The fuller passage is:
The malleability of intelligence
Intelligence has historically been conceptualized as a more or less fixed trait. This view perceives intelligence as something people are born with, and the function of development is to allow this genetic endowment to express itself. A number of investigators have taken the approach that intelligence is highly heritable, transmitted through the genes. Other investigators believe that intelligence is minimally heritable, if at all. Most authorities take an intermediate position.
Various methods are used to assess the heritability of intelligence. Notable among these is the study of identical twins reared apart. For a variety of reasons, identical twins are occasionally separated at or near birth. If the twins are raised apart, and if it is assumed that when twins are separated they are randomly distributed across environments (often a dubious assumption), then the twins would have in common all of their genes but none of their environment, except for chance environmental overlap. As a result, the correlation between their performance on intelligence tests can provide an estimate of the proportion of variation in test scores due to heredity.
Take 50 sets of twins, and give each person an I.Q. test:
1. Twin A: 130 Twin B:128
2. Twin A: 105 Twin B: 107
Can you see, if you extrapolate this trend (how close the I.Q. scores are for each pair), that they would have a high correlation coefficient - say .9
1. Twin A: 130 Twin B: 107
2. Twin A: 105 Twin B: 128
3. Twin A: 115 Twin B: 110
Here, you might find a correlation of .2, which when tested for significance, might not be greater zero - that is, not significant, and it cannot be concluded that heredity is a factor in determining intelligence.
.The correlation coefficient might be .2,.6, or .9. The higher the correlation coefficient, the more intelligence must be due to heredity, since they have dissimilar environmental upbringing.
Conversely, the lower the correlation, the less is this due to heredity and the more due to other factors.
In this way, the correlation coefficient takes the scores, shows how similar or not they are (how they vary: the 'varialtion' in tests scores - do they have a statistically significant correlation coefficient, that is, greater than zero (if not significantly greater than zero, then the scores bear no relation to each other); and how great this variation is: the 'proportion of variation', somewhere between no correlation, 'weak', 'moderate', and 'strong'.)
Need further clarification?
(What's also interesting here, is why a correlation coefficient is the appropriate statistical test. Can you see what would happen if you did a parametric test comparing the means, or the standard deviations?)
Student or Learner