When people are asked to ‘explain’ a time series consisting of population statistics, they will suggest factors responsible for later events rather than for earlier events. This is shown in Experiment 1 for pairs of events and in Experiment 2 for triads of events with one deviant member. When the deviant statistic is the most recent one, it will in most cases be singled out for explanation. When it comes first, it is rarely explained, but accepted as a given fact. This is seen as an instance of the temporal order effect, where the first event in a pair or a series is taken for granted, whereas later events are considered more ‘mutable’, i.e. they could have been different. In line with this, later statistics are considered to be more in need of support from additional information than earlier statistics (Experiment 3). The focus on temporally later events can be distinguished from other primacy and recency effects by being due to chronological order, rather than order of presentation. Copyright © 2004 John Wiley & Sons, Ltd.