Cagliari, Italy  15-18 September 2003

ALT V Conference

Michael Darnell & Michael Noonan

 An Investigation of Referential Density in the Languages of the World.
noonan@csd.uwm.edu

 

It has long been recognized that languages differ in the percentage of possible arguments that, in continuous text, are given overt expression versus those that are not.  The percentage of overtly expressed arguments to possible [i.e. notional] arguments is referred to as 'referential density' [RD].  We know, for example, that languages long resident in the northwestern littoral of Eurasia have high RD values, whereas languages in East Asia, in particular languages in the Sino-Tibetan family, have low RD values.

   Our research group is embarking on a project for which we will investigate differences in referential density in the languages of the world.  The goals of the project are to create a map of referential density values for the speech areas and language families of the world and to test hypotheses concerning the relation of RD values to other aspects of syntactic and rhetorical structure and hypotheses concerning RD as an areal and genetic feature.

   In our paper for the ALT conference, we will report on the results of a pilot study for the larger project.  The pilot study has three broad aims:

 1. to develop a methodology for assessing RD across a typologically

diverse set of languages,

2. to develop a set of hypotheses concerning the relation of RD values to other aspects of syntactic and rhetorical structure,

3. and to determine, to the extent our limited data will allow, whether speech area or language family is a stronger predictor of RD values.

In our paper, we will report on our progress in achieving these aims.

   In developing a methodology for assessing RD across a typologically diverse set of languages, we have been developing a set of protocols that can be applied consistently by trained personnel to any narrative discourse in any language.  For the larger project, we will attempt to apply this methodology to a consistent discourse frame, using either 'Frog Story' narratives (Berman & Slobin 1994; Slobin 1996) or Pear Story narratives (Chafe 1980) - more likely the former since Frog Stories are easier to elicit.  For the pilot, however, we do not have access to a sufficiently large and diverse body of glossed Frog Story or Pear Story narratives, nor do we have the resources to elicit either sort of narrative from an appropriate sample of languages.  We must, therefore, utilize existing narrative discourses that accompany published grammars for the bulk of our sample; where glossed Frog Story or Pear Story narratives are available, we will use them.

   In developing a set of hypotheses concerning the relation between RD values and other aspects of syntactic and rhetorical structure, we will compare RD values with the values for a set of syntactic and rhetorical features collected for each language in the sample.  Bickel (ms) has already discussed a number of such hypotheses.  We will attempt to expand on his work by utilizing an expanded database of features and languages.

   Finally, we will report on our preliminary findings concerning RD

as an areal feature.  We will attempt to determine, for example, whether RD values are relatively consistent within speech areas and whether, or to what degree, RD values are subject to influence across genetic lines.