Unpredictability of an unconditioned stimulus (US) is known to produce context conditioning in animals and humans. We modified the Martians task −a computer game in which learning of Pavlovian associations is measured using conditioned suppression− for assessing context conditioning in humans. Two experiments are reported: one between-subjects and one within-subjects study. Both experiments comprised four conditions. A predictable (Paired) condition in which the US was predicted by the conditioned stimulus (CS), a neutral condition (No-US) and two unpredictable conditions, one in which the CS did not predict the US (Unpaired) and one in which only unsignaled and temporally unpredictable USs were scheduled but no CSs were presented (US-only). The background color of the computer screen was changed to manipulate context. Results showed that at the end of acquisition, more conditioned suppression to the context occurred in the unpredictable conditions (Unpaired and US-only) compared to the predictable condition (Paired). For conditioned suppression to the CS, on the other hand, more cue conditioning was present in the Paired condition than in the Unpaired condition. Consistent with animal research, context conditioning was increased by unpredictability. These data support the Martians task as a promising tool to extend research on human context conditioning.