The rule of thumb that a thermometer measures the 'number of times molecules bump into it (per unit of time)' is borne out in practice (or the number of times and how intense those bumps are, but in this instance the intensity is constant and the number clearly falls).
So unsurprisingly, when he turns on the vacuum pump, the thermometer shows a lower temperature, and when he opens the valve to let air back in, the thermometer shows room temperature again (plus a bit for friction).
Sucking out some molecules doesn't actually reduce the average kinetic energy of the remaining ones, it reduces their number, so as a generalisation, we would expect there to be less 'heat' inside the chamber, however defined.
Debate Magazine
