78 PROCEEDINGS OF THE AMERICAN ACADEMY, 



settees were of ash, and their general style is shown in the accompany- 

 ing sketch. 



An investigation was then begun in regard to the nature of the pro- 

 cess of absorption of sound. The material chosen for this work was a 

 very durable grade of felt, which, as the manufacturers claimed, was all 

 wool. Even a casual examination of its texture makes it difficult to 

 believe that it is all wool. It has, however, the advantage of being 

 porous, flexible, and very durable. Almost constant handling for sev- 

 eral years has apparently not greatly changed its consistency. It is to 

 be noted that this felt is not that mentioned in the papers of six years 

 ago. That felt was of lime-treated cow's hair, the kind used in pack- 

 ing steam pipes. It was very much cheaper in price, but stood little 

 handling before disintegrating. The felt employed in these experi- 

 ments comes in sheets of various thicknesses, the thickness here 

 employed being about 1.1 cm. 



The coefficient of absorption of a single layer of felt was measured 

 for the notes from Ci 64 to C7 4096 at octave intervals. The experi- 

 ment was repeated for two layers, one on top of the other, then for three, 

 and so on up to six thicknesses of felt. Because the greater thicknesses 

 presented an area on the edge not inconsiderable in comparison with 

 the surface, the felt was surrounded by a narrow wood frame. Under 

 such circumstances it was safe to assume that the absorption was 

 entirely by the upper surface of the felt. The experiment was repeated 

 a great many times, first measuring the coefficient of absorption for one 

 thickness for all frequencies, and then checking the work by conducting 

 experiments in the other order ; that is, measuring the absorption by 

 one, two, three, etc., thicknesses, for each frequency. The mean of all 

 observations is shown in Figure 12 and Figure 13. In Figure 12 the 

 variations in pitch are plotted as abscissas, as in previous diagrams, 

 whereas in Figure 13 the thicknesses are taken as abscissas. The special 

 object of the second method will appear later, but a general object of 

 adopting this method of plotting is as follows : 



If we consider Figure 12, for example, the drawing of the line through 

 any one set of points should be made not merely to best fit those 

 points, but should be drawn having in mind the fact that it, as a curve, 

 is one of a family of curves, and that it should be drawn not merely as 

 a best curve through its own points, but as best fits the whole set. For 

 example, in Figure 12 the curve for four thicknesses would not have 

 been drawn as there shown if drawn simply with reference to its own 

 points. It would have been drawn directly through the points for 

 Ci 64 and C2 128. Similarly the curve for five thicknesses would have 

 been drawn a little nearer the point for C2 128, and above instead of 



