Each of the above values is the nearest whole number to the mean of many observations except in the case of silver, which is based on one set only.

Occasionally, though not often, deflections were observed which were considerably less than usual for the metal used, and sometimes unequally less for the two metals being compared, thus showing less or more contrast between them-usually less. But in no case was the contrast even nearly obliterated, nor its sign changed. The cause or causes of these occasional irregularities have not yet been ascertained, but are diligently sought. In view of them, however, the values given in the table must be regarded as fairly good first approximation only.

Pure iron was tried because, magnetically, it is the antithesis of bismuth. But while it caused deflections neither very large nor very small, they were quite irregular, doubtless due to local disturbance of the earth's magnetic field. The slight residual magnetism after the iron cylinder was subsequently magnetized horizontally caused it to give widely different deflections when differently oriented. But from certain pendulum experiments next to be described it is thought that iron should have a place in Table I, somewhat nearer zinc than bismuth, say about 87.

The zinc and tin used in these experiments are of commercial purity only; the bismuth, silver and lead are almost chemically pure, and the aluminum contains traces of silicon and iron.

It is interesting to note that bismuth exhibits much the smallest attraction per unit of weight, as was hopefully predicted; and although this may not be due to its exceptional diamagnetic qualities, the possibility that it is so due promises a fruitful field for future exploration. If further experimentation seems to warrant it, I shall, in a future paper, endeavor to expand the idea that diamag

netism is not simply a manifestation of negative magnetism (and) this view is supported by the fact that diamagnetism is not affected by varying strength of magnetic field), but is an inherent attribute, in varying degree, of many kinds of matter, often more or less masked by paramagnetic qualities.

It is also interesting to note the low atomic weight and small density of aluminum at one end of the scale of attractions, and bismuth with its very high atomic weight and large density at the other end.

It seems needless to say that the line of investigation indicated by the foregoing experiments is but barely begun. It is proposed to try many other metals, as well as alloys and chemical compounds, in the hope of finding some general law embracing all.

If such further experimentation confirms, in general, the findings already made, and I have yet found no reason to doubt that this will be the case, the scientific possibilities of the discovery are rather bewildering. It may mean, among many things, that we shall have to revise our notions of atomic weight values; that the second part of Newton's law of gravitation is not general in its application, and that if Professor Boys and others had used zinc or tin or bismuth for their large attracting spheres, instead of lead, their findings for the gravitation constant and mean density of the earth would have been materially different, though still erroneous.

As soon as the above-described experiments had proceeded far enough to afford reasonable assurance that the effects observed were not spurious, though very much larger than looked for,. I planned and began work on some pendulum experiments in the hope of again showing that weight and mass are not related in so simple a manner as heretofore supposed.

Of course weight is only an accidental attribute of mass and varies enormously in different localities, as on the sun, the earth, the moon and in interstellar space; while mass remains constant everywhere. But relative weight has always been taken as a true measure of relative mass in the same locality. The foregoing experiments seem to refute this; hence the pendulum experiments were undertaken for proof or further refutation.

Undoubtedly the commonly accepted measure of mass, viz., its resistance to a definite accelerating force, positive or negative, is reliable and safe.

If, then, we have two pendulums of exactly the same real length, one with a zinc bob and the other with a bismuth bob, and find that their periods are not the same, we may reasonably infer that the accelerating force of gravity acts more strongly per unit of mass on the one having the shorter period than on the other. This is the principle, though not exactly the method, of the pendulum experiments next described.

It was realized from the start that the difference, if any exists, must be very small, and not easy to detect, because the earth's field is so enormously preponderating that it does virtually all the attracting; and the supposed differences in the extremely weak fields of the zine and bismuth may be so nearly lost in the immensity of the earth's field as to be undetectable. Nevertheless, it was thought worth while to try the experiment in view of the importance of the subject.

The photograph, Plate VI., shows the pendulum apparatus as originally installed, together with driving clocks at the top, added later for long-continued observations.

A starting cradle, moving in guides on the low table just below the cylindrical zine and bismuth bobs, serves to start the pendulums swinging exactly together in any desired amplitude. After pushing the bobs sufficiently to the left, the cradle is suddenly withdrawn to the right, leaving the bobs free. This device is entirely satisfactory in performance.

A horizontal thick plate of hardened steel is very firmly bolted to the lower flange of a heavy iron I beam imbedded in the masonry of the ceiling and walls of the room. The plate is dropped 6.5 cm. below the beam by cylindrical iron spacers through which the bolts pass and is carefully leveled. Near one edge of the upper face of the plate is a long shallow V grove of 90° angle, with a slightly rounded bottom carefully ground straight and polished after the plate was hardened.

From this plate hang two exactly similar pendulums of about