Contradictory evidence exists relating to the presence of an attention bias to sleep-related stimuli in poor sleepers/insomnia using the emotional Stroop task (EST). These inconsistencies may be due to methodological issues related to the affective valence of the sleep-related stimuli. Thus, individuals may attend differentially to sleep-related stimuli not because of their ‘sleep’ properties, but their negativity. The current study addresses this by controlling the affective valence of sleep-related words. A total of 107 participants [mean age = 33.22 years, standard deviation (SD) = 12.31 years; 61.7% female] were recruited during an evening event at the Newcastle Science Festival. Participants completed the Pittsburgh Sleep Quality Index (PSQI) and a computerized EST containing 20 non-affective sleep-related, 20 neutral and 20 negatively valenced threat words. Good and poor sleepers were categorized using the PSQI. There were no significant differences between groups on response latency to sleep-related words (t(105) = –0.30, P = 0.76). However, the interaction between good versus poor sleepers and word-type on response latency was significant (F(2,210) = 3.06, P <0.05). Poor sleepers took longer to respond to sleep-related words (mean = 723.35, SD = 172.55) compared to threat words (mean = 694.63, SD = 162.17) than good sleepers (mean = 713.20, SD = 166.32; and mean = 716.65, SD = 181.14). The results demonstrate the presence of an attention bias towards sleep-related stimuli compared to threat stimuli in poor sleepers. Accordingly, poor sleepers may be consumed by stimuli relevant to their specific difficulties, as well as being more highly attuned to negative cues that signal anxious states. Thus, the present research suggests that there are two opposing forces at play: one which facilitates performance (non-specific threats) and one which hinders performance (personally relevant threats).