Some recent papers on dynamic investments (e.g. Pindyck (1988), Kort (1990)) adopt the very strong assumption that investments are completely irreversible. Here we relax this assumption by allowing the firm to disinvest. However, every time it disinvests the firm has to pay transaction costs which are linearly dependent on the size of disinvestment. This implies that, despite of dropping the irreversibility assumption, the problem still has a really dynamic structure: when taking its investment decision now the firm has to reckon with the fact that disinvesting later on is costly. It turns out that, depending on the size of the cash balance and the amount of assets, one out of five different policies is optimal and that disinvestments can be optimal for reasons of cash shortage and size reduction when the firm has grown beyond some optimal levels of capitalization.