We have measured the frequency of Tn9 transposition and cointegrate formation in several different ways and have examined the stability of the cointegrates. We have also physically analyzed the structure of 40 independently derived cointegrate molecules. We present evidence here that Tn9, unlike the transposable element Tn3, does not transpose via an obligate cointegrate intermediate. We suggest that transposition of Tn9 leads to two, mutually exclusive, end-products: either direct insertion of the element into a recipient replicon (transposition), or fusion between donor and recipient replicons (cointegrate formation). This conclusion is based on our observations that, while Tn9-mediated cointegrates are very stable, they are formed at a rate lower than the transposition frequency. This finding is discussed in terms of current models for transposition. We also present evidence that clearly demonstrates the compound nature of Tn9. We find that the individual flanking IS1 elements are more active than the entire Tn9 transposon in cointegrate formation. In addition, we find that one IS1 element that is proximal to the cam gene promoter, is more active than the other, and suggest that the difference in activity might be due to differences in nucleotide sequence at their extremities.