro Uemura, Reiko Sawada to advise us to recruit the participants in Kyoto and Keio Universities; Kanako Iriguchi for her help in stimulus construction; Riko Goto, Sana Kohmoto and Satomi Araya for their administrative works.

Biographies

Mayuko Iriguchi

is currently an assistant professor at Hamamatsu University School of Medicine. She completed her graduate studies at Kyoto University in Japan, focusing primarily on the psychological and cultural background of color perception and cognition in humans.

Sota Kikuchi

was a graduate student at the University of Tokyo in Japan when this research was conducted.

Takashi Morita

is a designated senior assistant professor at Chubu University, Japan. He received his Ph.D. in linguistics from Massachusetts Institute of Technology. His primary research interests include computational modeling of human language acquisition using machine learning techniques, with particular expertise in Bayesian inference and deep learning. He has also been working on computational studies of animal behavior, including vocalizations and biologged movement trajectories.

Hiroki Koda

is currently working at the University of Tokyo in Japan. He completed his undergraduate and graduate studies at Kyoto University in Japan. He has published several peer-reviewed papers, books, and chapters in books in the area of primatology and comparative cognitive science. His research focuses on understanding the primate origins of human uniqueness such as language, mind and culture from the perspectives of the biological and cultural evolution.

Author contributions

All authors contributed to the study conception and design. Material preparation and data collection were performed by Mayuko Iriguchi, Takashi Morita and Hiroki Koda. Analysis was performed by Mayuko Iriguchi and Sota Kikuchi. The first draft of the manuscript was written by Hiroki Koda and all authors commented on previous versions of the manuscript. The revision was led by Takashi Morita and Hiroki Koda, and all authors commented. All authors read and approved the final manuscript.

Funding

Open Access funding provided by The University of Tokyo. This study was funded by the JSPS KAKENHI (#17H06380/Evolinguistics, #21K18138 awarded to HK; #22K18454 to HK; #23H05428 to HK).

Data Availability

All painting data were shown in supplementary materials as thumbnail (i.e. miniature images), and analysis code was available in our repository: https://figshare.com/s/b62eb68df1027873883c. The original painting data would be available upon a reasonable request to authors.

Declarations

Ethics Approval

Our experiment complied with the Guidelines for Research in Human Participants, and was approved by the Human Research Ethics Committee of Primate Research Institute, Kyoto University (Permit No. 2018-03). Note that three of the authors (MI, TM, and HK) are now affiliated with different institution; however, all data reported here were collected during their affiliation with Kyoto University.

Consent to Participate

Written informed consent was obtained from all participants before the experiment commenced.

Competing interests

The authors have no competing interests to declare that are relevant to the content of this article.

Footnotes

1

Previous studies on colour perception have used greyscale images as their stimuli (Hansen et al., 2006; Olkkonen et al., 2008). However, such patterns are not suitable for our framework, utilizing cultural transmission, because greyscale images cause a mismatch with the patterns produced by participants (i.e., they have different supports of colour distributions), as “grey” was not included among the available painting options.

2

This implies that there was variation in elapsed time: some participants exceeded the time limit (particularly in earlier generations, where the stimuli were more complex and therefore harder to memorise/reproduce), whereas others completed the task more quickly (in later generations, when the stimuli had become simpler). We did not record the exact elapsed time for each individual participant and are therefore unable to analyse its potential effects on the results. Relatedly, our analyses did not take into account individual differences in memory ability, consequently treating participants as uniform agents.

3

This grouping reflects familiar colour diversity, as rose flowers vary in colour, whereas leaves and stems are uniformly green.

4

It should be noted that greater memorability does not necessarily lead to the preservation of a pattern. Rafferty et al. (2013) demonstrated that an irregular item within an otherwise regular set (e.g., “elephant” in a shopping list) was indeed more memorable, yet nonetheless disappeared from the transmitted information. Crucially, such irregular items do not re-emerge, owing to the low probability of their spontaneous reintroduction. A similar situation could in principle arise in our painting paradigm (e.g., a patch of an irregular colour surrounded by otherwise uniform regions would stand out and thus be memorable, consistent with the Von Restorff effect; von Restorff, 1933). However, quantifying such effects lies outside the scope of the present study.

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Mayuko Iriguchi and Sota Kikuchi contributed equally to this work.

References

  1. Aubert, M., Brumm, A., Ramli, M., Sutikna, T., Saptomo, E. W., Hakim, B., Morwood, M. J., van den Bergh, G. D., Kinsley, L., & Dosseto, A. (2014). Pleistocene cave Art from Sulawesi, Indonesia. Nature, 514(7521), 223–227. 10.1038/nature13422 [DOI] [PubMed] [Google Scholar]
  2. Bartlett, F., & Charles (1932). Remembering: A study in experimental and social psychology. Cambridge University Press.
  3. Bates, D., Mächler, M., Bolker, B., & Walker, S. (2015). Fitting linear Mixed-Effects models using lme4. Journal of Statistical Software, 67(1). 10.18637/jss.v067.i01
  4. Blei, D. M., Ng, A. Y., & Jordan, M. I. (2003). Latent dirichlet allocation. The Journal of Machine Learning Research, 3(null), 993–1022. [Google Scholar]
  5. Bradski, G. (2000). The OpenCV library. Dr Dobb’s Journal: Software Tools for the Professional Programmer, 25, 120–123. [Google Scholar]
  6. R Core Team (2023). R: A Language and Environment for Statistical Computing (Version 4.2.3). R Foundation for Statistical Computing. https://www.R-project.org/
  7. Deacon, T. (1998). The symbolic species: The co-evolution of Language and the brain. WW Norton & Company.
  8. Derex, M., Edmiston, P., Lupyan, G., & Mesoudi, A. (2024). Trade-offs, control conditions, and alternative designs in the experimental study of cultural evolution. Proceedings of the National Academy of Sciences, 121(48), e2322886121. 10.1073/pnas.2322886121 [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Feldman, N. H., Griffiths, T. L., Goldwater, S., & Morgan, J. L. (2013). A role for the developing lexicon in phonetic category acquisition. Psychological Review, 120(4), 751–778. 10.1037/a0034245 [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Griffiths, T. L., & Kalish, M. L. (2007). Language evolution by iterated learning with bayesian agents. Cognitive Science, 31(3), 441–480. 10.1080/15326900701326576 [DOI] [PubMed] [Google Scholar]
  11. Hansen, T., Olkkonen, M., Walter, S., & Gegenfurtner, K. R. (2006). Memory modulates color appearance. Nature Neuroscience, 9(11), 1367–1368. 10.1038/nn1794 [DOI] [PubMed] [Google Scholar]
  12. Hasegawa, Y., Tamura, H., Nakauchi, S., & Minami, T. (2024). Facial expressions affect the memory of facial colors. Journal of Vision, 24(5), 14. 10.1167/jov.24.5.14 [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Hoffmann, D. L., Standish, C. D., García-Diez, M., Pettitt, P. B., Milton, J. A., Zilhão, J., Alcolea-González, J. J., Cantalejo-Duarte, P., Collado, H., De Balbín, R., Lorblanchet, M., Ramos-Muñoz, J., Weniger, G. C., & Pike, A. W. G. (2018). U-Th dating of carbonate crusts reveals neandertal origin of Iberian cave Art. Science, 359(6378), 912–915. 10.1126/science.aap7778 [DOI] [PubMed] [Google Scholar]
  14. Jacoby, N., & McDermott, J. H. (2017). Integer ratio priors on musical rhythm revealed cross-culturally by iterated reproduction. Current Biology, 27(3), 359–370. 10.1016/j.cub.2016.12.031 [DOI] [PubMed] [Google Scholar]
  15. Jacoby, N., Polak, R., Grahn, J. A., Cameron, D. J., Lee, K. M., Godoy, R., Undurraga, E. A., Huanca, T., Thalwitzer, T., Doumbia, N., Goldberg, D., Margulis, E. H., Wong, P. C. M., Jure, L., Rocamora, M., Fujii, S., Savage, P. E., Ajimi, J., Konno, R., & McDermott, J. H. (2024). Commonality and variation in mental representations of music revealed by a cross-cultural comparison of rhythm priors in 15 countries. Nature Human Behaviour, 8(5), 846–877. 10.1038/s41562-023-01800-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Kimura, A., Wada, Y., Masuda, T., Goto, S., Tsuzuki, D., Hibino, H., Cai, D., & Dan, I. (2013). Memory color effect induced by familiarity of brand logos. PLOS ONE, 8(7), e68474. 10.1371/journal.pone.0068474 [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Kirby, S., Cornish, H., & Smith, K. (2008). Cumulative cultural evolution in the laboratory: An experimental approach to the origins of structure in human Language. Proceedings of the National Academy of Sciences, 105(31), 10681–10686. 10.1073/pnas.0707835105 [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Kirby, S., Tamariz, M., Cornish, H., & Smith, K. (2015). Compression and communication in the cultural evolution of linguistic structure. Cognition, 141, 87–102. 10.1016/j.cognition.2015.03.016 [DOI] [PubMed] [Google Scholar]
  19. Makuuchi, M., Kaminaga, T., & Sugishita, M. (2003). Both parietal lobes are involved in drawing: A functional MRI study and implications for constructional apraxia. Cognitive Brain Research, 16(3), 338–347. 10.1016/S0926-6410(02)00302-6 [DOI] [PubMed] [Google Scholar]
  20. Martin, A. D., Quinn, K. M., & Park, J. H. (2011). MCMCpack: Markov chain Monte Carlo in R. Journal of Statistical Software, 42(9). 10.18637/jss.v042.i09
  21. Miller, G. (1956). The magical number seven, plus or minus two: Some limits on our capacity for processing information. Psychological Review, 63(2), 81–97. [PubMed] [Google Scholar]
  22. Mithen, S. (1996). The prehistory of the Mind a search for the origins of art, religion and science. Thames and Hudson.
  23. Morita, T., & O’Donnell, T. J. (2022). Statistical evidence for learnable lexical subclasses in Japanese. Linguistic Inquiry, 53(1), 87–120. 10.1162/ling_a_00401 [Google Scholar]
  24. Nave, K., Carrillo, C., Jacoby, N., Trainor, L., & Hannon, E. (2024). The development of rhythmic categories as revealed through an iterative production task. Cognition, 242, 105634. 10.1016/j.cognition.2023.105634 [DOI] [PubMed] [Google Scholar]
  25. Olkkonen, M., Hansen, T., & Gegenfurtner, K. R. (2008). Color appearance of familiar objects: Effects of object shape, texture, and illumination changes. Journal of Vision, 8(5), 13. 10.1167/8.5.13 [DOI] [PubMed] [Google Scholar]
  26. Perfors, A., & Navarro, D. J. (2014). Language evolution can be shaped by the structure of the world. Cognitive Science, 38(4), 775–793. 10.1111/cogs.12102 [DOI] [PubMed] [Google Scholar]
  27. Rafferty, A. N., Griffiths, T. L., & Ettlinger, M. (2013). Greater learnability is not sufficient to produce cultural universals. Cognition, 129(1), 70–87. 10.1016/j.cognition.2013.05.003 [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Ravignani, A., Delgado, T., & Kirby, S. (2016). Musical evolution in the lab exhibits rhythmic universals. Nature Human Behaviour, 1(1), 0007. 10.1038/s41562-016-0007 [Google Scholar]
  29. Regier, T., & Kay, P. (2009). Language, thought, and color: Whorf was half right. Trends in Cognitive Sciences, 13(10), 439–446. 10.1016/j.tics.2009.07.001 [DOI] [PubMed] [Google Scholar]
  30. Saito, A., Hayashi, M., Takeshita, H., & Matsuzawa, T. (2014). The origin of representational drawing: A comparison of human children and chimpanzees. Child Development, 85(6), 2232–2246. 10.1111/cdev.12319 [DOI] [PubMed] [Google Scholar]
  31. Salakhutdinov, R., Tenenbaum, J., & Torralba, A. (2012). One-Shot Learning with a Hierarchical Nonparametric Bayesian Model. Proceedings of ICML Workshop on Unsupervised and Transfer Learning, 195–206. https://proceedings.mlr.press/v27/salakhutdinov12a.html
  32. Silvey, C., Kirby, S., & Smith, K. (2015). Word meanings evolve to selectively preserve distinctions on salient dimensions. Cognitive Science, 39(1), 212–226. 10.1111/cogs.12150 [DOI] [PubMed] [Google Scholar]
  33. Tamariz, M., & Kirby, S. (2015). Culture: Copying, Compression, and conventionality. Cognitive Science, 39(1), 171–183. 10.1111/cogs.12144 [DOI] [PubMed] [Google Scholar]
  34. Teh, Y. W., Jordan, M. I., Beal, M. J., & Blei, D. M. (2006). Hierarchical dirichlet processes. Journal of the American Statistical Association, 101(476), 1566–1581. 10.1198/016214506000000302 [Google Scholar]
  35. Tenenbaum, J. (1999a). A Bayesian framework for concept learning [Doctoral dissertation, Massachusetts Institute of Technology]. https://dspace.mit.edu/handle/1721.1/16714
  36. Tenenbaum, J. (1999b). Rules and Similarity in Concept Learning. In Advances in Neural Information Processing Systems. Vol.12. (pp. 59–65). MIT Press. https://papers.nips.cc/paper_files/paper/1999/hash/86d7c8a08b4aaa1bc7c599473f5dddda-Abstract.html
  37. von Restorff, H. (1933). Über die wirkung von bereichsbildungen Im Spurenfeld. Psychologische Forschung, 18(1), 299–342. 10.1007/BF02409636 [Google Scholar]
  38. Witzel, C., Valkova, H., Hansen, T., & Gegenfurtner, K. R. (2011). Object knowledge modulates colour appearance. I-Perception, 2(1), 13–49. 10.1068/i0396 [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Zeki, S. (1999). Inner vision: An exploration of Art and the brain. Oxford University Press.

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Material 1 (2.9MB, pdf)

Data Availability Statement

All painting data were shown in supplementary materials as thumbnail (i.e. miniature images), and analysis code was available in our repository: https://figshare.com/s/b62eb68df1027873883c. The original painting data would be available upon a reasonable request to authors.

Articles from Human Nature (Hawthorne, N.y.) are provided here courtesy of Springer

RESOURCES