It seems to be easy to get it wrong. Even the author of the article, who studies giftedness professionally, made such mistake (she tells that IQ 190 to IQ 130 is like IQ 130 to IQ 70).
What's that supposed to mean? The meaning of IQ 190 is that it is 6 standard deviations from the mean (rarity around 1 in billion), while both IQ 130 and 70 are 2 standard deviations from the mean (rarity around 1 in 50).
PS: It's not really completely accurate, but you can treat a 6 year old and a 12 year old with the same mental age in about the same way and see similar types of responses.
http://www.amazon.com/Measurement-Psychology-Critical-Histor...
by Joel Michell. IQ scores definitely do not, as bd correctly points out, constitute an interval scale. They are not measurements.
To respond to your original question, I have a hunch that intelligence variance (as opposed to IQ) is indeed quantitative.
It kind of make sense. Intelligence seems to correlate positively with simple stuff, like reaction speed or short term memory span.
Qualitative differences may arise as consequence of applying different "processing power" throughout your life.
For the first order effects, you can imagine that if you don't have enough "CPU-minutes", for some problems you simply time out.
For the higher order effects, it's like compound interest. For example, as written in the article, smarter kids start to speak and read sooner, thus they acquire more information and they also acquire it at higher rate. And as we know from financial compound interests, even tiny initial amounts can balloon into huge sums given enough time.
At any rate, is there quantitative metric for intelligence?
Simplified view: people solve a test with a number of questions. They are getting ordered by how many questions they get right.
Tests are constructed to produce Gaussian distribution of numbers of correct answers. Some questions are easy, almost everybody can solve them. Some questions are difficult, almost nobody can solve them. There should be progressively less and less people getting more questions right.
But in reality, it's not a perfect Gaussian distribution. There are some numbers of questions that have more people getting them right than would be predicted by normal distribution.
Now, about quantitativeness. We could get somehow quantitative metric, if questions would simple and uniform in structure, for example test composed just of "how fast you can multiply x-digit numbers" task, or test solely composed of "how fast you can arrange pieces into a particular shape" task.
But it's not like this. There are different tests, some of which have several thematically different sections. Even inside one section, questions do differ a lot (for example, some shapes are harder to compose of primitive elements, how would you quantify this difficulty?).
To further complicate matters, usually completely different tests are used for different intelligence ranges.
And we are not even speaking about normalizations. Your rank is computed just for your age group. For the same raw score, you get extra points if you are younger or older than the optimal age.
If I remember well, there is a fast ramp up of raw scores till 18 years, followed by a slow decline afterwards (rate of decline is slower for more intelligent people).
And then there are national/race differences. And Flynn effect. It's much more messy than it looks.
A reasonable analogy might be replacing a 200hp engine with a 300hp engine in the same car. For a wide range of day to day things they only change is their acceleration. However, at higher speeds increased drag comes into play. So the harder the task the longer it takes them to get there. Untill they just can't keep up. For some things there is little noticable difference but othertimes it's striking.
A good example might be an Algebra II homework mistake. We where doing 2 equations 2 unknowns in class which was simple and the homework was the end of chapter quiz. Well I sat down and the first 4 questions where like that but the next set had 3 equations and 3 unknowns and a few seconds later I figured out how to do that. It's seemed odd that the homework had something new but it was easy enough so moving on. The next day I realized I had not been paying attention and did the next chapters assignment. It took the teacher two weeks to get to that point. I remember thinking what are most of these people just dumb?
A more extream example: A few years before that my had an agument with my 3 year older sister and ended up doing some of her math homework. I do recall her say ok yea this is easy. She skipped 8th grade and was the salutatorian and took AP calculus so it's not like so was slow behind or anything it was just obvious based on the questions.
As to the statistics I don't think think there is much point in measuring IQ's over 150 or so. It's not that people don't become smarter, but it's harder to find how generally intelligent someone is. Plenty of people can solve an Rubik's Cube in under 45 seconds, but how long would it take them the first time they saw one? There are plenty of mental leaps that are easy after you have seen them, but they take can take a long time to show up and 3 people might each solve a different one first. According to several people I was 3 before I spoke my first word, but my first words where "Please pass the butter" so how do you measure that?
Also, do you meet more people than you would statistically expect who are around your level? Another interesting claim the article makes is some researchers say there are more exceptionally gifted people than their statistical distribution would predict. Since the IQ metric is based on assuming a normal distribution in the population, such a result would mean IQ scores are not accurate.