Conférences d'Andrew Spencer

Fri Mar 23 08:55:25 UTC 2007

ANNONCE

CONFERENCES D’ANDREW SPENCER

  Andrew Spencer, professeur à l’université d’Essex (Grande- 
Bretagne), invité par l’université de Paris 7, sera l’hôte du  
Laboratoire de Linguistique Formelle du 23 avril au 20 mai 2007. Il  
donnera une série de quatre conférences sur les représentations  
lexicales dans une morphologie à fonctions paradigmatiques  
généralisées.
PROGRAMME

Jeudi 26 avril          An overview of Paradigm Function Morphology

Jeudi 3 mai            Types of lexical relatedness

Jeudi 10 mai           Empirical issues in inflection

Mardi 15 mai           Generalized Paradigm Function Morphology

Lieu : LLF, 1er étage, 30, rue du Château des Rentiers, 75013 PARIS

Horaire : 14h-16h. Salle : sera précisée sur place.

Transports : Métro, bus PC2, T3: Porte d'Ivry; bus 83: Marcel Duchamp

Andrew Spencer est un spécialiste de morphologie mondialement connu.  
Il est l’auteur de Morphological Theory. An Introduction to word  
structure in Generative Grammar. 1991. Oxford: Basil Blackwell. Il a  
coordonné avec A. Zwicky The Handbook of Morphology. 1998. Oxford:  
Blackwell Publishers. Il s’intéresse principalement aux problèmes  
de morphologie flexionnelle (structure des paradigmes, déponence,  
cas, allomorphie) et aux rapports entre syntaxe et morphologie  
(clitiques).

Le programme des conférences se trouve aussi sur le site du  
laboratoire LLF. Il est détaillé plus bas.

Contact : bernard.fradin at linguist.jussieu.fr

PRESENTATION DES CONFERENCES

  MORPHOLOGY AND LEXICAL REPRESENTATION IN GENERALIZED PARADIGM  
FUNCTION MORPHOLOGY

  The four lectures will provide a survey of some of the aspects of  
morphology and morphosyntax that I have been researching on in recent  
years. I shall concentrate on developments within the so-called 'Word- 
and-Paradigm' class of models (also known as 'realizational- 
inferential' models), well-known from the work of Anderson, Aronoff,  
Beard, Corbett, Stump, Zwicky and others. This class of models  
undoubtedly represents the most influential approach to inflection  
and has had a noticeable impact on all other approaches, including  
Wunderlich's 'Minimalist Morphology', Halle and Marantz's  
'Distributed Morphology' and various approaches to morphology couched  
within an Optimality Theoretic framework. The realizational- 
inferential approach has also influenced lexicalist models of syntax,  
including LFG but especially HPSG.

I shall present an extension of Stump’s Paradigm Function  
Morphology, which I call Generalized Paradigm Function Morphology.  
Having summarized the classical model I shall present a number of  
empirical issues in need of resolution. I then sketch briefly the way  
that the GPFM model handles those issues.

Lecture 1         An overview of Paradigm Function Morphology

  This lecture sets the scene by discussing the conceptual issues  
underpinning contemporary work in morphology, taking as a starting  
point Stump’s model of Paradigm Function Morphology. I illustrate  
the notion of the ‘paradigm function (PF)’, a function which maps  
the root of a lexeme and a complete set of inflectional features to a  
cell in the lexeme’s paradigm: PF(<root, features>)=def  <word form,  
features>. I pay particularly attention to the role of defaults in  
the system and discuss the role played by the concept of  
‘paradigm’. I also summarize the devices used in PFM to ensure  
that morphs are placed in the correct linear order.

In addition, I draw attention to a number of problematical aspects of  
the model. First, it’s difficult to see how the model can handle the  
full array of types of lexical relatedness (there is only very brief  
discussion of derivational morphology in Stump 2001). Second, it’s  
difficult to see how to extend the model so as to handle clitic  
systems. This is of importance because it’s often very difficult to  
distinguish between highly grammaticalized clitic systems and affix  
systems which have not been fully integrated into the morphology.  
Third, there are certain redundancies in the way that PFM handles  
linear ordering.

  Lecture 2         Types of lexical relatedness

  In this lecture I examine the different ways that lexemes can be  
related to each other, basing myself on Spencer (2005, in press). I  
distinguish between four attributes for a lexical entry:

FORM (specification of the root form, irregular stem forms,  
morphological class, …)

SYNTAX (specification of syntactic class and/or argument structure,  
selectional/collocational restrictions…)

SEMANTICS (some sort of linguistic semantic/conceptual representation)

LEXEME INDEX (unique identifier serving to individuate lexemes from  
each other, e.g. in cases of polysemy).

The SEM attribute and some of the SYN sub-attributes may well be  
empty, e.g. in the case of auxiliary elements which have a purely  
grammatical function but no meaning (auxiliary verbs, grammatical  
prepositions, light verbs, …)

One important feature of the characterization of the lexical entry is  
the possibility for a lexeme to have a Morphological Class  
specification which is distinct from its Syntactic Class  
specification. A simple example is found when an adjective is  
converted to a noun, and retains all its adjectival morphology  
despite behaving like a noun in the syntax. Other, more complex,  
examples of such mismatches will be presented.

Lexical relatedness can now be thought of as a relation between the  
four principal attributes of lexical entries and their sub- 
attributes, in which some or all of the relations can be trivial.  
Where all four attributes are changed we have classical derivational  
morphology. Where only the FORM attribute is changed (so as to  
realize inflectional features) we have (contextual, functional,  
‘meaningless’) inflection. However, I show that there is a host of  
other types of relatedness that can be defined, including  
transpositions, such as deverbal participles or deverbal (action)  
nominalizations.

  Lecture 3         Empirical issues in inflection

  There are a number of ways in which the classical PFM model is  
insufficient to describe all the phenomena of interest to students of  
morphology and morphosyntax. Here I look at three types of system,  
each of which poses descriptive challenges to the original model. I  
first consider the clitic-affix distinction. Examples from European  
Portuguese (Luís & Spencer 2005) and Polish show that there's no  
clear-cut distinction between these notions. In particular, in  
Portuguese one and the same set of formatives behave like affixes in  
some contexts and clitics in other contexts.

Then I consider the stem-affix distinction, looking at the Hungarian  
case system. Again, the case formatives have both affix-like and root/ 
stem-like properties. Pronouns fail to take case endings. Instead,  
meanings such as ‘to-me’ or ‘from-them’ are expressed by  
inflecting the case ending with possessor agreement affixes.

Finally, I discuss question of stems. There are two aspects to this.  
First, many languages have meaningless, purely formal stems  
(Aronoff’s ‘morphomic stems’) which are nonetheless formed by  
regular rules of inflectional morphology, just like other inflections  
(indeed, this is true of Aronoff’s own example of the ‘third  
stem’ in Latin). Second, in many languages with complex position  
class morphology we find that word forms have complex and  
discontinuous stems in part or all of their paradigms. I illustrate  
this with examples from the Siberian language isolate, Ket.

The problem of discontinous stems and the affix-like clitics  
highlights the need to re-appraise the nature of linear order in PFM.  
The problem of regular stem formation rules highlights the need to  
allow rules of morphology to be responsible for pure forms as well as  
realizing feature values. The problem of root-like affixes highlights  
the need for the model to make reference to morphological objects  
rather than morphophonological operations.

  Lecture 4         Generalized Paradigm Function Morphology

  In this lecture I present a modification of PFM which is designed  
to answer all the questions raised so far, Generalized Paradigm  
Function Morphology. In GPFM we separate out the exponence/ 
realization of a set of features from the linearization and placement  
of the morphs (e.g. affixes) that realize those features. Thus,  
instead of a realization rule along the lines

Plural(N) ⇒ N-s

which specifies that -s is suffixed to the stem N, we split the rule  
into two basic parts:

		Plural(N):            Host             =             N

         	Exponence             =             s

             Linearization            =            suffix

The paradigm function is now replaced by a Generalized Paradigm  
Function (GPF). Instead of mapping <root, features> to a cell in the  
paradigm the GPF maps a complete representation of a lexeme and a set  
of features to another representation of a lexeme and that same set  
of features. Where we have completely regular derivational morphology  
(as opposed to idiosyncratic and non-productive lexical relatedness)  
the GPF maps the value of all four attributes of the lexeme to non- 
trivially changed values of those attributes, i.e. the GPF maps a  
lexeme onto a distinct, derived lexeme. Where we have meaningless  
inflectional morphology the GPF only induces a non-trivial mapping  
for the FORM attribute, in effect mimicking the classical paradigm  
function. The other three attributes, SYN, SEM, LI, are the values of  
an identity mapping. However, a great variety of other mappings is  
possible, corresponding to the various types of lexical relatedness  
discussed in Lecture Two.

The GPF approach allows us to make very general statements about  
linear ordering of morphemes, including discontinuous stems. It also  
allows us to integrate the stem-formation rules gracefully with the  
rest of the realizational morphology. Because the rules of exponence  
manipulate morphs, i.e. morphological objects, these can be defined  
independently of the realization rules, effectively being treated  
like meaningless ‘morphemes’. This allows us to do several things.  
First, we can readily handle descriptive problems such as affixes  
which double as roots (Hungarian case). Second, we can handle  
partially grammaticalized clitic clusters by judicious definition of  
what counts as a host for linearization. In this fashion we have a  
satisfactory descriptive framework for handling all of the  
problematical issues raised in the earlier lectures.

------------------------------
Cordialement.

Bernard Fradin
Tél.  33 (0) 1 57 27 57 84
Adresse postale / postal address
Laboratoire de linguistique formelle
Case 7031, 2 place Jussieu
F-75251 PARIS CEDEX 05
Adresse géographique /geographical address
30, rue du Château des Rentiers
F-75013 PARIS
Métro, bus PC2, T3: Porte d'Ivry; bus 83: Marcel Duchamp

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listserv.linguistlist.org/pipermail/parislinguists/attachments/20070323/e8a6c5be/attachment.htm>