Yeates, Jones, Wills, Aitken, McLaren and McLaren (2012) devised a serial reaction time (SRT) task that provided evidence for human learning without awareness. Adapting the SRT paradigm usually employed to investigate implicit learning, participants responded to two simple white circle fills on either side of a screen. Instead of these following a sequence that participants were unaware of (e.g. Willingham, Nissen & Bullemer, 1989) this task involved a separate stimulus, which was sometimes predictive of one of the circle fills. A square in the center of the screen would fill with one of eight colors before each circle fill: one of these colors predicted a right circle fill and the other a left on 80% of trials on which those colors occurred. When pressing the key that followed the consistent response trained with these two colors, participants were both faster and more accurate than when responding to either the inconsistent response or control colors. Participants demonstrated a lack of contingency awareness, performing at chance in identifying the predictive colors and on a suitably sensitive prediction task. On reanalyzing this result, this paper shows that it was confounded with a sequential artifact produced by the experimental design itself. Pilot studies demonstrated weak learning of color contingencies when the artifact was removed, thus we sought to improve learning by both increasing the amount of training and placing the predictive color cue on the circle fills. Without the sequential artifact, we can produce the same result, although we concede the effect is less robust than we first indicated. Thus, we are able to reiterate our original conclusion: that this task can demonstrate learning of color contingencies in the absence of awareness and can be used to investigate implicit learning in humans.