Morphology and language documentation

<p>What does it mean to document the morphology of a language, and how does one go about such a task? Most of the world’s languages are arguably underdocumented, yet morphological generalizations often require large amounts of primary data: thousands of word forms could be needed to establish basic patterns of allomorphy, for example, or the structure of an inflection-class system. Because of this, the major debates in the language documentation literature affect the field of morphology by shaping the nature of the data. A starting point is the idea that traditional methods of elicitation, often via translation from a contact language and inevitably requiring a patient speaker, can mask ingrained assumptions about the ontology of data and the wider context of linguistic research. Critical examination of these assumptions yields a wider range of possible approaches that can be drawn on to produce a corpus theorization (i.e., a rationale for the types of communicative events to be recorded) appropriate to each language situation. In particular, it has been argued that it is sometimes not ethical to collect language data in a decontextualized way that prioritizes (or appears to prioritize) the linguist’s goals above speakers’ goals, where those are not the same. Thus, in morphology, where virtually everyone agrees that some type of elicitation is essential, creativity and flexibility are sometimes needed to address or modify research questions. Fortunately, documentary linguistics has seen significant advances in the theory and practice of data management, making it possible to work efficiently with data from a wide variety of recording-session structures.</p> <p>Of equal interest are the reasons why a decontextualized approach may be undesirable, even for the linguist’s analytical purposes. The goal of ‘documenting morphology’ is an abstract one; one can only really document word forms, and morphological structure is a product of analysis. From this fact arise a few problems. First, and even independently of the ethical issues referred to above, it is not always obvious what methods are most reliable for getting speakers to produce word forms or for understanding speakers’ knowledge about them. Different methods have complementary pros and cons, so it is usually necessary to use a mix. When working with existing data, an appreciation of the complexities of the data gathering process is useful for developing a critical approach to the background contexts, strengths, and limitations of primary sources. Second, ‘documentation’ implies a reasonable level of comprehensiveness. For many semantically or functionally defined phenomena, it is possible to make a cross-linguistically robust checklist that ensures that one has more or less covered the relevant territory. It is much less straightforward to compile an inventory of structures in any formal domain, particularly given cross-linguistic variation in morphological vs. syntactic vs. prosodic encoding of similar functional categories. In morphology, the linguist’s inventory of phenomena often keeps expanding until nearly all grammatical constructions and large numbers of lexical items have been encountered. Again, this challenge can be addressed by using a mix of methods and genres to check that one has a correct understanding of at least the most commonly occurring patterns. Spontaneous speech tends to contain constructions that fail to show up in elicitation for reasons like pragmatics or interference from the contact language, while structured elicitation or metalinguistic work is needed to fully investigate the word-formation patterns within each of those constructions, or indeed (if the linguist is nonnative) to get enough of a foothold to work with spontaneous speech at all. Checklists from the viewpoint of morphological typology tend to initially be most useful for monitoring and organizing, and later for filling gaps at a more advanced stage of research.</p>