146 



THE STUDY OF SPEECH CURVES. 



ever, is of no use to us; we must know how it was actually produced. Now 

 we know that all speech curves are frictional ones; we have seen here 

 and before that a simple harmonic analysis of such curves gives results 

 that are utterly different from the truth. We are forced to the conclusion 

 that this kind of analysis— into harmonics or inharmonics— is not merely 

 inapplicable, but also distinctly misleading. It must cause some regret 

 to find this conclusion forced upon us. The various investigators of 

 speech curves in Germany, France, and America have, without hesitation 

 or critique, always appUed the simple harmonic analysis to vocal curves, 

 and, as explained above, have assumed a totally false theory of vocal action 

 on its suggestion. In fact, this theory'— the overtone theory— must be 



assumed in order to justify the analysis. For two years of the writer's 

 own work he employed the simple harmonic analysis with subsequent 

 calculation of the inharmonics, expecting that the discrepancies and incon- 

 sistencies that appeared in the results would disappear as more material 

 was accumulated. This did not occur. It was found that the results 

 for the same vowel under similar conditions differed often more than 

 the results for utterly different vowels. The untnistworthiness of the 

 methods did not become apparent until plotted artificial curves were 

 computed, whereby the results could be compared with the truth— a 

 method of procedure that every investigator ought to have used and still 

 ought to use parallel with his work on speech curves. The self-evident 

 fact that all vocal vibrations must have high coefficients of friction— 



