Scalar Categorization


Grazia Crocco Galèas

Università di Pavia


[Download this article (115K, Rich Text Format)]

(received October 1998, revised November 1998)

Copyright Notice:
First published in Web Journal of Modern Language Linguistics in association with the publishers (to be announced). © 1998 Grazia Crocco.
The moral rights of the author(s) to be identified as author(s) of this work are asserted in accordance with §§.77 and 78 of the Copyright, Designs and Patents Act 1988. This work may be reproduced without the consent of the author, in part or in whole in any manner and in any medium subject only to the two following conditions:
(a) no charge shall be made for the copy containing the work or the excerpt,
(b) a copy of this notice shall precede the work or the excerpt.


0. Introduction
1. Classical categorization and Prototype Theory.
Natural Morphology
Natural Morphology and Prototype Theory



0. Introduction

'If linguistics can be said to be any one thing it is the study of categories; that is, the study of how language translates meaning into sound through the categorization of reality into discrete units and sets of units.' Labov (1973: 342)

Categorization is a fundamental operation for any science since it constitutes an essential part of what I shall call a 'gnosiological' approach, and it is also a central problem in linguistics and, more specifically, in morphology. Essentially, my proposal is that current theories of categorization, which I see in terms of a binary opposition between classical categorization, based on the principle of discreteness, and prototypical categorization, based on the notion of prototype and basic-level categories, fail to account for a range of morphological data. On the one hand, the traditional view is tied to the classical theory that categories are defined in terms of common properties of their members. On the other hand, Prototype Theory has recently pointed out how human categorization is based on principles that extend far beyond those in the classical theory. In this paper, I intend to show that, although both theories can be interesting for linguistics, it is also necessary to adopt an intermediate approach, i.e. scalar categorization, which is profitably applied within the theory of Natural Morphology. Although the need for a gnosiological approach, which mediates between the objectivist view of classical categorization and the experimentalist view of Prototype Theory, is still far from being generally accepted in linguistics, I argue that the advantage deriving from an explicit assumption of the scalar perspective appears quite evident in the study of morphological systems.

In order to avoid misunderstandings I will mainly use the philosophical term 'gnosiology' to refer to the range of operations that human brains activate in the perception of reality or, in other words, when they try, in order to make sense of experience, to organise a conceptual system. A modern term for all this would be 'cognitivism'. However, I prefer to delimit the use of 'cognitivism' and 'cognitive linguistics' to denoting that branch of linguistic science that derives more or less directly from the psychological theory of prototype. Therefore, by adopting a philosophical term of ancient tradition, I mean by 'gnosiology' the set of operations that human minds exploit when building knowledge and, of course, the science that studies the forms of understanding, knowing, and thinking. Since classification (or categorization) of reality is the essential substratum of knowledge and, therefore, of science, I focused my attention on the questions related to the categorization of linguistic objects and, specifically, morphological operations, rules, and complex words. Therefore my main aim is to point out the importance of a scalar approach to morphology.

In what follows I will first briefly illustrate Prototype Theory in relation to the 'Aristotelian' theory (§ § 1. - 1.2.1.). Then I will concentrate on an exposition of the main tenets of Natural Morphology regarding particularly the notion of parameterisation (§§ 2. - 2.9.). After a section devoted to the discussion of the differences between the scales of Natural Morphology and the gradients of Prototype Theory (§ 3.), I will propose a new application of the principle of scalarity in the framework of Natural Morphology, i.e. the parameter/scale of indexicality. By introducing a new scale of morphological naturalness (§ 3.1. - 3.1.4.), I intend to show how scalar categorization can allow a better understanding of a given area of morphological phenomena, namely indices. The final observations of the last section (§ 4.) conclude the work.

The scale of indexicality presented in this paper is only posited as a hypothesis; in particular, it is elaborated according to a number of predictions that originate from the assumption of some semiotic principles. As in the case of other parameters and scales of Natural Morphology, I will show the validity of the predictions made in this contribution by providing a large corpus of data in a further article on indexicality.

1. Classical categorization and Prototype Theory.

The notion of prototype is largely due to Eleanor Rosch's psychological experiments (e.g. Rosch 1973a, 1975a, 1975b, 1977, 1978, 1981) on the organisation of conceptual categories. Rosch (1973b, 1975b) investigated the structure of categories such as 'furniture', 'fruit', 'vehicle', 'vegetable', 'bird', etc., by asking subjects to judge to what extent certain kinds of entity could be regarded as good examples of a category. On the basis of Rosch's results it is argued that members of the so-called classical (or 'Aristotelian') categories
[1] share all the same properties. This means that all entities, which on the basis of a limited number of common features, are classifiable as pieces of furniture, are all members of the category 'furniture' - they belong unequivocally to the category and all have equal status. However, Rosch's work showed that categories are structured in an entirely different way; members that constitute them are assigned in terms of gradual participation and the categorial attribution is made by human beings according to the more/less centrality/marginality of collocation within the categorial structure. Elements recognised as central members of the category represent the prototype. For instance, a chair is a very good example of the category 'furniture', while a radio is a less typical example of the same category. A chair is a more central member than a radio, which, in turn, is a rather marginal member.

1.1. An example: The categorization of colour.

According to the classical approach to categorization, colour systems in the languages of the world are a typical instance of a physical continuum (i.e. the colour spectrum) carved up in arbitrary ways by the different languages. In other words, it is the language system that arbitrarily cuts up the colour space into discrete categories (cf. Hjelmslev 1968:57-58). All categories, i.e. all colour terms in a system, have equal status. Black is black and white is white. And black is neither more nor less a colour than white. Therefore, there is no possibility of boundary colours, as all referents of a colour term have equal status. If two colours are both categorized as red, there is no sense in which one is 'redder' than the other. Alternatively, according to Prototype Theory, the role played by non-linguistic (perceptual and environmental) factors in the structuring of colour systems demonstrates that there are some basic colour categories (Berlin & Kay 1969) whose central members are the same universally. Further experimental research (Kay & McDaniel 1978) has found that universal focal colours represent the best examples of a given category. This means, for instance, that some members of the category 'red' are better examples of the category than others. Thus, colour categories have a centre and a periphery and members of a category do not all have the same status. Focal colours are, in fact, perceptually and cognitively more salient than non-focal colours. Scarlet is more salient than crimson and therefore constitutes a better example, namely a prototypical example of the category 'red'. In brief, we must conclude that colour categorization is neither all the same across cultures, nor is it by any means arbitrarily different across cultures. Reality is not a diffuse continuum, and our categorization of it is not merely an artefact of culture and language. Even in an area of experience as that of colour, where the reality-as-a-continuum hypothesis would seem to hold, the notion of categorization as language-dependent seems inadequate. The colour names do not just attach to the neurophysiologically determined distribution functions. There are mechanisms like fuzzy set union and intersection that reveal that the possible colour ranges depend upon limited gnosiological parameters. Therefore, both neurobiological (including psychological) and socio-communicational (including sociopsychological) features constrain -- but do not determine -- the structuring of colour systems in the languages of the world. Prototype Theory stresses that there is neither naturalistic nor cultural determinism in the organisation of categories. Thought is embodied, which is tantamount to saying that conceptual categories are directly grounded in perception, body movement, and experience of a physical and social character (Johnson 1987).

1.2. The weak and the strong version of Prototype Theory.

Rosch's interpretation of the experiments produced by her has not been always the same. At first she identified the prototype with a particular member of a category. Later she conceived the prototype as an abstract class or a conjunction of properties. Lakoff (1987:42-46) distinguishes three phases in Rosch's thinking about categorization. In my opinion, it is possible, without going into detail, to recognise only two fundamental phases, the weak version and the strong version respectively. In the first phase (from the late '60s to the early '70s) Rosch claims that prototypes, because of their perceptual salience, memorability and stimulus of generalisation, are interpretable as those entities which are first associated with the names of the categories when they are learnt. In the first period Rosch studied colours, shapes, and emotions and her conclusions led her to the definition of prototype as the most representative member of a given category. In my terminology this phase of research corresponds to the weak version of Prototype Theory. In the second phase (early to mid '70s), Rosch argues that the prototype effects of her experiments provide a characterisation of the internal structure of the categories. This second phase is influenced by information-processing-psychology and corresponds to what I call the strong version of Prototype Theory. By introducing basic level concepts, Rosch also suggests the comparison, or even the identification, of her theory with the notion of 'family resemblance' (see below §1.2.1.). The third phase (late '70s) with which Lakoff associates himself is certainly interesting as to the theoretical effort toward more mediation between objective reality and subjective reality of the categorizing human being. However, I do not take this last phase into account because I think that, by giving up the interpretations of her previous experimental data, Rosch basically denies her ideas and begins the so-called 'extended version' of Prototype Theory. Rosch claims that prototypes do not provide a theory of representation for categories, since prototypes can only constrain, but do not determine, models of representations (Rosch 1978). My view (cf. also Kleiber 1990) is that Rosch's studies from 1978 are in sharp contrast to her standard formulation of the Prototype Theory, although Lakoff (1987), Taylor (1989) and Rosch herself consider the standard version as a continuation of the first and second phase studies.

I argue that Rosch's first definition of prototype (weak version) is based substantially on the very same principle of discreteness which characterises classical categorial organisation. In fact, if we interpret the experimental data on the basis of the weak version of prototype, we can say, for instance, that a robin is, for the west-coast American speakers, the best example of the category 'bird' - it is the only prototype, rendering all the other types of birds non-prototypes. Therefore, if a prototype coincides with a well-established member of a category, all other members are non-prototypes, although to different degrees. On the other hand, Rosch's second definition of prototype (strong version) as an abstract concept rather than as a definite member entails a categorial structure which is absolutely bereft of any kind of discreteness. If, for instance, we extrapolate from the experimental data all the features that are attributed by American speakers to the 'typical bird', we need not conclude that an actual bird represents the prototype of the category. It is sufficient to shape an abstract, ideal, representative member of the category 'bird' that is not embodied by any specific, real bird. In other terms, the category 'bird' results from a gradual arrangement of bundles of properties along a continuum. Consequently, the boundaries between the categories are fuzzy and the continuum is not only intercategorial but also endocategorial. There is not only a gradient within a category like 'bird' but the continuum goes far beyond and entails what one could categorize as non-birds such as penguins, that, nevertheless, still share a very few features of the category 'bird'.

1.2.1. Prototype Theory and family resemblance.

Prototype Theory has been often compared, if not absolutely identified with, Ludwig Wittgenstein's notion of family resemblance (Familienähnlichkeiten). For Wittgenstein the proceedings of games constituted a paradigm of complicated nets of similarities which intersect and overlap: board games, ball games, Olympic games, and so on. What is common to them all? In his words (Wittgenstein 1978:32), '[...] we see a complicated network of similarities overlapping and criss-crossing: sometimes overall similarities, sometimes similarities of detail' . In order to shed light on the nature of such resemblances Wittgenstein put forward a comparison with the physical and psychical similarities that connect and at the same time distinguish the members of a family, for, as he wrote, 'the various resemblances between members of a family: build, features, colour of eyes, gait, temperament, etc., etc. overlap and criss-cross in the same way.' (Wittgenstein 1978: 33). In her strong version of Prototype Theory (e.g. Rosch & Mervis 1975) Eleanor Rosch tried to mould the notion of prototype on the Wittgensteinian notion of 'family resemblance'. However, it has been sharply argued that the two notions are not compatible (cf. Kleiber 1990). Prototype categorization has one and only one centre, whilst 'family resemblance' categorization does not have or has more than one centre. Wittgenstein does not claim that a particular game is a more representative example of the category 'game' than the others are. Thus, it is inconsistent to identify prototype structure categories in terms of 'family resemblance' classification even in the case of prototypes which are intended as clusters of properties (= strong version of Rosch's approach). The only trait which they have in common is the absence of obligatoriness in the characterisation of the categorial properties. Neither prototypical categorization nor family resemblance classification requires that the members have all the same features. For instance, not all pieces of furniture have four legs (e.g. a stool); not all games imply winning and losing or competition between players (e.g. patience). Still, unlike prototypical categorization, family resemblance does not posit any relationship of measure between a prototype (or a clusters of special properties) and all the other members of a family. Family resemblance does not entail a prototype.

2. Natural Morphology.

Natural Morphology[2] is the morphological theory of Natural Linguistics. The concept of naturalness is the fundamental feature of Natural Linguistics. Naturalness must not be equated with frequency, concreteness, simplicity, or intuitive plausibility. Instead, it is a basic principle of a linguistic approach that relies strictly on empirical evidence. The notion of 'naturalness' remains vague and pretheoretical until we connect it to that of 'markedness'. In order to explain the interrelation between naturalness and markedness let us start with five observations (cf. Mayerthaler 1981:2).

1) Not all morphological processes and structures are equally distributed in natural languages. For instance, as far as affixation is concerned, suffixes are cross-linguistically more frequent than prefixes (cf. Cutler et al. 1985). A language may lack prefixes altogether (as is the case of Turkish) but if it has just one kind of affix then it is the suffix.

2) Not all morphological structures are acquired by children at the same time. High-frequency suppletive paradigms, for instance, are acquired later then regular ones in first language acquisition (cf. Chini & Crocco Galèas 1995). Children first use irregular forms by rote-learning (cf. MacWhinney 1978); later, as their analytical capacities increase and regular paradigms are progressively organised, they drop those early irregular forms and substitute them with 'regularised' morphological structures e.g. Eng. goed instead of went (e.g. Berko 1958). Finally, they abandon the regularised structures (goed) and acquire the adults' suppletive forms (went).

3) Not all morphological structures are equally affected by language change. Suppletion is a relatively rare and unproductive morphological operation (Dressler 1985c, 1986). One would therefore expect language change to suppress suppletive forms. However language change does not always move in the direction of greater efficiency - it is typologically dependent. In fact, since inflecting languages show more suppletion than agglutinating languages, a language which changes from one type to the other increases or decreases its amount of suppletion accordingly. Estonian has become an inflecting language in contrast to cognate Finnish and Hungarian, and it has more suppletion. On the other hand, Tokharian has become an agglutinating language in contrast to cognate Latin, Greek, Russian, German and it has less suppletion.

4) Not all morphological processes and structures are equally impaired by language disorders. Aphasics handle more transparent complex words better than less transparent ones (Dressler & Denes 1988). For example, reader in the sense of 'someone who reads' is more transparent / compositional / descriptive than reader as a position in a British university. Indeed, the degree of morphosemantic transparency of a compound or derivative equals the degree to which the meanings of the parts (i.e. read, -er) yeld the meaning of the whole. We can thus predict that in aphasia the opaque (i.e. non-compositional) meaning of the complex word reader should be hard to process whether in production or perception.

5) Not all morphological structures are equally easy to decode. Non-biunique or ambiguous complex words, that is to say words not complying with the principle 'one meaning - one form' require much more effort in morphological processing. For instance, Italian has at least three suffixes to express agent nouns e.g. -tore : importare 'to import' -->importa-tore 'importer', -nte : militare 'to militate' -->milita-nte 'militant', -ino : imbiancare 'to whitewash' -->imbianch-ino 'whitewasher'. However each of these suffixes represents other meanings. For example, all of them can express instrument nouns. In particular, the suffix -ino also derives relational adjectives, nouns/adjectives denoting ethnic groups, and diminutives. Thus, Natural Morphology predicts that an ambiguous suffix like -ino should be decoded with much more difficulty than a biunique suffix like, for instance, It. -ificio (cf. calzature 'footwear' --> calzatur-ificio 'shoe factory'), which only has the meaning 'factory'.

Given this situation, naturalists make the following hypothesis: a morphological process or a morphological structure is natural if it is (a) widely distributed and/or (b) acquired relatively early and/or (c) relatively resistant to language change or develops frequently by language change and/or (d) is relatively less likely to be impaired by language disorders and/or (e) is relatively easy to decode. The degree of naturalness assigned to processes and structures is in inverse proportion to the degree of markedness: a morphological phenomenon is more natural the less marked it is, and vice versa.

2.1. Extralinguistic foundations of Natural Morphology.

Naturalists explain the complementary notions of 'naturalness' and 'markedness' in relation to extralinguistic bases of language. Extralinguistic foundations (= causa materialis) are best divided into two types: 1) neurobiological (including psychological) bases and 2) socio-communicative (including socio-psychological) bases. Extralinguistic factors either determine / prohibit or favour / disfavour conceivable properties of linguistic structure. Such extralinguistic constraints are relative, not absolute. They do not exclude marked (or unnatural) phenomena but predispose language users to avoid them. For instance, the highly unnatural morphological operation of suppletion is very limited in the languages of the world. It is relatively more frequent only in inflecting languages though restricted to very few morpholexical domains (e.g. few basic verbs, ethnical nouns, numerals from one to ten, etc.).

The first set of extralinguistic bases includes psychological limitations of perception and receptive processing, limitations of memory, restrictions on storage or on retrieval of information, on selective attention while producing and perceiving, etc. The second set has to do with the communicative function of language. For instance the relation between optimal perceptual contrast (or processing ease for the hearer) and articulatory effort presupposes the speaker's empathy with the hearer's receptive role (cf. Clark 1996).

The role of extralinguistic factors must be regarded from the perspective of the prototypical speaker. Indeed, the specific and universal properties of language users impose a certain number of constraints on the linguistic means that are available. This is tantamount to saying that extralinguistic factors delimit the range of possible morphological techniques and operations by enhancing some of them and inhibiting others. Given these premises, human capacities determine which operations are more or less natural according to a universal hierarchy. In this sense, what is easier for the potential language user is what is called natural. Naturalness is therefore a universal notion assumed by a universal linguistic theory based on the concept of markedness. In fact, morphological phenomena are natural if they are minimally marked or not marked at all. On the other hand, unnatural phenomena are morphologically marked. Nevertheless the theory of markedness is only a part of the theory of Natural Morphology.

2.3. The theory of Natural Morphology: The quintuple.

Natural morphologists agree on positing five levels of linguistic analysis (Dressler et al. 1987: 8-12): 1) The level of linguistic universals or the human language faculty; 2) the level of language types; 3) the level of language-specific competence; 4) the level of norm, and 5) that of performance. Natural Morphology deals particularly with the first three levels of analysis.

2.3.1. First level: The universals.

> The level of universals is modelled on markedness theory or a special case of preference theory (cf. Vennemann 1983). It comprises functions, operations, and principles, which can be assigned to parameters of naturalness. These parameters can be given the form of implicational scales from most to least natural. Since extralinguistic factors do not underdetermine linguistic structure, but limit the choice of linguistic (i.e. morphological) techniques open to languages, a linguistic preference theory must refer to extralinguistic facts. I will describe the level of universals in greater detail in § 2.6.

2.3.2. Second level: The morphological types.

The level of language types is modelled by typological theory (cf. Skalicka 1979 for the notion of language type as an ideal construct) and it filters the level of universals. In fact, universal properties are the basis of typological properties, i.e. they are restricted by the constellation of choices of a linguistic type. Language types (e.g. isolating, agglutinating, inflecting, introflecting, and polysynthetic) cannot always select the best morphological procedures of each naturalness scale due to the dialectical conflicts that characterise language systems. Thus, a language type is constituted by specific choices from the naturalness scales. Each language type 'sacrifices', as it were, naturalness in some parameters for the sake of greater naturalness in other parameters. For instance, agglutinating languages choose the most natural options from the parameters of diagrammaticity, morphotactic transparency, and morphosemantic transparency, but they 'sacrifice' the parameters of indexicality and that of size of signans. On the other hand, inflecting and introflecting languages often achieve optimal indexicality because of their fusionality, and the size of their complex words / word-forms is natural, i.e. non-marked. Yet, because of several typical phenomena: allomorphy, empty morphs, redundancy of exponence, etc., these languages do not normally select the first thresholds from the parameters of diagrammaticity, morphotactic transparency, and morphosemantic transparency.

2.3.3. Third level: System-adequacy.

The level of language-specific competence is modelled in a theory of system congruity (Wurzel 1984). Within any given language, a language type is realised according to the system-structural defining properties of each language. A morphological phenomenon (inflectional class, paradigm, a morphological form, marker, or rule) may be rather unnatural in terms of universal markedness theory, but at the same time it may be very 'normal' within the language-specific system in terms of system-adequacy. So, for instance, it is more 'normal' for German that a monosyllabic masculine is inflected according to the e-plural class (e.g. der Hund 'the dog' --> die Hund-e) and not the n-plural class (e.g. der Bär 'the bear' --> die Bär-en), and a monosyllabic feminine according to the n-plural (e.g. die Bahn 'the train' --> die Bahn-en) and not the e-plural class (e.g. die Maus 'the mouse' -->die Mäus-e) and not vice versa. Diachronic transitions, in fact, are from the n-plural class to e-plural class for the masculines (cf. der Hahn 'the cockerel' : die Hähn-e) and from the e-plural class to the n-plural class for the feminines (e.g. die Burg 'the castle : die Burg-en).

2.4. The semiotic metalevel.

Besides the universal, the typological, and the systemic level, Natural Morphology regards semiotics as a superordinate framework or metalevel. Language is a system of verbal signs serving two main functions: 1) It enables man to communicate better than with non-verbal signs, 2) it supports and guides cognition better than with non-verbal signs. Thus, because of the communicative and cognitive functions, man's verbal and non-verbal systems may be compared. Since semiotics is the theoretical and practical study of signs (both verbal and non-verbal), it represents a good candidate to supply a meta-theory of Natural Phonology, Natural Morphology, etc. Semiotics, therefore, serves as the basic framework underlying the notion of naturalness itself and the functionalist approach of naturalism (see below § 2.5.).

Naturalists draw particularly on the semiotic theory of Charles S. Peirce. According to Peirce, a sign consists of something (= signans) which stands to somebody (= interpreter) for something (= signatum) in some respect or capacity (= interpretant). Thus we have four aspects of a sign:

1) the interpreter is the user of the sign when inventing, producing, perceiving, processing, evaluating or storing it;

2) the signatum is what is expressed in the sign;

3) the signans is what expresses the signatum;

4) the interpretant is 'the idea to which a sign gives rise'.

When we use a complex word (e.g. (he) rewrites) the phones or allophones [riraits] are signantia of the phonemes /riraits/ and these, in turn, are the related signata. The phonemes and their respective (allo)phones are signs on the signs of morphemes whose morphs (e.g. re-, write, -s) are signantia and whose signata are lexical meaning /WRITE/, derivational meaning /REPETITION/, and inflectional meaning /3rd PERSON SINGULAR PRESENT/. Morphemes are again signs on the signs of words. In our example, the signans of the complex word is rewrites and its signatum is the meaning of (he) rewrites.

Peircean semiotics particularly focusses on the opposition 'natural / conventional'. This same opposition is also the crux of the theory of naturalness. Ferdinand de Saussure claimed that the relation between the two constitutive parts of the sign is arbitrary. Peirce, on the other hand, distinguished between various types of signs according to the relation that connects signans and signatum. Among the several classifications introduced by Peirce, the one that plays a crucial role in Natural Morphology is the triad of symbol, index, and icon.

According to Peirce, a symbol is a sign that refers to its object through a conventional or habitual link. The link or rule that connects a symbol to its referent must be known by the interpreter of the sign in order to be able to use and understand it. A linguistic symbol consists of a signans representing a signatum determined by convention. All linguistic signs are symbols. Nevertheless Peirce stresses at least two other semiotic aspects that are also identifiable in a number of linguistic signs. Indeed, besides the definition of symbol, Peirce also gives a semiotic definition of index and icon.

An index is a sign that directly points to its object without describing it. From an index, an interpreter can infer the existence of a given object. A linguistic index is therefore any sign whose primary function is to signal another sign. Demonstratives, pronouns, proper nouns, and grammatical morphemes are typical indices in language.

There is a third type of sign important for the theory of Natural Morphology: it is the icon. According to Peirce an icon is a sign exhibiting a resemblance with the object it denotes. An iconic sign in language is one whose signans shows a relation of similarity or analogy with its signatum. Icons are the most natural signs. As Peirce (1965.II: 158) states, 'the only way of directly communicating an idea is by means of an icon' because in icons there is an intrinsic connection between signans and signatum. In relation to the degree of similarity between signans and signatum, Peirce distinguishes three subtypes of icons: images, diagrams, and metaphors. Images are the most natural icons. Metaphors are the least 'iconic' icons.

An image is an icon representing directly the features of an object. A photograph is a typical image of what it represents. In language, images are mainly onomatopoeic words (e.g. to twitter), i.e. words imitating by the structure of their signans the sound of bird cries, objects, etc.

Diagrams are icons 'which represent the relations, mainly dyadic, or so regarded, of the parts of one thing by analogous relations in their own parts' (Peirce 1965 II: 157). For example, a paper pattern reproduces the parts of a suit or other article of clothing. The person who makes a suit, therefore, follows the scheme or model given by the paper pattern. He / she pays particularly attention to the relations among the paper pieces, in order to cut the cloth and sew it. Thus, a paper pattern is a good example of a diagram, since it reproduces by analogy the parts of an object, i.e. a suit that can be regarded as the referent of the diagrammatic icon. Similarly, a verbal diagram is an icon showing analogy of structure between signans and signatum. Its degree of iconicity lies between that of images and metaphors. Nevertheless diagrams are most important for Natural Morphology. Although their degree of iconicity is intermediate, it seems that all components of language share some fundamental traits of diagrammaticity. The general notion of structural isomorphism in language is referable substantially to the concept of diagrammatic relation between signans and signatum. In morphology, the role of iconicity is relevant (cf. Jakobson 1971). Starting from the observation that many morphological signs exhibit a relation of equivalence between signans and signatum, it is easy to find numerous examples in this respect. For instance, Indo-European languages express the three degrees of adjectives, i.e. positive, comparative, and superlative, by a gradual increase in the number of phonemes, e.g. Lat. clar-us 'famous' - clar-ior 'more famous' - clar-issimus 'most famous'. In this way the shape of the signantia reflects the intensity of the gradation conveyed by the signata.

A metaphor is an iconic sign characterized by similarity to its object. For instance, a tarot card showing sticks is a metaphor of a forest. In an analogous manner, a verbal metaphor is an iconic sign exhibiting some parallelism or partial similarity between signans and signatum. All types of morphological conversion (e.g. Eng. bottle - > to bottle, to run - > a run) are metaphors insofar as they show parallel signantia mapped onto different although morphologically related signata.

From each type and subtype of sign an adequate semiotic principle is deducible, expressing all the features that characterise the corresponding sign. Thus, the principle of symbolization derives from symbol, the principle of indexicality derives from index, the principle of iconicity derives from icon, etc. Natural Morphology derives its morphological parameters from a number of semiotic principles which mostly originate from the types of signs Peirce distinguishes. For instance, the morphological parameter of diagrammaticity derives from the semiotic principle of constructional iconicity, which in turn is elaborated according to Peirce's definition of diagram.

2.5. The functionalist approach of Natural Morphology.

The emphasis on a semiotic metalevel is closely related to the functionalist / teleological approach of naturalism. The theory of Natural Morphology is functionalist insofar as it explicitly distinguishes functions and operations serving these functions. In this respect the functionalist model of Natural Morphology diverges from that of André Martinet and is much more similar to the approach proposed by Hans-Jakob Seiler and his group in Cologne. In fact, analogously to Seiler's UNITYP model, Natural Morphology regards language as a problem-solving system and refers to the three levels of universals, typology, and language-specific system (cf. Seiler 1978a, 1978b, 1979).

According to the functionalist postulate, language is a tool for communication and cognition. Since communication is goal-oriented, linguistic theory must provide functional explanation of language aspects. In this regard, unlike Martinet's functionalist means-ends model, Natural Morphology takes into due account the phenomenon of multicausality. Therefore, in adopting functional explanation, adherents of naturalism recognise that one function can be served by several operations (multiple strategies) and one operation may serve several functions simultaneously (multifunctionality). For instance, the main function of word-formation is lexical enrichment. This is achieved through some morphological techniques such as derivation, compounding, conversion. We can therefore say that the function of lexical enrichment is served by numerous techniques and operations (e.g. the technique of derivation can be realised via a number of operations such as suffixation, prefixation, etc.). On the other hand, one operation like suffixation can fulfil not only the function of lexical enrichment but also that of deriving inflectional word-forms.

Language change is crucial in functional explanation. If one assumes that individuals use a determinate operation for communicative functions, then the speech community as a whole will try to improve the type of operation serving these functions. Thus, language change can be regarded as the inevitable result of such a tendency towards increased efficiency. Yet, not every diachronic change enhances communicative efficiency -- in other words, languages, as whole, do not become more and more efficient throughout their diachronic development. Therefore, the type of functional explanation invoked by Natural Morphology presupposes the existence of goal conflict. What is more efficient, i.e. more natural for some reasons, is less natural for other reasons. For instance, phonological naturalness comes into conflict with morphological naturalness and this in turn with lexical naturalness. Consequently, language change does not necessarily imply better serving of functions. After the change some functions are served better, some worse. Functional optimisation is in fact local, not global.

2.5.1. The functions of the morphological component.

There are four main functions that characterise the morphological component.
1) The first function of word-formation is lexical enrichment via morphologically derived words.

2) The first function of inflectional morphology is to express syntactic categories via morphologically derived word-forms.

3, 4) The second function of both word-formation and inflection is to motivate derived words / word-forms both morphotactically and morphosemantically.

Consequently, all morphological phenomena must be related, directly or indirectly, to at least one of these functions as well as to semiotic principles. These four functions and a small number of interconnected semiotic principles are the basis of the universal parameters of naturalness / markedness, i.e. constitute the universals of the first level of the Natural Morphology model.

2.6. The universals of the theory of Natural Morphology.

The universals comprise main functions (i.e. communicative and cognitive functions respectively), subordinate subfunctions specific to each language component (e.g. morphotactic and morphosemantic motivation of complex words), and some semiotic principles (e.g. the principle of diagrammaticity) largely deduced from the Peircean triad of legisigns: symbols, icons, and indices. The level of universals also corresponds to the operations realising the different functions (e.g. suffixation).

Functions are determined by the neurobiological and sociocommunicative constitutive traits of human beings. This means that universals have undeniable extralinguistic bases. Consequently, functional explanations of the type admitted by Natural Morphology consist of relating a given morphological phenomenon to either neurobiological (including psychological) or socio-communicational factors or indeed to both.

As we have seen, each level of the quintuple (see § 2.3.) is formalised through a specific subtheory. The level of universals is formalised in universal markedness theory. According to this universal theory, any linguistic phenomenon is said to be natural when it is unmarked or relatively less marked. More or less natural, as well as more or less unmarked, means more or less easy for the human brain. Therefore, naturalness is not a binary classificatory predicate (the opposite value being unnaturalness) but a gradient predicate. 'Natural' in the sense of Natural Morphology is not a value but a relation. We speak of naturalness and / or markedness relations. Similarly, (un)marked in the sense of Natural Morphology is an evaluative order relation, not a descriptive predicate of an idealized grammar.

Due to the gradient character of naturalness, universals of morphological naturalness are expressed in historical languages through preferences. This entails that some morphological forms, techniques, operations, and rules are preferred because they are natural. In this respect, I will give an example, which is often cited by naturalists. Assuming that the category of plurality is encoded morphologically in a given language, morphological naturalness predicts that it should be encoded by means of an overt marker or morpheme. This is a preference largely attested cross-linguistically. It is due to the relevant tendency to reflect iconically the addition of intensional meaning by the addition of phonological material. Therefore, if and only if plurality is indexed by a sign, the encoding will correspond to the naturalness preferences predicted by Natural Morphology. In other words, the encoding can be called natural or unmarked. On the contrary, if the requirement of overt sign is not fulfilled, then the type of encoding will be called unnatural or marked. A plural form like Eng. sheep is unnatural or marked because, unlike most plural forms in English, it is 'featureless' (= Germ. merkmallos). The regular plural, in fact, is always 'featured' (= Germ. merkmalhaft), e.g. ship-s, and for this reason it is natural or unmarked. From this example, it is also evident that 'marked' is not necessarily synonymous with 'featured'.

As I have pointed out, morphological universals are formalised in a theory of naturalness / markedness. The relationality or gradiency of the concept of naturalness is motivated by the existence of varying degrees of ease for human brain, namely that some things are easier to handle than others (e.g. suffixation is more natural than circumfixation, namely discontinuous morphs). Therefore, the theory of naturalness entails a theory of preference, which, in turn, forms the basis of a system of predictions. For instance, since according to morphological naturalness universals, featureless nominative forms are unmarked, we can predict (a) that featured nominatives will be acquired later, (b) that they will be cross-linguistically infrequent, (c) that there should be languages with featureless nominatives but without featured nominatives, (d) that in languages having featured nominatives their type and token frequency should be limited.

2.7. The parameters of universal naturalness.

The level of universals is modelled by the theory of naturalness / markedness in a number of parameters. Thus, the parameters of Natural Morphology represent the formalisation of functions and semiotic principles, which derive from the extralinguistic bases assumed by a functionalist and semiotic model of morphological theory.

The parameters of morphological naturalness / markedness express the preferences or tendencies of historical languages in the choice of morphological techniques, operations, and rules. Since the range of options within a single parameter goes from a maximum to a minimum of naturalness, the resulting gradient is a scale which has a most natural threshold and a least natural pole. In this sense, each parameter can be viewed as a scalarised factor of the morphological component.

A factor is, by definition, a morphological universal, whether deduced from a function or a semiotic principle or an operation. A factor is not only a universal but, at the same time, is also an essential point in the description and explanation of morphology. Any factor of morphological analysis is rooted in the morphological component, and this, in turn, due to the functionalist approach of Natural Morphology, is shaped in accordance with the system of features that characterise human beings. Hence, a parameterised factor constitutes both an object of reality and a tool of interpretation.

Parameterisation consists mainly of the elaboration of semiotic principles in the format of superordinate hermeneutic means of linguistic behaviour. The semiotic principles, which are among the most relevant parameterised factors, derive from Peircean semiotic theory.

Parameters are scalarised factors, i.e. gradually distributed sequences of options. Thus, scalarity instead of binarity is the direct consequence of the gradual and relational character of universal naturalness / markedness. In effect, a parameter arranges in a scalar dimension all the universal possibilities conceivable between two opposite poles. Thus, a parameter is an ordered series of morphological realisations elaborated in the format of an implicational scale of naturalness / markedness.

As I have already emphasised, languages do not always select the most natural realisations on every scale. This depends upon the language type to which a specific language conforms. Indeed, a type mediates between universal naturalness and system-adequacy. However, in every language type, the unnatural selections on some scales are balanced by the very natural options on other scales.

The parameters of morphological naturalness which are represented in the format of an implicational scale are eleven in number (cf. Crocco Galèas 1998).

1) Diagrammaticity. The parameter of diagrammaticity (or constructional iconicity) derives from the semiotic principle of diagrammaticity. Therefore it is a typical instantiation of the principle of iconicity. Diagrammaticity entails a relation of biuniqueness between segmentability of signans and compositionality of signatum. (For the relevant examples see below § 2.8.).

2) Morphotactic transparency. This parameter derives from the principle of semiotic transparency, which, in other terms, is the principle favouring the ease of both production and perception in the realisation of complex words. By morphotactic transparency naturalists mean the factor of boundary recognisability which relates with the morphemic segmentation of a complex word. For instance, Turk. çocuk 'child' - > çocuk-lar 'children', It. prendere 'to take' - > ri-prendere 'to take again' are transparent complex words, as one can easily segment the affix from the base. On the other hand, Eng. public - > public-ity is a less transparent derivative because the base public is blurred by the intervention of an allomorphic rule of palatalisation. Even less transparent, i.e. relatively opaque is Eng. to delude - > delus-ion with fusion at the morphemic boundary. Most opaque is suppletion, e.g. Eng. to be, am, is, are, etc.

3) Morphosemantic transparency. The semiotic principle underlying the parameter of morphosemantic transparency is the so-called Fregean principle of compositionality of meaning. On the basis of this principle, we assume that the meaning of a complex word is a function of the meaning of its constitutive parts. For instance, compound nouns like Eng. teacup, Germ. Hausarbeit 'housework', etc. are totally compositional, namely morphosemantically transparent. On the other hand, compund nouns like Eng. telephone box, Germ. Rosenkranz 'rosary' (literally 'rose crown') are morphosemantically opaque because they are lexicalised items.

4) Uniformity. This parameter (like (5) and (6) below) can be deduced from the semiotic principle of relational invariance, i.e. the relation between the two complementary faces of a complex sign. In particular, the parameter of uniformity regards the structure of sign from the point of view of signatum. A uniform sign is a sign whose signatum is expressed by a single signans. In English, for instance, the progressive aspect is encoded only by the gerund suffix -ing, which is therefore a uniform encoding from signatum to signans. On the other hand, in English the superlative degree of adjectives is not only expressed by the suffix -est (e.g. dear - > dear-est) but also by suppletive forms (e.g. bad - > worst) and analytical encoding (e.g. famous - > most famous). Thus, the signatum 'superlative' is not realised in a uniform manner.

5) Transparency of encoding. This parameter also derives from the semiotic principle of relational invariance. It is the inverse of the parameter of uniformity. Transparency of encoding implies that the semiotic perspective be the signans of a sign. There is transparent encoding if a given signans represents one and only one signatum. Instead, opaque encoding entails more than one signatum for one signans. In Italian, the inflectional suffix -ss- ([s:]) for the Imperfect subjunctive of all three conjugations (e.g. am-a-ss-i, ten-e-ss-i, dorm-i-ss-i) is a good example of transparent encoding - one signans represents one and only one signatum. On the other hand, in German the suffix -icht which is used to form neuter collective denominal (e.g. Rohr 'reed' --> Röhr-icht 'bed of reeds') or deverbal (e.g. spülen 'to wash up' --> Spül-icht 'dishwater') nouns, is not a transparent signatum, since the derivational semantic meaning of 'collectiveness' is also conveyed by the suffixes and suffixoids -heit (e.g. Mensch-heit 'humanity'), -schaft (e.g. Kollegen-schaft 'people from the office'), -tum (e.g. Bürger-tum 'citizens'), -werk (e.g. Laub-werk 'foliage'), and -wesen (e.g. Schul-wesen 'school-system').

6) Biuniqueness. Biuniqueness is the third parameter of relational invariance between signans and signatum. In the case of biuniqueness the viewpoint from which the relation is regarded is double or complementary. In fact, the semiotic relation of invariance is simultaneously evaluated form both signans and signatum. This implies that a complex sign is biunique if and only if each part of its signans expresses always and only its corresponding signatum. For instance, in Modern Greek the prefix is- conveys just one semantic matrix i.e. 'movement toward or in a place', vice versa, this semantic matrix is only rendered derivationally by the prefix is- (e.g. pnoí 'breath' (is-pnoí 'inhalation'). On the contrary, there is lack of biuniqueness in the case of the English suffix -en which can derive de-adjectival verbs (e.g. short --> short-en) but is also the past participle suffix of many strong verbs (e.g. to write --> writt-en).

7) Indexicality. (see below). The parameter of indexicality derives form that type of sign that Peirce calls index. The primary function of a verbal index is to signal another sign. (See the examples below in § 3.1. and ff.)

8) Metaphoricity. The semiotic principle of metaphoricity derives from that subtype of icon that Peirce calls metaphor. The parameter of metaphoricity (or morphometaphoricity) allows the evaluation of complex signs (i.e. morphological metaphors) characterised by a partial similarity between signans and signatum. As a matter of fact, a morphological metaphor can be defined as semantically complex but morphotactically unanalysable. Therefore, there is no diagrammatic relation between a signatum resulting from an intensional addition or a modification of meaning and a signans that cannot be segmented. In other words, a morphological metaphor does not show analogy of structure between signans and signatum. In English, de-adjectival abstract nouns such as young --> the young, absurd --> the absurd, etc. are good examples of morphometaphorical nouns. On the other hand, English abstract nouns denoting feelings can either be sources of metaphoric verbalisations (e.g. hate --> to hate) or targets of nominalisations (e.g. to hate --> hate). Therefore, lack of a good criterion of directionality renders these nouns less typical instances of morphometaphoricity.

9) Size of the signans. On the basis of the semiotic principle of distinctiveness and salience of signs, complex words can be analysed through the parameter of the natural size (i.e. length) of the signans. The parameter allows complex signs to be distinguished according to the number of syllables. Since there are three types of signantia (i.e. lexical, derivational, and inflectional morphemes) co-occurring in the structure of complex words, we need to identify three different implicational scales for this parameter. For instance, on the basis of the scale of the lexical morpheme, we can predict that monosyllabic lexical morphemes are more natural hence more frequent than polysyllabic ones. These, in turn, are more natural than lexemes shorter than a syllable. Bound lexemes whose length is less than one syllable are cross-linguistically rare, but the roots of introflecting languages are typical non-syllabic bound lexemes (e.g. Arabic /ktb/ "to write", /f'l/ "to do"). They represent an unnatural morphological choice, since there is no mapping between the morphological unit (i.e. the lexeme) and the phonological unit (i.e. the syllable). In adopting cononantal roots as their morphological bases, Semitic language prefer introflection besides mere inflection. This preference is a consequence of a radical differentiation of functions between consonants and vowels in shaping lexical and grammatical morphemes. The scalarization of the parameter of size of signans is complex (see Crocco Galèas 1998) and cannot be explained here. However it is very interesting and suggests further empirical research.

10) Morphological base. The semiotic principle underlying the parameter of the morphological base is that of lexical priority over morphology. In other words, lexical morphemes, whether free or bound, are primary signs because they are stored in the lexicon. Signs that constitute the semantic bulk of the lexicon are the best bases for morphological rules. The parameter of the optimal base allows different types of morphological bases to be distinguished according to criteria of universal naturalness. Free or bound lexemes are most natural as morphological bases (e.g. It. bar 'coffee-bar' --> bar-ista 'barman', Lat. equ-us 'horse' --> equ-in-us 'equine'). Complex words (e.g. Eng. function-al --> functional-ist) are less natural bases for morphological rules. Syntagmata (or phrases) are even more unnatural (e.g. It. pressappoco 'approximately, roughly' lit. 'almost close to little' --> pressappoch-ist-a 'careless, inaccurate person'). Non-lexical bound morphemes (i.e. affixes) represent the most unnatural type of morphological base (e.g. Germ. miss- [derivational prefix indicating disdain / contempt] + -lich [(in synchronic terms) derivational suffix for relational adjectives, cf. Eng. -ly] --> miß-lich 'most regrettable'.

11) Symbolicity. The parameter of symbolicity derives from the semiotic principle of symbolicity. From a morphological perspective, we can regard as symbol any word which is semantically complex but morphotactically unanalysable. A symbol is a sign truly acting as a label. It is unmotivated because it is not morphemically segmentable and its signans is not articulated into a base and one or more affixes. In general, a symbolic sign is bereft of internal structure, and therefore, its signatum is not diagrammatically reflected by its signans. Prototypical symbols are (inflectional or derivational) suppletive words.

2.8. An example of parameterisation: diagrammaticity.

The parameter of diagrammaticity (or constructional iconicity) is most important for the organization of the morphological component. It derives from the semiotic principle of diagrammaticity. Therefore, it is a typical instantiation of the principle of iconicity. In particular, diagrammaticity entails a relation of biuniqueness between segmentability of signans and compositionality of signatum. A diagrammatic word is perfectly segmentable and semantically motivated. In other terms, the transparency of its complex signans reflects the compositionality of its signatum.

For instance, singer is a diagrammatic derivative. It is an agent noun consisting of the lexical morpheme sing, which is properly a verbal base, and the derivational suffix - er, denoting an agent. The addition of signans, i.e. the suffix -er to the verbal base sing parallels the addition of meaning, i.e. the notion of agentivity to the action expressed by the verb to sing. Therefore, the morphological rule of affixation combines two levels. On the one hand, the derivational process adds intensional meaning (e.g. agency) to the meaning of the base (e.g. to sing). By this combination, the global meaning of the complex word is perfectly motivated. On the other hand, the addition of expression (e.g. the suffix -er) reflects the semantic composition of the word. Thus, we can say that morphotactic transparency diagrams (i.e. reflects analogically) semantic compositionality.

Let us denote the semantic operation of addition with (A+B) and the corresponding morphotactic operation with (a+b). Then we can say that A (= the meaning of "sing") is conveyed symbolically / conventionally by a (= the morph sing), whereas B (= the meaning of "agency") is conveyed in an analogous way by b (= the suffix -er). By and large, the word singer is a symbol. Nevertheless, since it is semantically and morphotactically motivated by its verbal base and the agentive suffix, it is an iconic sign or, more specifically, a diagram.

Similarly, Turk. kanunlarimizda 'in our laws' is an inflected form characterised by a biunique structure. The word is analysable in the following manner:

1) a lexical base, kanun 'law';

2) the Plural suffix -lar-;

3) the Possessive suffix -imiz 1st Pl. Pers. 'our' (which can be further segmented into -im-, 1st Sg. Pers. 'my' and -iz- Plural suffix for Possessive endings);

4) the Locative case ending -da.

The sequence of morphs corresponds to the semantic composition. Thus, there is a biunique link connecting each morph to its meaning. The agglutinating morphological structure of this Turkish word fully exhibits analogy of signans and signatum. Turk. kanunlarimizda is therefore a good example of a diagrammatic word.

On the contrary, the degree of diagrammaticity is lower if a derivative and / or an inflected form show mere modification of the lexical base, i.e. without addition of signans. In this case, the addition of phonological sequence does not parallel the intensional addition of meaning in the morphological process of derivation or inflection. In fact, it is only a modification in the shape of the base morpheme, which realises the morphotactic aspect of the operation. For instance, denominal English verbs voicing the fricative consonant of the base are relatively less diagrammatic derivatives than affixed complex words, e.g. advice --> to advise.

2.8.1. The scale of diagrammaticity.

> The scale of diagrammaticity comprises six degrees and some subdegrees.

I. Agglutinative affixation
e.g. Eng. pig --> pig-let
Eng. sing --> sing-er

II. Affixation + modification
e.g. a) Eng. shelf --> shelv-es
b) It. Vicenza (toponym) --> Vicent-ino (ethnical noun)
c) It. Chieti (toponym) --> Teat-ino (ethnical noun)

III. Modification
e.g. Germ. Vater 'father' --> Väter 'fathers'.

IV. Metaphoricity
e.g. Eng. bottle --> to bottle

V. Total suppletion
e.g. Eng. go --> went

VI. Subtraction
e.g. Russ. logika 'logic' --> logik 'logician' First degree: Agglutinative affixation.

The most natural degree of the scale of diagrammaticity is represented by agglutinative affixation, i.e. direct suffixation or prefixation onto the lexical base. Examples of this process are some Eng. suffixed action nouns, e.g. to annul --> annulment, to conceal --> concealment , etc. or some prefixed deverbal verbs, define --> to predefine, to judge --> to prejudge, etc. These words are morphotactically transparent and semantically motivated, hence they are diagrammatic complex words. Second degree: Affixation + modification.

Both a process of affixation and a modification of the lexical morpheme realise the words of the second degree of the scale. Cases of apophony, metaphony, umlaut, consonant changes, and any kind of introflective phenomenon exemplify modification, whether in inflection or derivation. In Eng. shelf --> shelv-es there is a consonant modification, i.e. a morphonological rule affecting the final consonant of the lexeme. Moreover, the inflected word shows suffixation for plural formation. Therefore, diagrammaticity is still preserved by the operation of suffixation whereas morphotactic transparency is disturbed by the modification of the base signans. As diagrammaticity is the result of a parallelism of transparency between the sequence of morphs and the composition of meanings, complex words like shelves belong to a lower degree of diagrammaticity.

Weak suppletion (see ex. IIb in § 2.8.) constitutes a subtype of this degree. Suppletive weak forms that are also affixed belong to this level of the scale. For instance, a number of ethnical nouns in Italian are both characterized by weak suppletion of the lexeme and transparent suffixation, e.g. Vicenza --> Vicent-ino, Arezzo --> Aret-ino (vs. Agnadello --> Agnadell-ino).

Another subtype of the second degree of diagrammaticity is given by suppletive strong alternations (see ex. IIc in § 2.8.) which, nevertheless, are combined with direct affixation. Again, Italian ethnical nouns can provide several examples of strong lexical suppletion and parallel transparent suffixation, e.g. Chieti --> Teat-ino. In addition, verbal inflection in Italian also presents some significant cases of strong suppletive suffixed forms. For instance, andare 'to go' --> vad-o 'I go', va-i 'you go' (2nd Sg.) beside non-suppletive forms parlare 'to speak' --> parl-o 'I speak', parl-i 'you speak' (2nd Sg.). Third degree: Modification.

The third degree of the scale is less natural than the previous degrees (including all subtypes) because the quantity of diagrammaticity is minor. Words of the third degree are formed only via modification without affixation. For example, German has some plural forms characterised by umlaut, e.g. Vater 'father' --> Väter, Mutter 'mother' --> Mütter. On the other hand, German has also plural nouns derived by both umlaut and suffixation, e.g. Stadt 'town' --> Städt-e, Baum 'tree' --> Bäum-e (cf. Tür 'door' --> Tür-e 'doors', etc.). These suffixed forms are more diagrammatic and, therefore, represent typical instances of the second degree of the scale. Fourth degree: Metaphoricity.

All words derived by the morphological technique of metaphoricity are referable to the semiotic principle of metaphoricity, i.e. a different instantiation of iconicity (see Crocco Galèas 1997, 1998). Diagrammaticity (or constructional iconicity) requires a relation of biuniqueness between signans and signatum. Metaphoricity, on the other hand, entails the absence of any operation of affixation or modification. The inalterability of the base is a constant of all morphometaphorical rules. The output of a morphological metaphor has the same signans as the input. There is no intervention of any change with respect to the lexical base. Thus, the morphotactic level does not reflect the semantic modification. The numerous cases of denominal converted verbs in English represent a typical operation of morphometaphoricity, e.g. bottle --> to bottle, carpet --> to carpet. Conversion in English is not an aniconic operation. It is rather a metaphorical iconic operation deriving from the technique of morphometaphoricity. Fifth degree: Total suppletion.

Total suppletion[3], i.e. morphological irregular alternation without affixation is an aniconic operation subsumable under the technique of modification. Being aniconic it shows no diagrammaticity. This is a consequence of the lack of both morphemic segmentation and semantic compositionality. Total suppletion is altogether a rare and unproductive phenomenon, whose occurrence is cross-linguistically very restricted and limited to a few morpholexical domains only. Strong verbs in English are not quite good examples of total suppletion. Although they do not show segmentation into morphs (e.g. to buy --> bought), they are adequately classifiable into types or schemes of weak suppletion (see Crocco Galèas 1991a). The verb to be, however, can be regarded as a genuine case of total suppletion, e.g. to be --> was, were. Sixth degree: Subtraction.

The last degree of the scale is the least natural because it is represented by an anti-iconic operation, i.e. subtraction. Subtraction is a peculiar operation of the technique of modification. It consists in a modification of the lexical base via deletion of phonological material. The derivative or the inflected form resulting through subtraction shows a reduction of the phonological shape of the base. The countericonicity of subtraction is decisive for its morphological unnaturalness. Subtracted word-forms contravene the principle of biuniqueness of form and meaning since the semantic addition is paralleled by a reduction in the phonological shape. Due to its intrinsic antidiagrammaticity, this operation of modification is rarely used in languages and is unproductive too. Besides the denominal nouns of Russian, belonging to a restricted domain (i.e. animate nouns denoting people engaged in scientific disciplines, e.g. logika 'logic' --> logik 'logician', mathematika 'mathematics' --> mathematik 'mathematician'), subtraction is also exemplified by some geographical nouns, e.g. It. Sardegna --> Sardo, Liguria --> Ligure. Also the so-called singulative formations in Welsh can be viewed synchronically as cases of subtraction, e.g. ader-yn 'a bird'--> adar 'birds'. However, these subtracted plurals, which in diachronic perspective constitute the morphological bases of their singular forms, represent a closed and unproductive class of nouns in Modern Welsh. On the other hand, "normal" (or system-congruent) plural patterns in Welsh are diagrammatic, namely derived via addition of phonological material, e.g. trên "train" --> tren-au, llestr "dish" --> llestr-i, afon "river" --> afonydd, etc.

2.9. Discussion.

Each parameter represents a constitutive factor, that is a selected criterion of analysis of the morphological component. In other words, the parameters of Natural Morphology are factors which are used in the morphological analysis of complex words and word-forms, i.e. words having structure. On the basis of a number of factors, rooted in a semiotic superordinate level, morphological rules (i.e. language-specific realisations of typologically definable operations subsumable under a few universal techniques)[4] and complex words (i.e. objects upon which morphological rules operate)[5] are evaluated. Indeed, each factor can be expressed by a scale, along which different rules together with their corresponding complex words are located in a delimited range of degrees. According to a determined selected factor, a morphological phenomenon is 'measured' on a scale. A parameter is therefore a scalarised factor, and, as a factor, is a criterion for morphological analysis; the parameterisation of factors is nothing else but the elaboration of morphological scales of naturalness. The threshold of any morphological scale always represents the most highly preferred, that is to say the most natural solution for language users. Such a solution is better realised by natural morphological rules i.e. rules that are formally and semantically relatively less or not marked at all. Therefore, Natural Morphology claims that the most natural degrees of all morphological scales refer to those rules, i.e. those formalised processes, expressing in a higher 'quantity' a given morphological factor. For instance, agglutinative affixation, which is the first-degree-operation of the scale of diagrammaticity, gives rise to the most natural type of morphological rules according to the universal principle of diagrammaticity.

The following two points are most interesting for the present discussion.

1) The concept of naturalness is neither vague nor pretheoretical within the theory of Natural Morphology[6]. This means that, although having on the one hand extralinguistic bases depending on the psychophysical structure of human beings[7], morphological naturalness is, on the other hand, an entirely linguistic notion defined by the synergy of a semiotic level and three filters of analysis (i.e. the levels of universals, language type, and language-specific-competence respectively). This synergy gives rise to a selection of constitutive factors. Moreover, the notion of naturalness is not only rooted in the biological, psychological, and gnosiological nature of language users and linguistically defined as well, but it is also constantly verified by the analyst on the background of both internal and external evidence. Thus, the relation between extralinguistic factors, theoretical assumptions and data develops triadically by a process of continuous feedback from the most different domains of language.

2) Since parameters are scalarised factors, Natural Morphology makes scalarity the fundamental feature of its theoretical model. In this sense naturalists can reasonably claim that the relation between data and theory is neither naive nor preconceived. In other words, the model of scalarity neither excludes the many-sided variety of phenomena nor leaves the analyst dumbfounded by a welter of features. On the other hand, this very important advantage of perspective is not achieved by negation of reality which is independent of human classification, nor is it the result of superimposing an arbitrary yes-or-no categorial membership attribution.

3) As a consequence of 2) all morphological phenomena that are reducible to formalised processes[8], that is to say rules

a) are described by each and every scale of naturalness,

b) show in a computable, hence predictable way (Dressler 1985b) the incidence of a determined factor.

What a) implies is the possibility of linguistic and, in the case in point, morphological analysis. What b) implies is the possibility of functional explanation which is a necessary part of scientific and, in the case in point, morphological theory (cf. Dressler 1985a: 262-279).

4) The implicational scales of Natural Morphology allow probabilistic predictions (cf. Dressler 1985b) on distribution, frequency, and implications of morphological phenomena cross-linguistically. Because Natural Morphology takes psychological reality into account, we can claim that predictiveness is a constitutive part of morphological analysis. In fact, morphological analysis supported by empirical testing cannot be restricted to mere description. Analysis implies explanation and this, in turn, requires the possibility of making predictions. However, since no total or complete explanation is possible in science, both explanations and predictions must be partial and probabilistic.

3. Natural Morphology and Prototype Theory.

Dressler (1989) has pointed out the analogies between Natural Morphology and Prototype Theory. According to Dressler categorization within Natural Morphology may be compared generally with gradient classifications and, particularly, with Prototype Theory itself. Since a lot of morphological criteria are prototypical, not discrete, Natural Morphology and Prototype Theory should have much in common. For instance, one might observe that both theories insist on extralinguistic bases for categorization. Therefore, Dressler has proposed a multifactorial and scalar definition of derivation and inflection which, in his opinion, could well show the usefulness of Prototype Theory for Natural Morphology. On the basis of twenty criteria differentiating derivational morphology and inflectional morphology Dressler identifies both the prototypical nucleus of derivation and the prototypical nucleus of inflection. With respect to the opposite thresholds of this gradient, he recognises an intermediate domain in which he places phenomena and morphological rules either as derivationally non-prototypical or as inflectionally non-prototypical. The continuum consists therefore of four degrees:

1) prototypically derivational morphology;
2) non- prototypically derivational morphology;
3) non-prototypically inflectional morphology;
4) prototypically inflectional morphology.
According to Dressler, these degrees are comparable with the main clusters of properties situated along the gradient of a prototype category. Therefore the morphological component is regarded as a continuum. Most inflectional categories are prototypical ones, e.g. case, gender, definiteness, and possessive inflection in the noun; person, number, gender, tense, voice, mood inflection in the verb. Prototypical derivational categories are deverbal result nouns (e.g. approv-al), denominal adjectives (e.g. silk-y), and de-adjectival nouns (e.g. small-ness). Non-prototypical inflectional categories are gradation (comparative, superlative, excessive), verbal aspect, infinitive, gerund, participle, etc. There are only very few non-prototypical derivational categories, e.g. in English agent nouns (e.g. writ-er), action nouns (e.g. writ-ing), and -able adjectives (e.g. respect-able).

In my opinion, however, the following crucial difference between Rosch's prototype categories and Natural Morphology scales must be stressed. The adoption of the principle of scalarity in the case of derivation-inflection categorization and, more generally, in the whole theoretical model of Natural Morphology allows the unambiguous identification of definite points within a category; it is in relation with such prototypical points that the non-prototypical ones are defined. No one has so far underlined that the scales of morphological naturalness are constructed on the basis of two extremes which are both characterised by the presence / absence of one criterion (i.e. factor) or more than one (as in the case of the derivation / inflection scale). Besides, it must be added that morphological naturalness scales have two nuclei, i.e. two opposite degrees which constitute, in a certain sense, the limits of one category. As for the scale of diagrammaticity, the two opposite poles are represented by agglutinative affixation and subtraction respectively. In relation to these two poles , the intermediate degrees (i.e. modification + affixation, modification without affixation, morphometaphoricity, and total suppletion) are identifiable insofar as they differ from the prototype (the extent of which is fully characterised by the presence of a given factor) and gradually approximate to the opposite extreme (the extent of which is fully characterised by the absence of the same factor). That is tantamount to saying that the prototype implies the existence of an anti-prototype; thus, it is just in regard to the two determined endparts of the scale that it becomes possible to define all other phenomena which somehow relate with the selected factor converted into a parameter via scalarisation.

Nonetheless it has to be observed that the most typical structure of a naturalness scale is triggered by one single factor: each single parameter (i.e. scalarised factor) classifies a range of morphological evidence. One could therefore object that the scales of Natural Morphology repropose the discreteness of classical categorization, where categorial membership is a yes-or-no question. On the other hand, it is essential to Prototype theory that categorial attribution be realised via clusters of properties, i.e. ranges of factors. From Rosch's viewpoint a Natural Morphology scale that should be considered close enough to prototypical categorization would be the derivation / inflection scale (based on twenty criteria) rather than the diagrammaticity scale (essentially based on the unique criterion of iconicity).

In relation to these arguments there is one main point that has to be evidenced: the means of categorization in the theory of Natural Morphology is not prototypicality but scalarity. One can thus claim that prototype categorization is only one of the modi classificandi of the human mind. Cognitive linguistics, which is mostly based upon prototype categorization, is to be correctly viewed as only one of the linguistic and, more in general, gnosiological perspectives at our disposal. Moreover, both the limit and the weak point of an indiscriminate application of the prototype model to linguistic science consist in the unconstrained explanatory power or non-falsifiability of such a gnosiological theory. Since any realium can / may become in principle a member of any category, the possibility of predictability, hence explanation, is zero. With these premises, no theory, linguistic or not, can function. To sum up this argument, as far as morphological naturalness scales are concerned, we must take the following crucial feature into account: i.e. it is not absolutely decisive that there may be one or more criteria playing the role of common denominator for a given area of morphological phenomena to be analysed. Even the assumption of one single attributional factor does not exclude gradience distribution vs. discreteness. On the contrary, it is remarkably more significant that the principle of scalarity, which differentiates a positive (prototype) and a negative (antiprototype) pole, be rigorously applied. Between the prototype and the anti-prototype the intermediate degrees are distinguishable.

A theoretical approach that stresses the scalar structure of a series of morphological phenomena allows considerable predictiveness. This, in turn, makes it possible to falsify and hence to calibrate the model itself, by diminishing the generical power of prototype categorization, i.e. of this one specific gnosiological procedure.

3.1. A new parameterisation: the scale of indexicality.

Natural Morphology elaborates the semiotic principle of indexicality from the sign that Peirce calls index. Nonetheless, up to this moment there has been no proposal of a morphological parameter of indexicality. More precisely, although naturalists mention also the so-called parameter of indexicality (Dressler 1985a: 284; 308-312; 1987: 110-111), they do not refer themselves clearly to any kind of scale derivable as a consequence from a coherent definition of indexicality.

3.1.1. Premises.

According to the pattern of the other parameterised semiotic principles, I have put forward a scale of indexicality which presupposes a precise definition of morphological index. Thus, in order to elaborate a scale of indexicality the following points will have to be discussed.

1) The parameter of indexicality is based on the sign that Peirce calls index. Therefore, we must observe that from a semiotic point of view an index is a sign that focuses the attention of the interpreter on the object it denotes. The relation between an index and its object is a relation of contiguity. For example, smoke is usually considered as an index of fire; symptoms, fever, specific pains are indices of a disease. In contrast to an icon, there is neither similarity nor analogy between an index and its indexed entity. In contrast to a symbol there is no conventional link between the two entities of an indexical connection. In other words, we can claim that whereas an index is determined by a relation of contiguity, an icon and a symbol are determined respectively by an analogical and an allegorical disposition between sign and referent[9].

2) According to Natural Morphology, an index is a less natural sign than an icon because of the lack of analogy between signans and signatum. In fact, we can observe that an indexical relation is mainly based upon a certain link of contiguity between the index and the indexed element, while on the contrary an iconic relation exhibits similarity or parallelism between signans and signatum. This is the reason why for some naturalists indexicality does not represent a crucial parameter as, for instance, constructional iconicity (or diagrammaticity) does. However, there is a general disagreement among naturalists about the precise number and the different roles of naturalness parameters. In my model of Natural Morphology, rather than assigning different values to the various morphological naturalness parameters, I prefer to adopt a solution that underlines the conflictual dimension (cf. Wurzel 1994a) in the co-existence of all of them within the morphological component of a linguistic system. I therefore consider both the iconic and the indexical relation as two of the most important tendencies that shape any morphological system. For instance, let us assume that in the word explorations the lexical morpheme explore is the indexed element while the derivational suffix -ation and the inflectional suffix -s are both indices whose function is to signal the semantic content of the base explore. Given this situation, we are allowed to view the relation between the elements of a complex word as a relation of contiguity, i.e. a semiotic connection that lies upon a temporal-spatial dimension. On the other hand, if we analyse the nature of any iconic relation we find that in most cases it is describable just as a type of analogy, whether in the form of an image, a diagram, or a metaphor. One can say that what is at stake in an icon, although at different degrees, it is the principle of analogy. For instance, we speak of a diagrammatic type of analogy in the case of complex words whose morphotactic sequence mirrors biuniquely the morphosemantic sequence: e.g. in swimmers the basic content of the word is given by the base swim; to this is added the suffix -er which diagrams formally the addition of meaning in order to obtain via a derivational rule an agentive noun (swimmer) from the basic verb (to swim); furthermore, the addition of the inflectional suffix -s diagrams directly the semantic and grammatical change from singular to plural.

The two examples above show the different role of both analogy (see the analysis of swimmers), or in other terms diagrammatic iconicity, and indexicality (see the analysis of explorations) in morphology. However, I do not think it is arguable that an indexical relation is less natural than an iconic relation. One should limit oneself to recognising that analogy (i.e. iconicity) is a largely exploited device in the morphological structure of language, while spatio-temporal contiguity (i.e. indexicality) is a different kind of dimension also playing a crucial role in the morphological component - a dimension whose effects still need to be examined.

3) On the basis of the above premises, the parameter of indexicality is a scalarised morphological factor that measures the capacity of a given sign to refer itself to another sign.

4) As an indexical relation is a relation between an index and an indexed element, we must define what represents a morphological index and what its indexed referent. Any lexical morpheme is a sign which, due to its primary semantic value, is indexed by any affix, whether derivational or inflectional, or any functional (i.e. grammatical) word referring to that primary value. Thus, from the point of view of indexicality the lexical morpheme is always the indexed element, whereas derivational or inflectional affixes and any grammatical word or, more generally, any determinant is the index. Therefore, on the basis of the parameter of indexicality it is possible to recognise the indexical force as a function of the major / minor proximity between index and indexed sign within the morphemic sequence and the hierarchy of content.

5) As an index signals its lexical morpheme, it follows that the function of an index consists in indicating a semantic primary content, i.e. the lexical component of a signatum of a word. In this sense we can consider an index as a signans whose signatum is represented by the lexical morpheme to which it refers. Therefore naturalists use the relative denominations of indexical signans and indexical signatum to denote respectively the index and the indexed sign of an indexical relation.

3.1.2. The scale of indexicality.

The scale of the parameter of indexicality is a hierarchy of distance between the lexical morpheme and the relative index. It consists of four degrees and a number of subdegrees. Each degree subsumes a typology of indices. The most natural indices are the derivational morphemes, which are the most direct insofar as they indicate the lexical value of the indexical signatum. On the other hand, the inflectional affixes, which refer to the grammatical content of the indexical signatum are relatively less natural indices and so they are more peripheral to the lexical morpheme. The derivational affixes stand for the first degree of the scale, the inflectional affixes stand for the second degree. The third degree is represented by all types of functional words (e.g. articles, clitics, classifiers, etc.) which, being kinds of words and not affixes, are situated in a more distant position in relation to the indexed element and, what is most decisive, their content is more similar to that of inflectional affixes than to derivational ones. The fourth degree of the scale is represented by the syntagmatic context, that is to say by the collocation of the lexical morpheme within the phrase-marker.

I. Derivational affix
suffix: e.g. Eng. child --> child-hood
prefix: e.g. Eng. to make --> to re-make

II. Inflectional affix
suffix: e.g. Eng. to bring --> (he / she / it) bring-s
prefix: e.g. Luganda tu-li-laba ki-tabo 'we will see a book' vs. ba-li-laba bi-tabo 'they will see some books'

III. Functional word
pre-determinant position: e.g. Eng. the player
post-determinant position: e.g. Alb. gjalpë 'butter' vs. gjalpë-t 'the butter'

IV. Syntagmatic context
E.g. Eng. give me a smileN vs. I will smileV First degree: Derivational affix.

Given its proximity to the base, a derivational affix represents the best index. Since affixes vary as to the signatum they express and their position in relation to the base, we can distinguish two progressively less natural subdegrees for the first degree of the scale.

a) Morphotactically and morphosemantically transparent derivational suffix - this is the most natural index, given its relational semantic content and its psycholinguistically salient collocation in the morphemic chain, e.g. Eng. child --> child-hood;

b) Morphotactically and morphosemantically transparent derivational prefix - although this is less natural than the suffix because of its collocation to the left of the lexical morpheme, it is nevertheless cross-linguistically more favoured than other affixes (i.e. infixes, interfixes, and circumfixes) being a signal of new (rhematic) lexical information with respect to the 'given' (thematic) lexical information of the lexical morpheme, e.g. Eng. to make --> to re-make. Second degree: Inflectional affix.

As its content does not substantially modify the lexical signatum, i.e. the signifié, an inflectional affix is a weaker index. Its collocation to the right of the base is more natural from the point of view of the distance, hence as far as the indexical force is concerned. On the other hand, its collocation to the left of the indexical signatum is more salient as to the knowledge structure of the morphemic chain.

a) Morphotactically and morphosemantically transparent inflectional suffix, e.g. Eng. to bring --> (he / she / it) bring-s.

b) Morphotactically and morphosemantically transparent inflectional prefix, e.g. Lug. tu-li-laba ki-tabo 'we will see a book' vs. ba-li-laba bi-tabo 'they will see some books', where laba 'to see' and tabo 'book' are lexical morphemes, tu- and ba- correspond to the 1st and 3rd pl. pers. pronoun respectively, and -li- is an exponent of future tense. Third degree: Functional word.

Similarly to the second-degree affixes the content of the third-degree indexical signantia does not modify the meaning of the indexical signatum to which they refer. However, unlike the previous degrees, functional words do not behave as more / less natural indices according to their pre-determinant or post-determinant collocation within the syntagma whose head is the indexical signatum. In fact, pre- or post-determination is connected with the word-order type to which a language conforms.

a) Pre-determinant position. It is typically represented in the definite article of many European languages (e.g. Eng. the play-er), the classifiers of many south-eastern Asiatic languages (ex. Mal. orang lit. 'person, man' in tiga orang budak-budak 'three boys', where tiga is the numeral and budak the noun) and, generally speaking, all kinds of determinants that precede the indexed morpheme (e.g. Eng. three adapt-ation-s, some work-er-s);

b) post-determinant position: for instance, the postponed definite article in Albanian (e.g. gjalpë 'butter' vs. gjalpë-t 'the butter'), the clitic possessive adjective of Modern Greek (e.g. to pukámisó-mu lit. 'the shirt-of-me' = 'my shirt')[10]. Fourth degree: Syntagmatic context.

The weakest index corresponds to the syntagmatic collocation. Complex words that typically recur to the syntagmatic context for indexical purposes are morphological metaphors (e.g. Eng. give me a smileN vs. I will smileV). In this example the lexeme smile is not indexed by an affix but the type of syntagmatic collocation that disambiguates its word-class undeniably signals it (cf. Crocco Galèas 1990, 1991b, 1997, 1998).

3.1.3. Some remarks on indexical signatum.

In the previous section, I have assumed that the indexical signatum is the lexical morpheme signalled by derivational and / or inflectional affixes, functional words, pre- / post-determinants, and syntagmatic collocation.

As the lexeme is the indexed sign, the indexical force of all other signs is established in accordance with the type of meaning bearing upon the lexeme and the degree of proximity to it. Thus, derivational affixes are necessarily the most prototypical (or natural) indices insofar as they consistently affect the meaning of the lexeme. Furthermore, due to their position in the morphemic chain, they are efficient indexical signantia. Still, in assigning the role of indexical signatum to the lexeme, I have inevitably neglected a complementary indexical function specifically performed by inflectional affixes. These, indeed, point not only to the lexical morpheme but also to the syntactic categories expressed in the phrase or sentence and, therefore, can be seen as good indexical signantia in relation to the morphosyntactic interface. In other words, their peripheral position within the word is a consequence of their simultaneous signalling both to the lexeme and to the other words of the syntagmatic and syntactic environment. However, since morphology is the domain of word, in the sense of monomorphemic or polymorphemic free form, the parameter of indexicality can only refer to the indexical signatum within the complex word, i.e. the lexeme. As for all other parameters of morphological naturalness / markedness, indexicality must be evaluated within the word. Hence, inflectional affixes are necessarily less natural indices than derivational affixes. In particular, inflectional morphemes modify the meaning of the lexeme in a different way to derivational morphemes, that is to say in a more abstract manner. In effect, their content is more similar to that of functional words, whereas the semantic content of derivational affixes is closer to the lexical meaning of roots / stems / bases.

Since the double signalling toward lexemes and syntactic categories characterizes inflectional affixes only, the scale of naturalness that I have proposed takes into account one type of indexical signatum and, consequently, arranges the indexical signantia that point to it in a scalar manner.

3.1.4. Synthesis.

The scale of indexicality is based upon the gnosiological procedure of scalarity. At this point I will briefly sum up the process that led me to the formulation of a new scale of morphological naturalness. First, there is an area of morphological phenomena to describe, i.e. the formal and semantic links between lexemes and affixes. Second, the factor which is internal to the analysis of these phenomena is that of interrelation among signs and within complex signs. Third, the factor is expressible in a semiotic principle, i.e. the principle of indexicality. Fourth, the semiotic principle constitutes the basis for the parameterisation of the factor, i.e. the elaboration of the parameter of indexicality. Fifth, the scale is the direct consequence of the parameterisation of the factor. Finally, the introduction of the criterion of parameterisation is the result of the application of the gnosiological principle of scalarity. This principle allows to locate on a definite gradient objects and relations that morphologists observe.

4. Conclusion: Three types of categorization.

Scalarity is a fundamental gnosiological concept which represents also a substantial principle more or less explicitly recurrent in very many classificatorial operations. Linguists have not yet paid this principle due attention. However, some consequences should already necessarily follow from its recognition. The first and most important of all consists in the theoretical assumption of three ways of categorizing alternatively adopted by the human mind:

1) classical categorization, which is based upon the principle of discreteness of the categorial properties attributed to reality;

2) prototypical categorization, which is based upon a non-discreteness principle of attribution of properties to a categorial continuum;

3) scalar categorization, which is indifferent to discreteness / non-discreteness membership properties and is rather based upon the assignment of category boundaries (the prototype and the antiprototype) and the related identification of an intermediate gradient.

These three ways of categorizing belong to the human mind and are therefore applicable and actually applied in the apprehension of reality. According to the varying cognitive necessities realia appear to be adequately categorizable on the basis of one of the aforementioned classification models.

Given the theoretical recognition of both classical and prototypical categorization, the third modus classificandi, which can be properly called scalar categorization, coexists with the other two, even though it has not as long as a scientific tradition. Nevertheless, this further procedure of categorization is actually used in science and, primarily, by human beings. Its history needs only to be outlined as was the case with the prototype theory of categorization.

Finally, we can say that Prototype Theory is neither an absolute substitute for classical categorization nor the unique alternative to it. Prototype Theory just represents one of the three modi classificandi used by human beings; therefore, cognitive linguistics is one of the possible theoretical main perspectives of linguistics and, generally, of gnosiology. The extremely powerful explanatory power of prototype categorization, constitutes both the most salient feature and also the weak point of a scarcely falsifiable theory.

If we move from the presupposition that classical and prototype categorization are two opposite forms of pertinentisation of reality, it results that, in a certain sense, scalar categorization stands for an eclectic solution. As a matter of fact, this third category type mediates between the rigidity of the discreteness of the classical principle and the elusive vagueness of the prototypicality principle. However, the character of eclecticism is not to be intended as a superficial compromise with little or no scientific and gnosiological foundation. A scale, i.e. an opposition of extremes, which is applied to an area of realia allows the fixation of a number of definite points, i.e. degrees, without giving up the awareness of osmotic pressure among the various forms of phenomenological reality.

By explicitly using the principle of scalar categorization in the apperception of morphological phenomenology, Natural Morphology can avail itself of a theoretical model endowed with both greater flexibility, in relation to classical categorization, and less powerful generation, in relation to Prototype Theory. That is tantamount to admitting the actual possibility of predictability and functional explanation beyond mere descriptiveness of data.


Berko, Jean, 1958. The child's learning of English morphology. In: Word 14.150-177.

Berlin, Brent & Paul Kay, 1969. Basic Color Terms: Their Universality and Evolution. Berkeley: University of California Press.

Bittner, Andreas, 1988. Reguläre Irregularitäten. Zur Suppletion im Konzept der natürlichen Morphologie. In: Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 41/4.416-25.

-- 1990. Eine unendliche Geschichte? Nochmal zum Verhältnis von Suppletion und Natürlichkeit. In: N. Boretzky / W. Enninger / Th. Stolz (eds), Spielarten der Natürlichkeit - Spielarten der Ökonomie. Bochum: Bochum-Essener Beiträge zur Sprachwandelforschung 8/2.227-48.

Chini, Marina & Grazia Crocco Galèas, 1995. Verbalparadigmen, Prototypizität und Natürliche Morphologie: Überlegungen am Beispiel des Erstspracherwerbs des Italienischen. In: Quaderni Patavini di Linguistica 14.73-110.

Clark, H.H., 1996. Using Language. Cambridge, New York: Cambridge University Press

Crocco Galèas, Grazia. 1990. Conversion as Morphological Metaphor. In: J. Méndez Dosuna & C. Pensado (eds). Naturalists at Krems. Salamanca: Ediciones de la Universidad de Salamanca. 23-32.

-- 1991a. Gli etnici italiani. Studio di morfologia naturale. Padova: Unipress.

-- 1991b. La metafora morfologica in italiano. In: Quaderni Patavini di Linguistica 10.73-104.

-- 1995. Alcuni aspetti della teoria di Morfologia Naturale. In: M. Dardano, W. U. Dressler, C. Di Meola. (eds). Parallela 5. Roma: Bulzoni. 113-56.

-- 1997. Metafora morfologica. Saggio di morfologia naturale. Padova: Unipress.

-- 1998. The Parameters of Natural Morphology. Padova: Unipress.

Cutler, Anne, Hawkins, J.A., Gilligan, C. 1985. The suffixing preference: A processing explanation. In: Linguistics 23.723-58.

Dressler, Wolfgang U. 1985a. Morphonology: The Dynamics of Derivation. Ann Arbor: Karoma.

-- 1985b. On the Predictiveness of Natural Morphology. In: Journal of Linguistics 21.321-37.

-- 1985c. Suppletion in Word-formation. In: J. Fisiak (ed.). Historical Semantics, Historical Word-formation. Berlin: Mouton de Gruyter. 97-112.

-- 1986. Inflectional Suppletion in Natural Morphology. In: B. F. Elson (ed.). Linguistics in Global Perspective. Dallas (Texas): The Summer Institute of Linguistics. 97-111.

-- 1987. Naturalness in Word-formation. In: Dressler et al. (1987:99-126).

-- 1989. Prototypical differences between inflection and derivation. In: Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung. 42.3-10.

Dressler, Wolfgang U. & Gianfranco Denes. 1988. Word-Formation in Italian-Speaking Wernicke's and Broca's Aphasics. In: Dressler & Stark (1988:69-81).

Dressler W.U. & J. Stark (eds.), 1988. Linguistic Analysis of Aphasic Language. Berlin: Springer.

Dressler, Wolfgang U. / Mayerthaler, Willi / Panagl, Oswald / Wurzel, Wolfgang U. 1987. Leitmotifs in Natural Morphology. Amsterdam: John Benjamins.

Hjelmslev, Louis. 1968. I fondamenti della teoria del linguaggio. Transl. by Lepschy, G.C. Torino: Einaudi. (Prolegomena to a Theory of Language).

Jakobson, Roman. 1971. Quest for the essence of language. In: R. Jakobson, Selected Writings. Vol. 2: Word and Language. The Hague / Paris: Mouton. 345-59.

Johnson, Mark, 1987. The Body in the Mind. The Bodily Basis of Meaning, Imagination and Reason. Chicago and London: University of Chicago Press

Kay, Paul & McDaniel, Chad. 1978. The Linguistic Significance of the Meanings of Basic Color Terms. Language 54. 610-46.

Kilani-Schoch, Marianne. 1988. Introduction à la morphologie naturelle. Berne: Peter Lang.

Kleiber, George. 1990. La sémantique du prototype. Catégories et sens lexical. Paris: Presses Universitaires de Françaises.

Labov, William. 1973. The boundaries of words and their meanings. In: Bailey, C.-J. N. & Shuy, R. W. (eds), New Ways of Analysing Variation in English. Washington: Georgetown University Press. 340-373.

Lakoff, George. 1982. Categories: An essay in Cognitive Linguistics. In: Linguistics in the Morning Calm. Seoul Hanshin: Linguistic Society of Korea. 139-93.

-- 1987. Women, Fire and Dangerous Things. What Categories Reveal about the Mind. Chicago: The University of Chicago Press.

Lieb, Hans-Heinrich. 1992. Integrational Linguistics: Outline of a Theory of Language. In: H. H. Lieb (ed.), Prospects for a New Structuralism. Amsterdam: John Benjamins. 127-82.

Luschützky, Hans C., forthcoming. Jüngere Entwicklungen in der Natürlichen Morphologie. In: S. Auroux et al. (eds), Geschichte der Sprachwissenschaften. Berlin: de Gruyter.

MacWhinney, Brian. 1978. The Acquisition of Morphophonology. Chicago: Chicago University Press. (Monographs of the Society for Research in Child Development 43.)

Mayerthaler, Willi. 1981. Morphologische Natürlichkeit. Wiesbaden: Athenaion.

Mel'cuk, Igor. 1976. On Suppletion. In: Linguistics 170.45-90.

-- 1994. Suppletion: Toward a Logical Analysis of the Concept. In: Studies in Language 18/2.339-410.

Peirce, Charles S. 1965. Collected Papers. Cambridge: Harvard University Press. (Edited by C. Hartshorne & P. Weiss.)

Ronneberger-Sibold, Elke. 1988. Entstehung von Suppletion und Natürlicher Morphologie. In: Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 41/4.453-62.

Rosch, Eleanor (Eleanor Heider). 1973a. Natural Categories. In: Cognitive Psychology 4.328-50.

-- 1973b. On the internal structure of perceptual and semantic categories. In: Moore, T.E. (ed.), Cognitive Development and the Acquisition of Language. New York: Academic Press. 111-44.

-- 1975a. Cognitive Reference Points. In: Cognitive Psychology 7.532-47.

-- 1975b. Cognitive Representations of Semantic Categories. In: Journal of Experimental Psychology: General 104.192-233.

-- 1977. Human Categorization. In: N. Warren (ed.), Studies in Cross-cultural Psychology.. London: Academic Press.

--1978. Principles of Categorization. In: E. Rosch & B. B. Lloyd (eds), Cognition and Categorization. Hillsdale (N.J.): Lawrence Erlbaum Associates. 27-48

-- 1981. Prototype Classification and Logical Classification: The Two Systems. In: E. Scholnick (ed.), New Trends in Cognitive Representation: Challenge's to Piaget's Theory. Hillsdale (N.J.): Lawrence Earlbaum Associates. 73-86.

Rosch, Eleanor & Mervis, Carolyn B. 1975. Family Resemblances: Studies in the Internal structure of Categories. In: Cognitve Psychology 7.573-605.

Seiler, Hans-Jakob (ed.) 1978a. Language Universals. Tübingen: Narr.

-- 1978b. The Cologne project on language universals. In: Seiler (1978a:11-25).

-- 1979. Language universals research, language typology and individual grammar. In: Acta Linguistica Hungarica 29.353-367.

Skalicka, Vladimir. 1979. Typologische Studien. Braunschweig: Vieweg.

Taylor, John. 1989. Linguistic Categorization. Oxford: Clarendon Press.

Vennemann, Theo. 1983. Causality in language change: Theories of linguistic preferences as a basis for linguistic explanations. In: Folia Linguistica Historica 6/1.5-26.

Wittgenstein, Ludwig. 1978. Philosophical Investigations. Oxford: Basil Blackwell. (English translation by G.E.M. Anscombe.)

Wurzel, Wolfgang U., 1984. Flexionsmorphologie und Natürlichkeit. Berlin: Akademie Verlag.

-- 1985. Die Suppletion bei den Dimensionsadjektiven. In: Linguistische Studien des ZISW der AdW der DDR. Reihe A. 126.114-43.

-- 1990. Gedanken zu Suppletion und Natürlichkeit. In: Zeitschrift für Phonetik, Sprachwissenschaft und Kommunikationsforschung 43/1.86-91.

-- 1994a. Skizze der natürlichen Morphologie. In: Papiere zur Linguistik 50/1.23-49.

-- 1994b. Natural Morphology. In: R. E. Asher & J. M. Y. Simpson (eds), The Encyclopaedia of Language and Linguistics. Oxford: Pergamon Press. 2590-8.


[1] As to the differences between classical and prototypical categorisation cf. Lakoff (1982, 1987), Taylor (1989).

[2] Among the most significant illustrations of Natural Morphology see Mayerthaler (1981), Wurzel (1984), Dressler (1985a), Dressler et al. (1987), Kilani-Schoch (1988), Wurzel (1994a, b), Crocco Galèas (1995, 1997, 1998), Lusch[[cedilla]]tzky (to appear).

[3] Within the framework of Natural Morphology scholars have frequently tackled the topic of suppletion. Dressler (1985c, 1986) has dealt with both derivational and inflectional suppletion; Crocco Galèas (1991a: 109-69) has proposed a scalar classfication of suppletion based on a number of criteria, e.g. the concept of scheme and that of submorpheme; other naturalists who have treated this morphologically unnatural phenomena are Wurzel (1985, 1990), Bittner (1988, 1990), Ronneberger-Sibold (1988). See also Melcuk (1976, 1995).

[4] For this definition of rule and the related definitions of operation and technique see also Crocco Galèas (1997: 14-15).

[5] For this interpretation of linguistic rule I refer to Lieb (1992).

[6] Cf. Mayerthaler (1988:1), Dressler et al. (1987:3-4).

[7] Several types of linguistic evidence, such as cross-linguistic frequency, simplicity, salience, ease of processing and learning, diachronic persistence, etc. derive from the extralinguistic bases of Natural Morphology.

[8] For an exhaustive discussion on rules and processes see Crocco Galèas (1991a: 128-36).

[9] 9 A symbol stands for, represents something else by some accidental or conventional relation. In characterising the symbol and its opposition to both the icon and the index, I recur to the notion of allegoric relation because I intend to stress that a symbol always establishes a conventional, hence abstract, unpredictable correspondence between two levels or two entities.

[10] The possessive adjectives of Modern Greek are primarily personal oblique pronouns.

Return to The Issue 3 Contents Page |