Zgryźliwość kojarzy mi się z radością, która źle skończyła.
12. Naturalness and Morphological Change : The Handbook of Historical Linguistics : Blackwell Reference Online
12/11/2007 03:36 PM
12. Naturalness and Morphological Change
WOLFGANG U. DRESSLER
Subject
Linguistics
»
Historical Linguistics
Key-Topics
morphology
DOI:
10.1111/b.9781405127479.2004.00014.x
During the last decades, “natural” has often been used by linguists in an inductive or even anecdotal way as
a synonym of “intuitively plausible” or of “cross-linguistically frequent,” in reference to both synchrony and
diachronic change. In more theoretical views, it often overlaps with cognitively simple (cf. Anttila, this
volume), elementary and therefore universally preferred, and with Praguian (especially Jakobson's) notions of
markedness (where unmarked loosely corresponds to natural).
Naturalness is a relative, gradient concept: a phenomenon X is more or less natural than Y. For example,
within English plural formation, the modern plural
cow-s
is a more natural plural of
cow
than its precedent
correspondents
cyne/kin(e
). Change from a less natural to a more natural morphological phenomenon may
then be called “natural/preferred/unmarked morphological change.” Thus, naturalness studies in diachronic
change usually do not deal with absolute constraints on change but minimally with tendencies or maximally
with “soft constraints” or defaults. Preferences are of a functional nature (cf. Heine and Mithun, this volume)
and ultimately founded in extralinguistic bases.
Tendencies of morphological change have been investigated by many with recourse to some notion of
naturalness, often with a shady transition from notions of naturalness to those of simplicity, that is, to views
that natural morphological change results in simplification. But only few have done that in any systematic
way, notably Bailey (1982) in his “Developmental Linguistics,” Plank (1981), and Keller (1990). Most
systematic work, however, has been done within the framework of “Natural Morphology” (NM) or in reference
to it. A second reason for using the NM framework in this chapter is the important role that diachrony has
always played in this approach.
This chapter will center on grammar-initiated, natural change, first according to universal, system-
independent morphological naturalness/markedness (section 2.1), in regard to the parameters of
constructional iconicity, morphosemantic, and morphotactic transparency (including preferences for
continuity and word bases), and binarity, whereas the parameters of (bi)uniqueness, indexicality, optimum
shape are dealt with in sections 2.2 and 3. Conflicts between universal parameters (section 2.2) either within
morphology or with other components explain unnatural changes. After type-adequacy as a filter on change
(section 3), language-specific, system-dependent naturalness (system-adequacy; section 4.1) is dealt with,
followed by interaction between the three subtheories of universal versus typological versus system-
dependent naturalness (section 4.2). Finally, in section 5, work on change initiated by grammarexternal
factors is briefly mentioned, viz. on contact-induced change, language decay, creolization, and language
planning in terminology.
1 The Framework of Natural Morphology (NM)
The theory of NM
1
originated with the integration of concepts of Praguian markedness and phonological
naturalness (cf. Stampe 1969) into the study of morphology and the conception of naturalness conflicts by
http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747914
Page 1 of 8
12. Naturalness and Morphological Change : The Handbook of Historical Linguistics : Blackwell Reference Online
12/11/2007 03:36 PM
naturalness (cf. Stampe 1969) into the study of morphology and the conception of naturalness conflicts by
Mayerthaler (1977) and Dressler (1977). In the same year, they, along with Wurzel and Panagl, formulated, at
the 1977 LSA Summer Institute at Salzburg, a common platform later extended into Dressler et al. (1987),
where morphological change occupies a prominent place. More recently, naturalness has become a cover
term for a set of more specific terms to be defined in specific subtheories and to be derived from more
general semiotic, cognitive, and/or psychological concepts.
These subtheories proposed since 1977 are those of universal markedness or system-independent
morphological naturalness (cf. Mayerthaler 1981: 3), and of type-adequacy (cf. Dressler 1985a, 1988a:
section 4), preceded by a theory of system-dependent naturalness or system-adequacy (cf. Wurzel 1984:
section 5). Subtheories for interfaces with other areas of morphological naturalness were established for
morphonology (cf. Dressler 1985b) and morphopragmatics (cf. Dressler and Merlini Barbaresi 1994), and
Mayerthaler is establishing one for morphosyntax, within his theory of Natural Syntax (cf. Mayerthaler et al.
1994).
The focus of diachronic investigations has been on those types of morphological change where the above
subtheories can explain most, that is, where no external theories appear to be crucially involved (but see
sections 5 and 6). This appears to be the case when change is supposed to be mainly triggered by forces
which lie within grammar or become manifest in first language acquisition. Such change has been called
“grammar-initiated change” (the title of Wurzel 1994b). This term is justified in view of the (admittedly
simplified) dichotomy between origin and spread of change. Whereas social factors are of great importance
in spread (cf. Guy, Pintzuk, and Wolfram and Schilling-Estes, this volume), they are of little or, maybe, even
no importance in certain types of change which are subsumed under the term grammar-initiated change.
2 Universal, System-Independent Morphological Naturalness/Markedness
This subtheory is a preference theory (cf. Vennemann 1983; Dressler 1999), which establishes deductively
degrees of universal preferences on a restricted number of naturalness parameters. Here naturalness refers
specifically to what is universally preferred on one given parameter. Parameters and their preference degrees
are deduced from their extralinguistic bases. For each parameter, the two main diachronic predictions are:
i the more natural a phenomenon is on a given morphological parameter, the more stable, that is, the
more resistant it should be to morphological change (but not necessarily to phonological or syntactic
change);
ii if, of two comparable morphological options X and Y, X is more natural than Y on a given parameter
Z, then natural /unmarked change of X to Y should be more likely to occur than the reverse,
unnatural/marked change Y to X. This predicted direction of change does not imply the absurd
position of overall change toward more and more naturalness, but represents the hypothesis of local
improvement on just one parameter (cf. Vennemann 1990), which goes back to Jespersen's (1949)
idea of local efficiency of change (more in sections 2.2 and 4.2 below).
Empirical testing of the predictions of (i) and (ii) should reach statistical significance (weak hypothesis) or
they should even predict the default (strong hypothesis).
2.1 Universal naturalness parameters
Iconicity is the best-known semiotically derived parameter. Most important for morphology is its
subparameter of constructional iconicity (cf. Mayerthaler 1981). According to Peirce's (1965) subdivision of
icons, the various types of English plural formation can be classified as follows:
oaf-s
is diagrammatic, that
is, most iconic, because there is an analogy between morphotactic addition of a plural marker and addition
of the morphosemantic feature of plurality; umlaut plural
feet
with vowel modification (from
foot
), instead of
addition, is only metaphoric (i.e., with weaker iconicity);
loav-es
(from sg.
loaf
) lies in between;
sheep
is
non-iconic; the counter-iconic operation of subtraction can be illustrated with subtractive plural
hon
of sg.
hond
‘dog’ in a Franconian German dialect: naturalness decreases accordingly on this scalar parameter of
constructional iconicity.
If we test the predictions (i) and (ii) of section 2 with recent diachronic change, then the diagrammatic type
oaf-s
is the only productive and stable one in English. It acquires new items from the type
loav-es
, as
attested by the variation
roof-s/roov-es
, cf.
leav-es
versus
The Toronto Maple Leaf-s;
similarly the umlaut
http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747914
Page 2 of 8
12. Naturalness and Morphological Change : The Handbook of Historical Linguistics : Blackwell Reference Online
12/11/2007 03:36 PM
attested by the variation
roof-s/roov-es
, cf.
leav-es
versus
The Toronto Maple Leaf-s;
similarly the umlaut
plural
lice
has a recent variant
louse-s
(with new meaning and diagrammatic plural), and, earlier on, pl.
cyne
has been replaced by
cow-s;
the anti-iconic Franconian plural type
hon
has become unproductive and loses
items to diagrammatic additive plural formation.
From the semiotic preference for transparency, the two parameters of morphosemantic and morphotactic
transparency
2
are derived: on the parameter of morphosemantic transparency, full transparency means fully
compositional meaning, as is generally the case with inflectional meanings. For example, the meaning of
cow-s
equals the meanings of
cow
and of plurality. There is, however, morphosemantic opacity in cases of
parasitic formation (cf. Aronoff 1994: 33), such as in the formation of the Latin periphrastic future
canta-t-
urus sum
‘I'll sing,’ formed via the stem of the past participle
canta-t-us
, whereas there is no meaning of
past in the periphrastic future. Since, in general, there is an iconic preference for a pairing of transparent
meaning and form, this morphosemanti-cally opaque periphrastic future of Latin has been replaced by more
transparent ones in the Romance languages, especially the type Fr.
chanter-ai
, Sp.
cantar-é
, etc., which still
shows its origin from Inf. (Lat.
cantare
, Fr.
chanter
, Sp.
cantar
) and the auxiliary Lat.
habeo
, Fr.
ai
, Sp.
he
‘I
have.’
On the parameter of morphotactic transparency, the most natural forms are those where there is no
opacifying obstruction to ease of perception. Purely phonological processes opacify very little, for example,
phonological surface palatalization in the Polish pejorative nom.pl.
Polak-i
of sg.
Polak
‘Pole.’ More
morphotactic opacity occurs in the frequent intervention of morphonological rules, such as in Polish
morphonological palatalization, for example, normal nom.pl.
Polac-y
. Most opaque is suppletion, as in E.
am, is, are, was
. Diachronically, opaque forms are less stable than more transparent ones and easily
replaced, unless high token frequency facilitates memorization of opaque forms and hence helps to preserve
them. Thus in English and German, morphotactically opaque strong verbs have been increasingly replaced by
transparent weak verbs.
3
Most resistant to change are the very frequently used auxiliaries, modal verbs, etc.
Another way of rendering plurals more transparent (and often more iconic) is hypercharacterization (cf.
Malkiel 1957), similar to children's plural
feet-s
, where a morphological category is doubly marked, the
second time in a more iconic and/or transparent way. Diachronic examples are Middle English pl.
child-er /
child-re > child-ren;
comparative
worse >
colloquial
wors-er
.
A consequence of the preference for morphotactic transparency is also the preference for continuous (rather
than discontinuous) elements. Therefore suf-fixation and prefixation is preferred over infixation
(discontinuous base) or circumfixation (discontinuous affix). The diachronic instability of infixes is evident in
the history of Indo-European languages: for example,
-n
- infixes of the Latin present stem, as in
fra-n-g-o
‘I break,’ perfect
freg-i
, past participle
frac-tus
, have become part of the immutable verb stem, as in It.
frang-o, fransi, franto
(cf. Klausenburger 1979: 49–54). Also, diminutive/hypocoristic suffixes are generally
preferred to comparable infixes, such as in Sp.
Cesar-ito
versus
Ces-ít-ar
, hypocoristics of the name
Cesar
(cf. Méndez Dosuna and Pensado 1990), from the late Latin suffix
-ittus.
4
Related to the preference for morphotactic transparency is also the word-base preference (cf. Dressler
1988b): the most natural base of a morphological rule is a word, because this is, in semiotic terms, a
primary sign and thus a very transparent unit. Smaller bases (stems, roots) or more complex bases (phrases,
sentences) are dispreferred. This makes, among the two German plural variants of
Pizza
, word-based
Pizza-
s
more natural than root-based
Pizz-en
. As predicted, the type
Pizza-s
is now the preferred one (cf. Wurzel
1984; Janda 1991, 1999a, with extensive literature).
The word-base preference also renders compounds like
do-it-yourself movement
(with a sentence as first
member) less natural than
eye movement
(where the first member is a word). Such sentence-based
compounds seem to originate only in literary languages as marginal neologisms or often only occasionalisms
(ad-hoc /nonce formations) and are socioculturally motivated, relatively unnatural complications, which may
be compared to the rise of complicated politeness forms.
A similar motivation can be found for violations of the binarity preference: grammatical relations are
preferentially binary (based on the binary nature of neurological information transmittance). In syntagmatic
relations, the preferred patterning consists in concatenating one element to one base. This preference holds,
for example, for compounding, including coordinate/copulative compounding, as in
queen-mother
or
prince-consort
. This preference is violated, due to entirely extralinguistic reasons, in denominations of flags,
for example,
red-white-red
for the Austrian and Peruvian flags. Similarly, Sanskrit coordinate compounds
http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747914
Page 3 of 8
12. Naturalness and Morphological Change : The Handbook of Historical Linguistics : Blackwell Reference Online
12/11/2007 03:36 PM
for example,
red-white-red
for the Austrian and Peruvian flags. Similarly, Sanskrit coordinate compounds
start to have more than two members only in later, stylistically marked texts or due to extralinguistic
reasons, as in the denomination of the four main castes
brãhma
a-k atriya-vi -śū
drãh
‘the set of
brahmins, warriors, Vaiśyas, Śudras.’
The preference parameters of bi-uniqueness (uniform symbolization), indexicality, and optimum
shape/extension of morphological word forms will be dealt with in sections 2.2, 3 and 4.
2.2 Conflicts between universal parameters
Unnatural/marked changes can be partially explained by recourse to conflicts between parameters either of
the same grammatical component (here morphology) or of different components (morphology versus
phonology/syntax). Let us start with morphology-internal parameter conflicts.
First, we must introduce another parameter derived from Peircean semiotics, the parameter of indexicality.
On this parameter, adjacency is preferred to distance, both syntagmatically and in terms of rule application.
This favors the diachronic change of rule telescoping (cf. van Marle 1990: 270; Dressler 1996a: 97), insofar
as the morphological surface form can be immediately derived from the base instead of having to be
preceded by intermediate rules and false steps. An example is the genesis of German circumfixes, as in the
poetic occasionalism
Ge-khaki-t-e
‘having a khaki-colored uniform’ (= GIs, Arno Schmidt). The intermediate
steps are: noun
Khaki
→ verb
*khaki-en
, past participle suffixation of this verb →
*khaki-t
, prefixation →
*ge-khaki-t
, conversion of past participle → adjective. Rule telescoping allowed direct derivation of the
adjective from the noun via circumfixation and a morphosemantically transparent relation between nominal
base and derived adjective. This, however, created the class of morphotactically opaque circumfixes.
Another, better-known, semiotically based parameter consists in the preference for bi-uniqueness, or at
least uniqueness, as opposed to ambiguity. Bi-uniqueness
5
holds if one and the same form has always the
same meaning and vice versa, uniqueness if this holds in only one of the two directions. Biuniqueness is
difficult to achieve because of economy of sign shapes and is thus often violated, for example in cases of
hypercharacterization (see section 2.1), where one and the same meaning, for example plurality, is
expressed twice instead of once only. In hypercharacterization, morphotactic transparency wins out over bi-
uniqueness.
Since conflicts between parameters of morphology and those of either phonology or syntax are dealt with,
albeit from a different perspective, in Janda's and Joseph's contributions to this volume, my discussion can
be limited to indicating specifics of the naturalist approach.
In interaction with phonology (cf. Dressler 1985b, 1996), in a first stage, sound laws apply with little or no
respect for morphology. These purely phonological processes (postlexical, postcyclic rules in the terminology
of Kiparsky's Lexical Phonology) opacify morphotactics very little, as in the surface palatalization example
from Polish, in section 2.1. Similarly, more morphotactic opacity occurs when morphonological rules
intervene, as in the morphonological palatalization example in Polish. This increasing morphologization of
phonological rules represents, on the one hand, an unnatural /marked change on the parameters of
morphotactic transparency and constructional iconicity (cf. section 2.1) as well as on the parameter of
phonological (bi-)uniqueness, because Polish [k'] uniquely derives from underlying /k/, whereas [c] may
derive from either /c/ or /k/. On the other hand, a context-sensitive phonological process possesses much
phonological indexicality (since it refers to its phonological context) and little morphological indexicality
(since it refers to a following suffix), whereas for morphonological rules morphological indexicality is much
more important than phonological indexicality, and morphological rules possess only morphological
indexicality. Thus increasing morphologization of phonological rules represents a shift from phonological to
morphological indexicality, explainable by semiotic priority of morphology over phonology. This explains the
unidirectionality of change from phonological to morphological rules (cf. Dressler 1996). The questions of
the conditions under which natural change of indexicality may outweigh unnatural change on several other
parameters will be reconsidered in sections 3 and 4.
A similar approach applies to interaction with syntax and to unidirectionally increasing grammaticalization
from syntax to morphology, a natural change on the parameters of morphotactic transparency (continuous
forms are preferable to discontinuous ones) and of indexicality, insofar as fixed morpheme order is
preferable to alternating order of words or clitics, as in the contrast between periphrastic constructions of
the type I
was read-ing, Wasn't I read-ing?
and their Italian equivalents
Legge-v-o, Non legge-v-o?
http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747914
Page 4 of 8
12. Naturalness and Morphological Change : The Handbook of Historical Linguistics : Blackwell Reference Online
12/11/2007 03:36 PM
Morphologization of phonological and syntactic patterns has been understood as the basic source of
morphological patterns by Wurzel (1984: 102ff, 212, 1987: 69) in his claim that truly morphological change
is only reactive (criticized by Dressler 1997a, 1999). Of diachronic relevance is also Wurzel's (1996a, 1996b)
perspective on “the age of morphological constructions,” which is correlated to stages of development with
different properties on morphological parameters.
3 Type-Adequacy
Inspired by Skalička's (1979) views on ideal language types which consist of properties which favor (or
disfavor) one another, we can reinterpret language types as (alternative) sets of consistent responses to
naturalness conflicts (section 2.2).
6
Since not all of the most natural options on all parameters can be
combined within one language, naturalness on certain parameters must be, so to say, sacrificed for greater
naturalness on others (cf. Dressler 1985a, 1985c, 1988a; Sgall 1988). Thus the agglutinating type (as best
represented by Turkish) has the advantages of much constructional iconicity, morphosemantic and
morphotactic transparency, and bi-uniqueness, but deviates with its often very long word forms from the
optimal shape of morphological words (one prosodic foot - another universal preference) and does not fully
achieve fixed morpheme order (unnatural on the parameter of indexicality), whereas the opposite holds for
the inflecting-fusional type. In this way, universally rather unnatural options may be typologically adequate if
they fit the properties of the respective language type.
A morphological change is type-adequate if one of the two following conditions is met:
i change does not modify typological properties - this is a typologically conservative change;
ii change correlates with other changes which implement an overall typological change of the
respective language - this is a typologically innovative change.
The second type of change can be exemplified with correlated changes from Latin to Romance inflectional
morphologies, that is, from a strong inflecting language to weak inflecting languages (with a greater role for
the isolating language type). Also, Estonian has changed from an agglutinating type to a predominantly
inflecting-fusional language, with less constructional iconicity, morphosemantic and morphotactic
transparency, and (bi-)uniqueness, but with fixed morpheme order and greater approximation to the optimal
shape of morphological words (di- and trisyllabicity), when compared with Finnish.
Type-adequacy affects the solution of conflicts between parameters. In section 2.2 we have discussed
morphologization of phonological rules and the question of why unnatural change on the parameters of
morphotactic transparency, constructional iconicity, and (bi-)uniqueness can be outweighed by other factors.
This happens especially in languages of the inflecting-fusional type where naturalness on these parameters
is sacrificed in favor of naturalness on other parameters, such as indexicality. In other words, relative
unnaturalness on these parameters obtains little weight and can therefore be outweighed by greater
naturalness on the parameter of indexicality. Similarly, the morphologically fairly unnatural class of infixes
(section 2.1) originates due to phonological factors and is typologically restricted (cf. Moravcsik 1977;
Méndez Dosuna and Pensado 1990).
Also suppletion, the most unnatural option on the parameter of morphotactic transparency, originates in
inflecting-fusional rather than in agglutinating languages. The many origins of suppletion (cf. Ronneberger-
Sibold 1990) must be strictly differentiated from the factors of maintenance (i.e., stability) of suppletion:
those suppletive forms are best preserved which have high token frequency (thus storage is more
economical than composition and decomposition by rule, e.g., with auxiliaries), have idiosyncratic meanings
(e.g., learned connotations, such as Fr.
Fontainebleau
, adjective
Bellifontain
of artificial humanistic origin),
are not natural members of large classes (e.g., auxiliaries in contrast to main verbs), or support each other
analogically, as in antonyms (e.g.,
good, bad
, comparatives
better, worse
).
4 System-Dependent Naturalness
4.1 System-adequacy
Language-specific, system-dependent naturalness, as conceived by Wurzel (1984) for systems of inflectional
morphology (modifications in Dressler 1997b), represents what is normal or system-congruous (system-
http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747914
Page 5 of 8
zanotowane.pl doc.pisz.pl pdf.pisz.pl hannaeva.xlx.pl
12/11/2007 03:36 PM
12. Naturalness and Morphological Change
WOLFGANG U. DRESSLER
Subject
Linguistics
»
Historical Linguistics
Key-Topics
morphology
DOI:
10.1111/b.9781405127479.2004.00014.x
During the last decades, “natural” has often been used by linguists in an inductive or even anecdotal way as
a synonym of “intuitively plausible” or of “cross-linguistically frequent,” in reference to both synchrony and
diachronic change. In more theoretical views, it often overlaps with cognitively simple (cf. Anttila, this
volume), elementary and therefore universally preferred, and with Praguian (especially Jakobson's) notions of
markedness (where unmarked loosely corresponds to natural).
Naturalness is a relative, gradient concept: a phenomenon X is more or less natural than Y. For example,
within English plural formation, the modern plural
cow-s
is a more natural plural of
cow
than its precedent
correspondents
cyne/kin(e
). Change from a less natural to a more natural morphological phenomenon may
then be called “natural/preferred/unmarked morphological change.” Thus, naturalness studies in diachronic
change usually do not deal with absolute constraints on change but minimally with tendencies or maximally
with “soft constraints” or defaults. Preferences are of a functional nature (cf. Heine and Mithun, this volume)
and ultimately founded in extralinguistic bases.
Tendencies of morphological change have been investigated by many with recourse to some notion of
naturalness, often with a shady transition from notions of naturalness to those of simplicity, that is, to views
that natural morphological change results in simplification. But only few have done that in any systematic
way, notably Bailey (1982) in his “Developmental Linguistics,” Plank (1981), and Keller (1990). Most
systematic work, however, has been done within the framework of “Natural Morphology” (NM) or in reference
to it. A second reason for using the NM framework in this chapter is the important role that diachrony has
always played in this approach.
This chapter will center on grammar-initiated, natural change, first according to universal, system-
independent morphological naturalness/markedness (section 2.1), in regard to the parameters of
constructional iconicity, morphosemantic, and morphotactic transparency (including preferences for
continuity and word bases), and binarity, whereas the parameters of (bi)uniqueness, indexicality, optimum
shape are dealt with in sections 2.2 and 3. Conflicts between universal parameters (section 2.2) either within
morphology or with other components explain unnatural changes. After type-adequacy as a filter on change
(section 3), language-specific, system-dependent naturalness (system-adequacy; section 4.1) is dealt with,
followed by interaction between the three subtheories of universal versus typological versus system-
dependent naturalness (section 4.2). Finally, in section 5, work on change initiated by grammarexternal
factors is briefly mentioned, viz. on contact-induced change, language decay, creolization, and language
planning in terminology.
1 The Framework of Natural Morphology (NM)
The theory of NM
1
originated with the integration of concepts of Praguian markedness and phonological
naturalness (cf. Stampe 1969) into the study of morphology and the conception of naturalness conflicts by
http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747914
Page 1 of 8
12. Naturalness and Morphological Change : The Handbook of Historical Linguistics : Blackwell Reference Online
12/11/2007 03:36 PM
naturalness (cf. Stampe 1969) into the study of morphology and the conception of naturalness conflicts by
Mayerthaler (1977) and Dressler (1977). In the same year, they, along with Wurzel and Panagl, formulated, at
the 1977 LSA Summer Institute at Salzburg, a common platform later extended into Dressler et al. (1987),
where morphological change occupies a prominent place. More recently, naturalness has become a cover
term for a set of more specific terms to be defined in specific subtheories and to be derived from more
general semiotic, cognitive, and/or psychological concepts.
These subtheories proposed since 1977 are those of universal markedness or system-independent
morphological naturalness (cf. Mayerthaler 1981: 3), and of type-adequacy (cf. Dressler 1985a, 1988a:
section 4), preceded by a theory of system-dependent naturalness or system-adequacy (cf. Wurzel 1984:
section 5). Subtheories for interfaces with other areas of morphological naturalness were established for
morphonology (cf. Dressler 1985b) and morphopragmatics (cf. Dressler and Merlini Barbaresi 1994), and
Mayerthaler is establishing one for morphosyntax, within his theory of Natural Syntax (cf. Mayerthaler et al.
1994).
The focus of diachronic investigations has been on those types of morphological change where the above
subtheories can explain most, that is, where no external theories appear to be crucially involved (but see
sections 5 and 6). This appears to be the case when change is supposed to be mainly triggered by forces
which lie within grammar or become manifest in first language acquisition. Such change has been called
“grammar-initiated change” (the title of Wurzel 1994b). This term is justified in view of the (admittedly
simplified) dichotomy between origin and spread of change. Whereas social factors are of great importance
in spread (cf. Guy, Pintzuk, and Wolfram and Schilling-Estes, this volume), they are of little or, maybe, even
no importance in certain types of change which are subsumed under the term grammar-initiated change.
2 Universal, System-Independent Morphological Naturalness/Markedness
This subtheory is a preference theory (cf. Vennemann 1983; Dressler 1999), which establishes deductively
degrees of universal preferences on a restricted number of naturalness parameters. Here naturalness refers
specifically to what is universally preferred on one given parameter. Parameters and their preference degrees
are deduced from their extralinguistic bases. For each parameter, the two main diachronic predictions are:
i the more natural a phenomenon is on a given morphological parameter, the more stable, that is, the
more resistant it should be to morphological change (but not necessarily to phonological or syntactic
change);
ii if, of two comparable morphological options X and Y, X is more natural than Y on a given parameter
Z, then natural /unmarked change of X to Y should be more likely to occur than the reverse,
unnatural/marked change Y to X. This predicted direction of change does not imply the absurd
position of overall change toward more and more naturalness, but represents the hypothesis of local
improvement on just one parameter (cf. Vennemann 1990), which goes back to Jespersen's (1949)
idea of local efficiency of change (more in sections 2.2 and 4.2 below).
Empirical testing of the predictions of (i) and (ii) should reach statistical significance (weak hypothesis) or
they should even predict the default (strong hypothesis).
2.1 Universal naturalness parameters
Iconicity is the best-known semiotically derived parameter. Most important for morphology is its
subparameter of constructional iconicity (cf. Mayerthaler 1981). According to Peirce's (1965) subdivision of
icons, the various types of English plural formation can be classified as follows:
oaf-s
is diagrammatic, that
is, most iconic, because there is an analogy between morphotactic addition of a plural marker and addition
of the morphosemantic feature of plurality; umlaut plural
feet
with vowel modification (from
foot
), instead of
addition, is only metaphoric (i.e., with weaker iconicity);
loav-es
(from sg.
loaf
) lies in between;
sheep
is
non-iconic; the counter-iconic operation of subtraction can be illustrated with subtractive plural
hon
of sg.
hond
‘dog’ in a Franconian German dialect: naturalness decreases accordingly on this scalar parameter of
constructional iconicity.
If we test the predictions (i) and (ii) of section 2 with recent diachronic change, then the diagrammatic type
oaf-s
is the only productive and stable one in English. It acquires new items from the type
loav-es
, as
attested by the variation
roof-s/roov-es
, cf.
leav-es
versus
The Toronto Maple Leaf-s;
similarly the umlaut
http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747914
Page 2 of 8
12. Naturalness and Morphological Change : The Handbook of Historical Linguistics : Blackwell Reference Online
12/11/2007 03:36 PM
attested by the variation
roof-s/roov-es
, cf.
leav-es
versus
The Toronto Maple Leaf-s;
similarly the umlaut
plural
lice
has a recent variant
louse-s
(with new meaning and diagrammatic plural), and, earlier on, pl.
cyne
has been replaced by
cow-s;
the anti-iconic Franconian plural type
hon
has become unproductive and loses
items to diagrammatic additive plural formation.
From the semiotic preference for transparency, the two parameters of morphosemantic and morphotactic
transparency
2
are derived: on the parameter of morphosemantic transparency, full transparency means fully
compositional meaning, as is generally the case with inflectional meanings. For example, the meaning of
cow-s
equals the meanings of
cow
and of plurality. There is, however, morphosemantic opacity in cases of
parasitic formation (cf. Aronoff 1994: 33), such as in the formation of the Latin periphrastic future
canta-t-
urus sum
‘I'll sing,’ formed via the stem of the past participle
canta-t-us
, whereas there is no meaning of
past in the periphrastic future. Since, in general, there is an iconic preference for a pairing of transparent
meaning and form, this morphosemanti-cally opaque periphrastic future of Latin has been replaced by more
transparent ones in the Romance languages, especially the type Fr.
chanter-ai
, Sp.
cantar-é
, etc., which still
shows its origin from Inf. (Lat.
cantare
, Fr.
chanter
, Sp.
cantar
) and the auxiliary Lat.
habeo
, Fr.
ai
, Sp.
he
‘I
have.’
On the parameter of morphotactic transparency, the most natural forms are those where there is no
opacifying obstruction to ease of perception. Purely phonological processes opacify very little, for example,
phonological surface palatalization in the Polish pejorative nom.pl.
Polak-i
of sg.
Polak
‘Pole.’ More
morphotactic opacity occurs in the frequent intervention of morphonological rules, such as in Polish
morphonological palatalization, for example, normal nom.pl.
Polac-y
. Most opaque is suppletion, as in E.
am, is, are, was
. Diachronically, opaque forms are less stable than more transparent ones and easily
replaced, unless high token frequency facilitates memorization of opaque forms and hence helps to preserve
them. Thus in English and German, morphotactically opaque strong verbs have been increasingly replaced by
transparent weak verbs.
3
Most resistant to change are the very frequently used auxiliaries, modal verbs, etc.
Another way of rendering plurals more transparent (and often more iconic) is hypercharacterization (cf.
Malkiel 1957), similar to children's plural
feet-s
, where a morphological category is doubly marked, the
second time in a more iconic and/or transparent way. Diachronic examples are Middle English pl.
child-er /
child-re > child-ren;
comparative
worse >
colloquial
wors-er
.
A consequence of the preference for morphotactic transparency is also the preference for continuous (rather
than discontinuous) elements. Therefore suf-fixation and prefixation is preferred over infixation
(discontinuous base) or circumfixation (discontinuous affix). The diachronic instability of infixes is evident in
the history of Indo-European languages: for example,
-n
- infixes of the Latin present stem, as in
fra-n-g-o
‘I break,’ perfect
freg-i
, past participle
frac-tus
, have become part of the immutable verb stem, as in It.
frang-o, fransi, franto
(cf. Klausenburger 1979: 49–54). Also, diminutive/hypocoristic suffixes are generally
preferred to comparable infixes, such as in Sp.
Cesar-ito
versus
Ces-ít-ar
, hypocoristics of the name
Cesar
(cf. Méndez Dosuna and Pensado 1990), from the late Latin suffix
-ittus.
4
Related to the preference for morphotactic transparency is also the word-base preference (cf. Dressler
1988b): the most natural base of a morphological rule is a word, because this is, in semiotic terms, a
primary sign and thus a very transparent unit. Smaller bases (stems, roots) or more complex bases (phrases,
sentences) are dispreferred. This makes, among the two German plural variants of
Pizza
, word-based
Pizza-
s
more natural than root-based
Pizz-en
. As predicted, the type
Pizza-s
is now the preferred one (cf. Wurzel
1984; Janda 1991, 1999a, with extensive literature).
The word-base preference also renders compounds like
do-it-yourself movement
(with a sentence as first
member) less natural than
eye movement
(where the first member is a word). Such sentence-based
compounds seem to originate only in literary languages as marginal neologisms or often only occasionalisms
(ad-hoc /nonce formations) and are socioculturally motivated, relatively unnatural complications, which may
be compared to the rise of complicated politeness forms.
A similar motivation can be found for violations of the binarity preference: grammatical relations are
preferentially binary (based on the binary nature of neurological information transmittance). In syntagmatic
relations, the preferred patterning consists in concatenating one element to one base. This preference holds,
for example, for compounding, including coordinate/copulative compounding, as in
queen-mother
or
prince-consort
. This preference is violated, due to entirely extralinguistic reasons, in denominations of flags,
for example,
red-white-red
for the Austrian and Peruvian flags. Similarly, Sanskrit coordinate compounds
http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747914
Page 3 of 8
12. Naturalness and Morphological Change : The Handbook of Historical Linguistics : Blackwell Reference Online
12/11/2007 03:36 PM
for example,
red-white-red
for the Austrian and Peruvian flags. Similarly, Sanskrit coordinate compounds
start to have more than two members only in later, stylistically marked texts or due to extralinguistic
reasons, as in the denomination of the four main castes
brãhma
a-k atriya-vi -śū
drãh
‘the set of
brahmins, warriors, Vaiśyas, Śudras.’
The preference parameters of bi-uniqueness (uniform symbolization), indexicality, and optimum
shape/extension of morphological word forms will be dealt with in sections 2.2, 3 and 4.
2.2 Conflicts between universal parameters
Unnatural/marked changes can be partially explained by recourse to conflicts between parameters either of
the same grammatical component (here morphology) or of different components (morphology versus
phonology/syntax). Let us start with morphology-internal parameter conflicts.
First, we must introduce another parameter derived from Peircean semiotics, the parameter of indexicality.
On this parameter, adjacency is preferred to distance, both syntagmatically and in terms of rule application.
This favors the diachronic change of rule telescoping (cf. van Marle 1990: 270; Dressler 1996a: 97), insofar
as the morphological surface form can be immediately derived from the base instead of having to be
preceded by intermediate rules and false steps. An example is the genesis of German circumfixes, as in the
poetic occasionalism
Ge-khaki-t-e
‘having a khaki-colored uniform’ (= GIs, Arno Schmidt). The intermediate
steps are: noun
Khaki
→ verb
*khaki-en
, past participle suffixation of this verb →
*khaki-t
, prefixation →
*ge-khaki-t
, conversion of past participle → adjective. Rule telescoping allowed direct derivation of the
adjective from the noun via circumfixation and a morphosemantically transparent relation between nominal
base and derived adjective. This, however, created the class of morphotactically opaque circumfixes.
Another, better-known, semiotically based parameter consists in the preference for bi-uniqueness, or at
least uniqueness, as opposed to ambiguity. Bi-uniqueness
5
holds if one and the same form has always the
same meaning and vice versa, uniqueness if this holds in only one of the two directions. Biuniqueness is
difficult to achieve because of economy of sign shapes and is thus often violated, for example in cases of
hypercharacterization (see section 2.1), where one and the same meaning, for example plurality, is
expressed twice instead of once only. In hypercharacterization, morphotactic transparency wins out over bi-
uniqueness.
Since conflicts between parameters of morphology and those of either phonology or syntax are dealt with,
albeit from a different perspective, in Janda's and Joseph's contributions to this volume, my discussion can
be limited to indicating specifics of the naturalist approach.
In interaction with phonology (cf. Dressler 1985b, 1996), in a first stage, sound laws apply with little or no
respect for morphology. These purely phonological processes (postlexical, postcyclic rules in the terminology
of Kiparsky's Lexical Phonology) opacify morphotactics very little, as in the surface palatalization example
from Polish, in section 2.1. Similarly, more morphotactic opacity occurs when morphonological rules
intervene, as in the morphonological palatalization example in Polish. This increasing morphologization of
phonological rules represents, on the one hand, an unnatural /marked change on the parameters of
morphotactic transparency and constructional iconicity (cf. section 2.1) as well as on the parameter of
phonological (bi-)uniqueness, because Polish [k'] uniquely derives from underlying /k/, whereas [c] may
derive from either /c/ or /k/. On the other hand, a context-sensitive phonological process possesses much
phonological indexicality (since it refers to its phonological context) and little morphological indexicality
(since it refers to a following suffix), whereas for morphonological rules morphological indexicality is much
more important than phonological indexicality, and morphological rules possess only morphological
indexicality. Thus increasing morphologization of phonological rules represents a shift from phonological to
morphological indexicality, explainable by semiotic priority of morphology over phonology. This explains the
unidirectionality of change from phonological to morphological rules (cf. Dressler 1996). The questions of
the conditions under which natural change of indexicality may outweigh unnatural change on several other
parameters will be reconsidered in sections 3 and 4.
A similar approach applies to interaction with syntax and to unidirectionally increasing grammaticalization
from syntax to morphology, a natural change on the parameters of morphotactic transparency (continuous
forms are preferable to discontinuous ones) and of indexicality, insofar as fixed morpheme order is
preferable to alternating order of words or clitics, as in the contrast between periphrastic constructions of
the type I
was read-ing, Wasn't I read-ing?
and their Italian equivalents
Legge-v-o, Non legge-v-o?
http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747914
Page 4 of 8
12. Naturalness and Morphological Change : The Handbook of Historical Linguistics : Blackwell Reference Online
12/11/2007 03:36 PM
Morphologization of phonological and syntactic patterns has been understood as the basic source of
morphological patterns by Wurzel (1984: 102ff, 212, 1987: 69) in his claim that truly morphological change
is only reactive (criticized by Dressler 1997a, 1999). Of diachronic relevance is also Wurzel's (1996a, 1996b)
perspective on “the age of morphological constructions,” which is correlated to stages of development with
different properties on morphological parameters.
3 Type-Adequacy
Inspired by Skalička's (1979) views on ideal language types which consist of properties which favor (or
disfavor) one another, we can reinterpret language types as (alternative) sets of consistent responses to
naturalness conflicts (section 2.2).
6
Since not all of the most natural options on all parameters can be
combined within one language, naturalness on certain parameters must be, so to say, sacrificed for greater
naturalness on others (cf. Dressler 1985a, 1985c, 1988a; Sgall 1988). Thus the agglutinating type (as best
represented by Turkish) has the advantages of much constructional iconicity, morphosemantic and
morphotactic transparency, and bi-uniqueness, but deviates with its often very long word forms from the
optimal shape of morphological words (one prosodic foot - another universal preference) and does not fully
achieve fixed morpheme order (unnatural on the parameter of indexicality), whereas the opposite holds for
the inflecting-fusional type. In this way, universally rather unnatural options may be typologically adequate if
they fit the properties of the respective language type.
A morphological change is type-adequate if one of the two following conditions is met:
i change does not modify typological properties - this is a typologically conservative change;
ii change correlates with other changes which implement an overall typological change of the
respective language - this is a typologically innovative change.
The second type of change can be exemplified with correlated changes from Latin to Romance inflectional
morphologies, that is, from a strong inflecting language to weak inflecting languages (with a greater role for
the isolating language type). Also, Estonian has changed from an agglutinating type to a predominantly
inflecting-fusional language, with less constructional iconicity, morphosemantic and morphotactic
transparency, and (bi-)uniqueness, but with fixed morpheme order and greater approximation to the optimal
shape of morphological words (di- and trisyllabicity), when compared with Finnish.
Type-adequacy affects the solution of conflicts between parameters. In section 2.2 we have discussed
morphologization of phonological rules and the question of why unnatural change on the parameters of
morphotactic transparency, constructional iconicity, and (bi-)uniqueness can be outweighed by other factors.
This happens especially in languages of the inflecting-fusional type where naturalness on these parameters
is sacrificed in favor of naturalness on other parameters, such as indexicality. In other words, relative
unnaturalness on these parameters obtains little weight and can therefore be outweighed by greater
naturalness on the parameter of indexicality. Similarly, the morphologically fairly unnatural class of infixes
(section 2.1) originates due to phonological factors and is typologically restricted (cf. Moravcsik 1977;
Méndez Dosuna and Pensado 1990).
Also suppletion, the most unnatural option on the parameter of morphotactic transparency, originates in
inflecting-fusional rather than in agglutinating languages. The many origins of suppletion (cf. Ronneberger-
Sibold 1990) must be strictly differentiated from the factors of maintenance (i.e., stability) of suppletion:
those suppletive forms are best preserved which have high token frequency (thus storage is more
economical than composition and decomposition by rule, e.g., with auxiliaries), have idiosyncratic meanings
(e.g., learned connotations, such as Fr.
Fontainebleau
, adjective
Bellifontain
of artificial humanistic origin),
are not natural members of large classes (e.g., auxiliaries in contrast to main verbs), or support each other
analogically, as in antonyms (e.g.,
good, bad
, comparatives
better, worse
).
4 System-Dependent Naturalness
4.1 System-adequacy
Language-specific, system-dependent naturalness, as conceived by Wurzel (1984) for systems of inflectional
morphology (modifications in Dressler 1997b), represents what is normal or system-congruous (system-
http://www.blackwellreference.com/subscriber/uid=532/tocnode?id=g9781405127479_chunk_g978140512747914
Page 5 of 8