Many experiments have already been made to find out how much weight bodies gain or lose when they are heated.  These experiments are so very delicate and are susceptible to so many errors--not only because of the imperfections of the instruments used, but also because of unexpected disturbances like currents in the surrounding air--it is not at all surprising that there has been so much disagreement on the results.

It's been a long time since I first began to think about this subject, and I have made many of my own experiments, trying hard to avoid as many sources of error as possible.  But though all my research has convinced me more and more that a body acquires no additional weight upon being heated, or, rather, that heat has no effect whatever upon the weights of bodies, I am aware of the delicacy of the investigation, and so I was, for a long time, afraid to form a decided opinion on the subject.

I was surprised to read about Dr. Fordyce's experiments in which he measured a gain in the weight of water after it was frozen.  So, as soon as it became cold enough in the winter of 1787, I set about to repeat his experiments in order to see the results for myself.

I selected two glass bottles which appeared to be so much alike as to be indistinguishable from each other.

I put equal amounts of liquid in each bottle.  In one bottle, which I shall call A, I put pure distilled water.  Into the other, B, I put an equal amount of weak spirit of wine.  I sealed both bottles, washed them, and wiped them perfectly clean and dry on the outside.

Then I put each bottle on one arm of a balance that I had placed in a large room.  For several weeks, this room had been regularly heated every day to a temperature of 61°F by a German stove.  I left the filled bottles in this room until I felt they had reached the temperature of the surrounding air.  I wiped them again with a very clean, dry handkerchief, and then adjusted the balance until the two were precisely in equilibrium.

I left the whole apparatus sit like this for twelve more hours and then checked the balance again.  Finding them still in perfect equilibrium, I now moved everything into a large uninhabited room, facing north, in which the air, which was very quiet, was at the temperature of 29°F.  I left the room, locked the door behind me, and left the bottles undisturbed for 48 hours.

At the end of that time, I entered the room, using the utmost caution not to disturb the balance.  To my great surprise, I found that the bottle A was noticeably heavier!  The water in this bottle was completely frozen into one solid body of ice, but the spirit of wine, in bottle B, showed no signs of freezing.

I very cautiously measured the difference in the bottles' weight to see how much weight bottle A had gained.  It was a very small amount--less than .003% of its original weight.

I soon found out, however, how very dangerous it is to draw conclusions from a single experiment.

I returned the apparatus to the warm room.  The ice in bottle A gradually thawed, and turned to water again.  Weighing bottles A and B showed that they were once again exactly the same weight.

I repeated the entire experiment.  I put the apparatus back in the cold room and got nearly the same result as before--the water appeared to be heavier when frozen than when it was liquid!

But something seemed strange about this second trial.  First, the water had gained a different amount of weight when it froze.  Second, there was an irregularity in the way in which the water lost its additional weight as it melted.  I began to suspect that the experiment was not a reliable one.

I checked the accuracy of my balance--making sure that the changes in temperature were not affecting it, or that there wasn't some mechanical problem with it.  My confidence in the balance was confirmed.

I conducted a number of experiments using water and spirit of wine which convinced me further that my first experimental design was not reliable.

I then decided to conduct the experiment comparing a bottle of water with a bottle filled with an equal weight of mercury.  This time, there was no difference in the weights of the two bottles after the water was frozen solid.

From these experiences, I concluded that if heat is in fact a substance, it must be something so infinitely rare, even in its most condensed state, as to baffle all our attempts to discover its weight.  And if heat is nothing more than a vibrating motion of the small particles in bodies, it is clear that the weights of those bodies cannot be affected by such motion.

The experiments I did with water and ice did not involve anything special.  The amount of heat that water is known to part with when it freezes is very large, so that if this loss has no effect upon its apparent weight, it may be that we shall never be able to contrive an experiment by which we can measure it.