80 TransactioJis of the Boyal Society of South Africa. 
takes place as a rule about once in every two days, but occasionally the 
flooding of the Fish Eiver cuts off communication for four or five days. 
I have come to the conclusion that as a rule the railway time is known on 
the farm within 5 minutes, but that occasionally the error may have 
amounted to 20 minutes. Such occasional errors will, of course, be little 
appreciable on a record whose scale is such that 1 hour is represented by 
a length of curve amounting to inch. 
Even after all efforts had been made to reduce the friction, the records 
gave evidence that much more friction remained than in the case of the 
1905 machine. The upward and downward movement of the recording 
pen was often accomplished in a series of small jerks of about inch 
each, and occasionally of as much as J inch. The machine was not sen- 
sitive to the larger gas- bubbles, and the general aspect of the line traced 
resembled the line traced by an ordinary barograph. There was little 
appearance of seich frill, and no hair lines due to impact of gas-bubbles on 
the float. Moreover, the machine did not seem to be capable of recording 
much of the semi-diurnal fluctuations when the amplitude was small — i.e., 
about the times of moon's first and last quarters. The semi-diurnal fluc- 
tuations were clearly apparent, however, at all seasons of new and full 
moon, and thus one of the principal objects of taking the long record 
was successfully attained, as will be seen on Diagram No. 7, which 
is a photographic reduction of a tracing of all the records of these 
fifteen weeks. 
The fifteen record sheets were fitted together in order, and carefully 
traced by me as a continuous record. The junctions of the original week 
records are indicated by vertical dotted lines. 
The observed data from which the clock rate for each week was 
calculated is shown in Schedule VI. Each record was divided up into 
24-hour periods in accordance with the estimated average clock rate for 
its week, and the curved ordinates drawn to indicate the position of 
9 a.m. for each day. The clock rate was always very near 12 inches 
per week. 
As the recording lever had been divided by the fulcrum pivot in the 
proportion of 7 : 11, the amplitudes of the fluctuations on the record were 
less than the actual water fluctuations. 
Theoretically if the machine had been free of friction and backlash, 
and if the float had been perfectly sensitive to small variations of water- 
level, the mechanical reduction of the amplitudes would have left the 
recorded amplitudes in a scale of nature. As a matter of fact the 
imperfections of the machine conspired to reduce the amplitude scale still 
further to an unknown extent. 
The barograph records were obtained on a time scale of about 
II inches per week, but this scale is sufficiently near the time scale of 
