to get an instrument accurate to within the sphere of other unavoidable errors incident to its uses. Paying for a greater degree of accuracy than this is only so much money wasted. Properly accurate thermometers are always to be had at prices which pretty closely correspond to their true value, but they never can be had at the prices at which the makers are obliged to sell a large proportion of the instruments which are sold and in use, simply because the prices are below the cost of the time, skill and labor essential to the construction of a properly accurate instrument. Prices paid to dealers, however, are not always indications of either true value or of the prices realized by makers. A thermometer good-looking enough to be bought cheaply and sold dearly is the primary object of many dealers. It becomes very necessary to know how to select a good thermometer independently of its appearance and price, and to aid in such knowledge is the object of this note.

Beside the errors which arise from imperfect tubes, a low degree of skill, and the use of incorrect standards in making thermometers, there is one very important source of large error that is not so generally known, namely, the contraction of the glass of which the bulb is made.

It was long ago shown that glass undergoes a process of contraction after having been melted, which, from being comparatively rapid at first, becomes slower and slower during many years; and that glasses of different composition undergo different degrees of contraction, and at different rates. Regnault's observations (given from memory) showed that the contraction noticeable by very careful experiments continued for six years or more, but that after the first three or four years it was very slight and very slow, hardly appreciable to the finest measurements, and therefore not at all important to any but the very finest standard instruments.

It was also shown that if during the course of this contraction, or at any time after it had been completed, the glass was re-heated, even to a moderate degree, and far short of melting, it again expanded to an extent proportionate to the heating; but that after re-heating it contracted much more rapidly than at first, and so regained the degree of contraction which might have required two years to attain, within a few months.

From these facts it follows that a thermometer which is graduated as soon as it is made, though made with proper care and skill, and graduated correctly from a good standard, will have its mercury column pushed up so that the readings from it will all be

too high, and this error will increase through many years. This contraction, however, becomes so slight and so slow that after three years, perhaps, it could not be detected in any ordinary thermometer by any ordinary means of observation. And farther, it may be stated as being probable that during the last six months of the three years the contraction in any ordinary thermometer glass would not exceed two-tenths of a degree. Then, as this two-tenths of a degree is within the sphere of other common errors of construction and uses of clinical thermometers, it might be disregarded in view of the additional cost of eliminating this with the other equally small errors of construction and application.

Hence clinical thermometers should always be "seasoned.” That is, after being made so far as heating processes are necessary, they should be put away for three years,-or at least for two years and a half, before they are graduated, because after three years, no change of practical importance will occur.

It is probably only the makers of cheap instruments who graduate them as soon as made, and such thermometers being untrustworthy from other causes, are most so from this cause also. It may perhaps be stated,-though rather at random, as they do not tell the details of their business,-that makers who have moderate reputations to keep up, do not as a rule graduate their tubes until they have been seasoned for a year or a year and a half. But as only about one-half of the total contraction will have occurred in a year and a half, such thermometers in three years will be found to read about five-tenths of a degree too high. Hence it occurs that from this contraction alone thermometers which have remained unsold. in the hands of dealers, or have remained unbroken in the hands of physicians for two years or more, are often half a degree out of the way, and always too high, while in cheap thermometers it is not uncommon to find them a degree or more out of the way.

A majority of observers will be well served if they get thermometers which are accurate to within two-tenths of a degree. But fairly close observers should not be satisfied with an error of more than one-tenth, unless the error be accurately known, and be applied as a correction to each observation made with the instrument. Of course an otherwise well-made instrument which is old enough to have reached its maximum contraction,-or say over three years old,—if its error be accurately known, and always applied, is as good as if it had no error, and the common practice of all close observers should be to get an instrument of known age and error,

and keep it carefully, -not for daily use, but to compare other thermometers by, and then buy cheaper thermometers for daily use and accidental breakage. Every clinical thermometer bought by a physician should be accompanied by a certificate stating when and where it was compared. For many years, and until 1880, the only thermometers having these certificates were compared at the Royal Observatory at Kew Gardens, London, and the certificates of comparison with their standards were supplied by the Kew Observatory at a shilling each. These certificates gave a very great advantage to the English clinical thermometers with which they were sold; but the great demand for these thermometers caused the English makers to send them out unseasoned, though certified, and thus the certificate lost much of its utility, and often accidentally came to be misleading. About 1880, however, the Winchester Observatory of Yale College, at New Haven, Conn., established a thermometric bureau, and placed it under Dr. Leonard Waldo, the astronomer in charge, and now this bureau takes the place, for this country, of the Kew Observatory in England. Any physician can now send his thermometer to the Yale College Observatory and for fifty cents obtain an accurate certificate of its error throughout the scale; and should his thermometer be old enough when sent, the certificate will be good for the entire lifetime of the instrument. It should be a source of great satisfaction to all who use thermometers for any purpose that this well-known Observatory has assumed this important work, and now it will be the fault of observers if they have not corrected instruments. If the results realized in Great Britain from the Kew Observatory should be attained here from the Yale Observatory, it will not be many years before the grossly inaccurate instruments scattered all over this country, for meteorological as well as for medical uses, will be very much diminished in number. At least, if they are not, it will be the blamable fault of the observers who use them. The Yale Observatory also undertakes, for a very small consideration, to seal up packages of thermometers for makers, so that it may issue with them certificates of the age or seasoning of each thermometer.

In selecting a clinical thermometer several important points are entirely within reach of the physician who selects.

First, the index or register being shaken well down, the bulb is taken between the thumb and finger, and the ascending column of mercury carefully observed. If it rises very slowly the glass of the bulb is too thick, or the tube too large. If it rises very rapidly

indeed, the glass of the bulb is too thin and the thermometer will be easily broken. This defect from thinness of glass is, however, comparatively rare, while those in which it is too thick are very common. When an instrument is found which does not rise very slowly, let it be cleansed, and then placed under the tongue, for about eight or nine minutes by the watch. Then let the temperaature be read, and the bulb be at once returned to its position under the tongue for ten minutes more and again read. If the temperature has increased since the first reading, the instrument should be rejected, because a thermometer which does not reach its maximum indication in nine minutes is unfit for clinical uses. It will tire out the patient, and use up a great deal of the physician's time unnecessarily. If the reading has not increased, allow it to cool for ten minutes and again place the bulb under the tongue for six minutes and again read. If now the reading is lower than it was before, then the time required to reach its maximum indication is longer than six minutes but shorter than nine, and it must be again cooled, and again placed under the tongue for seven or for eight minutes according to the difference between the former readings; and in this way a fourth or a fifth trial will determine the point of maximum indication. But if the six minutes reading corresponds within onetenth of a degree with the nine minutes reading, then it is to be again cooled and placed under the tongue for four minutes,-and so on until the time for reaching its maximum indication is established. Then in all his observations throughout the lifetime of this instrument the physician must always keep it in place for this length of time in order to get a trustworthy observation. The average of good thermometers may be given as from six to eight minutes. Those of three, four and five minutes are exceptional and rare, and are very valuable when found, if well taken care of, for they save a great deal of time. But while an eight minute thermometer may fall on the carpet either in or out of its case and only exceptionally be broken, a similar accident will be almost surely fatal to one of four minutes.

The next point in the selection is to see that the register shakes down easy enough without moving too easily. This will depend a good deal upon the method used by each individual. If the method used by thermometer makers be used, it will with much ease get any register down, and that without much risk of losing it, but some skill and practice is needed in acquiring this method, and it is not susceptible of being clearly explained. They hold the thermom

eter rather loosely by the extreme upper end and then give it a quick flirt or sling outward from them somewhat as if cracking a whip, the arm representing the whip-handle and the thermometer the whip-lash. The methods in almost universal use are all much inferior to this, and are too well known to need description, and the observer must select his instrument to accord with his own method if he does not choose to acquire a better one. If the register moves down too easily it will in all thermometers, except one variety, be liable at any time to be lost by causing it to coalesce with the body of mercury below it; and if lost, the thermometer, as a rule, is hopelessly spoiled. The register should be separated from the mercurial column below it by a distance almost inappreciable to the naked eye when the instrument is at the lower end of its scale. The thermometer with "indestructible index" is made on the principle of Phillips' maximum thermometer, namely, by a narrow contraction near the bulb. But in this case the contraction must be so much narrower than in the Phillips instrument that the mercury passes through it by jumps, and of course the whole column rises by jumps so that it does not indicate with accuracy to tenths of a Fahrenheit degree. These thermometers differ somewhat in the length of their jumps, and probably the makers will improve upon them in future, but at present they are not adapted to very close observations, though perhaps accurate enough for ordinary indiscriminate use. It may be safely said, however, in regard to the ordinary clinical thermometer that a register is never lost except by mismanagement, and such mismanagement is much more frequent with thermometers which shake down easily. To this it may be added that physicians, as a rule, are dissatisfied with instruments which do not shake down easily, and this is the reason for so many lost registers. Registers are often, if not generally, lost at the first or second shaking down, by supposing they are hard to shake down when they are not, or by using some bad method of shaking. Hence a new thermometer should be shaken down with great care until the observer knows how much force is required.

Almost all clinical thermometers are graduated to fifths of a degree, but are easily read to tenths, the only difficulty being with the shorter ones, where the marks for fifths are so close that the distortion by thickness of glass on the stem makes it difficult to hold the instrument exactly at a right angle with the line of accurate vision. Any one of these short thermometers can be easily read one or twotenths out of the way on either side of its true indication by this