Bound pronouns and non-local anaphors:

the case of earlier English[1]

Elly van Gelderen

Languages employ fully referential, somewhat referential and non-referential nominal expressions. They act differently where anaphora is concerned. In Modern English, full Noun Phrases such as the manatee cannot be coreferential with another argument in a sentence; personal pronouns such as me and him are barred from being bound, i.e. referring, to arguments in a local domain; and specially marked forms such as myself are bound in a local domain, i.e. refer to other arguments (cf. Chomsky 1981; Koster 1993; Reinhart & Reuland 1993). The latter elements will be referred to as anaphors or reflexives (no distinction between these is made in this paper). When a referring item is not an argument (e.g. not a direct or prepositional object position), it is referred to as an emphatic (cf. König & Siemund 1997).

In this paper, I will be concerned with cases where pronouns can be used anaphorically in a local domain (i.e. the opposite of what is expected) and I will argue that, when they do, they do not have full referential features (to be made precise later). I will also be concerned with cases where specially marked anaphors occur that are bound to arguments outside the local domain (again unexpected).

Locality is defined through governing category (Chomsky 1981), or the domain of AGR(eement) (Koster 1993), or the domain of a predicate and its arguments (Reinhart & Reuland 1993). This locality captures a generalization across languages (cf. Faltz 1977) that direct objects are more likely to be specially marked, for instance, through -self, than adjuncts or indirect objects. Thus, one expects (1) and (2) to be grammatical but not (3). Similarly, there are varieties of English where (4) is grammatical because the anaphorically used pronoun is an indirect object (see also Baker 1995; Haiman 1995; van der Leek 1994):

(1) I see myself.

(2) I saw a snake near me.

(3) *I saw me.

(4) I 'll buy me a dictionary.

Older versions of English do not display this complementarity. In fact, the reverse is true. For instance, in Gawain and the Green Knight and in Chaucer, (a) pronouns are locally bound in direct object position but (b) forms marked with `self' appear in prepositional object position. Thus, the Middle English domain within which reflexivity is licensed seems very different from the Modern English one. However, rather than doing away with the notion of domain for Middle English and yet to account for (a), I examine the possibility that (Old and) Middle English objects have inherent Case (as in Chomsky 1986). In Old English, objects can be argued to have inherent Case and, in accordance with Reinhart & Reuland's (1993) Chain Condition, pronominal objects are thus not fully specified and can function anaphorically. Once structural Case is introduced in Middle English (perhaps activating AGRoP), first and second person pronouns continue to be used reflexively, again in accordance with the Chain Condition, because they are also less specified in terms of phi-features (i.e. person, number and gender features); third person ones are not. The evidence for the difference in feature strength comes from pro-drop and lack of agreement. With respect to question (b), I argue that `self' marked forms cannot appear in direct object position, i.e. in structurally Case marked position, because the reflexive forms are not fully specified for structural Case due to the change that takes place in `self' from adjective to pronoun. Thus, the pronominal form is genitive rather than accusative. This means the Case features continue to be inherent and the `self'-marked forms only occur in prepositional and indirect object position, i.e. in non-structurally Case marked positions. First and second person forms lag behind here too.

After a brief theoretical discussion, the outline of the paper is chronological. I start by discussing two Old English texts (Beowulf and Junius) where specially marked reflexives do not occur, and two where they start to (Alfred and Aelfric). After turning to early Middle English Layamon's Brut where the morphological change in `self' from adjective to (pro)noun is taking place, I examine two fourteenth century works (Gawain and the Green Knight and Chaucer). Then, the fifteenth century Paston Letters are addressed as well as some subsequent texts (Shakespeare's 2 Henry IV and Hume's Enquiry). The historical texts are examined synchronically rather than diachronically. Thus, I make very few claims as to why one stage would develop into another.

1 Some theoretical background

In this section, I discuss the theoretical background relevant to anaphora as well as to Case.

Chomsky (1981; 1986) formulates three well-known Binding principles: (A) an anaphor must be bound in its governing category, (B) A pronoun must be free in its governing category, and (C) An R-expression must be free. The governing domain for an element is the minimal domain containing the governor, the element itself and a subject. The formulation of A and B assumes that anaphors and pronouns are in complementary distribution. One of the problems with principles A and B and one that has prompted reformulations is the cross-linguistic variation with respect to the famous `snake' sentences. It is well-known that in English, as in (2) above, the pronoun can be coreferential to the subject; in German, as in (5), ihr cannot and the reflexive sich is needed; and in Dutch, as in (6), both are possible (even though some speakers do not accept haar):

(5) Sie sah eine Schlange neben sich/*ihr,

`She saw a snake next to herself/her'.

(6) Zij zag een slang naast zich/haar,

`She saw a snake next to herself/her'.

Other languages display a similar variety (cf. de Jong 1995; 1996 for Romance) which is problematic since the governing category for an element should not be so different for different languages. Chomsky's approach is also problematic for (4).

To remedy this, different types of solutions have been proposed. Reinhart & Reuland (1993) argue that Binding Theory should be formulated as a condition on predicates rather than as a condition on anaphors and pronouns. In (1), the predicate is reflexive-marked (one of its arguments has -self) and therefore two of its arguments must be coindexed. This condition is met. Condition B is stated such that a predicate that has two coindexed arguments must have reflexive marking. Hence, (3) is ungrammatical. If, in (2) and (4), me is not a proper argument to the predicates see and buy respectively, reflexive marking is not needed and the sentences should be grammatical. In addition, they claim there is a Chain Condition that allows pronouns to be locally bound if they are not fully marked for Case and phi-features. Even though they do not mention (4), the Chain Condition might allow locally bound me if one argues that indirect objects do not check structural Case but have inherent Case connected to thematic structure and would not be fully specified. Similarly, one could argue that prepositions as in (2) assign inherent Case and that is the reason the pronoun can be locally bound[2].

As to structural Case assignment, since Pollock (1989) and Chomsky (1989), structural Case is assumed to be checked in a Specifier-Head relationship. Thus, nominative is checked with AGRs and objective is checked with AGRo as in (7):

(7) AGRsP

Spec AGRs'

AGRs AGRoP

Spec AGRo'

AGRo VP

Spec V'

V NP

Zoya saw Bela

Languages differ as to whether or not movement of the subject and object to the Specifier of AGRsP and AGRoP respectively is overt. English is said to have overt movement to the Spec of AGRsP. Inherent Case is assigned under government; it is a lexical Case.

Koster (1993) reformulates the notion of governing category in Minimalist terms (cf. Chomsky 1995) and crucially uses Case checking. He argues that morphologically marked anaphors are strong and must be checked with AGR(eement). Languages differ as to where the feature is located: with AGRs as in German (and Slavic) or with AGRo as in English. Thus, in German, the entire sentence is a binding domain and within it, bound pronouns would violate Principle B; in English, there is a smaller domain and pronouns may function anaphorically if they are not direct arguments of the verb. In (5), the reflexive is in the domain of AGRs, i.e. the entire sentence, and checks its feature; in (2), it is not and a pronoun appears. The anaphor in (1) is in the domain of AGRo and checks its structural Case there. Since inherent (or oblique) Case is not checked in AGR, obliquely marked pronouns can function anaphorically.

Thus, the domain is crucial for both Reinhart & Reuland and for Koster: languages either choose a domain that only includes direct arguments, or the domain is extended to include adjuncts as well. The Chain Condition, however, focusses on pronouns and the conditions where they are referential. I will present some data that are the opposite of the data in (1) to (4) where specially marked forms and simple pronouns are concerned and then use the Chain Condition to deal with the problems that Middle English poses. I start with some unproblematic Old English texts.

2 The data

2.1 Old English

As is well-known, Old English lacks a specially marked reflexive (cf. Penning 1875; Farr 1905; Hermodsson 1952; Ogura 1989). In Beowulf, one of the earliest Old English texts (the manuscript is 10th century but the composition is earlier), simple pronouns function anaphorically as direct, indirect and prepositional objects, as in (8), (9) and (10) respectively:

(8) Beowulf, 677-8[3]

No ic me an herewæsmun hnagran talige

guþgeweorca, þonne Grendel hine

not I me regarding prowess smaller consider

wardeeds than Grendel him

`I think of myself for my prowess and wardeeds no less than Grendel does of himself'.

(9) Beowulf, 932-3

þæt ic ænigra me weana ne wende

that I any-GEN.P me hope not expected

`that I expected any hope for myself'.

(10) Beowulf, 2523-4

forðon ic me on hafu bord ond byrnan

therefore I me on have shield and coat-of-mail

`therefore I shall have on me a shield and a coat of arms'.

There is (even though Visser 1963: 420 and Mitchell 1985: 189-90 deny this) an early form of `self' that marks a reflexive object as in (11). This instance of `self' is assumed to be an emphatic since it is the only instance; the other forms mainly refer to subjects. If hyne sylfne in (11) were an early instance of a reflexive, this occurrence would be in accordance with the observation that if any anaphors are specially marked, they will be the direct object ones. Thus, Beowulf presents no special problems even though the use of `self' is probably emphatic to the reflexive object, due to the rarity of (11). `Self' in these early texts is an adjective inflected for person, number and gender modifying the argument pronoun (cf. König & Siemund 1997 for a discussion on the origin of emphatics). For instance, sylfne in (11) is marked for accusative, masculine singular (indefinite declension) and sylfe in (12) for nominative, masculine plural indefinite (cf. Quirk & Wrenn 1955):

(11) Beowulf, 2875

þæt he hyne sylfne gewræc

that he him-ACC self-ACC.M.S avenged

`that he avenged himself'.

(12) Beowulf, 1995-7

þæt þu ... lete Suð-Dene sylfe geweorðan guðe wið Grendel

that you ... let Danes self-NOM.M.P fight against Grendel

`that you let the Danes themselves fight against Grendel'.

In other Early Old English texts, the same is not true. In the Junius Manuscript (c1000 but composed earlier), `self' does not mark anaphoric direct objects but prepositional objects as in (13). There are 3 possible reflexives, given here:

(13) Genesis 438

Sittan læte ic hine wið me sylfne

remain let I him-ACC with me-ACC self-ACC.M.S

`I let him remain with myself'.

(14) Genesis, 2628

heht hie bringan to him selfum

ordered her-ACC bring to him-DAT self-DAT

`ordered (them) to bring her to himself'.

(15) Genesis 885-6

Nu ic þæs tacen wege

sweotol on me selfum

now I the token carry evident on me-DAT self-DAT

`Now I clearly carry the sign upon me'.

Assuming the forms in (11) to (15) are emphatic, not anaphoric (cf. Faltz 1989), no problems occur. If they were anaphoric, it would be strange that the specially marked form occurs outside the immediate domain in (13).

As mentioned above, Visser (1963: 421) says that by the time of Alfred, the reflexive pronoun is often followed by `self'. The examples he mentions (pp. 421-3) have inflected forms of `self' and modify genitive and accusative objects as in (16) and (17):

(16) Alfred, Pastoral Care 34.7

mon forgit his selfes

man forgets his-GEN self-GEN.S

`Man forgets himself'.

(17) Idem, Orosius 166.23

[he] hiene selfne ofslog

him-ACC self-ACC.M.S killed

`he killed himself'.

In Alfred's Boethius, there are many others, for instance, `self' modifies a prepositional object as in (18). Wülfing (1894: 358), in his two-volume syntax of Alfred's works, lists many with all kinds of endings, e.g. modifying an indirect object in (19):

(18) Boethius 13.13-4

Wastu oþres bi þe selfum to secganne

know-you other by you-DAT self-DAT to say

`Do you yourself know to say anything else'.

(19) Pastoral Care, 4.22

Ac ic þa sona eft me selfum andwyrde

but I then soon after me-DAT self-DAT answered

`But soon after, I soon answered myself'.

The simple pronoun remains used reflexively as in (20). Wülfing (1894: 356) claims that this is the preferred way of expressing the reflexive and provides several pages of instances, not only where the pronoun is a direct object as in (20) but also as an indirect or prepositional object as in (21):

(20) Pastoral Care 409.33

ðu ðin scamige

you you-GEN shame

`Be ashamed of yourself'.

(21) Orosius 154.15

hie namon heora fultum mid him

`they took their support with them'.

Thus, Alfred's reflexive use of pronominals is the same as that in other Old English texts even though more forms of `self' serve as reinforcements of reflexive pronouns (cf. also Wülfing's 1901: 2-18 lists of reflexive verbs).

The late Old English works of Aelfric indicate that `self' is regularly used to reinforce a reflexive pronoun and could be said to be part of it. Looking through the instances in Aelfric's Homilies, sylf is used emphatically with singular nominatives as in (22), sylfe with plural nominatives, sylfes with genitives, sylfne with accusatives as in (23) and sylfum with datives as in (24), occur frequently. Accusative and dative forms are possibly reflexive in (23) and (24):

(22) Hom II 8.173

He sylf clypode to me

he self-NOM said to me

`He himself said to me'.

(23) Hom II 93.51

Þa beðohte he hine sylfne

then bethought he him-ACC self-ACC

`Then he reconsidered'.

(24) Hom II 45.117

Se ðe him sylfum leofað

The that him-DAT self-DAT loves

`Who that loves himself'.

It is interesting that modification of third person pronouns by forms in -ne or -um predominates. Thus, sylfne almost exclusively occurs with hine and sylfum does with him; both modify direct as well as prepositional objects. In both volumes of the Homilies, there are 121 instances of hine followed by sylfne; 9 of me with sylfne; 86 of him and sylfum and 10 of me with sylfum. In comparison, there are 934 instances of hine, 1608 instances of him, and 516 of me. Noticeable first al all is that even though there are almost twice as many instances of him than of hine, the latter is more often modified by a form of `self'. Counting him and hine together and comparing these to me, third person singular pronouns are followed by `self' in 8% of the cases; first person singular is in 3.7%. This third person preference, especially of the accusative form, is related to what is the case in later texts, namely that third person pronouns are the first to develop specially marked reflexives.

In conclusion, the situation in Old English is one where pronouns can be used reflexively. I will argue in section 3 that they can be used this way because they are not fully specified and do not violate the Chain Condition of section 1. The reflexive pronoun, however, is increasingly modified by a form of `self'. This occurs in all contexts by the time of Alfred and Aelfric, i.e. in direct, indirect and prepositional object positions.

2.2 Early Middle English

In this section, I discuss Layamon's Brut, which is from the early half of the 13th century[4]. The points of interest in this text, two versions of which exist, is (a) that `self' is grammaticalizing from an adjective into a noun, (b) that the innovations regarding first and second person pronouns lag behind those of third person pronouns, and (c) that the introduction of a reflexive is in oblique position (unexpected in the framework sketched in section 1).

In the thirteenth century, there is evidence (cf. van Gelderen 1996a) that the category of `self' changes from adjective to (pro)noun. The endings on `self' in both versions (Caligula and Otho) of Layamon's Brut `simplify' and are reanalyzed as Case markers (non-nominative in (30) and (26) below; cf. Diehn 1906: 60). The reason for this may be the general loss of endings on adjectives. In addition, `self' is merged with the pronoun (i.e. written as one word) which is genitive rather than accusative. In the early version of Layamon (beginning of the thirteenth century), there are some adjectival endings such as -ne in (25) and 16 `self' variants preceded by an accusative me occur as in (25); in the later version (second half of the same century), the endings are zero or -e and only 1 form occurs preceded by me:

(25) Caligula 4156

ah hit wes þurh me seolfne

but it was through me-ACC self-ACC

`but it was through myself'.

In Caligula, there are 9 forms of mi-self/mi-seolf, 2 of þi-`self', and 80 of him-`self'[5]. In the somewhat later Otho, the same numbers are 21, 12, and 54. (In Caligula, there is one miseolf, one himseolf and one himsuluen; in Otho, none). Otho is a text that is severely damaged and hence fewer lines are left, but the change from accusative to genitive pronoun is obvious. For instance, (25) becomes (26), (27) becomes (28). Some of the ones that disappear in Otho are (32), (34) and (36), as compared to (31), (33) and (35) in Caligula. Emphatics change as well in pronominal form, as from (29) to (30):

(26) Otho 4156

ac hit was þorh mi-seolue

`but it was through myself'.

(27) Caligula 4165

7 me sulfne heo þencheþ quellen

and me self-ACC they think to kill

`and they plan to kill me'.

(28) Otho 4165

and þench(eþ) mi-seolue cwelle

(29) Caligula 1594

þu seolf wurð al hisund

you self-NOM become all healthy

`you yourself become healthy'.

(30) Otho 1594

þou þi-seolf far hol and (sunde).

(31) Caligula 5466

7 he seolf him wolden specken wið

and he self-NOM him wanted speak with

`and he himself wanted to speak with him'.

(32) Otho 5466

he wolde come and speke him wiþ.

(33) Caligula 6195

7 heo seolf liðden forð

and they self-NOM went forth

`and they themselves slipped away'.

(34) Otho 6195

and hii flowe forþ.

(35) Caligula 10151

and wraðe hine sulfne

and angered him-ACC self-ACC

`and he angered him self'.

(36) Otho 10151

and wreþþede him swiþe[6].

I now examine the shape and function of pronouns followed by `self'. First and second person pronouns followed by `self' usually function emphatically. Third person pronouns continue to be accusative in form (himself rather than hisself) and about half of these forms are reflexive. I start with first person, then proceed to second and third.

In the early, Caligula, version, there are 16 forms with an accusative pronoun followed by `self' as in (37), (38) and (39), but none are reflexive objects. There are 8 forms of mi-seolf as in (40), one of miseolf as in (41), and one of mi-self, i.e. forms where a genitive pronoun precedes. Seven of these are emphatic as in (41), two are reflexive adverbials, shown in (40), and one is a reflexive following a copula in (42). However, Binding Theory around copulas is different. For instance, in Modern English, Binding Theory exhibits exceptions around copulas: He is Hamlet; Let Clinton be Clinton.Thus, mi-`self' is introduced in oblique position:

(37) Caligula 4156

ah hit wes þurh me seolfne

but it was through me-ACC self-ACC

`but it was through myself'.

(38) Idem, 12939

a uolden he me laiden. and lai mid me seoluen

but wanted he me lay and laid with me self-ACC

`but he wanted to lay me and he lay with me'.

(39) Caligula 14012

and þe leo i þan ulode. iwende wið me seolue

and the lion in the water went with me self

`and the lion went into the water, taking me with her'.

(40) Caligula 14004

Buten mi-seolf ich gon atstonden

outside myself I started stand

`I myself stood outside'.

(41) Caligula 8511

miseolf ich habbe inowe

myself I have enough

`I myself have enough'.

(42) Caligula 4397

Ah ich mi-seolf neore

but I myself not-was

`But I wasn't myself'.

Rather than using `self' for reflexives, simple pronouns as in (43) and (44) are used. Their functions are both direct and prepositional object. Looking through the entire text, I found 12 such cases, but this is not an exhaustive list:

(43) Caligula 9500

and ich me wulle ræsten

and I me want rest

`And I want to rest myself'.

(44) Caligula 10967

swa ich here biuoren me. mid æ3enen bihælde

such I here before me with eyes saw

`such as I saw here before me with my own eyes'.

First person plural pronouns are also used reflexively as in (45) and (46); there are 4 instances of a combination with a `self' form, as in (47) (and 2 of these are reflexive) and two with the dual as in (48) (one of which is reflexive):

(45) Caligula 2999

Wrake we us on Bruttes

Revenge we us on Brits

`Let us revenge ourselves on the Brittons'.

(46) Caligula 9176

7 leten we us ræden. of ure misdeden

and let we us council of our misdeeds

`and let us consider our misdeeds'.

(47) Caligula 1656

Vs selve we habbet cokes

us self we have cooks

`Ourselves, we have cooks'.

(48) Caligula 11809

þat fehten wit scullen unc seoluen

that fight we-DUAL shall us-DUAL self

`That we shell fight each other'.

With second persons, simple pronouns continue to be used reflexively; there are only two singular forms preceded by a genitive pronoun as in (49) but both are emphatic. The 10 forms preceded by an accusative as in (50) are mainly emphatic as well, even though some are ambiguous as in (51) and 2 are reflexive as in (52). There are 9 pronouns that I found (using the same method as with first persons) that function reflexively; 2 are given in (53) and (54):

(49) Caligula 8963

þat weore þu Uðer þi-seolf

that was you Uther yourself

`That was you Uther yourself'.

(50) Caligula 14048

and ich æm icumen to þe seoluen

And I am come to you self

`I have come to you'.

(51) Caligula 4907

7 þe seoluen 7 þin folc. falleð to grunde

and you-ACC self-ACC and your people fall to ground

`and you yourself and your people fall to the ground'

(52) Caligula 9915-6

a brutten þe seoluen.

halden la3en rihte

in britain you-ACC self-ACC hold law right

`In Britain, (you) hold yourself to the right law'.

(53) Caligula 8089

Nu þu scalt þe warmen þer

Now you-NOM shall you-ACC warm there

`Now you shall warm yourself'.

(54) Caligula 8596

7 þat weorc þu scalt bringen. mid þe to þissen londe

and that work you-NOM shall bring with you-DAT to this land

`and that work you shall bring with you to this land'.

With second person plural, I found 2 combinations with `self' and one between second dual and `self' but all are emphatic. Thus, with first and second person singular forms, `self' marks reflexivity in possibly 5 cases; whereas simple pronouns do this at least 21 times.

The situation with third person pronouns is different. Simple pronouns continue to be used in both prepositional and direct object positions (I found 4 instances of him and 16 of hine used reflexively in Caligula[7]). As in Beowulf, hine is used as direct object as in (56) and him as prepositional object as in (55). If hine represents the morphologically inherent Case, this fits with Reinhart & Reuland's (1993) Chain Condition: only pronouns not completely specified (in this circumstance for Case) function anaphorically. If him is already the structural Case form, it fits that it does not function anaphorically in the direct domain of the verb. Some instances are:

(55) Caligula 8908

naðeles he hafede mid him

nonetheless he had with him

`nonetheless he brought with him'

(56) Caligula 3302

he hine vncuð makede

he-NOM him-ACC unknown made

`he made himself unknown'

(57) Caligula 2291

he hine bi-ðohte

he-NOM him-ACC thought-about

`he considered'.

In the later Otho version, the numbers are different because the specially marked accusative is disappearing. In this text, 11 instances of him are used reflexively and 8 instances of hine. The total number of hines in Caligula are 682 and in Otho, 430. I will argue in section 3 that the decline of the use of hine is related to the loss of inherent Case.

Unlike with first and second persons, there are more hims (but not hines or his forms[8]) combined with a form of `self' that are used reflexively (mainly as prepositional objects as in (58) and (59) but also as beneficial object as in (60) and (61) and as direct object as in (62)[9]. The number of instances as in (58) to (62), i.e. reflexive forms of him followed by a form of `self', is 30, out of a total of 80 such forms (and 2 fused ones). There are 16 third person plurals and they pattern with the singulars in that half of the forms that are combined with `self' are reflexive:

(58) Caligula 1454

he heo lette nemnen; efter him-seoluan

he-NOM it-ACC let name after himself

`and had it named after himself'.

(59) Caligula 770

Corineus com quecchen. 7 to him-seolfe queð

Corineus came collect and to himself said

`Corineus came collecting spoil and to himself said'.

(60) Caligula 5839

he makede him-seluen muchel clond

he made himself much pain

`He made for himself much pain'.

(61) Caligula 5604

halde him-seolf þisne dom

held himself this doom

`held for himself this authority'.

(62) Caligula 5856

Maximien ... to resten hine seolue

`Maximilian ... to rest himself'.

The third person singular forms are summarized in Table 1 for the two versions. The table shows (a) that the special accusative form, i.e. hine, starts to disappear and becomes the same as the dative, i.e. him, in the later Otho, and (b) that there is an increase of `self' marked reflexives, mainly in prepositional object position (Note that the hyphen is inserted by Brook & Leslie, see footnote 4):

Caligula Otho

refl `him' 4 11

refl hine 16 8

him`self' 2 0

him-`self' 80 (30 refl) 85

hin-seolf 1 0

hine `self' 13 0

TABLE 1: Third person anaphors

So far, I have shown that first and second pronouns continue to be used reflexively in Caligula and Otho. There are only 5 combinations of me or þe combined with `self' that are reflexive, whereas there are 21 reflexively used pronouns. With third person, the figures are 30 with `self' as against 20 `simple' forms. To summarize the positions in which possibly reflexive compounds with `self' occur in Layamon: (a) after prepositions as in (63) to (68), (b) in oblique contexts as in (60) above and (70), (c) as ethical dative as in (69) and (72), (d) the direct objects as in (62) above, repeated here as (71), these are rare:

(63) Caligula 214

he heihte his folc sumunen. 7 cumen to him-seoluen

he ordered his people together and come to himself

`he ordered his people together to come to him'.

(64) Caligula 770 (is (59) above),

Corineus com quecchen. 7 to him-seolfe queð

`Corineus came collecting spoil and to himself said'.

(65) Caligula 977

Ah scupte him nome; æfter him-seluan

but created him name after himself

`But (Brutus) gave him a name after him'.

(66) Caligula 1382

iholden mid himself

held with himself

`(he) held with himself'.

(67) Caligula 1454 (is (58) above)

he heo lette nemnen. efter him-seoluan

`he it let be named after himself'.

(68) Caligula 1470

He seide to himsuluen

he said to himself

`He said to himself'.

(69) Caligula 309

him-self mid his fenge. he to wode ferde

himself with his caught ones he to wood went

`He himself with his hostages went to the woods'.

(70) Caligula 13951

and seide þat he wolde. him-seolue þat lond holde

and said that he wanted himself that land hold

`and seid that he wanted to hold the land for himself'.

(71) Caligula 5856-7

Maximien ... to resten hine seolue

Maximien ... to rest him-ACC self

`Maximilian ... to rest himself'.

(72) Caligula 1102

7 him-seolf ... ferde into ane watere

and himself ... went into once water

`and (he) himself went at once into the water'.

Thus, in Caligula (and Otho), the introduction of special reflexives is most common with prepositional objects (adjunct as well as complement). This can also be seen in the pronouns accompanied by `self' that are introduced in Otho. The cases where Otho has special reflexives where Caligula has simple pronouns are in prepositional object position:

(73) Caligula 1026

hehte heo nemnen Kaerlud. æfter þone kinge

called it name Kaerlud after that king

`(he) called it (the city) Kaerlud after the king (i.e. himself)'.

(74) Otho, idem

hehte nemny hine Kairlud. after him-seolue.

With the morphological change between Caligula and Otho (shown in table 2), one might expect a decrease of the reflexive use of the simple pronoun. This is NOT the case between Caligula and Otho. There are 545 instances of me in Caligula and 12 instances are clearly reflexive. These are direct as well as oblique object. In Otho, the situation does not change, except that the manuscript is damaged and that, as a result, there are only 402 instances of me and 10 reflexive uses:

Caligula Otho

me refl 12: DO + PO 10: DO + PO

me `self' 16: emph >> 1

mi-`self' 9: emph + PO >> 21: emph + PO

mi`self' 1: emph 0

TABLE 2: First person anaphors

The situation in Layamon's Caligula version is perhaps not surprising: if the specially marked form is introduced, it makes pragmatic sense to do so in ambiguous contexts, i.e. third person. One might, however, expect that the introduction would be limited to the direct domain as in (62). This is not the case because it mainly occurs in the oblique (prepositional and indirect object) domain as in (58). In 3.1 below, I argue that inherent Case is lost last for third person pronouns. Since a special accusative third person hine is still frequent in Caligula (and Otho), this might still be an inherent Case and that might be the reason hine seolf occurs less frequently whereas him-seolf is very frequent. In addition, in Caligula, hine is used reflexively many more often (16 times) in contrast to him (4 times). As for Reinhart & Reuland's Chain Condition, one could argue that him is becoming the marker of structural Case and can therefore no longer function anaphorically. As to why first and second person simple pronouns continue to function this way, I develop an account in 3.3.

2.3 Middle English

Fourteenth century texts such as Gawain and the Green Knight use reflexive pronouns and present a challenge to Binding Theory. Checking first person pronouns, seven simple pronouns occur reflexively as in (75) to (81). Out of a total of five `self' compounds, two are used reflexively as in (82) and (83):

(75) Gawain, 402

And I shal ware alle my wyt to wynne me þeder

`And I schall employ al my wit to get myself there'.

(76) Gawain, 474

I may me wel dres

`I may prepare myself well'.

(77) Gawain, 1009

I pyned me parauenture

`I troubled myself perhaps'.

(78) Gawain, 1215

For I 3elde me 3ederly

`Because I surrender myself promptly'.

(79) Gawain, 1964

I 3ef yow me for on of youre3

`I give to you myself for one of yours'.

(80) Gawain, 2121

And I schal hy3 me hom a3ayn

`And I shall hasten myself home again'.

(81) Gawain, 2159

And to hym I haf me tone

`And to him I have committed myself'.

(82) Gawain, 1540

Bot to take þe toruayle to myself to trwluf expoun

But to take the hard-task to myself to expound true-love

`But to take on the task of interpreting true love'.

(83) Gawain, 2434

When I ride in renoun, remorde to myseluen

`When I ride in glory, (I) call to mind with remorse to myself'.

Thus, the simple first person pronoun occurs in object position, except for (80) which is different because it is an ethical dative (cf. Mustanoja 1960), never as prepositional object. Myself/uen only occurs as object of a preposition. This means the simple pronoun is still the reflexive but that the introduction of the `self'-compound occurs in oblique (i.e. non-direct object) position. These forms are not emphatic since emphatics are no longer formed by adding `self' but as in (84). If (83) had been emphatic, it might have looked like (85). Such sentences are unattested as `%' indicates:

(84) Gawain, 1052

For I am sumned myselfe to sech to a place

`For I myself am summoned to seek a place'.

(85) %remorde to me myselfue.

There are no first person plural `self' forms and (only) 19 instances of we occur of which none is followed by a reflexive us.

The situation is similar with second person singular. The only reflexive form marked with `self' is given as (86) and is in prepositional object position. The simple pronoun is used reflexively in direct object position in (87):

(86) Gawain 2141

þat þou wylt þyn awen nye nyme to þyseluen

that you want your own harm bring upon yourself

`that you want to take all your trouble on yourself'.

(87) Gawain 2341

halde þe wel payed

`hold yourself well paid'.

The same is true for second person plural, except that the reflexive form in (89) is used without referring to an NP in the same clause. This is possible in impersonal constructions:

(88) Gawain 1267

Hit is þe worchip of yourself

(89) Gawain 1964

if yowreself lykez

`If you would like'.

(90) Gawain 1394

Where 3e wan þis ilk wele bi wytte of yorseluen

where you won this kind wealth by intelligence of yourself

`where you acquired wealth of this kind through your wisdom'.

(91) Gawain 1547-8

As I am hy3ly bihalden, and euermore wylle

Be seruaunt to yourseluen, so saue me dry3tyn!

`For I am highly beholden and evermore shall be the servant of you, so save me God'.

Third person `self' forms as in (92) are different in that even though more hymselfs are prepositional objects than direct objects, there are 3 direct objects out of 24 `self' forms as in (93). There is also an impersonal subject as in (94); many are non-anaphoric as in (95). The simple pronoun remains in some use:

(92) Gawain 1198

Bot 3et he sayde in hymself

`But still he said to himself'.

(93) Gawain 2040

Bot for to sauen hymself

`But to save himself'.

(94) Gawain 976

To be her seruaunt sothly, if hemself lyked

`To be their faithful servant, if it would be pleasing to them'.

(95) Gawain 1085

Þer watz seme solace by hemself stille.

`There was fair pleasure by themselves privately'.

Concluding, the data in Gawain and the Green Knight indicate two problems. (a) The domain in which specially marked anaphoric forms appear is not within the immediate domain of the verb. This presents problems for Reinhart & Reuland's conditions on predicates, as well as for Chomsky's notion of governing category and for Koster's AGR-domain. (b) There is a difference between first and second person anaphors on the one hand and third person ones on the other: third person reflexives such as hymself are used as direct objects.

I now turn to Chaucer and will show that the data are very similar to Gawain for first and third but not for second person anaphors. With first person singular reflexives, there are more simple pronouns than specially marked ones: 71 forms of `myself' as in (96) to (98) (including emphatics), but at least 125 reflexive me as in (99) to (105):

(96) Knight's Tale 1813

I woot it by myself ful yore agon

`I knew it by myself long ago'.

(97) Pardoner's Tale 841

this tresor to myself allone

`this treasure to myself alone'.

(98) Boece Bk 1 P4, 105

I ne reservede nevere nothyng to myselve

`I never reserved anything for myself'.

(99) Clerk's Tale 145

I me rejoysed of my liberte

(100) Knight's Tale 2052

I wol me haste

(101) Wife of Bath's Tale 1231

I put me in your wise ...

(102) Melibee 1058

if I governed me by thy conseil

(103) Romaunt of the Rose 1807

thanne I avysede me

(104) Rom 6297

If I may passen me herby.

(105) Troilus and Criseyde, II, 12

Forwhi to every lovere I me excuse

The simple pronoun me is only used in direct object position (except in the expression sayde for me); myself is mainly used in oblique position[10]. The same distribution occurs in the case of us as in (106) and (107) of which around 20 cases occur. Ourself and us selven are used reflexively in oblique contexts in (108) and (109):

(106) Melibee, 1765

we putten us and oure ...

(107) Melibee, 1821

we submytten us to the ...

(108) Bo Bk 3 P12

ben asschamid of ourself

(109) Prol WBT, 812

acorded by us selven two

Second person `self'-forms are different. Many are emphatic as in (110); a few are in subject position by themselves as in (111); many are direct objects as in (112) to (114) and many are objects to prepositions as in (115). The same seems true for the second person plural:

(110) Boethius Bk 3, P4

and thou thiself hast ysought it mochel

(111) Troilus 369, Bk 3

so loth was that thiself it wiste.

(112) Merchant's Tale 1385

Thou lovest thyself.

(113) Troilus 528 Bk 4

Why nylt thiselven helpen don redresse

(114) Boethius Bk 4 P4

thow hast joyned thiself to the most excellent

(115) Troilus 620 Bk 4

Have mercy on thiself for any awe

The second person simple pronoun thee is used reflexively but as mentioned above much less frequently than with first persons (27 times as opposed to 125 times with first person). A number of simple pronouns function anaphorically, mainly as direct objects. The vast majority of these are reflexive verbs such as repent, shryve `confess', bithink `reflect'. When discussing the Paston Letters below, I come back to these.

Third person reflexives pattern with second person ones. Even though the majority of the reflexive `self' forms occurs as object of a preposition as in (116), there are quite a lot of direct objects as in (117):

(116) The Knight's Tale 1773

And softe unto hymself he seyde

(117) The Parson's Tale 1042

and helpen hymself the ofter with the orisoun

Again, the instances of him used reflexively occur with reflexive verbs such as shryve.

Summarizing Chaucer, the first person simple pronoun is used reflexively in direct object position. Outside that immediate domain, a special indicator, i.e. a `self'-marked form, is needed. This is not true for second and third person where even though specially marked forms predominate in prepositional object position, some occur as direct objects. The use of second and third person simple pronouns is reduced to reflexive verbs.

2.4 Later Developments

In The Paston Letters (PL), written by various people throughout the 15th century, the simple pronoun ceases to be used reflexively except with what one could call inherently reflexive verbs such as repent. These verbs cannot be other than reflexive and therefore a specially marked reflexive is least necessary. They occur throughout the history of English as (99) above shows. Reinhart & Reuland (1993: 663) assume that a predicate is reflexive-marked if the predicate is lexically reflexive. Hence, a simple pronoun can be coindexed with the subject without having an ill-formed predicate. The Chain Condition could account for it as well if one argued that the Case assigned by reflexive verbs is inherent and does not fully specify the pronoun. Checking over a hundred instances of hym in the immediate environment of he, I find 6 reflexive hyms, namely (118) to (122):

(118) PL, #310 (1478)

he repentyd hym

`He repented'.

(119) PL, #129 (1448)

he xuld repent hym

`He should repent'.

(120) PL, #143 (1452)

he shall repente hym

(121) PL, #143 (1452)

for he shall ell repent hym

(122) PL, #165 (1461)

he schold bryng wyth hym

The several hundreds of instances of I in the immediate environment of me involve almost exclusively the verb recommand.

Reflexives such as hym-self are used in all possible environments, for instance in (123) and (124):

(123) PL #14 (1445)

ho so euer schuld dwelle at Paston shulde have nede to conne defende hymselfe

`whosoever should dwell at Paston should be able to defend himself'.

(124) PL #116 (1461)

fore he is not bold y-now to put forthe hym-selfe

`because he is not bold enough to put forth himself'.

Thus, the situation in the Paston Letters is close to that in Modern English and can be accounted for the same way.

In a later text, however, the First Folio Edition (1623) of Shakespeare's 2 King Henry IV, simple pronouns as in (125) to (128) as well as specially marked ones in (129) to (132) function anaphorically in both direct and indirect domains:

(125) 2 Henry IV, I, iii

He that buckles him in my belt

(126) II, ii

(sayes he) that takes vpon him not to conceiue?

(127) II, iv

I feele me much to blame.

(128) IV, i

I take not on me here as a Physician

(129) I, iv

That thou prouok'st thy selfe to cast him vp.

(130) II, iv

I dresse my selfe handsome

(131) V, i

An honest man sir, is able to speake for himselfe

(132) II, ii

as hee hath occasion to name himselfe.

This situation is very different from the one in Gawain and Chaucer and will not be accounted for here. Notice, however, that my and thy are separate from the forms in `self' whereas himselfe is not. There are 28 instances of independent selfe as in (129) and (130) and all of these involve first, second or neuter pronouns. The 4 forms of independent selues also involve first or second. All instances of third person are a unit as in (131) and (132). Checking the entire First Folio Edition, I find only 2 forms of myselfe but hundreds of my selfe; with second person and third neuter and feminine, all instances are morphologically separate. The total numbers of self is 14; of selfe 1405; of selfes 1. With third person masculine, however, there are 4 instances of himself, 417 of himselfe, 7 of him-selfe, but there is never an instance of him selfe. This indicates that even in Shakespeare, where simple pronouns are still used reflexively, the third person (masculine) ones are more grammaticalized.

A century and a half after Shakespeare, there is no trace of the reflexive use of simple pronouns in the texts I examine and simple pronouns as in (133) are free. `Self'-forms as in (134) are used exclusively in, for instance, Hume's Enquiry Concerning Human Understanding (1748). Typical instances are:

(133) Hume, Enquiry IV, i, 23

Adam ... could not have inferred from the fluidity and transparancy of water that it would suffocate him, or from the light and warmth of fire that it would consume him.

(134) Hume, Enquiry XII, ii, 128

And though a Pyrrhonian may throw himself or others into a momentary amazement ...

Checking some Austen texts a century later, the `modern' English situation prevails and simple pronouns are not used reflexively.

Concluding section 2, I have indicated several problems for Binding Theory: the introduction of specially marked reflexives is in prepositional object position, and there is a difference between the different pronouns.

3 Middle English Binding

Using Reinhart & Reuland's (1993) Chain Condition, I will entertain a number of ways to account for the Middle English data, through (a) inherent Case; and (b) feature content of pronouns, and (c) the change of `self' from adjective to noun. The Chain Condition allows pronouns to be used anaphorically (i.e. they are not referential) if they are not fully marked for structural Case and/or phi-features. I will argue that inherent Case is lost first with first and second person pronouns and later with third person. This renders pronouns with inherent Case anaphoric. The phi-features of first and second are not fully marked as opposed to third. In Old English, inherent Case makes a pronoun less referential and enables it to function anaphorically; once inherent Case is lost for third person, the pronoun becomes referential and ceases to be reflexive. First and second person pronouns continue to do so longer due to the less fully marked nature of their phi-features. I am not focussing on putting referentiality in Minimalist (Chomsky 1995) terms but this can readily be done by arguing that Inherent Case and less marked phi-features are Interpretable and not relevant at the level of chains. They are relevant at LF. As the language changes from one with morphological licensing to one with positional licensing (see Kiparsky 1997), the features become Non-Interpretable.

3.1 Case

As shown above, Old English has no specially marked reflexive. However, using the insights the Chain Condition, as in Reinhart & Reuland (1993), or Koster (1993) provide us into inherent Case, this is not surprising. If inherent Case makes a pronoun less referential, it can function anaphorically[11].

In Old English, as is argued in e.g. van Gelderen (1996b) and others, the Case of the object is not structural as many Cases are distinguished morphologically: the first person nominative, genitive, dative and accusative forms are ic, min, me, me(c); the third person forms are he, his, him and hine. During the Old English period, the special accusative forms for first and second person disappear (e.g. mec `me-ACC' and þec `you-ACC-SG' are no longer present in Late Old and Early Middle English), but third person special forms (e.g. hine `him-ACC') remain in use in texts of the middle of the 13th century.

Many of these Cases are thematically predictable. For instance, in Beowulf, some verbs as forgripan `seize' in (135) and misbeodan `do wrong' have dative objects (cf. Visser 1963: 280ff; Mitchell 1985: 454ff.) because their objects have Goal theta-roles; others such as seon `see' have accusative because the theta-role is Theme; or genitive such as feon `rejoice in' because of the Cause theta-role. This thematic relationship indicates that Case is inherent, rather than structural (cf. Chomsky 1986):

(135) Beowulf 2353

ond æt guðe forgrap Grendeles mægum

and at battle seized Grendel-GEN kinsmen-DAT

`and he crushed Grendel's kinsmen to death in battle'.

Another piece of evidence for inherent Case is that Old English has passives as in (136)[12]. In Beowulf, instances of (136) to (142) can be found where the passivized object him retains its Case rather than getting nominative Case (all the instances found after checking the 200 hims). I take these to be evidence of inherent Case:

(136) Beowulf 140

ða him gebeacnod wæs

then him indicated was

`when he was shown by means of a sign'.

(137) Beowulf 1192

Him wæs ful boren

him was cup given

`He was given a cup'.

(138) Beowulf 1269

þær him aglæca ætgræpe wearð

there him monster grabbed became

`here he was grabbed by the monster'.

(139) Beowulf 1330

Wearð him on Heorote to handbanan

became him in Heorot to slayer

`He was slain in Heorot'.

(140) Beowulf 1356-7

hwæþer him ænig wæs ær acenned

dyrna gasta

whether him any was before born bad spirit-GEN

`whether to him before this a bad [ghost] had been born'.

(141) Beowulf 2682

Him þæt gifeðe ne wæs (þæt...)

him that given not was

`It was not given to him'.

(142) Beowulf 2696

swa him gecynde wæs

so him taught was

`as he was taught'.

It has been argued that Old English lacks transformational passives (cf. Lightfoot 1979 and others, but see Lightfoot 1991 for arguments against this position). Instances where the `object' has nominative Case are argued to be adjectival. Looking at the 284 instances of he in Beowulf, there are 3 such instances (over half fewer than the 7 passives with him as in (136)), listed as (143) to (145). A fourth instance must be disregarded as the parentheses in (146) indicate "conjecturally inserted letters" (Klaeber in his note to the 1922 edition).

(143) Beowulf 693

þær he afeded wæs

there he grown-up was

`where he had grown up'.

(144) Beowulf 1539

þa he gebolgen wæs

then he angred was

`then he became angry'.

(145) Beowulf 2692-3

He geblodegod wearð

sawuldriore

He bloodpoured became lifeblood-DAT

`He became stained with blood'.

(146) Beowulf 723

ða (he ge)bolgen wæs

then he angred was

`then he was angry'.

Not being able to apply tests (cf. Wasow 1977) to native speakers of Old English, e.g. whether the past participle appears after raising verbs and whether it can have an un- prefix, makes it hard to decide. However, looking at the meaning of (136) to (142) versus (143) to (145), one gets a sense that in the former (except for (140)), an agent is involved or overtly mentioned but not in the latter. For instance an agent cannot `grow someone up' in (143). Reading the text around (144), it becomes clear that the anger is not caused by an external agent but by `himself'. Thus, I assume that (136) to (142) are true `transformational' passives, rather than (143) to (145), where an underlying object becomes the subject but where the Case remains the original objective Case.

A third piece of evidence that structural Case marking does not occur is the absence of constructions where a verb of the main clause `assigns' Case to the subject of the embedded clause, e.g. in Accusative-with-Infinitive (hence ACI) constructions. ACIs are constructions where theta-marking is not connected to Case. Thus, a `subject' can get accusative Case from a verb not related to it in theta-marking. ACI-constructions occur much less frequently in Old English and with other verbs than they do in Modern English. As has often been noticed, e.g. Callaway (1913) and Zeitlin (1908), ACI-constructions in Old English occur with verbs of command (e.g. hatan), permission, sense perception and causation and it might be possible to analyze such sentences differently e.g. as double object constructions, like persuade in Modern English:

(147) Alfred Pastoral Care 451, 8

he us het ðæt we hit beforan monnum dyden

he us commanded that we it before men did

`he commanded us that we it in the presence of men did'.

(Visser 834)

Hatan `command' in (147) can be argued not to have an ACI and hence, presents no evidence for the existence of structural Case. Let can be regarded the same way even though I have not found an instance of let with a NP-CP complement in Visser. `Real' ACI-constructions, e.g. with believe and want, start to appear in the late fourteenth century, e.g. in Wyclif and Chaucer. In an early Middle English text such as Layamon, they do not yet occur (cf. Funke 1907: 25-6). Thus, the fact that they do not occur is accounted for if Case dependent on a structural rather than a thematic relationship does not (yet) occur.

If Old English objects have inherent Case and if me in (8) above has inherent Case, it can form a chain with ic because me is not fully specified. Prepositional objects as in (9) and (10) can also be argued to receive inherent Case. For instance, as in Modern German, certain Cases go with certain prepositions, dependent on their meaning. The early Middle English Caligula version of Layamon's Brut retains inherent Case and simple pronouns continue to be used anaphorically.

The situation in Middle English is different. If Case in Middle English were inherent as well, i.e. tied to theta-marking rather than to structural position, the pronoun could form a Chain with the antecedent without violating Reinhart & Reuland's (1993) Chain Condition; it would also not need to be marked in Koster (1993) because it would not check its (inherent) Case in Spec AGRo. The reason `self' would be necessary in prepositional object position might be that prepositions cease to assign inherent Case. However, the existence of inherent Case in Middle English cannot be demonstrated. On the contrary, in van Kemenade (1987), it is argued that inherent Case is lost in Early Middle English. In van Gelderen (1993: 171ff; 1996b), the date is put around 1250, based on morphological and thematic Case marking being lost. Thus, in the mid-fourteenth century Gawain, the direct object is not assigned morphologically inherent Case. If this is true, Case to the object is checked in AGRo (as in Kayne 1989 or under government by the verb) and a pronoun should not be able to function anaphorically. In addition, as mentioned above, even though the third person is the last to lose inherent Case, it is the first to develop special reflexives. Thus, Case cannot be directly responsible for the changes with the reflexive.

3.2 Underspecified Phi-Features

I will argue (but see Collins & Thráinsson 1996: 423 for a different view) that first and second persons have underspecified or less fully marked phi-features (e.g. for number and gender) and can therefore continue to function anaphorically even though they lose inherent Case; third person pronouns, on the other hand, have fully marked phi-features and when inherent Case disappears, they cease to function anaphorically. The difference between Gawain and Chaucer shows that the features of second person become specified before those of first person. This idea is `translatable' in different frameworks. Thus, Givon (1983) argues for a difference between null-subjects, pronouns, full NPs in terms of topic shift possibilities. Arguing that the person features are weaker or less specified means that they would serve less as topic shifters than third person pronouns. This is borne out by the pro-drop data below, namely third person pronouns are dropped but not first and second in Old English. The weakness or unspecified nature may also show up in prosodic features. Here Dutch is a good instance where phonologically reduced elements function anaphorically, e.g. in (148) and (149). Phonologically reduced third person never loses enough to become weak as (150) shows. In (148) and (149), me and je can be coindexed (or form chains) with their antecedents, but third person `m in (150) cannot:

(148) Ik waste me

I washed me

(149) Jij waste je

You washed you

(150) *Hij waste `m (hij and `m coindexed)

He washed him

The pronouns me, je and `m are morphologically weak and their strong counterparts are mij, jou and hem. The latter forms are used as regular pronouns but when as in (148) and (149) they are used reflexively, they become ungrammatical (and (150) remains so):

(151) *Ik waste mij

I washed me

(152) *Jij waste jou

You washed you

(153) *Hij waste hem (hij and hem coindexed)

He washed him

Thus, Dutch has a set of pronouns weak in phi-features, as well as one strong in those features and this distinction may have phonological ramifications. Everaert (1986: 206) formulates this observation in terms of phonologically unmarked. Reuland (1997) argues that me and je have no number specification and are therefore non-referential. He (p.c.) notices that (152) is worse than (151) and that this indicates that the phi-features rather than the phonological strength is relevant. It is interesting that the grammatical form used for third person in Dutch constructions such as (153) is zich and that this form in Yiddish, i.e. zikh, is generalized across the paradigm, i.e. used for first, second and third. I assume the reason is that, in Yiddish, it is unspecified for phi-features. Taraldsen (1996: 201) claims that sig/seg in Icelandic and Faroese is unspecified for number. What these forms show is that anaphors are typically less specified than pronouns; and that reflexively used pronouns lack certain features as well. Burzio (1991: 87) puts it in the following terms: "An NP with no features is an Anaphor". He makes it clear (p. 96) that what is meant by `no features' is referential rather than morphological underspecification.

As mentioned, one could also argue that there is a split between first/second and third in terms of Interpretable and Non-Interpretable features respectively (cf. Chomsky 1995): the former need not check their phi-features whereas the latter do. Thus, languages and stages vary as to which features are Interpretable. I will not elaborate on Interpretable features here. It is sufficient to focus on phi-features and Case.

Some additional evidence for the claim about the strength or specifiedness of pronouns comes from referential pro-drop data. Following a rich tradition (e.g. Rizzi 1982; Jaeggli & Safir 1989), Ura (1994) argues that pro is licensed by Case or phi-features. Iatridou & Embick (1997) argue the same; pro is specified for person and number and cannot refer to an element not specified for these. If, as I argue above, third person is in fact more specified than first and second, one would expect pro-drop with third but not with first and second. This is in fact the case in older varieties of English. Here I rely more on secondary literature and on looking through Old English texts than on absolute figures.

Berndt (1956) argues that pro-drop in late Old English occurs more with third person than with first and second (as opposed to Modern English). In his examination of late Old English verbal inflection, Berndt also tabulates the increased use of personal pronouns. His tables indicate a clear first/second versus third person split. For instance, in the early 10th century Durham Ritual, which shows fewer pronouns than the other texts examined, 87% of the first person singular pronouns appear; 78% second person singular; 7% of the third singular; 98% of first person plural; 93% of second person plural and 17% of third person plural. Berndt divides The Lindisfarne Gospels and The Rushworth Glosses in two parts each because one part of the latter is from a different dialect area than the other. The figures for indicative constructions for the presence of first person singular are 96%, 99%, 97% and 96%; for second person singular 87%, 93%, 88%, 90%; for third singular 21%, 15%, 54%, 16%; for first plural 100%, 99%, 98%, 98%; for second plural 95%, 95%, 89%, 83%; third plural 29%, 20%, 52%, 19%. Intuitively, the same seems true in Beowulf as well. For instance, in the first 20 lines, there are 5 third person instances of pro-drop but none with first person. Representative instances of both are (154) and (155)/ (156) respectively:

(154) Beowulf 7-11

He þæs frofre gebad

weox under wolcnum weorðmyndum þah

oð þæt him æghwylc þara ymbsittendra

ofer hronrade hyran scolde

gomban gyldan.

He that-GEN consolation-GEN waited grew under clouds-DAT honor-DAT accepted/grew until him every those-GEN people-GEN around-GEN across sea obey should tribute pay

`He was consoled for that; grew up; his honor grew until everyone of the neighboring people on the other side of the sea had to obey him; had to pay tribute'.

(155) Beowulf 292-3

Ic eow wisige

swylce ic maguþegnas mine hate ...

I you will-lead as I men my command (to ...).

`I will lead you as I command my men'

(156) Beowulf 335-8

Ic eom Hroðgares

ar one ombiht. Ne seah ic elþeodige

þus manige men midiglicran

Wen ic þæt ge ...

I am Hrothgar-GEN messenger and officer. Never saw I foreign-warriors so many men more courageous. hope I that you ...

`I am Hrothgar's messenger and officer. I never saw in foreign warriors so many courageous men. I hope that you ...'.

Thus, pro-drop in Old English provides evidence that third person features are more specified and that the pronoun can therefore appear as pro[13]. More work is needed, however, on pro-drop in later stages.

If phi-features of first and second person pronouns are less specified than of third person ones, there might be a difference in verbal agreement too. Indeed, agreement reduction occurs with first and second but not with third. I will list some here but see Quirk & Wrenn (1955) and van Gelderen (1997). In (157), (158), (159) and (161) an -að ending is expected whereas in (160) an -st ending is. Since the -e ending is unspecified, I will not gloss it for person or number:

(157) Ælfric, Hom I, 88.32

Nelle we ðæs race na leng teon

not-want we that story not long teach

`We do not want to teach that story long'.

(158) Idem, 280.4

Nu hæbbe ge gehyred ...

now have you-NOM heard

`Now you have heard'.

(159) Idem, 286.15

Ac wite ge ðæt nan man ...

But know you-NOM that no man

`But do you know that noone'.

(160) Exeter, Christ 1487

For hwon ahenge þu mec

Why hang you-NOM me-ACC

`Why are you hanging me'.

(161) Caligula 537

Nulle we noht þis on-fon. ah we faren wlle[ð].

Not-want we not this accept. but we go want

`We do not want to accept this but we want to go'.

None of the cases of reduced inflection have a null-subject since pro needs to be licensed by strong features.

In 3.2, I argue that first and second person pronouns continue to function reflexively because they are less specified; third person ceases to do so because, once the Case becomes structural, they are referential.

3.3 Grammaticalization

Even though the loss of inherent Case does not explain why pronouns are used in direct object position, it may explain the use of `self' marked pronouns in oblique positions. As mentioned in connection with Layamon's Brut, around 1250, the adjectival nature of `self' is lost. It is lost earlier in those connected with third person than in those connected to first and second person pronouns. This means that the structure of `self' changes from modifying adjective to nominal head as in (162):

(162) [ NP[ him] AP[ self]] or, using a DP and N-to-D movement:

DP[ him AP[ [t] self]] --> DP[ his/m NP[ self]].

The fact that himself occurs regularly in the Otho version of Layamon's Brut but not hineself indicates that, once the form is grammaticalized into a pronoun, only the form associated with structural Case (i.e. him not hine) appears in that complex, cf. Table 1. Thus, the third person form himself can check structural Case due to the pronominal part in the later, Otho version. In a Minimalist framework, Modern English himself has structural Case features that are attached to it. In Middle English, the Case on the first and second person pronominal part is not clear. As mentioned, at the time the change in (162) is taking place, the form changes from `me self' to `mi self' in many instances. For instance, in the earlier Caligula edition of Brut, there are 16 mes followed by `self' as in (27) and (29) and these change to mi-seolf in the later text as (28) and (30) show. Another set is (163) and (164), where in the later Otho version, me has been replaced by mi when it precedes `self':

(163) Caligula 11309

her ich sette þe an hond. me seoluen and mi kine-lond

here I place you in hand. myself and my kingly lands

(164) Otho 11309

ich sette þe her an hond. mi-seolfe and myn kinelond..

If the Case of miself and thyself (and early on even of himself) is not accusative Case, it cannot be checked in AGRo. In first and second person, the Case on the pronominal part is more clearly genitive than accusative and hence the checking in Spec AGRo is completely impossible in Gawain. It is confirmed by the confusion that Visser (1963: 95) notes about verbal agreement when myself or thyself are the subject of the sentence. It can either agree or be third person. The OED provides some interesting examples with third person endings, as in (165) and (166):

(165) Chaucer, Wife of Bath, Prol, 175

My self haþ ben þe whippe

`I have-3.S been the whip'.

(166) Shakespeare, Titus Andronicus IV, iv, 74

My selfe hath often heard them say

`I have-3.S often heard them say'.

In Modern English, himself and herself can on occasion be found in subject position. With first and second person, this is harder. However, if a speaker is forced to select the verbal ending she or he will choose the third person indicating that the appropriate person and nominative Case features are not connected with myself and yourself, even though they presumably check the Case. The person features come from the head self.

This unclarity in terms of Case provides a possible account for the appearance of specially marked reflexives exclusively as objects to prepositions: this is a position where they can check non-structural Case. Additional evidence can be found in sentences such as (89) and (94) where the impersonal subjects yowreself and hemself check dative and not accusative. In Modern English, the situation is not unclear: anaphors such as himself do check structural Case (even though they are anaphors[14]).

Concluding this section, I have shown that the reflexive use of simple pronouns in Old English is accounted for under a version of Reinhart & Reuland's Chain Condition. The data in Middle English indicate a difference between first, second and third person pronouns and these can be accounted for by a difference in the strength of phi-features. The introduction of specially marked reflexives in oblique position can be accounted for through Case incompatibilities between the first part of the compound and structural Case. The Paston Letters and Hume present no problems for the theory of reflexives since, except for inherently marked reflexive verbs in the former, the situation is as in Modern English. The question remains why the situation could remain `unstable' as late as Shakespeare. In the Shakespeare text examined, first, second and third person simple and `self'-marked pronouns function both as direct objects and as prepositional objects, even though third masculine forms are more grammaticalized. I leave this stage for further research.

4 Conclusion

I examine stages of English that are very different from Modern English. In Old English, simple pronouns are used anaphorically in all environments. This can be explained using the insight that inherent Case is different from structural Case and that it makes a pronoun into an anaphor, i.e. not fully specified structurally. In Middle English, the situation is more complex. Specially marked anaphors are introduced after 1250 (NB this claim is valid for the texts examined in this paper, but there may be other varieties that have them earlier) but their distribution is unexpected: specially marked anaphors occur outside the immediate domain whereas simple pronouns are used in direct object position. The account valid for Old English cannot hold for e.g. Gawain and the Green Knight since inherent Case is lost in the thirteenth century. I argue that the reason for the unexpected distribution must be sought in the change of `self' from adjective to noun and its not being connected with accusative Case features that must be checked in the structural position (i.e. Spec AGRo). With respect to the anaphoric use of first and second simple pronouns, I argue that their phi-features are incomplete. Additional evidence for this is provided by the absence of first and second person pro-drop. Table 3 below summarizes the changes in pronouns (where `phi 1' stands for the phi-features of the first person pronoun and `C 1' for the Case features; w and s stand for weak or not fully marked and strong or fully marked):

OE eME: Layamon ME: Gawain ME: Chaucer

phi 1 w w w w

phi 2 w w w ---> s

phi 3 s s s s

C 1 w ---> s s s

C 2 w ---> s s s

C 3 w w ---> s s

TABLE 3: Feature Changes

References

Baker, C. L. 1995. "Contrast, Discourse Prominence, and Intensification, with Special Reference to Locally Free reflexives in British English". Language 71.1: 63-101.

Benson, Larry. 1987. The Riverside Chaucer. Boston: Houghton Mifflin Company.

Berndt, Rolf. 1956. Form und Funktion des Verbums im nördlichen Spätaltenglischen. Halle: Max Niemeyer.

Blake, Norman. 1971. Paston Letters and Papers of the Fifteenth Century, Part I. Oxford: Clarendon.

Brook, G.L. & R.F. Leslie, eds. 1963. Layamon: Brut. Oxford: Oxford University Press, [EETS 250].

Burzio, Luigi. 1991. "The Morphological Basis of Anaphora". Journal of Linguistics 27: 81-105.

Callaway, Morgan. 1913. The Infinitive in Anglo‑Saxon. Washington, DC: Carnegie Institution.

Chomsky, Noam. 1981. Lectures on Government and Binding. Dordrecht: Foris.

- 1986. Knowledge of Language. New York: Praeger.

- 1989. "Economy of Derivation and Representation", chapter 2 in Chomsky (1995).

- 1995. The Minimalist Program. Cambridge: MIT Press.

Collins, Chris & Hoski Thráinsson 1996. "VP-Internal Structure and Object Shift in Icelandic". Linguistic Inquiry 27.3: 391-444.

Diehn, Otto. 1901. Die Pronomina im Frühmittelenglischen. Heidelberg: Carl Winter.

Everaert, Martin. 1986. The Syntax of Reflexivization. Dordrecht: Foris.

Faltz, Leonard. 1985. Reflexivization: A Study in Universal Syntax. New York: Garland.

- 1989. "A Role for Inference in Meaning Change". Studies in Language 13-2: 317-331.

Farr, James. 1905. Intensives and Reflexives in Anglo-Saxon and Early Middle English. Baltimore: J. H. Furst.

Funke, Otto. 1907. Kasus-Syntax bei Ormm und Layamon. Wien.

Gelderen, Elly van 1993. The Rise of Functional Categories. Amsterdam: John Benjamins.

- 1996a. "The Emphatic Origin of Reflexives". BLS talk.

- 1996b. "Case of the Object in the History of English". Linguistic Analysis 26: 117-133.

- 1997. "Inflection and movement in Old English". In Werner Abraham and Elly van Gelderen (eds), Problemsyntax. Tübingen: Niemeyer, 71-82.

Givon, Talmy. 1983. "Topic continuity in discourse: the functional domain of switch reference". In John Haiman & Pamelo Munro (eds.), Switch Reference and Universal Grammar. Amsterdam: John Benjamins: 51-82.

Haiman, John. 1995. "Grammatical Signs of the Divided Self". In Werner Abraham, Talmy Givón, Sandra Thompson (eds.), Discourse Grammar and Typology. Amsterdam: John Benjamins.

Hermodsson, L. 1952. Reflexive und Intransitive Verba. Uppsala: Almqvist & Wiksell.

Iatridou, Sabine & David Embick 1997. "Apropos pro". Language 73.1: 58-78.

Jaeggli, Oswaldo & Ken Safir eds. 1989. "Introduction", The Null Subject Parameter. Dordrecht: Kluwer.

Jong, Jelly Julia de 1995. "Local Binding and the Minimalist Framework". In Ale de Boer et al. (eds), Language and Cognition 4. Groningen.

- 1996. The Case of Bound Pronouns in Peripheral Romance. Groningen dissertation.

Kayne, Richard. 1989. "Facets of Romance Past Participle Agreement". In Paola Beninca (ed.), Dialect Variation and the Theory of Grammar. Dordrecht: Foris, 85-103.

Kemenade, Ans van 1987. Syntactic Case and Morphological Case in the History of English. Dordrecht Foris.

Kiparsky, Paul. 1997. "The rise of positional licensing". In Ans van Kemenade & Nigel Vincent (eds.), Parameters of Morphosyntactic Change. Cambridge: Cambridge University Press, 460-494.

Kökeritz, Helge. ed. 1954. Mr. William Shakespeares Comedies, Histories & Tragedies: A Facsimile Edition. New haven: Yale University Press.

König, Ekkehard & Peter Siemund 1997. "Intensifiers and Reflexives: A Typological Perspective", paper presented at the International Symposium on Reflexives and Anaphors, Boulder, CO.

Koster, Jan. 1993. "Towards a New Theory of Anaphoric Binding", ms Groningen University.

Klaeber, F. ed. 1922. Beowulf. Boston: D.C. Heath [1950 edition].

Krapp, G.P. 1931. The Junius Manuscript. New York: Columbia University Press.

Leek, Frederike van der 1994. "Dutch Pronominals", talk, Stanford.

Lightfoot, David. 1979. Principles of Diachronic Syntax. Cambridge: Cambridge University Press.

- 1991. How to Set Parameters. Cambridge: MIT Press.

Mitchell, Bruce. 1979. "Old English Self". Neuphilologische Mitteilungen 80: 39-45.

- 1985. Old English Grammar. Oxford: Oxford University Press.

Mustanoja, Tauno. 1960. A Middle English Syntax. Helsinki.

Ogura, Michiko. 1989. Verbs with the Reflexive Pronoun and Constructions with Self in Old and Early Middle English. Cambridge: D. S. Brewer.

Penning, Gerhard. 1875. A History of the Reflexive Pronouns in the English Language. Bremen.

Pollock, Jean-Yves. 1989. "Verb-Movement, Universal Grammar, and the Structure of IP". Linguistic Inquiry 20: 365-424.

Quirk, Randolph & Christopher Wrenn 1955 [1977]. An Old English Grammar. London: Methuen.

Reinhart, Tanya & Eric Reuland 1993. "Reflexivity". Linguistic Inquiry 24.4: 657-720.

Reuland, Eric. 1997. "Encoding Anaphoric Relations" (this volume?).

- & Tanya Reinhart 1995. "Pronouns, Anaphors and Case". In Hubert Haider et al. (eds) Studies in Comparative Germanic Syntax. Dordrecht: Kluwer, 241-268.

Rizzi, Luigi. 1982. Issues in Italian Syntax. Dordrecht: Foris.

Sedgefield, Walter. 1899. King Alfred's Old English Version of Boethius. Oxford: Clarendon.

Selby-Bigge, L.A. ed. 1902. Hume Enquiries. Second Edition. Oxford: Clarendon.

Solá, Jaume. 1996. "Morphology and Word Order in Germanic Languages". In Werner Abraham et al. (eds) Minimal Ideas. Amsterdam: John Benjamins, 217-251.

Sweet, Henry. 1871 [1934]. King Alfred's West-Saxon Version of Gregory's Pastoral Care. London: Oxford University Press.

Taraldsen, Tarald. 1996. "Reflexives, Pronouns and Subject/Verb Agreement in Icelandic and Faroese". In James Black & Virginia Motapanyane (eds.), Microparametric Syntax and Dialect Variation. Amsterdam: John Benjamins.

Tolkien, J.R.R. & E.V. Gordon. eds. 1925. Sir Gawain and the Green Knight. Oxford: Clarendon.

Ura, Hiroyuki. 1994. Varieties of Raising and the Feature-based Bare Phrase Structure Theory. MIT Occasional Working papers in Linguistics 7.

Visser, F.Th. 1963. An Historical Syntax of the English Language I. Leiden: Brill.

Wasow, Tom. 1977. "Transformations and the Lexicon". In Peter Culicover, Tom Wasow & Adrian Akmajian, Formal Syntax. New York: Academic Press.

Wülfing, J.E. 1894-1901. Die Syntax in den Werken Alfreds des Grossen I and II. Bonn: Hanstein.

Zeitlin, Jacob. 1908. The ACI and Some Kindred Constructions in English. Columbia PhD.

Notes