Desktop version

Home arrow Language & Literature

Paradigms and families

Distinguishing grammatical words from lexemes introduces many of the same kinds of issues that confront lexicographers when they attempt to distinguish primary ‘word entries’ from ‘word senses’. These issues bear in a direct way on the delineation of inflectional and derivational processes. From a classical WP perspective, inflectional processes are said to define (or, more generally, relate) forms of a lexeme (i.e., grammatical words), whereas derivation derives (or relates) lexemes. It is typically assumed that the grammatical words that comprise a lexeme must belong to the same word class, preserve a core lexical semantics and argument structure, and even share a set of ‘intrinsic’ features that are invariant for a given item. Inflectional processes are regarded as monotonic processes that add features to the invariant properties that can be associated with stems. In contrast, processes that alterclass or argument structureare treatedasderivational, sincethese changes create new lexemes.

Classical descriptive practices are not fully consistent with this split, as noted in recent works such as Haspelmath (1996) and Spencer (1999). In particular, conjuga- tional paradigms often include categorial hybrids, such as participles, which exhibit adjectival properties, and gerunds or masdars, which exhibit nominal properties. The inclusion of passive participles also introduces argument structure variation into conjugational paradigms. This inconsistency would appear to undermine the coherence of the notion ‘lexeme’, along with associated conceptions of the inflection/derivation divide, if, as these authors assume, the notions ‘lexeme’ and ‘inflectional paradigm’ are taken to coincide.

However, it is also possible to interpret classical descriptions more charitably, as inexplicit rather than inconsistent. Pursuing the line of analysis outlined above, lexemes can be reconstructed as sets that occupy a position between inflectional paradigms and morphological families, as in Figure 3.5.

This interpretation also resolves a systematic ambiguity in the use of the term ‘paradigm’. Nouns typically have a unified set of forms, inflected for case, number, definiteness, etc., which can be consolidated in a single declensional paradigm. Hence the lexeme mreza ‘network’ contains the single paradigm in Table 3.5, and there is no real harm in conflating paradigm with lexeme.

In contrast, the inflected forms of verbs often exhibit a sub-organization into sets of forms that share the same tense/aspect/mood features and vary in their agreement features. This is illustrated by the conventional classification of the

Lexical organization of grammatical words

Figure 3.5 Lexical organization of grammatical words

Table 3.5 Paradigm of Croatian mmza ‘network’ (Baric et al. 2005:157)
























Table 3.6 Synthetic forms of Croatian zeleti ‘want’ (Baric et al. 2005:257)































forms of zeleti ‘want’ in Table 3.6 into present, aorist or imperfect sets. The term ‘paradigm’ is applied to these smaller sets of forms in references to ‘the present indicative paradigm of zeleti’ but to the full set of inflected forms in references to ‘the (conjugational) paradigm of zeleti’. The distinction between paradigm and lexeme in Figure 3.5 provides a simple resolution. The term ‘paradigm’ is reserved for the first, smaller, set of forms, and ‘lexeme’ is applied to the complete set of forms in the paradigms. With this clarification, the ‘present indicative paradigm of zeleti’ is a subset of the ‘the lexeme zeleti’

A general characteristic feature of Slavic verbs is that they come in imperfec- tive/perfective pairs. Corresponding to the forms of imperfective zeleti there is a parallel series of forms of perfective pozeleti. The relation between imperfective and perfective verbs is standardly regarded as ‘word formation’ or, following Aronoff (1994) and Beard (1995), ‘lexeme formation’ Processes that relate distinct lexemes define the third member in the classification in Figure 3.5, the ‘morphological family’ In the present case, the morphological family of zeleti will contain the imperfective lexeme zeleti itself, the perfective lexeme pozeleti, and other lexemes derivationally related to zeleti.

This organization provides a structure within which to reconstruct standard classifications without inconsistency, and also without prejudging answers to the questions of analysis that arise. Reconciling grammatical descriptions with WP models and other ‘lexeme-based’ approaches involves determining, for any set of related forms, at what level they are related, whether internal to the paradigm, lexeme or family. Traditional grammatical descriptions tend to draw a primary split between inflectional processes, which relate grammatical words belonging to the same lexeme, from derivation, which relates distinct lexemes. Over time, the terms ‘inflection’ and ‘derivation’ have become overloaded with other associations and assumptions, to the point that accounts such as Haspelmath (1996) appear to identify inflectional with productive processes. Although it might be best to replace these terms altogether, it will suffice if inflection is understood as applying to ‘endocentric’, within-lexeme processes and derivation to ‘exocentric’ cross-lexeme processes.[1]

This formulation in turn requires a classification of processes based on the nature of the relations they establish rather than on their productivity or transparency. Usually, this classification is framed in terms of the properties shared by the members of a lexeme. Traditionally, inflectional processes relate forms that share a set of ‘intrinsic’ properties of a lexeme, whereas derivational processes relate forms that may have different intrinsic properties. As mentioned directly above, these shared properties are normally taken to include aspects of meaning, recurrent units of form, a set of ‘inherent’ features and a common or similar argument structure. There is nothing theory-specific about this choice of‘constant’ properties; essentially the same set of attributes are associated with a ‘lexical entry’ in a Post-Bloomfieldian approach.[2]

A coherent split between inflection and derivation can then be maintained by imposing a consistent classification on the problematic elements discussed by Haspelmath (1996) and Spencer (1999). Consider categorial hybrids, beginning with participles. One type of analysis avoids the problem of classifying verbal participles altogether by treating them not as forms of a paradigm, but as components of a larger periphrastic form. The idea that synthetic and periphrastic forms both realize the paradigm cells of a morphological system underlies the comprehensive descriptions of Curme (1922, 1935), and is developed in a more formal setting in Ackerman and Stump (2004). On this account, verbal participles are no more members of a paradigm than the stem of a synthetic form is.[3] Alternatively, active participles can be included in the lexeme of a verb if, as has often been suggested, participles are treated not as a distinct word class but as a word class that neutralizes the categorial distinction between verbs and adjectives. On this analysis, active participles can be included in the lexeme of a verb, provided that the lexeme is defined as a set of grammatical words with noncontrastive (rather than identical) values for word class and other intrinsic features. A similar analysis extends to other ‘hybrids’ that are analyzable as neutralizing rather than altering word class.

However, one might also want to reconsider the inflectional status of participles. Although participles are sometimes included in ‘paradigms’, they tend to be associated more with whole lexemes than with specific subsets of inflected forms. This association can also be expressed at the level of the morphological family. In languages where participles inflect exactly like adjectives, an analysis at the level of the family removes a source of morphological heterogeneity from verb lexemes. Active participles are thereby assimilated to ‘masdars’ and other varieties of derived nominals which, irrespective of their productivity, comprise separate lexemes in the morphological family of a verb. The regular and predictable members of this family can even be regarded as forming a kind of ‘derivational paradigm’, along the lines suggested in the treatment of ‘paradigmatic derivation’ in Austerlitz (1966) and Blevins (2001).

Table 3.7 Classification of Georgian screeves (Aronson 1990:462)








Present Indicative


Conjunctive Present



Future Indicative


Conjunctive Future



Aorist Indicative




Present perfect


The classification of passive participles and other forms depends on the treatment of argument structure. Passive forms can be included in the lexeme of a verb if the forms in a lexeme need only share a common logical (or thematic) argument structure. This assumption underlies standard descriptions of a language like Georgian that organize the ten paradigms of the language into three inflectional ‘series’, each with distinctive surface valence demands.[4]

If, however, the words that make up a lexeme are required to have common surface valence (subcategorization) demands, then active and passive words will belong to different lexemes but again to a common morphological family. On this alternative, Series I and II in Table 3.7 would belong to a different lexeme from Series III but fall within a common family. Either choice restores coherence to a standard perspective in which lexemes occupy a position between inflectional paradigms and morphological families.[5]

  • [1] Classical WP accounts do not distinguish processes that apply within single paradigms from thosethat apply across paradigms, though this distinction corresponds to the contrast between ‘contextual’and ‘inherent’ inflection in Booij (1993,1996).
  • [2] This correspondence reflects the fact that the Post-Bloomfieldian lexical entry is, from a classicalWP perspective, an abstraction over a set of grammatical words.
  • [3] On this analysis, the participial forms that do not occur as parts of complex tenses would, whereverpossible, be assigned to determinate verbal or adjectival categories.
  • [4] Briefly, in Series I the logical subject is nominative and the logical object is ‘dative’; in Series II thelogical subject is ‘narrative’ and the logical object is nominative; and in Series III the logical subject isdative and the logical indirect object is nominative. For discussion and analysis of these patterns, see,among others, Tschenkeli (1958), Harris (1981), Anderson (1986) and Blevins (2015b).
  • [5] On either alternative, ‘morphosemantic’ alternations (Sadler and Spencer 1998) like causativiza-tion will relate distinct lexemes within a larger morphological family.
< Prev   CONTENTS   Source   Next >

Related topics