Forward serial recall is affected by a diverse range of phonological factors that are readily replicated and relatively well understood. In contrast with backward recall, these phonological effects are not consistently replicable in that some studies show that the effects are present and some show the effects are absent or severely attenuated. Moreover at the theoretical level there is no consensus about how participants perform backward recall. The current research was aimed at understanding the differences between forward and backward recall by using meta-analytic techniques on 16 previously published experiments that examined the effects of benchmark phonological factors on both forward and backward recall. In each of the studies, recall was decomposed into 2 components, the first 2 items output and the remaining later responses. A consistent pattern emerged in the data. Each effect was present in both the early and late output positions in forward recall. The effects were present in the late output positions in backward recall, but the effects were weaker than in forward recall. The phonological variables had little impact on early output in backward recall (with the exceptions of articulatory suppression). The presence of qualitative differences between forward and backward recall and quantitative differences between studies have implications for the theoretical understanding of direction of recall in immediate memory tasks.