There are many issues which need to be studied in regards to shape-based classifiers and the schematizations they enforce on native or near-native speakers of a classifier language. We are also interested in the study of classifiers from a practical point of view. Some directions for future research are enumerated below.
In this paper, we have not speculated how exactly ideal meanings or usages of classifiers are adapted to specific objects. What kinds of schematizations apply to specific objects? Can these be represented as well-specified geometric functions? In other words, given an arbitrary object, how can we decide (computationally or otherwise) which classifier to use based on the physical properties of the objects.
A question that needs to be answered pertains to how children learn the proper use of shape-based classifiers? Is it based on exhaustive examples? Most probably not. How do adults who learn Assamese as a foreign language learn to use the appropriate classifiers? It is well-known that the proper use of Assamese classifiers is very difficult to acquire by non-native speakers just like in Burmese [Bur65] and Thai [Haa42]. It is frequently possible to pick out a non-native speaker who speaks Assamese fluently and without an accent when he/she uses wrong classifiers. It will also be interesting to study how a native speaker of Assamese, when he/she encounters a new object for the first time (e.g., objects newly introduced by industrial and manufacturing processes), decides on the appropriate classifier. It has been reported that native speakers usually demonstrate the ability to classify new objects consistently and easily based on observed characteristics in languages such as Burmese [Bur65], Navajo [EL63], and Dyirbal, an Australian language [Dix68]. There also have been some recent studies in the acquisition of classifiers by children, and native as well as non-native speakers [Hu93,Zha93].
Another question which may be of interest is why Assamese seems to be so complex with respect to the use of classifiers compared to other Indo-European languages, particularly Indo-Iranian and Dravidian languages of the Indian subcontinent? What do classifiers say about how native speakers of Assamese categorize objects in their mental representation based on shape? Is it different from speakers of other languages? It seems to be at least to some extent culture/language-dependent.
It is clear that shape perception and categorization of objects based on shape is play an important role in language understanding and usage. The case of nominal classifiers in Assamese makes it evident at a superficial and low level. It is ``superficial'' because a nominal phrase in Assamese is considered syntactically malformed without the proper classifier. It can be considered ``low-level'' because noun phrases are among the simplest of phrases that we usually find in a language. This leads us to conclude that representation of shape (at least schematization of an object's shape into a closed set of shapes) must be a part of a word's lexical meaning. In the case of English, researchers such as Pustejvosky [Pus91] have claimed that the representation of spatial and other characteristics may need to specified in the lexical entry for a noun for proper semantic and pragmatic processing. We claim here that for a language like Assamese, the representation of an (idealized) shape is required for syntactic processing as well.
Proper understanding of Assamese classifiers will help us construct translation programs from English to Assamese. In fact, this was one of the primary motivations which led us to the study reported here. Instead of storing the appropriate classifier list with each noun, we should ideally be able to use general geometry-based rules or functions to infer the correct classifier. Note that it is possible to use more than one classifier correctly with many words. What characterizes such situations? Which one is the most preferred classifier in such cases and why?
In summary, the research presented here is preliminary, but very interesting from the point of view of study of schematization which has been studied mostly in the context of English spatial prepositions such as in, on, at, behind, across and around [Tal83,Her86,Her95]. This study illustrates that shape schematization readily occurs in many natural languages in constructs other than the use of prepositions. We think the study of spatial classifiers and any other linguistic phenomena that are influenced by schematization of shape and related spatial characteristics will provide us with additional insights regarding the conceptualization of shape and space.