Previous research has claimed that evaluative conditioning (EC) effects may obtain in the absence of perceptual identification of conditioned stimuli (CSs). A recent meta-analysis suggested similar effect sizes for supra- and subliminal CSs, but this was based on a small body of evidence (k = 8 studies; Hofmann, De Houwer, Perugini, Baeyens, & Crombez, 2010). We critically discuss this prior evidence, and then report and discuss 6 experimental studies that investigate EC effects for briefly presented CSs using more stringent methods. Across these studies, we varied CS duration, the presence or absence of masking, the presence or absence of a CS identification check, CS material, and the instructions communicated to participants. EC effects for longer-duration CSs were modulated by attention to the CS–US pairing. Across studies, we were consistently unable to obtain EC for briefly presented CSs. In most studies, this pattern was observed despite above-chance perceptual identification of the CSs. A meta-analysis conducted across the 27 experimental conditions supported the null hypothesis of no EC for perceptually unidentified CSs. We conclude that EC effects for briefly presented and masked CSs are either not robust, are very small, or are limited to specific conditions that remain to be identified (or any combination of these).