Background: Cognitive flexibility deficits are present in patients with schizophrenia and are strong predictors of functional outcome but, as yet, have no pharmacological treatments. Aims: The purpose of this study was to investigate whether the phosphodiesterase type-4 inhibitor, roflumilast, can improve cognitive flexibility performance and functional brain activity in patients with schizophrenia. Methods: This was a within-subject, randomised, double-blind, placebo-controlled, three-period crossover study using a version of the Intradimensional/Extradimensional (ID/ED) task, optimised for functional magnetic resonance imaging (fMRI), in 10 patients with schizophrenia who were scanned after receiving placebo, 100 µg or 250 µg roflumilast for 8 consecutive days. Data from an additional fMRI ID/ED study of 18 healthy participants on placebo was included to contextualise the schizophrenia-related performance and activations. The fMRI analyses included a priori driven region of interest (ROI) analysis of the dorsal frontoparietal attention network. Results: Patients on placebo demonstrated broad deficits in task performance compared to the healthy comparison group, accompanied by preserved network activity for solution search, but reduced activity in left ventrolateral prefrontal cortex (VLPFC) and posterior parietal cortex for attentional set-shifting and reduced activity in left dorsolateral prefrontal cortex (DLPFC) for reversal learning. These ROI deficits were ameliorated by 250 µg roflumilast, whereas during solution search 100 µg roflumilast reduced activity in the left orbitofrontal cortex, right DLPFC and bilateral PPC, which was associated with an improvement in formation of attentional sets. Conclusions: The results suggest roflumilast has dose-dependent cognitive enhancing effects on the ID/ED task in patients with schizophrenia, and provides sufficient support for larger studies to test roflumilast’s role in improving cognitive flexibility deficits in this clinical population.