What about this model: Say programming skill is more or less proportional to the time spent programming. Further say that at any moment a person has a certain probability of leaving the industry. In that case the skill curve over the population should be a geometric, hockeystick like curve. It wouldn't make sense that the majority of people had experience in the 5-8 years interval say.

It is still not likely to be bell shaped.

Of course there is a /distribution/ and of course it is not shaped like a vertical line -- but it is very unlikely to be a normal distribution.

Why?

In general one would assume that a distribution of this type would be normal absent evidence to the contrary right? What is that evidence to the contrary?

If we take robinrob's point, and measure only "people who have programmed", then I would still expect a very positively skewed distribution. There are really a lot of people in the "I tried it and didn't like it" category.

For almost any boundary you choose, a majority of people will sneak in just over that boundary, leaving a long one-sided tail.

This skewed distribution would appear even if you cut a Normal distribution.

A Normal distribution occurs when everyone is expected to be at the same average level but could move up and down equally due to many random factors. When learning skills, people generally start at a 0 level and generally move up. That won't be Normal.

If you take a cohort of programmers with exactly the same experiences, that could plausibly follow a Normal distribution of skill. If you take e.g. all professional programmers, that won't.

> A Normal distribution occurs when everyone is expected to be at the same average level but could move up and down equally due to many random factors. When learning skills, people generally start at a 0 level and generally move up. That won't be Normal.

Aha, yes that's an excellent point.

If you chop off the left side of a bell curve, is the remaining part suddenly bell shaped?