We investigated linguistic prediction in second language (L2) English speakers using the Visual World Paradigm (VWP) while using a modified version of the Corsi Block Tapping task to vary cognitive load within participants. Participants completed three blocks of trials with either (1) no, (2) a low, or (3) a high cognitive load manipulation. Each trial began with the presentation of 9 squares in a random grid. Depending on the condition, either 0, (no load) 2, (low) or 4 (high) of the squares would, sequentially, turn gray for 500ms. Participants had to remember the location and order of the squares during the VWP task and recreate the sequence after the VWP task. During the VWP task, participants saw 4 pictures while listening to a sentence and needed to click on the picture mentioned during the sentence. The correct picture was either predictable or unpredictable based on the agent of the sentence. Thus, the experiment followed a 2 (predictable/unpredictable) x 3 (no load, low load, high load) design. Preliminary data (n = 7) support that L2 speakers may have graded predictive abilities based on cognitive resource availability. During the low cognitive load block, predictive eye movements to the target were present but started later than in the no load condition. During the high cognitive load block, there was no evidence of predictive eye movements. These findings provide evidence that the formation of linguistic predictions is both modulated and gated by the availability of nonlinguistic cognitive resources.