The nineteenth century produced hundreds of devices for alarming and extinguishing automatically. All depended upon the principle of the expansion or melting of metal in the increased temperature arising from a fire. At one time the circuit-closing thermometer was popular on account of its simplicity. "Its drawback," says Mr. Oatway, "is the smallness of its heat-collecting surface, its isolation, and, last and worst of all, its fixity of operation. In thermometer or fuse-alarm practice it is usual to place the detectors at intervals of about ten feet or so, so that a room of any size will contain a number. If a fire breaks out, the ceiling is blanketed with heat, and every detector feels its influence. Each is affected, but none can give the alarm until some one of the number absolutely reaches the set point or melts out. Having no means of varying the composition of the solder or shifting the wire, an actuating point must be selected which is high enough to give a good working margin over the maximum industrial or seasonal heat of the year; and thus it comes about that if the fire breaks out in winter, or when the room is at its lowest temperature, the amount of loss is considerably and quite unnecessarily increased. In a device set to fuse at 150° Fahrenheit, it will be clear to every one that the measure of the damage will depend upon the normal temperature of the room at the time of the outbreak. If the mercury is in the nineties, there is only some sixty degrees of a rise to wait for; whilst if it happens to be a winter's night, the alarm is held back for a rise of perhaps 120°. What chance is there in this case for a good stop?"
Mr. Oatway has examined the fuses under different conditions, and his conclusions are drawn from practical tests. Great intelligence will not be required to appreciate the force of his arguments. Inasmuch as the rise of temperature caused by a fire is relative, during the early stages at least, to the general heat of the atmosphere, it becomes obvious that an automatic fire-alarm should be one which will keep parallel, as it were, with fluctuations of natural heat. Thus, if the "danger rise" be fixed at 100°, the alarm should be given on a cold night as certainly as at midday in summer. It was the failure of early patterns in this respect that led to their being discredited by the fire-brigade authorities.