Second edition

§ 2. Principles of Morphemic
Analysis. Types of Word
Segmentability

As far as the complexity of the morphemic structure of the word is concerned all English words fall into two large classes. To Сlass I belong segmentable words, i.e. those allowing of segmentation into morphemes, e.g. agreement, information, fearless, quickly, door-handle, etc. To Сlass II belong non-segmentable words, i.e. those not allowing of such segmentation, e.g. house, girl, woman, husband, etc.
The operation of breaking a segmentable word into the constituent morphemes is referred to in present-day linguistic literature as the
89
analysis of word-structure on the morphemic level. The morphemic analysis aims at splitting a segmentable word into its constituent morphemes — the basic units at this level of word-structure analysis — and at determining their number and types. The degree of morphemic segment-ability is not the same for different words.
Three types of morphemic segmentability of words are distinguished: complete, conditional and defective.¹
Complete segmentability is characteristic of a great many words the morphemic structure of which is transparent enough, as their individual morphemes clearly stand out within the word lending themselves easily to isolation.
As can be easily seen from the examples analysed above, the transparent morphemic structure of a segmentable word is conditioned by the fact that its constituent morphemes recur with the same meaning in a number of other words. There are, however, numerous words in the English vocabulary the morphemic structure of which is not so transparent and easy to establish as in the cases mentioned above.
Conditional morphemic segmentability characterises words whose segmentation into the constituent morphemes is doubtful for semantic reasons. In words like retain, contain, detain or receive, deceive, conceive, perceive the sound-clusters [ri-], [di-], [кэn-] seem, on the one hand, to be singled out quite easily due to their recurrence in a number of words, on the other hand, they undoubtedly have nothing in common with the phonetically identical morphemes re-, de- as found in words like rewrite, re-organise, deorganise, decode; neither the sound-clusters [ri-] or [di-] nor the [-tern] or [-si:v] possess any lexical or functional meaning of their own. The type of meaning that can be ascribed to them is only a differential and a certain distributional meaning: ² the [ri-] distinguishes retain from detain and the [-tern] distinguishes retain from receive, whereas their order and arrangement point to the status of the re-, de-, con-, per- as different from that of the -tain and -ceive within the structure of the words. The morphemes making up words of conditional segmentability thus differ from morphemes making up words of complete segmentability in that the former do not rise to the full status of morphemes for semantic reasons and that is why a special term is applied to them in linguistic literature: such morphemes are called pseudo-morphemes or quasi-morphemes. It should be mentioned that there is no unanimity on the question and there are two different approaches to the problem. Those linguists who recognise pseudo-morphemes, i.e. consider it sufficient for a morpheme to have only a differential and distributional meaning to be isolated from a word regard words like retain, deceive, etc. as segmentable; those who deem it necessary for a morpheme to have some denotational meaning qualify them as non-segmentable words.
Defective morphemic segmentability is the property of words whose component morphemes seldom or never recur in other words. One
¹ The Russian terms are: живое, условное и дефектное морфологическое членение слов.
² See ‘Semasiology’, §§ 13-16, pp. 23-25.
90
of the component morphemes is a unique morpheme in the sense that it does not, as a rule, recur in a different linguistic environment.
A unique morpheme is isolated and understood as meaningful because the constituent morphemes display a more or less clear denotational meaning. There is no doubt that in the nouns streamlet, ringlet, leaflet, etc. the morpheme -let has the denotational meaning of diminutiveness and is combined with the morphemes stream-, ring-, leaf-, etc. each having a clear denotational meaning. Things are entirely different with the word hamlet. The morpheme -let retains the same meaning of diminutive-ness, but the sound-cluster [hæm] that is left after the isolation of the morpheme -let does not recur in any other English word with anything like the meaning it has in the word hamlet.¹ It is likewise evident that the denotational and the differential meaning of [hæm] which distinguishes hamlet from streamlet, ringlet, etc. is upheld by the denotational meaning of -let. The same is exemplified by the word pocket which may seem at first sight non-segmentable. However, comparison with such words as locket, hogget, lionet, cellaret, etc. leads one to the isolation of the morpheme -et having a diminutive meaning, the more so that the morphemes lock-, hog-, lion-, cellar-, etc. recur in other words (cf. lock, locky; hog, hoggery; lion, lioness; cellar, cellarage). At the same time the isolation of the morpheme -et leaves in the word pocket the sound-cluster [роk] that does not occur in any other word of Modern English but obviously has a status of a morpheme with a denotational meaning as it is the lexical nucleus of the word. The morpheme [роk] clearly carries a differential and distributional meaning as it distinguishes pocket from the words mentioned above and thus must be qualified as a unique morpheme.
The morphemic analysis of words like cranberry, gooseberry, strawberry shows that they also possess defective morphemic segmentability: the morphemes cran-, goose-, straw- are unique morphemes.
The oppositions that the different types of morphemic segmentability are involved in hardly require any comments with the exception of complete and conditional segmentability versus defective segmentability. This opposition is based on the ability of the constituent morphemes to occur in a unique or a non-unique environment. In the former case the linguist deals with defective, in the latter with complete and conditional segmentability. The distinction between complete and conditional segmentability is based on semantic features of morphemes proper and pseudo-morphemes.
Thus on the level of morphemic analysis the linguist has to operate with two types of elementary units, namely full morphemes and pseudo-(quasi-)morphemes. It is only full morphemes that are genuine structural elements of the language system so that the linguist must primarily focus his attention on words of complete morphemic segmentability. On the other hand, a considerable percentage of words of conditional and
¹ Needless to say that the noun ham denoting ‘a smoked and salted upper part of a pig’s leg’ is irrelevant to the ham- in hamlet.