In Old English, as in modern Dutch and German, there were a series of prefixes which were unstressed and phonologically constrained; some of them, because they determined the word-class of the derivatives they formed, were typologically unusual. If we trace these prefixes through into modern English, we find that they have lost ground. Partly they have been replaced by corresponding learned prefixes, partly they have simply become marginalised in the system of English. At the same time, if we look at those prefix-like items which are most productive today, we see that they carry their own stress, are phonologically unconstrained, and many of them are semantically much more lexeme-like. We can interpret these observations as a shift from a largely compounding Germanic basis through a long period of English history where prefixes were a norm, and with signs now starting to appear that a return to a more compound-oriented stage of the language is under way. In retrospect, we have no difficulty in explaining the various shifts of type that have occurred. What is interesting is the method by which the compound-orientation is being re-established, and the possible effect of typological pressures on such a shift. The more compound-oriented modern stage is being achieved not through any simple change, but through a conspiracy of different changes which have the combined effect of leaving erstwhile prefixal elements looking more like lexemes. The changes can be seen as being influenced by the pressures which give rise to the so-called suffixing preference across languages: replacing prefixes with lexemes increases the number of items to be recognised by the listener, but allows maximal use of word-initial cues.