Schizophrenia is a severe mental disorder characterized by a diverse set of symptoms, including positive (hallucinations, delusions) and negative (lack of motivation, social withdrawal, blunted affect) symptoms and cognitive impairment. Cognitive deficits are recognized as forming part of the core psychopathology of schizophrenia, with impairments in learning, memory and attention being common (Weickert et al. 2000; Palmer et al. 2009). Brain-derived neurotrophic factor (BDNF) is a member of the neurotrophin family of growth factors that potentiates synaptic strength and plasticity underlying learning and memory (Lu & Chow, 1999; Poo, 2001; Egan et al. 2003; Hariri et al. 2003; Lu, 2003; Dempster et al. 2005). A single nucleotide polymorphism in the human BDNF gene produces a valine to methionine substitution in the protein prodomain that influences intracellular trafficking, activity-dependent release of BDNF, learning and brain activity (Egan et al. 2003; Hariri et al. 2003). Lower levels of both brain and blood BDNF have also been implicated in the majority of studies of the pathophysiology of schizophrenia (Weickert et al. 2003; Xiu et al. 2009; Rizos et al. 2010; Green et al. 2011). Peripheral blood can provide an important reservoir of BDNF, reflecting BDNF secreted from the brain and BDNF that may be available to influence brain function (Schmidt & Duman, 2010). However, the extent to which circulating BDNF correlates with human brain activity during learning has not been demonstrated and the extent to which lower circulating BDNF may be indicative of the neural correlates of cognitive impairment in schizophrenia has not been shown. Low peripheral BDNF is reported in first-episode psychosis (Buckley et al. 2007) and is associated with cognitive impairment in chronically ill people with schizophrenia (Zhang et al. 2012b). Peripheral BDNF levels are also significantly decreased in people with mild cognitive impairment (Yu et al. 2008). Thus, peripheral BDNF levels may be useful as a biomarker in health and disease. However, to more fully interpret what any change in a potential biomarker may mean for brain function, a relationship between peripheral BDNF and brain activity should be established in the healthy and diseased states. One reason that general predictions about blood–brain relationships in schizophrenia are difficult to make is that studies of peripheral BDNF levels in schizophrenia have reported inconsistent results. Although the majority of studies report decreased circulating BDNF levels (Pirildar et al. 2004; Tan et al. 2005; Grillo et al. 2007; Rizos et al. 2008), some find increased levels (Gama et al. 2007; Reis et al. 2008), and one study reports no significant difference in plasma BDNF levels in people with schizophrenia compared to healthy controls (Lee & Kim, 2009). A recent meta-analysis supports an overall reduction in peripheral BDNF levels in schizophrenia (Green et al. 2011). However, the extent to which alterations in blood BDNF levels relate to abnormal learning-related brain activity in schizophrenia, as compared to healthy individuals, remains to be demonstrated. Healthy cortical neurons innervate and supply BDNF to several brain regions, including the striatum (Alexander et al. 1986; Altar & DiStefano, 1998). As striatal neurons are trophically dependent on BDNF supplied through anterograde transport from the frontal cortex (Hofer et al. 1990; Altar et al. 1997; Rauskolb et al. 2010; Dieni et al. 2012), higher levels of BDNF would be expected to relate to stronger frontal–striatal activity in a healthy brain, and a reduction of prefrontal cortical BDNF, as found in schizophrenia (Weickert et al. 2003; Hashimoto et al. 2005), would be expected to negatively impact striatal function in schizophrenia. There is abundant evidence from neuropsychological and neuroimaging studies demonstrating significant learning impairments and frontal–striatal dysfunction in schizophrenia (Pantelis et al. 1997; Meyer-Lindenberg et al. 2002; Murray et al. 2008; Howes et al. 2009; Weickert et al. 2010; Morris et al. 2012). Probabilistic association learning (requiring gradual learning of probabilistic-based cue–outcome associations) elicits frontal–parietal–striatal activity in healthy people (Poldrack et al. 1999; Fera et al. 2005) and is related to reduced activity in a neural network that includes the dorsolateral prefrontal cortex (DLPFC), parietal cortex and caudate nucleus in schizophrenia (Weickert et al. 2009). Probabilistic association learning is impaired in people with schizophrenia (Weickert et al. 2002; Foerde et al. 2008; Horan et al. 2008), the offspring of people with schizophrenia who are at high risk of developing schizophrenia (Wagshal et al. 2012, 2013) and in unaffected siblings of people with schizophrenia (Weickert et al. 2010; Wagshal et al. 2014), with deficits in overall performance and learning rate. In the present study, we sought to define the relationship between peripheral BDNF levels and the neural processes underlying probabilistic association learning in healthy people and people with schizophrenia. First, we confirmed the reduced brain activation in people with schizophrenia versus healthy controls during a probabilistic association learning test using functional magnetic resonance imaging (fMRI). Second, we tested the extent to which plasma BDNF levels are related to brain activity in healthy people during probabilistic association learning. Third, we determined whether a similar relationship was also present in schizophrenia. We predicted a positive relationship between plasma BDNF levels and brain activity in a neural network consisting of prefrontal–parietal cortices and striatum, which has previously been shown to be relevant for probabilistic association learning in healthy adults. Based on the majority of work showing dysregulation of either BDNF levels or prefrontal cortex activity along with an increase in the truncated BDNF receptor (Weickert et al. 2003; Hashimoto et al. 2005; Wong et al. 2010, 2013; Ray et al. 2014), which is thought to block trophic effects in schizophrenia, we further predicted an abnormal relationship between plasma BDNF levels and brain activity in schizophrenia.