A rapid paradigm shift has taken place over the past decade that fundamentally alters the field of behavioral genetics. We have moved from viewing DNA sequences as somehow providing the fixed, immutable, predetermined basis to behavior to viewing these as interacting components, albeit critical ones, in the complex systems that underpin physical and behavioral development (Jaenisch & Bird Reference Jaenisch and Bird2003).

The specialism of epigenetics has developed rapidly (Bird Reference Bird2007; Jablonka & Lamb Reference Jablonka and Lamb2008) and did not even warrant mention in a major behavioral genetics text as recently as 2003 (Plomin et al. Reference Plomin, DeFries, Craig and McGuffin2003). Genetics and epigenetics interact – epigenetic processes are in part regulated by genetic mechanisms (Yuan Reference Yuan2012), whereas gene functions are themselves epigenetically regulated (Trakhtenberg & Goldberg Reference Trakhtenberg and Goldberg2012).

Prenatal and perinatal exposures alter development, and the effects may persist over generations (Bloomfield 2011; Gicquel et al. 2008; Gluckman & Hanson Reference Gluckman and Hanson2005; Gudsnuk & Champagne Reference Gudsnuk and Champagne2011). Epigenetic marking and gene methylation are keys in this process (Chatterjee & Morison Reference Chatterjee and Morison2011; LaSalle Reference LaSalle2011; Szyf Reference Szyf2011; van Ijzendoorn et al. 2011) and can help to explain developmental commonalities and individual differences (Mehler Reference Mehler2008). Placental nutrition and intrauterine environment influence the fetal phenotype. Charney rightly highlights this as one important aspect that has not received its due attention in genetic analyses and draws our attention to other factors such as parent-of-origin effects.

A further limitation in classic genetic studies is that data analyses often minimize complexity. While Occam's razor has its place, there is seldom appropriate consideration of issues like multicollinearity (Rahbar et al. Reference Rahbar, Samms-Vaughan, Loveland, Pearson, Bressler, Chen, Ardjomand-Hessabi, Shakespeare-Pellington, Grove, Beecher, Bloom and Boerwinkle2012).

Many genetic issues influence behavioral development, including differences in microsatellite DNA (Hammock & Young Reference Hammock and Young2005) that are themselves affected by factors like diet (McGowan et al. Reference McGowan, Meaney and Szyf2008). Various genetic anomalies have been linked to specific developmental psychopathologies (Aitken 2010; Geschwind Reference Geschwind2009).

Social behavior directly affects gene expression, with implications for infant development and animal models (Branchi et al. 2011; Meaney Reference Meaney2001). The ways in which such influences operate are complex (Lee et al. Reference Lee, Raygada and Rennert2012) and are interlinked with many other aspects of the biopsychosocial environment. Where genetic anomalies are known, or in laboratory environments where effects can be evaluated ceteris paribus, the strength of specific contributions may be readily demonstrable. In most real-world situations, however, the effects of specific factors are harder to deconstruct (Robinson et al. Reference Robinson, Fernald and Clayton2008).

Some brief examples indicate the range and complexity of issues that can influence developmental processes. In humans, (1) maternal immune status can affect the likelihood of seizures developing in a child with fragile-X syndrome (Chonchaiya et al. Reference Chonchaiya, Tassone, Ashwood, Hessl, Schneider, Campos, Nguyen and Hagerman2010); (2) particular patterns of nutrient intake during pregnancy can alter the course of fetal brain development (Brantsæter et al. Reference Brantsæter, Birgisdottir, Meltzer, Kvalem, Alexander, Magnus and Haugen2011); (3) maternal consumption of foodstuffs (such as licorice) can affect the human fetus, impairing placental barrier protection by lowering 11ß-HSD2 levels against circulating maternal glucocorticoids (Räikkönen et al. Reference Räikkönen, Pesonen, Heinonen, Lahti, Komsi, Eriksson, Seckl, Jarvenpaa and Strandberg2009); and (4) the effects of maternal stress on the fetus during pregnancy partly results from differences in placental gene activation (Gheorghe et al. Reference Gheorghe, Goyal, Mittal and Longo2011). In other animals, (1) dietary supplements can prevent fetal hypomethylation otherwise resulting from exposure to neurotoxic factors like bisphenol-A (Dolinoy et al. Reference Dolinoy, Huang and Jirtle2007); and (2) lab chow differences can have major effects on gene expression in mice (Kozul et al. Reference Kozul, Nomikos, Hampton, Warnke, Gosse, Davey, Thorpe, Jackson, Ihnat and Hamilton2008).

Twin studies have been a fundamental source of information in human behavioral genetics (Bell & Saffery Reference Bell and Saffery2012; Boomsma et al. Reference Boomsma, Busjahn and Peltonen2002). The validity of simple monozygotic–dizygotic (MZ–DZ) comparison studies that form the bulk of this work has been questioned for various reasons. For example, fetal growth differences result from monoplacental twinning, and there are implications of mono versus dichorionicity (Nikkels et al. Reference Nikkels, Hack and van Gemert2008). Simple mechanical factors like chord entanglement and more complex issues like transfusion syndromes are only found with monochorionic pregnancies (Machin Reference Machin2004), and they are consequently far more likely to affect MZ than DZ twins. In more general ways multiparity increases risks to offspring. They are typically born earlier and have a higher rate of other perinatal complications, with a three- to sevenfold increased rate of morbidity and mortality as compared to singletons (Sherer Reference Sherer2001). Many twin similarities lessen with age, resulting from factors such as divergence of methylation status (Wong et al. Reference Wong, Caspi, Williams, Craig, Houts, Ambler, Moffit and Mill2010). Methylation status is, however, unlikely to account for most such MZ twin discordance (Chatterjee & Morison Reference Chatterjee and Morison2011). Overall, such findings show that the MZ–DZ gene research is unduly simplistic and question the generalizability of many results from twin populations. By inverting the typical paradigm and investigating sources of discordance, research on twins may help us to identify epigenetic mechanisms (Bell & Spector Reference Bell and Spector2011).

Knowledge of these mechanisms is an important component to understanding developmental psychopathologies. This is complementary to and highlights the limitations of purely “gene-centric” models. These neurobiological processes are relevant to the assessment and focused treatment of many conditions where diagnosis and management have more traditionally been based only on history and presenting behavior. This point applies particularly to people with attention-deficit/hyperactivity disorder (ADHD; Archer et al. Reference Archer, Oscar-Berman and Blum2011) and autism spectrum disorder (ASD; Coleman & Gillberg Reference Coleman and Gillberg2012; Grafodatskaya et al. 2010).

Rett syndrome is a paradigm condition (Chahrour & Zoghbi Reference Chahrour and Zoghbi2007), having been identified as a disorder linked to dysfunction of the methyl-CpG binding protein MeCP2 involved in chromatin remodeling and RNA splicing (Amir et al. Reference Amir, Van den Veyver, Wan, Tran, Francke and Zoghbi1999). Knowledge of the mechanism has enabled a reversible loc-cassette model to be developed (Guy et al. Reference Guy, Jian, Selfridge, Cobb and Bird2007).

In both ASD and ADHD there are a wide range of genetic and epigenetic anomalies and a range of comorbidities. Many overrepresented differences such as mitochondrial abnormalities demonstrate compromised metabolism (Rossignol & Frye Reference Rossignol and Frye2012; Weissman et al. 2008). Other neuro-metabolic disorders such as sleep difficulties (Aitken Reference Aitken2012), neuromuscular dysfunction (Marrone & Shcherbata Reference Marrone and Shcherbata2011), and seizure disorders (Qureshi & Mehler Reference Qureshi and Mehler2010) are also more common than in the general population.

Despite the broadening range of sufficient neurobiologies, clinical classification is becoming steadily more broad-based and homogenized (“lumping,” not “splitting”). In part, this facilitates large-N research on otherwise heterogeneous conditions like ASD (Hus et al. Reference Hus, Pickles, Cook, Risi and Lord2007). Through this process, however, there is a danger of marginalizing the relevance of much of the rich information on subgroup-specific neurobiology being generated through research. I have tried to illustrate this in the examples above. This obvious divergence should serve as a wake-up call before the increasing specificity of bioscience and the broadening generalization of clinical practice become an irreparable barrier to progress. It should help us to focus on the need for more informative multidisciplinary approaches, increasing the scope for collaborative, focused research relevant to clinical management and treatment. Recognizing and acknowledging such complexities should enable us to progress from the current unrealistic emphasis placed in clinical practice on an increasingly incompatible research-evidence base to one with a greater focus on practice-relevant evidence (Green Reference Green2006). This should in turn inform a more valid and fit-for-purpose research paradigm that would mutually enhance both academic research and clinical practice (Ioannidis 2006; Sidman Reference Sidman2011).