The limits of unconscious semantic priming

Semantic information can be accessed unconsciously, yet it remains unclear to what extent unconscious semantic information spreads across association networks. We compared conscious and unconscious semantic priming among different levels of semantic associations: direct, cross-form, and metaphoric associations. Chinese words associated with thermal qualities (cold or warm) were adopted as the primes, followed by a discrimination task regarding the target being associated with coldness or warmth. The targets were (1) words with thermal qualities, (2) illustrations representing thermal qualities, or (3) words describing personality traits that are metaphorically associated with thermal qualities. We first demonstrated the typical semantic priming effect in the three types of semantic associations when the prime was visible (Experiment 1). We then rendered the primes invisible using the continuous flash suppression paradigm and found a reversed semantic priming effect for the direct association and yet no priming effects for the cross-form and metaphoric associations (Experiment 2). These results suggested that unconscious semantic priming only occurs between directly associated stimuli while consciousness is necessary for higher-level associations and facilitatory interactions, delineating the contrast between unconscious and conscious semantic processing.