Transmission Chains as a Measurable Engine for Abstracting Heritage Motifs into Water-Shedding Surfaces

Heritage crafts often fail commercially for one of two reasons. Products copy ornate traditional imagery that younger buyers read as rustic, or they become decorative souvenirs with no use in daily life. Design scholars agree that the remedy is "abstraction": distilling a motif into a few geometric primitives that still carry its identity. In practice, though, abstraction is done by one designer's intuition. Nobody can say how much of a motif can be stripped away before it stops being recognizable. We propose to make abstraction a measurable process by borrowing a tool from cognitive science: the transmission chain. In a transmission chain, person 1 sees a motif briefly and redraws it, person 2 redraws person 1's drawing, and so on. Each copy loses detail in a direction set by human perceptual biases. The result is a sequence of progressively simpler versions of the motif. Along that sequence we can measure, for each version, how recognizable it remains and how well it serves a physical function.

The function we choose is self-cleaning water repellency. The motif is laser-written into a superhydrophobic surface: one so finely textured that water beads up and rolls off, the way it does on a lotus leaf. Here the motif is not decoration printed on top of the surface. Its lines become raised ridges that interrupt the micro-texture, and each ridge is a place where rolling drops can catch. Our central bet is that the same property that makes a motif simple for people, short total outline with few line ends, also makes it shed water well. If that holds, cultural abstraction and surface physics pull in the same direction, and a transmission chain can find the best compromise between them, which a designer working alone cannot locate systematically. The product target is a tea tray or saucer, where spilled liquid should drain cleanly. This is a daily-use object rather than a fridge magnet.

Core Claim

Iterated reproduction of a heritage motif produces a family of abstractions ordered by complexity. Within that family there is a measurable sweet spot: a generation at which recognition by both tradition holders and naïve viewers is still high, while the motif's outline length per unit area is low enough that a laser-textured superhydrophobic surface carrying it keeps its low sliding angle. We predict that chain-derived abstractions will dominate designer-made abstractions of matched complexity on recognition, and that sliding angle will be predicted mainly by outline-length density, not by motif identity. Both predictions can fail, and the experiments below are built to show which way they go.

How This Relates to Prior Work

Prior work supplies the method. Iterated-learning experiments (Iterated learning: Intergenerational knowledge transmission reveals inductive biases) passed one-dimensional functions through nine-generation chains and found that 28 of 32 chains collapsed to a positive linear function within three to five generations. Work on social sampling (Overcoming Individual Limitations Through Distributed Computation) showed that multigenerational populations can pool evidence beyond what any individual could. Both studies used abstract laboratory stimuli. We apply the same chain method to culturally loaded two-dimensional motifs. The 28-of-32 result also acts as a warning: left unchecked, chains erase content into generic priors. We therefore treat the generation number as a dial to be tuned, not something to run to convergence.

A second group of papers motivates the gap. Design studies of Huangmeixi opera visuals (Research on the Development and Visual Image of Opera Culture Creative Products) decompose motifs into points, lines, colors and the circular "group flower" archetype. The textile-heritage framework of Visual Transformation of Textile Intangible Cultural Heritage Patterns separates cultural invariants from modern reorganization and calls for functional daily objects. The work on cattle idiom graphics (Modern Creative Expression of Chinese Traditional Graphic "Cattle") reports that heavy abstraction leaves untrained viewers unable to recognize the source without explanatory text. All three are qualitative and judged by the designer. We keep their decomposition vocabulary but replace intuition with measured recognition curves. The textile paper also notes that nobody theorizes the physical "needle" side of craft. Our physical-texture coupling is one concrete answer to that gap.

The evaluation design comes from the purple clay teapot study (Research on Quantification and Design Evaluation of Visual Perception Beauty of Purple Clay Teapot). That study paired measured geometry with semantic-differential ratings, a survey method in which people rate an object on scales between opposite adjectives, collected from 109 participants. It found that the Golden Ratio scored lowest, which shows that design theory and perception can disagree measurably. We reuse its rating method and its stance that preference claims must be tested, not assumed.

The physics side constrains the design. Reviews of femtosecond-laser fabrication (The Design and Analysis of the Fabrication of Micro- and Nanoscale Surface Structures) establish that ultrafast lasers can write lotus-like hierarchical textures with small heat-affected zones, but slowly. Machine-learning surrogates for wetting (Computational Intelligence-Assisted Understanding of Nature-Inspired Superhydrophobic Behavior) showed that sliding angle and adhesive force conflict, with Pareto solutions spanning sliding angles of roughly 5–12° and adhesion of 0–3 mN on SiO₂/PVC composites. The convex-hull work on harvester tracks (Design for Copying Grouser and Bionic Convex Hull Patterns) shows that mesoscale geometry decides whether liquid films are broken or anchored. That is exactly the role motif ridges play on our surfaces. None of these papers asks what happens when the mesoscale geometry is a culturally meaningful figure rather than an engineered lattice.

The closest work that crosses directions, the study of Diao Cui kingfisher-feather jewelry (Modern Improvement and Development of Traditional Jewellery Embellishment Craft), links bio-derived surface properties, market repositioning and curricular transmission. It does so qualitatively and without engineered micro-texture or any model of transmission.

Research Program

We will select about twelve source motifs from documented traditions that have accessible tradition holders: Huangmei pick-flower embroidery, opera group-flower roundels, and lantern motifs of the kind catalogued for the Qianjiao Lamp (The protection and publicity of Qianjiao Lamp in Dongguan). We will run transmission chains of ten generations per motif, with several independent chains per motif. Each participant views the previous drawing for a few seconds and redraws it on a tablet in a single stroke-based canvas. Chains come in two types. "Youth" chains contain only young adults outside the tradition. "Mixed" chains place a tradition holder at the first and fifth positions. This tests whether reinjecting expert knowledge slows the loss of identity, which is the transmission question.

A third chain variant adds a fabrication filter. Before a drawing is passed on, it is rasterized and checked against laser-writing rules: minimum ridge width and spacing, and maximum outline-length density. Violations are snapped or pruned automatically, so later generations inherit only manufacturable geometry. The filter thresholds will be calibrated in a pilot of uniform-line test coupons rather than fixed in advance.

Every drawing is scored on three axes. Complexity is total outline length per unit area, plus counts of line ends and closed regions. Recognition is measured with forced-choice identification among motif foils, judged separately by tradition holders and by naïve viewers. Preference uses semantic-differential ratings following the teapot study, collected from a younger consumer panel.

For the physical side, selected generations are femtosecond-laser written onto stainless steel coupons as ridges on a uniform laser-textured, hydrophobized field. On each coupon we measure static contact angle, sliding angle, contact-angle hysteresis (the gap between the angles at a drop's advancing and receding edges, which indexes how strongly it sticks), and retention of these after standardized abrasion and tea-staining cycles. A regression of sliding angle on complexity measures, with motif identity as a factor, tests whether outline-length density alone explains wetting. Finally, two or three sweet-spot designs are made as tea-tray prototypes for a hands-on preference session.

Baselines

The naive reference is a uniform superhydrophobic coupon with no motif. It sets the best achievable sliding angle and durability, and every motif coupon is reported relative to it. The SiO₂/PVC Pareto range of 5–12° sliding angle from the computational-intelligence wetting study is a sanity reference only, because it concerns a different material. On the abstraction side, the naive reference is the unabstracted source motif written at full detail.

The strongest existing method for abstraction is expert designer decomposition into points, lines and archetypes, as practiced in the opera-products study. We will recruit three professional designers and give each the same source motifs, the same fabrication rules and the same time budget. Each produces a ladder of five abstractions at increasing simplicity. Their designs are then scored by the same recognition panels and fabricated on the same coupons. A chain-derived abstraction beats the designers if it is recognized more often by naïve viewers at matched outline-length density, within a tolerance band, with no loss among tradition holders. No reported result found on Lacuna for recognition of designer-abstracted heritage motifs, so this comparison sets the first number.

The generative-model baseline uses an image generator conditioned on the source motif and prompted for "minimal line-art versions" at controlled stroke counts. Its outputs are passed through the same fabrication filter, and its sampling budget is matched to the chains' total number of drawings. The only reported figures in the file for generative design concern shoe images: the conditional StyleGAN2 system in A Trend-Driven Fashion Design System for Rapid Response Marketing in E-commerce reached 0.79 trend accuracy, against 0.82 for a GPT+VQ-VAE2 baseline. That metric does not transfer to motif recognition, and no reported result found on Lacuna for generative abstraction of heritage motifs.

Two ablations isolate the key ingredients. Removing the fabrication filter from the chains tests whether embedding constraints during transmission matters, or whether pruning afterward works just as well. Replacing ridge-written motifs with motifs engraved flush, which changes color and reflectance but not topography, separates the visual effect from the wetting effect.

Why not just…?

Why not just write a uniform water-repellent texture and print or engrave the motif flush on top? This is the most serious threat. It separates function from imagery entirely, and the motif may then cost nothing in wetting. We take it seriously by including it as an ablation. If flush motifs match the uniform baseline on sliding angle and match ridge motifs on preference, our coupling claim fails, and the project reduces to a clean measurement study of chain abstraction. That would still be useful, but it would be smaller. We expect ridges to matter for two reasons. The convex-hull results show that mesoscale relief decides whether liquid films are broken or anchored. In addition, a tactile, light-catching relief is part of what buyers perceive. Both effects are measured directly rather than assumed.

Why not just let a skilled designer abstract the motif and check wetting afterward? Designers are the incumbent, and they may well produce more attractive single designs. What they do not produce is a complexity-ordered family with measured recognition at each step. The cattle-idiom study's finding that abstraction can silently destroy recognizability is exactly the failure that a curve exposes and an intuition hides. The matched-complexity comparison decides the question. If designers equal the chains, the chain method's value narrows to cost and the ability to involve communities.

Why not just ask an image generator for simplified versions? Generators are fast, but they simplify toward whatever is typical in their training data, not toward human perceptual priors or the motif's cultural invariants. The textile-heritage paper's warning about "transplanting shape" rather than meaning applies directly. The budget-matched generator baseline with an identical fabrication filter tests this. If generated abstractions match chains on tradition-holder recognition, we would recommend the generator for practice.

Why not just apply a hydrophobic spray to an ordinary smooth product? Sprays deliver beading but not durable roll-off. The ethics-of-materials argument in the Diao Cui study also weighs against chemical-heavy finishes. Even so, a sprayed smooth coupon is cheap to add, and we will include it in the durability tests so that the extra cost of laser texturing must justify itself against it.

Decisive Checks

The project succeeds on the cultural side if recognition-versus-complexity curves from chains sit above the designer and generator curves for naïve viewers, without dropping for tradition holders. It succeeds on the physical side if sliding angle on ridge-motif coupons is well predicted by outline-length density across motifs, and if some sweet-spot generation stays close to the uniform baseline after abrasion. It succeeds on the transmission side if mixed chains hold recognition longer than youth chains at equal generation. The cleanest falsification would be any of three outcomes: chains collapse to generic shapes before recognition plateaus, which is what the 28-of-32 linear-collapse result warns of; motif identity rather than outline-length density drives wetting; or flush motifs perform as well as ridges. Each of these would redirect the work, and each is measured.