One of the phenomena accompanying a grammaticalization process is the expansion of the contexts in which a grammatical formative is used. This was first observed in Kuryłowicz 1965 and then, under the term ‘host-class expansion’, elevated to the most important factor in grammaticalization in Himmelmann 2004, § 4.

Here are some examples of the expansion accompanying grammaticalization:

• When the verb have is grammaticalized to a perfect auxiliary, then due to its origin as a transitive verb, it first only combines with transitive verbs. Only in a second step does its distribution expand to include intransitive verbs, too.
• When be going to V starts being grammaticalized, V is an intentional motion. Later on, the periphrase is used with verbs that do not designate a motion, like write, and when V is not intentional, like be born. Finally, V can even be stative, like be difficult.
• When the Old English demonstrative determiner is grammaticalized to a definite article, it first only occurs in anaphora. Secondarily, it combines with other nominals whose referent the hearer can identify. Their heads are initially concrete individual common nouns. With advancing grammaticalization, the article becomes possible with nouns that do not form a plural, such as earth. It is also found with certain proper names like the Himalayas. In other Germanic languages, this expansion goes on, so that the definite article is also compatible with abstract nouns and with anthroponyms (colloquial German der Peter, lit. “the Peter”).
• The Spanish indefinite article un first only combines with singular nouns, in a second phase also with plural nouns.

Expansion is an aspect of an ongoing process of grammaticalization, esp. in its initial phases. That is, it is not a static property of a sign or construction with a given degree of grammaticality. This is why it does not have a place in the system of six parameters which measure the degree of grammaticality of a linguistic sign. Instead, if a more and a less grammaticalized variant of a linguistic sign are compared, the former presents a more advanced degree of expansion, which means it is found in more different contexts. This is essentially captured by its advanced desemanticization (as already remarked by Kuryłowicz 1965).

The less specific the meaning of a sign, the more freely it combines with other signs. This principle is valid for just any sign, including lexical items. As a simple lexical example, consider the two verbs deform and melt. Both designate, in their transitive use, an act where an actor affects an undergoer by changing its form. Among other things, one can deform a piece of butter, a ball, a face and a statue. But one can only melt a piece of butter, not the other things. And we are not here talking about physical possibilities, but about possible combinations in a clause. The lesser combinatory potential of melt if compared with deform is directly related to its more specific meaning, including importantly its selection restrictions. The same goes for the grammaticalized items reviewed above.

There is a frequent misunderstanding concerning ‘expansion’ which is the reason why Himmelmann retermed it ‘host class expansion’: Expansion is not simply increased frequency of a linguistic sign in discourse. Some semantic classes of words, especially evaluatives like cool or discourse particles like exactly, are constantly subject to drastically increased frequency when they become fashionable and to complete oblivion when the fashion is over. These processes have nothing to do with grammaticalization. Expansion of a grammaticalized item is an increase in the set of distinct contexts in which it appears.