Studies on talker normalization have reported that contexts would affect how a vowel/consonant/tone is perceived, which suggests that listeners use cues in the context as a reference for speech perception. However, it is unclear how the cues of the context are encoded in memory, and to what extent the context effects are influenced by interruption in sensory memory and reduction in attentional resources. To fill in this gap, this study examined the effects of noise interruption and a secondary visual task on the identification of words carrying level tones in context by native Cantonese speakers. Experiment 1 compared the tone identification performance between a block where a 300ms noise was presented immediately after the context to interrupt the normalization effects of the context on the following target word and a block without noise interruption. Experiment 2 compared the tone identification performance between a block where participants had to perform a secondary visual task (picture same/different discrimination) and a block without a secondary visual task. Results suggest that context cues are likely encoded both in sensory memory and in short-term categorical memory and that reduction in attentional resources has marginal influence on tone normalization.