Steady-state visual evoked potential (SSVEP)-based brain-computer interface (BCI) is one of the most promising non-invasive BCI paradigms due to its relatively fast and reliable performance. Nevertheless, it has been reported that a considerable amount of people cannot attain effective control of this kind of BCI, i.e. achieving low classification accuracy of their intentions. Dozens of signal processing algorithms and visual stimulator designs have been attempted to solve this problem. Some of these methods have been proven successful in improving the participants’ performance. However, if the participants cannot generate classifiable signals, even the best methods cannot provide satisfactory results. In extreme cases, if the quality of the SSVEP signal is exceptionally low, it may even prevent the participants to use the SSVEP-based BCI systems. Therefore, in this work, we studied the enhancement of the SSVEP signal by focusing on the user. More specifically, we wanted to investigate whether neurofeedback training (NFT), a technique that uses real-time displays of brain activity to teach participants to self-regulate brain functions of interest, could be used to enhance the participants’ SSVEP quality, causing subsequent increase in the participants’ SSVEP-based BCI performance. We considered three main problems: finding the key training parameters to increase the SSVEP quality, developing a training procedure to improve such parameters, and evaluating experimentally such procedure and its effects on the SSVEP quality as well as SSVEP-based BCI performance. In order to achieve our goal, we firstly investigated the relationship between the SSVEP signal and ongoing background electroencephalography (EEG). Experimental iii results from seven healthy participants showed a negative correlation between the SSVEP and alpha activity more prominently at the occipital cortex. Based on these results, thirty-three healthy participants were screened for their SSVEP-based BCI performance and SSVEP signal-to-noise ratio (SNR). From these participants, ten with “low” SSVEP-based BCI performance (classification accuracy < 80%) and “low” SNR performed NFT for individual alpha band (IAB) amplitude decrease. Experimental results showed that, during the NFT, the participants were able to successfully decrease their IAB amplitude over the sessions. In comparison to a nonneurofeedback control group of ten participants, the NFT group showed an average increase of around 16.5% of their SSVEP SNR, which led to an average increase of around 20.3% of their SSVEP-based BCI performance. Lastly, we attempted to identify the most effective mental strategies in alpha decrease NFT in order to provide guidelines for more efficient NFT designs. The grounded theory methodology (GTM), a systematic methodology that allows the construction of a theory through analysis of data, is adopted to assess how the participants’ selfreported mental strategies during NFT related to training effects. It was found that effective strategies for alpha activity decrease varied among individuals and the most successful mental strategies were related to negative thinking and high demanding cognitive tasks.