A History of English Reflexive Pronouns
Elly van Gelderen
2000: John Benjamins, Amsterdam
In the course of the history of English, morphological case and agreement gradually disappear. At the same time, auxiliaries, determiners, pronouns and prepositions become more frequent, and the word order becomes fixed. These changes are often described as transforming the language from synthetic (lacking auxiliaries, etc.) to analytic (having auxiliaries, etc.), the strictness of word order compensating for the lack of case and agreement endings, and the introduction of auxiliaries compensating for the loss of inflection for tense and aspect. Thus, word order and morphological agreement are seen as different ways of expressing grammatical relations: Old English has morphological case and agreement but relatively few auxiliaries, determiners, pronouns and prepositions; Modern English has very little case and agreement but many auxiliaries, determiners, pronouns and prepositions that occupy fixed positions. In addition, in Old English (as well as in many other synthetic languages), specially marked reflexives do not occur, whereas in Modern English (and in other analytic languages), they do. This book links the changes in reflexives to the transformation of English from a synthetic to an analytic language.
In the generative framework of the 1980s, a fixed word order such as that of Modern English comes about because lexical elements move to certain positions (functional categories), as in Chomsky (1986b). Since agreement and case are situated in these positions, movement is seen as motivated by the need to check case and agreement. However, this framework provides no explanation for why case and agreement would be less present in Modern English than in Old English, or why word order would be more constrained in Modern English than in Old English, or why the subject can be left out in Old English. I will account for these phenomena and base my account on an argument advanced in van Gelderen (1993) that certain functional categories (for the auxiliary, the determiner, etc.) are introduced or activated in the course of the history of English. If this is the case, word order becomes fixed because auxiliaries such as will and be occupy functional categories and lexical items move to these projections to check their case and agreement features. I examine the changes in case and agreement features of pronouns as English becomes more analytical, using the insights provided by reflexively used pronouns and special reflexive pronouns. I argue that the feature composition is responsible. I also describe the relationship between word order and overt inflection. Developments with null subjects, agreement, and case can be argued to parallel those of reflexives closely. Hence, I not only connect the loss of case and agreement to the introduction of strict word order, but also relate the introduction of specially marked reflexive pronouns to the increasingly analytic character of the language.
Reflexive pronouns such as myself in Modern English are not fully referential: they need to refer to another element in the sentence. Personal pronouns such as me, on the other hand, are somewhat referential and are barred from referring to arguments `locally' (cf. Chomsky 1981; Koster 1993; Reinhart & Reuland 1993). The reason for the difference, I argue in this book, lies in the feature composition of the pronouns, namely the fact that the features of myself are not as specific as those of me. Work by Rizzi (1990) and Woolford (1999) explains the lack of reflexive forms in subject position through the lack of agreement features, which also indicates that reflexives are underspecified. In Old English, unlike in Modern English, there are no special reflexive forms; instead general pronouns are used reflexively, as they are in Old Egyptian (cf. Gardiner 1927: 40) and Middle Dutch (cf. Hermodsson 1952: 263ff.), for example. I relate this to Reinhart & Reuland's (1993) condition on antecedent-anaphor links: certain case and person/number features render a pronoun reflexive. The book examines what these features are, how they change, and how this change is related to the larger changes in the language.
In a number of Old and Middle English constructions involving pronouns, including those with pronouns functioning reflexively, there is a person split. For instance, first and second person pronouns continue to be used as reflexives (I see me) long after third person ones cease to do so. There are other differences as well. In certain Old English texts, the inflection of `self' after third person pronouns is definite, whereas it is indefinite in other cases, indicating that `thou self' and `Beowulf self' are different from `him self'; case on first and second person pronouns disappears earlier than on third person ones; pro-drop is more common with third person pronouns; and verbal agreement is marked more on third person verbs. There is a similar split between plural and singular: plural has `conservative' reflexives, no morphologically specified case, less pro-drop, and less agreement than singular. To explain the split(s), I argue that two changes occur in pronominal features, at different rates for different features: (a) case becomes structural rather than inherent (or theta-related), and (b) person and number features become checked in functional categories. These changes can be phrased in terms of Chomsky's (1995; 1998ab) distinction between features that are Interpretable, i.e. relevant to the interpretation, and those that are not. Structural case features are Uninterpretable and must be checked before Logical Form since they are irrelevant there. Thus, in Modern English case is Uninterpretable while in Early Old English it is Interpretable as are some of the person and number features. These person splits are also obvious from different frequency of use, a phenomenon not usually dealt with in a generative approach. My account is that changes in feature composition are gradual, i.e. a particular pronoun has either Interpretable or Uninterpretable features for a period of time.
The explanations provided in this book are Minimalist, but the data are described in general terms so as to be accessible to linguists working in other frameworks. The outline of this introductory chapter is as follows: in the first section, I provide some background on functional categories and Minimalist features; in the second section, I outline the theory of Binding I use; in the third section, for ease of reference, I list the general personal pronouns in Old English with some instances (in later chapters, other paradigms are listed where they become relevant); and in the fourth section, I provide a structure for pronouns. In the last section, I justify the selection of texts I have made and provide a short outline of the book.
0.1 Background on Functional Categories (FCs) and Features
In the first subsection, I outline some basic notions on phrase structure and functional categories (FC) in a pre-Minimalist (Chomsky 1986b) and Minimalist (Chomsky 1992; 1995) framework. In the second one, I focus on the role of +Interpretable features.
0.1.1 Functional Categories and Checking
Chomsky (1998b: 123) writes that "[f]rom the origins of generative grammar, the fundamental operations were taken to be formation of the lexicon and recursive operations of two kinds that make use of lexical items: phrase structure and transformational rules". Much of the effort of making the formalism less language specific and more universal was aimed at generalizing phrase structure rules (e.g. through X'-theory) and reducing transformations to one (move-alpha). The Minimalist framework continues that: phrase structure rules become `bare', i.e. no intermediate levels appear, and lexical items are combined by `merge' and moved if necessary. In Chomsky (1995: 235 ff.), lexical items are taken from the lexicon (and inserted in the numeration) fully inflected, but see Halle & Marantz 1993 for alternatives. For the sake of convenience, a lot of work still assumes explicit phrase structure rules (rather than bare ones) and, hence, I outline those here.
In Chomsky (1986b: 2-4), FCs such as the complementizer and the auxiliary are considered on a par with lexical categories and head their own projections. Thus, in (1), a C(omplementizer) such as that and an auxiliary such as will project to full maximal projections, namely CP and I(nflection)P, which also contain a specifier position. The specifier of CP can be used when wh-elements move in questions and, when C is not occupied by that, the auxiliary can move there in questions as well. The grammatical subject occupies the specifier of IP position and the Specifier-Head (Spec-Head) relationship accounts for nominative Case and verbal agreement between the NP in Specifier position and the verbal element in the Head I position (from now on, whenever Case is used in a technical sense, it will be capitalized):
that Zelda will see Bela
Chomsky (1992: 173), based on Pollock (1989) and Chomsky (1989), i.e. `early' Minimalism, argues that all Case is checked in a Spec-Head relationship. For this purpose, several FCs, such as AGRs and AGRo, are introduced. NPs move to the Specifier and verbs to the Head positions. Nominative Case is checked against AGRs and objective against AGRo. Verbal agreement is checked in a Head-Head relationship between V and AGR after the verb incorporates into the AGR Head. The person and number features of the head are given determined by the NP (just as the verb determines the Case: if in AGRs, nominative; if in AGRo, accusative). The checking of Case and agreement occurs either overtly or covertly, depending on whether the features in the functional head are strong or weak. In Chomsky (1992: 196; 199), there are two types of features: N-features and V-features. The former are responsible for triggering NP-movement and for checking Case; the latter for triggering V-movement and for checking agreement. Overt checking of the NP takes place in a Spec-Head relationship as in (2) before SPELL-OUT (or at s-structure in earlier frameworks); covert movement will mean that the element must wait until LF to check its features because this is `cheaper.' English is generally assumed to have weak V-features and the verb does not move overtly. Therefore, in (2), a French example is given where both the NP and V move overtly:
Zorai AGR XP
V AGR ....
arriveraj [sV] VP
will-arrive [sN] ti V'
Other FCs are introduced as well. Thus, T(ense)P accompanied by V- and N-features is included in (3), which is a typical tree structure. Categories such as ASP(ect)P, VoiceP, Perf(ect)P, Num(ber)P, PersonP and others are also possible:
Chomsky (1995: 349ff.) argues that there is no direct evidence for AGRs and AGRo in English. Since agreement, tense and Case features would be checked through Head-Head and Spec-Head agreement in the IP in (1) above, the tense and agreement features are not necessarily connected to one particular functional head and consequently not all projections need be present.
As mentioned above, trees such as (3) are no longer assumed in Chomsky (1994; 1995). After taking elements from the lexicon, they `merge' into phrases such as (4). Through merge, the Head I(nflection) and Specifier are added to (4) after which the V and N move into these positions. Thus, the tree is built from bottom to top:
In the remainder of this book, and especially in chapter 4, I assume a version of (1), i.e. a tree structure with C and I. For the sake of clarity, my structures will not be bare as in (4), but nothing hinges on this.
0.1.2 Interpretability of features
For each linguistic expression, a grammar makes available two kinds of information, phonetic and semantic, or a Phonetic Form (PF) and a Logical Form (LF), in older terminology. The PF representation gives information to the Articulatory-Perceptual system and the LF one to the Conceptual-Intentional system. Legibility must be ensured at these interface levels (Chomsky 1998b: 119). Features are therefore divided as to whether they are phonetic, i.e. not allowed at LF, or semantic, i.e. not allowed at PF. Thus, a derivation splits into two parts. There are, however, features in language that are neither phonetic nor semantic, thereby violating legibility. These features are Uninterpretable and do not "enter into interpretation at LF" (Chomsky 1995: 277); they exist "to force movement, sometimes overtly" (p. 278) to a higher FC.
In the generative framework, movement has always been seen as problematic. As Chomsky (1998a: 42) puts it, "[w]hy language should have this [movement] property is an interesting question, which has been discussed for almost 40 years without resolution". Verbal agreement and Case are problematic as well since they are not relevant to the interpretation in Modern English. Chomsky (1998a: 42-8) proposes to connect both of these problems: the `offending' Uninterpretable Case and agreement are eliminated through movement.
Thus, Uninterpretable features trigger movement but Interpretable ones do not. Interpretable features are relevant at LF and do not erase or delete but can be `used over.' Uninterpretable features explain several phenomena earlier treated as separate; for instance, (a) an NP has one and only one Uninterpretable Case feature, as (5) shows, and (b) the features justify the inclusion of FCs in the numeration and the ensuing movement into the heads and specifiers of these projections. In (5), Zoya cannot check the Case in both subject positions:
(5) *Zoyai seemed ti was annoyed with Amir.
According to Chomsky (1995: 283), the person and number, i.e. phi-, features of Nouns are Interpretable because they can be reused. The example given by Chomsky is (6) where John moves to the subject of the IP to check its Case, checking phi-features along the way:
(6) Johni is [ti AGR [ti intelligent]].
However, in (6), there is no agreement between intelligent and John, and AGR may not have been activated. Alternatively, the movement to Spec AGRP may have to do with the categorial N-features in AGR that must be checked. Thus, there is no empirical evidence that the phi-features are Interpretable. In languages other than English, there is such evidence since the number features appear twice, both on the verb and on the adjective or past participle. An instance is French where the number features in (7) appear on both the finite verb sont `are-3P' and the past participle parties `left-FP.' The person features are only marked on the finite auxiliary and the gender features only on the past participle (both in bold):
(7) Les femmes sont parties.
the women are-3P left-FP
`The women have left.'
There are other languages that have number and gender marked on the past participle. In Spanish, for instance, the passive participle as in (8), inflects for number and gender, but not for person; and in Swedish, number is marked on the past participle in (9) (there is no gender in the plural on Swedish verbs and finite verbs show no inflection). The data in (7) to (9) might indicate that person is not Interpretable and cannot be checked twice:
(8) Las casas son vendidas.
the houses are-3P sold-FP
(9) Tre bilder blev målade.
three pictures were painted-P
There are, however, languages where person is marked on both auxiliary and participle. Van Driem (1987: 119) gives instances from Limbu, a Tibeto-Burman language spoken in Nepal. Anderson (1999) cites other cases of what he calls `split inflection', namely Gorum (an Austroasiatic language) and Venda (a Bantu language) display person on both the auxiliary and the past participle. This is expected if the interpretability of features varies cross-linguistically.
With object agreement, as in (10) from O'odham, person features (in bold) do occur on the participle. Here, person appears as well as number, and so, nothing blocks person marking on participles. It just does not seem to be the case that person is `re-used', i.e. Interpretable, in (7) to (9):
(10) Ceoj 'o 'añi: ñceggia.
boy is/was me 1S-fighting
`The boy is/was fighting me.' (Zepeda 1983)
I therefore argue, contra Chomsky (1995), that person features in a number of languages (including Modern English) are Uninterpretable and are checked only once. Number can be re-used as in (7) to (9) above.
There is some dialectal evidence from Belfast English that the features of pronouns are checked differently from those of full NPs. Henry (1995: 16) describes Hiberno English constructions as in (11) and (12) where the number features of the full noun in (11) are not checked but the ones of the pronoun in (12) are:
(11) The eggs are/is cracked.
(12) They are/*is cracked.
In standard English, the phi-features of both pronouns and full nouns must be checked before LF, again an indication that person might be Uninterpretable.
According to Chomsky (1995: 232 ff.), overt NP- and V-movement occurs if the features of the FC are strong; covert feature attraction takes place when the features of the FC are weak. Thus, the subject moves to Spec IP overtly to check the strong categorial feature ([D]) in I. The phi-features and Case move along and are checked in due course. For English, assuming the non-listed, categorial V-features are weak, movement of the subject to check [D] is the only overt movement necessary. The phi-features of the verb or Auxiliary are attracted to I without overt movement of the verb. The object NP (or its features) moves to the specifier of (the small) v and the main verb adjoins to v. The result is that all the Uninterpretable features are checked:
[D] . .
[Case] v VP
[D] NP V'
[Case] [phi] [phi]
[phi] [Case] [Case]
The woman sees a javelina
If only strong features trigger overt movement, there is a possibility that Uninterpretable features are not checked by LF. However, in Chomsky (1998ab), this is no longer a possibility and features can be attracted through feature-attraction even if the lexical element does not itself move. Feature-attraction is more economical and involves only head-movement of the features (Chomsky 1995: 271); it is formulated to "have the following property: an uninterpretable formal feature UFF in the extended lexical item ELI seeks the closest matching feature F in its c-command domain and attaches it to ELI, UFF then erasing if the match is successful" (1998b: 124). Thus, the modification from Chomsky's (1995) analysis is that it is not only strong features that must be checked before LF is reached, but all Uninterpretable features since only Interpretable features are visible at LF. Hence, the strong/weak distinction is replaced. The evidence for this is (14), in which the expletive there does not check the Case features, since otherwise the Case features of the postverbal five javelinas would not be attracted. If this happened, the Uninterpretable Case features of the NP would remain unchecked and the sentence would not be well-formed:
(14) There are five javelinas in our backyard.
As Chomsky (1995) notes, if the expletive were present to check the phi-features, the Interpretable plural phi-features of the noun would not be attracted to I(nflection) and again, (14) would not converge. Since (14) is grammatical, there is only inserted to check the Uninterpretable categorial features. The problem now is to explain why the subject position in (14) must be lexically filled and why attracted D-features do not suffice in (15). Some stipulation for D-features must be made:
(15) *e are five javelinas in our backyard.
So far, the Case discussed in (13) is grammatical or structural Case, dependent on the nominal's position in the sentence. There is another kind of Case, namely inherent Case, dependent on the thematic structure. Chomsky (1986a: 193) "distinguish[es] the `structural Cases' objective and nominative, assigned in terms of S-structure position, from the `inherent Cases' assigned at D-structure. . . . Inherent Case is associated with [theta]-marking, while structural Case is not". Inherent Case is relevant at LF. As mentioned, in Old English, there is more evidence for inherent Case than in Modern English (cf. also van Gelderen 1996b). For structural Case, there is a one-to-one relationship between Cases and nominal elements. Belletti (1988) and Mahajan (1990) assume that inherent Case is optionally assigned/checked. The nominal, when it does not have inherent Case, may check its structural Case, if available. Thus, in many languages, nominals have either structural or inherent Case. The structural Case features are Uninterpretable but the inherent ones are not. The former make it necessary for a lexical element to move to an FC; the latter do not.
In conclusion, I assume that linguistic expressions have a phonetic and a semantic component. In the `ideal case', all features would be relevant at either LF or PF. This is, however, not true since there are features that force movement that are neither semantic not phonetic. These are the Uninterpretable Case and agreement features. They force movement but are not relevant to the interpretation. Above, and throughout the book, I argue that languages and different stages of the same language differ as to which features are Interpretable. In Modern English, Case features and the person and number features of verbs are Uninterpretable but, I argue, there is no direct evidence (cf. (6) versus (7)) that all nominal phi-features are Interpretable. In other languages, number features on nominals are Interpretable, but not person. Case features are Uninterpretable in Modern English (i.e. structural) but Interpretable in Old English (i.e. inherent). Thus, the status of features ultimately accounts for differences in word order, Case and agreement across languages, and for whether a language is synthetic or analytic. It will also account for the referential or non-referential nature of pronouns.
0.2 Background on Binding
In this section, I outline several theories that have been formulated to account for binding phenomena, namely Chomsky (1981; 1986a), Reinhart & Reuland (1993), and Koster (1993). These attempt to explain the binding domain and differences in this domain across languages. I also discuss Burzio (1996) since he provides an account of person differences. I adopt and adapt Reinhart & Reuland's account by focussing on the Chain Condition. The main points I want to explain with this condition in the remainder of the book are: (a) why pronouns function reflexively in Old English, (b) why special reflexives appear first outside the argument domain of the verb, and (c) what explains the person split. I will not be concerned with Long Distance Anaphora (cf. Koster & Reuland 1991; Brinton 1995) or `irregular reflexives' (cf. Cantrall 1974).
Chomsky (1981: 220; 1986a: 166) formulates three well-known Binding principles:
(16) (A) an anaphor must be bound in its governing category
(B) a pronoun must be free in its governing category
(C) an R-expression must be free
An instance of an anaphor is myself in (17). According to A, an anaphor must be bound in a particular domain, defined in Chomsky (1981) as its governing category. The governing category includes the anaphor, its governor, and a SUBJECT (either the subject of a non-finite clause or the AGReement part of a finite clause). Thus, (18) is ungrammatical since myself is not bound inside the domain that includes it, the governor saw, and the finite AGReement:
(17) I see myself.
(18) *I thought that [she saw myself].
Before the inclusion of SUBJECT in the domain, the Specified Subject Condition (e.g. Chomsky 1973: 90) and Tensed Sentence Condition (e.g. Chomsky 1973: 98) excluded sentences such as (18), (19), and (20):
(19) *I want [her to see myself].
(20) *I thought that myself had won.
In (19), myself cannot be bound outside the subordinate clause since the subject her blocks this, and in (20), myself cannot be bound outside the tensed subordinate by I. Both (19) and (20) can be accounted for by including SUBJECT in the definition of governing category since her would be the SUBJECT in (19) and the finite AGReement would be in (20). Cross-linguistically, there is variation as to what constitutes a governing category. For instance, the Korean counterparts of (19) and (20) are correct since Korean does not include SUBJECT. Instead, anaphors must be bound in the root clause (see Wexler & Manzini 1987 for more on cross-linguistic parameters).
According to Condition B, pronouns such as me in (21) must be free. In (21), the pronoun is bound and hence the sentence is ungrammatical. Thus, anaphors and pronouns are frequently in complementary distribution:
(21) *I saw me.
R-expressions, where R stands for Referential, such as the second NP in (22), must be free in the entire sentence:
(22) *Rolando noticed that Rolando left early.
In Chomsky (1995: 211), the three conditions are recast as interpretative principles at LF but the basic insights remain. In this book, anaphors as in (16A) will be referred to as reflexives. When a referring item is not in an argument position (e.g. not a direct, indirect, or prepositional object position), it is referred to as an emphatic (cf. König & Siemund 1997; 1998 for conditions under which these appear).
Other cross-linguistic and cross-dialectal variation in the binding domain exists in the famous `snake'-sentences. It is well-known that in English, as (23) below shows, the pronoun can be coreferential with the subject; in German, as in (24), ihr cannot and the reflexive sich is needed; and in Dutch, as in (25), both are possible (although not all speakers accept haar):
(23) I saw a snake near *myself/me.
(24) Sie sah eine Schlange neben sich/*ihr.
`She saw a snake next to herself/her.'
(25) Zij zag een slang naast zich/haar.
`She saw a snake next to herself/her.'
Other languages display similar variation (cf. de Jong 1996 for Romance), which is problematic since the governing category for an element should not be so different for different languages. Condition (B) is also problematic for (26) and for languages such as Old English where sentences such as (21) are grammatical (see also Baker 1995; Haiman 1995):
(26) I 'll buy me a dictionary.
To account for the `snake'-sentences, different types of solutions have been proposed. Reinhart & Reuland (1993) argue that Binding Theory should be formulated as a condition on predicates (the verb and its arguments) rather than as a condition on anaphors and pronouns. Their conditions are listed in (27) (I will ignore the distinction between syntactic and semantic predicates):
(27) (A) A reflexive-marked syntactic predicate is reflexive.
(B) A reflexive semantic predicate is reflexive-marked. (p. 678)
In (17), the predicate is reflexive-marked (one of its arguments has -self) and therefore its two arguments must be coindexed. This condition is met since I and myself corefer. In (21), the predicate is reflexive since two of its arguments are coindexed but it is ungrammatical since it is not reflexively marked. In (23), me is not part of the predicate and hence the predicate need not be marked as reflexive. In sentences such as (28), the reflexive is an argument (a benefactive object) and, hence, part of the predicate, as opposed to (23), where the pronoun is part of an adverbial:
(28) I bought it for myself.
Reinhart & Reuland's definition differs from the three conditions in (16) in that the governing category, i.e. the binding domain, is reformulated as the predicate. Indirect or beneficial objects as in (26) and (28) are problematic because they are sometimes treated as obligatory arguments and sometimes as optional ones. This has consequences for Binding Theory. Reinhart & Reuland's Condition (B) is stated so that a predicate with two coindexed arguments must have reflexive marking. Hence, (21) is ungrammatical. In (23) and (26), if me is not a proper argument to the predicates see and buy respectively, reflexive marking is not necessary and the sentences are correctly predicted to be grammatical.
In addition to (27), Reinhart & Reuland claim there is a Chain Condition that allows pronouns to be used anaphorically if they are not fully marked for Case and phi-features. This allows for variation among languages and between stages of a language as I show in chapters 1 and 2. The Condition on A-Chains can be formulated as in (29):
(29) Condition on A-Chains: a maximal A-chain contains exactly one link -- α1 --that is both +R and Case marked (cf. Reinhart & Reuland, p. 696).
The property R involves referential independence and is defined as having "a full specification for φ-features and structural Case" (p. 697). In Reuland & Reinhart (1995: 255ff.), a full specification includes having a nominative/accusative contrast. Nichols (1997: 79-84) also shows that inherently Case marked arguments in Zuni are not visible to other arguments, which means that inherently Case marked elements are not referential for purposes of binding. So, only by being checked, i.e., as I argue, by having Uninterpretable features, can an element function referentially.
Even though Reinhart & Reuland do not mention (26), the Chain Condition allows anaphoric 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. Likewise, the pronoun in (23) could be argued to have inherent Case (even though pseudo-passive constructions do not provide evidence either way: both `She was looked next to/near' and `Her was looked next to/near' are ungrammatical). So, Reinhart & Reuland's theory contains both a condition on predicates and one on arguments.
In Old English, predicates are not reflexively marked, i.e. there is no argument marked by self. This means that Reinhart & Reuland's Condition on Predicates does not apply. The Chain Condition does, however. I will show that, in Old English, the Chain Condition is relevant with respect to Case features, and in Middle English, with respect to person and number features. The question then arises if the Condition on Predicates can be reduced to the Chain Condition for other languages as well. I argue it can for Modern English, for example, if one considers forms such as myself unspecified for person features, and hence unable to be referential. In Old English, `self' is an adjective, but in Middle English, it becomes a noun and the head of the reflexive pronoun complex. Due to the lack of person features of self in (Middle and) Modern English, the features of the pronoun complex are unspecified and the complex can function reflexively. This lack of person features is similar to the situation in Yiddish, for example, where zikh (Weinreich 1949 : 100) can be used as a first, second or third person reflexive. The reason me is ungrammatical in (21) can also be explained using the Chain Condition since the Case is structural and renders the pronoun referential. Hence, condition (29) is sensitive to the distinction between Interpretable and Uninterpretable features.
There are two related issues about (29) that neither Reinhart and Reuland (1993) nor Reuland & Reinhart (1995) address: (a) what role inherent Case, i.e. an Interpretable feature, plays in the interpretation so as to make an element non-referential, and (b) what role structural Case, i.e. a Uninterpretable feature, plays in making an element referential. One could argue that the unspecified or Uninterpretable features of the object in (17) are checked on the verb (through movement to an FC) before LF but make the predicate `reflexive' at LF. The relations between the subject and object with Interpretable features in (23) are `calculated' at LF and hence do not obey the Chain Condition. Thus, pronouns with Interpretable features can be either referential or reflexive and the interpretation of an Old English pronoun can be reflexive, as in (30), or referential, as in (31):
(30) Beowulf 1799
Reste hine þa rumheort
rested him the big-hearted one
`The big-hearted one rested himself.'
(31) Beowulf 447
gif mec deað nimeð
if me-ACC death takes
`If death seizes me.'
Koster (1993) reformulates the notion of governing category in Minimalist terms and crucially uses Case checking. He argues that morphologically marked anaphors are strong and must be checked with AGR(eement) (assuming a split IP as in (32), which Chomsky 1995 no longer does). Languages differ as to where the feature is located. If it occurs with AGRs, the position responsible for subject agreement, as in German (and Slavic), non-argument pronouns cannot function as anaphors; if it occurs with AGRo, the position responsible for object agreement, as in English, non-argument pronouns function anaphorically. Thus, in (24), the reflexive is in the domain of AGRs and checks its feature; in (23), it is not and a pronoun appears. The anaphor in (17) 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:
. AGRo' ... neben ihr/next to her
Thus, both Reinhart & Reuland and Koster argue that domains can vary in terms of whether or not adjuncts are included and that inherent Case marking enables a pronoun to serve as an anaphor. Since the presence of inherent Case varies from language to language, pronouns function anaphorically in some but not in other languages.
Another approach to Binding is provided by Burzio (1996). He argues that the antecedent is important and that anaphora is a kind of agreement between the anaphor and the Subject/Inflection complex. If verbal agreement is strong (as in many Indo-European languages), pronominal reflexives are less likely than if it is weak (as in East Asian languages). However, agreement in languages such as Modern English, with no pronominal reflexives, is weaker than in Old English (see chapter 4 below), a language with pronominal reflexives. Hence, this cannot be correct.
Burzio also makes a claim that is possibly relevant to the person split found in Old and Middle English. He argues that, in Italian, "[c]ertain reflexives . . . are morphologically invariant for all gender, number, and person, a fact which [he] interpret[s] as actual lack of morphological features. . . . If correct, this means that these reflexives cannot truly agree with their antecedents . . . but can only `pseudo' agree, in the sense of not bearing distinct features" (pp. 4-5):
(33) Ioi parlo di *sei/mei.
I talk about self/me
`I talk about myself.' (Burzio 1996: 4)
In (33), se is partly specified and is not allowed in Italian, probably because the first person features are stronger in Burzio's terms (however, me-stesso `myself' is also allowed (Burzio 1996: 6)). So, as in Reinhart & Reuland (not cited in Burzio), the features of the reflexive play a role: se is unspecified and must be anaphoric.
The person split could be accounted for in structural terms as well, assuming that first and second person pronouns check their features in different FCs, as in Rice & Saxon (1995) and Ritter (1995). For instance, one might argue that first and second person pronouns are checked as in (34) but that third person ones, as in (35), need not be. I will not entertain these structures for Old and Middle English because (a) there is no person split in Old or Modern English, and (b) there is no structural evidence for such FCs in Middle English (cf. van Gelderen 1993):
In short, I will use (29) extensively in chapters 1 and 2 to explain the behavior of reflexives. I argue that it explains the facts in Modern English and that Reinhart & Reuland's additional Condition on Predicates is not necessary. Chapters 3 to 6 provide support for using the Chain Condition.
0.3 Old English Morphology and Inherent Case
In this section, I list the basic pronominal paradigm since this will be helpful in the chapters that follow. Partly on the basis of the morphological richness of this paradigm and the thematically predictable nature of the (object) Cases in sentences such as (31) to (40) below, I argue in chapter 5 that (object) Case is inherent. There are also other arguments, namely lack of verbal passives, and exceptional Case marking. First, however, a brief note on the status of pronouns in Old and Middle English.
Pronouns in Old English act like clitics in that they occupy fixed positions in the sentence (more in 2.5.1), as has been argued by Traugott (1972) and van Kemenade (1987). For instance, in (36) to (40) below, object pronouns are fronted. My focus, in this book, will be on the feature content of pronouns rather than on their position in the sentence. I argue that even though first, second and third person pronouns all occupy clitic positions, their internal structure is different. Since I do not focus on their position, I continue to refer to these elements as pronouns. The same is true for the distinction between weak and strong pronouns (cf. also 2.5.1). For instance, in Dutch and Middle English, first and second person weak pronouns act positionally like third person ones but differ in terms of feature content.
Campbell (1959: 288-9) lists the basic Old English pronominal paradigm as one where me and þe are used for dative and accusative even though some texts are said to have a distinct accusative form. Quirk & Wrenn (1957: 38) also list the basic paradigm without special forms for first and second person accusative but note that "[e]arly texts sometimes have distinctive a.sg. forms for the 1st and 2nd pers". The paradigm given below lists separate forms for first and second person accusative since they are quite frequent in texts such as Beowulf, Vespasian Psalter and The Lindisfarne Gospels.
Individual texts vary a great deal both in forms and in orthography, especially for third person pronouns (e.g. hiene, hie, hio, hiere, hyne, hym, hyre and hy). Gericke & Greul (1934: 85ff.) comment on dialect differences regarding these pronouns. In general, if variants exist, I use single quotation marks around the word, e.g. `him' for hem, him, hym when referring to the third person dative pronoun:
Table 0.1: Old English Pronouns
Singular Dual Plural
First NOM ic wit we
GEN min uncer ure
DAT me unc us
ACC mec uncet usic
Second NOM þu git ge
GEN þin incer eower
DAT þe inc eow
ACC þec incit eowic
Third NOM he/heo/hit -- hi
(M/F/N) GEN his/hire/his -- hira
DAT him/hire/him -- him
ACC hine/heo/hit -- hi
Instances of some of these in Beowulf are (31) above, (36) to (40), where mec, þec and hine are accusatives, dependent on niman `take' in (31), oferswyðan `overpower' in (36), and teodan `prepare' in (40); ic, þu, hio and hi are nominatives; and eow and us are datives, dependent on wisian `show' in (37), and a benefactive connected to be god `be good' in (38). Full nouns are also marked, e.g. as in (39), through an -e ending for the dative:
(36) Beowulf 1768
þæt ðec dryhtguma deaþ oferswiþeþ
that you-ACC mighty-ruler death overpowers
`that death overpowers you, mighty ruler.'
(37) Beowulf 292
Ic eow wisige
I-NOM you-DAT show
`I will lead you.'
(38) Beowulf 269
Wes þu us larena god
be you-NOM us-DAT teaching-GEN good
`Give us good counsel.'
(39) Beowulf 623-4
þæt hio Beowulfe ...
... medoful ætbær
that she-NOM Beowulf-DAT ... meadcup brought
`that she brought Beowulf the meadcup.'
(40) Beowulf 43
Nalæs hi hine læssan lacum teodan
no-less they-NOM him-ACC less gifts-DAT.P prepared
`They made him no fewer gifts.'
Table 0.1. shows that there is a difference between the nominative forms and the others, especially in the first person and the second person non-singular (ic vs m-; we vs u-; ge versus eow-). This is a remnant of the Indo-European split between active and non-active which is later reanalyzed as a Case split: nominative against the others (cf. Lehmann 1993). The third person develops late and a demonstrative is used (cf. Beekes 1990: 250). Hence, no suppletion occurs in the paradigm.
0.4 The Structure of Pronouns
Since Abney (1987), it has been assumed that the structure of a phrase such as the house is a D(eterminer)P, as in (41):
An advantage of this structure over the traditional NP, as in (42), is that the head the heads its own phrase in (41), just like the head house does, and that the head the does not occupy the position of a maximal projection:
The D head expresses in/definiteness and is considered an FC, on a par with the I position, as in (10) above. The DP has been split into many other FCs, e.g. a Num(ber)P (cf. Ritter 1995), an n(oun)P (Bejar 1999), and a P(er)s(on)P. There is also evidence that indefinite and definite NPs have different structures (e.g. Zamparelli 1995). I abstract away from these splits and use DP.
Modern English pronouns are generally considered DPs but it is undecided whether the pronoun is base generated in D or if it moves to D. The reason it is argued to occupy D is that other Ds cannot co-occur with pronouns, e.g. *the he, *her she, and *that me. In older versions of English, this is not so clear. Mustanoja (1960: 120) gives examples of þe he `the he' and Wood (p.c) finds examples in Early Middle English of sum heo `some they.' The status of articles is not clear either and they occur much less frequently than in Modern English (e.g. Traugott 1972: 85-7). Thus, the evidence for the presence of a D(P) is not as straightforward for Old English. This is not special to Old English. Kornfilt (1991) argues that Old Turkish lacks a D and Philippi (1997) argues that indefinites do not have a DP in Early Germanic.
The position of adjectives is controversial as well: should the adjective appear in Spec NP or as a separate FC? In Old English (and Modern German, Dutch, Swedish, to name but a few), adjectives have either definite or indefinite endings depending on whether or not they are preceded by no article or a definite one. The structure could therefore be as in (43), with the adjective moving to D if there is no article or if the article is indefinite:
^ A N
`the fierce attack' (Beowulf, l. 1148)
In the case of a definite NP such as se grimma gæst `the fierce spirit' (Beowulf l. 102), this movement does not take place and the adjective has an indefinite ending. Most of the time (in e.g. Beowulf), the adjective occurs without noun and it has been argued that the indefinite inflection is a nominal inflection (see Curme 1905). When there is more than one adjective, only the first has a definite ending (Spamer 1979: 245). This fits with (43) because there is only one D position where definiteness can be checked.
In this section, I have indicated a number of questions about the status of the DP. Is the DP universally present (as in Progovac 1998) or does it have to be activated by the language learner on the basis of language data? Does the pronoun move to D or is it base generated there? In this book, I assume that there is some representation of definiteness in Old English, probably as D, as in (43). However, I will argue that pronouns are not always DPs, for instance, first and second person Old English ones are not.
0.5 Justification of the Texts used and Outline of the Book
I have selected the texts used in this book by looking through a variety of texts representing different time, dialect area, and text type. In addition, I have searched the Helsinki Corpus (Kytö & Rissanen 1988) for texts with `interesting' instances of reflexives. With some texts, I have examined and listed every possible reflexive, but doing that for all texts would have been too time and space-consuming. I have used the computer-readable versions of Beowulf, The Vespasian Psalter, The Junius Manuscript, The Exeter Book, The Lindisfarne Gospels, The Rushworth Glosses and works by Alfred and Aelfric. Please see Appendix A for a description of the texts and the standard editions. The computerized editions are readily available from the Oxford Text Archive and Dictionary of Old English Project (Toronto). I have used TACT as a Concordance builder. Middle English material is less readily available in computerized form. I have used the computerized versions of Layamon's Brut (both Caligula and Otho), Gawain and the Green Knight, and Chaucer's entire works. I have also examined texts from the Katherine Group (with the help of the Penn-Helsinki annotated computer version), The York Plays (with the help of Kinneavy's Concordance) and Cursor Mundi. For Early Modern English texts, I have relied on The Paston Letters and the First Folio Edition of Shakespeare's works (both available from Oxford Text Archive). On occasion, where relevant, I have used additional examples from other texts which I did not systematically examine for all aspects dealt with here.
The outline of the book is as follows: In chapter 1, I describe the reflexive constructions in Old English where the vast majority of reflexive elements are simple pronouns. According to the Chain Condition (cf. (29) above), this is not surprising since Old English has a system of inherent Case. In Middle English, described in chapter 2, the situation changes and a special reflexive is introduced, especially with third person pronouns. I argue that changes in the Case, person and number features are responsible for this. If Case becomes structural, it is checked in an FC, which become activated in Early Middle English independently. This is one step in the direction of becoming an analytic language. The first and second person features become un(der)specified or weak (e.g. phonologically), which enables first and second person pronouns, such as `me' and `thee', to continue to be used reflexively (in accordance with the Chain Condition). The specially marked pronoun (e.g. `himself') is first introduced outside the direct domain of the verb and with third person pronouns. I argue that due to the loss of overt Case marking, the adjectival `self' is changed into the head of the complex pronoun. Once `self' is the head, the complex form, i.e. `himself', can function reflexively since it has no features of its own (like Yiddish zikh).
Chapters 3 to 6 provide support for the claims in chapters 1 and 2. In chapter 3, I argue that the underspecification of certain person features can also be seen in the lack of pro-drop with those persons: less pro-drop with first and second than with third. This means the verbal agreement features cease to license an empty subject, another step toward an analytic language. In chapter 4, the underspecification is examined with respect to agreement on the verb: again less agreement with first and second than with third person. Chapters 5 and 6 show that Old English has a system of inherent Case which is first lost in first and second person. This can be seen from the morphology as well as from certain constructions, such as passives and impersonals (discussed in chapter 6), that show a person split. Third person inherent Case is the last to be lost.
. In Chomsky (1995: 233), a D-feature is assumed to trigger NP movement whereas in Chomsky (1992) it is an N(P)-feature.
. As in Chomsky (1995: 349), I assume that Functional Categories do not have phi-features but that I has Case (when finite). In (13), Chomsky allows for the subject being in Spec vP rather than being in Spec VP.
. Reuland & Reinhart (1995: 255ff.) argue that if there is a Case distinction between he/him then Case is fully specified. Since heself is impossible, anaphors in English lack the distinction for Case and that enables them to function anaphorically.
. I will not go into all the details of the Chain-Condition. For instance, even though German has inherent Case in (24), the third person phi-features render ihr pronominal, but not mich or mir in (i) and (ii):
(i) Ich wasche mich
I wash me-ACC
`I wash myself'.
(ii) Ich wasche mir die Hände
I wash me-DAT the hands-ACC
`I wash my hands'.
This is similar to the case of Dutch that I discuss in 2.5.1.
. There are others that focus on the antecedent as well: for example, Authier & Reed (1997) show that whether or not the antecedent is a quantifier has consequences for binding. I will not be concerned with such instances here.