Examination of the Phenomena with a Spectroscope of Wide Dispersion.

Green blindness was produced by the spectroscopic method already described. When the exposure was complete the slit was closed until the Fraunhofer lines were sharply defined. All sensation to green was lost, the red appearing to meet the blue in the centre of the field. The position of the junction of these two colours could be varied considerably by exposing the eye to strong red or strong blue light, thus showing that the red and blue overlap. But violet light had no effect upon the position of the junction of red with blue.

Similarly, during blue-blindness, the green and violet were seen to overlap, exposure to green light shifting the junction towards the green and vice versa. Red light had no effect on the position of the junction of green with violet during blue-blindness.

The phenomenon of flickering* visible between the red and the green of a highly magnified spectrum, is also seen at the junction of red with blue during green-blindness, and of green with violet during blue-blindness, as well as at the junctions of green with blue and of blue with violet under normal conditions.

The author has succeeded by an exposure of three minutes to light from between H and K in blinding the eye to violet without affecting the blue, the real hue of which is thus seen, unaccompanied by any other colour sensation.

These experiments lead to the conclusion that no one coloursensation is related to any other in the sense indicated by Hering. Each may be exhausted without either weakening or strengthening the others. The observed facts are, in the author's opinion, more in accordance with the Young-Helmholtz theory, but they imply the existence of a fourth colour sensation, namely, blue.

Examination of the Colour Sensations of 109 Persons.

The tests employed were Holmgren's wools, supplemented by gelatine films stained with various colours, Hering's method of coloured. shadows, and the author's spectroscopic method, which was applied to seventy normal cases in the following manner.

Using the large spectroscope referred to, with the slit narrow so as to give a comfortable degree of illumination, the observer selects those portions of the spectrum at which he sees a marked change of hue. He then looks at the red between A and B for thirty seconds, and at a given signal traverses the spectrum rapidly, stopping at the first of these changes.

Next he looks at the green for thirty seconds before turning to the

* Physiol. Soc. Proc.,' June, 1897.

he seeks the third violet to its limits. again, fatiguing the

second change of hue. Again, after looking at blue for thirty seconds change of hue. The next step is to trace the After this he works through the spectrum back eye with violet before finding the blue, and so on, ending with the determination of the limits of the red. The degree of fatigue is so slight that he is quite unconscious of it.

The seventy cases examined in this way agree as to the number and mean position of the changes of hue, but they may be divided broadly into those whose colour sensations overlap and those whose colour sensations do not overlap, i.e., those who find the changes of tint occur in the same place when working from red to violet as when returning from violet to red.

The first class includes persons both educated and uneducated whose avocations require them to compare colours. The second comprises all who fail with the closer shades of Holmgren's wools. Details are given of some in whom the green and violet are so far extended into each other that they see practically no pure blue, and it is suggested that these, and other differences in the relative intensity and extent of the colour sensations may account for the divergence of opinions among writers on the subject. The paper concludes with an account of five cases of red-blindness.

"On the Connection between the Electrical Properties and the Chemical Composition of different kinds of Glass." By Professor ANDREW GRAY, LL.D., F.R.S., and Professor J. J. DOBBIE, M.A., D.Sc. Received February 7,-Read February 17, 1898.

The experiments and results described in the following paper are a first instalment of work we have undertaken with a view to finally determining, if possible, the circumstances which affect the conductivity and specific inductive capacity of glass. It appeared from some experiments which were carried out by Professor T. Gray and ourselves some years ago,* that it might be of interest to have a number of glasses specially made up with a view to testing some of the conclusions then arrived at.

A result previously obtained by Professor T. Gray had shown that potash and soda lime glasses have a higher conductivity than flint glasses; this result had also been arrived at by Dr. Hopkinson. In particular it seemed desirable to ascertain whether by increasing the amount of lead oxide and diminishing the amount of soda, the conductivity would go on diminishing. We have experienced great

* 'Roy. Soc. Proc.,' No. 231, 1884.

difficulty in getting glasses made according to our own specifications. We endeavoured to make the glasses ourselves, and several experiments were made accordingly, both in the laboratories here and at the Ogwen Tile Works, where a large furnace had been erected for the construction of tiles from slate dust. Some success was achieved, but it was found impossible, without the expenditure of far more time than could be spared, to obtain the glasses in a condition suitable for the experiments we wished to carry out.

Through the kindness, however, of Messrs. Schott & Co., of Jena, and of Messrs. Powell & Sons, Whitefriars, London, we have recently obtained a number of specimens of glass all richer in lead than the specimens formerly available, and, further, in some cases practically free from soda. We have also had made to order by Messrs. Schott specimens of their own glass, used, we believe, chiefly in the construction of thermometers, as well as of a barium crown glass, which have not hitherto, so far as we are aware, been experimented with.

Determination of Conductivity.-The method of experimenting followed was practically the same as that described in the paper already referred to, but its nature may perhaps here be indicated.

Owing to the large percentage of lead oxide in some of the glasses prepared for us by Messrs. Schott, it was found impossible to blow them into flasks, and they were therefore cast into plates; the arrangements therefore required some modification for their case.

The specimens which were in the form of flasks were filled up with mercury to the bottom of the stem (which in most cases was about 8 or 9 inches long), and the flask thus filled was sunk in a bath also containing mercury, so that the mercury was at the same level inside and outside. One terminal of a circuit containing a battery of about thirty secondary cells and a very sensitive galvanometer was connected to the mercury within the flask by a wire passing down the neck, while the other terminal was connected to the mercury in the bath.

The galvanometer was the instrument formerly used and described in 'Roy. Soc. Proc.,' vol. 36, p. 287. It was carefully insulated, as was also the reversing key, and all necessary precautions were taken to make sure that the current passing through the galvanometer was that passing through the walls of the flask between the mercury coatings. Thus it was always verified that no deflection took place when the wire was withdrawn from the flask and placed round the outside of the neck. This test obviated the possibility of the existence of any disturbing film of moisture on the surface of the glass. The bath could be heated to any temperature required in. the experiments.

The conductivity was calculated from deflection of the galvano

meter produced by the current through the glass, the area and thickness of part of the specimen in which the current flowed, as described in the former paper.

In the capacity measurements the plate or flask, as the case might be, was supported as described above. A quadrant electrometer was kept connected to the plates of one of Lord Kelvin's air leydens. This was charged with twelve secondary cells, and therefore to a difference of potential of about 24 volts. After the battery had been removed a reading was taken of the electrometer deflection, and then the specimen was connected for a very short interval of time as a condenser in parallel with the leyden.

This connection was made by means of a myograph pendulum which, when freed, swung over a considerable arc to a catch which prevented it from returning. At its lowest point a metal piece projecting below the bob touched the top of a tongue projecting upwards from a hinge at its lower end, and leaning against the point of an adjustable screw. The connection between the two condensers thus only endured while the three pieces, the screw, tongue, and bob, were in contact. This was only the interval of time required for a pulse of flexure to travel about a centimetre in a bar of steel about half a centimetre thick, and about a centimetre broad. The interval was reckoned as at most about 1/30,000 of a second.

The plates were originally rather over a quarter of an inch thick, and after some observations of capacity had been made on some of them, and it had appeared that their resistance was too great to be measurable, they were cut down on a turning-table used for cutting slates, to a thickness of about 3 mm.; they were then fixed on a bed of pitch, and ground down by hand to a thickness of about 0-24 cm. They were polished and properly cleaned, and then covered on both sides with a dense and thoroughly adherent coating of silver. This was cut away for a space of about half an inch round the edges. Great care was taken to remove every trace of silver, and to make the edge thoroughly clean.

While being experimented on, the glass plate was laid with one coating of silver resting on a plate of copper at the bottom of an iron bath. Another plate of copper was laid on the upper coating of silver, and kept down with a weight, and the connections of the battery circuit, described above, were made with the copper plates. The iron bath was placed within a larger bath partly filled with sand, so that the temperature could be raised by heating the outer bath from below.

The results of the experiments are exhibited in the table which follows. We have there given the density, specific resistance, specific inductive capacity, and chemical composition of each specimen experimented on.