cumstance, the mean temperature so deduced would always be below the truth. It was also inferred that the readings of the dry bulb would not be at all influenced by the neighbourhood of the wet bulb, the evaporation from which would also have the effect of causing the dry-bulb readings to be slightly below the truth. It was further inferred, that the rise and fall of temperature was uniform between each two-hourly period at which the readings were taken; so that, when tables were drawn up to suit Greenwich time, the corrections to be applied were calculated on the principle of uniform rise and fall, which we all know is far from being the case. It was also inferred, that a period of five years would suffice to deduce a satisfactory mean basis on which all the calculations should be founded,-a fallacy which strikes at the root of the whole superstructure.

Now, any one reflecting on these circumstances will at once see that serious errors must arise from constructing tables of correction on such imperfect data-Firstly, Because a five years' series of observations is far too short a period to give a satisfactory mean result for either daily, monthly, or annual mean temperature; and, secondly, Because the highest temperature of the day not being noted at all by these two-hourly readings, we not only obtain a mean result below the truth, but we have no data by which to calculate the period of the day when the highest temperature is attained; and hence the 3 o'clock readings are rendered perfectly useless for the purpose of estimating mean temperature, when three readings only are taken in the course of the day, of which that at 3 o'clock is always one.

First, then, it may be demonstrated that a series of observations for a period of five years only could never elicit a true mean, either for days, months, or years; and yet, as the basis of all tables of correction must be founded on a fixed mean, it is absolutely necessary that that basis be extended over a much longer period of years before any reliance could be placed on its deductions. Let us look for a moment at this point of the subject.

Scotland has only had the advantage of having its meteoroNEW SERIES.-VOL. XI. NO. II.-APRIL 1860.

2 E

logical data collected and published since 1855. Let us see how the last four years stand with regard to the one element of mean temperature :

It is seen at once from this table, that during this period of four years there has not been the most distant approach to uniformity of even mean annual temperature in Scotland; but, on the other hand, there is the striking fact that during that period the mean temperature of one of the years differed from that of another by nearly 4 degrees. Now it would be a most unsafe conclusion to deduce from this table that the mean temperature of Scotland was 46°1. The period of time is evidently far too short to afford even an approximation to the truth.

If we look at the mean temperature of a few of the months during these several years, we shall find a still greater diversity, as is shown in the following table:

By this table it is seen that during a period of five years the range of mean temperature during the months has been so varied in different years as quite to preclude the idea of a five years' average giving even the most distant approach to a true mean. Thus, in July and September, the range of mean temperature during five years was 4 degrees; in June, 5.6°;

in March, 7°; in May, 7°9; while in February it was no less than 12° 6.

Secondly, It may next be shown that the two-hourly periods of observation could not give the true mean temperature, as they did not include the highest temperature of the day, which was from two to six degrees higher than any of the two-hourly readings, during the warmer periods of the year. In proof of this position is subjoined one of the Greenwich Summary Tables for the year 1844, one of the years on which Mr Glaisher's Tables of Correction are founded; and this table speaks a language which one would think no one can misinterpret.

This table shows the highest and the lowest degrees of temperature at Greenwich during every month of the year 1844, both as noted at the two-hourly readings of the dry-bulb thermometer, and also as indicated by the self-registering thermometers; and a comparison of the highest readings of the drybulb, with the highest readings of the self-registering thermometer, and likewise the comparison of the lowest readings of the dry-bulb with the lowest readings of the self-registering thermometer, fully bear out the position which has been stated. Thus, every meteorologist knows, that while the curve of temperature is very great during the day, the curve of temperature during the night is so much less that for several hours of the night the temperature varies but little. This table, then, exhibits also this fact; for while it shows that the two-hourly readings approached the lowest temperature during the night within a few tenths of a degree, on the other hand, these same two-hourly readings never, during any month, reached the highest temperature, the very mean of the twelve months showing that the highest readings of the dry-bulb thermometer at the two-hourly periods were 2°-4 below the actual highest temperature which occurred, as indicated by the self-registering thermometer.

This table, then, at once proves, what was stated at the outset, that readings at two hourly periods must fail to give the true mean temperature, and must indicate a mean temperature below the truth. But what is true of the months is also true of the days, and this, of course, to a greater extent than the months themselves. Let us take an example from the Greenwich observations of the highest temperatures by each kind of instrument on consecutive days of the same month of 1844. On the 2d of September, the highest of the two-hourly readings of the dry-bulb thermometer was 73°1, but the actual highest degree of temperature which occurred that day was 78°, as registered by the self-registering thermometer. On the 3d September, the highest two-hourly reading was 68°7, but the actual highest degree of temperature that day was 71°4. On the 4th September, the highest two-hourly reading was 72°-8, but the actual highest was 73°7. On the 5th September, the highest two-hourly reading was 69°1, but the actual highest was 71°-3. On the 6th September, the highest two-hourly reading was 72°-2, but the actual highest was 73°8. On the 7th September, the highest two-hourly reading was 72°5, but the actual highest degree of temperature was 74°.

These, then, may

serve as examples of the statement made, and at once account for the mean of the two-hourly readings being below the mean of the self-registering thermometers, and of course below the truth.

Had Mr Glaisher, however, confined his corrections to the dry-bulb readings alone, no great harm might have resulted, provided these corrections were applied to their legitimate use -viz., to enable a person from one reading of his thermometer daily to deduce the probable mean temperature of his locality for each month. But Mr Glaisher has gone far beyond this; for he not only alters by these tables every observation made with the dry-bulb thermometer, so as to let the mean of no series of observations be published as observed, but, from finding that the mean of his two-hourly readings was always below the mean of the observations made with the self-registering thermometers, he alters their means also, by deducting from the mean of each month a quantity which is intended to reduce their mean to the mean value of his two-hourly readings. He thus commits two great errors-first, in assuming that the two hourly readings of the dry bulb, taken in the faulty circumstances above noticed, give the only true mean temperature; and, secondly, that the mean of the self-regising observations must be erroneous. He hence asserts that the mean temperatures procured by taking the strict mean of the maximum and minimum readings of the self-registering thermometers are too high. I shall however prove, I trust to your perfect satisfaction, that the strict mean of the maximum and minimum self-registering thermometers (provided these instruments be of proper construction) is far nearer the true mean when unaltered, than when altered by Mr Glaisher's tables.

Now, to prove this point, I am prevented from referring to the Greenwich observations, first, because the original observations, from which Mr Glaisher drew his conclusions, were made with that most untrustworthy instrument, Six's registering thermometer, so that I cannot refer to the observations made with it; and, secondly, because when, at Greenwich, instruments of proper construction were procured, Mr Glaisher has published no results which could permit any one to test