Visual perception can be affected by training mental representations. However, it remains unclear if training procedures can also affect the quality of mental representations. To investigate if training enhances the fidelity of mental representations retrieved from visual long-term memory (VLTM), we used a task including object-color associations with a continuous response-space. We tested 15 participants in a training group, and 15 participants in a control group. Training consisted of six training runs executed on three consecutive days. Before and after training, we assessed accessibility and fidelity of mental representations in VLTM, and of single objects in visual short-term memory (VSTM). Not only accessibility to mental representation but also their fidelity increased across training and transferred to novel object-color associations in VLTM and VSTM after training. At the end of the training, fidelity of VLTM representations were virtually identical to fidelity of VSTM representations. We conclude, that training objectcolor associations does not only improve the accessibility of VLTM representations, but also their fidelity based on perceptual plasticity of the visual system.

The levels of processing (LOP) account has inspired thousands of studies with verbal material. The few studies investigating levels of processing with nonverbal stimuli used images with nameable objects that, like meaningful words, lend themselves to semantic processing. Thus, nothing is known about the effects of different levels of processing on basic visual perceptual features, such as color. Across four experiments, we tested 187 participants to investigate whether the LOP framework also applies to basic perceptual features in visual associative memory. For Experiments 1 and 2, we developed a paradigm to investigate recognition memory for associations of basic visual features. Participants had to memorize object–color associations (Experiment 1) and fractal–color associations (Experiment 2, to suppress verbalization). In Experiments 3 and 4, we extended our account to cued recall. All experiments revealed reliable LOP effects for basic perceptual features in visual associative memory. Our findings demonstrate that the LOP account is more universal than the current literature suggests.

Synaesthetic consistency is the hallmark of synaesthesia and plays an important role in the definition and validation of synaesthesia. It has been hypothesised that the acquisition of initially unspecified synaesthetic associations is based on consolidation processes. Thus, we investigated in non-synaesthetes whether repeatedly engaging with grapheme-colour associations mimics the developmental trajectory of synaesthetic consistency in genuine grapheme-colour synaesthesia. This was the case for the two tested experimental groups, irrespective of whether they were instructed to memorize their chosen associations, but not for the passive control group. Moreover, consolidated associations of the experimental groups resembled those frequently found in genuine synaesthesia. Furthermore, the acquisition of consistent graphemecolour associations resulted in a transfer of benefits to performance in recognition memory for abstract stimuli, as also found in genuine synaesthesia. Our findings suggest that consistent synaesthetic associations are based on consolidation processes due to repeated engagement with graphemes and colours.