quently there are but few places where we have the requisite data. The former of these instruments it has been found difficult to construct so as to work satisfactorily.

My first effort as to the problem was made in the year 1837, at Ogdensburg, N. Y., where I erected a self-registering vane, and made arrangements for frequent observations of the barometer. The definiteness of the result surprised me. It divided the horizon into two perfectly distinct portions, the winds from one of which were attended with an average rise of the barometer, and those from the other with a fall. And although my vane registered from 32 points of the compass, there was no intermingling of the points in the result. But was the law that I had thus discovered, a general one? Or, was it owing to something peculiar to that locality? To test this, I proceeded to make similar computations for twelve other places in this country and elsewhere, according as the observations to which I had access furnished data applicable to the purpose, and while so employed I found that Prof. Dové, of Berlin, had done the same for five places more, which I united with my own, making eighteen in all. eighteen in all. [These are delineated in Plate 23.]

"Early in these investigations the question arose whether the results favored the rotary or centripetal theory of storms; the indications were that the motion was both rotary and centripetal. I was not then fully prepared to submit what I had offered for publication, except in outline, and I deferred to do so. Since 1853 I have added nothing to it, except the results of Dr. Louis Berlandier's observations at Matamoras in Mexico. The following gives in a tabular form the results of all the observations since that date:

TABLE I.

SHOWING THE AVERAGE RATE OF RISE OR FALL of the Barometer, in DECIMALS OF An Inch, per TWENTY-FOUR HOURS DURING WINDS FROM DIFFERENT POINTS OF COMPASS.

66

Regarding the rate of rise or fall in the barometer during winds from each point of compass, given in the preceding table, as the measure of the force that produces it, and reducing these forces to a single force, in the usual way, we obtain the results in the second, third, and fourth columns of the following table; to which I have added, in the fifth column, the mean direction of the wind.1 The arrows within the inner circle of the Barometrical Wind-roses [Plate 23] exhibit these results to the eye.

"The results shown in the foregoing tables and diagrams confirm all that I had previously adduced, and establish conclusively, I think, the following facts, at least in the zones of westerly winds.

"1st. That the horizon is divided by nature into two well-defined portions, the winds from between the division points on the one side being all attended with a rise in the barometer, and on the other with a fall. This is found true at all the stations where there are reliable observations. Even where they are taken for thirty-two points of the compass, there is no intermingling.

"2d. That in the northern hemisphere, one of these points lies in a southwesterly direction, and the other in a northeasterly. Barnoule in Siberia, and Sitka in Alaska, look like exceptions; but at both these places the results were computed

The observations at sea were taken in various latitudes, and those on the direction of the wind not reported; so that it was impossible to know accurately what mean direction to assign. But taking into account the circumstances of the voyages during which they were taken, I have assumed, as approximately correct for the southern hemisphere, one that I computed from a zone on Lieut. Maury's charts, extending from lat. 40° to 45° S., and from long. 20° E. to 120° W.; and for the North Atlantic, one deduced from about twelve years' observations, taken north of lat. 36°

for a short time only,' and might be somewhat modified by making use of a longer series of observations. It is probable, moreover, as I have shown elsewhere, that Sitka lies without the zone of westerly winds, and where a different law may prevail.

3

"3d. That the line of its approach generally makes an angle, more or less acute, with one drawn to the point of maximum pressure. The only exception is at Hamilton, Bermuda, where it is slightly obtuse (92° 40′). Nor is the result different, if, instead of regarding the mean resultant of all the forces which raise the barometer as the point of maximum pressure, we (perhaps more properly) regard each fall as a negative rise, and vice versa, and then obtain one mean resultant for the whole. The fourth column in Table II. was computed in this way, and the results are shown on the Barometrical Wind-roses [Plate 23] by a broken line. [For application of this discussion to the storm-curve, see author's article on pp. 89-101, Proceedings of the American Association for the Advancement of Science, Cleveland, Ohio, 1853.]

"The plan of the 'Winds of the Globe' contemplated giving resultants at each place, for each month and season, with monsoon influences for the seasons. The work would be much more perfect, if this could be done in all cases, but the magnitude of the labor forbade it. For a like reason, as well as to render it possible to represent the results on maps, it was thought judicious to group the places of observation by districts, where they were numerous, instead of making computations for each place separately. With the facilities we have devised, in the way of special tables to aid in the computations, we have found that where observations, recorded for 16 points of the compass, have been collected and properly arranged for computation, an active computer can calculate about 35 resultants in a day. When the observations are recorded for 32 or more points of compass, the labor is of course much greater, but there are comparatively few such. On the other hand, there are many where they are recorded for only 8 points. If we include the calculation of the monsoon influences, which has been done chiefly by plotting, the average per day will not exceed the number just named.”

[The exact state of forwardness of the work at the time of Professor Coffin's decease is fully related in the preface.]

1 One month at Sitka, and two at Barnoule.

Further on, in the same article from which these conclusions are quoted, and which may be found on page 89 of the Proceedings of the American Association for the Advancement of Science, 1853; Prof. Coffin determines this angle as 65°; and a reference to the article plainly shows that this determination was reached, without any knowledge by him of its having been accomplished, or even attempted, at that time, by any other writer on the subject, although the reference on page 664, of this work, conveys the intimation that this principle is generally referred to in Europe as "BUYS-BALLOT'S LAW OF THE WINDS". But it does not there appear at what date Prof. Ballot had made the announcement, with which he is so accredited.

It is worthy of remark that here, too, the angle is acute, if, instead of the mean direction of the wind observed at Hamilton, we employ that at Ireland Isle, another island in the same group, or even the mean between the two.

D December, 1875.