The presence of investment cycles demonstrates the long-run policy of firms investing in particular periods (investment spikes) with lower or zero investment levels in between, which contradicts the smooth pattern predicted by a convex adjustment model. This paper investigates the spells between investment spikes in a discrete-time proportional hazard framework to estimate the probability of observing lumpy investment and factors underlying lumpy and intermittent patterns of investment. Duration models were estimated on two datasets: on an unbalanced panel and on average data of 10 'firm size' groups of Dutch greenhouse firms over the period 1975¿1999. Two specifications of the model were estimated: one includes only theoretically grounded variables, and the other specification is extended by empirically grounded variables. Theoretically based models can explain the occurrence of investment spikes. Both specifications of model show an investment cycle of six years. This is also confirmed for the average firm, which exhibits a higher hazard ratio in the 6th, 12¿13th and 21st years of duration.