Temporal grouping can provide a principled explanation for changes in the serial position curves and output orders that occur with increasing list length in immediate free recall (IFR) and immediate serial recall (ISR). To test these claims, we examined the effects of temporal grouping on the order of recall in IFR and ISR of lists of between one and 12 words. Consistent with prior research, there were significant effects of temporal grouping in the ISR task with mid-length lists using serial recall scoring, and no overall grouping advantage in the IFR task with longer list lengths using free recall scoring. In all conditions, there was a general tendency to initiate recall with either the first list item or with one of the last four items, and then to recall in a forward serial order. In the grouped IFR conditions, when participants started with one of the last four words, there were particularly heightened tendencies to initiate recall with the first item of the most recent group. Moreover, there was an increased degree of forward-ordered transitions within groups than across groups in IFR. These findings are broadly consistent with Farrell's model, in which lists of items in immediate memory are parsed into distinct groups and participants initiate recall with the first item of a chosen cluster, but also highlight shortcomings of that model. The data support the claim that grouping may offer an important element in the theoretical integration of IFR and ISR.