Samples of government-supplied marijuana from NIDA were puffed in a smoking machine in a manner designed to mimic human marijuana smoking. The smoke was collected in Cambridge glass fiber filters designed to capture particles > 0.1 microns, which are used to separate solid particles or 'tars' from gaseous smoke components, such as carbon monoxide. The filtered solids also include all of the cannabinoids. The filtered residue was weighed to measure total tar content and quantitatively analyzed for three cannabinoids, THC, CBD and CBD, by means of a GCMS.
As expected, all of the devices produced a reduction in tars relative to the control: 33% for the filter, 89%-98% for the waterpipes, and 56%-97% for the vaporizers. However, only the vaporizers achieved an improvement in the ratio of tars to cannabinoids. The cigarette filter performed worse than the unfiltered joint, producing 30% more tar per cannabinoids...
...A likely explanation for the poor performance of physical filtration systems is that THC molecules are especially sticky and apt to adhere to other smoke particles. Any attempt to screen out the latter is therefore apt to pick up the former as well. Indeed, to the extent that cannabinoids are relatively stickier than other compounds, particles containing them may be more likely to be trapped by filters.