Both facilitative and competitive interactions occur simultaneously among plants, and the net balance between them can vary over time. Despite this, recent model-fitting studies have found that negative interactions predominate. This suggests that more complex models may be necessary to uncover facilitation. Here we fitted models including seasonality, interannual variation, and time lags to survey data to test for patterns in positive and negative interactions among plants in a Michigan dry sand prairie. We hypothesized that interactions would be generally facilitative in this dry environment. Results indicate that most immediate (direct) interactions among dominant species are actually competitive, although interactions were more facilitative over the drier summer season. Interestingly, lagged density dependence was strong for all species in both seasons; it was positive for conspecific interactions and both positive and negative for heterospecific interactions. Observed lagged density dependence is likely due to effects from litter and/or past storage in rhizomes. Conspecific immediate and lagged interactions tended to be stronger than heterospecific interactions, suggesting that population dynamics in this community are driven mostly by conspecifics. Overall, the presence of strong lagged density dependence in this system suggests that it may be more widespread in plants than previously thought.