See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/240756693

The history of English modals; a reanalysis

Article  in  Folia Linguistica Historica · January 1992

DOI: 10.1515/flih.1992.13.1-2.143

CITATIONS READS

17 3,063

1 author:

Ans Van Kemenade

Radboud University
90 PUBLICATIONS   1,207 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

LONGDALE View project

Information structure of historical English View project

All content following this page was uploaded by Ans Van Kemenade on 21 August 2014.

The user has requested enhancement of the downloaded file.

The History of English Modals; A Reanalysis
Ans van Kemenade, Free University, Amsterdam

The development of English modal from main verb in Old English to

auxiliary in early Modern English, has been analysed on the one hand as a
categorial reanalysis from V to Aux, on the other as a grammaticalization
process. It is clear that the historical changes affecting the modal verbs
take place very gradually, which would seem on the surface to argue for a
grammaticalization approach. However, it is unclear what the
grammatical correlates are for the various stages in this
grammaticalization process. It will be argued that the graduality of the
changes is to some extent illusory; modals in Old English are lexical verbs
that select a propositional VP complement, as opposed to IP or CP with
other verbs. A number of syntactic changes, themselves reanalyses, con-
spired to induce a categorial reanalysis from V to Aux. Thus we will see
that the gradual process can be viewed as a series of discrete and indepen-
dently motivated steps.

In his wellknown diatribe against Lightfoot's (1979) version of the modals story, Plank (1984)
argues that the nature and function of modals -expressing the attitude of the speaker towards
what is expressed in the proposition/locution- is such that they are doomed by their very
nature to 'grammaticize' into auxiliaries. On the basis of the history of English modals, this
seems to me untenable. In the gradual process by which modal verbs changed to auxiliaries, a
number of factors and changes were at play that in themselves have nothing to do with
inherent properties of modality. For comparison, modals in Dutch are still main verbs,
whereas at least a number of developments said to have played a role in the auxiliarization of
English modals took place in the history of Dutch as well. It seems that in the history of
English a number of factors 'conspired', some of which did not play a role in Dutch. This is
not attributable to inherent properties of modality. The notion of grammaticization allows one
to characterize the changes of the modals, appropriately, it seems, as a very gradual one.
However, it glosses over a number of discrete steps that can be seen primarily in the syntactic
behaviour of modals. In this article, I will therefore consider primarily their syntactic
behaviour, and in doing so, I will compare the syntax of modal + complement with that of
other verbs + complement. Not surprisingly, the status of bare infinitive and to-infinitive will
play a part in this.

Given recent GB theorizing (Sportiche (1988)), there are three types of propositional
complement: VP, IP, CP. These are indicated under (1) in the order primarily relevant for OE:
(1) CP

specC' C'

C IP

specI' I'

VP I

specV' V'

... V

V to I

V/I to C

Spec V' and spec I' are subject positions: specV' is where the subject is base-generated; specI'
is where under given conditions the subject may move. My assumption for ModE, again
according to recent theorizing, is that modal verbs are not V in any sense, i.e. they do not
select a propositional complement; rather they are base-generated under I as inherently finite
verbal elements. For ModE this can be contrasted with the complementation of causative and
perception verbs, which I assume are main verbs that take a VP complement in the sense of
structure (1), where the subject remains in situ and receives case from the matrix V; other
main verbs in ModE always take a to-infinitival complement, which I take it, reflects either an
IP complement (with believe-type or seem-type verbs; always infinitival) or a CP complement
(finite or infinitival). Thus, the modals are exceptional in that, though they are verbal, have no
selectional properties, which I take to reflect the inability to assign a theta-role.

As far as the modals are concerned, I assume that the difference between ModE and OE is
that whereas ModE modals are INFL elements, OE modals represent V with a propositional
complement VP in the vast majority of cases; but in OE there is already a subset of modals
with auxiliary status. But let us first take a closer look at those aspects of OE clause structure
that are relevant for the finite verb, and the processes affecting it.
1. Old English

The relatively uncontroversial background assumption I make for Old English (OE) is that it
is SOV with fronting of the finite V to second constituent position in root clauses. I give an
illustration of this in (2)

(2) On twam þingum hæfde God þæs mannes sawle gegodod (AHThorpe,I, 20)
in two things had God the man's soul endowed
'with two things God had endowed man's soul'

The basic word order is verb-final, the finite verb appears in second constituent position. In
terms of the structure (1), this is analyzed as movement of the finite V to I, and subsequent
movement of V/I to C. Some other sentential constituent, subject, object, PP is topicalized to
specC'.

Lightfoot (1979) argues that OE modals (premodals as he calls them) are main verbs with
perfectly ordinary complementation. Thus, in the structure (1), they would figure as V, a head
of VP. This stretches reality in a very obvious way: though OE modals appear with direct
object, PP and tensed clause, they are unique among OE verbs in canonically (as canonically
as one can get in OE) selecting a bare infinitive complement. Also their morphology is
defective to some extent even in OE. Plank (1984) is aware of this and makes a categorization
based on a form-function relationship: though they have a number of main verb properties,
OE modals function as auxiliaries, because even when they have a direct object, it is possible
to infer a verbal concept other than the modal itself. Plank (op. cit. 310-11): "Given a
premodal and its object nominal, it used to be possible to infer particular verbal concepts,
which in later English are required to be present on the surface: 'do' or 'have' rather than
'forget' in if it had beene the pleasure of him who may all things (1597 Morley" (and more
examples of this kind AvK). Thus, the object all things would be an object of the inferred
infinitive rather than of the modal verb. However, most of the examples that Plank gives here
do not seem to me to illustrate his point. In the example quoted here from Plank there is no
question of an 'inferred' verbal concept. The modal may should translate in Modern English as
'can do, having the power to do', similar to Dutch archaic 'vermogen', and the meaning of
ModE 'do' is inherent in may, not in some 'inferred concept'. Notice that the notion of inferred
concept is based on the behaviour of Modern English (ModE) modals. In ModE this extra
verb has to be added because the modal in itself is syntactically and perhaps also semantically
unable to express the 'inferred concept'. It is precisely this inability that makes ModE modals
auxiliaries rather than main verbs.

Let us consider in more detail the properties and behaviour of OE modals. In section 1.1 I
will consider what the status of the complement of modals is; in 1.2 and 1.3 we will look at
various aspects of the syntactic behaviour of modals.

1.1 To infinitive vs. bare infinitive

Where they take an infinitival complement, modals are quite unique among OE verbs in
selecting as canonically as one can get in OE a bare infinitive complement.1 However, before
we can regard this as significant, we must see how modals contrast with other verbs in this
respect, and how bare infinitive contrasts grammatically with to-infinitive.

For ModE, it has become standard to analyse the to-infinitive marker as an inflectional
element base-generated under I in the structure (1). Thus, this marker is characteristic of
sentential (IP or CP) complements only. For OE, the standard assumption among traditional
grammarians, recast in EST terms in Lightfoot (1979), is that the infinitive with to + dative
inflection is a Noun, comparable to a gerund. This argument is not very convincing, cf.
Fischer and van der Leek (1981). In particular, all the distributional criteria Lightfoot
mentions for NPhood of to-infinitive, are equally valid for ModE to-infinitives which are
standardly analysed (also by L) as clauses.

Fischer (to appear) claims, following Callaway (1913), that verbs selecting bare infinitives
correspond to the set of verbs governing accusative case; those selecting to-infinitives to those
governing dative case; verbs of double regimen select either. She also suggests that there is a
clear-cut semantic distinction between a bare infinitive and a to-infinitive complement;
whereas a V + bare infinitive reflects one event or two simultaneous events, a V + to
infinitive reflects two events that do not occur simultaneously. While this might be correct for
late Middle English, the main subject of Fischer's paper, it is not tenable for OE: it is very
easy to find examples that argue against Fischer's claims, and I will now discuss a few.

If the correspondence between bare/to and accusative/dative is to hold, one would not expect
to find examples like the following, occurring with some frequency:

1
. Callaway (1913: 80-82) discusses in full the very few counterexamples to this. Convincing examples of to
infinitives occur only with agan, the precursor of ModE ought.
(3)a. Æt ærestum lyst ðone monn unnyt sprecan be oðrum monnum
(CP,38.279.5)
at first pleased the man(A) frivolity speak about other men
'at first it pleased the man to speak frivolously about other men'
b. ... & hine ne lyst his willan wyrcean (HomS 14 (Bl Hom 4))
and him (A) not pleased his will perform
'and it didn't please him to perform his will'
c. ... cwædon, þæt him þæt licode eallum to healdenne
(LawAfel 49.10)
said, that him(D) that pleased all to hold
'said that it pleased him to keep all'

In (3a) we find the verb lystan 'please' with two arguments: an accusative object _one monn
and a bare infinitive complement. Similarly in (3b) an accusative hine and a bare infinitive.
On Fischer's claims, one would expect an infinitive with to here, since the other object is
accusative. (3b) is a different counterexample; the verb lician 'please' has a dative him, and an
accusative or conceivably nominative þæt which anticipates a to-infinitive.

It is easy, too, to find counterexamples to the claim that V + bare infinitive reflects one event,
and V + to-infinitive two events. Consider the following examples of the verb think, the first
with a bare infinitive, the second with a to-infinitive:

(4)a. he ... þohte mid þam bigan ætberstan þam deaðe (ASL,31, 1059)
he thought by the doubling escape the death
'he... thought he might escape death by the doubling'
b. and geðencaðto awendenne eowerne freond (AHThorpe,II,454)
and you think to pervert your friend
'and you think you can pervert your friend'

In both examples, the verb think is used in exactly the same way, as far as I can see. Both
involve planning something that has not yet taken place. There is no temporal difference
between them, so that it is hard to see how the relation between V + to-infinitive is
semantically less direct than that between V + bare infinitive.

On the basis of the examples discussed above, it seems at least very doubtful whether one can
equate an accusative object with a bare infinitive and a dative object with a to-infinitive. I will
now argue that to-infinitives reflect a genuine clausal complement IP or CP; whereas a bare
infinitive represents a propositional VP complement. It is clear that at least some to-infinitives
in OE must be analysed as CP, notably instances of tough-movement constructions and
infinitival relatives:

(5)a. forðon ðe heo is hwiltidum smylte and myrige on to rowenne

(AEHom Thorpe, I, 182, 31)
because that she is sometimes serene and pleasant on to navigate
'because it is sometimes serene and pleasant to navigate on'
b. gif ðær ðonne sie gierd mid to ðreaganne ... (Greg 127,1)
if there then be rod with to beat .....
'if there will be then a rod to beat with ...'

This construction evidences preposition stranding, and preposition stranding in OE only

occurs in constructions with wh-operators of a specific sort, cf. van Kemenade (1987)).
Therefore one must assume a clause with a wh-operator position, i.e. S'/CP. What one may
conclude from this so far is that in OE to is at least compatible with clausal status, though this
does not necessarily mean that OE to can be straightforwardly compared with ModE to as
being an INFL marker.

There is some interesting evidence that to acts as a case marker of the infinitival verb. To see
this, consider the following quantitative data: (extrapolated from Callaway (1913), who has
looked at all the OE texts).

(6) obj cl subj cl compl to A compl to N

U I U I U I U I
84% 16% 30% 70% 3% 97% 2% 98%

These percentages are not entirely clean, but even if we allow for some dirt, they are very
striking: only in the position where the infinitive clause is selected and governed by a verb,
i.e. a case marker, i.e. as object clause, the bare infinitive constitutes the convincing majority
of 84%. (this excludes modals and other auxes) In subject clauses the to-infinitive is predomi-
nant with 70% 2; as complement to N and A it is as canonical as one can get with OE data.

2
. This percentage is greatly increased if one takes a good look at what is listed as 'Subject clauses' by
Callaway. For instance the uninflected infinitive is found in 'subject infinitives' with 'active' verbs like becuman
'happen'; gelystan 'please'; ge_yncan 'seem good; geweor_an 'happen'; and beon + adjective. These are all
The way I read this is that when the infinitive is governed by an objective (perhaps potentially
oblique) case-marker, the infinitive verb is governed and to-insertion is not triggered; when
this is not the case, to varies from strongly predominant (subject clauses) to canonical
(complement to N or A). The high numbers for to in complements to N or A presumably
indicate that the head of VP is ungoverned in a very absolute sense; it cannot be governed by
N or A because these are typically instances of complements with CP status; tough movement
(compl to A) and infinitival relative (complement to N).

Let us assume that verbs, like nouns, need to be case-marked, in the spirit of Fabb (1984). In a
position governed by V (the case of object infinitive clauses in (6)), this can be done directly
by V, apparently regardless of whether the infinitive clause is a direct or an indirect object, cf.
(4). In positions not governed by a case-marker (subject clauses and complements to A and
N), to ranges from predominant to canonical. This leads to the conclusion that to is a case-
marker, presumably a preposition. It is worthwhile to consider what the precise status of this
preposition is and what the relative positioning of this preposition and its complement at D-
structure is. I don't think that these questions can be answered separately. Suppose to is a true
preposition with a VP complement. This is the position adopted by Fischer (to appear):

(7) PP

P VP
to
... V

We would have to adopt a preposition structure, since postpositions in OE are exceptional in

several respects and there is no reason to assume infinitives are among them. To-infinitive
complements predominantly have the order (XP)-(XP)-to-V. To derive these surface orders,
one would have to assume a rule lowering P to a position immediately preceding V. This
would be a very problematic process by any current standards. Therefore it seems reasonable
to treat to as a preposition designated as INFL marker, which assigns case to the infinitival
verb. This amounts to the following structure:

instances of subjectless constructions, where the infinitival complement originates in object position.
(8) IP

spec I'

VP I
to
spec V'

... V

V moves to I (against standard assumptions for nonfinite clauses, but see Sigur_sson (1991)
for motivation on the basis of Modern Icelandic), resulting in a final cluster to-V. Because to
is a preposition assigning dative case, the infinitive V has a dative ending -ne.

I conclude then, that to is an infinitival INFL marker even in OE, and that to-infinitives reflect
IP or CP complements. This leads to the inevitable conclusion that the complement of modals
is a propositional VP complement. I will now consider the various types of modals in OE.

1.2 Three types of modals

In this section, I argue that there are three structures available for modals in OE: first, there
are the 'main verb modals' like volitional will, ability can and may etc., which have their own
subject and correspond to subject control structures; second, there are 'main verb modals' like
hypothetical may, can, must etc. which do not have their own thematic subject and correspond
to subject raising verbs. Third, there is a small group that corresponds to auxiliaries.

1.2.1 Root modals as main verbs

Lightfoot (1979) argues that OE modals (premodals as he calls them) are ordinary lexical
verbs with a full range of complementation and a full inflectional paradigm. This stretches
reality somewhat. Though OE modals undoubtedly have a number of main verb properties
and the range of complementation is fuller, infinitival complements of modals show quite
exceptional behaviour; also the inflectional paradigm of modals is not really complete. The
main verb properties are the following: OE modals have a wider range of verbal inflections
than their modern English counterparts; although examples are scarce, infinitives, present and
past participles are just occasionally found (though not for mot and sceal), and the range of
morphological present/past and indicative/ subjunctive contrasts they have is certainly fuller
than in ModE. Syntactically, they can take a full range of complements: NP object, tensed
clause, infinitival clause:

(9)a. Ne drincð nan man eald win, & wylle sona þæt niwe
(Luke, 1, 319)
not drinks no man old wine, and WILL suddenly the new
'noone drinks old wine and will suddenly want the new'
b. ... ic wolde þæt þa ongeaten þe ... hwelc mildsung siþþan
I WOULD that they perceived who ... what blessing since
wæs, siþþan se cristendom wæs. (Oros, 38, 10)
was, since the christianity was
'I would want those who... to perceive what blessing there has been since the rise of
christianity'
c.... þæt he mehte his feorh generian (Oros, 48,18)
that he COULD his life save
'so that he could save his life'

On the other hand, it seems clear that already in OE, modals were at least to some extent a
separate group. Morphologically they stood isolated in the class of preterite present verbs
whose most conspicuous characteristic is that its verbs lack the usual 3sg. eþ ending.
Syntactically, OE modals form a separate group too. It is certainly the case that in very many
instances modals are followed by some complement NP or clause; in such cases they have
hardly any modal meaning (when in the indicative mood), and where they do, the meaning is
clearly deontic (e.g. sculan in the sense of 'owe'; motan in the sense of 'obligation'; willan in
the sense of 'want'; magan in the sense of 'have the power to'; cunnan in the sense of 'know',
to mention the 'core' cases of modals). When such verbs appear in the subjunctive mood, their
basic meanings acquire more modal 'overtones', as the following examples indicate:

(10)a. Binnan þrimnihtum cunne (subj) ic his mihta (Metr. Charms, 9, 14,)
within three nights CAN I his powers
'may I know his powers within three nights'
b. Butan tweo, gif hie pa blotan mehten(subj) (Oros,115, 14)
except two, if they them sacrifice MIGHT
'except two, if they could (be able to) sacrifice them'
On the assumption that these verbs are main verbs, with deontic meaning, sentences with an
infinitival complement such as (10) represent control structures in the sense of Chomsky
(1981), and have a structure like (11)3

(11) C''

SpecC' C'

C I''
gif
specI' I'
hie
V'' I

V'' V
mehten
specV' V'
PRO
NP V
pa blotan

That is, the modal is a full verb that has its own subject and a propositional VP complement
with a PRO subject that is coreferential with the matrix subject.

1.2.2 Epistemic modals as main verbs

The modal verbs mæg, mot, sceal, can express epistemic modality, as the following examples
show:

3
. This presupposes that it is possible for PRO to be governed and case-marked, against standard
assumptions. See Koster (1984); Sigur_sson (1991) for proposals along these lines.
(12)a. þonne magon hie ryhtor cweþan þæt þæt wæren pa
then MAY they more accurately say that that were the
ungesælgestan (Oros,113,18)
most unhappy ones
'then it would be more accurate for them to say that those were the unhappiest'
b. eallehie þæt anmodlice wilnodan þæt hie his word gehyran
all they that unanimously desired that they his words hear
moston (Bl.Hom., 219,34)
MUST
'they all unanimously desired that they might hear his words'
c. ... æfre woldest ðæt seo wyrd swa hwyrfan sceolde (Bo,10,18)
ever wanted that the chance so turn SHOULD
'... ever wanted that fate should turn like this'

Particularly striking about the epistemic use of modality in these examples, and perhaps of
epistemic modality in general, is that the modal in the matrix clause does not appear to have
its own subject. In the sentences (12), the subject (hie in a; hie in b; seo wyrd in c) is not,
thematically speaking, the subject of the modal, but rather the subject of the infinitive verb.
This can be shown most clearly by the fact that they can cooccur with subjectless impersonal
verbs as main verbs in the infinitival complement, and with passive participles. Some
discussion of these can be found in Denison (1990); Warner (1990). The following examples
(from Denison and Warner) are illustrative:

(13)a. þonne mæg hine scamiganþære brædinge his hlisan (Boeth, 46. 5)
then MAY him shame the spreading of his fame
'then he may well be ashamed of the extent of his fame'
b. be þam mæg þæt apostolice word cweden beon (Bede, 472, 10)
about whom MAY the apostolic word said be
'...concerning whom the apostolic word can be repeated'
c.... þæt we þa þing don þeus to ecere hælu gelimpan
that we those things do that us to eternal salvation happen
moten (HomS, 25, 412)
MUST
'that we do those things that may lead to eternal salvation for us'
d. ða cwæð ic: Hwy ne sceolde me swa ðyncan? (Boeth, 38, 119, 9)
then said I: why not SHOULD me so seem
'then I said: why should it not seem so to me?'
e.Me mæig... gif hit mot gewiderian, mederan settan (Law Ger 12)
one can ... if it MUST weather, madder plant
'one can, in case of fair weather, plant madder'

Denison (op.cit.) and Warner (op. cit.) both analyse the instances of modal + impersonal as
instances where the modal is an auxiliary rather than a main verb; they argue that the fact that
the modal does not have a thematic subject, and the cooccurrence with an impersonal, make
the modals semantically bleached at best, and that therefore these instances of modals are
arguably auxiliaries. 4 In van Kemenade (1989) I accepted this conclusion, but further research
has led me to believe that it is not correct for at least a number of these modal + impersonal
combinations. Though the precise meaning of mot in (13c) and (13e) is rather elusive and
seems 'bleached', the meaning of the modals in the other examples retains a strong element of
the root meaning, except that there is no agent. The same applies to wile 'will' followed by an
impersonal. Both Denison and Warner discuss instances of wile + impersonal, which they
argue involves no volition because there is no subject to have the volition. A priori this seems
sound reasoning, but a number of their examples do seem to me to involve volition on the
part of some impersonal entity. Consider (14):

(14) ic wat, þæt hine wile tweogan, hwæðer... (HomU 21 (Nap 1) 3.6)
I know that him WILL doubt whether
'I know he can't help doubting whether ...'

Denison (op cit) and Warner (op cit) interpret wile here as exclusively futural, resulting in a
translation 'I know he will doubt whether...'. This is not convincing as far as I can see.
Subjectlessness does not preclude volition, as can be seen in Modern Dutch. A sentence
parallel to (14) would be:

(15) er wil voor hem twijfel bestaan of ...

there WILL for him doubt be whether ...

where the meaning of wil involves not the least element of futurity, but clearly impersonal
volition , resulting in a meaning roughly like the one given above for the OE parallel. Many,
though not all, of the examples of will in Denison and Warner translate more naturally if
interpreted in this way. The same goes for passives.

4
. Warner (1992) weakens his original conclusion considerably, noting that cooccurrence of modal +
impersonal may reflect a modal main verb with subjectless syntax.
It is clear that in the above examples there is no thematic subject for the modal verb. Yet it
seems that the instances as discussed (except perhaps mot in (13)) do not qualify for auxiliary
status, as argued by Denison and Warner. Rather, they should be treated as main verbs with
subjectless syntax, i.e. raising verbs. 5 Denison (op cit) makes a distinction between instances
of modal + impersonal with epistemic meaning (corresponding with auxiliary status) and root
meaning with subjectless syntax. He gives a number of examples of each. I can't help
disagreeing with a number of the meaning distinctions apparent from his glosses. For
instance, he cites (16a) (corresponds to (12a)) as an instance of epistemic meaning, and (16b)
as an instance of a subjectless root modal:

(16)a. þonne mæg hine scamiganþære brædinge his hlisan (Boeth, 46. 5)
then MAY him shame the spreading of his fame
'then he may be ashamed of the extent of his fame'
b. Ne mæg nænne mon þæs tweogan þætte ... (Boeth. 75.13)
not MAY no man that doubt that ...
'No man may doubt that...'

However, it seems to me that there is at least an element of permission (i.e. a root meaning) in
the meaning of MAY in i), translating like 'he will do well to feel ashamed...'. On the other
hand, I would sooner translate ii) as 'no man can reasonably doubt that...', i.e. possibility
MAY. This might, for these two examples, result in a classification which is the reverse of
Denison's. But that is not what I want to argue. I would argue that there is no principled
distinction between the two classes, not in meaning and not in the syntax. They are both
instances of subjectless modals, resulting in epistemic meanings. Only they happen to have an
infinitive that is subjectless, too. Thus, the only syntactic difference between (12) and (14) is
that in (12) the infinitive has a thematic subject, whereas this is not the case in (14). Thus (17)
would be the appropriate D-structure for such sentences, illustrated with (12a):

5
. The idea that epistemic modality corresponds with subjectless syntax is a widely held one, but, surprisingly,
there are not many written sources for it. Koster (1984) specifically elaborates this idea in an analysis of modern
Dutch modals.
(17) CP

spec C'

C IP
þaet
spec I'

VP I

spec V'
e
VP V
moston
spec V'
hie
NP V
his gehyran
word

The D-structure subject hie is subsequently raised to spec,IP, where it becomes the syntactic
(nominative) subject of the modal.

1.2.3 Modals as auxiliaries

Some of the modal verbs are used in what seem to be auxiliary constructions even in OE. The
evidence for this for my purposes is thinner than suggested by Denison (op cit) and Warner
(op cit), since it was suggested above that an important piece of evidence, colligations of
modal + impersonal, are not instances of auxiliary usage but of raising verbs. But even apart
from this evidence, wile and sceal can be used in an exclusively futural interpretation. Such
examples can be found in clauses embedded under verbs expressing promise, intention,
expectation, that can only be interpreted as purely futural. I give some of these in (18)
(18)a. ða gebeotode an his ðegna þæt he mid sunde þa ea
then vowed one his soldiers that he with swimming the water
oferfaran wolde mid twam tyncenum (Oros,43, 7)
cross WOULD with two caskets
'then one of his soldiers vowed that he would cross the water swimming with two caskets'
b. & hie him geheton þæt hie þæt gefeoht ærest mid him selfum
and they him promised that they the fight first with themselves
þurhteon wolden (Oros, 47,16)
perform WOULD
'and they promised him that they would first try out the fight among themselves'
c. .. wenen and wilnian ðæt gelange libban scylan her on worulde
(Boethius 46, 31)
think and want that you long live SHALL here on world
'think and wish that you will live long in this world'

It is clear from the contexts in these sentences, as in others, that a futural meaning of the
modal is the only plausible or possible one. In the a) sentence the volition that is potential for
wile is expressed already much more strongly by the matrix verb; similarly in b) voliti-
on/intention is expressed by geheton; in c) the modal cannot be interpreted in any way other
than futural. If the use of shall and will in an exclusively futural sense typically signals
auxiliary status, as is usually assumed for Modern English, this should carry over to Old
English.

The above examples suggest that among the modal attestations there is, already in OE, a
group of auxiliaries, though it is not impossible to analyse these, too, as subjectless main
verbs. Warner (1992) argues on the basis of the occurrence of modals in constructions with
postverbal ellipsis and pseudo-gapping, that it seems likely that OE modals could appear in an
auxiliary configuration, though it is difficult to prove. To quote him: 'it is verisimilitudinous
rather than true'.

So far then, the conclusion is that there is a big group among the modal attestations that has
main verb characteristics; the deontic modals as discussed above, which appear in control
structures with an embedded PRO subject. There is a second group in which the modals act as
main verbs with subjectless syntax. Third, there is a small group in whci the modals appear to
act as auxiliaries. If so, these should presumably be base-generated under I, as in the
following structure
(19) C''

specC' C'

C I''

specI' I'

V'' I
modal Aux
specV' V'

... V

This distinction made here between main verb modals and auxiliary modals, is consonant
with recent ideas on the status of modals. For instance, Scholten (1988) argues that modals
can be INFL elements or V elements by UG definition. This does not necessarily represent an
either/or choice.

1.3 Further syntactic properties in OE

In this section I shall briefly consider the further syntactic properties of modals in OE. This is
a summary of the analysis presented in van Kemenade (1985; 1987; 1989), to which I refer
for more arguments. OE modals evidence a rather free form of V-raising, i.e. clustering of the
modal verb with the embedded infinitival verb. I give some examples in (20):

(20)a. ðæt he Saul ne dorste ofslean (CP,199,2)

that he Saul not DARED murder
'that he did not dare to murder Saul'
b. þæt he þæs gewinnes mehte mare gefremman (Oros,47,14)
that he the victory COULD better achieve
'so that he could achieve the victory all the better'
 c. Cwæð ðæt he gelyfde, ðæt he ðurh ðæt meahte
said that he believed that he through that MIGHT
hælo onfon ðurh Drihtnes gife (Bede, 156, 24)
salvation receive through God's gift
'said that he believed that by that he might be salvaged through God's gift'
d. þæt he mehte his feorh generian (Oros, 48,18)
that he MIGHT his life save
'so that he could save his life'

With (21) as a structure for (20a), I illustrate the process of V-raising:

(21) CP

spec C'

C IP
þæt
spec I'
he
VP I

spec V'

VP V --------> V OR V
mehte
spec V' V V V V'
mehte gefremman mehte
NP V' ADV V
þaes mare gefremman
gewinnes ADV V
mare gefremman

The modal main verb mehte triggers VR. VR is essentially a clause union phenomenon that
destroys the projection of the VP complement. The infinitival verb, or a projection of it,
adjoins to the modal V mehte to form a verbal cluster. Thus in (21) the infinitival verb
gefremman can form a cluster with mehte, resulting in [V V V]; or the V' mare gefremman
can do so, resulting in [V V [V' ADV V]]; or the V' paes gewinnes mare gefremman can do
so, resulting in [V V [V' NP [V' ADV V]]]. The second instance is attested in the example
(20b), but the other options are in principle possible. V-raising applies potentially to all the
main verb modals, not to the auxiliary modals. The main verb and auxiliary patterns can be
kept apart on the basis of word order differences, cf. van Kemenade (1989).

Summarizing, OE modals , unlike other verbs, occur exclusively with a propositional VP

complement. They occur in three syntactic configurations:
- as main verbs with an agent subject, selecting a propositional VP complement that has a
PRO subject coreferential with the matrix subject; this corresponds with deontic meanings;
- as subjectless main verbs that select a propositional VP complement, and subject to subject
raising; this corresponds with epistemic readings;
- as auxiliaries base-generated under I. Here too, they are followed by a propositional VP, but
this is not a complement of the modal. This corresponds with 'auxiliary' meanings such as
futurity.
Main verb modals trigger a rather free form of V-raising.

Let's reconsider at this point Plank's observation that modals are doomed by their very nature
to grammaticize into auxiliaries. Considering the analysis presented above of the three
constructions into which OE modals enter, it is perhaps here that we touch on what Plank
means by 'the very nature of modals'; the same verbs enter into three rather different basic
structures with concomitant meaning differences. But on the surface, the syntactic differences
are not so great and there seems to be a fair amount of fuzziness between the two types of
main verb modals on the one hand, and between epistemic and auxiliary usage on the other.
However, this still does not mean that they are 'doomed' to become auxiliaries. They became
auxiliaries because a rather large number of syntactic changes took place in the course of the
ME period, which touched the modals in a variety of ways. We will now consider some of
these.

2 Middle English

Main verb modals in ME evidence a wider range of main verb characteristics than in OE.
Beside the complementations they already had in OE (object, tensed clauses, infinitival
clauses), they occur with some frequency in infinitive forms and participial forms, and thus
also in each others complementation. Some examples are given in (16)
(22)a. Wultu kastles and kinedomes? (AR,398) (+ object)
WILL-you castles and kingdoms
'do you want castles and kingdoms?'
b. Ichulle þet 3e speken selde (AR,72) (+tensed cl)
I-WILL that you speak should
'I want you to speak'
c.... þatt I shall cunnen cwemenn Godd (Orm,2958) (+inf)
that I SHALL CAN please God
'that I will be able to please God'

The examples in (22) make several things clear. First of all, of course, the main verb status of
the modals in these and similar examples. They take direct object, tensed clause complement,
infinitival complement. Cunnen in (22c) appears in the infinitive. If we want to see, however,
whether there is still V-raising, we run into a problem. In OE, V-raising patterns are, to the
historical linguist, primarily recognisable on the basis of word order; we saw above that in OE
infinitival clause material sometimes 'spreads' on both sides of the finite (main) verb. This is
clear in OE because OE is an SOV language. But in ME, the language has become SVO;
presumably this change from SOV to SVO was completed around 1200. This means that from
that time onwards (nearly) all complement material is on the right of the verb and that such a
check is no longer available. The only cases where it is clearly visible are instances of V-
raising resulting in modal sequences as in (22c); these occur up to the sixteenth century. Other
visible cases are where we find V-raising with a to-infinitive: this is exemplified in (23):

(23) for swa muche pe hokerluker him punche_ to beon ouercumen

(HM,7,22)
for so much the more shameful him seems to be overcome
'for however much the more shameful it seems to him to be overcome'

From examples such as this one can see that V-raising still exists, although it is hard to check
whether modals take part in it because the language has become SVO, so that we do not find
complements of the infinitive to the left of the finite modal. The only clear evidence with
modals is the fact that they occur in clusters as in (22c).

Epistemics of the kind discussed for OE continue to be found in eME. Some examples:
(24)a. He moste conne wel mochel of art þat þou woldest 3if þerof
he MUST can well much of cunning that you would give thereof
ani part (Fl.& Bl.,in Bennett & Smithers,l.259)
any part
'he must be very astute to make you give any part of that'
b. and gif he leng moste liuen alse he mint to don (PC,1137)
and if he long MUST live as he meant to do
'if he might live long, as he intended to'

In the development to and through the Middle English period, there are several changes that
are important for the development of the modals, beside those mentioned by Lightfoot (1979)
and Plank (1984).

The first is the change from SOV to SVO, whereby the distinction (resulting from V-raising)
between the word order patterns of the main verb modals and the auxiliary modals becomes
more difficult, since now they are all SvVO patterns (in embedded clauses at least; the picture
in main clauses is complicated further by Verb Second). As a result, what fuzziness there was
in distinguishing main verb from auxiliary modals, greatly increased.

The second is the gradual erosion of the subjunctive mood, noted among many others by
Plank (1984), as a result of which the modality that was previously expressed by the
subjunctive mood, comes to be expressed by epistemic modals or auxiliary modals, as argued
by van Kemenade (1989). As an illustration, consider the following example:

(25)a. Butan tweo, gif hie þa blotan mehten (Oros,115, 14)

except two, if they them sacrifice MIGHT (subj)
'except two, if they could (be able to) sacrifice them'

It seems clear that in OE this form of 'might' is basically deontic, expressing ability, overlaid
with a hypothetical meaning resulting from the subjunctive mood. Once the subjunctive is
lost, the hypothetical character of this type of modal expression comes to be expressed by an
epistemic modal, and it was argued above that these correspond to subjectless syntax, i.e.
raising verbs. Thus, the loss of the subjunctive mood increased the number of main verb
modals used epistemically.

The third is the gradual emergence of a firm distinction betwee bare infinitive and to-
infinitive, as noted by Lightfoot (1979). It was argued in section 1.1 that in OE to is a
preposition assigning dative case, designated as INFL marker. As a result of the loss of
inherent and morphological case early in the ME period, to ceased to assign dative case to the
infinitival verb. Recall that in OE modals were followed canonically by a bare infinitive
complement. The use of the bare and the to infinitive vacillates for object clauses with all
other verbs, but the bare infinitive is quite predominant. In the course of the ME period we
can see a gradual evolution towards the establishment of to-infinitives with all verbs except
modal, causative and perception verbs. Fischer (to appear) gives a list of figures of the use of
the various types of infinitival complement for the complete works of Chaucer (late
fourteenth century) and the Paston Letters (last three quartes of the fifteenth century). It is
clear from this list that for nearly all the verbs except causatives and perception verbs, the use
of the to-infinitive becomes predominant. This reinforces the distinction between modals and
other verbs.

When at the end of the ME period modals loose a number of morphological properties that
are crucially main verb properties (cf. Lightfoot (1979); Plank (1984)), the auxiliary pattern is
readily available to be adopted by all the modals. But there were two other changes that
facilitated this process, the loss of verb movement, and the establishment of obligatory
nominative subjects.

2.1. The loss of V2

Recall the analysis assumed above of Verb Second in OE. I repeat the structure (1) here as
(26), adapted for eME (SVO)

(26) C''

specC' C'

C I''

specI' I'

I V''

specV' V'

V ...
Verb Second in this structure entails movement of the finite verb to I, and subsequent
movement of V/I to C. Evidence for this was adduced above, and can be found further in van
Kemenade (1987). Although around 1200 the change from SOV to SVO was completed, this
V2 situation remains stable until well into the fourteenth century, as argued by van Kemenade
(op cit). Some examples of this are given in (27)

(27)a. On þis gear wolde þeking Stephne tæcen Rodbert (PC 1140,1)
in this year wanted the king Stephen seize Robert
b. Thus may thine instrument last perpetuel (EP,42)
thus may your instrument last perpetually
c. Ofte schal a womman have thing which... (Gower,CA,I,3206,123)

These examples show that in spite of SVO order, there was still V2 movement to I and C.
This general V2 is lost in the late ME period; from the early part of the fifteenth century,
topicalization no longer cooccurs with fronting of the finite verb, in other words subject-verb-
inversion was lost. In terms of the sentence structure (26) this means that the finite verb still
moves to I, but V/I no longer moves to C. I do not wish to go into the motivation for this
change, for details see van Kemenade (1987).As a result of this change, the first constituent
position in ordinary declaratives comes to be reserved exclusively for the subject. Hulk and
van Kemenade (1988) give an analysis of the relationship between the loss of V2 and the
emerging obligatoriness of nominative subjects.

2.1 Changes in subjectless constructions

The late fourteenth and early fifteenth century is a time when yet more changes took place. In
this period, nominative subjects became obligatory. This is apparent in a number of
subjectless constructions:

The impersonal construction is lost at this stage; by this I mean the construction with a psych
verb accompanied by two arguments, none of which has nominative case:

(28)a. þonne ofþyncþ him (D) þæs ilcan (G) þe he ær forbær OE

(CP 225,19)
then regrets him the same that he before endured
'then he regrets what he endured before'
b. him (obj) wile sone longe þarafter (Trin., 148,19) ME
 him wile soon long after that
he will soon long for that'

After ca. 1450 this construction is almost nonexistent, cf. Elmer (1981). Some of the verbs
featuring in the construction dropped out of the language. Others, such as 'like', came to be
interpreted as having nominative experiencers, cf. Allen (1986).

Zero expletives become ungrammatical at this stage. A well-documented example is the easy-
to-please construction in van der Wurff (1990: ch. 3). Adjectives like 'easy' occur from the
oldest times in three constructions; van der Wurff calls them:

(29)a. the it-type ('it is hard to do this');

b. the zero-type ('is hard to do this');
c.the NP type ('this is hard to do')

The zero-type disappears in the late fourteenth/early fifteenth century.

Butler (1980) reports similar findings for constructions with extraposed clauses: it-insertion
became obligatory in late ME. And in existential sentences there-insertion became obligatory
in late ME.

Let us consider what influence the changes dealt with so far must have had on the develop-
ment of main verb modals used as subjectless verbs: Raising modals are subjectless main
verbs with a propositional VP complement. Due to the reanalysis touching impersonal verbs,
the only other subjectless verbs with an infinitival complement are verbs like semen 'seem'.
However, these verbs take a to-infinitive (IP) almost canonically in late ME. The only other
verbs that select a VP complement are 'subject-ful' verbs: modals used deontically; causative
verbs; perception verbs. Modals by this time appear very frequently in an auxiliary
configuration. A reinterpretation along the lines of that of like is semantically inconceivable. I
propose therefore that the changes in infinitive marking and the emerging obligatoriness of
nominative subjects converged to force a reinterpretation of modals used in raising
configurations as auxiliaries.6

2.3 The loss of V to I

6
. This suggestion was made in Roberts (1985), though not substantiated.
The loss of V2 entails that V/I to C was lost. Roberts (1985) has argued that some time later,
movement of V to I was lost as well. These two developments should be separated by over a
century. The loss of V2 should be dated as beginning around c. 1350 and completed in the
mid-fifteenth century. Kroch (1989) argues on the basis of the relative positioning of finite
main verbs and adverbs that the loss of V to I starts in the second half of the fifteenth century,
and is completed in the second half of the sixteenth century. As a result, lexical verbs are no
longer moved to I. Modals, by this time, are so firmly associated with I that they come to be
designated as I elements.

3 Conclusion

In this article I have tried to present an analysis of OE modals, and of subsequent syntactic
changes in the later history of English, that makes it clear that a large number of clearly
identifiable and independently motivated syntactic changes conspired to force a categorial
analysis of the modals from V to Aux. To analyse this process as one of grammaticization due
to the inherent properties of modals, is to do an injustice to the grammar in which they figure.
Bibliography
Allen, Cynthia L.
1974 "Old English modals" UMass Occasional papers in linguistics 1: 99-106.
Butler, Milton Chadwick
1980 Grammatically motivated subjects in early English (Texas linguistic forum 16).
Callaway, Morgan
1913The infinitive in Anglo-Saxon (Washington: The Carnegie Institution).
Chomsky, Noam
1981Lectures on Government and Binding (Dordrecht: Foris).
1986Barriers (Cambridge Mass.: MIT Press).
Denison, David
1990"Auxiliary + impersonal in OE" Folia Linguistica Historica IX.1: 139-166.
Elmer, Willy
1981Diachronic grammar; the history of Old and Middle English subjectless
constructions (Tübingen: Max Niemeyer Verlag).
Fabb, Nigel
1984"Verbs need case" (Unpublished ms. MIT).
Fischer, Olga
to appear "Factors conditioning infinitive marking in late Middle English".
(Dutch Working Papers in English Language and Linguistics).
Fischer, Olga and Frederike van der Leek
1981"Optional vs. radical radical reanalysis: mechanisms of syntactic change", Lingua
55: 301-359.
Haegeman, Liliane and Henk van Riemsdijk
1986"Verb projection raising, scope, and the typology of rules affecting verbs",
Linguistic Inquiry 17.3: 417-467.
Kemenade, Ans van
1985"Old English infinitival complements and West-Germanic V-raising" Papers from
the fourth international conference on English historical linguistics, edited by
Roger Eaton et al (Amsterdam: Benjamins), 73-85.
1987Syntactic Case and Morphological Case in the History of English (Dordrecht:
Foris).
1989"Syntactic Change and the History of English Modals" Dutch Working Papers in
English Language and Linguistics 26. Abbreviated version to appear in Rissanen
et al (eds.) Proceedings of the fifth international conference on English historical
linguistics (Berlin: Walter de Gruyter).
Koopman, Willem
forthcoming "Old English constructions with three Verbs", Folia Linguistica His
torica.
Koster, Jan
1984"On binding and control", Linguistic Inquiry 15.3: 417-461.
Kroch, Anthony
1991"Reflexes of grammar in patterns of language change", Journal of language
variation and change, 1.3:.
Lightfoot, David W.
1979 Principles of Diachronic Syntax, (Cambridge: CUP).
Plank, Frans
1984"The modals story retold", Studies in Language 8.3: 305-364.
 Roberts, Ian
1985"Agreement parameters and the development of English modal auxiliaries", Natural
Language and Linguistic Theory 3: 21-59.
Scholten, Tineke
1988Auxiliary Verbs and Universal Grammar, (unpublished PhD. dissertation, U. of
Maryland).
Sigurðsson, Halldór
1991 "Icelandic case-marked PRO and the licensing of lexical Arguments".
Natural Language and Linguistic Theory 9.2: 327-365.
Sportiche, Dominique
1988"A theory of floating quantifiers and its corollaries for constituent structure",
Linguistic Inquiry 19.3: 425-451.
Visser, F. Th.
1963-1973 An Historical Syntax of the English Language (Leiden: Brill).
Warner, Anthony
1983"Review article on Lightfoot (1979)", Journal of Linguistics 19: 187-209.
1989"Elliptical and impersonal constructions: evidence for auxiliaries in Old English?",
Evidence for Old English, edited by Fran Colman (=Edinburgh Studies in the
English Language 2) (Edinburgh: John Donald), 178-211.
1990"Reworking the history of English Auxiliaries", Papers from the fifth international
conference on English historical linguistics, edited by S. Adamson et al
(Amsterdam: Benjamins), 537-559.
Weerman, Fred
1989The V2 Conspiracy (Dordrecht: Foris).

View publication stats