A dynamic view of changes in learner grammars

Assumptions about the role of variation in learners’ structural development indicate that the acquisition of grammar rests largely on the interaction of information from different linguistic sub-domains. Notice that this implies not only a modular view of grammar in the sense outlined previously; it also involves learning processes regulating the information flow.

Crucially, modularity as implicit in UG theory involves not only the idea that a grammar is composed of autonomous sub-components but also that there are multiple interfaces. Because the information of the different linguistic levels cannot be reduced to each other, correlations across levels need to be established. The complex organisation of language systems bears a potential of change that has not received much attention. Instead, much attention has been paid to the identification of external elements, that is, the triggers that might serve as agents of change. For example, it has been argued that the acquisition of inflectional morphology, notably the inflectional ending -st (2SG), serves as the trigger for the implementation of the target German Verb-Second (V2) property (in German, finite verbs obligatorily appear in the second position in declarative main clauses, see section 4.1) (cf. Clahsen 1988). The acquisition of lexical complementers, in turn, would involve the projection of a new structural layer needed to accommodate embedded clauses. This notion of external triggers is confronted with the question of why the target-like parameter-setting is not immediately triggered if the relevant external information is part of the child’s input (Triggering Problem, cf. Borer & Wexler 1987; Lightfoot 1991). What is interesting about the notion of a trigger is that it implies not only that specific external data may affect the linguistic system, but also that there is a reaction of the system to such impetus. Therefore, in earlier work (Plaza-Pust 2000; 2008a) we proposed to revise the notion of triggering in line with current assumptions in the area of dynamic systems theory (DST), so as to embrace “ . . . the reaction of a given system to the introduction of new units able to multiply and take part in the system’s processes” (Prigogine & Stengers 1984: 189, our emphasis).

In order to conceive of how new units might affect a given system such as a learner language system, we suggested to consider the complex information flow that characterises the organisation of complex systems as conceived of in DST. From a dynamic systems perspective, the information flow is modelled by internal and external feedback processes (Briggs & Peat 1990; Ebeling 1991; Cramer 1993). Notice that feedback processes are elementary mechanisms in such different domains as ecology, society, and mathematics, and that they are at the heart of the relation between order and chaos. Minimal disturbances, as exemplified in the much cited butterfly effect (Lorenz 1972), may have major effects if they get amplified via feedback. From the theory of self-organising systems, we know that the conditions under which local changes may turn into global rearrangements are met when the system is far from the equilibrium (Prigogine & Stengers 1984; 1993). In fact, what appears on the surface as disorder, contains a high degree of implicit correlation. It is in such a state that the system becomes sensitive to the new information from “outside”.

The question of whether and how minimal disturbances may turn into more global changes in the development of grammars leads us to reconsider their modular organisation in terms of a complex information flow. From a dynamic systems perspective, we conceive of the information flow at the interfaces in terms of a complex interplay of internal and external feedback processes (Plaza-Pust 2008a). If we consider, additionally, the linguistic environment, the following three dimensions of interaction strike us in their similarity to the organisation of other open complex functional systems (see Figure 2.1):

• - the interaction of the system’s sub-components (i.e. a well-formed structure is the product of the interaction of the principles of the different grammatical sub-theories),
• - the interaction of a part with the whole (i.e. linguistic elements fulfil a function in relation to the overall structural context), and,
• - the interaction with the environment (i.e., the linguistic input).

Figure 2.1: Language input, feedback processes, and the modularity of grammar (Plaza-Pust 2008a: 255).

The dynamic view not only departs from linear cause-effect conceptions of change implicit in traditional learning theories. As the new information may not affect all the sub-parts of the system at the same time, coexistence and competition occur on the different levels of linguistic analysis (Karpf 1993; Tracy 1991). This means that variation in learner grammars is expected. Feedback processes in grammars may fulfil a regulating function so that new information might only have a local impact and retain a marked or residual status. Yet they also bear a potential for change, which is given by their ability to amplify new information (Plaza-Pust 2000; 2008a). Grammars, like other complex dynamic systems, may remain stable over long stretches of time. At the individual level, the apparent stability of mother tongue grammars in adults is commonly conceived of as a steady state of knowledge. But it is also true that they change over time. What is important for present purposes is that such changes do not derive completely different systems (cf. also Larsen-Freeman 1997).

In the development of grammars, a mirror-world of stability and instability can be observed in relation to parameter setting. The empirical evidence gathered in the course of the last years shows that the implementation of FCs and fixation of parametric properties takes time, as manifested in the form of precursor structures or the apparent coexistence of alternative grammatical options. Example (6), produced by an L2 learner of German (cf. Plaza-Pust 2000: 258), illustrates the alternation of target-deviant and target-like verb placement in embedded clauses, which reflects the availability of both a head-initial and a head-final IP.

From a dynamic perspective, the variation encountered would seem to result from multiple parametric options exerting their influence on the system at the same time. Grammars in such states appear to be “undecided” between different states of order or, to use a common dynamic term, attractors. Such unstable states have also been described as bifurcation regions in DST as a system’s transition through them may lead to the convergence toward a new order or attractor. An intriguing aspect about a dynamic system’s transition through bifurcation regions that has been described in DST is that these transitions may throw it into chaos or they might render it into a stable state (Prigogine & Stengers 1984: 206). Self-organising processes in the evolution of dynamic systems thus correspond to a “delicate interplay between chance and necessity” (Prigogine & Stengers 1984: 1976). The system’s bifurcations represent the milestones in the history of its development: it is here that the variety of possibilities of change is displayed, and it is here where a minimal influence will have its major effect (Briggs & Peat 1990: 213).

The evidence gathered in different developmental situations reveals that grammars go through unstable states at which crucial bifurcations take place (cf. Hohenberger 2002 for child L1 acquisition; Tracy 1994/5 for child bilingual L1 acquisition; Plaza-Pust 2000, 2008a for adult L2 acquisition and diachronic language change). In learner grammars, oscillations between elementary and more advanced structures (that is, old and new grammars), or between alternative grammatical options, are tied to structure-building processes, on the one hand, and, on the other hand, to the necessary specification of the associated properties of FCs. System-internal conflicts resulting from competing linguistic representations have been found to arise in different situations, as, for example, in the mapping across representation levels, in the merging or linking of syntactic trees or in the reconstruction of a derivational relationship such as a path of movement (cf. Tracy 1994/5: 147). To the extent that the eventual convergence toward the target grammar is preceded by such developmental crises it seems plausible to assume that the potential for change unfolds in these very conflict situations and that it is in unstable states that something new may emerge (i.e. FCs), where the (self-)organising principles of the system come into play.