Words with spatial associations (e.g., "bird") can hinder identification of an unrelated visual stimulus at the implied location (i.e., at the top of a display). This spatial interference effect has been demonstrated many times by several independent research groups (e.g., Estes, Verges, & Barsalou, 2008), and it fits within the theoretical framework of grounded cognition. Petrova et al. (in-press) reported a series of experiments that varied in similarity to Estes et al. (2008), successfully replicating the spatial interference effect only once, and concluding that the effect is unreliable. Here we report a comprehensive meta-analysis of 37 tests of the spatial interference effect. We show that the effect is moderately large and highly reliable among studies that evoke semantic processing of the linguistic cues and that assess interference at suitably short delays. In conclusion, the spatial interference effect occurs reliably with semantic processing, and varies systematically with temporal and linguistic constraints.