likely to give correct indications, may really be in error two or three degrees Fahrenheit at some part of their scales. Thus the "fever" thermometers in general use by physicians are almost invariably too high in their readings. An analysis of the results of sixty-eight thermometers of this description, verified in June of this year, will show how great this error may come to be: one fifth had errors less than 0·1°; one fifth had errors less than 0·4° but more than 0·2°; two fifths had errors less than 0·7° but more than 0·4°; one fifth had errors less than 1·0° but more than 0·7°; and occasionally a thermometer was found which had errors exceeding 1° and more rarely one exceeding 2°. The thermometers on which the above deductions rest were chosen to represent seven makers, and may be fairly taken to indicate the liability to error in using fever-thermometers which have not been compared with authoritative standards. It is not unlikely that members of the medical profession have been sometimes misled by the readings of inaccurate thermometers, and they may have made such unfavorable statements regarding the chances of recovery of patients whose temperatures were high, that the patient, under the influence of his imagination, has given up the struggle for existence when a little more hopeful view of the case might have imbued him with fresh courage and led to ultimate recovery.
The work at the Yale Observatory divides itself into two parts—the establishing of the standard thermometers with which thermometers sent to the observatory are to be compared, and the-work of comparing thermometers. The investigation of the standards themselves is by far the most tedious of the two; and as the methods used in studying the observatory standards are also the methods used, with greater or less detail, in investigating the higher grades of thermometers sent to the observatory, the methods will be briefly outlined.
It will be necessary to recall some of the fundamental principles of thermometry, however, in order to properly comprehend the methods of procedure in the case of standards:
1. Glass mercurial thermometers slowly increase their freezing-point readings as their age increases after the heating they undergo in filling with mercury in their manufacture.
2. The readings of the boiling-points are also increased, but in a much less degree—perhaps not more than one fifth as much as the freezing-point.
3. Whenever the thermometer is heated at all, the freezing-point is lowered, and the amount of this depression is very nearly proportional to the square of the temperature to which the thermometer is heated.
4. It follows from 3, that if a thermometer is kept at the ordinary temperatures, the freezing-point of water will be indicated by a lower scale reading than if the thermometer is kept at a low temperature. Now, if we suppose the thermometer has been kept at the freezing-point of water for a period of several days, and that the progressive
