Disambiguating Complexity: From CAF to CAFIC: A Commentary on “Complexity and Difficulty in Second Language Acquisition: A Theoretical and Methodological Overview”

In their position paper “Complexity and Difficulty in Second Language Acquisition: A Theoretical and Methodological Overview,” the authors propose to reserve the term complexity for absolute structural complexity and difficulty for other measures typically labelled as “complexity” in the complexity–accuracy–fluency (CAF) triad. We agree with the authors that there are excellent arguments for making these distinctions and that we should avoid “conceptual ambiguity of the key notions and the proliferation of measures, with little attention to issues such as their construct validity and redundancy.” At the end, the authors suggest adding two new constructs to the triad: difficulty (the ability to comprehend and produce difficult linguistic structures) and appropriateness (the ability to choose from within one's repertoire the alternatives that are most adequate for a given communicative context), so that the acronym may become CAFDA. We agree that it might be useful to add new constructs; however, not only are difficulty and adequacy difficult to operationalize in terms of quantifiable linguistic features, but, more importantly, they do not do justice to many new useful measures that have emerged over the last decades and that need to be subsumed under separate constructs. At a time when structural linguistic views of language were the default, the CAF paradigm was originally intended as a way of understanding interaction among the constructs in psycholinguistic research (Skehan, 1998). With the main focus on structural properties of language (rather than its appropriate and effective use in context), the CAF constructs were operationalized as measures of linguistic features that can be objectively quantified. However, as the authors themselves point out, “language development also occurs along dimensions that are neither complexity nor difficulty (nor do they concern accuracy or fluency), such as appropriateness and adequacy.” As far as adequacy is concerned, the functional adequacy construct proposed by Kuiken and Vedder (2017) is very useful in writing research in that it focuses on the most important aspects of effective communication, namely content, organization, and coherence as related to the task. However, it is a holistic measure assessed through human judgment that can be used as point of reference for quantifiable measures, but it cannot be directly quantified and is therefore not a typical CAF construct. On the other hand, appropriateness could be related to two important aspects of spoken or written texts that have quantifiable features: (a) linguistic constructions and multiword constructions related to idiomatic use, and (b) discourse features related to achieving coherence. Therefore, we would like to propose extending the CAF acronym to CAFIC (mainly because this is easy to pronounce), in which the “I” stands for “idiomaticity” and the “C” for “coherence.” Appropriateness in terms of conventionalized use is reflected in terms of two major strands of studies, one in formulaicity and the other one based on a usage-based linguistic view that has emerged in our field with quantifiable measures such as verb–argument constructions (VACs). These measures gauge to what extent the text has features similar to those employed conventionally by an expert target language user. This is very much in line with expressions such as “native-like selection” and “idiomatic way” (Pawley and Syder, 1983), and was operationalized in a recent article (Köylü et al., 2024) as holistic formulaicity, defined as targetlike language use of intensifiers, fillers, multiword sequences, lexical constructions, verb–argument constructions, pragmatic and discourse features, and so on. As all these features interact and also have to be used appropriately, the authors argued that the degree to which a text was “formulaic” was best judged by humans. However, specific features such as formulaic sequences and VACs are definitely quantifiable. These linguistic features are not necessarily structurally complex, more sophisticated, nor more difficult to learn, but they are measures related to conventionalized—idiomatic—use in the target language community. Because such features do not clearly fit into the CAF paradigm, we propose to add the construct idiomaticity (rather than “formulaicity,” which would create a sequence of “F”s in the acronym). Appropriateness may also be tied to linguistic features that are related to coherence. Coherence, like functional adequacy and holistic formulaicity, is quite subjective and depends on the reader's interpretation and background knowledge and on the context in which the text is read, so it cannot be easily quantified. However, subsumed under coherence is cohesion, which refers to the grammatical and lexical connections between sentences and parts of a text, such as the use of conjunctions, pronouns, and lexical ties. These elements can be identified and measured, making cohesion quantifiable, and some of these indices have been shown to grow along with language development (Crossley et al., 2017). Thus, we propose to add the construct coherence to the CAF paradigm to subsume all indices related to affecting text flow. In conclusion, we are grateful to the authors for pointing out the ambiguity of several key notions related to complexity, and these definitely stand, and we do need new constructs. We suggest adding idiomaticity and coherence to encompass a large number of new useful measures in our field of research that clearly are not related to complexity nor necessarily to difficulty. If our colleagues agree, we would gladly work with them to develop a more precise delineation of the new constructs.