The opposite shift, viz. the downgrading of a controlled action to an automatic process, is automation. It is much more frequent and important in human life than the gain of control over an automatic process, since it is essentially involved in the learning of a skill. Consider such examples as riding a bicycle, driving a car, reading, writing and, on top of the latter, type-writing. All of these and many other skills are learnt by repeated practice. In the beginning, the learner is aware not only of the task, but also of each component operation that he needs to execute, controlling every single step. By and by, he coordinates component processes into complex schemata which he acquires as wholes. He integrates the single steps into programs which run by themselves, so that they no longer require individual attention. In the course, monitoring decreases, less and less intellectual effort needs to be summoned, performance becomes faster and errors fewer (Schneider 1985:475f).

For example, riding a bicycle involves the simultaneous execution of many different operations and processes. One has to pedal and, at the same time, keep one’s balance. For the incipient learner, these are two different operations which he has to control. Automating them not only implies mastering each of them, but also combining them into one complex action: one keeps the balance by pedaling. Automation is, thus, the choice method to achieve parallel processing.

Automation of some action depends on two conditions: First, it presupposes (Schneider 1999:63) that “there is a consistent mapping ... between the stimuli and responses at some stage of processing.” In other words, if a given kind of problem is reliably amenable to a uniform solution, the solution can be automated; otherwise, it will always require attentive processing. Second, there must be sufficiently frequent occasion to practice the problem-solution pair. It is important to note that neither of these two conditions is in itself sufficient; it is their combination which leads to automation.

Usage-based approaches to language have imported the concept of entrenchment into linguistics in order to account for the gradual fixing of a feature of the language system in the mind of the speaker.¹ This concept is neutral both to the kind of entity entrenched – it may be a linguistic unit (typically, a more or less schematic representation) or an operation – and to the nature and place in life of the learning process, viz. to primary language acquisition or linguistic change. On the account sketched in Schmid 2015, §4.2, entrenchment properly includes routinization (besides association and schematization). Here, routinization is the same as automation.

The top half and the bottom half of a teleonomic hierarchy as visualized in the preceding section can thus be assigned to two different modes of processing information, controlled vs. automatic processing. These differ in many respects, and in all of these they differ gradually. The two modes are confronted in the following table.²

In general, declarative knowledge is processed consciously, while procedural knowledge is processed automatically. Consequently, the acquisition of procedural knowledge involves automation, while the acquisition of declarative knowledge involves reflection.

Maybe the most far-reaching conclusion to be drawn from this is the following: Increasing automation and increasing formation of consciousness are the same evolutive process (Givón 1989:260): Processing of what is up to a point the highest level of the teleonomic hierarchy is automated at the next evolutionary step, and at the same time, consciousness is freed to reflect on this level from a meta-level.

¹ E.g. Tomasello 2003:300: “Entrenchment simply refers to the fact that when an organism does something in the same way successfully enough times, that way of doing it becomes habitual and it is very difficult for another way of doing that same thing to enter into the picture.” Some of the aspects mentioned recur in the table below.

² These properties of controlled vs. automatic processing have been ascertained in psychological research reported on in Schneider 1985, Levelt 1989 ch. 1.4 and Schneider & Chein 2003. The contrast between the two modes is summarized thus in Schneider & Chein 2003:554f: “Automaticity leads to fast, parallel, robust, low effort performance, but requires extended training, is difficult to control, and shows little memory modification. In contrast, controlled processing is slow, serial, effortful and brittle, but it allows rule-based processing to be rapidly acquired, can deal with variable bindings, can rapidly alter processing, can partially counter automatic processes, and speeds the development of automatic processing.” The set of properties was brought into a tabular form in Givón 1989, ch. 7, esp. 256f.