Jay Joseph is a psychologist particularly keen on showing the flaws underlying behavior genetics and twin studies in particular. His landmark book is pretty old (2004) but a new book by him is coming out in December. It'll be interesting to see what he's got to say about the evidence that has accumulated over the last 10 years.
His main criticism is that twin studies rely on the equal environments assumption (that MZ twins do not share a more similar environment than DZ twins, and even if they do, this is due to underlying genetic causes) which he regards as false.
His new book “The Trouble with Twin Studies: A Reassessment of Twin Research in the Social and Behavioral Sciences,” will be published in December, 2014

http://www.madinamerica.com/author/jjoseph/

He acknowledges that his criticism does not affect studies of reared apart twins, such as Bouchard's Minnesota study, but claims that these have another set of issues. Hopefully these will be discussed in his book, but I've written on his blog asking him for more details.

It's odd that he focuses on EEA. I think the assumption has been proven a while ago.

Loehlin, J. C. (1989). Partitioning Environmental and Genetic Contributions to Behavioral Development. American Psychologist, 44, 1285-1292.
Derks, E. M., Dolan, C. V., & Boomsma, D. I. (2006). A test of the equal environment assumption (EEA) in multivariate twin studies. Twin Research and Human Genetics, 9(03), 403-411.

He acknowledges that his criticism does not affect studies of reared apart twins, such as Bouchard's Minnesota study, but claims that these have another set of issues. Hopefully these will be discussed in his book, but I've written on his blog asking him for more details.

Bouchard (2009) put the matter well:

Multiple independent lines of evidence support the conclusion that human intelligence is under strong genetic influence when assessed in the ordinary range of environments in advanced industrialized societies. None of the lines of evidence is either perfect or foolproof. Nevertheless, as Scarr (1981) has argued in response to earlier criticism of behavior genetic studies, ‘the most important fact is that the flaws of one study are not the same as those of another, there are non-overlapping cracks in the evidence. . .Each study can be criticized for its lack of perfection, but laid on top of one another, the holes do not go clear through.’ (p. 528). Picking apart individual studies and pointing out possible flaws in the face of a coherent body of evidence within which the hypotheses has already been tested using alternate designs is common, but it is a regressive approach to science (Urbach 1974a,b; Sesardic 2005) and deserves a name. I have called it pseudoanalysis (Bouchard 1982). To put the argument in another way, it is possible to subject the various proposed causes of IQ score variance (often posed as criticisms of behavior genetic designs) to multiple empirical tests (strong inferences strategy) and the consistent result has been failure to support the hypothesis that environmental factors are the sole cause of variance in IQ. Scientists in most disciplines typically chose the most powerful model system available to test scientific hypotheses. The two most powerful model systems available to study human individual differences are the URT design (no genes in common, highly similar rearing environments) and its inverse the MZA design (almost all genes in common, dissimilar rearing environments). These designs show that, in the range of environments under discussion, growing up in the same family is at best a very modest source of adult IQ similarity and that genetic factors are a major source of similarity. All the other behavior genetic designs support this finding. The most parsimonious conclusion that can be drawn from this large and continually expanding body of evidence is that human intelligence, when assessed in the ordinary range of environments in advance industrialized societies, is more highly heritable than psychologists have previously thought.

Yes, all of the methods have limitations, but their strengths compliment each other.

Edit: For traits which are poorly studied and for which there are not multiple overlapping lines of evidence (e.g., the heritability of sexual orientation or political orientation X) such critiques are valid.