"The decoding of brain states using machine learning techniques has become a widely used approach in many areas of neuroscience research. The performance of the classifier under different conditions, as well as the "by products" of the decoding process, can be used to make inferences about how the brain encodes information. Decoding methods are also central to the development of non-invasive brain-computer interfaces (BCI) where decoding is done in real-time. Non-invasive real-time decoding techniques have obvious control applications, but less obvious is their potential as a research tool. One of the most exciting prospects for real-time decoding in basic neuroscience research is the potential to modify the sensory context in response to brain events as they happen, or perhaps even slightly before they happen, in order to study the correspondence between brain states and subjective states in a very direct way. This points to a key distinction in real-time decoding: the discrimination of one brain state from another given that the time of occurrence is fixed or known, versus the asynchronous detection of a particular brain state – the decoder remains idle, but attentive, and responds only when the particular state is detected. Recent advances in non-invasive BCI research along the latter front (high detection rates, very low false-alarm rates, very short latency) demonstrate that such an experimental approach is indeed feasible. Our aim is to use asynchronous real-time decoding to study the onset of movement intention and the onset of conscious sensory events. The former will find applications in the study of forward models and in schizophrenia research. The latter can be developed into a novel diagnostic and clinical tool. In both cases we will study the performance of the decoder across a range of latencies in order to gain insight into the buildup toward a conscious neural event, in the context of evidence-accumulation models of perception and decision-making."