Abstract
Memory is often superior for pictures relative to words. Dual-coding theory (Paivio, 1969) proposes that this is because pictures lead to imagery plus verbal labelling, taking advantage of two codes, whereas words provide only a verbal representation in memory. We investigated whether common symbols (e.g., !@#$%&) are processed with dual codes, like pictures, or a single code, like words. Participants’ memory were tested for symbols or words (e.g., $ or ‘dollar’). We predicted that symbols are processed using imagery, much like pictures, and as a result memory for symbols should be superior to words. Our prediction was supported across four experiments: Symbols were consistently better remembered than words, regardless of setting, design, or retrieval test type. In a fifth experiment, memory for symbols was driven in-part by participants' familiarity with the stimuli as well as the highly memorable visual properties that symbols possess (as estimated by the ResMem neural network). These findings are consistent with the idea that symbols benefit memory by eliciting distinct representations at encoding.