Reconstruction as Inference to the Best Explanation

from a conversation with Claude AI

Comparative reconstruction (Proto-Indo-European etc.) is abduction — positing the ancestor form that best explains the attested daughters under known sound-change patterns. Structurally similar to inference in historical geology or cladistics: no direct access to the object, only convergent evidence.¹

Interesting epistemic question: what makes a reconstructed form more than a useful fiction? PIE speakers left no writing — the proto-language is a model, not a discovery, and yet linguists treat starred forms (*ph₂tḗr) with real confidence.

Open threads:

• Where does reconstruction’s confidence come from vs. textual criticism’s stemma-building (same logical shape, manuscripts instead of daughter languages)
• Note Saussure’s laryngeal theory as a case where reconstruction predicted a fact (Hittite ḫ) before it was found — a real confirmed prediction, rare in historical linguistics

References

Footnotes

1. Generally I’m somewhat skeptical both of if this is really how it was, it can of course change, and the value of this knowledge, (at least the funding put toward doing these things extensively). ↩