Constructing a typological database for inflectional morphology:
the SMG database for syncretism

Dr Dunstan Brown
Surrey Morphology Group
University of Surrey

Typological databases tend to register broad generalizations about languages (e.g.
word order). Inflectional morphology may not appear to be readily accessible to
database oriented study, because of its apparent idiosyncrasy. We discuss three
important issues for constructing a database for a morphological phenomenon,
syncretism: coverage, terminology and structure.

For coverage there is a trade-off between quality and quantity. Broad coverage
may mean that quality of analysis suffers, with no guarantee that similar phenomena
are being investigated cross-linguistically. Hence for inflectional phenomena, such as
syncretism, it is better to start with a smaller sample of diverse languages, but to be
sure that there is a great deal of detail. We have constructed a detailed relational
database of inflectional syncretism from 30 languages. The aim has been to enter
every type of syncretism found within each of the chosen languages.

The next issue for databases on inflectional morphology is that of terminology.
For example, in current theoretical linguistics, syncretism would be defined along the
lines of Matthews: "The relation between words which have different
morphosyntactic features but are identical in form." Furthermore, if one has an
empirical orientation, it should be required that a language is defined as syncretizing
features, when those features occur elsewhere in that language. Other uses of the term
'syncretism' treat a language as syncretizing features, precisely because they are not
present in that language. The latter definition would make it difficult to create a
database, as one would need a list of potential feature values a priori. Furthermore,
these are two definitions of quite different things. So it is important that any
typological database should come with a detailed definition. This explicitness
increases the databases value for other researchers, and is important for others who
may aggregrate information from a number of databases.

The structure of the database may reflect a particular theory for practical
reasons. There are a variety of theoretical approaches to syncretism:
underspecification, referrals (syncretism as a binary pair of
morphosyntactic combinations) and indexing (syncretism as a set of
morphosyntactic combinations).  Underspecification fails to encode all of
the information required for searching the database, but there may be
little to choose between the referrals and indexing approaches. The
structure of our database reflects the referrals approach. Hence, we have
developed a typological database where the phenomenon is clearly defined,
and the data structured such that a potential user knows what they are
getting.