Research Paper I-Derivational Paradigm in Word Formation

Text-only Preview

Research Paper I

Henry Beecher
May 14, 2004

Derivational Paradigm in Word Formation

Main Reader: Farrell Ackerman

Ancillary Reader: Sharon Rose
Ancillary Reader: Chris Barker


The morphological paradigm as a theoretical construct has a central role in explaining inflectional
word formation, particularly in Word & Paradigm approaches to morphology. In contrast, research
into the role of paradigm in derivation has been programmatic as well as fragmentary. In this paper
I investigate the extent to which patterns of relationships among derived words constitute a
derivational paradigm. A more comprehensive and precise characterization of derivational paradigm
is provided based on a unified treatment of inflection and derivation. Previous diachronic-oriented
research (van Marle, 1984; Pounder, 2000) is complemented by a dynamic model I propose for
representing the role of derivational paradigms in synchronic word-formation. The proposed model
captures local vs. non-local derivational associations as identified via what I refer to as the
Paradigmatic Derivate Generalization (PDG). Using the model and a generalized notion of
paradigm derived from the rule-patterns of Bochner (1993), a variety of derivational processes in
English, French, Tigre and Lithuanian are analyzed to identify the paradigmatic relationships upon
which word formation in those languages is dependent. The suitability of the model for
implementation using word-formation rules is also demonstrated through an analysis of a fragment
of the Lithuanian data using a novel adaptation of Paradigm Function Morphology (Stump, 2001).
While inflectional and derivational paradigms are not parallel in function and may differ in their
range of applicability, they do share many of the same characteristics. Furthermore, these findings
indicate that generalizing the notion of paradigm as a theoretical construct to derivational word
formation is supported by cross-linguistic evidence.

Derivational Paradigm in Word Formation*


In the domain of morphology the regularities associated with the participation of words in
paradigms has a long and venerable tradition originating with classical Latin, Greek and Sanskrit
grammarians. The concept of paradigm is generally restricted to representing patterns of
relationships among inflected words such as depicted for the French verb finir ‘to finish’ in figure
(1)a. In contrast, patterns of relationships among derived words like the French examples in 1(b)
are generally not considered to be a paradigm.

a. Present forms of finir ‘to finish’
b. Adj (m) / Adj (f) / Adverb ‘gloss’ .

Je finis
Nous finissons
beau / belle / bellement ‘beautifully’

Tu finis
Vous finissez
certain / certaine / certainement ‘certainly’

Il finit
Ils finissent

franc / franche / franchement ‘frankly’

(1) Inflectional vs. derivational paradigm.

It is my claim that the concept of an inflectional paradigm is deservedly generalizable to
derivational morphology; that is, patterns of relationships among derived words constitute a
derivational paradigm. While inflectional and derivational paradigms are not parallel in function, I
show how both share many of the same characteristics. Based on a unified treatment of inflection
and derivation, I formulate a more comprehensive and precise characterization for derivational
paradigm as well as a model appropriate for the requirements of synchronic word-formation and
suitable for implementation using word-formation rules (WFR).

Recognizing a central role for paradigm in lexeme formation, Hockett (1954) coined the
label ‘Word and Paradigm’ (WP) to distinguish word-based morphology from morpheme-based
‘Item and Process’ (IP) or ‘Item and Arrangement’ (IA) models. In WP as further elaborated by

* To Jan, my second mom, and the only person I know of Lithuanian descent. I am also indebted to all in the
Linguistics Dept at UCSD for their valuable comments and assistance. Any errors or inaccuracies remain my own.

Beecher, 2

Robins (1959), Matthews (1972, 1994), Anderson (1992), Aronoff (1994) and Stump (2001 among
others, the paradigm is generally restricted to inflectional processes. Van Marle (1985), Bochner
(1993), Pounder (2000) and ongoing work by Bauer (1997), Booij (1996, 1997, 2002) and Stump
(2001), in viewing the distinction between inflection and derivation as a cline rather than a
categorical break, suggest an equal role for paradigm in derivational processes. Derivational
paradigm in much prior research is focused on diachronic aspects of word formation. I complement
this paradigm-oriented research with additional cross-linguistic evidence bearing on the synchronic
analysis of derivational morphology. Bochner’s concept of a rule-pattern network is used to
formulate a generalized notion of paradigm which together with the dynamic model I propose are
applied to derivational processes in several languages to capture the paradigmatic relationships
upon which word formation in those languages is dependent.

Several criteria for distinguishing inflection and derivation are examined in §1, as
differentiating them is critical to generalizing a notion of paradigm for derivation on analogy with
inflection. Using an example of derivational phenomena entailing paradigm, key advantages of
adopting a WP framework for generalizing paradigm are also identified. My proposed model for
representing derivational paradigm and some related research by Bochner are presented in §2. A
generalized notion of paradigm based on a set of features common to both inflection and derivation
is formulated in §3. Cross-linguistic data representing a variety of derivational processes in English,
French, Tigre and Lithuanian are presented and analyzed in §4. Some merits of Paradigm Function
Morphology (PFM) over Finite State Morphology (FSM) are shown in §5, and a formal
implementation of the proposed model demonstrated through an adaptation of PFM. To conclude, a
summary of the findings and some future directions are discussed in §6.


1.1 Differentiating inflection and derivation.
Any effort to develop a generalized notion of derivational paradigm as a theoretical construct on
analogy with inflectional paradigms must provide criteria for distinguishing between inflectional
and derivational phenomena, despite the admittedly problematic nature of a sharp division between
them. Inflection is conventionally construed as the production of a word form suitable for use in a

Beecher, 3

particular syntactic context. In English the forms takes, took, and taken are inflected forms of the
verb take associated with a particular set of morphosyntactic properties restricting their grammatical
distribution. Thus, took may be used with any subject to express the simple past tense; whereas
takes is restricted to use with 3rd person singular subjects only to express the present tense. In each
case the lexical meaning of the verb, noun or adjective1 is not changed. In other words, the
inflected forms of take represent particular values assigned to syntactically relevant properties of the
verb such as tense or agreement without modifying or altering the lexical semantic meaning
associated with the verb.
In contrast, derivation is described as the creation of one word from another in which the
lexical meaning of the created word is in some way distinct from that of the base word. For
instance singer is a word derived from sing in which the original meaning of sing is altered in
singer to become “one who sings”. Moreover, in the case of sing~singer the change in meaning
also entails a change in word class, namely from verb to noun. The derived noun singer is
subsequently subject to any inflection which may be imposed on nominal forms by the morphology
of the language. In this sense, while derivation can be said to feed inflection, it is claimed inflection
cannot feed derivation2.

Table (1) Common criteria used to distinguish inflection and derivation with counter-examples.

Lexical meaning
Never changed
Always changed
cyclic~cyclical or
‘I rejected his leaving’
‘I rejected his leavings’

No change in
Past participles
Change in
Prefixation uniformly
functioning as
category preserving
in English

1 In languages with richer inflectional morphology nouns and adjectives are similarly inflected for case. The only
remnants of case inflection in English are found in the pronominal system as in the nominal form he, possessive his and
accusative him.
2 Booij (1993) argues that this claim is actually more nuanced and points to a distinction between ‘inherent’ versus
‘contextual’ inflection. Also, Kiparsky (1982) proposes incorporating a derivational/inflectional distinction into levels
in Lexical Phonology.

Beecher, 4

Defective verbs like
Formation of gerunds
French frire ‘to fry’
as deverbal nouns is
lacks simple past,
exceptionless in
imperfect, subjunctive

Semantic regularity
Breton merc’h-ed
Not maintained
Deverbal verbs
‘girls’ (simple plural)
formed by re- are
vs. merc’h-ed-ou ‘girls’
extremely regular
(double plural) convey-
semantically in
ing affectionate scorn4
(Trépos, 1957)

No further
Inherent inflection can
Further inflection ‘double plurals’ in
inflection or
feed derivation
or derivation
(Booij, 1993)

Based on these descriptive generalizations, it might be assumed that this is a categorical
distinction confirmable with sufficient evidence. However, table (1) summarizes several findings
showing any such distinction to be equivocal at best. While not an exhaustive examination,
these facts do provide a compelling argument against positing any single criterion as both sufficient
and necessary. Consequently, rather than rely on some arbitrary combination of criteria, a unified
treatment (described in §1.2) is adopted in which inflection and derivation are both considered
modes of word-formation differing only in their respective roles in the grammar.

1.2 A unified treatment of inflection and derivation.
The lack of any single dimension along which to categorically divide inflection and derivation has
suggested to some linguists (Plank 1994) that their separation is more appropriately characterized as
different ends of single continuum or cline. Along these lines, a central premise for treating both in
a unified fashion, as argued by Bochner (1993), lies in the fact that they involve the same sorts of
formal operations such as prefixation, suffixation, infixation, circumfixation, reduplication, etc.
These are all attested cross-linguistically for both inflection and derivation as well as other non-
affixal processes involving apophony, metathesis, or tone. Languages use these strategies to mark
property-exponence (i.e. content to form) pairings, with some form of affixation the most common.
In addition numerous deviations in the property-exponence pairing found in inflectional markings

3 For present purposes productivity may be understood as the ability to generalize a process of inflection or
derivation to all members of a given lexical category.
4 The double plural does not have this connotation with all Breton nouns.

Beecher, 5

also occur in derivation. To illustrate, table (2) provides both inflectional and derivational examples
of various deviations in four different categories. Although property-exponence deviations occur
more frequently in connection with inflection, these facts nevertheless support the claim that
inflection and derivation are alike in their morphotactic properties.

Table (2) Deviations in the marking of property-exponence pairings.


Cumulation Latin /re:ksisti:/
/ti:/ realizes
-ress realizes
one marking

rxisti ‘you had
2nd person and
female gender and realizes two or
sing. num.
more properties

Latin /re:ksisti:/
/s/, /is/, /ti:/
-e and -ate both
two or more
rxisti ‘you had
all realize
realize transitivity markings realize
one property

Latin cornu ‘horn’ -u and –ua
-er realizes either

or cornua ‘horns’
realize either
agentivity or

nominative or
accusative case

English to be
am realizes 1st
gubernatorial realizes adjectival

p. singular and

form of governor
is realizes 3rd
p. singular

In summary, a unified treatment of inflection and derivation recognizes that morphological
processes of either sort make the same contribution, namely the formation of words. Crucially, any
formal operations available to one are equally available to the other, as attested cross-linguistically.
Where their respective roles in the grammar differ is with respect to what each serves to
functionally encode. Inflection encodes morphosyntactic properties such as plurality or non-
finiteness. These are phrase-level properties to which syntactic relations like agreement are
sensitive. Derivation encodes lexicosemantic properties such as agentivity or stativity. These are
word-level properties which determine how a word enters into the semantic composition of larger
constituents. Languages do not necessarily treat all the same properties as either phrasal-level or
word-level in a uniform fashion. Consequently, as some properties may be encoded inflectionally
in one language versus derivationally in another, the particular status of specific phenomena is not
always clear. Following Anderson (1992), the data examined in §4 is considered to be derivational

5 Both inflectional examples using rxisti originate with Matthews (1972).

Beecher, 6

in the absence of any evidence that the syntax is sensitive to the particular processes or forms

1.3 Dutch toponyms: an archetype for derivational paradigm.
To demonstrate empirical phenomena supporting derivational paradigm, this section briefly looks at
the derivation of adjectives and terms for inhabitants corresponding to some Dutch toponyms. In
originally presenting this data and analysis, Booij (1997) argues that a paradigmatic perspective is
paramount to achieving an adequately explanatory account of the facts and relevant generalizations.
This exemplary case also identifies several key issues which any theoretical framework must
address. Table (3) illustrates the particular forms involved and their relationships to each other.

Table (3) Dutch toponyms, inhabitant terms and toponymic adjectives.











The significant generalizations regarding these forms are 1) in each case the four forms are formally
and semantically related to each other; 2) the form of an inhabitant term is not predictable from the
corresponding toponym; 3) inhabitant terms serve as a base to which either –isch or –s is suffixed to
form toponymic adjectives; 4) toponymic adjectives serve as a base to which –e is suffixed to form
female inhabitant terms; and 5) there is no single form which can serve as a common base from
which all the other forms may be directly derived. The processes involved are also clearly
derivational since each creates one word from another in which the lexical meaning of the created
word is distinct from that of the base word. Furthermore, nothing in the syntax is sensitive to the
nouns or adjectives so formed. That is to say, any inhabitant term may be used syntactically
anywhere any other noun may be used and the same is true ceteris paribus for any toponymic

Beecher, 7

Capturing and adequately accounting for these generalizations raises several non-trivial
issues. A theory assigning inherent lexical meaning6 to affixes is faced with explaining why the
term for female inhabitant should incorporate an adjectival notion like the ‘quality/characteristics of
being from toponym X’ unlike the term for inhabitants, who are simply ‘from X’. Alternatively
ische or –se might be a sort of single ‘synaffix’, however that would still not explain why
specifically each adjective with –isch or with –s (coincidently) has a corresponding female
inhabitant term with –ische or –se, respectively. In contrast, a theory which recognizes and uses
inter-relationships among the four categories as a regular means of deriving all the forms would not
be confronted by the limitations of accounting for forms in a strictly compositional fashion. Such a
paradigmatic approach would also predict the generalizability of the observable patterns to other
Dutch toponyms. Attention is given in the next section to evaluating and adopting an appropriate
framework for formulating a generalized notion of paradigm which is able to suitably address these

1.4 Adopting a word-based over a morpheme-based theoretical framework.
In formulating a generalized notion of paradigm I assume a WP approach to morphology. In WP
the whole word as opposed to any constituent part is the fundamental unit of analysis and the
minimal unit with which meaning is associated. Thus only whole words and not individual
morphemes are listed in the lexicon. Notions of ‘root’, ‘stem’ or ‘affix’ are referenced, however
only in the context of the relationships between and among individual words. For example in the
Dutch terms Belgie, Belg, Belgisch, Belgische from table (3), Belg- is identifiable as the root or
minimal portion which they have in common. Portions exclusive of the root such as -isch may be
identifiable as affixes. Whether recognizable as derivational or inflectional, affixes themselves are
crucially not meaning-bearing units. It is only in being suffixed to a nominal stem that the
adjectival property ‘quality of X’ is associated with (i.e. ‘realized by’) the affix –isch in Dutch.
Consequently, a WP account is not confronted with reconciling any particular ‘meaning’ associated
with –isch when it appears in a term such as Belgische.
This same example of Dutch Belgische is much more problematic in an IA or IP
framework. The primary object of morphological description in these frameworks is the

6 For example, –isch or –s mean the ‘quality of being from X’ because they convert some toponym X into an adjective.

Beecher, 8

Bloomfeldian concept of morpheme, or the smallest morphological unit establishing a one-to-one
Saussurean association between form (signifiant) and meaning (signifie'). Consequently, the Dutch
word Belgische is viewed as decomposable into the individual segments, Belg-isch-e, where the
form of each segment is a ‘morph’ bearing a unique meaning or ‘morpheme’. Word formation in
IA/IP is the structured concatenation of these segments beginning with a base form to which in
some ordered fashion are added any prefixes and/or suffixes. So, the root, Belg-, is a base for the
suffixation of -isch which in turn produces another stem for the further suffixation of -e. Crucially
each morpheme has its own entry in the lexicon containing such information as its form, meaning,
and in the case of derivational affixes placement restrictions and possibly grammatical category7.
Thus the lexical entry -ISCH identifies it as a denominal suffix in Dutch creating an adjective,
having the semantics ‘quality/characteristics of X’ where X is the base noun. This then creates a
problem when an affix like -isch appears as part of noun such as Belgische, the semantics of which
do not support its inclusion.
In addition to the occurrence of ‘unmotivated’ affixes, so called ‘zero morphology’ is yet
another challenge confronting IA/IP. Zero morphology refers to the absence of any overt realization
for one or more properties associated with a word. This is illustrated by the unaffixed inhabitant
terms such as Belg in table (3). An even more striking example of this phenomenon is found in the
plural forms of some toponyms in Tigre8 as shown in table (4).

Table(4) Some toponymic plurals in Tigre.
Masculine Singular
Feminine Singular
Gender-neutral Plural
‘Bilin speaker’
‘Hamasien speaker’
‘Tigrinya speaker’

In accounting for either the unaffixed Dutch inhabitant terms or the plural Tigre forms, IA has no
choice but to posit affixing a null or ‘zero’ morph to the form in question with which to associate
the relative property of gender or number. Not to do so would deprive a form like ha‹ma‹sen from
being plural at all, as it lacks any morphemes with which the property of plural number may be
associated. IP on the other hand could associate the property of plural number to truncation, in

7 Complex words (both form and meaning) are assembled from their constituent morphemes and so are not individually
listed in the lexicon.
8 Tigre is a Semitic language in Eritrea. Plural formation in Tigre is described by Palmer (1962).

Beecher, 9

which ha‹ma‹sen is formed by an anti-iconic process arbitrarily removing either the masculine or
feminine suffix. To avoid being ad hoc this would require independently motivated evidence,
especially as cross-linguistically truncation is an infrequently attested morphological phenomenon,
particularly in regards to the formation of plurals.
WP is distinguished from morpheme-based frameworks by two central tenets: 1) the
separation of form and content, also referred to as the Separation Hypothesis (SH) of Beard (1986);
and 2) the association of meaning exclusively with whole words. A further distinction, shared by
many lexicalist theories including the variant of WP I assume, is the premise that all word
formation, both inflectional and derivational, occurs only in the lexicon. This is claimed under the
Strong Lexical Hypothesis (SLH) and specifically denies that derivation is restricted to the lexicon
while inflection occurs in the syntax.

Given the primacy of words in WP, having an operational characterization of ‘word’ is most
important. Per the Separation Hypothesis, two theoretical notions of ‘word’ need to be
differentiated: 1) lexeme representing all morphosyntactic property sets or ‘content’ with which a
word may be underlyingly associated (e.g. {1st person, singular, present, indicative} versus {3rd
person, plural, imperfect, subjunctive}); 2) grammatical word referring to the ‘form’ realized by the
combination of a lexeme and any exponence (e.g. French finis versus finissent which realize the
above two feature sets, respectively).

Using these distinct senses for the term word, an inflectional paradigm within WP is defined
as the full set of grammatical words realizing a particular lexeme. Moreover, the structure of such a
paradigm is determined by the inventory of morphosyntactic properties inflectionally encoded in a
given language. Less abstractly, part of the paradigm for the future tense of French finir ‘to finish’
may be schematically represented by a table having six cells as depicted in figure (2).

{1st sg, future}
{1st pl, future}
{2ndsg, future}
{2nd pl, future}
{3rd sg, future}
{3rd pl, future}
Figure (2) Partial schema for inflectional paradigm of French finir
In this way a paradigm captures the formal and semantic relationships among inflected forms
representing a particular class of lexemes (i.e. conjugation or declension) and is generalizable to all