Having established some criteria for transitivity in minimal radicals there remain to be considered the problems posed by extensions to the radical. First, some preliminary remarks on the concept of extension are necessary.1 All radicals in Swahili may be characterized as either minimal or extended. A radical is extended if it may be contracted, e.g. -vunjik-> -vunji-, and minimal if it can not. Many radicals, however, look as though they are extended, e.g. -ambi-, - andik-, -lema-, -gaw-, -nunu-, -simam-, -tukush-, but in default of a shorter form in current use must be regarded as minimal. These I regard as pseudoextended2 radicals and their transitivity patterns have already been discussed. As Miss Bryan has pointed out, however, pseudo-extended radicals are of different kinds, and two major types may be distinguished; those in which the extensionlike element is commutable with other such elements, and those in which it is not. Thus:-
(a) Radicals which, historically at any rate, represent extended forms, e.g. -ambi-/-amb-
(b) Radicals which can plausibly be regarded as extensions, e.g. - andik-/-anda-; -koma-/-kom-
(c) Radicals which are only fortuitously similar to extended forms, e.g. - nunu-, -gaw-, -limuk-
In all cases the important fact is that the extension-like element will not commute with zero: -anik-/-anu-; -bandik-/-bandu-
Minimal radicals may be extended by one or more of a number of extensions, which in the grammars and dictionaries are usually treated as derivative affixes and labelled with such terms as ‘prepositional’, ‘stative’, ‘causative’, ‘reciprocal’ and the like. Such labels are useful provided that it is recognized that the semantic range of several of these is both wide and opaque. Indeed, the problems of identifying and describing a particular extension are often extremely difficult: if the criterion is morphological then the range of associated meanings is often impossibly wide: if the criterion is semantic then one is liable to have a proliferation of apparently similar morphological units. The problem becomes progressively more difficult as extensions are added; in a radical extended three or four times only rarely will it be possible to regard the meaning of the extended radical as a sum of the meanings of the constituent extensions-something which those trying to operate the language learn to their dismay. One answer might be to regard each extended form as a completely new unit, but while this is undoubtedly useful in some cases it is not generally helpful. While many of my informants would be quite happy to regard -oko-/-okot-and -fa-/-fany-as separate units, they would be quite unhappy to see -limik-and -limi-as anything other than extensions of -lim-.