The executive functions of prefrontal cortex (PFC) are regulated by ascending monoaminergic systems. In particular, the dorsal PFC is associated with spatial working memory (SWM), linked to persistent neuronal activity, and it receives anatomical projections from the brainstem monoamine centres. Catecholamine regulation of SWM has been studied at mechanistic and computational levels. However, the role of serotonin (5-HT) in SWM is unclear. Thus, while 5-HT is clearly associated with orbitofrontal cortex functions, its association with SWM is controversial, with positive and negative reports. An indirect association between PFC 5-HT and SWM is suggested by the pro-cognitive effects of antipsychotic drugs targeting 5-HT receptors. Several neurocognitive problems are core deficits of schizophrenia and SWM deficits mark a genetic predisposition for the disease. These weak associations between 5-HT and SWM are in contrast with neurobiological observations predicting a stronger link. Indeed, 5-HT receptors are densely expressed in various PFC subfields, and serotonergic projections from the raphe nuclei target specifically PFC and modulate neuronal activity. Moreover, 5-HT2A receptor activation/blockade in dorsal PFC reduces persistent neuronal activity during SWM tasks. Here, we sought to develop a computational model of serotonergic modulation of the PFC network in order to address the possible reasons for this discrepancy. We used a computational network model of spiking neurons [1] which is consistent with PFC neurophysiological data from behaving monkeys in SWM. This model falls within the general class of attractor models proposed to account for behavioural deficits of schizophrenia, and provides a rich set of behavioural output predictions that can be experimentally tested. We incorporated 5-HT receptor mechanisms in the model following electrophysiological evidence, and we tested the effects of 5-HT receptor ligands on network simulation performance during SWM. We found that 5-HT modulated network’s SWM performance non-monotonously via 5-HT2A and especially via 5-HT1A receptors, following an inverted-U shape. This may partly explain the weak behavioural effects of serotonergic agents in previous SWM studies. Our simulations showed that errors committed at low and high 5-HT concentrations are due to different network dynamics instabilities, suggesting that these two conditions can be distinguished experimentally based on their dependency with delay length, on the response confidence declared in error trials and on PFC activation contrasts in neuroimaging studies. We derived specific predictions regarding the expected behavioural effects of serotonin agents in two classic working-memory tasks: an oculomotor delayed-response task and a delayed matchto-sample task. In brief, the probability of memory storage in the oculomotor task and the rate of misses in the match-to-sample task will be most sensitive to serotonin modulation, depending on delay duration for increases but not for decreases in baseline 5-HT. Our study underscores the relevance of identifying different error types in SWM tasks in order to reveal the association between neuromodulatory systems and SWM. Testing these model-derived predictions in electrophysiological and psychophysical studies of SWM will help advance our understanding of the neural basis of SWM and its neuromodulation by 5-HT receptors.