Why do languages tolerate heterography? An experimental investigation into the emergence of informative orthography

It is widely acknowledged that opaque orthographies place additional demands on learning, often requiring many years to fully acquire. It is less widely recognized, however, that such opacity may offer certain benefits in the context of reading. For example, heterographic homophones such as ⟨knight⟩ and ⟨night⟩ (words that sound the same but which are spelled differently) impose additional costs in learning but reduce ambiguity in reading. Here, we consider the possibility that—left to evolve freely—writing systems will sometimes choose to forego some simplicity for the sake of informativeness when there is functional pressure to do so. We investigate this hypothesis by simulating the evolution of orthography as it is transmitted from one generation to the next, both with and without a communicative pressure for ambiguity avoidance. In addition, we consider two mechanisms by which informative heterography might be selected for: differentiation, in which new spellings are created to differentiate meaning (e.g., ⟨lite⟩ vs. ⟨light⟩), and conservation, in which heterography arises as a byproduct of sound change (e.g., ⟨meat⟩ vs. ⟨meet⟩). Under pressure from learning alone, orthographic systems become transparent, but when combined with communicative pressure, they tend to favor some additional informativeness. Nevertheless, our findings also suggest that, in the long term, simpler, transparent spellings may be preferred in the absence of top-down explicit teaching.

Published paper: https://doi.org/10.1016/j.cognition.2024.105809
Preprint: https://osf.io/preprints/psyarxiv/qnz5k
GitHub repository: https://github.com/jwcarr/hethom

众所周知，晦涩难懂的字母文字系统对学习提出了额外的要求，通常需要数年才能完全掌握。然而，这种晦涩难懂的特性在阅读语境下可能带来某些益处，这一点尚未得到广泛认同。例如，异形同音字如“knight”和“night”（发音相同但拼写不同）虽然在学习过程中增加了成本，但减少了阅读时的歧义。在此，我们探讨了这样一种可能性：在功能压力下，若书写系统得以自由演化，它们有时会为了信息的丰富性而牺牲一定的简洁性。我们通过模拟从一代传至下一代的书写系统演化过程，无论是出于避免歧义的压力还是出于沟通的需要，来检验这一假设。此外，我们还考虑了两种可能导致信息丰富性异形书写系统被选中的机制：一是通过区分意义创建新的拼写（例如，“lite”与“light”），二是异形书写作为语音变化副产品出现（例如，“meat”与“meet”）。仅在学习压力下，书写系统趋向于透明化，但当与沟通压力相结合时，它们倾向于偏好额外的信息丰富性。然而，我们的研究也表明，在长期看来，如果没有自上而下的明确教学，简单、透明的拼写可能更受欢迎。