Scalp EEG data were recorded while participants listened to monophonic music. Forward ridge regression models were fit to assess what features of the stimulus were encoded in the low-rate EEG signal. The link between music features and EEG was assessed by using these models to predict unseen EEG data. (A) Prediction correlations were greatest when the stimulus was described by a combination of acoustic information (A: envelope Env, and its half-way rectified first derivative Env’) and melodic expectations (M: SP, SO, HP, HO). This effect of expectations was significant on the average prediction correlation across all 64 EEG electrodes (*p<10−6). The error bars indicate the SEM across participants. (B) The enhancement due to melodic expectations emerged at the individual subject level. The gray bars indicate the predictive enhancement due to melodic expectation. Error bars indicate the SEM over trials (***p<0.001, **p<0.01, *p<0.05, permutation test). (C) Predictive enhancement due to melodic expectations (AM-A; y-axis) increased with the length of the local context (in bars; x-axis) used to estimate the expectation of each note (ANOVA: *p=0.0003). The boxplot shows the 25th and 75th percentiles, with the whiskers extending to the most extreme data-points that were not considered outliers. Circles indicate outliers. (D) The effect of melodic expectations (rAM-rA) emerged bilaterally on the same scalp areas that showed also envelope tracking. (E) Ridge regression weights for TRFAM. Red and blue colors indicate positive and negative TRF components respectively.