In estimating the annual cost of lighting, the renewals of lamps must be taken into account; and although some lamps have worked 3000 or even 4000 hours, a life of 1000 working hours is the highest average it is safe to assume in practical work under even the best conditions, that is, using secondary batteries and never over running. The average life of 130 lamps on H.M.S. troopship _Malabar_ was 3799 hours each, the shortest life being 638½ hours for 18 yard-arm lamps of 32 candle-power. If the current is allowed to fluctuate, the average life would be very much less; it is an unsettled question whether long-lived lamps are really economical, by reason of the blackening of the globes, which takes place after the lamp has been worked some time, and is probably due to small particles of carbon thrown off from the filament being deposited on the glass. It has been suggested that attrition of the filament is going on all the time the lamp is at work, and that the heated atoms striking against the filament may account for the blackening, in that the mean free path of the atoms would be greater in a perfect vacuum than in the air, consequently they would abrade the filament with considerable force. If lamps were sold at 1_s._ each instead of 3_s._ 6_d._, which is now the price for not less than a thousand, it would be more economical to change them at the first signs of blackening, even if the life did not exceed 500 hours. The diagram, Fig. 6, has been so arranged that the amount of light required in a given district can be ascertained for any period of the day or night; it has been calculated from the observations taken daily at one of the Berlin central-stations by the engineer to the company. Six hundred and forty watts are assumed, for the purposes of the diagram, to be the equivalent of a horse-power, instead of 736, as the German electrical horse-power is 736 watts instead of 746 watts. [Illustration: FIG. 6.] The table, Fig. 6, has two vertical scales, A and B, each giving the kilowatts[3] and corresponding horse-power. A is drawn to a scale ten times greater than B, with the object of noting the smaller amount of lights required for street illumination. The horizontal line is divided into hours, and represents a day’s lighting in the middle of December and the end of July, so as to show the maximum and minimum amount of current that will be required. In the lighting of a town there are two classes of illumination, the amount taken by the public, which is uncertain, and that employed for street lighting, which is a known quantity. [3] 1000 watts. The curves, II and II A, represent the private lighting of houses, hotels, theatres, and shops of different kinds in December and in July, the curve, II A, being in dotted lines clearly shows what a vast difference there is in the amount of light, and consequently the amount of energy required in the generating station, as compared with curve II, which is taken when the days are longest.