In all the examples that we have examined so far, the whole of each component root (or base) is reproduced in the compound. Sporadically, however, we encounter a kind of compound where at least one component is reproduced only partially. These are known as blends. A straightforward example is smog, blended from smoke and fog; a more elaborate one is chortle (first used by Lewis Carroll in Through the Looking Glass), blended from chuckle and snort.

Examples of partial blends, where only one component is truncated, are talkathon (from talk plus marathon) and cheeseburger (from cheese plus hamburger). The ready acceptance of cheeseburger and similar blends such as beefburger and vegeburger may have been encouraged by a feeling that hamburger is a compound whose first element is ham – scarcely appropriate semantically, since the meat in a hamburger (originally a kind of meat pattie from Hamburg) is beef.

The most extreme kind of truncation that a component of a blend can undergo is reduction to just one sound (or letter), usually the first. Blends made up of initial letters are known as acronyms, of which well-known examples are NATO (for North Atlantic Treaty Organisation), ANZAC (for Australian and New Zealand Army Corps), RAM (for random access memory), SCSI (pronounced scuzzy, from small computer systems interface), and AIDS (from acquired immune deficiency syndrome). Intermediate between an acronym and a blend is sonar (from sound navigation and ranging).

The use of capital letters in the spelling of some of these words reflects the fact that speakers are aware of their acronym status. It does not follow that any string of capital letters represents an acronym. If the conventional way of reading the string is by pronouncing the name of each letter in turn, as with USA and RP (standing for the ‘Received Pronunciation’ of British English), then it is not an acronym but an abbreviation.

It is clear from these examples that blending and acronymy are in active use for the creation of new vocabulary. However, they differ from derivational affixation and normal compounding in being more or less self-conscious, and are concentrated in areas where the demand for new noun vocabulary is greatest, such as (currently) information technology.