Recent advances in blood oxygen level-dependent-functional MRI (BOLD-fMRI)-based neurofeedback reveal that participants can modulate neuronal properties. However, it is unknown whether such training effects can be introduced in the absence of participants' awareness that they are being trained. Here, we show unconscious neurofeedback training, which consequently produced changes in functional connectivity, introduced in participants who received positive and negative rewards that were covertly coupled to activity in two category-selective visual cortex regions. The results indicate that brain networks can be modified even in the complete absence of intention and awareness of the learning situation, raising intriguing possibilities for clinical interventions.neurofeedback | training | reward | spontaneous activity | functional connectivity T here has been a growing interest in the field of neuroscience in the use of neurofeedback (NF) as a tool to both study and treat various clinical conditions. The uses of NF are diverse, ranging across a variety of motor and sensory tasks (1-4), investigation of cortical plasticity and attention (5-9), to treatment of chronic pain, depression, and mood control (10-13).Recent advances in functional MRI (fMRI) techniques and hardware have made real-time fMRI (rtfMRI) a viable method for NF (14). This enables more anatomically specific training compared with methods such as EEG. This enhanced localization additionally allows to provide feedback to differential activation patterns (6, 15, 16), beyond simple up/down-regulation of a specific region/frequency.Another advance in the field of NF is the finding by several recent studies that participants are able to learn to successfully perform the NF paradigm, even without being given an explicit strategy (8,16). This form of implicit learning is intriguing, both because there have been reports indicating certain advantages to implicit over explicit learning (17,18), but mostly because this opens up previously unidentified pathways for therapeutic intervention, for cases for which there are no specific explicit strategies available (for instance, control over complex networks, such as in epilepsy, or over brain regions whose function is not fully elucidated).However, an important common factor in all previous NF studies was the fact that participants were aware that they were being trained, and received specific goals for this training. A fundamental question that therefore remains unanswered is whether targeted brain networks can still be modulated even in the complete absence of participants' awareness that a training process is taking place. Theories of closed-loop learning provide evidence that such implicit learning through reward cues is possible (19,20). This is an important issue, because it may open the way for NF training even in severe clinical cases such as minimally conscious or vegetative state, where such awareness is absent.In the present study, we examined this question in fMRI experiments in which participants were i...