By varying the probabilities that a stimulus would appear at particular times after the presentation of a cue and modeling the data by the theory of visual attention (Bundesen, 1990), Vangkilde, Coull, and Bundesen (2012) provided evidence that the speed of encoding a singly presented stimulus letter into visual short-term memory (VSTM) is modulated by the observer's temporal expectations. We extended the investigation from single-stimulus recognition to whole report (Experiment 1) and partial report (Experiment 2). Cue--stimulus foreperiods were distributed geometrically using time steps of 500 ms. In high expectancy conditions, the probability that the stimulus would appear on the next time step, given that it had not yet appeared, was high, whereas in low expectancy conditions, the probability was low. The speed of encoding the stimuli into VSTM was higher in the high expectancy conditions. In line with the Easterbrook (1959) hypothesis, under high temporal expectancy, the processing was also more focused (selective). First, the storage capacity of VSTM was lower, so that fewer stimuli were encoded into VSTM. Second, the distribution of attentional weights across stimuli was less even: The efficiency of selecting targets rather than distractors for encoding into VSTM was higher, as was the spread of the attentional weights of the target letters.