1. Introduction
Suppletion has elicited two sharply contrasting reactions in the literature. One view sees it as marginal and of minimal significance to the working morphologist. Thus Halle & Marantz (1993: 113) write: '… suppletion is not of central importance in the morphology of English or of any other language …', and therefore dismiss its evidential value in deciding questions of morphological theory. The alternative view, articulated among others by Carstairs-McCarthy (1994) and to which we also subscribe, acknowledges the relative marginality of suppletion in morphological systems but nonetheless recognises its considerable diagnostic value in understanding the structure of those systems in which it is found.
If there has been diversity amongst scholars as to the significance of suppletion, there has been unity in seeing it as an exclusively morphological phenomenon. We will suggest instead that suppletion in its classic sense shares important properties with periphrasis, and that the study of suppletive patterns in this wider sense can have important consequences for the overall architecture of a grammar.
After a brief theoretical prelude (§2), we will first argue
the case for suppletion as a morphological diagnostic (§3),
then show how the same line of argument can be extended to periphrastic
formations (§4), before concluding with a more complex case
in which both suppletion and periphrasis are interwoven in an
intricate historical scenario (§5). The burden of the empirical
data adduced in §§3-5 will be to support a model of
morphology and syntax based on features rather than projections,
and our concluding §6 will elaborate this theoretical consequence
in more detail.
2. The syntax-morphology interface
Suppletion is one of a number of phenomena that have led researchers over the years to challenge the validity of a universally applicable morpheme-based theory of morphological representation (cf Matthews 1972, Anderson 1992, Beard 1995). The latter is more appropriate the nearer a language approaches the agglutinative ideal, in which complex morphological items are readily segmentable into smaller parts ('items' in the famous terminology of Hockett 1954) whose co-occurrence is guaranteed by distributional statements ('arrangements'). Within early transformational grammar, some of the problems posed by non-agglutinative morphology could be handled by a two-stage model which attributed a key role to the phonology. First, the syntax delivered essentially IA representations. Then, a phonological component equipped with enough power in terms of boundaries, extrinsically ordered rules, and abstract underlying representations effected the necessary changes to yield the final surface phonetic representations. This is essentially the position adopted in The Sound Pattern of English and much work of that era.
More recently, attention has concentrated on the syntactic side of the problem, and a model has evolved in which seemingly every morphosyntactically relevant feature heads its own functional projection. Items belonging to the lexical categories N, V, and A then move through a succession of functional heads picking up all the morphologically relevant properties as they go. (See Williams 1996 for an effective diagnosis and critique of the analytical strategies involved.) Either way the goal is to reduce morphology to syntax or to phonology or to some interaction of the two rather than recognising it as an independent domain with its own structure and organizing principles (see Déchaine 1996 for a particularly vigorous endorsement of this research strategy, and Börjars, Vincent & Chapman 1996: §1 for a brief response).
The present paper is in part a sequel to Börjars, Vincent & Chapman (1996), and offers further support for the conclusions reached there, namely:
(i) that it must be recognised that some, if not all, morphological systems require to be analysed in terms of morphosyntactc representations which consist of feature bundles rather than X-bar projections;
(ii) that nonetheless there is an inevitable continuity between morphology and syntax both synchronically and diachronically;
(iii) that, in consequence of (i) and (ii), the best model of
the morphology-syntax interface will be featural rather than configurational.
Put another way, we seek to exploit the shared recognition by
word-and-paradigm morphology and feature-based syntax that the
common and universal metalanguage of morpho-syntactic description
is one involving values and attributes (f-structure). By contrast,
as has consistently been recognised in work within LFG, configuration
in syntax and linear arrangement in morphology (c-structure) are
aspects of the realizational system. They may be appropriate to
(some parts of) some languages but there is neither evidence nor
need for them to be generalized to all aspects of all languages.
3. Suppletion
Two examples will serve to show how suppletion may reveal the dimensions of a morphological system in a way more regular formations do not.
3.1 The English verb 'be'
Consider Table I, which sets out the forms required for
an adequate
morphosyntactic description of the English verb.1
| walk | take | go | be | |
| base | walk | take | go | be |
| pres non 3sg | walk | take | go | am/are |
| pres 3 sg | walks | takes | goes | is |
| past | walked | took | went | was/were |
| past part | walked | taken | gone | been |
| -ing form | walking | taking | going | being |
|
| ||||
A regular verb such as WALK does not formally distinguish the
Past Participle and the Past Tense forms, but both irregular verbs
like TAKE and the suppletive GO do so. However, when it comes
to distinguishing the base form from the present non-3rd person
singular, the only verb in the language which realises that distinction
in overt morphology is the suppletive BE (Huddleston 1984: 85).
3.2 The Romance verbs 'go'
Romance languages and dialects show a
wide range of suppletive
patterning in the verb meaning 'go', built up in various waves
of morphological change and replacement from three etymologically
distinct verbs:2 ire, vadere, and a
third verb of controversial
origin which shows up in French as aller and Italian as
andare.3 The first stage of the
change was phonologically
driven and involved replacement of those forms of ire that
regular sound changes had reduced to monosyllables by forms of
vadere. Hence, we find in Old Spanish, for instance, the
present indicative sub-paradigm as follows:
(1)
| Old Spanish | Proto-Romance | ClassicalLatin | Latin | ||||
| 1PSG | vo | < | vado | < | jo | < | e.o |
| 2PSG | vas | vadis | is | ||||
| 3PSG | va | vadit | it | ||||
| 1PPL | imos | imus | imus | ||||
| 2PPL | ides | itis | itis | ||||
| 3PPL | van | vadunt | junt | < | e.unt |
At this stage the non-finite forms ire (infinitive) and itum (past participle), being polysyllabic, are not affected and come through into Spanish as ir and ido.4
In the next stage of the development, the surviving i-stem
forms are replaced - in Spanish by v-stem forms but
only in the finite series, and in French and Italian by a-stem
forms for both finite and non-finite series. Hence the following
(partial) patterns in Modern Spanish, Italian and French:
(2)
| Spanish | Italian | French | |
| 1PSG.PRES | voy | vado | vais |
| 2PSG.PRES | vas | vai | vas |
| 3PSG.PRES | va | va | va |
| 1PPL.PRES | vamos | andiamo | allons |
| 2PPL.PRES | vais | andate | allez |
| 3PPL.PRES | van | vanno | vont |
| INF | ir | andare | aller |
| PPART | ido | andato | allé |
These changes suggest that what were in origin
independent lexemes
with their own PRED features have merged into a single lexeme
with a single PRED feature, say for French PRED = 'aller
<(SUBJ)>'.5
It is also clear that the viability of the morphosyntactic representations
(MSR) of the forms in question are not affected by the changes
in their phonological realization (PR). It is this independence
of MSR and PR which lies behind the arguments of Anderson (1992),
Aronoff (1994) and Beard (1995) for so-called 'separationist'
morphology (Aronoff 1994: 8), a conclusion in morphology which
is very much in the same spirit as the syntactic arguments within
LFG for the independence of f-structure and c-structure.
The diagnostic value of suppletion is revealed in a further aspect
of the development of the Romance 'go' verbs. Compare the future
forms set out below:
| italian | spanish | french | |
| infinitive | andare | ir | aller |
| future | andrò, etc | irè, etc | irai, etc. |
Etymologically, we know that this series derives from the infinitive
plus the present tense of the Latin verb HABERE 'to have':
(3)
amare + habeo > aimerai 'I will love'
amare + habes > aimeras 'you (sg) will love'
amare + habet > aimera 'he/she will love'
amare + habemus > aimerons 'we will love'
amare + habetis > aimerez 'you (pl) will love'
amare + habeunt > aimeront 'they will love'
The transparency of the connection between infinitive and future
stem in Modern Spanish reflects, as we have seen, their shared
historical source. The fact that this historical connection is
still discernible in Modern Italian reveals that, for that language,
at the point at which the a-stem replaced the i-stem,
the construction must still have been synchronically analyzable
as one in which the future was built out of the infinitive either
as a true periphrasis or as what Matthews (1972: 85-6) calls a
'Priscianic' or 'parasitic' formation. Now the French
data are
crucial. The fact that the future stem remains as ir- even
though the a-stem replaces the infinitive and the past
participle exactly as in Italian shows that the French future6
was at the relevant time an independent morphological formation
and not a periphrastic or Priscianic form. In other words, the
suppletive form is a valuable indicator of an earlier periphrastic
formation which has become, as it were, fossilized inside a separate
sub-paradigm and thus reveals - in a way non-suppletive forms
could not - the historical links between periphrastic and inflectional
forms. The synchronic consequence of the existence of such links
will be developed in the next section.
4. The analysis of periphrases
In Börjars, Vincent & Chapman (1996) we used the pronoun system of certain English dialects and the Latin passive to illustrate the way in which syntactic and morphological elements can share crucial properties. Here we will make a similar point by discussing some striking data from Kashmiri.
Most of the modern Indic languages and also Kashmiri (a member of the Dardic family) provide examples of inflecting postpositions (Payne 1995). In these languages, two types of case are usually distinguished, traditionally referred to as Layer I and Layer II case. Using the Hindi examples discussed by Payne (1995:283-287) the phenomenon can be schematically illustrated as in (4) below. In this example, the Layer II case marker is the Possessor, and the Layer I cases Direct (DCT) and Oblique (OBL) are used.
(4) a. Ra#m ke bha#i# ka# bha#ti#ja#
Ram GEN=OBL.SG.M brother=OBL.SG.M GEN=DCT.SG.M cousin.DCT.SG.M
'Ram's brother's cousin' (direct case)
Here we can see how the postposition of the modifying possessor phrase agrees in case (and also in number and gender, but that is not our concern here) with the modified noun - for NP1 this is DCT, and for NP2 it is OBL. So, the postposition which is a Layer II case marker is itself marked for a Layer I case. The Layer I cases can be described as grammatical, DIR is the case used for instance for subjects of intransitive sentences and indefinite objects, OBL noun phrases are governed by a Layer II case marker/ postposition. Hence the case of NP1 in (4b) indicates the NP's grammatical function within the sentence and the case of NP2 results from the governing postposition. The Layer II cases can be described as semantic cases, defining more subtle relationships, like instrumental, location or possessor.
However, it is not the phenomenon of double case marking in itself - or suffixaufnahme - fascinating though it is, which is the main focus of the present paper.7Our point here rather is the way in which syntax and morphology interact in the paradigm for the inflected postpositions in one of the languages discussed by Payne (1995), namely Kashmiri.
In Kashmiri, the picture sketched in (4b) is made more complex by the fact that the form of the genitive marker depends not only on the case, number and gender of the modified noun, but also on the type of noun which heads the possessor noun phrase. Four genitive markers are relevant (Payne 1995:290):
(5) un with animate masculine singular proper nouns
uk with inanimate masculine singular common nouns
und with animate masculine singular common nouns
hund with feminine and plural masculine nouns
Examples of the use of these case markers are provided in (6). In (6a), ˆk' is a form of uk, the choice of marker being motivated by the fact that s‡ahr is an inanimate singular count noun. In (6b), we find und since the possessor is an animate masculine singular common noun.8
>(6) a. sFACE="SILDoulosIPA-Regular">‡ahr-FACE="SILDoulosIPA-Regular">ˆk' s‡ur'9 town(OBLII)-GEN=DCT.PL.MASC children=DCT.PL.MASC
'children of the town'
b. m´#l'ˆs und ka#r
father=OBLI GEN=DCT.SG.MASC case
'father's case'
The four markers in (5) can then be said to fill the same semantic function, they are all case markers, particularly, they are case markers which can be said to be semantic in that they have a PRED feature (cf. discussion in Andrews (1995)). This PRED feature can be captured by the simplified representation in (7).
(7) 
The markers differ in one important respect, however, as indicated in (6). Und and hund show the syntactic behaviour of postpositions. The behaviour of un and uk, on the other hand, appears to be governed by morphological rules. The clearest evidence for this distinction is found in noun phrases where the modifying noun phrase consists of a co-ordinated phrase. Und and hund can modify the entire noun phrase, as in (8a). When the head nouns of each conjunct are taken from different categories of nouns, the postposition is selected on the basis of the last one. Hence in (8a), hund is used, since this is the appropriate form for the second noun m´#jˆ 'mother'. Un and uk, on the other hand, behave more like genuine affixes in that they have to be repeated on each conjunct (Payne 1995:290-2).
(8) a. m´#l'ˆs tˆ m´#jˆ hund ka#r
father=OBLI and mother=OBLI GEN=DCT.SG.MASC case
'father and mother's case'
b. s‡ahr-ˆk' tˆ ga#m-ˆk'
town(OBLII)-GEN=DCT.PL.MASC and village(OBLII)-GEN=DCT.PL.MASC
s‡ur'
children=DCT.PL.MASC
'children of the town and the village'
What we have here is then another example of how a morphological formation, namely a noun and -uk or -un, and a syntactic construction, a noun phrase and und or hund, can fulfil the same function. The existence of these constructions thus reinforces the need for a syntactic theory to allow morphological elements and syntactic constructions to be represented, at least partly, in terms of the same theoretical devices. An obvious way to do this within a 'Principles & Parameters' type of framework would be to identify both -uk/-un and und/hund as heads of similar functional projections so that configurationally the two examples (8a) and (8b) would look the same. Indeed, the functional parallels Payne notes between the two constructions would be taken as strong motivation within such a framework that this was indeed the right way to proceed. An alternative would however be to provide equivalent featural characterisations of the two sets of data, something which is easy to do within a unification-based approach like LFG. We would propose the representation in (9a) for example (8a) and (9b) for (8b). (9b) is an f-structure of a familiar kind for a single inflected word form, whereas (9a) contains two separate f-structures for the two syntactically independent constituents of the periphastic, postpositional construction. Assuming an appropriate P-S rule (not stated here) to ensure that the structures in (9a) are unified however, as in (9c), the parallelism between (9c) and (9b) is immediately obvious, and hence the functional equivalence of the two forms in (8) is accounted for.
(9) a. m´#l'ˆs
und 
b. s‡ahrˆk'
c. m´#l'ˆs
und 
The Kashmiri data illustrate how there can be equivalence between
a periphrastic and an inflectional formation. In §3 we also
saw how there could be equivalence between a suppletive formation
and an inflectional one. What we shall see in §5 is the missing
link, namely equivalence between a suppletive and a periphrastic
formation. Given the obvious undesirability of representing suppletions
configurationally, it will then follow that, faced with the choice
between c-structure and f-structure as the locus of the grammatical
equivalences we have identified, there is little doubt but that
we should pick f-structure.
5. Comparative adjectives and adverbs in Latin and Romance
In this section we return to Romance and consider a complex historical
scenario in which periphrasis and suppletion (in the narrow sense)
intertwine, and thus where the modelling of the stages needs free
passage in both directions across the morphology-syntax boundary.
This example therefore combines the lesson of the Romance 'go'
verbs that suppletive and non-suppletive formations must be treated
as equivalent with the lesson of the Kashmiri data that inflectional
and periphrastic forms must likewise find a common mode of representation
in the grammar. We will characterise the developments from Latin
to Romance in a series of stages.
STAGE I: The Latin morphological system for adjectives and adverbs
A Latin base such as alt- 'high, deep' could enter into
a four-way morphological opposition as indicated in the following
table.
| positive | comparative | |
| adjective | altus, -a, -um | altior, -ius |
| adverb | alte | altius |
Establishing f-structures for these forms gets us into two very difficult areas, namely the characterization of the adjective-adverb distinction and the proper treatment of comparison, to neither of which can we do justice here. In what follows, in the interests of a succinct presentation of the general argument, we will simply assume an archi-category called ADCAT with two values, AJT and AVB. We will also assume a feature DEG(ree) with a value COM(parative). This will permit the following f-structures:
(10) a. altus10 
b. altior 
c. alte 
d. altius11 
Now for some adjectives the comparative series is suppletive -
e.g. bonus 'good', which forms the comparative from what
is in origin a different base, viz. mel-, cognate with
multus 'many'. There is also a semi-suppletive, lexicalised
vowel shift in the adverb form, thus:
| positive | comparative | |
| adjective | bonus, -a, -um | melior, -ius |
| adverb | bene | melius |
The f-structures here will correspond exactly to those set up
for altus. In particular, they will all have the PRED feature
value 'bon-' since the distributions of the suppletive and lexicalised
forms are from a morphosyntactic and a semantic point of view
precisely parallel to those for the regular formations alte,
altior and altius. In the change from Latin to Romance
two changes affect the state of affairs we have just described.
STAGE II: Emergence of periphrases
First, the inflectional comparative yields place to a
periphrastic
one in which the basic form of the adjective combines with a degree
word, either plus (whence French plus and Italian
più) or magis (whence Spanish mas
and Rumanian mai).12 Second, the
inherited pattern of forming
adverbs by bound suffixes is gradually replaced by a periphrasis
involving the ablative case form of the noun mens 'mind,
spirit': e.g. sana mente 'lit. with a healthy mind' (Cicero);
caeca mente 'blindly, lit. with a blind mind' (Gregory
of Tours) (for further examples and discussion see Tekavi1980:
Ch 14).
STAGE III: Morphologization of the adverbial periphrasis
With the passage of time the originally independent noun form
mente undergoes the classic stages of grammaticalization
(Hopper & Traugott 1993: ch 1). Its meaning generalises so
that, whereas the Ciceronian (1st cent BCE) sana mente
still means 'with a healthy mind' and not 'healthily', Gregory's
(6th cent CE) caeca mente means 'blindly'. It also passes
from being an independent item to being a bound suffix. This difference
is reflected very clearly in the contrast between Modern Spanish
or Portuguese where a co-ordinated adverb is expressed with only
one occurrence of -mente - e.g. Port. intensa e constantemente
'intensely and constantly' - and Italian or French where
there must be a suffix for each adverb - e.g. Ital. intensamente
e costantemente. Old French and Old Italian by contrast exhibited
the same pattern as the modern Ibero-Romance languages.
Table V sets out the results of these changes, using examples
from Italian.
| positive | comparative | |
| adjective | alto, -a | più alto, -a |
| adverb | alta + mente >
altamente | più altamente |
Following the proposals set out in Börjars, Vincent & Chapman (1996), according to which a periphrastic form has a PRED feature for the main lexical item but not for the grammatical word, and assuming that the value of the whole periphrasis is achieved by unification, we can envisage the following f-structures:
(11) a. più 
alto 
b. alta 
mente 
The two representations in (11a) unify in a straightforward manner to yield an f-structure equivalent to that given above for altior, thus expressing the morphosyntactic equivalence between periphrasis and inflected form. At first sight, however, the two f-structures in (11b) will not unify since they have conflicting values for the feature ADCAT. What is needed here, rather than simple unification is a formal operation such as 'priority union' (Kaplan 1987/1995: 365). This in effect makes the adverb suffix the head of the complex, so that its ADCAT value takes precedence over the ADCAT value of the adjectival base. The feature GEND is required since the input to this morphological formation is the feminine form of the adjective, a synchronic residue of the fact that the originally independent noun mens in Latin was feminine.
Granting the preceding analysis of più alto and altamente, it is a straightforward matter to put them together to generate an f-structure for più altamente 'more highly', namely:
(12) 
This in turn, as desired, corresponds to the structure set up for the inflected Latin form altius.
To sum up, so far we have seen a system for regular formations which has developed from being purely affixal to one involving two independent periphrases replacing the functions of earlier affixes, after which one of the periphrases has in turn re-morphologized. To model this we have sketched a system of f-structures which can express the necessary morphosyntactic equivalences.
While all these changes were taking place in the
regular sub-system,
the remarkable thing about the bonus/melior series is that
nothing changed.13 Hence, what at the outset
was a relation between
a suppletive and non-suppletive series within the bound morphology
becomes a relation between a bound suppletive series and two periphrastic
series, one of which subsequently returns to being bound.14 This
means that whatever stage in the development of a given periphrasis
or grammaticalization has been reached, there will always be at
least one corresponding item in the language where the same f-structure
information will be derived from an autonomous lexical entry.15
The parallel situations patterns are represented in synoptic form
in Table VI:
| I altus/alte/altior/altius | bonus/bene/melior/melius |
| II altus/alta mente
plus altus/plus alta mente | bonus/bene/melior/melius |
| III altus/altamente
plus altus/plus altamente | bonus/bene/melior/melius |
In fact, the development is even more striking due to the cumulative
effects of sound change. Table VII displays the modern Italian
forms that derive from Latin bonus/bene/melior/melius.
| positive | comparative | |
| adjective | buono, -a | migliore |
| adverb | bene | meglio |
While the morphosyntactic changes described above were taking
place, one of the consequences of regular sound change was to
make the form meglio even more morphologically opaque than
it had previously been. Compare the status of Latin melius
'better (adv)' vis-à-vis its Italian reflex meglio.
The base of the former may be suppletive but the
morphological
formation is in other respects regular. Thus, the suffix -ius
is common to all comparative adverbs, and so the word is susceptible
of internal morphological segmentation along the same lines as
alt-ius. By contrast, Italian meglio, while still
realizing the same constellation of morphosyntactic features,
has no internal morphological structure,16
but is simply an unanalyzable
string of phonemes following the entirely regular operation of
sound changes (loss of final [-s], palatalization of the medial
cluster [-lj-] and raising of final unstressed [-u] to [-o-]).
It is at one extreme therefore of morphological
opacity, while
a double periphrastic formation such as plus alta mente
represents maximal syntactic transparency. The morphosyntactic
equivalence of two such different types of formation provides
a strong argument for seeking a shared mode of representation,
one that we would argue only a feature-based system such as LFG
can offer.17
6. Theoretical conclusion
Viewed in this way, suppletion can neatly be fitted into a larger typology of ways in which the expected correlations between MSRs and PRs do not hold, all of which may be regarded as 'suppletions' in the etymological sense that the expected form is 'supplied' from elsewhere (Vincent 1987). The various possibilities are:
(i) syncretisms ('referrals'; cf Zwicky 1985)
(ii) minor or major irregularities of various kinds
(iii) suppletion proper (cf Plank 1996)
(iv) defectiveness (cf Morin 1996)
(v) periphrastic formations
Suppletion in the traditional sense is the lexical limit case, where a form may have no internal structure of any kind to match its morphosyntactic feature composition, cf Italian meglio. Periphrasis can be seen as the syntactic limit case where the requisite form is imported from what is often considered a separate module, namely the syntax (Matthews 1991: 219-221). Finally, defectiveness is the absolute limit case in which there is simply no available mechanism to supply the missing forms.
What this kind of patterning shows is that there needs to be
a single metalanguage in which information can be freely passed
between the lexicon, the morphology and the syntax. Viewed from
this perspective, the argument for features over X-bar configurations,
i.e. for f-structure over c-structure, as the appropriate universal
grammatical metalanguage seems overwhelming. Only in this way
can the commonalities be stated while preserving the independent
differences that inhere in morphology as opposed to syntax, and
in the lexicon as opposed to either. Pace Déchaine
(1996), the non-universality of morphology does not require it
to be handled as a kind of pseudo-syntax. Paradoxically, it is
suppletion - surely one of the least universal, most idiosyncratic
aspects of linguistic structure - that provides a crucial piece
of evidence for undermining the hegemony of configurationality
not only in morphology but also, by parity of reasoning, in syntax.
Authors' address
Department of Linguistics
University of Manchester
Manchester
M13 9PL
UK
nigel.vincent@man.ac.uk
kersti.borjars@man.ac.uk
Footnotes:
1. See for example Huddleston (1984:77-88) for
detailed justification of this analysis. There may be a case for extending
this list, e.g. by distinguishing the base form from what is sometimes
called the subjunctive (as in I demand that he answer my
question) or by distinguishing various uses of what we have here called
the -ing form such as gerund, present participle, etc. Our point is
simply that the list in Table I is the irreducible number of
morphophonologically distinguishable forms. (Return to main
text)
2. Inevitably, in presenting our argument here, we
will not be able to reflect the full historical and dialectal complexity of
the problems raised by Romance 'go'. For recent discussion and references
to the philological literature, see the excellent study by Aski (1995). (Return to main text)
3. To avoid the etymological cruces here, we will refer
to this in what follows as the a-stem. (Return to
main text)
4.Note that the subsequent reduction to monosyllabic
status within Spanish is not relevant to the argument here. This is an
internal development within the history of Spanish and does not impinge on
the proto-Romance stage at which suppletive replacement takes place. (Return to main text)
5. This situation is rather different from the case of
English go/went. There the past of WEND was somehow
transferred to GO, but WEND as an independent lexical item with its own
PRED feature survives to the present day. In Romance, by contrast, the
i-stem, v-stem and a-stem forms of modern ALLER/ANDARE/IR
converge into a single lexical item and cease to exist as separate lexical
entries. (Return to main text)
6. What goes for the future also goes for the
conditional, which is etymologically simply the past of the future. (Return to main text)
7. We refer the reader to Plank (1995) for excellent
descriptions of suffixaufnahme in a wide variety of languages. LFG
accounts of the phenomenon have been provided for instance by Simpson
(1991) and Andrews (1995). (Return to main text)
8. Payne (1995) translates und and hund as
'of' for reasons that will become clear later, but since both are referred
to as case markers and in order not to prejudge the issue, we translate all
four markers in (5) as GEN in these examples. (Return to
main text)
9. Un and uk are traditionally said to
govern an oblique case which we refer to as OBLII. Payne (1995:291-292)
proposes an alternative analysis, but for the purpose of our example here,
we will adopt the traditional view. (Return to main
text)
10. Adjectives but not adverbs agree with nouns in
gender, number and case. For simplicity we will not indicate these feature
values here since they are not germane to the argument we are presenting.
(Return to main text)
11. Readers familiar with Latin will know that the
form altius is also the neuter singular of the comparative
adjective. It is presumably non-coincidental that this form doubles for
both functions, and one way to express this would be to treat altius
as the unmarked, default form in the same way that third singular is the
default verb form for weather and impersonal verbs. This is supported by
the fact that a perfect passive impersonal takes the neuter form of the
past participle: pugnatum est 'it was faught, i.e. there was a
battle'. However, this solution will not generalize to the positive
(non-comparative) where the relevant form is, as we have seen, alte
and not altum. Another solution might therefore be a rule of
referral triggered by the constellation of features [AVB, COM]. We leave
the matter unresolved in what follows and simply treat adverbial
altius as an autonomous form. The nature of our argument is not in
any case affected. (Return to main text)
12. For those whose background is not in Romance, it is
as if English had passed from a stage in which all adjectives formed their
comparative in -er to a stage in which all adjectives - except
suppletive forms like better and worse - expressed the
comparative by means of the more-periphrasis. (Return to main text)
13. One is reminded of Sherlock Holmes' famous
observation in another context that the important thing was that the dog
did not bark! (Return to main text)
14. Again to give a pseudo-English parallel, it is as
if the past of English regular verbs was replaced by the do
construction, so that instead of she walked we said she did
walk, but we still continued to say we went and not *we did
go. (Return to main text)
15. In fact there are some other forms which show
similar developments, though each with additional complexities. Thus, Latin
peior 'worse (adj)' and peius 'worse (adv)' survive in
Italian as peggiore and peggio respectively, though they
would now be treated as the suppletive forms of different bases, vis
cattivo 'bad (adj)' and male 'badly (adv)'. There are also
modern reflexes of maior 'bigger', viz. Ital. maggiore, Fr.
majeur. etc., and of minor 'smaller' and minus
'less'. (Return to main text)
16. We assume that there is no case for identifying the
final -o of meglio as an adverbial suffix; it certainly does
not have the same status as other adverbial items ending in -o -
e.g. certo, poco, troppo, molto, etc. - to all of which there is a
corresponding adjective and for which it is therefore reasonable to see the
adverb as being the default masculine, singular form (cf the argument re
the Latin neuter in footnote 8 above). (Return to main
text)
17 The alternative within a Minimalist-style account
would presumably be to provide meglio with the same kind of functional
suprastructure as the periphrasis, and then to raise meglio through a
number of layers of functional heads in order for it to pick up/check the
relevant morphosyntactic features, a solution for which there is not a
shred of independent evidence. (Return to main
text)
References
Anderson, S.R. 1992. A-morphous morphology. Cambridge: CUP.
Andrews, A. 1990. "Unification and morphological blocking". Natural Language and Linguistic Theory 8, 507-557.
Andrews, A. 1995. "Semantic case-stacking and inside-out unification." Australian Journal of Linguistics 15, 127-181.
Aronoff, M. 1994. Morphology by itself. Cambridge, Mass.: MIT Press.
Aski, J. 1995. "Verbal suppletion: an analysis of Italian, French and Spanish to go". Linguistics 33, 403-432.
Beard, R. 1995. Lexeme-morpheme base morphology. N. Y.: SUNY Press.
Börjars, K., N. Vincent & C. Chapman. 1996. "Paradigms, periphrases and pronominal inflection: a feature-based account". To appear in Yearbook of Morphology 1996.
Carstairs-McCarthy, A. 1994a. "Suppletion". R.E. Asher (ed), Encyclopedia of language and linguistics. Oxford: Pergamon Press, 4410-4411.
Déchaine, R.-M. 1996. "Compositional morphology". Paper presented at The 7th International Morphology Meeting, Vienna, Feb 16-18, 1996.
Halle, M. and A. Marantz. 1993. "Distributed morphology and the pieces of inflection". In K. Hale and S.J. Keyser (eds), The view from building 20: essays in honor of Sylvain Bromberger. Cambridge, Mass: MIT Press, 111-76.
Hockett, C. 1954 "Two models of grammatical description" Word 10: 210-31
Hopper, P. & E. Traugott. 1993. Grammaticalization. Cambridge: CUP.
Huddleston, R. 1984. Introduction to the grammar of English. Camb: CUP.
Kaplan, R. 1987/1995. "Three seductions of computational psycho-linguistics". In M. Dalrymple et al, eds, Formal issues in Lexical-Functional Grammar, Stanford: CSLI, 339-367.
Matthews, P.H. 1972. Inflectional morphology. Cambridge: CUP.
Matthews, P.H. 1991. Morphology. An introduction to the theory of word-structure. 2nd ed. Cambridge: CUP.
Morin, Y. 1996. "The simplicity and acquisition of defective paradigms: a historical perspective". Paper presented at The 7th International Morphology Meeting, Vienna, 16-18 February, 1996.
Payne, J. 1995. "Inflecting postpositions in Indic and Kashmiri." In F. Plank (ed.), pp. 283-298.
Plank, F. (ed.) 1995. Double case: agreement by Suffixaufnahme. Oxford: OUP.
Plank, F. 1996. "Patterns of suppletion in inflection". Paper presented at The 7th International Morphology Meeting, Vienna, 16-18 Feb, 1996.
Pollard, C. and I.A. Sag. 1995. Head-driven Phrase Structure Grammar. Chicago: University of Chicago Press.
Simpson, J. 1991. Warlpiri morpho-syntax: a lexicalist approach. Dordrecht: Kluwer.
Tekavi, P. 1980. Grammatica storica dell'italiano. Vol II: Morfosintassi. Bologna: il Mulino
Vincent, N.B. 1987. "The interaction of periphrasis and inflection: some Romance examples". In M. Harris and P. Ramat (eds), Historical development of auxiliaries. Berlin: Mouton de Gruyter, 237-256.
Vincent, N.B. 1994. "Zero". In R.E. Asher (ed), Encyclopedia of language and linguistics. Oxford etc: Pergamon Press, 5081-5083.
Vincent, N.B. 1995. "Exaptation and grammaticalisation". In H. Andersen, ed. Historical linguistics 1993. Amsterdam: John Benjamins, 433-445.
Williams, E. 1996. "Three models of the morphology-syntax interface". Paper presented at The 7th International Morphology Meeting, Vienna, 16-18 February, 1996.
Zwicky, A. 1985. "How to describe inflection". BLS
11: 371-386.