In his seminal dissertation, Henson (1996) identified a number of constraints on theories of serial order memory. Two constraints, the fill-in constraint, in which an item that is erroneously recalled early is likely to be followed by its predecessor rather than its successor (recall of ACB is more likely than ACD), and the protrusion constraint, in which prior list intrusions are likely to be recalled in the same output position as their previous serial position, were considered evidence against chaining theories. We present results from two experiments which investigate the extent to which these effects are dependent on experimental methodology. When participants are given an open set of items, an equal ratio of fill-in and in-fill errors was observed and a protrusion effect was obtained. However, when a reconstruction of order task was used, a fill-in effect was observed. Implications for theories of serial order memory are discussed.