According to Sinnemäki (2008, 2011), a common problem in defining and understanding complexity is how to measure complexity in a reliable manner. Since there is no unified definition among researchers in the sciences of complexity including linguistics, it can be difficult to understand complexity and the “question of complexity is bound to remain elusive” (Sinnemäki, 2008, 2011). Hence in studying and approaching linguistic complexity, there may be a few points to consider as well as a number of clarifications that should be made such as:

• Complexity is separated from difficulty (Dahl, 2004);
• Complexity is classified in different types (Rescher 1998);
• Local complexity is separated from global complexity (Miestamo 2008);
• Complexity is measured as the description length of an object’s structure (Gell-Mann 1995).

Firstly, Dahl (2004) claims that it is important to keep complexity separated from difficulty as different users experience a different level of difficulty, and perhaps also between different usage events of the same user.

The second point to note is that there is also an unexpected unity behind the different formulations for complexity. To explain this, Lloyd (2001) classifies complexity metrics under three broad types, whereas Rescher (1998) classifies it in only a few modes. This shows that many researchers do agree that there are different notions of complexity. While it is impossible to devise a “correct” metric, it is feasible to approach language complexity by classifying it into different categories.

The third point is that local complexity should be separated from global complexity. Local complexity is about the complexity of some part of an entity, whereas global complexity is about the overall complexity of that similar entity. It is therefore rather impractical to measure the overall complexity of a particular linguistic system, as it is quite impossible to devise a fully comprehensive description of the grammar of any single language (Miestamo 2008). It is, however, both possible and practical to measure the local complexity of a system for example, the complexity of the numeral system, the case system, or the verb’s argument structure (e.g. Sinnemäki 2011).

The last point is complexity is intuitively situated between order and disorder. To put it simply, the general intuition is that when a language has more “structural units/rules/representations” (Hawkins, 2009: 252) in its system, it would be more complex. Since there has been no unified metric system for complexity, the best way to capture this is to measure the length of description of an object’s structure (effective complexity, Gell-Mann 1995) instead of the length of description of the object itself (e.g. Kolmogorov complexity). This is because the latter is usually related to randomness with complexity, while the former, randomness with low complexity.

According to Sinnemäki (2008, 2011), he argued that while it is impossible to compare the overall complexity of one language to that of another, it is possible to compare complexity across languages when focusing on particular types of effective complexity in their local contexts.

 

Why separate complexity & difficulty?

As mentioned in the first point made by Sinnemäki, it is vital to treat complexity and difficulty as two separate concepts. There are a few reasons as to why one should separate complexity and difficulty, as explained below:

• Description and operation are two separate tasks;

These two tasks can be done independently of one another. Native speakers are naturally able to talk fluently. They often do this without thinking about language description. To some degree however, description is possible without fluency. For example, it is possible for a new learner of a language to be able speak the language and yet make minimal mistakes, even without being fluent in the language.

• The problem of finding a user-type neutral definition for complexity (Miestamo, 2008:24-29);

Different user-types (eg. speakers, hearers, 1st language acquirer, 2nd language learner) may have experienced different linguistic patterns, hence varying degrees of difficulty of using and processing a particular language. This in turn, can cause conflicting results in measuring complexity. A more objective and theory-based perspective should therefore be adopted in approaching complexity to avoid conflicting results in the measurement of complexity.

• To avoid the problems in the evaluation measure of early generative grammar (Chomsky, 1965; Chomsky & Halle, 1968);

Researchers have long struggled with the evaluation measure of early generative grammar. This evaluation measure was used in child language acquisition as it was assumed that the framework that had a more brief description of the system would therefore allow a more accurate link to language acquisition. What researchers failed to realise was that “the shortest description was not necessarily the most plausible one psychologically” (Kiparsky 1968). It is thus important to separate complexity and difficulty in order to avoid the problems that were faced by researchers in the past.

• Possible to determine independently the processing responses of the different types of complexity (Hawkins, 2004 & 2009).

Separating complexity and difficulty could show the varying degrees of processing responses of the different types of complexities.