Categories in language descriptions and linguistic typology – Melbourne Linguistics in the Pub May 2017

Stefan Schnell (University of Melbourne) recaps last month’s Linguistics in the Pub (Melbourne)

Leading the discussion was Ana Krajinović (University of Melbourne / Humboldt University)


The relationship between language-specific descriptive-analytical categories and categories figuring in cross-language comparative studies, and in particular the nature of the latter, have been subject of intensive and recurrent debate over the years, most recently in a dedicated discussion at last year’s SLE conference in Naples, and a focused discussion in the last October issue of Linguistic Typology (Vol 20, issue 2, 2016). In this LiP session, we focused on the research-practical aspects of the issue at hand from a descriptive point of view, asking questions about how researchers go about in identifying relevant categories in the languages they describe, and how they capture and describe their functions and label the categories. But what criteria and concepts do researchers apply when going about these tasks? A notoriously difficult area is research into systems of tense-mood-aspect (TMA) which illustrate some of the points during our discussion.

Thus, on the one hand, linguists seeking to describe the grammatical system of a language, or a sub-system thereof, will resort to some conceptions of specific linguistic categories when confronted with a hitherto unknown linguistic system. While these conceptions are in principle multi-faceted, and associated with different traditions in language research, including philological traditions, most linguists today will have received considerable knowledge about linguistic typology as part of their training as linguists. Therefore, categories figuring in the typological literature will naturally be a point of departure for descriptivists.

On the other hand, however, individual language descriptions, and the description of particular language-specific categories are the essential input into linguistic typology. Hence, mere adoption of typologically identified categories can lead to tendencies in researchers to see systems under investigation for the first time as essentially conforming to what is known from other languages, and this tendency potentially prevents insights into linguistic system that work unlike any other systems encountered thus far. For linguistic typology, this bears the danger of a confirmation feedback loop, and an unwarranted highlighting of universality, at the cost of overlooking diversity.


To what extent should grammatical descriptions be explicit about their contribution to typology, and be typologically informed? And what does this mean?

The first problem mentioned is that of describing word classes. In work by Dixon (and Aikhenvald), we find statements that adjectives can be either nouns or verbs, and the like. But then what is the class of adjectives?

We should take formal criteria into account, as part of comparative concepts. See Wierzbicka’s conception of adjectives: a class of words that include words meaning ‘big’, ‘small’ etc. and that share formal properties distinct from those of nouns and verbs (also defined as classes with members like ‘person’, ‘thing’ etc.).

Hence, formal aspects of linguistic expressions are part of comparative concepts, but are left under-specified, as having same / different properties from another class.

In other areas, for instance TMA markers or demonstratives where functions are notoriously difficult to pin down, a descriptive solution can be to merely number categories, thus refraining from any functional-semantic label that is potentially misleading or descriptively inadequate. But this only addresses the problem of describing individual semantics of categories, not their identification, that is it leaves open the question about the number of categories, and whether there are subcategories, etc. Also, numbers are of course not informative, and can only be understood in the context of the specific linguistic system at hand, which hampers typological comparison.

Typologically informed can mean the following:

  1. considering typological literature, what is possible? design space of human languages
  2. appreciation of linguistic diversity, different from universalism in the generative tradition
  3. avoiding ideologies about language

A further dimension of the issue at hand is the existence of philological traditions in specific linguistic areas that often uses well-established terminology that is not necessarily typologically informed or informative. An example is terminology like “focus structure” in Philippinist linguistics, which does not translate straight-forwardly into more general conceptions of “focus”, and how this is structurally marked. Researchers need to be aware of the differences in usage, and clarify what they mean when they use such terms. As a general rule of thumb, it is useful to remind descriptivists that they need to be transparent in their use of terminology and identification of categories, and do not presume that everyone knows / understands them.

One can see this as a difference between global, large-scale and more areal typology, with the latter often resorting to traditional terminology. It may be wise to use this transitional terminology, depending on the audience of a descriptive grammar, where there are often well-established conceptions. A good example in this regard is the terminology used for possession in Oceanic languages, direct vs indirect. This is useful for the Oceanist audience, but it requires some comments for a wider audience with a typological, but no Oceanist background.

What role do more recent usage-based approaches play? How can they refine linguistic categorisation?

Example: corpus-based TMA project (MelaTAMP) at Humboldt-University in Berlin, where formal categories can be brought back to specific contexts of usage, which can – via the use of comparable corpora – serve as something like an “etic grid”, so that comparison can be undertaken with minimal presumptions about system properties of the categories involved.

What about phonetics? Here, the articulatory and acoustic space(s) should be ideal “etic grids” for comparison, but there are complicated issues here too, in particular problems of grouping and dissection in e.g. prosodic typology – traditional categories like “stress” are not tenable in the light of latest research. Although generally applicable analytic tools like the IPA chart are useful as a starting point, it is not very accurate. There has been a move towards finer-grained categories, and for instance stress research nowadays goes beyond simple questions of placement, e.g. what articulatory and acoustic features give rise to perception of “stress”? Are these perceptions universal? (Himmelmann’s more recent work, but also Janet Fletcher’s and other’s work over the years). There are also methodological problems: how do we ask people to identify elements that are stressed?

What is the impact of typological discussions over time on the content of descriptive grammars?

Descriptive linguists following various typological discussions will perceive certain themes as more relevant than others. This may shape their focus during language description, so that a particular phenomenon widely discussed in typology at a given point in time will also be reflected in grammars, and dealt with in some detail, whereas other structures and functional categories will receive less attention, although this is not necessarily warranted by the specificities of any given language under investigation. Also, what is typologically interesting is not necessarily relevant for a specific language. So descriptivists should be aware of the typological discussion, but be critical and selective.

What should be part of a descriptive grammar?

Aside from the core grammatical system, where certain semantic categories are treated – what about, for example, systems of spatial orientation – if they are not strictly speaking part of the grammatical system of language? Shouldn’t a grammar cover much more than just the grammaticalised system?

Different approaches: some recent approaches develop ideas of massive coverage of linguistic systems plus its use in different contexts, etc., essentially incorporating discourse structure and aspects of ethnography of communication and the like. This is not necessarily desirable, and others have advocated slim grammars that are a gateway into a language. On this view, a grammar is not necessarily a reservoir for everything about a language, and there are other forms of publication. Moreover, nowadays any descriptive grammar is tied to a language documentation project (whether part of a dedicated program in language documentation or not), and authors can always store material other than the grammar in their language archive, as part of the respective language documentation; this could also be a place for any information that does not fit into either the grammar or any academic paper.

But then where do typologists obtain basic information about language systems, including areas in semantic typology?

Grammars may be too detailed, and information is not necessarily easy to find. There were attempts to make grammars maximally typological, Smith & Comrie (1978) and the Lingua Descriptive Series. Imaginable would be an intermediate form of publication, short 3-5 page accounts of a structure X in a Language A, purely descriptive, minimising theoretical ideologies. This could also be a good place for subsystem description like spatial orientation. The problem is of course that this type of publication does not give credit to researchers, which will be a problem in particular for ECRs.

What is the role of theory in grammars?

If a grammar is not meant to be theory-driven, but papers are, what is the difference between audiences? Given that no linguistic work could ever be entirely theory-free, would it be imaginable to go in the other direction and have “theory-alternative grammars”? One could imagine five different grammars of the same language, each from a different theoretical point of view. Though this may not be practicable for entire linguistic systems, one could at least imagine doing this for subsystems, say accounts of e.g. complementation or control constructions from different models’ perspective, like MP, HPSG, LFG, RRG etc.

Leave a Reply