I don't know about this. However, one thing that I have no doubt about is that levels are reflective of character ability. It's certainly possible to create a useless 12th level character or an amazing 3rd level character, but the vast majority of characters at level X will be able to deal with roughly the same level of challenge. Your 12th level fighter may be a little bit weak, but unless you intentionally screwed him up or fell for some kind of power-gamer trap (if I take one level of wizard, I can cast shield, if I take one level of cleric, I can cast cure spells and get a re-roll. If I take on level of druid, I'll be faster; oops, now I'm 6th level with a +3 base attack bonus and a bunch of equally useless spells--darn), he'll be able to hang with the other 12th level characters. Your 7th level wizard may be a little strong for his level, but except for a few exploits, nobody is going to mistake you for a 9th level wizard.
Classes play a big part in this by establishing a set progression of combat abilities that you get whether you like it or not as long as you don't multiclass.
This enables a module writer to know roughly what kinds of characters a module is appropriate for and write it down in an easy succinct sentence--"designed for a party of 6th level characters." It also enables a DM to quickly gauge the level of his party "Level 4+level 5+level 5+level 3=average level 4.25" and compare what they are capable of to what a published module might require. If I were to try this with a true point based system that didn't include the automatic upgrade paths that D&D classes do, the easiest method would require the DM to keep track of how many points he had given out (I think I've seen champions games advertise "600 point characters" and such), but to get the same degree of very approximate accuracy, I would probably need to create a separate chart based on character abilities (a Delta class hero can reliably hit AC X; average save Y; Z hit points) to approximate level and calculate what class the PCs are in.
So, how did non-level based games successfully publish modules? The impression that I get from my shadowrun experience and what I've seen written is that they tend to have a much less dynamic power curve. You can gain power, but my impression is that the same system does not generally support the grim mercenary company struggling to survive in a gritty campaign against the orcish hordes and the paragons of martial virtue who ride to slay the great wyrm of the north. In D&D, you could start as the first level mercenary and become the dragonslayer--in fact, that is the normal progression. My impression is that your power level is far more static in non-level based gaming systems (perhaps because the more flexible a character creation system is the more it is generally possible to power game it).
howandwhy99 said:
2. Back in college a friend told me levels were more reflective of actual ability. I'm not sure all will agree, but I do see how practice can go from uneffective to degrees of effectiveness. For example, a gymnast learns certain tricks until they become competent to perform them in competition. Further proficiency in more difficult tricks shows greater ability. Partial learning does not count as a partial flip dismount still results in landing prone. Only true proficiency can be trusted (a new level).