Interlocutors construct sense by executing operations. There are different types of such operations:

1. Formal operations on signs are semiotic operations. The basic ones are selection and combination.
2. Operations of cognition and communication:
• Operations regarding the message as a whole and controlling its relationship to the interlocutors, specifically their shared intentionality, are illocutionary operations. Basic illocutions constitute declarative, directive and interrogative speech acts.
• Operations articulating a message work at two levels:
  • At the higher level, discourse is structured into topic and comment.
  • At the lower level, a proposition is composed by reference and predication.

First utterances are holophrastic, i.e. they consist of one sign (Mithen 2006). Such an utterance can already constitute a speech act.

• Examples of utterances with directive illocutionary force are ‘Come!’, ‘Go away!’, ‘Let's go!’, ‘Look!’
• Examples of utterances with declarative illocutionary force are ‘There is fire!’, ‘There is a bear!’

Since such utterances consist, ex hypothesi, of one sign, none of the operations applied is yet one of a language system. In fact, the only operations applied are illocutionary operations. These are signalled by gestures and prosody and for the rest result from the speech situation. At more recent stages of the evolution of languages, basic illocutions are converted into grammatical construction schemata in the form of sentence types. They may, however, continue to be signalled by prosody and non-linguistic cues down to contemporary languages.

Assume the inventory contains the holophrastic utterances listed in the table. The beginnings of grammar appear when two such utterances are combined into a complex utterance.

Now we can compose some elementary messages. The first message can already have a discourse structure, with sign_1 as topic and sign_2 as comment. At the same time, sign_1 and sign_2 become words meaning ‘there’ and ‘bear’. An analogous topic-comment structure underlies the third message. This presupposes a further step: From utterances provided by an illocutionary force – here, sign_4 –, the notions that compose it may be abstracted and used in utterances with a different illocutionary force – here, a declarative utterance.

Each of the three messages consists of a pair of utterances connected by a semantic relation. In a next step, such a pair is welded into one complex utterance. The meaning of such an utterance is now a proposition involving the propositional operations of reference (‘the bear’) and predication (‘look’ or ‘is coming’). Combinations brought about by propositional operations are at the basis of syntax.

Sign_4 is selected from a set comprising ‘come’ and ‘go’ and combined with sign_2 into one proposition articulated in a referential and a predicative expression. A certain set of signs is categorized for use for reference, while another set is categorized for use for predication. This founds the categories of noun and verb.

These are the beginnings of grammar. Grammar is that part of the language system which constrains the selection and combination of signs. The selection is constrained by organizing the inventory of signs in categories. The combination is constrained by providing a set of construction schemas in which signs are fitted. The function of such constraints is to allow an analytic access to complex messages and to introduce redundancy by rendering their structure more predictable.

The two basic functional classes of language signs are lexical and grammatical items. While the primary function of the former is to contribute meaning to the message, the primary function of the latter is to render its structure transparent, as a kind of signposts. These become increasingly useful as messages get more complex. For instance, as long as a verb only takes an actor and an undergoer, it suffices to have these precede and follow the verb in order to distinguish them ().

If the verb has a third direct dependent, as in , this may be marked by a grammatical formative like to in English. Such formatives are created by grammaticalization.

The development of grammar on the primal basis is a build-up of complexity (Dahl 2004). Further grammatical constructions and formatives are created by grammaticalization. Those which bear a markedness relation to each other are introduced in obeyance to the principle of unilateral foundation.