Trial-level bias score versus mean bias score: Comparison of the reliability and external validity using dot-probe task among daily smokers

Introduction: Attentional bias (AB) is an individual difference risk factor that represents the extent to which cigarette cues capture one's attention. AB is typically indexed by mean bias score (MBS), theoretically assuming that AB is static. However, poor reliability of MBS has threatened valid interpretation of the results on AB. Based on observed trial-by-trial temporal fluctuation and variability of attentional allocation, trial-level bias score (TLBS) has been introduced as an alternative index with evidence of better psychometric properties in various populations, as compared to MBS. However, such evidence is limited among daily smokers. The current study aimed to replicate and extend extant findings in a sample of daily smokers by hypothesizing that TLBS, as compared to MBS, would demonstrate superior reliability and external validity. Methods: Forty-eight daily smokers completed self-reports, ad-libitum smoking, and a dot-probe task three times, which was comprised of 36 pairs of pictorial stimuli of cigarette and neutral cues, yielding 144 total trials. Results: The TLBS demonstrated superior internal (range intra class correlation [ICC] = 0.79–0.95) and test-retest reliability (range ICC = 0.64–0.88) compared to MBS (range ICC = 0.31–0.40 and 0.06–0.16, respectively). However, few significant relations between either the MBS or TLBS and measures of biobehavioral and self-report indices of smoking reinforcement were observed. Conclusions: The current findings demonstrate that TLBS, as compared to MBS, is a more reliable measure of AB among daily smokers, while evidence of its external validity is limited.