You are on page 1of 27


2 Criteria for the distinction between derivation and inflection

It has already been suggested that the crucial difference between word-formation and
inflection stems from the semiotic (naming) vs. relational functions of the two fields of
language, respectively. For this reason, the following overview takes Dresslers
semiotically grounded criteria, proposed within the context of Natural Morphology, as a
basis for discussion. Dressler (1989) proposes 20 criteria for a distinction between
inflectional morphology and derivational morphology. These are completed in this section
with criteria proposed by Scalise (1988),1 Plank (1994) and other authors. It will be
shown that, while prototypically the criteria are effective, nearly all face counterexamples and/or various degrees of validity indicating their cline-like nature. In other
words, our overview will confirm Planks account (1994) that the majority of the
derivation vs. inflection criteria are based on the principle of continuum with typical cases
of inflection at one end and typical cases of derivation at the other, and with numerous
intermediate cases illustrating both derivation and inflection.
1. Derivational morphology has the function of lexical enrichment, inflectional
morphology does not.
In Dresslers view, exceptions in derivational morphology are notably occasionalisms
formed for the textual purpose of anaphora and cataphora. In accordance with what was
suggested above, we wish to put emphasis on the naming function, with the lexical
enrichment being its (not necessary) consequence. By implication, word-formation
always produces a naming unit, a linguistic sign, even if not all of them comply with
Dresslers requirement of lexical enrichment. As proposed in tekauer (2002), all new
naming units, including those that merely fulfil the anaphoric/cataphoric reference, have
the same status at the time of their coming into existence: they are neologisms coined in
accordance with (more or less) productive rules of word-formation. Therefore, anaphoric
and cataphoric ad hoc formations are not considered here any exceptions and equally
differ from inflectional morphology like any other new linguistic sign.
2. Inflectional morphology has the function of serving syntax or marking syntactic
constructions with special word forms.
If criterion 1 lays emphasis on the semiotic function of word-formation, criterion 2 does
on the syntactic function of inflection which can be labeled as the relational function of
inflection. In other words, inflection, unlike derivation, is required by syntactic
The syntax-focussed criterion was proposed, among others, by Bybee (1985: 813) and Matthews (1991: 50) but, as noted by Booij (2006: 655), [t]his does not mean
[] that inflection is always governed by syntax. As an example, Booij refers to the
accusative form of the Latin word Roma Rome, i.e., Romam, which can express either a
direct object, or an adverbial phrase (to Rome). The use of the adverbial form is not
required by syntax. Similarly, the plural form of book in the sentence John read these
books does not follow from syntactic relations. On the other hand, as further maintained
by Booij (ibid.: 655-656), derivation can also have syntactic relevance because it can
change a syntactic category, and this has consequences for syntactic relations. For

Scalise proposed 15 criteria but a number of them overlap with Dresslers.

example, causatives derived from adjectives, such as to whiten, are transitive verbs
requiring a direct object. We can generalize Booijs position by assuming that, while
inflection is affected by syntax, derivation affects syntax.
3. Inflectional morphology is obligatory within a syntactic construction (unless the
morphological rule is optional); derivational morphology is not.
This criterion follows from criterion 2. As an example, Dressler points out that in a
language with morphological case each noun must appear in a specific case form, but it
may be either derived or a simplex word. In other words, no morphosyntactic
environment requires the expression of a particular derivational category, i.e., the position
of a derivative can be assumed by a simple word.
In Plank (1994: 1673), the observations about inflectional morphology, formulated
here in criteria 1. through 3, take the form of a principle saying that the function of the
inflectional category is of a relational kind (i.e, they relate syntactic constituents to one
another), that of the derivational category of the nonrelational kind. For example, English
3rd person sg. relates finite verbs to subjects by virtue of agreement.
However, Booij (1994) and van Marle (1995) argue that one type of inflection, the
so-called inherent inflection (numerous examples of plural nouns feeding word-formation
in various languages; infinitives, participles and comparatives in Dutch) can feed word
formation, and that inherent inflection
expresses like derivation, a certain amount of independent information, whereas the
information expressed by contextual inflection is redundant, and only reflects certain
aspects of the syntactic structure of the sentence.' For instance, the marking of a noun as
plural is not predictable on the basis of syntactic structure, whereas the plural marking of
the finite verbis, once the subject has been identified (Booij 1994: 30) .

Booij notes that that certain categories of inherent inflection have incomplete, defective
paradigms (cf. singularia tantum, pluralia tantum, mass nouns, proper names; nongradable adjectives; certain participles in Dutch lacking a verbal base thus resembling
complex words without an identifiable word-formation base) and therefore they are not
compulsory in the above sense; that nominal plural tends to be restored in the case of
deflection in contrast to verbal plural which belongs to contextual inflection and, as
such, provides redundant information which need not be restored; that plural forms of
nouns (but not of verbs) as well as comparatives and superlatives of adjectives, verbal
participles and infinitive (but not tensed forms of verbs) may have lexicalized meanings,
etc. (Booij 1994: 32ff).
Furthermore, van Marle examples from Dutch show that derivational forms may
develop inflection-like properties [] and that inflectional forms may display derivationlike properties (ibid: 78) as illustrated in (7):

[[X]Quant ___]NP,

where the slot can be filled by adjectives compulsorily taking -s:


iets groen-s
niets waar-s

something green
nothing true

een heleboel slecht-s a lot of bad things

The -s suffixation is automatic, and the resulting forms do not occur as independent
words: they can only occur as a part of constructions of the above-given structure, that is,
they occur neither as independent words (or very rarely do) nor in other constructions.
These are typical inflectional characteristics. The point is, however, that the suffixed
forms have striking nominal features: (a) at least in some cases, nouns (without -s) may
fill the empty slot in (7), the -s forms are sometimes used as nouns, and some of them
became lexicalized as nouns (lekkers, sweets, lieuws news). This suggests that
inflectional forms may develop properties which are usually primarily associated with
derivation: the ability to change category. The -s forms indicate, then, that category
changing operations can be found in the realm of inflection as well [] (ibid: 74)
Booij and van Marle thus provide strong arguments against the theory of split
morphology (Anderson 1992, Perlmutter 1988).
4. Obligatoriness connected with criterion 3 and with biuniqueness is more characteristic
of inflectional than of derivational morphology.2
While this holds in general, obligatoriness also has some gaps, for example, in incomplete
paradigms (the poor, the rich; the absence of comparative/superlative forms for some
adjectives, like concrete, dead, round, absence of plural for some nouns, like police,
luggage, information, knowledge; or, as noted by Stein 1977, tense and person constraints
on some back-formed verbs in English, like *he housekept, *he speedreads, etc.).
5. As a consequence of criteria 4. and 3., rule variation/competition is typical of
derivational morphology but rare in inflectional morphology.
This criterion is connected with the semiotic nature of word-formation. It has been
suggested elsewhere (tekauer et al. 2005) that each act of naming may be considered as
an act of creativity within productivity constraints. In other words, at the individual
levels of the naming act a coiner may select from a certain number of options: the coiner
may choose one of five onomasiological types as a general conceptual basis for each act
of naming, and may select from a stock of units (lexical morphemes, and affixes) which
may be used to express the semantic components constituting the underlying
onomasiological structure of a naming unit. 3 By implication a relatively high creativity of
word-formation processes is contrasted with strict discipline in the field of inflection
where the function-form relation follows from fixed paradigms.

Dressler (2005: 274) notes that biuniquness holds if one and the same form always has the same meaning
(and vice-versa) which is more natural than uniqueness (one-to-many relations) and especially ambiguity
(many-to-many relations).
Cf. tekauer et al. (2005b) for details of the creativity within the productivity constraints conception
outlined within the onomasiological theory framework. For an onomasiological theory of word-form ation
and the relevant terminology see tekauer (1998, 2005a, 2005b). Individual choices may be affected by
ones language experience, family language background, age, education, profession, and other
sociolinguistic factors. A detailed discussion of the diverse factors influencing decision-making in wordformation is given in Grzega (2002, 2004).

6. Inflectional categories form a small universal set, whereas the language specific
categories of derivational morphology are much more numerous and vary widely crosslinguistically.
While true in principle, this assumption needs to be perceived in the broader context of
Beards theory of Indo-European categories (1981, 1995), and a small digression seems
necessary at this place. Beard lists 44 functions which are identical to those of nominal
Case functions. This is expressed by the Unitary Grammatical Function Hypothesis
(UGF) (Beard 1995: 306):

The number and nature of Case property functions and the functions of the functional Lderivation rules are both determined by the universal set of primitive grammatical
relations [...]

This set of functions is universal, even though not all of them may occur in individual
languages in which more specific constraints apply. Beard thus suggests that nominal and
adjectival derivations are closely interconnected with the original Indo-European case
system and, through this, they are connected (diachronically) with inflectional and
syntactic ways of their expression. The meaningful elements added to the bases are
grammatical relations such as Subject, Object, Means, Locus, etc., which also represent
the fundamental functions of the Case categories (Nominative, Accusative, Instrumental,
Locative). This correlation is supported by four correlation principles:
(i) L-derivations and I-derivations have the same subclassification system
The 44 Case functions are divided into primary, secondary, and spatial functions, some of
them marked by case endings, some by adpositions plus endings, or complex endings, in
some languages by word-order. The same system is then expected for L-derivations.
Beard shows that in some Indo-European languages some L-derived words are marked by
a suffix alone, others by a suffix and a prefix, the prefix often being the same morpheme
as the preposition marking the corresponding inflectional function. Beards English
examples (1995: 308) are given in (10):


Grammatical function
co-worker, co-author
fore-tell, pre-dispose
over-fly, super-class
cross-walk, over-pass
between-class, inter-national Intermediate
in-house, in-state, input

(ii) The same morpheme marks L-derivations and inflectional categories sharing
identical function
This principle means that both are marked by a single Morphological Spelling component
which ignores the distinction between L- and I-derivation and marks only the function in
question. A relevant English example concerns the active participle which serves as a
marker of Subjective nominalization. The Active participle unlike the Passive one
binds the Subject function of the underlying verb syntactically to the head noun it
modifies; for example, the Subject of drench in drenching rain is rain. A Subjective

nominalization such as drench-er also binds the Subject of the base. Importantly,
however, since participles distinguish Tense, which is an inflectional category, they must
be I-derivations. On the other hand, the Subjective nominalization is lexical because it
does not preserve any verbal inflectional functions and its product possesses all the
features of nouns: number, animacy, gender, and noun class. Despite the crucial
differences between the participles and the Subjective nominalization, the Present Active
participle suffix is a common marker of Subjective nominalizations.
(iii) Parallel polysemy, where one affix marks the same set of L-derivation functions that
are marked by some single Case marker even though the affixes differ
For illustration, the Russian Subjective (i.e. AGENTIVE) and INSTRUMENTAL (Modalic
Nominalization) derivations are marked identically by the suffix
-te: ita-te (reader) vs. podogreva-te (heat-er). Identical situation is in other Slavonic languages. So,
for example, Slovak AGENTIVE noun kov- blacksmith uses the same suffix as the
INSTRUMENTAL noun ohriev-a heater.4 Similar examples abound. Analogically, a single
Instrumental affix marks both Passive Subject and Active Modalic functions (Beard 1995:

a. Ivan porazil Boris-a

Ivan stunned Boris-ACC REFLEX-INST wit-INST
Ivan stunned Boris with his wit
b. Boris byl poraen Ivan-om
Boris was stunned Ivan-INST
Boris was stunned by Ivan

(iv) Typologically and historically parallel development

An example of historical correlations is the Slavic Locative function. *IE Locative did not
distinguish Locus in terms of the Inessive function (in) and the Adessive function (on).
The Slavic Locative, however, introduced such a distinction by means of prepositions (for
example, Serbo Croatian Inessive u + Loc in and the Adessive na + Loc on: u park-u
in/at the park vs. na polj-u on/at the field). In accordance with prediction (iv) of the
UGF, the same kind of change emerged in the derivational system of Slavic languages,
and the Locative nominalizations split in two, for example (Beard 1995: 319):

a. raditi u work in
a. raditi na work on
b. kupati se u bathe in
b. kupati se na bathe on

rad-io-nica workshop
rad-il-it-e work site
kup-ao-na bathroom
kup-al-it-e bathing beach

To account for the principle that both L- and I-derivations are based on the same set of
functions despite the fact that lexicon and syntax represent two independent components,

The forms -a and - are variants of the same morpheme.The short variant in the case of the Instrumental
noun is conditioned by the Rhythmical Law in Slovak preventing the occurrence of two long syllables next
to each other.

Beard adapts Bothas Base Rule Theory. Botha (1980, 1981) maintains that the base
generates syntactic structures whose nodes bear morphosyntactic categories that are
subject to lexical as well as inflectional rules. Beard (1981, 1988) adjusted Bothas Base
Rule Theory to the framework of the Lexeme-Morpheme-Base Morphology. In his view,
the base is a general grammatical component, a categorial component which accounts
for all the grammatical relations of language common to syntax, inflection, and the
lexicon. Only after lexical selection are lexical and syntactic structures distinguished
(Beard 1995: 328).
In other words, and in accordance with his Lexeme-Morpheme-Base Morphology,
Beard assumes that a general base structure may undergo either a lexical or a syntactic
operation yielding different results.
7. The uniformity is greater in inflectional morphology than in derivational morphology.
This criterion is derivable from criterion 4 concerning the obligatory nature of inflectional
rules, and means that exceptions in inflection are in principle exceptions to the rule, while
gaps in the application of derivational rules are common. The criterion of uniform
semantic contribution of inflectional rules vs. diversity characterizing derivational rules
was also proposed by Plank (1994: 675).
However, Stump (2005: 55-56) suggests possible counterexamples to the
exceptionless nature of inflectional rules by pointing to cases of irregular past tense
formation in English and defective paradigms, 5 on the one hand, and to complete
word-formation rules, such as -ing deverbal substantives in English, on the other. He
aptly notes that this issue depends on the definition of the notion of completeness.
In this connection, we find Lacas claim (2001: 1218) that irregularity in
derivational morphology manifests itself in the many-to-one and one-to-many
relationships too strong. This type of relationships seems to be equally typical of
inflectional morphology, at least for fusional languages which are characterized by
numerous cases of cumulative exponence and syncretism. Also, as noted by Anderson
(1992: 75), portmanteau morphs, which generally involve the conflation of two or more
otherwise independent inflectional categories in a single formative, never occur in
A highly interesting example of the one-to-many relationship in inflectional
morphology is provided by our informant for Zulu (van der Spuy, pers. comm.). He
mentions the verbal suffix -a which has a kind of default meaning: there are several
verbal suffixes in Zulu, e.g. -e SUBJUNCTIVE, -ile PERFECT, -i PRES.NEG with more precise
meanings. The suffix -a is used when no other suffix is available, and could therefore be
said to have a multiplicity of meanings, but it could also be analyzed as having a neutral
8. The meanings of inflectional morphology are more abstract than those of derivational

See also some examples of defective paradigms given within Criterion 4.

This claim is also emphasized by Adams (1973) but should be very carefully examined,
also in view of Beards Unitary Grammatical Function Hypothesis, which seems to
relativize any considerations concerning the abstract-concrete opposition.
9. Grammatical agreement is typical of inflectional morphology (especially if the
language belongs to the inflectional type) and rare in derivational morphology.
This criterion seems to raise no objections, and is a natural consequence of criterion 2,
emphasizing the syntactic, relational nature of inflectional morphology.
10. Inflectional morphology is usually morphosemantically more transparent than
derivational morphology.
This criterion does not hold generally. The major problem with morphosemantic
transparency is posed by fusional languages, in particular, by cumulative exponence and
extended exponence so characteristic of these languages. Let us take, for example, one of
the numerous paradigms of the Slovak substantival system represented by the pattern
chlap man, fellow:



Each inflectional morpheme is a portmanteau morpheme (see point 7 above) expressing

three fundamental meanings: case, number, and gender (masculine in this case), and it
also reflects the distinction between animate-inanimate objects. Obviously, the
considerable degree of cumulative exponence works against morphosyntactic
transparency, so characteristic for agglutinating languages. In addition, there is extensive
syncretism in this and the other Slovak (and also Slavonic) paradigms. The inflectional
morpheme -om, for example, can mean either singular, instrumental, masculine animate
noun of the pattern chlap or plural, dative, masculine, animate noun of the same pattern.
This extended exponence runs across paradigms, and so -om, for instance, does not only
indicate singular, instrumental, masculine noun of the pattern chlap but also the singular,
instrumental, masculine inanimate noun of the pattern dub oak. This situation, so
common in Slavonic languages, contradicts the general validity of Dresslers criterion no.
On the other hand, Plank (1994: 1675) aptly relates the lack of transparency in
some derived words to the process of lexicalization to which inflected forms are usually
not exposed. Plank gives the following examples of exceptions to plural transparency:
air-s unnatural manner or action intended to impress; damage-s money claimed from a
person for causing damage; work-s moving parts of a machine. They indicate that an
inflected form, too, may undergo a process of semantic shift typical of a basic form of a

11. Morphological rules of inflectional morphology are typically more productive than
those of derivational morphology.
This criterion reflects the deep-rooted belief of a number of morphologists and
syntactitians, at least since Chomskys Remarks (1970), that productivity of wordformation processes is much lower than that of syntactic and inflectional processes.
However, it has been called into question in recent decades (cf., for example, Anderson
1982, Strauss 1982b, Di Sciullo and Williams 1987, tekauer 1998). In any case,
productivity remains one of the most contested areas in the study of word-formation
(Bauer 1983: 62). And, in contradiction to the above-mentioned criterion, Plag (1999: 2)
emphasizes that derivational processes are much more regular than previously
conceived, and Anderson (1992: 78) aptly notes that even a completely productive
process can still, arguably, be derivational, such as the formation of English nominals in
-ing from verbs.
Bybee (1985) speaks in this connection of generality (see Point 15 below) and
points out numerous restrictions imposed on derivational processes; she notes that even
highly productive derivational rules are subject to certain productivity constraints (a view
now widely accepted).6 By contrast, inflectional categories must be present in every word
of a particular class.
The previous sentence indicates the unequal criteria used for the evaluation of
productivity in inflection and in word-formation. While productivity of inflectional
morphology is usually assessed on the basis of morphological categories, i.e., from the
perspective of categorial meanings, like the formal means used to express the categories
of gender, plural number, case, tense, etc., productivity in word-formation is judged from
word-formation processes. In other words, the mainstream generative morphology
evaluates productivity in word-formation from a purely formal viewpoint. This engenders
a biased view of limited productivity in word-formation. This point has already been
mentioned by Bauer (2000: 37) in a different context when he maintains that
[] inflection appears to be defined in terms of certain categories such as number, tense,
person, etc. As far as I know, nobody has attempted to define derivation in terms of the
categories involved [] This raises the question of whether there are any categories
which we can view as derivational in the same way as tense is seen as being inflectional

This brings us, on the one hand, to the point made by Beard (see above) concerning the
Indo-European system of basic case functions and, on the other, invites to question the
assessment of both fields on the same footing, the footing of categories. This was
proposed in tekauer (1998) when pointing out that the same sort of argument as that
proposed by Bybee should equally be applied to word-formation: it is not important in
what way (by suffixation, compounding, conversion, etc.), for example, AGENTIVE
meaning is expressed. What matters is that, if required (not by a syntactic slot but rather
by a conceptual/cognitive slot). such an AGENTIVE name can be produced for any
AGENTIVE concept. Furthermore, while it is true that the suffix -ion does not combine
with all verbs, it is also true that not all verbs can be used in the sentence structure noun

For an overview of restrictions on productivity of word-formation rules, see Rainer (2005). For a
discussion on various approaches to productivity, see Bauer (1983, 2001, 2005), Plag (1999), and tekauer
et al. (2005).

verb object. The limitation permits only transitive verbs to be inserted. Both restrictions
(syntactic and derivational) are based on the same principle of combinability of structural
units. And, to give one more example from the field of inflection, the productivity of, for
example, Slovak plural suffix -ovia (as in hrdinovia heroes) as one of the relatively
high number of plural affixes is restricted to one of 12 different substantival paradigms,
notably plural of masculine animate nouns ending in a vowel).
Our claim actually corresponds to Bybees (1985: 84) argumentation concerning
This does not mean that all expressions of an inflectional category must be regular or
productive it does not matter if an English verb forms its Past Tense by suffixation or
vowel change it just means that there must be some way to form the Past Tense of every
English verb.

From this point of view, however, the assumption of full generality of inflection vs.
significantly restricted generality of derivational processes does not seem to be so
12. Prototypical inflectional morphology does not change word class, derivational
morphology often does.
It may be added that one and the same derivational affix may attach to words of different
word-classes. Furthermore, the change of word-class has crucial consequences for the
paradigmatic and syntagmatic characteristics of a new word. First, the set of inflectional
morphemes (inflected forms) of a word completely changes and, second, the function of
the word in a sentence is completely different from that of its motivating counterpart.
Prototypically, inflectional processes do not have this sort of influence upon paradigmatic
and syntagmatic features of words (see also Malkiel 1978: 128).
While one might object along with Stump (2005: 53) that this criterion is limited
to class-changing derivation, Scalises position is even stronger than Dressers when he
claims that:
[e]ven when a noun remains a noun (as in man manhood), it is reasonable to assume
that the suffix has changed the entire list of information attached to the base [] The
-hood rule, for example, changes the features <abstract> and <+countable>:

N, <abstract>, <+countable>
N, <+abstract>, <countable>

[] there is no derivational rule which leaves unchanged both the lexical category and
the features associated with the base (1988: 564).

Regarding Stumps (1995: 54) remark in 7 above, note that the evaluation of completeneness of
inflectional paradigms depends on the definition of this notion. He exemplifies his argument by the
incompleteness of the -ed past tense rule and the incompleteness of various irregular patterns of past
tense formation in irregular verbs. To avoid this incompleteness, one must postulate a higher level of
generalization, i.e., that for virtually every verb in the language, there is an operation defining the pasttense form (Stump 2005: 54). This assumption brings us to our insistence on using the same criteria for
both derivation and inflection.

Scalise (1988: 565) therefore suggests a Unitary Output Hypothesis: the output of a rule
of suffixation is always the same, independent of the base, which means that it is not
important what the specific base of, for example, the rule attaching the suffix -hood is: the
resulting word (output) is always an abstract, uncountable and common noun. Scalise
adds a list of features that can be changed by derivational rules but not by inflectional
rules, including syntactic category, conjugation/declension class, subcategorization
features, selectional features, inherent features, count, animate, abstract, common, etc.
Stump (2005: 53ff) objects that also an inflectional category can affect the
category of word-class. He refers to cases like present participles (i.e., inflected forms of
verbs) which can also function as attributive adjectives never occurring without
inflectional morphemes (in Slovak, for example, these forms are always fitted with
gender/plural morphemes). Moreover, Stump refers to Sanskrit, where the derivation of
causative verbs is a productive process, but this process is simply based on shifting a verb
from one conjugation category to another. Thus, the derivation of the causative verb from
the verb dvis hate is simply based on its shifting from the second to the tenth
Our sample provides us with several such examples. In Swahili, for example, both
Diminutives and Augmentatives are formed very productively by the change of paradigm
(noun class) (Contini-Morava). A change of inflectional paradigm is a device used in
Datooga where causatives are formed by conversion of verb from class 1 to class 2

(class 2) stick, fasten < (class 1) be stuck

A prototypical case of the change of word-class without derivational affixes is conversion

in Slovak, Polish, Russian and other Slavonic languages (Sl. beaV > behN, ervenAdj >
ervenNpl). According to Smirnickij (1953), conversion, as a word-formation process
whose most striking feature is the change of word-class, is based on the change of
inflectional paradigm. A more radical position is taken by Haspelmath (1996: 50):
the myth that word-class changing inflection does not exist is not more than a myth [] it
has to do with the fact that grammatical theory has been dominated by thinking about
English grammatical structure for the past decades and continues to be dominated by
Anglophone linguists. English has very little morphology, and although it has several
cases of transpositional [i.e., class-changing] inflection (participle, gerund, adjectival
adverb), these were not sufficient to direct grammatical theory in the right direction [] it
may be that the myth discussed above is another example of an error introduced into the
mainstream of grammatical theory due to insufficient consideration of linguistic diversity.

Haspelmath (1996: 52) distinguishes between word-class properties related to external

syntax determined by word-form word-class, and to internal syntax determined by lexeme
word-class. Thus, for example, the German participle singende singing, as in der im
Wald laut singende (Jger) (a hunter) singing in the forest, behaves like an adjective
from the point of view of external syntax, and as a verb in terms of internal syntax
(because a verb cannot be modified by an adverb of manner or location). Haspelmath
emphasizes and illustrates that the preservation of internal syntax is not a matter of all or

nothing, but that different languages may preserve fewer or more properties of internal
syntax (ibid.: 60).
From this it follows that a word is transposed (in the sense of Marchands 1967
definition of transposition as a use of word in the function deviating from its default
function) in some of its uses. While our previous discussion confirms the basic idea of a
class-changing inflection, the criteria used by Haspelmath may be called into question.
His main criterion for the distinction between inflection and derivation is that inflection is
regular, general and productive, while derivation is irregular, defective, and unproductive
(1967: 47). This view, no doubt, echoes Chomsky (1970), but it has been demonstrated by
a number of morphologists that this assumption is another unjustified myth, as stressed in
our discussion on criterion 11 above. Furthermore, Haspelmath maintains that inflectional
forms are described exclusively in grammatical paradigms, whereas derivational
formations are described by listing them individually in a dictionary and that derivation
is never organized in paradigms [] (ibid.). In Point 16 below we give important
arguments in favour of the existence of derivational paradigms which come from a
background other than Anglophone linguistics.
Regarding individual examples, we do not share Haspelmaths view that, for
instance, the formation of adverbs from adjectives is an inflectional process just because
it is highly productive and regular in English. Cognitively, they represent two different
conceptual categories (QUALITY and CIRCUMSTANCE, respectively), which implies the
derivational status of -ly adverbs in English. Furthermore, a major difference between
inflection and derivation is that derivational categories can be replaced by non-derived
words of the same word-class, while inflectional forms are always dependent for their
existence on their basic forms. Thus, this existence of adverbs does not depend on the
existence of adjectives: many adverbs exist independent of the derivation process. On the
other hand, present participles in English functioning analogically to adjectives do depend
on their basic verbs and are therefore their inflectional forms. This, in principle, is in
accordance with Haspelmath distinction between external vs. internal syntax.
All in all, we do believe that inflection can be class-changing but, in general, this
feature is a prototypical feature of word-formation.
13. Criterion 12 is the basis for the structuralist criterion of substitutability within the
same slot.
Plank (1994: 1673) notes that the basic postulate for derivations is that they can be
replaced in all syntactic contexts (unlike inflected forms). For illustration, he mentions the
3rd person sg. of English verbs which can only be replaced by basic verbs in certain
subordinate clauses and in jussive main clauses (such as It is essential that he comes/come, God save-s/save the Queen). Derivations, on the other hand, are in principle,
substitutable by simple words in general.
14. Morphological rules of derivational morphology are easier to reapply than
morphological rules of inflectional morphology.
This criterion has been strongly supported by our cross-linguistic data: out of 52 suffixing
languages in our sample, 40 admit multiple (recursive) suffixation; also, out of 43
prefixing languages, 34 admit multiple prefixation (although these languages differ in
terms of productivity of these processes).

One and the same derivational affix can occur more than once in one word. Let us
mention Bauers (1983: 67ff.) examples for English, such as re-remake, meta-metatheory, and semi-hemidemisemiquaver for prefixation, even if, as noted by Bauer, there
are (semantic, pragmatic, etc.) restrictions on the recursiveness of affixation processes.
Negative prefixation, for example, is rarely recursive, with few exceptions such as
In suffixation, recursiveness seems to be much more productive. As an example,
let us mention institutionalization, with two -ion suffixes; containerizer with two -er
suffixes (Bauer, ibid.: 69-70); diminutive formations such as Span. chiqu-it-it-ito very
very tiny, Ital. panc-in-ino little belly, or with different suffixes poch-ett-ino a tiny bit
(Laca: 2001: 216), Turkish causatives (l-dr-t-t caused to cause to die) (ibid.).
Let us also note that West Greenlandic features considerable recursiveness of
derivative affixation in this language: There are around 400 productive affixes in use
[] They may combine with each other iteratively, producing a prodigious potential for
the derivational expansion of simple stems; up to ten or more affixes in succession before
the inflectional ending is not particularly unusual [] (Fortescue 1984: 313). In our
questionnaire Fortescue adds that these affixes can be used recursively to build up
complex verbs and nouns, with possible switches back and forth between verbal and
nominal base within a single word. He illustrates the repetition of the same suffix -siur
look for (ibid.: 316):

coal future look-for place future look-for intr.-part. only be intr.-part
who is the only one looking for a place to look/prospect for coal

Affixal recursiveness is also very productive in Slavonic languages in the field of

evaluative morphology, especially DIMINUTIVENESS. A case is point is Slovak infix -li:

mal small
mali-n-k very small
mali-li-nk very very small
mali-li-li-li-li-nk very very [] small

This infix can be repeated, in principle, ad infinitum. A similar example is given by

Szymanek for Polish (1998: 73):
(17) a. nos nose
nos-ek nose-DIM
nos-ecz-ek nose-DIM-DIM
b. dom house
dom-ek house-DIM
dom-ecz-ek house-DIM-DIM
(18) gives examples of combinations of diminutive affixes in Slovak:

(18) a. maco teddy bear (childish)

mac-ko teddy bear + DIMINUTIVE
mac-in-ko teddy bear + DIMINUTIVE + DIMINUTIVE
b. medve
medved--ek bear + DIMINUTIVE + DIMINUTIVE
Certainly, there is always an upper limit to the number of affixation processes per
language. According to Hardman (2000: 88), [a]s many as four derivational suffixes can
occur per verb root [] in Jaqaru. Similarly, Ljung (1970: 3) states that the maximum
number of prefixes in English is three, and the maximum number of suffixes is four. 8 The
latter seems to be disputable as adding -al to institutionalization gives a word with five
Furthermore, as noted by Scalise (1988: 571, 577-9), the cumulative nature of
derivation manifests itself in the fact that the order of derivational affixes reflects the
order of semantic operations, whereas the order of inflectional affixes is fixed and/or
semantically irrelevant: only one order is available. Scalise (1988: 571) illustrates this
criterion with an example from Italian:

automobile + ista + ico relative to motorists

stor + ico + ista historicist

This observation can be supported by an example from West Greenlandic. Fortescue

maintains that [r]eversing the relative order of two successive affixes will generally
change the meaning entirely [] (1984: 313). His example is as follows:
(20) a. urnik-kusun-niqar-puq
come-to want passive 3s-indic.
Somebody wanted to come to him

In this respect, consider the Redundancy Restriction formulated by Lieber (2004: 161) that restricts the
possibilities of recursiveness: Affixes do not add semantic content that is already available within a base
word (simplex or derived), even if there are some exceptions to this rule, for example, double AGENTIVES
checkerist, consumerist, tympanister, collegianer, musicianer, physicianer or English relational adjectives
like arithmetical, geographical, etc. (Lieber 2004: 164). Similar examples emerged in our experimental
research (tekauer et al. 2005: 43): butter-inner, hanger-onner, butter-innist; weberer, shoe-tier-upper,
grass-cutter-upper, bird-fisherman, or shoe-tierman. Cf. also section 2.8.2 for an interesting type of
recursiveness in Kwakwala.
An extreme case which at the same time indicates an immense capacity of (English) derivation is
illustrated by the following example which I received from an American lector in Slovakia several years
ago who insisted on it being an authentic text:
At present, gentlemen, we live with an apparently stable balance of terror. But that balance may at any
time be de-stabilized by our opponents. As the leaders of the peace-loving state, our objective must be an
un-de-stabilize-able balance. But now, just as we have begun to un-de-stabilize-able-ize the situation, our
opponents have bent all their efforts to de-un-de-stabilize-able-ize our precarious balance. In our current
negotiations, it will not be enough to require an un-de-stabilize-able balance: we must aim to create an unde-un-de-stabilize-able-ize-able balance.

b. urnin-niqa-rusup-puq
some-to passive want 3s-indic.
He wanted somebody to come to him
In Ket the finite verb is actually a positional formula (Vajda, pers. comm.). To form
transitive/intransitive/causative/frequentative involves not only the addition of this or that
prefix, or set of prefixes, but also often a rearrangement of the position of the roots in the
stem: the basic verb root may go from the end of the verb and into the incorporate
position, while the original root position is filled by a marker of transitivity or aspect.10
That the prediction about the non-existence of recursiveness in inflection,
analogical to the cyclic application of derivational rules, is not absolute follows from
Dresslers (1989: 8) example taken from Turkish:

ev house
ev-ler Plural
ev-ler-de Loc. in the houses
ev-ler-de-ki that/which/who is in the houses
ev-ler-de-ki-ler Plural
ev-ler-de-ki-ler-de Loc. in those which are in the houses

15. Inflectional morphology typically involves smaller meaning changes than

derivational morphology.
As noted by Dressler (1989: 8), the meaning change in, for example, plural formation is
smaller than in semantically related collective formation, for example professor-s vs.
At this place, another digression is necessary, in this case, in reference to Bybees
(1985) theory. Tatevosov (2006: 284) argues:
Bybee (1985) makes significant cross-linguistic generalizations about the distribution of
inflectional and derivational morphological items. Having examined verbal morphology
across 50 languages, she proposed that derivational/inflectional status of morphological
items correlates with their semantic relevance for the root and generality of their

In Bybees approach [a] meaning element is relevant to another meaning element if the
semantic content of the first directly affects or modifies the semantic content of the
second (1985: 13). Bybee illustrates the point with the verb to walk in its meaning to go
on foot by taking steps. As she notes, the additional meaning through water is inherent
in the verb to wade which thus combines the two meanings in one word because
whether one has ones feet on dry land or in water is quite relevant to the act of walking
(ibid.) in contrast to, for example, a limited relevance of a sunny or cloudy sky for the
act of walking. Therefore, logically, a language does not have any special word to express
the meaning walk on a sunny day or walk on a cloudy day
Bybee distinguishes two types of derivation: class-changing and classmaintaining. They behave differently in terms of the relevance principle. Classmaintaining derivations are characterized by considerable semantic changes. For

Cf. an example from Kwakwala in Section 1.2.1.

example, with verbs they result in the changed valence. Importantly, valence-changing
categories produce large meaning changes in verbs, since an event can be changed
substantially if the number of participants and the nature of their roles change. Thus kill
differs from die, and send differs from go in the events being described. So it is not
surprising that in the cross-linguistic survey, valence was found to be frequently
mentioned as a derivational category for verbs. (ibid.: 83).
The class-changing derivation results in a changed syntactic category of a word
which is a highly relevant information about a word. In any case, the class-changing
derivations make varying amounts of semantic change, depending on how much
semantic content they contribute along with the category change (ibid.: 83).
In view of our topic, Bybees important observation concerns some border-cases
between derivation and inflection due to limited semantic contribution of some
morphemes. A case in point is English gerundial nominalizations in -ing which tend to
describe the same situation from the perspective of a different word-class: Bill reads in
bed and reading in bed is fun the semantics of reads and reading is very similar.
As also indicated by this example, the change in meaning is of a scalar nature, and
thus it cannot provide an unambiguous distinction between derivation and inflection.
Therefore, Bybee introduces a second criterion, in particular, meaning generality, i.e.,
consistent application of meaning across a particular category of roots with the same
semantic effect. Derivational processes are identified by a high degree of relevance for
the meaning of the root and low level of generality, while the opposite characteristics of
relevance and generality apply to inflectional morphology. As a final criterion, Bybee
applies the amount of semantic change resulting from affixation: the greater the
difference between the meaning of the derived word and the meaning of the base, the
greater the likelihood that the affix is derivational (ibid: 5).
In this connection, let us mention Scalises (1988) criterion according to which
derivational rules change the conceptual meaning of their base, while inflectional rules
change the grammatical meaning of their base. Anderson (1992: 79) speaks of
inflectional meaning in opposition to more genuine semantic meaning. This claim may
be related to the above-mentioned discussion of the semiotic foundations of wordformation because Scalises observation is necessarily a direct consequence of the
semiotic nature of derivational rules, and their close interconnection with extralinguistic
16. Inflectional morphology is typically organized in paradigms and inflectional classes,
whereas the paradigmatic organization of derivational morphology is much weaker.
While this criterion reflects another deep-rooted belief (see also, for example, Katamba
1993) we should like to refer to Bauers discussion of derivational paradigms. Bauer
(1997a: 254) admits that derivational paradigms are not prototypical paradigms but, at the
same time, since there is a cline between typical cases of inflection and typical cases of
derivation, it is possible to expect that there also is a cline in degrees of paradigm
coherence and applicability between the two. Based on this analysis, Bauer concludes that
there are arguments for the value of the derivational paradigm in morphological study,
even though it appears that a derivational paradigm allows a lesser degree of prediction
than an inflectional one does (1997a: 255).

Similarly, van Marle rejects the view that paradigmatics is of little or no

importance to derivational morphology. The only conclusion which is justified is that, in
derivation, paradigmatic structure may manifest itself in a fundamentally different way
from the way it does in inflection (1994: 2929). This stance may be further supported by
Valls view that word formation patterns emerge from paradigmatic relations (2003:
14), and Booijs position that native speakers competence to create new compounds and
derived words is based on abstractions over sets of existing complex words and the words
that are paradigmatically related to them (to appear).
Dokulil (1962: 12ff) views paradigmatic relations in word-formation in an even
more comprehensive way: he points out that word-formation relations cannot be reduced
to those between the motivating and the motivated words. Rather, each such pair of words
is the basis for much more complex relations, including clusters relating a motivating
word with a set of motivated words, as illustrated in (22):


lstek (small leaf)

lst (leaves)
listina (document)
listov (leafAdj)
list (dossier)
listov (foliage)
palist (stipule)
listovka (leaf-

oliten (provided with leaves)

LIST (leaf)
listovat (browse)
listovn (concerning a letter)
listnat (deciduous)
listopad (November)
listn (deciduous tree)

or a motivated word with a set of motivating words:


list leaf

list leaf

listnatec deciduous tree

listnat leafy

lst leafage


listopad November




or a series of words whose neighbouring members are related by motivation:


A simplified representation of Dokulils example.


list lstek lstkov lstkovit lstkovitost

leaf leafletN leafletA characterized by leafletsA property of
having leafletsN

These clusters and series constitute word-formation families. A relatively simple family
exemplified by Dokulil (1962: 13), based on the motivating word med (honey), includes
61 words.
Furthermore, each of the words in the family is a member of one of several series
on the basis of its affix (or the right-hand member of a compound). An example is given
in (28)

listn belongs to a series, including words like slamn, hlinn, vlnn,

bavlnn, lnn, kostn, mdn, olovn, drtn, etc., all of them having the
meaning MADE OF leaves, straw, clay, wool, cotton, linen, bones,
copper, lead, wire, etc.

This brings us to what might be called derivational paradigm in a narrow sense,

corresponding to the concept of proportional series:




smrkovit etc.

(The motivating words are leaf, flower, tooth, corner, and spruce, respectively;
the second column includes their substantival diminutives, the third column diminutive
adjectives, and the fourth column words carry the meaning characterized by what is
expressed by the diminutive).
Furdk (2004) developed Dokulils ideas and defines the concept of derivational
paradigm as an ordered set of motivants (i.e., motivating words) constituted
immediately from a single motivant with no motivating relations between the comotivants as in the following example (2004: 74):

kola school

kol-k schoolboy
kol-nk janitor
partial set of co-motiovants
kl-ka kindergarten of the motivant kola
kol-stvo education system
kol-ika small school

In Furdks view, derivational paradigms constitute, as it were, a system of wordformation cases, even if this system is not so fixed as an inflectional paradigm.
The approach proposed by Dokulil and Furdk stresses a highly important role of
derivational paradigms of different degrees of complexity for the system of wordformation. They show that the derivational system is in fact based on paradigmatic
relations reflecting the internal structure of the word-formation system of a language as a
whole. This fact obliterates to a considerable degree the difference between derivational

and inflectional morphology in terms of the role of paradigm in their organization, and
partly contradicts Dresslers and Haspelmaths view of the paradigm as a crucial criterion
for drawing this distinction. We use the term partly because, as also emphasized by
Rainer (pers. comm.) paradigm has two different meanings/organisations in derivation
and inflection.
17. Analogical leveling is much easier in inflectional morphology than in derivational
morphology, because the members of a close-knit paradigm can influence each other
more easily.
This criterion seems to be unproblematic, but at the same time not of much use for the
distinction of the individual problematic cases.
18. Accepted derived words are likely to be stored as wholes in memory, whereas
inflected word forms are unlikely to be so.
Booij (2006: 659) refers to contradictory results of recent psycholinguistic
experiments some of which suggest that derived words, even regular ones, are always
stored while only irregular inflections are stored in the lexical memory (Clahsen et al.
2003). Plag (1999: 11) maintains the view, supported by psycholinguistic and structural
linguistic arguments, that regular complex words can also be stored in the lexicon. By
contrast, experiments by Stemberger and MacWhinney (1988), and Baayen et al. (1997)
indicate the storage of regular inflectional forms of high frequency of occurrence.
Views within the field of generative word-formation are not homogeneous either.
Aronoff (1976) and Anshen & Aronoff (1988) maintain that the lexicon only contains
unpredictable, idiosyncratic units; words that are based on a predictable and regular
compositional meaning are formed anew by a rule each time they are needed by syntax. A
similar view is presented by Bauer (2005: 320-1).

19. Inflectional affixes typically have a more peripheral position in the word form than
derivational affixes.
This criterion is, in fact, a paraphrase of Universal 28: If both the derivation and
inflection follow the root, r they precede the root, the derivation is always between the
root and the inflection (Greenberg 1966: 93). This preference has several reasons
according to Dressler (1989: 8):
i) Derivational morphology forms words, inflectional morphology does not. Due to
the word base preference, outputs of derivational rules must be preferred as bases
for inflectional morphology than vice versa. Then affixes are typically stacked
from center outwards, with the exception of (dispreferred!) infixes and interfixes
which are inserted medially. This results in the prototypical position of inflectional
suffixes outsides of derivational suffixes.
ii) Roots have the most concrete meanings, affixes of derivational morphology less
concrete ones, affixes of inflectional morphology the most abstract ones (see Point
8 above). This is reflected in the relative position of morphemes.

iii) Affixes of inflectional morphology are indices of syntactic functions, i.e., they
indicate (indexically) how the words are related within a syntactic construction.
Indices are the more effective the closer the signantia are to the elements indicated
(their signata). Thus, a peripheral position is more effective than an internal
This criterion, formulated from the position of Natural Morphology, corresponds to the
basic assumptions of the theory of split morphology (e.g., Anderson 1992) which, as
noted by Booij (2006: 658), predicts this order of derivational and inflectional affixes by
assuming that new lexemes are generated by a pre-syntactic component and inflected
forms by a post-syntactic component of the grammar.12
The peripheral position of inflectional affixes with regard to derivational affixes is
captured by Greenbergs Universal 28 (1963), which predicts that productive rules of
word-formation apply before regular inflection, a claim made by various models within
the framework of lexical phonology. Kiparsky (1982a,b), for example, puts #-boundary
(regular) inflection at level 3, preceded by #-boundary derivation and compounding at
level 2; a three-level model with similar distribution of derivation and inflection is
proposed by Anderson (1992). Mohanan (1986) also distinguishes four strata, with
regular inflection at stratum 4, preceded by both class 1 and class 2 derivations and
compounding. By contrast, Katamba (1993) distinguishes only two strata: while irregular
inflection and derivation take place at stratum 1, regular derivation, inflection, and
compounding take place at stratum 2.
This point also follows from Andersons observation (1992: 75) that derivational
morphology is stem-based (= surface words minus inflectional material), which clearly
documents the difference between derivational and inflectional units. Moreover, derived
words are often built on stems that contain other derivational affixes, but not inflectional
Certainly, neither this criterion is absolute, and there are important exceptions,
either exceptions to the system of a language, or the exceptions follow from the existence
of languages whose system as such is an exception to the general expectations. It is this
latter type that suggests that not all exceptions belong to the periphery of a language
system as believed by Scalise (1988: 566-567).
As for the first type of exception, let us mention well-known English examples
like worsen, betterment, unhappier, or Hebrew imahut (motherhood) formed from the

The model of split morphology has a number of opponents, for example, Halle (1973), Williams (1981),
Lieber (1981, 1992), Guerssel (1983), Kiparsky (1982a, b), Mohanan (1986), van Marle (1995), and Booij
(1994, 1995). Van Marle (1995), for example, argues against the strict division of derivation and inflection
based on examples of Dutch which, in his view, demonstrate that derivational forms may develop
inflection-like properties, and inflectional forms may display derivation-like properties. Therefore, he
speaks of the interwoven character of derivational and inflectional properties, and maintains that
derivation and inflection bear upon two distinct aspects of words: their lexical-semantic dimension and
their syntax-oriented dimension. Typical of word structure seems to be, then, that these two dimensions are
often entangled [] Evidently, this means that both dimensions of the word i.e. the lexical semantic and
the syntax-oriented constitute a unity which is much closer than is often assumed (1995: 78-79). Van
Marle concludes that inflection and derivation should be dealt within one and the same component of

plural of em (mother), i.e., pl. imahot by means of the suffix -ut (Schwarzwald 2001:
For the latter type of exception let us cite Vajda (pers. comm.) on Ket. For Vajda,
it is difficult to separate derivational and inflectional affixes in the finite verb: they are
interspersed in between one another in a rigid template of eight prefix position classes:

subject 8
incorporated root 7
subject or object 6
derivational consonant 5
tense/mood affix 4
neuter class subject or object 3
tense/mood consonant 2
subject or object affix 1
verb root 0

The choice of position of subject or object affixes also acts as a sort of stem-deriving
element. There are nine productive classes of subject/object agreement marker
configurations. Each verb belongs to one of these, but it is not possible to know to which
based on grammatical principles. It is a lexical choice, like the 1st vs. the 2nd conjugation
in Russian. So in the Ket verb it is not possible to separate derivation from inflection in a
linear fashion. Affixes are not concatenated; rather, they are placed in pre-existing
position class slots. Thus, Ket does not comply with the criterion of linear ordering of
derivational and inflectional affixes. An example is given in (29):

she dragged it (using a conveyance)

Rubino gives an example of interspersed derivational and inflectional affixes in Ilocano:




is derivational (AV- AGENTIVE voice)

is inflectional based on ag- (Perfective)
is derivational (Comitative)
is inflectional (Reciprocal)

Another example is provided by the agglutinating verbal morphology of Kujamaat Jola.

Fudeman (in Aronoff & Fudeman 2005: 140-141) gives the following description:
The core of the structure of a verb is the lexical stem, either simple or derived. The stem
is followed by position 1 suffixes, including aspectual and negative markers, as well as a

derivational directional suffix and the second members of the past subordinate and first
person plural inclusive circumfixes. Position 1 suffixes are followed by the position 2
suffixes: the passive marker, object pronominals (direct, indirect, or both), and noun
emphasis marker. Finally, the third position is filled by verb reduplication and the simple
subordinate marker. Immediately preceding the verb stem are the subject markers and
relative pronouns (position 1 prefixes); the leftmost position (position 2) is filled by the
resultative, the resultative negative, the negative imperative, and the past subordinate

In addition, Fudeman (ibid.: 96) notes that the derivational suffix -u from can follow
inflectional markers, as exemplified in (31):

nri- e
3AGR- arrive- HAB- from- REDUP
He habitually arrives from

In Udihe, the integrity of a verbal compound may be violated by the negative auxiliary
which is placed between the constituents of the compound as in the case of the compound
zemui ba (get hungry) in the following example (Nikolaeva & Tolskaya 2001: 326):

He is not hungry


In the same language, the DIMINUTIVE suffix ziga is preceded by the plural affix
(Nikolaeva & Tolskaya 2001: 183):


The Plural-Diminutive affix order is more frequent than one might expect. A good crosslinguistic overview is provided by Derzhanski (2005).
Stump (2005: 57-58) provides another counter-example from Breton where
inflected plural forms can serve as a basis for the derivation of denominal verbs and
(34) sg. delienn leaf
sg. goz mole
sg. maen rock
sg. Loer sock

pl. delio

deliauoi to grow leaves

pl. gozed
pl. mein
pl. lero

gozeta to hunt for moles

meinek full of rocks
dilero sockless


Furthermore, Stump (ibid.) also gives examples of plural DIMINUTIVES in which two
plural number exponents appear, each on either side of the DIMINUTIVE morpheme:

bag-o-ig- o

little boats
As he stresses, similar constructions occur in other languages:


tmt little trees

mazivarume big men
zasanimlex little bridegrooms13

Let us also mention productively formed Noun+Noun compounds in Telugu (Pingali,

pers. comm..) with a plural suffix inside them, for example:


In such cases, the first noun is plural if the interpretation can be plural. Similarly, the first
constituent in Finnish Adjective+Adjective compounds is in Genitive (Koivisto, pers.


Finally, let us refer to the formation of exocentric compounds of the garde-manger type in
Spanish. Chung (1994: 4) shows that one of two productive types of exocentric
compounds in Spanish has the form combining third person singular indicative of a verb
plus a plural noun, a form which contradicts the principles of level-ordering, because
inflection precedes the word-formation process of compounding:14

(el) cuentagotas
eye dropper
(el) espantapjaros

20. Whereas roots have the most varied shapes within any given language, affixes of
derivational morphology show less variation, affixes of inflectional morphology least,
i.e., their morpheme structure rules/constraints are most restrictive.

For more counter-examples of a still different nature, see Stump (2005) and Rainer (1996).
On inflection inside derivation in Spanish and Portuguese, see also Rainer (1996).

This criterion seems to be typologically conditioned because, as noted by Dressler

himself (1989: 9), it is much less clear for agglutinating languages than for inflecting
and introflecting (Semitic) languages []. The data from our core sample of languages
show that 71% of suffixing languages and 74% of prefixing languages are characterized
by allomorphy, that is to say, a fairly high number.
This list of criteria proposed by Dressler may be extended by the following of
Scalises (1988):
21 The criterion of headedness, assuming that only derivational suffixes can function as
heads while inflectional morphemes are not heads.
This has been much discussed in the literature, ranging from Marchands (1967) idea of
determinantdeterminatum structure of the word-formation syntagma, through Williams
(1981) Righthand Head Rule, Selkirks (1982) Revised Right-hand Rule or Di Sciullo &
Williams (1987) concept of relativized head, with the latter two admitting inflectional
morphemes to function as heads. It should be noted, however, that morphologists differ in
their concept of headedness.15
22. The readjustment rules which apply to the output of derivational rules are different
from the readjustment rules which apply to the output of inflectional rules.
The point is that there are readjustment rules which cannot adjust both derived and
inflected words with the same structural description, as in the following example (Scalise
1988: 573):

dialo[g]o >


blind PL. blindness
dialo[d]ico dialog PL. dialogic

23. Derivational suffixes and inflectional morphemes behave differently in relation to the
atom condition.
Scalise (1988: 575) shows that the addition of a derivational suffix (in contrast to an
inflectional suffix) in Italian may be conditioned by the presence of a prefix:

conciare *conciamento to tan

acconciare acconciamento to adjust adjustment

Furthermore, in the structure of WORD] +A] +B] the addition of a derivational suffix
does not depend on the properties of WORD (in accordance with the Atom Condition). In
contrast to it, Scalise points out Carstairs example (1984: 83) of deponent verbs in
Latin, such as sequor to follow whose form differs from those of non-deponent verbs:



sequ] + e] + ba] + mur] we were following

rege] + ba] + mus] we were ruling,

For the diversity of views of headedness see, for example, Marchand (1967), Williams (1981), Lieber
(1981), Zwicky (1985), Bauer (1990), and tekauer (2001).

where the selection between mur/mus depends on whether a verb is deponent as in (42a.)
or non-deponent as in (42b.).
24. Inflectional structures are different from derivational structures.
Scalise (1988: 577) gives the following representation of inflectional affixes:


a b c

It means that if we have two inflectional morphemes to the right of a word, the simple
attachment of the first does not give us an existing word (ibid: 577), which is
exemplified by the Latin ama + v + o I was loving, where v = tense and o =
person/number. From this it follows that the form without the final o cannot exist by
itself. Since Scalise believes, in our view correctly, that it is the node INFL which is
preassociated with the word rather than specific morphemes, he prefers the following
representation of inflection:

where INFL varies according to the specific word-class.

This observation can be related to Booijs (2006: 657) remark concerning IndoEuropean languages, in which one and the same morpheme frequently cumulates several
grammatical meanings (cumulative exponence). This feature is not typical of derivational
morphemes. This view corresponds with that by Anderson (1992: 76), notably that
portmanteaux are much rare in derivation (if indeed such elements exist at all) and that
there do not ever seem to be elements which combine inflectional and derivational
categories in the same portmanteaux.
While it is true that cumulative exponence is chracteristic of inflectional systems,
predominantly those of fusional languages, Ricca (2005) shows that portmanteaux
morphemes combining two derivational categories as well as derivational and inflectional
categories do exist. The existence of portmanteaux morphemes cumulating inflectional
and derivational categories is an important argument againstr strict separation of
inflectional morphology and derivational morphology. Our data also provide us with this
sort of examples. Zwarts (2007 pers. comm.)(pers. comm.) shows that in Endo, a
Southern Nilotic language, the derivation process based on changing a vowel from [ATR]
to [+ATR] results in Agent nouns in Plural:

kwaang to cook kwng cooks

Furthermore, if conversion is conceived as zero-derivation, then English conversions like

poorN illustrate a combination of Patient and Plural.
Riccas considerations are based on Romance languages; he provides several
types of fusion of derivational and inflectional categories, such as the French denominal
adjectives in -[al] as in national national; its masculine plural is formed by means of [o] (nationaux). The suffix not only derival denominal adjectives (derivational category)
but also combines the inflectional categories of masculine and plural Ricca 2005: 207).
Examples of fusion of two derivational categories, such as Agent + Male/Female
and and Agent + Augmentative/Excessive are given in () and (), respectively:
() a. It. gioca-(re) gioca-tore
to play
b. Fr. vend-(re) vend-eur
to sell

female player
female seller (2005: 200)

() It. mangi-one heavy eater < mangiare eat (ibid.: 202)

Derzhanski (2005) provides additional cross-linguistic evidence of the
derivation+inflection type of cumulative exponence, in particular, by giving examples of
the Plural+Diminutive cumulation in languages of various families (e.g. Bulgarian
(Slavonic), Fula (Atlantic-Congo), Swahili (Bantu), Asmat (Trans-New Guinea).
25. Perlmutter (1998) and Anderson (1992) speak of split-morphology. In their view,
word-formation is pre-syntactic and inflection post-syntactic, which suggests a clear-cut
difference between these two processes. This view contradicts the then established
lexicalist hypotheses according to which all morphological processes (inflectional and
derivational) are pre-syntactic, and take place in the Lexicon. But the split-morphology
hypothesis has been called into question by a number of linguists (cfr. Note 16).
26. Aphasia of the type called agrammatism seems to involve (for at least one class of
patients) a deficiency in ability to construct and manipulate syntactic structure and
inflectional morphology, while the rest of the lexicon (including derivational
morphology) remains relatively intact (Anderson 1992: 75).
27. Adams (1973: 13) observation that the number of derivational affixes in English is
much higher than that of inflectional affixes is language-specific and cannot be
generalized. The proportion between the derivational and inflectional affixes heavily
depends on the morphological type of a language. We are not aware of any empirical
cross-linguistic research into this issue.
Let us finally add some of Planks (1994) criteria not mentioned by the previous
authors. Each of them is based on the postulate of a continuum with two poles of
typicality: the inflectional and the derivational ones, and with many less typical cases in
both inflection and derivation.

28. Inflectional morphemes are typically added to existing bases. Derivation can also
make use of non-existing bases.
Thus, 3rd person singular present indicative in English is always based on a combination
of an existing verbal base plus suffix. On the other hand, within the realm of inflection,
there are plural forms without any corresponding singular (belonging-s, outskirt-s,
particular-s, etc.). Our sample provides us with this kind of derivations, too. In Hausa, for
example, some AGENT nouns are based on non-existing verbs. The verb *zinaata does not
actually occur as such, but one has to postulate it as a step in the derivation to make any
sense of what is happening (Newman, pers. comm.).

maznac adulterer < zn adultery

The Tzotzil suffix -tik (its a place of) combines with intransitive stems which never
occur independently (Cowan 1969:102-103) as in (46):

nkaxtik they are sitting down < *nkax

Furthermore, in the section on noun incorporation we give examples of stems in Slavey

and Kwakwala that do not exist outside noun incorporation. Finally, Bauer (to appear)
mentions compounds in Danish whose one element has no independent existence:
(47) a. bom-uld


As noted by Rainer (pers. comm..), similar examples can be found in English,

such as bett-er.
29. The positioning and the segmental, suprasegmental, syllabic, and morphemic
structure of the exponents of inflectional categories are relatively similar, those of
derivational categories are dissimilar (Plank 1994: 1676).
Plank maintains that (a) exponents of inflectional categories, if expressed by affixes, are
exclusively suffixes; (b) they are phonetically similar; (c) they are mostly expressed by
monosyllabic or nonsyllabic exponents; (d) they neither carry nor change the position of
While this is true of English, it does not hold universally, and there are languages
which express inflectional categories by prefixes too: Cutler, Hawkins & Gilligan (1985:
747) admit the existence of languages with inflectional prefixes, and this book cites
examples of prefixal inflection which exists next to suffixal inflection. An example from
Ket (Vajda, pers. comm.) is given in (48):


She dragged it (using a conveyance)

A similar example comes from Zulu (Van der Spuy, pers. comm.):

We were no longer going to call him/her

The issue of phonetic similarity does not hold universally either. Inflectional languages
are good counter-examples. On the other hand, derivational affixes may show homonymy
too: consider (not only) English polysemantioc/homonymous suffixes standing for
30. When considered in isolation, some exponents of derivational categories may
resemble free morphemes in terms of their internal structure (Plank 1994: 1676).
Plank illustrates this shape-based criterion with derivational affixes -ize, -(i)fy, -(at)ion,
-ling, and mini- that look like genuine words (ibid.).