I think that it is generally accepted that diverse studies on intelligence or IQ inheritance among identical twin brothers, with identical genes, have shown correlations in intelligence of 80% or close to this figure.
For me, this argument regarding identical or monozygotic twin studies is definitive, because it would not make much sense if intelligence had a genetic component that was so powerful in some cases and in others just the opposite.
Other question would be why it does not appear always like that. I believe mendelian genetics is the cause of the apparent inconsistence.
An interesting aspect of monozygotic or identical twin studies on intelligence or IQ inheritance is that if the correlation of the twin's IQ with their respective parents were studied; the explained variance would probably be noticeably less than the mentioned percentage due to the genetic combination derived from Mendel's laws.
One of the authors that is most well known for his IQ studies with twin brothers and his articles in favor of genetic influence on intelligence quotients is Arthur Jensen.
In their book, The Bell Curve; Charles Murray and Richard J. Herrnstein present an intermediate analysis regarding the heritability of IQ and intelligence; it brings together articles about twin brothers, adopted and normal siblings and works with different, even contradictory conclusions, some of them with monozygotic, dizygotic or non-identical twins. Its ideas are categorized as sociology and the consequences on education, basically say that the genetic and environmental influence are fairly correlated and could generate pockets of populations with slower development.
From the measurements of intelligence carried out on siblings or dizygotic or non-identical twins, we can make two independent comments.
On the one hand, in statistical twin studies on intelligence or IQ inheritance, if the observed correlation in one case is 40% and, taking into account mendelian genetics, the expected is 50%, the degree of heritability will be determined by the ratio between both correlations; that is 40% / 50% = 80%
To establish the expected correlation, in the case of siblings and dizygotic or non-identical twins, we would have to start with some theoretical hypotheses. It would not be the same if we knew the rules for determining the supposed dominant and recessive gene, or if there were various genes or chromosomes intervening in the characteristic being studied, it would be quite complicated in this ultimate case.
The second comment refers to if environmental circumstances were really important, it would be worth waiting for a greater resemblance between the IQ and intelligence quotients in siblings and dizygotic and non-identical twins than those actually observed. I think that these circumstances are fundamentally equal within the same family, except if we pay disproportionate attention to the influence that having a different math teacher or of any other subject or circumstance could have; we might find that the sum of all the parameters would be greater than the unit.
There is not very much of this type of work on intelligence or IQ inheritance, or it is less known; normally the results on the correlation of the IQ are fairly low. If the indicated correlation were made at the previous point regarding the observed and expected correlations, the results may not have been so low.
The genetic affinity in intelligence between parents and children will be, as a maximum, equal to that of siblings or dizygotic and non-identical twins.
The EDI Study – Evolution and Design of Intelligence included in the annex belongs to this group. If the same results are maintained in additional studies, the debate over the heritability of intelligence could be brought to a close, at least in its current scale. At the same time, it is possible that more profound debates are becoming more important.
The obtained correlation was higher than 80% in many cases, reaching 96% and 99% in some of them.
The key of success doubled. On one hand, the GTCEL model was incorporated, and on the other, the information was grouped so that it would compensate for variations due to the random component of the combination.
In fact, with the grouping, it is not necessary to correct the rising correlation observed according to the expected correlation. There is the advantage that the expected correlation does not need to be known; and also that other possible variable, which could affect both intelligence and the problems in its manifestation and measurement, of small intensity and random distribution can be compensated for.
The multiple dimensions that the different groupings imply, has allowed an analysis of sensitivity to be carried out in respect to the function being studied; altering partial aspects of the model's structure and the parameters involved with a reasonable guarantee that the results are not due to somewhat random coincidences of the sample information.