The dopamine hypothesis of schizophrenia has been one of the most enduring ideas in psychiatry. Initially, the emphasis was on a role of hyperdopaminergia in the etiology of schizophrenia (version I), but it was subsequently reconceptualized to specify subcortical hyperdopaminergia with prefrontal hypodopaminergia (version II). However, these hypotheses focused too narrowly on dopamine itself, conflated psychosis and schizophrenia, and predated advances in the genetics, molecular biology, and imaging research in schizophrenia. Since version II, there have been over 6700 articles about dopamine and schizophrenia. We selectively review these data to provide an overview of the 5 critical streams of new evidence: neurochemical imaging studies, genetic evidence, findings on environmental risk factors, research into the extended phenotype, and animal studies. We synthesize this evidence into a new dopamine hypothesis of schizophrenia—version III: the final common pathway. This hypothesis seeks to be comprehensive in providing a framework that links risk factors, including pregnancy and obstetric complications, stress and trauma, drug use, and genes, to increased presynaptic striatal dopaminergic function. It explains how a complex array of pathological, positron emission tomography, magnetic resonance imaging, and other findings, such as frontotemporal structural and functional abnormalities and cognitive impairments, may converge neurochemically to cause psychosis through aberrant salience and lead to a diagnosis of schizophrenia. The hypothesis has one major implication for treatment approaches. Current treatments are acting downstream of the critical neurotransmitter abnormality. Future drug development and research into etiopathogenesis should focus on identifying and manipulating the upstream factors that converge on the dopaminergic funnel point.

The hypothesis that dopamine and dopaminergic mechanisms are central to schizophrenia, and particularly psychosis, has been one of the most enduring ideas about the illness. Despite a relatively inauspicious start—dopamine was initially thought to be a precursor molecule of little functional significance—the idea has evolved and accommodated new evidence to provide an increasingly sophisticated account of the involvement of dopamine in schizophrenia. This review summarizes the evolution of the dopamine hypothesis, which we characterize as having 2 main prior incarnations (version I, the original incarnation, and version II, which was articulated in 1991 and has been the guiding framework since). The main effort in this article is to synthesize the evidence since version II and articulate what we call “The Dopamine Hypothesis of Schizophrenia: Version III,” which represents the most parsimonious account of the current state of knowledge. We call it version III—because we expect it to be revised. However, we highlight features of version III that we believe are sufficiently well established that they are likely to be constant in future revisions, as well as aspects that are still in evolution. Finally, we review the explanatory power of the hypothesis—indicating the known aspects of schizophrenia that it can and cannot explain.

The first version of the dopamine hypothesis could be entitled the dopamine receptor hypothesis. It emerged from the discovery of antipsychotic drugs¹ and the seminal work of Carlsson and Lindqvit who identified that these drugs increased the metabolism of dopamine when administered to animals.² Further evidence came from observations that reserpine, which is effective for treating psychosis, was found to block the reuptake of dopamine and other monoamines, leading to their dissipation.³ Studies showing that amphetamine, which increases synaptic monoamine levels, can induce psychotic symptoms (reviewed in Lieberman et al⁴) provided additional evidence. It was not until the 1970s, however, that the dopamine hypothesis was finally crystallized with the finding that the clinical effectiveness of antipsychotic drugs was directly related to their affinity for dopamine receptors.^5–7 The focus at the time was on excess transmission at dopamine receptors and blockade of these receptors to treat the psychosis (eg, Matthysse⁸ and Snyder⁹). While version I accounted for the data available then, it was seen as a hypothesis of schizophrenia as a whole without a clear articulation of its relationship to any particular dimension (eg, positive vs negative symptoms) and no link was made to genetics and neurodevelopmental deficits (understandably as little was then known about them), and there was little clear indication of where the abnormality was in the living brain—this would require the later application of in vivo imaging techniques. Additionally, dopamine was thought of in isolation, with little consideration of how it might relate to known risk factors for schizophrenia, and finally there was no framework for linking the dopaminergic abnormality to the expression of symptoms.

Much has changed since version II. There have been more than 6700 articles and 181000 citations to the topic of “dopamine and schizophrenia” since 1991. It is not possible to provide a comprehensive review of all the new findings since then, much less try to weave them into a coherent hypothesis. So, the focus of our effort is to identify the 5 most critical streams of new evidence, briefly summarize what we see as the key findings from these, and use them to develop the most parsimonious understanding of the role of dopamine in schizophrenia—version III.

The dopamine hypothesis ‘version II’ was published before the Human Genome Project and the huge advances in genetic research in schizophrenia. After over 1200 studies, it seems clear that no one gene “encodes” for schizophrenia.⁶³ Rather, in common with many other complex diseases, there are a number of genes each of small effect size associated with schizophrenia.⁶³ The gene database on the Schizophrenia Research Forum (http://www.schizophreniaforum.org) provides a systematic and regularly updated meta-analysis of genetic association studies. As of autumn 2008, 4 of the top 10 gene variants most strongly associated with schizophrenia are directly involved in dopaminergic pathways. The strongest association is with a gene variant affecting the vesicular monoamine transporter protein (rs2270641, odds ratio 1.63). This protein acts to accumulate dopamine and other monoamines into vesicles, which fits with the PET studies that show elevated radiolabeled dopamine accumulation into striatal vesicles in schizophrenia. Additionally, other gene variants in the list of the strongest associations, such as in the genes for methylenetetrahydrofolate reductase and V-akt murine thymoma viral oncogene homolog 1, indirectly affect the dopaminergic system among other effects.⁶⁴ Many of the other gene variants in the top list are involved in brain development, such as the gene for dysbindin, or influence more ubiquitous brain transmitters such as glutamate or γ-aminobutyric acid (GABA).^63,64 While recent findings have breathed great interest in the copy number variations in schizophrenia—the early evidence there also suggests that they are rare, tend to be unique to families, and are unlikely to account for more than a few percent of schizophrenia.^63,65–67 It would be premature to try and synthesize these genes into a pathway leading to dopamine abnormality because the precise number, nature, function, and association of these genes to schizophrenia is evolving. The most parsimonious statement that can be made today is that while a number of genetic associations have been identified, none of them accounts for the majority of schizophrenia and most of them are likely to be susceptibilities. Of the ones that have been identified, some have already been tied to altered dopamine transmission.⁶⁸ However, the functional relevance of most of them to dopamine function is not known.⁶⁸ This view of schizophrenia genetics then reemphasizes a critical role for other interacting factors—particularly the environmental risk factors for schizophrenia.

Striatal dopamine elevation is not seen in mania, depression, or other psychiatric disorders without psychosis^142–147 and not related to measures of anxiety or depression in people with psychotic symptoms.^20,148 Thus, it is not a nonspecific indicator of psychiatric morbidity. However, striatal dopamine elevation is seen in psychosis associated with psychosis in at least one disorder other than schizophrenia.²² Furthermore, dopamine blockade with antipsychotic drugs does not respect diagnostic boundaries either—it is effective for psychosis related to mania, depression, or Parkinson disease^149,150 as well as for psychosis in schizophrenia. While further studies and direct comparisons are required, dopamine elevation appears specifically related more generally to psychosis proneness and not just to psychosis in schizophrenia.

If a neurochemical hypothesis (based on dopamine or any other neurotransmitter) is to explain a psychiatric illness defined by its clinical expression, it has to link the 2. A major shortcoming of the first 2 versions of the dopamine hypothesis was the total silence on the issues of how dopaminergic abnormalities led to the clinical expression of the disease. Since version II of the dopamine hypothesis, developments in neuroscience have provided increasing evidence of dopamine's role in motivational incentive salience. The experiments and syntheses of data by Berridge and Robinson,¹⁵¹ Robbins and Everitt,^152,153 and Schultz and others^154–158 have implicated a distinct role for subcortical dopamine systems in incentive or motivational salience and reward prediction, respectively. These conceptualizations provided a framework to link neurochemical dysfunction to clinical expression using concepts of salience and reward. According to one such extension of the dopamine hypothesis,^159,160 the abnormal firing of dopamine neurons and the abnormal release of dopamine leads to an aberrant assignment of salience to innocuous stimuli. It is argued that psychotic symptoms, especially delusions and hallucinations, emerge over time as the individual's own explanation of the experience of aberrant salience. Psychosis is, therefore, aberrant salience driven by dopamine and filtered through the individual's existing cognitive and sociocultural schemas—thus allowing the same chemical (dopamine) to have different clinical manifestations in different cultures and different individuals.^159,160 Incentive salience models also provide a plausible explanation for negative symptoms: dopamine dysregulation may increase the noise in the system, “drowning out” dopaminergic signals linked to stimuli indicating reward, eg, Roiser et al¹⁶¹ and Seamans and Yang.¹⁶² The net result would be reduced motivational drive that would lead over time to negative symptoms, such as social withdrawal, and neglect of interests. As an explanation, this has face validity, and there is some evidence that schizophrenia is associated with reduced ventral striatal activation to reward, and greater reduction is related to higher levels of negative symptoms.¹⁶³ However, this proposal and the hypothesis linking low frontal dopamine levels to the cognitive impairments in schizophrenia both need to be tested by further in vivo studies of neurochemical function in patients.

The hypothesis that the final common pathway is presynaptic dopamine dysregulation has some important clinical implications. Firstly, it implies that current antipsychotic drugs are not treating the primary abnormality and are acting downstream. While antipsychotic drugs block the effect of inappropriate dopamine release, they may paradoxically worsen the primary abnormality by blocking presynaptic D2 autoreceptors, resulting in a compensatory increase in dopamine synthesis. There is some evidence from healthy volunteers that acute antipsychotic treatment does increase presynaptic dopamine synthesis capacity,¹⁶⁶ and while successful subacute treatment can reduce this,¹⁶⁷ it is nevertheless elevated in patients who have received antipsychotic treatment for many years.¹⁷ This may explain why patients relapse rapidly on stopping their medication, and if the drugs may even worsen the primary abnormality, it also accounts for more severe relapse after discontinuing treatment. This suggests that drug development needs to focus on modulating presynaptic striatal dopamine function, either directly or through upstream effects.

An attractive feature of version II was that it proposed a dysfunction in the dopamine system as a complete explanation for schizophrenia: a prefrontal hypodopaminergia leading to a subcortical hyperdopaminergia. We depart from this parsimony in version III mainly because in the last 2 decades there has been little convincing evidence for this sequence of dopamine dysfunction. On the other hand, the last 2 decades have provided substantially more evidence about the multiple routes (genetic, neurodevelopmental, environmental, social) that lead to the striatal hyperdopaminergia, as discussed earlier. Furthermore, the appreciation of the dimensional nature of symptoms of schizophrenia also speaks for partial independence of the different features (cognitive, negative) from psychosis.¹³⁹ There is of course correlational evidence that striatal dopamine abnormalities are associated with poor performance on cognitive tasks^17,20,168 and suggestion that higher striatal dopamine synthesis capacity is linked to functional abnormalities in the cortical regions engaged by these tasks.^168,169 However, it should be noted that recent data suggest that these frontal/cognitive changes need not necessarily be primary but instead may arise as a consequence of striatal dysfunction.¹⁷⁰ Thus, in contrast to version II, which proposed a single pathway, we propose that changes in multiple transmitter/neural systems underlie the cognitive dysfunction and negative symptoms of schizophrenia, and in many cases these dysfunctions precede the onset of psychosis. It is when these pathways, in convergence with other biological or environmental influences, lead to striatal dopamine hyperfunction that psychosis becomes evident and the label of schizophrenia is assigned. Thus, rather than being a hypothesis of schizophrenia—version III is more accurately a “dopamine hypothesis of psychosis-in-schizophrenia.” It remains to be tested whether this is specific to psychosis of schizophrenia or is seen with psychosis in other disorders too.

Because so much is unknown, it is given that the hypothesis will be revised as more data become available. The more intriguing question is whether one can envisage evidence that would lead to a wholesale rejection of the hypothesis. The 2 central claims of version III are the primacy of the presynaptic abnormality and the claim that dopamine is the “final common pathway.” Two different kinds of evidence could lead to a complete rejection of the hypothesis. PET studies directly implicating presynaptic dopamine dysfunction are a major foundation of this new version of the hypothesis. PET data require to be modeled to provide estimates of L-dopa uptake or synaptic dopamine levels—and the results are inferred rather than direct measurements. Thus, if it turns out that the body of evidence based on PET imaging is a confound or an artifact of modeling and technical approaches, this would be a serious blow for version III, though the data behind versions I and II would still stand strong. While possible, we think this to be highly unlikely. What is perhaps more likely is that a new drug is found that treats psychosis without a direct effect on the dopamine system. In other words, the dopamine abnormalities continue unimpeded, and psychosis improves despite them. A good example of such a new drug might be LY2140023, an mGlu 2/3 agonist.¹⁷¹ If this were to be an effective antipsychotic and it could be shown that the new pathways do not show any interaction with the dopamine system, then the fundamental claim of version III, that it is the final common pathway, would be demolished. A similar situation would arise if a pathophysiological mechanism that does not impact on the dopamine system is found to be universal to schizophrenia. Much more likely is the possibility that the hypothesis will be revised but with a stronger version IV. The next decade will provide more information on the role of dopamine, particularly how genetic and environmental factors combine to influence the common pathway, and better drugs will be developed that directly influence presynaptic dopaminergic function—both logical successors to the idea of a final common pathway.

A considerable body of new evidence has amassed in the last 2 decades that is not compatible with reconceptualization of Davis and colleagues of the dopamine hypothesis of schizophrenia. To account for these developments, we have elaborated the dopamine hypothesis of schizophrenia: version III—the final common pathway. This hypothesis accounts for the multiple environmental and genetic risk factors for schizophrenia and proposes that these interact to funnel through one final common pathway of presynaptic striatal hyperdopaminergia. Furthermore, it provides a framework linking the abnormal neurochemistry to symptoms and explains both why many disparate risk factors and functional and structural abnormalities are associated with schizophrenia but are not specific to schizophrenia. It provides an explanation for overlapping findings in people with risk factors for schizophrenia and explains eventual diagnosis not in neurochemical terms but as the result of individual factors interacting with the sociocultural milieu. In addition to funneling through dopamine dysregulation, the multiple environmental and genetic risk factors influence diagnosis by affecting other aspects of brain function that underlie negative and cognitive symptoms. Schizophrenia is thus dopamine dysregulation in the context of a compromised brain. It follows from this that future drug development should focus on the systems acting on the funnel points leading to the final common pathway.