My analysis was directed toward purely physical conditions, such as defective wiring, presence of lack of air spaces between metal flues and woodwork, etc., and the results were presented in these terms.

Collected Papers on Metalinguistics (ed. 1952)