On the roof of the Octagon Room we find, beside the two turrets already referred to, a small wooden cabin built on a platform several feet above the roof level. This cabin and the north-western turret contain the wind-registering instruments. Opening the turret door, we find ourselves in a tiny room which is nearly filled by a small table. Upon this table lies a graduated sheet of paper in a metal frame, and as we look at it, we see that a clock set up close to the table is slowly drawing the paper across it. Three little pencils rest lightly on the face of the paper at different points. One of these, and usually the most restless, is connected with a spindle which comes down into the turret from the roof, and which is, in fact, the spindle of the wind vane. The gearing is so contrived that the motion on a pivot of the vane is turned into motion in a straight line at right angles to the direction in which the paper is drawn by the clock. A second pencil is connected with the wind-pressure anemometer. The third pencil indicates the amount of rain that has fallen since the last setting, the pencil being moved by a float in the receiver of the rain gauge. [Illustration: THE ANEMOMETER TRACE.] An objection to all the mechanical methods of continuous registration is that, however carefully the gearing between the instrument itself and the pencil is contrived, however lightly the pencil moves over the paper, yet some friction enters in and affects the record: this is of no great moment in wind registration, when we are dealing with so powerful an agent as the wind, but it becomes a serious matter when the barometer is considered, since its variations require to be registered with the greatest minuteness. When photography, therefore, was invented, meteorologists were very prompt to take advantage of this new ally. A beam of light passing over the head of the column of mercury in a thermometer or barometer could easily be made to fall upon a drum revolving once in the twenty-four hours, and covered with a sheet of photographic paper. In this case, when the sensitive paper is developed, we find its upper half blackened, the lower edge of the blackened part showing an irregular curve according as the mercury in the thermometer or barometer rose or fell, and admitted less or more light through the space above it. Here we have a very perfect means of registration: the passage of the light exercises no friction or check on the free motion of the mercury in the tube, or on the turning of the cylinder covered by the sensitive paper, whilst it is easy to obtain a time scale on the register by cutting off the light for an instant--say at each hour. In this way the wet and dry bulb thermometers in the great shed make their registers.