Many, many more details in Heller's book and in this Econtalk interview:
http://www.gridlockeconomy.com/
http://www.econtalk.org/archives/2009/11/heller_on_gridl.htm...
it's the NIH and universities (via government funding) that actually discover the drugs, then they sell the exclusive rights to manufacture it to big pharma (who then spends a lot to get it approved and market it and such...)
(this is the book, i haven't read it) http://www.amazon.com/Truth-About-Drug-Companies-Deceive/dp/...
so in this case (if true) patents fail again.
Since manufacturing, testing and marketing a drug is hysterically expensive, the patent seems to serve a purpose there -- even if pharma companies are more akin to specialized investment bankers than scientists.
If we take the familiar example of the ICE, simple dimensions like the bore and stroke of the cylinder aren't arbitrarily chosen. Altering the ratio of the bore and stroke changes the revving characteristics of the engine (a somewhat larger bore than stroke will cause the engine to have a higher rev limit, for example), and increasing the size of the bore requires disproportionately heavier cylinder heads (because cylinder heads have to have some depth and lateral reinforcement).
The size and weight of the cylinder heads, combined with the stroke effect the strength and dimensions required of the connecting rods and the crankshaft, so on and so forth.
Basically, a lot of mechanical "inventions" are big physics and engineering puzzles where the mere dimensions of each part was worked out through a lot of equation crunching (and probably a lot of plain old tinkering and testing as well). So for someone to just go and copy someone else's design, well that's kind of a big deal.
Metallurgy, semiconductors, superconductors, ...