Have we any clew to the cause of the opacity? I think we have. Villejuif is close to Paris, and over it, with the observed light wind, was slowly wafted the air from the city. Thousands of chimneys to windward of Villejuif were discharging their heated currents; so that an exceeding non-homogeneous atmosphere must have surrounded that station.[91] At no great height in the atmosphere the equilibrium of temperature would be established. This non-homogeneous air surrounding Villejuif is experimentally typified by our screen, with the source of sound close behind it, the upper edge of the screen representing the place where equilibrium of temperature was established in the atmosphere above the station. In virtue of its proximity to the screen, the echoes from our sounding-reed would, in the case here supposed, so blend with the direct sound as to be practically indistinguishable from it, as the echoes at Villejuif followed the direct sound so hotly, and vanished so rapidly, that they escaped observation. And as our sensitive flame, at a distance, failed to be affected by the sounding body placed close behind the cardboard screen, so, I take it, did the observers at Montlhéry fail to hear the sounds of the Villejuif gun. Something further may be done toward the experimental elucidation of this subject. The facility with which sounds pass through textile fabrics has been already illustrated,[92] a layer of cambric or calico, or even of thick flannel or baize, being found competent to intercept but a small fraction of the sound from a vibrating reed. Such a layer of calico may be taken to represent a layer of air, differentiated from its neighbors by temperature or moisture; while a succession of such sheets of calico may be taken to represent successive layers of non-homogeneous air. [Illustration: FIG. 3.] Two tin tubes (M N and O P, Fig. 3) with open ends were placed so as to form an acute angle with each other. At the end of one was the vibrating reed _r_; opposite the end of the other, and in the prolongation of P O, the sensitive flame _f_, a second sensitive flame (_f′_) being placed in the continuation of the axis of M N. On sounding the reed, the direct sound through M N agitated the flame _f′_. Introducing the square of calico _a b_ at the proper angle, a slight decrease of the action on _f′_ was noticed, and the feeble echo from _a b_ produced a barely perceptible agitation of the flame f. Adding another square, _c d_, the sound transmitted by _a b_ impinged on _c d_; it was partially echoed, returned through _a b_, passed along P O, and still further agitated the flame _f_. Adding a third square, _e f_, the reflected sound was still further augmented, every accession to the echo being accompanied by a corresponding withdrawal of the vibrations from _f′_, and a consequent stilling of that flame.