[Illustration:  FIG. 4.]

This source of error was avoided by fixing a polished steel rod or ‘guide’ at right angles to the vertical part of the stylus, just in front of the stylus; the stylus trailed against this rod, and could not spring out of position.  The friction of the rod did not modify the record, and the rod gave much greater certainty to the details of the sound curve, by fixing the position of the vibrating point.  This rod or guide is shown in Fig. 3 (g).

The disc was driven directly from the phonograph by a very simple method.  A fine chain was fixed to the shaft carrying the disc, and wrapped around a pulley on the shaft.  The chain was unwound by the forward movement of the recording apparatus of the phonograph against the constant tension of a spring.  When the phonograph apparatus was brought back to the beginning of a record which had been made, the spring wound up the chain, and the disc revolved back to its original position.

A T from the speaking-tube near the diaphragm box was connected by a rubber tube with the phonograph recorder, so that the voice of the speaker was recorded both on the smoked glass plate and on the phonograph cylinder.  The advantages of such a double record are that the possible error of a transcription process is eliminated, and yet there is an original record to which it is possible to refer, and by which the record measured may be checked.

An important feature in the method was the rate at which the disc revolved.  The disc turned so slowly that the vibrations, instead of being spread out as a harmonic curve, were closely crowded together.  This had two great advantages; the measurements were not so laborious, and the intensity changes were much more definitely seen than in the elongated form of record.  Each syllable had an intensity form, as a ‘box,’ ‘spindle,’ ‘double spindle,’ ‘truncated cone,’ ‘cone,’ etc. (cf. p. 446).

The disc was run, as a rule, at a rate of about one revolution in two minutes.  The rate could be varied to suit the purposes of the experimenter, and it was perfectly possible to procure the usual form of record when desired.  As a result of the low rate, the records were exceedingly condensed.  The records of the 300 stanzas measured are on two glass discs of about 25 cm. diameter, and as much more could still be recorded on them.

The diaphragm and the speaking tube were the great sources of error.  For measurements of time values the particular component of the tone to which the diaphragm happens to vibrate is not important, but the record of intensities depends on the fidelity with which the diaphragm responds to a given component, preferably the fundamental, of the tone.  The speaking tube has a resonance of its own which can be but partly eliminated.  For the records here recorded either glass or goldbeater’s skin was used as a diaphragm.  Goldbeater’s skin has the advantage of being very sensitive, and it must