The article “Continuous Reaching and Grasping with a BCI Controlled Robotic Arm in Healthy and Stroke-Affected Individuals” by Forenzo et al. adds a clicking component to the traditional 2D movement paradigm for motor control based on EEG. The “click” signal allows for users to move the cursor or robotic arm to the target area and then choose to initiate the movement. Previous alternative methods have used a hovering mechanism where the user hovers over the target for 2 seconds to initiate the action, which makes it more difficult to determine the correct location of action and does not allow the user to rest without triggering an action. Forenzo et al’s method also seeks to enable continuous control of movement, rather than users selecting from a discrete list of actions. Subjects from two groups, healthy and recovered stroke patients, underwent two training sessions with a cursor task, three study trials with the same cursor task, one session with a robotic arm, and one session that performed a reach-and-grasp task with cups using a robotic arm. The movement decoder for each subject consisted of two deep learning models: one for two-dimensional movement output, and one for the click output. Forenzo et al. also wanted to examine whether refinement, where the deep learning model is recalibrated during a session, is beneficial, and so during each session the model was recalibrated and fine-tuned after 5 runs. Overall, subjects performed better than chance, but there were high levels of inter-subject variability, with some subjects achieving substantial control and others performing similarly to chance.