Algorithms and Philology in Dialogue: Reconstructing the Haṭhapradīpikā

The Haṭhapradīpikā is among the most influential premodern yoga texts, and its transmission is notoriously complex. With over two hundred manuscripts preserved in multiple recensions (ranging from three to ten chapters), the text presents an editorial challenge: pervasive contamination, numerous variant readings, and transposed verses complicate the construction of a traditional stemma.

In this study, I present the results of an editorial approach that combines classical philological analysis with computer-assisted stemmatics to reconstruct the most plausible visual representation of the text's transmission history. While computational trees are highly effective at grouping manuscripts and highlighting patterns of similarity—and here the computer is indeed much faster and more precise than the human mind—they often diverge from the true underlying stemma, especially under the influence of contaminated branches of transmission. The crucial insight is that no algorithm alone can resolve these cross-connections; careful philological scrutiny remains essential.

I will demonstrate how iteratively comparing algorithmically generated trees with the insights of traditional stemma-building methods of textual criticism resulted in a stemma that plausibly reflects both the primary lines of transmission and the complications introduced by contamination.

The presentation will highlight the interpretive insights gained from this hybrid approach, with particular attention to the final stemma of the new critical edition of the Haṭhapradīpikā. By showing how computational analysis and traditional philology can converge, this case study illustrates the potential of such methods to clarify intricate transmission histories and to strengthen the foundations of critical editions of highly complex manuscript traditions.