The status of to in Old English to-infinitives: A reply to Kageyama

The status of to in Old English to-infinitives: A reply to Kageyama

ELSEVIBR Lingua 99 (1996) 107-133 Review article The status of to in Old English to-infinitives: A reply to Kageyama Olga Fischer Department of Eng...

2MB Sizes 25 Downloads 29 Views

ELSEVIBR

Lingua 99 (1996) 107-133

Review article

The status of to in Old English to-infinitives: A reply to Kageyama Olga Fischer Department of English, Universio

of Amsterdam, Spuistraat 210, NL-1012 VT Amsterdam

Received 14 March 1995; revised version 27 March 1996

1. Introduction In this paper I would like to discuss a number of problems I had with the interpretation of Old English to in to-infinitives as described by Taro Kageyama (henceforth TK) in an earlier volume of this journal (Lingua 88, 91-128, 1992).’ Let me express first of all my admiration for this article, which is well researched, very carefully put together, and contains a lot of important insights both as far as the theory as well as the data are concerned. What is especially attractive about TK’s proposal is its attempt, on the one hand, to link together synchronically four ‘peculiarities’ of Old English, that is, the absence of a passive infinitive with to, and of constructions like the man tofw the sink, the lack of a lexical subject with to-infinitives (i.e. the so-called Exceptional Case Marking constructions), and the frequent occurrence of retroactive infinitives (i.e. infinitives of the type, these things we to do), which are active in form but passive in meaning; and, on the other, to explain diachronically the simultaneous appearance of three new, related constructions in Middle English. However, even though the author offers cogent argumentation for treating the above Old English constructions together, I believe that the basis on which the whole argument rests, namely that Old English to should be interpreted as an element positioned in the AGR node, is shaky and too meagre to build onto further.

’ I would like to thank Teun Hoekstra, George Jack, Wim Klooster, Willem Koopman, Bettelou Los, Najib Narad, and a number of anonymous reviewers for their constructive criticism of an earlier version of this paper and for their many helpful suggestions. My very special thanks here go to Wim van der Wurff, who has given so freely of his time to discuss numerous problems with me. I am grateful to Chris van der Linden for his extensive help with the history of the ‘stand and wait’ constructions in High German, and to Max N&my, Jutta Christiansen and Andreas Jucker for their willingness to provide me with their native speaker judgments of this construction in Modem High German.

0024-3841/96/$15.00 Copyright 0 1996 Elsevier Science B.V. All rights reserved PIf SOO24-3841(96)00012-5

108

0. Fischer

I Lingua

99 (1996)

107-133

Another problem I had was that semantic (in the sense of ‘text-interpretative’) considerations are almost completely ignored, i.e. a number of constructions are equated (details will follow below), which to my mind cannot be used in the same context due to semantic divergence. Since the historical linguist cannot have access to native speaker judgments and cannot rely on his own intuitions concerning the structure of a particular construction, a semantic consideration of the structures discussed is particularly important in order to make the correct (in as far as possible) syntactic distinctions. Context, therefore, should play an important role in deciding how a particular construction should be analysed. My main problem with TK’s analysis in this respect is that he believes that the presence or absence of to before a coordinated infinitive is optional in Old English, i.e. that it has no function and no meaning. I cannot agree with him. The main part of my argument, therefore, will be to show that the presence (or absence) of to does make a difference semantically. For this purpose, I will look at a number of constructions in which both ro and zero infinitives occur, show how they differ, and how this can be stated in synctactic terms2 The paper is structured as follows. In section 2, I will go into Kageyama’s most fundamental piece of evidence, on which the rest of the argument is built, namely that to should be interpreted as positioned in the AGR node. As we will see, TK

* As a historical linguist, I have also some difficulty with the rather abstract and sometimes idiosyncratic analysis of syntactic constructs, especially the notion of the preposition to functioning before the infinitive as its external argument. TK goes one step further here than the more generally accepted proposal, cf. Vanden Wyngaerd’( 1988). that passive morphology bears the external argument role and that affixes such as -en function as the head of AGRP. I also find problematic the idea that ro assigns dative case to tbe infinitive as a sign of subject verb agreement (why, one wonders, dative case?). I am quite aware of the fact that in generative theorising it is or has become acceptable for categories that started off as ‘cognitively’ recognisable elements, to become more and more abstract (cf., e.g. Haiman, 1993, on the development of PRO). This no doubt accounts for the relative eaSe with which it is suggested that ro should be interpreted as an external argument in Old English, which seems a long way from the ‘surface’ function of ro as a preposition and/or infinitival marker. Of course, this suggestion is part of a natural development towards greater abstraction, leading to greater theoretical elegance and economy, and thus ultimately to a better understanding of UG. But the coin of elegance and economy has another side. It has been shown often that in language change “speakers pay much more attention to their immediately apparent data, in particular to surface forms of words and expressions, than they do to more ‘distant’ data of the sort that a linguist - or a perfect speaker - might be aware of” (Joseph, 1992: 139-140, who also quotes other linguists hem who have presented evidence to this effect). In other words, what may be more economic and/or elegant synchronically or theoretically, may not be so diachronically or from the point of view of the theory of language change. As Joseph puts it (1992: 140). “the typical types of evaluation metrics that linguists use to argue for the proper formulation of a fragment of a grammar cannot (always) be maintained; the grammars linguists construct, therefore, ought to be allowed to reflect uneconomical ‘solutions’, at least in diachrony, but also, given the relation between synchrony and diachrony argued for here, in synchronic accounts as well”. This is not the place to go into such ‘large’ questions in detail. However, I will try to show that TK’s analysis of to, although it seems on the outside certainly more economic and elegant as far as the grammatical system of Old English and as far as an explanation of the relatedness between later diachronic changes is concerned, creates more problems than it solves, and does not do justice to other facts of Old English not paid attention to in his article. I hope to show that a less abstract, and theoretically perhaps a more “uneconomical” analysis of to (which happens to be also a more traditional one), fits the Old English data better and more. comfortably.

0. Fischer I Lingua 99 (1996) 107-133

109

bases this interpretation on the behaviour of coordinate infinitives, i.e. on “the phenomenon of the optional reduction of to in coordinate structures” (p. 96) of the type, (la) Me is geseald anweald to ofsleanne and to edcucigenne me is given power to kill and to revive @Elf.L.S.XXXIV 321-322, TK p. 96) (1 b) Me is neod to farenne and bone geseon Me is need to go and that see (Elf. Horn.11 372,18, TK p. 96) I will try to show - and this will be the main part of this paper - how difficult it is to interpret coordinate constructions of this kind, even in present-day languages, and present evidence that in Old English the constructions with and without to are not syntactically equivalent. In section 3 I will briefly look at TK’s interpretation of the constructions which are said to follow from the placement of to in AGR. If to should not be in AGR, as I believe is the case, the rest of TK’s argument really falls away. However, I will go into some of the facts adduced by TK in order to show that there are problems here too, and that alternatives are available. I will also touch upon some of the related new diachronic developments, which according to TK are more neatly captured by his proposal. Finally, in section 4, I will consider again very briefly the status of to and present an alternative (and less cognitively indifferent, cf. note 2) interpretation, which to my mind covers the facts of Old English more successfully.

2. The interpretation of to in Old English 2.1. TK’s analysis of the status of to and its syntactic position in Old English TK first shows that the Old English to-infinitive has “the appearance of a ‘word’ rather than of a phrase” (p. 95), by which he means that the to-infinitive is a unit that cannot be broken up by intervening elements. Thus, split infinitives are not yet possible in Old English, nor can to be stranded, as in Modem English: You may come if you want to. It is for this reason, too, that van Gelderen (1993) argues that Old English to is not in COMP, but part of the infinitival VP, and that it becomes situated in COMP (TP in van Gelderen) only in Middle English. This accounts for the later changes mentioned by TK, and also for the introduction of ECM (or AcI) constructions in Middle English. In other words, the usual arguments given for the higher level position of to in Present-day English (cf. also Pullum, 1982: 185, who uses the same arguments) cannot be applied to Old English, which still clearly lacks these constructions. So what indications do we have for the status of to, as not part of a PP but part of the AGR node? TK writes that the categorial status of the to-infinitive provides a “clue to choosing between the lexical and syntactic derivations” (p. 99). The lexical analysis, according to TK, “would predict that the entire infinitival clause should be a projection of V, because its head, to + V, is a verb” (p. 99). He shows with the help of examples like (2) (his (IS)),

110

0. Fischer

I Lingua

99 (1996)

107433

(2) Ut eode to his gebede o66e to feorniunne mid his gefentm. to study with his friends Out went to his prayer or ‘He went out to say his prayers or to study with his friends.’ (Bede 162,7, Callaway, 1913: 139; more exx. in Mitchell, 1985: $965)3 that IO V cannot be verbal because it occurs in coordination with PPs, in other words the to-infinitive must be a PP as well. However, the PP analysis according to TK also has its problems because in an example like (lb) “the surface structure . . . contravenes a parallel structure condition that would be required of any reasonable theory of coordination because the two conjuncts . . . are not symmetric” (p. 97). His solution to this is “to generate to in a position that governs the coordinated VP’s” (p. 98) as shown in (3) (TK’s (13)) (note that for TK the VPs are in symmetric coordination - “&P”, he writes, is a “mnemonic representation of a coordination phrase”, p. 98; below, see (12a), I will suggest that the two VPs are not symmetric), AGRP

(3) /\ Spec

AGR /\ AGR I to

Y\ VP >L

I v I

& I

faran

and

>L

NP I

bone

V I

geseon

The advantage of this analysis is according to the author that we “can automatically derive the PP status of the ro-infinitive because the infinitival marker to, though inserted under an AGR node, is lexically specified as a preposition” (p. 99). It seems to me indeed obvious that to + infinitive together cannot be interpreted under TK’s ‘lexical analysis’ as a word, but that does not necessarily mean that TK’s syntactic analysis is the only possible one. The to-infinitive can also be interpreted on a par with nominal to PPs, in which to cannot be separated from the infinitive just as this is not possible with PPs containing a preposition and an NP. It would mean that the infinitive itself is not verbal but nominal. This creates its own problems (see further section 4), but these are perhaps more easily surmountable than the problems created by TK’s proposal. TK’s hypothesis entails that to can be separated from the infmitive in underlying structure, as his rules of to-lowering and V-raising illustrate respectively, ’ The references to Old English texts will be as given in the source-texts. References to Callaway, 1913, Visser, 1963-73, Mitchell, 1985, will be abbreviated to Call., Viss. and Mitch. respectively.

0. Fischer I Lingua 99 (1996) 107-133

V-raising : to-lowering:

111

TO

[NP V] j to-V, [NP ti] to [NP V] + ti [NP tOi_V] (TK’s (14) on p, 99)

Note that in each case the result of the rule is that to and the infinitive appear on the surface next to one another. What empirical evidence then is there that to is not part of a PP but positioned in AGR (apart from TK’s evidence given in (lb), which will be extensively discussed below)? Examples like (2) do not provide evidence for to as AGR, only for the status of to as a preposition. Indeed, I would suggest that examples like (2) argue more strongly for an underlying PP analysis of the to-infinitive than a VP analysis (if the to-infinitive is a PP, to would automatically be a preposition). According to Pullum (1982), the coordination of a nominal PP and a to-infinitive (which is now definitely more verbal) is no longer possible in Present-day English, showing that the status of the to-infinitive has changed.4 It is important, too, to keep in mind the remarks made by Callaway (1913: 60-71), that to-infinitives tend to occur with matrix verbs that take a prepositional object or an object in the dative or genitive, whereas bare infinitives tend to occur with matrix verbs that take an accusative, thereby emphasising the close relation between infinitives and Case in Old English. There are some exceptions to this. Most of these Callaway seems to be able to explain away, but it may also show an incipient moving away of the infinitives from their nominal status.5 Since the prepositions in PPs are always tied to the NP both underlyingly and on the surface, it is rather idiosyncratic, to say the least, that in TK’s analysis to in the to-infinitive is a preposition, but yet one that is not part of a PP underlyingly (TK writes that the verbs in (3) “are both generated in the form of bare infinitives”, p. 98), thus making it quite different structurally from nominal PPs. 2.2. The evidence provided by coordinating constructions for the status of to As already mentioned, the only piece of evidence that TK adduces for the AGR status of to is the behaviour of coordinate infinitives, i.e. the optionality (this is TK’s claim) of to in the second infinitive in the examples in (1) above. It is important to point out how crucial this evidence is for TK, because all his other arguments about the status of to are built up from here (i.e. to as an agreement marker, as an external argument, as an anaphor, etc.; see the discussion in section 3). I would first like to show that coordinate infinitives are an extremely dicy topic, generally, before I go into the more specific characteristics of Old English coordinate constructions. 4 Generally, I think, this is correct. An example like (2) above sounds odd when literally translated into Modem English. Only when the first part of the coordination is also verbalised, as has been done in the translation provided, does the sentence read well. George Jack (personal communication) suggested that this may not be such a hard and fast rule in Present-day English given the occurrence of examples like, He did it for profit and to fulj?ll a long standing ambition. Notice, however, that a different preposition is involved here, so this is a less parallel instance of coordination. 5 I would like to stress that, interesting though Callaway’s suggestion is, there are nevertheless exceptions. It is quite clear that this topic needs further research before any more definite conclusions can be drawn.

112

0. Fischer I Lingua 99 (1996) 107-133

Coordinate constructions are difficult to analyse even in contemporary languages, where we can use intuitions that we do not have for dead languages. This is quite clearly shown, for instance, in Zwart’s (1993) attempt at an analysis of the status of Modem Dutch te (the equivalent of to) before infinitives. Zwart believes, like TK for Old English, that Dutch te in re-infinitives is different syntactically from prepositions in nominal PPs. He first shows that re can be separated from the infinitive. However, the only facts he cites in support of this analysis are taken from Dutch dialects. Although I believe that dialects are linguistically as important as the standard language, I am of the opinion that grammatical structures occurring in dialects should not be mixed with those of the standard when providing a description of the standard. Thus, constructions like (4) should not be taken into account when analysing te in Standard Dutch, where (4) is quite impossible: (4) Zai begon te toavel schoon maken (Groningen dialect) *Zij begon te tafel schoon maken (Standard Dutch) She began to table clean make ‘She began to clean the table.’ (Zwart, 1993: 103) Zwart’s other piece of evidence for the syntactic independence of to comes from coordination structures. He notes on p. 104 that “one re suffices for two coordinated bare infinitives’* given certain syntactic conditions, and gives the following examples, (5a) (L.A. is een mooie stad) om in te leven en (te) sterven (L.A. is a beautiful city) for in to live and (to) die ‘(L.A. is a beautiful city to live and die in.’ (5b) . . . om boeken te kopen en (te) lezen . . . for books to buy and (to) read ‘ . . . in order to buy and read books’ I have checked these examples with a good number of native speakers of Dutch (none of them linguists) of different ages and different sex, and so far they all agree that (5a) without te may be marginally acceptable (they would never say it themselves!), but that (5b) sounds very strange as it stands. For all informants the repetition of te makes the sentences totally acceptable. It seems to be the case that a second bare infinitive works marginally only if the two infinitives can be interpreted as a simultaneous activity, or as an idiomatic unity (‘leven en sterven’ would be an example of the latter). Note that (6a) is judged as completely impossible, whereas (6b) seems to be slightly more acceptable, and even more so (6c), where gaan en man is an idiomatic phrase: (6a) Hij ging naar de winkel om &ken te kopen en lezen He went to the shop for books to buy and read ‘He went to the shop to buy books and read them.’

0. Fischer I Lingua 99 (1996) 107433

(6b) Hij He ‘He (6~) Hij He ‘He

113

wil veel geld verdienen om boeken te kopen en lezen for books to buy and read wants much money to earn wants to earn a lot of money to buy books and read them.’ besloot te gaan en staan waar hij Wilde decided to go and stand where he wanted decided to go wherever and do whatever he wanted.’

In (6a) the two infinitives are to be interpreted as consecutive because one does not normally buy books in a shop and read them there as well. This makes (6b) perhaps more acceptable because here the two activities can be pragmatically interpreted as a unity with respect to the matrix verb. We now turn to Old English, where similar differences may be observed, as far as this is possible without native speakers. Concerning coordination facts in Old English, it has to be noticed first of all that the use of a second bare infinitive in coordination with a to-infinitive (i.e. type (lb) given above) is extremely rare,6 as rare in fact as the construction with fo left out but with the dative infinitival ending (-(e)nne) expressed, *to furenne and done geseonne, which according to TK (p. 97) should not occur. I have checked the complete Old English section of the Helsinki Corpus,’ and found 61 instances of two coordinated to-infinitives (some examples are given in (7)), 1 example of a coordinate infinitive without to but with the inflection -enne (8), and 1 example of a coordinate bare infinitive (9). Of the instances where two or more coordinated to-infinitives are used, 45 occur in positions where a to-infinitive is obligatory (i.e. after nouns, adjectives, etc.), 15 after a main verb, many of which can take to- as well as bare infinitives.

6 Callaway (1913) has found very few examples of coordinated bare infinitives in positions (i.e. after adjectives and nouns) where the to-infinitive is to all intents and purposes obligatory. After adjectives he has found three instances (pp. 1X)-151), given below in (1 la to c), and after nouns two (p. 174), one of which is the one quoted by TK (see (17b) below). One more example after a noun was discovered by Visser ($967), see (1 Id) below (in my opinion this is not a noun but a verb; it is not listed as a noun in any of the Old English dictionaries, but cf. Mitchell $292). It is to be noted that the two single bare infinitives (so not as part of a coordinate structure) mentioned by Callaway, that occur after a noun, helle asendan Be anweald hcf8, se%3an he ofslyh& on into hell send fear (him) who power has, when he kills, ‘fear the one who has the power, when he kills, to send you into hell’ (L.12.5, Call. p. 173) myne fleon fealone stream @I wits him ut was them away intent flee dark StreiUll ‘their inclination was to flee away from the dark stream” (Andreas, Call. p. 173)

(a) adraeda6 bone

both indicate the content of the noun rather than some purpose, which presumably accounts for the quite exceptional use of a bare infinitive here (see also the discussion below). Moreover, one is used in a gloss and the other is from poetry, which also makes them-suspicious as examples of current Old English. ’ For more information on this historical corpus see Kytij (1991). 02, 03, 04 at the head of the reference indicate the period from which it is taken, i.e. 850-950, 950-1050 and 1050-l 150 respectively. More information on the editions used can be found in Kytii.

0. Fischer I Lingua 99 (1996) 107433

114

(7a) Wib eagena sare, haran lifer gesoden ys god on wine to Against of-eyes sore, of-hare liver boiled is good in wine to drincenne, & mid pam brope Ba eagan to bepianne drink, and with the broth the eyes to bathe ‘Against eye-sore, a boiled hare’s liver is good to drink in wine, and to bathe the eyes with the broth.’ (02/3 Quadrupedibus de Vriend, 1972: 27, also de Vriend, 1984: 252) (7b) Witodlice swa oft swa we ymbe o&a manna neode hogiab. we as often as we about of-other men need think, we truly geefenl~ca8 mar-than. and swa oft swa we to godes huse gaB imitate Martha, and as often as we to God’s house go his lof to gehyrenne. and us to gebiddene we geefenlaxab his praise to hear, and ourselves to pray we imitate marian. Mary. ‘Truly, as often as we think about the needs of other men, we imitate Martha, and as often as we go to God’s house to hear him being praised and to say our prayers, we imitate Mary.’ (03 A3fr. Horn. II, 29 Pope, 1968: 258) (7c) & on bone ilcan deaZI hie wilniaB eal moncynn to forspananne & and to the same death they wish all mankind to allure and to forlledanne

to for-lead. ‘and to that same death they want to allure all mankind and to lead them into destruction.’ (02 Cura Past., Sweet, 1871: 249) (7d) swa swa se laxe, Bonne he bam siocan ne truwa21, & wenb Baet just as the leech, when he the sick not trusts, and thinks that his gehelpan ne mange, bonne [aliefB] he him eal baet &et hine his help not can, then allows he him all that that him lystr to donne & to aycganne . . , pleases to do and to eat . . . ‘just as the doctor, when he is not confident about the sick man, and thinks that he cannot help him, allows him all that he likes to do and to eat . . .’ (02 Cura Past., Sweet, 1871: 391) (8) ueni enim separare hominem aduersus patrem suum et filiam aduersus matrem suam, ic cwom forbon to delanne vef sceadenne monnu wi8 fazder his 8z dohter wib moder hire, ‘I came therefore to part or separate man from his father and daughter from her mother,’ (03 Rushworth Gosp., Skeat, 1871-87: 89) (9) Georius ic eom gehaten, and ic hazbbe ealdordom on minum earde, George I am called, and I have earldom country, in my be is gehaten cappadocia, and me bet licatr to forhtenne nu which is called Cappadocia, and me better likes to leave now pisne hwilwendlican wurBmynt, and @es wuldorfullan Godes this transitory honour, and of-the glorious God

0. Fischer / Lingua 99 (1996) 107433

115

cynedome gehyrsumian on haligre drohtnunge. kingdom obey in holy way-of-life ‘I am called George, and I have a noble state in my country, which is called Cappadocia, and it pleases me better to abandon now this transitory honour, and obey the kingdom of the glorious God in a holy way of life.’ (03 ,Elfr, Lives of Saints 14, Skeat, 1881-1900: 308) As far as (8) is concerned, it is clear that a second inflected infinitive without to is exceptional. (8) comes from a gloss, and in this example, as can be seen from the use of vel, the second infinitive is merely a synonym of the first (the Latin text only has one infinitive, separare). The example TK gives of this construction (note 2, p. 97) is rather similar in that also two synonyms are involved, to cydenne and secgenne ‘to make known and say’, translating just one Latin infinitive referre ‘report’ (see Plummer, 1896: 276). Both Visser (1963-73: $967) and Mitchell (1985: $5929, 935, 956) give examples of a bare second infinitive, following a to-infinitive, but most of these examples depend on matrix predicates which in Old English can take both to and zero (exceptions will be discussed below). Yet it is of interest to look at some of these instances because they show, as we will see, that a bare infinitive in the second conjunct expresses something different than a to-infinitive. It must be stressed that this difference can only be deduced by looking at the context and the semantic content of the infinitives, showing again that semantic considerations do play a role in the acceptability of the chosen coordinate construction.

(104 Da sende he Gearaman bone biscop

. . . to gerecenne bone gedwolan then sent he Gearaman the bishop . . . to correct the error 7 heo to SoBfaestnesse geleafan eft gecegan. and her to truth of-faith again recall. ‘then he sent bishop Gearaman to correct the heresy and call it [the tribe] again to the true faith.’ (Bede, 250.18, Mitch. $956) ac us gedafenab swyZIor mid geswince to campigenne for barn but us befits more-strongly with toil to fight for the undead-licum cynincge and pe ofer-swidan.8 undeadly king and thee overpower. ‘but it befits us more to fight with toil for the immortal king and overcome you.’ (Saints’ Lives 11,30, Viss. $967)

(lob)

_ * The quotation in Visser runs as follows, ac us gedafenap swydor mid geswince eode Pam folce to mcessigenne and godesmcersunge don. I have checked this against the edition Visser has used. This line is not found there, instead the one given in (lob) is found. Visser seems to have confused two references, since the latter part of the above quote is found in, . . . corn se erce-diacon, and cwae8 baet hit tima wsere &et he into cyrcan eode Pam force to mcessigenne and godes mersunge don. ‘ . . . the archdeacon came and said that it was time that he should go into church to say mass for the people and do honour to God.’ (Elfr. Lives of Saints, Skeat, 1881-1900, 31: 917) Both (lob) and the above example illustrate the use of a second bare conjunct, however.

116

0. Fischer I Lingua 99 (1996) 107-133

(10~) He alysde us of mum feondum . . . myldheortnysse to wyrcanne . . . he delivered us from our enemies mildheartedness to work . . . and gemunan hys halgan cyPInesse and remember his holy covenant. ‘He delivered us from our enemies to perform deeds of mercy . . . and remember his holy sayings.’ (OE Gasp. Luke 1,72, Viss. $967) The examples in (10) are similar to (9) in that in all these cases the second conjunct does not express a different activity from the one in the first, or an activity subsequent to it; rather it expresses the content of the first activity, or the way through which the first activity may be achieved. (Contrast this with the examples in (7) where the second to-infinitive does indicate a separate activity.9) Thus, it is noteworthy that the second infinitives in (9) and (10) can all be translated by -ing forms (e.g. (9) . . . by obeying . . ., (10a) by calling, (lob) thus overcoming, (10~) thus remembering, etc.), which is not possible for the examples in (7) where only the toinfinitive fits. The same is true for all six instances (given in (11) and (17a,b)), in which a bare second conjunct follows an adjective or noun. Thus, in (1 la), for the saints to suffer persecution usually entails their death, and in (1 lc), the acceptance of the monastic way of life simultaneously involves the giving up of all worldly pleasures, etc. (This also links up with the way in which bare single infinitives are (by exception) found after nouns, cf. note 6.) hi bodedon gearowe waeron and for the truth which they preached ready were ehtnysse to boligenne, and deabe sweltan. persecution to suffer, and death die ‘and for the truth which they proclaimed were ready to suffer persecution and die’ (AX Horn ii.130.3, Mitch. $929) biddende mid wope wurf3e wieron for criste to &et hi praying with weeping that they worthy were for Christ to Browiganne and becuman to his halgum. suffer and come-to his saints ‘praying and weeping that they deserved to suffer for Christ and come together with his saints.’ (AZ&.L. S. 138.353, Call.: 151) ic eom gearo to gecyrenne to munuclicere drohtnunge, and woruldlice I am ready to turn to monastic way-of-life, and worldly beawas ealle forhan. practices all leave-off. ‘I am ready to turn to a monastic way of life, and forgo all worldly practices.’ (AZ8 Horn. 1.534, Call.: 150-151)

(1la> and for &zre sobfaestnesse Be

(1lb)

(llc)

9 It seems rather likely that the to-infinitive is the default case, i.e. when the second infinitive is not clearly a simultaneous part of the first, then the to-infinitive is used. As Bolinger (1968: 122-123) remarked, “It should go without saying that potential contrasts are not always operative contrasts. It is enough that the resources are there, part of the speaker’s competence, to be used when he needs them”.

0. Fischer 1 Lingua 99 (1996) 107433

(lld)

117

and se haefde gemynt mynster to arav-enne and he/who had decided [had the intention?] monastery to establish and mid munecum gesettun. and with monks settle. ‘And who had decided to establish a monastery and settle it with monks.’ @ES 6.126, Viss. $957, Mitch. $292) (for the interpretation of gemynt, see note 6)

Syntactically, the difference between a second bare infinitive and a second to-infinitive could be represented as follows: (12a) second conjunct: bare infinitive (12b) second conjunct: to-infinitive

P(redicate) [to [inf, [and inf2]]] P(redicate) [[to inf,] and [to inf,]]

where the second to-infinitive (12b) is shown to be on the same level as the first infinitive, i.e. coordinated to it, both governed by the matrix predicate, whereas the second bare infinitive (12a) is one level deeper, in the domain of the first infinitive: it is dependent on the first infinitive and not coordinated to it. In other words, and does not serve as a marker of coordination, but as a marker of simultaneity.‘O Callaway (1913: 158, 181) - and in this he is halfheartedly followed by Mitchell ($$925, 929) - suggests that it is the ‘remoteness’ of the second infinitive that may be responsible for the lack of to. This is difficult to prove, and does not seem so very likely. First of all, it has to be noted that in many instances the infinitive is not really remote, e.g. the example in note 8 and (1 la,b,d). Secondly it would be more usual for a ‘delayed’ infinitive to receive extra marking rather than less marking: Ohlander (1940-41) shows this for Middle English, and see also Jespersen (1940: 159). Visser (1963-73: $967) suggests it may have to do with ‘rhythm or metre’, but his illustrations all come from Middle English not from Old English. Above, I have suggested that there might be a semantic difference between the use of the to- or the bare infinitive. This difference links up nicely with what I have found with respect to the use of (single) bare or to-infinitives in early and late Middle English (see Fischer, 1995, 1996, forthcoming a). In these articles I indicate that the bare infinitive signals ‘directness’, i.e. it indicates the actuality of an event and the simultaneity of tense domains of matrix verb and infinitive. The to-infinitive, on the other hand, signals ‘indirectness’, i.e. a separation between the activity of the matrix verb and that of the infinitive, either in terms of time, or in terms of reality. This correlates well with the use of the bare second conjunct in (10) and (1 l), which

lo The use of and as a marker of simultaneity is not as strange as it sounds. Ramat (in press) shows, when looking at the polyfunctionality of coordinating conjunctions used in spoken language, or in languages that have only recently developed a written form - as is the case with Old English - that the Irish conjunction agus ‘and’ is typically used to convey simultaneity. He gives the following example, ThCnig st agus m&la leis he and a-bag with-him come ‘He came while carrying a bag’

0. Fischer I Lingua 99 (1996) 107433

118

function ‘directly’ in respect to the first infinitive in that they elaborate upon the content of the first infinitive, i.e. the second infinitive is entailed by the first, it is part of the same domain. In other words, it is on the basis of semantic considerations that I believe that the second infinitive should be seen as dependent on the first, and not as dependent on the matrix verb in coordination with the first infinitive.” For similar proposals with regard to the difference between to and zero, see also Mittwoch (1990) and Duffley (1992).‘* It is useful in this respect also to bring to mind that coordination reduction has long been under attack as a meaning-preserving transformation (since TK sees the “reduction of to” in the second conjunct as “optional” (p. 96), it follows that he sees no semantic difference between the two choices). Haiman (1983: 808) for instance already showed that conjunction reduction, which decreases the linguistic distance between two coordinated elements, decreases the conceptual distance between them, so that examples such as (13a,b) and (14a,b) do not have to mean the same thing, (13a) (13b) (14a) (14b)

red red We We

ribbons and white ribbons and white ribbons can do it quickly and we can do it well can do it quickly and well

Only in (13b) may the colours appear on the same ribbon. And Haiman writes about (14) that this is “A Sovjet bureaucratic joke . . . [which] asserts (a) and denies (b) without contradiction” (Haiman, 1983: SOS). We find a similar decrease in conceptual distance with coordinated PPs in Old English (I have found only one Old English example in the Helsinki Corpus), cnitton hi rapas mid re&tm anginne hire to handum and forum (15) pa then knotted they ropes with cruel design her to hands and feet ‘then, in their cruel attempt, they tied her, hands and feet, with ropes’ (03 B@-. Lives of Saints 9, Skeat, 1881-1900: 216) Compare (15) to instances where the preposition is repeated,

‘I In my work (referred to above) on the difference between the to- and bare infinitives, I have shown that the semantic directness versus indirectness (or actuality versus generality/potentiality) is also indicated in more formal ways by the presence or absence of locative phrases (see also below), of modal verbs, negatives, conditional clauses, etc. I2 Langacker (1992) is of interest here too, although he writes on the variation between to-infinitives and gerunds in modem English (I have shown, however, that modem English gerunds can be seen as a replacement of some Old and Middle English bare infinitives, see Fischer, 1995). Langacker argues against the idea that to in infinitival~constructions is used purely grammatically, i.e. with a loss of semantic content and of its prepositional status. Concerning to, he writes, “the to-marked process receives a holistic construal vis-a-vis the main clause relationship” (p. 305). that is, the to-infinitive is seen as separate from the main clause process, while -ing, according to Langacker, involves temporal overlap with the main clause process.

0. Fischer I Lingua 99 (1996) 107433

119

fo undernsunge and fd middqsange and to nonsange and for midday service and for nones, for tierce, syn to &an pry capitulas gesungen for each three chapters should be sung’ (03 Bened. Rule. Schroer, 1885-88: 42)

(16a) On

monandaege

‘On Monday,

(16b) Pa asprang his hlisa geond pa land wide, ‘Then his fame spread widely throughout those lands, and eat swilce to irlande and eat sup to franclande and also likewise to Ireland and also south to the land of the Franks’ (03 &lfr. Lives ofSaints 26, Skeat, 1881-1900: 140) In (15) the two PPs refer to one and the same concept or action (note also that we have an idiomatic expression here); in (16), however, the actions must be seen as separate or consecutive to one another. Note that the constructions of (13)-(14) and ( 15)-( 16) are different from the ones discussed in ( 12) because ( 13)-( 16) all involve parallel or symmetic coordination, all be it on different levels, whereas (12a), in contrast to (12b), involves no coordination, it is not symmetric. The point I wish to make, however, is that leaving out or not expressing the second to (or some other element as in (13)-( 14)) is not optional, but may serve a difference in meaning. 2.3. Some further evidence for the non-parallel analysis of and +V in Old English

It is noteworthy that two of the six bare second conjuncts found after adjectives and nouns involve the verb fur-an as a first infinitive (and I found one more example of this combination in Visser, see (17c)), (17a) Ic bohte aenne tun, and me is neod to farenne and bone geseon and to-me is need to go I bought a place, and it see ‘I bought a farm, and now I must go and see it’ (&v Horn. II 372.18, Call.: 174, TK p. 96) (17b) me is neod to farenne and aone sceawian ‘I must and look at it’ (Napier’s Ai’to Th. 102.35, Call.: 174) (17~) Drihten, alyfe me arest to farenne and bebyrigean minne fader ‘Lord, allow me first to go and bury father’ my (OE Gosp. Mt. 8.21, Viss. $967) Farun is a special verb, which, like other verbs of movement and rest, can be found

in a number of expressions all involving a second verb, where the two verbs together express one activity rather than two. Van der Gaaf (1934) describes the patterns for the verbs of ‘rest’ (‘lie’, ‘sit’, ‘stand’), and Mitchell (1985: $967) follows van der Gaaf’s set-up for the verbs of motion: (18) A. finite verb + and + finite verb B. finite verb + present participle

120

0. Fischer I Lingua 99 (1996) 107-133

C. finite verb + inflected infinitive D. finite verb + simple infinitive Van der Gaaf (1934: 82) writes about these constructions, “the A, B, and D constructions evidently had pretty much the same function. In all of them the second verb denoted what the agent was doing while lying, sitting or standing . . . the C construction, too, was already in Old English sometimes used to express simultaneousness of the two actions”. I quite agree with van der Gaaf about constructions B and D (for simultaneousness in the latter, see also Callaway, 1913: 194, 197). As far as C is concerned, I concur with Mitchell, who writes (1985 : $968), after he has quoted van der Gaaf’s Old English examples of this construction, “I detect an element of purpose in all of these and so remain unconvinced that the C-type necessarily implies ‘simultaneousness’, despite the Dutch and M[iddle] E[nglish] evidence adduced by van der Gaaf”. Indeed, most, if not all, of even van der Gaaf’s Middle English examples can be interpreted as expressing purpose in context. I have also checked the Chaucer Concordance and the Middle English part of the Helsinki Corpus, and here too, to as a rule expresses purpose. The presence of purpose is usually shown by the fact that the sentence contains a locative phrase, as in, (19) Whan hit was nyght to slepe I luy/Ryght ther as I was wont to done,/ And fil on slepe wonder sone (Chaucer HF 112-114, Benson, 1988: 349) (cf. van der Gaaf’s exx. on p. 89) Or an adverbial expression emphasising the manner of the standing/lying, present.

etc. is

cm She

[Criseyde], this in blak, lykinge to Troilusl Over alle thing, he stoodfor to biholde;l Ne his desir, ne wherfore he stood thus,/ He neither there made, ne word tolde; (Chaucer TC I 309-12, Benson, 1988: 477) (cf. van der Gaaf, P. 90)

Troilus’ standing still as an activity noteworthy in itself is also highlighted by the fact that he is described as walking before he stands still (1.267), and walking again afterwards (1.317). So clearly he stands still in order to look at Criseyde. I have found only one instance among van der Gaaf’s examples where the two activities really seem to be simultaneous: (21) in my barm ther lith to wepe / Thi child (Gower ConfAm. III 302, van der Gaaf p. 90) It is possible that English to some extent shared the development taking place in Dutch, where after verbs of rest the A, B, D constructions all disappear, and C is left over to express simultaneity (in contrast, after verbs of motion, the D construction is always used in Dutch): I3 I3 In Modem German, too, there has been a reduction in the number of constructions available. Here only the ‘and’ construction has survived to express simultaneity: Ich stehe und warte ‘I stand waiting’.

0. Fischer I Lingua 99 (1996) 107-133

121

(22) Ik sta te machten I stand to wait = ‘I stand waiting’ However, even if the C construction could express simultaneity in Middle English, it could have lasted only for a short time because the to-infinitive has completely disappeared again here, only A (I stood and waited) and B (I stood wuiring) have survived. It is likely that this difference in development in Dutch and German is related to the divergent ways in which the to-infinitive came to be used: the toinfinitive becoming more verbal and acquiring a tense-node in English - as indeed TK himself proposes on p. 125 - the te- and zu-infinitives in Dutch and German respectively, retaining more the original nominal value (for more details see Fischer, forthcoming b). I have checked the occurrence of the four constructions after verbs of motion and rest in one period of the Old English part of the Helsinki Corpus (i.e. covering the years 950-1050) and have found the following numbers (see Table 1). All seven instances of the C construction must be read as purposeful, while all B and D instances are clearly circumstantial (simultaneous).14 Concerning the A-type,

However, as my informants tell me, stehen has strong referential meaning still, and the expression is most regular when there is also some locative present. It is interesting to note, though, that in Middle High German, stehen could be bleached in the ‘and’ expression, as examples such as the following show, als ich stein unde denke, wl ich si hab’ gesehen ‘as I stand and think where I have seen you’ (minnes. 1,336a, 3 v.d. Hagen, Heyne et al., 1919: 1631) lasz darumb nit ab, sonder sfee und hof, so werden deine werck gulden ‘therefore do not leave,off, but remain hopeful (stand and hope), thus your works will become golden’ (Luther 10,3,231,21 Weim., Heyne et al., 1919: 1631) The Grimm’s Dictionary also gives examples with stehen followed by a bare infinitive, a zu-infinitive and a present participle. In all these cases stehen could be bleached, but with constraints as to the time of usage and as to genre, ein b&e magd vol arger list gem mit den knechten reden steht ‘an evil maiden full of mean tricks eagerly stands talking to the (male) servants’ (only in poetry and northern dialects; lauter warh. 316, Heyne et al., 1919: 1633) ir mundlein zu lachen stat stets her little-mouth always to laugh stands ‘her mouth is always laughing’ yhsfn. sp. 265.23, Heyne et al., 1919: 1633) nachdem ich . . . mein ungltick beweinende sfunde ‘after which I stood weeping over my misfortune’ (schausp. engl. komiid. 219,6, Heyne et al., 1919: 1632) The ‘and’ construction was also the usual one still in Middle Dutch (see Stoett, 1909: §$lO, 11). Stoett, moreover, shows by means of examples such as, Amand, die syn ghetiden sat ende las Amand, who his book of hours sat and read ‘Amand, who sat reading his book of hours’ that the verbal construction must be looked upon as a unit, because otherwise the object of las could not have been raised to the position before the verb sat. I4 The A, B and D types all come close to so-called serial verb constructions, where two syntactically separate verbs express one notion simultaneously. With serial verbs, however, the two verbs are in the

0. Fischer I Lingua 99 (1996) 107-133

122

instances here can be interpreted as either consecutive or simultaneous, and it is often difficult to decide which of the two it is. When there is a clear locative or manner adverbial accompanying the first verb (which by the way is never present with D, only once with B, and always (sometimes implicitly) with C), the second verb must be interpreted as consecutive, Table 1 Frequency of the constructions involving verbs of ‘motion’ and ‘rest’ in the Old English part of the Helsinki Corpus.

faran cuman gan standan sittan licgan Total

A (and)

B (pr. part)

10 17 II 14 5

2 7 5 7

C (to)

D (bare)

4 3

14 5

7

19

I

I 57

23

(23) & hie on sunde to pare and they by swimming to that bzre ea to baem eglande the river to the island ‘and they went swimming to that island’ (02/3 Alex. Letter, Rypins, 1924:

byrig foron

&

city

and swam

went

swumman ofer

aefter across along

city and swam across along the river to the 15)

When there is no such adverbial, both readings are usually possible, (24) Lazarus ure freond slzepb ac ic wyllegan Lazarus, our friend, sleeps, but I will go of slaepe out of his sleep. (03 West-Saxon Gasp. Skeat, 187 l-87 : 104)

& awreccan and awake

hyne him

Only in a few cases is it clear that simultaneity is present, see (25). I have found no examples like the Middle Dutch one mentioned in note 13, where the raising of the object of the second verb to a position before the first verb makes the simultaneity-

same tense and person, whereas in the B and D types only the first verb is inflected, and in the A type the two verbs are connected by and. It is interesting, though, that, especially in American English, the and is sometimes dropped, as in I’ll go see what happened, bringing this closer to a serial verb.

0. Fischer I Lingua 99 (1996) 107-133

123

interpretation clear. (I have found three of these for the D construction. but these are all in poetry where word order may be looser.i5) bzem neahdunum & scrafum hider (25a) Cwomon ba wyrmas of Came the worms from the near-hills and caves there to bon baet hie bret wazter drincan woldon. &don Zla wyrmas & because they that water drink would. Went the worms and scluncon wundorlice, . . . crawled strangely, , . . ‘The reptiles came down the neighbouring hills and out of the caves because they wanted to drink the water. They went along crawling strangely, . . . ’ (02/3 Alex. Letter, Rypins, 1924: 20) & henu monige gaefelhroefe & synnfulle cwomon & hfionudun mi8 h&end & (25b) leomeras his (et ecce multi puplicani et peccatores venientes discumbebant cum iesu et discipuli eius) ‘and lo, many publicans and sinful [men] came and rested/stayed with the saviour and his disciples’ (03 Rushworth Gasp. Skeat, 1871-87: 89) In (25a), the second verb stresses the manner of their going and must therefore be simultaneous with the first. In (25b), the use of a present participle in Latin (venientes) may indicate the simultaneity of the two verbs. All in all, I found 31 examples with a locative or manner adverb of type-A, and 26 examples without such adverbials. On the basis of the above facts I would suggest that the examples with farm in (17) should be interpreted like the examples in (25), i.e. with the second bare infinitive directly dominated by farm and not by the matrix verb. In other words “the reduction of to in coordinate structure” is not “optional” as TK asserts (p. 96) because the two infinitives are not coordinate to one another. This means that TK’s analysis, whereby to is generated “in a position that governs the coordinated VPs” (p. 98), so outside the VP proper (in TK’s case in the AGRP) is not necessary. Instead we have a PP where to governs the first infinitive and the first infinitive the second, just as in Modem English He wanted to go and stand in the corner, and stand is an elaboration of to go and does not depend on want. I should like to end this section with some instances I have found in the Helsinki Corpus where the first infinitive is alsofurun, but where the second infinitive has TO. In all the cases it is clear that furun and the infinitive are coordinated and have an equal relation to the matrix verb, because farm is a clearly separate activity, it has its own locative phrase.

I5 An example of this is, We purh holdne hige hlaford pinne, sunu Healfdenes, secean cwomon We through loyal thought lord thine, son of-Healfdene, seek come . . . ‘We have come seeking your lord, the son of Healfdene, out of loyal purpose’ (OX/3 Beowulf, Dobbie, 1953: 10)

124

0. Fischer I Lingua 99 (1996) 107433

(26a) on XIII1 nihte monan is god . . . on niwne hired to f&v-enne & on 14th night of-moon is good . . . to new household to go and preost to hafgiene & nunnan halig ref to anfone priest to hallow and for-nun holy vestment to receive ‘on the fourteenth night of the moon it is good . . . to go to a new house(hold) and to consecrate priests and for a nun to receive the holy vestment’ (OX/4 Prognostications: 44 (see Kto, 1991: 214) there is one more similar ex. on the same page) (26b) . . . byb god to f@ranne on oaer & wyf to onfonne lond . . . will be good to go to another country and to receive a wife to riht life for a right way of living[?] (OX/4 Prognostications: 43) (26~) similar examples with gun and standan can be found in 01 Hist. Dot. Harmer 1914: 2, 3; 02 Bede Hist. Miller, 1890-91: 436. 2.4. Conclusion To sum up, the examples in section 2.2 have shown that it is likely that to is part of the infinitival phrase, that it cannot be moved as TK maintains (as in his rule of to lowering on p. 99), since there is no evidence coming from ‘coordinated’ bare infinitives that to can be moved. I have tried to show that where there is no to before the second infinitive, it was actually never there and can also not be there, and that there is a difference in syntactic level between a to-infinitive and a ‘coordinated’ bare infinitive. In other words, the problem noted by TK, i.e. “the apparent violation of the category selection and the parallel structure condition” (p. 97) in example (lb) is not a problem because there is no parallel structure. The second and + infinitive expresses simultaneity with the first infinitive, and is dependent on it. It was further shown in section 2.3 that and +V performs a similar function in other constructions.

3. Is the syntactic interpretation of to as AGR in Old English necessary? If we decide that there is no need for a separate, higher functional node in which to put to because the evidence on which this is based is not conclusive; in other words, if we decide that to has as its domain only the first infinitive and not the second, and that to + infinitive is equivalent to a nominal PP, it automatically follows that there is no need to talk about the properties of to in the AGR Phrase. However, because the greater part of TK’s article concerns the properties of to in AGRP, and especially how these tie up nicely with certain grammatical peculiarities of Old English and the emergence of a number of new constructions in Middle English, it would be helpful to briefly show that these same properties can also be otherwise explained.

0. Fischer I Lingua 99 (1996) 107433

125

3.1. To as AGR feature First of all, it should be noted that TK produces very good arguments to show that to is neither part of the COMP phrase, nor of Tense, so that, if it needs a separate node, in fact only the AGR phrase is left as a suitable one. However, this in itself shows that the functional categorial position of to is not very strong, in that it only seems to be responsible for agreement and not for the other properties of INFL. What evidence is there that to functions as an agreement marker? TK argues that if to is an agreement marker, then the inflected infinitive (unlike the uninflected one where AGR is lexically null, and can therefore pass on the phi-features) is not compatible with complements that exhibit overt number and case inflections. The problem with the evidence one needs for this is that the facts are hard to check. Only the complements of verbs like be would be able to exhibit agreement features,i6 and the problem is that to beonne (unlike beon) does not occur with any frequency at all in Old English simply because the frequency of bare infinitives is much higher than that of to-infinitives (Callaway, 1913: 28 notes a ratio of 84% bare against 16% toinfinitives here). However, one may wonder whether the two examples that TK gives as evidence for the lack of agreement features, do not in fact show overt agreement inflections. (27a) [swa fela swa hine underfengon [PLUR], pam he forgeafl anweald them he gave] the power [as many as received him, Godes beam to beonne God’s children to be (&F Horn. 1352, TK (20b) p. 103) (27b) He tiold ungelic to bionne paem o&urn He tries to be unlike the others (Bo. 39.135.4, TK (2Oc) p. 103) As can be seen from the first part of the quotation in (27a) (not given in TK, here added in square brackets), bearn is in fact a nominative plural (neuter) and not a nominative singular as indicated by TK in his gloss. The neuter plural in Old English happens to have no visible ending when the stem-syllable was metrically long (the original -u ending was dropped in this case in primitive Old English, see Campbell, 1951: $571). Now, should one state that, because the ending is not visible for phonological reasons, therefore it is not overt? The same is true for (27b), where ungelic is a nominative masculine strong adjective, which again by phonological accident bears a zero ending. In the Helsinki Corpus, I did not find a single example where the subject complement should show overt case, but George Jack (p.c.) directed my attention to one he had found in the Toronto Concordance,

I6 Teun Hoekstra, in his comment on an earlier version of this paper, notes, and I quite agree with him, that it is not clear whether cases of predicate agreement, as in the examples with beon, are theoretically similar to other phenomena involving case and agreement. This it itself weakens the strenith of TK’s argument on this point.

126

0. Fischer I Lingua 99 (1996) 107-133

(28) AC monige sindon me su& onlice on ungelaerednesse, But many are for-me very similar in unleamedness, wilniaa Beah &ah& hi naefre leomingcnihtas meren, not-were, they-wish yet although they never disciples lareowas [NOM.PL] to beonne teachers to be ‘But there are many who seem to me very similar in want of learning, for (Cura Pasalthough they never were disciples, yet they wish to be teachers’ toral& Sweet, 187 1: 25.7-9) Mitchell also notes with respect to such predicative endings: “I have found no nouns or adjectives which do not carry an inflexion when one is appropriate” (1985 : 53767, and see also his 83772). So, although the argument concerning predicate agreement is in itself not a very strong one (see note 16), I see no reason to observe a difference between to- and bare infinitives in terms of a lexically filled or unfilled AGRP. 3.2. The absence of a syntactic subject in to-infinitives According to TK, another difference to be noted between bare and to-infinitives is that bare infinitives can take a syntactic subject whereas to-infinitives cannot. We do find, for instance, constructions in Old English like (a) They saw the child lie/lying in the manger ((22a) in TK), but not ECM constructions like (b) I believe him to be a hero. TK believes that (a) is an ECM construction and that “paradigm cases of the ECM construction in PE [Present-day English]” also occur in Old English with verbs like findun ‘find’, witun ‘know’, etc., always with a bare infinitive. In other words, TK believes that the only difference between the (a) and (b) constructions in Old English is the presence or absence of to, or, to put it differently, that I believe him be a hero, without to, is a regular construction of Old English. The reason why to cannot be present there being explained by the fact that to is itself a subject, and therefore another lexical subject is not possible. In Fischer (1989), however, I have shown in some detail that the constructions following verbs of physical perception and causatives are different from ‘believe’-type constructions (or TK’s ‘paradigm’ cases) because semantically they are quite different, and, moreover, if we equate them, we cannot explain why (b) does not also occur without to in Old English. TK gives some examples of ‘paradigm’ ECM constructions after the verbs witun andfindan (see (29), TK’s (23)), but does not note that all his examples are slavish translations from Latin, (29a) I hazbbe afandod be habban gode geferan. good comrades I have found thee have ‘I have found thee to have good comrades’ @Elf. Coll. 203, TK p. 105) (29b) God . . . we lyfaa aefre beon andwyrde answer [sic, this should be ‘present’] God we believe always be ‘We believe God always to be present’ (Interlinear Rule St. Benet 30, 10, TK p. 105)

0. Fischer I Lingua 99 (1996) 107-133

127

(29c) Dara cynna monig he wiste in Germanie wesan of-the tribes many he knew in Germany be ‘He knew many of the tribes to be in Germany’ (Bede 40821, TK p. 105) That we cannot use these as evidence of regular Old English must be clear from the fact that in the same (Latin-based) sources, we also find ECM constructions with to, which TK disallows, (30a) . . . witende for pi hi to underfonne mede gode reward good . . . knowing for this them to receive ‘ . . . knowing that for this they will receive a good reward’ (BenRGL 64.3, Fischer, 1994: 94) (30b) . . . forpon ic wat me to beonne unscadwis on swa deorwur&a unskillful in such worthy . . . because I know me to be spraeca languages (LS29[Nicholas]7, Fischer, 1994: 94) (and note also TK’s own example in note 5, p. 106, which he explains away as “truly exceptional”). TK does not quote the much more regular examples (found in ‘native’ Old English) of what he calls ECM constructions after witun and findan. I have discussed these at great length in Fischer (1989: 191-201, and also in Fischer, 1994), and shown that examples such as the following, (31a) . . . funde hiene amne be wege licgan . . . found him alone by way lie ‘ . . . found him lying by himself along the road’ (Or 3 9.12813, Fischer, 1989: 192) (31b) Ic wat eardfzestne anne standan, deafne, dumban . . . I ‘know’ firm deaf, dumb . . . . one stand, ‘If see someone stand firm, deaf, dumb . ..’ (Rid 49. 1, Fischer, 1989: 201)17 are in fact exactly like the constructions following verbs of physical perception.r8 With these they share the identity of tense domain between matrix verb and infinitive, the existence of an entailment relation, the infinitive expressing a concrete activity that can indeed be observed, the presence of a locative phrase, and the fact that the matrix verb does not only govern the accusative NP but also gives it its 8role, all of which is not the case in ‘paradigm’ ECM constructions of the type illustrated in (29) above. My conclusion, therefore, was that Old English only allowed the ‘lexical subject’ construction after physical perception verbs and causatives (just

” The original meaning of Old English witan was ‘to see’. It is a preterite-present verb, which in the preterite developed the meaning of ‘know’, from which a new infmitive meaning ‘to know’ was formed. ‘s This is also noted in Lightfoot (1991: 81). However, Lightfoot refers to Gorrell(l985) on this point, where this phenomenon is not really clearly discussed.

128

0. Fischer I Lingua 99 (1996) 107433

as in Present-day Dutch and German), and that they therefore they must be distinguished from the ‘paradigm’ cases and thus constitute a different type of structure altogether. How they should be syntactically analysed is another matter, and has been subject of debate for a long time. This is not the time to go into that here. But it clearly is too simple to state that the only difference between (a) and (b) above is the presence of to, as TK has done. 3.3. Other properties I have a problem with the status of to as an external argument (disoussed by TK on p. 108ff). To is said to represent the external O-role of the infinitive, which would explain why “to-infinitives appear to be restricted to agentive verbs” (p. 110). This is probably not correct as the examples in (32) show, (32a) ba waeteru the waters (32b) Ic ongimre I begin

begunnon to wanigenne to wane (L#Z~Hept. Gen. 8.3, Call.: 51) began to blacigenne to grow pale (&lfr. Cr. 212.7, Call.: 53)

TK notes as typical examples of non-agentive verbs (i.e. the verbs not appearing in to-infinitives) “weather verbs (.. .), verbs of time (. . .), verbs of happening (. . . growan ‘grow’, hyngriun ‘become hungry’)” (p. 110). It seems to me that among’the latter category also belong verbs like (32)‘s wuniun ‘to wane, grow less’ and bluciun ‘to grow pale’. Another point that TK makes is that to “absorb[s] accusative Case in addition to the external argument role” (p. 114) because he wishes to account for the fact that only to-infinitives may act as retro-active passives of the type These thingsi are to do ti [ACC] (absorption of accusative case is necessary in TK’s account in order to get movement of the object NP to the matrix subject position; we will see below that an alternative analysis can dispense with movement altogether). In this respect, Beukema and van der Wurff (1993) note that to also occurs with a verb like beon (their example (5)), (33) He tiolab ungelic to bionne paem o&urn He tries unlike to be the others ‘He tries to be unlike the others’ (Bo. 39.1354) and that beon cannot be assumed to have an external e-role. They conclude that “[tlo would then be an argument without a O-role, a flagrant violation of the &riterion” and also that “Kageyama cannot provide a reason for his claim that to absorbs the external O-role” (Beukema and van der Wurff, 1993: 34). For TK the later appearance of ECM constructions with a to-infinitive is entirely dependent on the change in the status of to (see p. 125), which change according to him also licensed two other new constructions, i.e. morphologically passive toinfinitives and infinitival relatives in subject relation “at about the same time” (p. 124). It is a problem, however, that the latter two constructions are encountered

0. Fischer I Lingua 99 (1996) 107433

129

much earlier (from the 12th century onwards) than the ECM constructions with to, which appear only from the end of the 14th century onwards. I have shown in some detail (in Fischer, 1988, 1992, 1994) that the rise of these ECM constructions is connected with other factors, such as the change in basic word order from SOV to SVO, the influence of Latin, and especially the development of passive infinitives after control or ‘persuade’-type verbs. The latter cleared the way for the ‘paradigm’ ECM constructions with verbs like ‘believe’. In other words, TK’s analysis of lo is not sufficient to explain the relation between the absence of lexical subject constructions, passive infinitives with to, and subject infinitival relatives in Old English, and their later ‘simultaneous’ emergence in Middle English. I have another problem with the passive interpretation of retro-active infinitives. According to TK’s proposal only the to-infinitives possess the possibility of being syntactically passive. This puts a clear demarcation between constructions like I heard a song sing (with a syntactically active bare infinitive) and retro-active ‘passive’ infinitives like These things are to do. Jespersen (1940: 221) notes that these constructions are clearly related, and I have shown (Fischer, 1991) that both are being replaced by formally passive infinitives at about the same time in the Middle English period. If bare infinitives are analysed differently from to-infinitives in this respect, this interesting coincidence becomes a mere accident. Secondly, TK believes that the construction These things are to do must be analysed as syntactically passive because of the occurrence of truly active constructions like, (34) And ure drihten is to cwe6enne . . . And our lord is to say . . . (AS. Horn. & L.S.11, TK (30); p. 113)19

Klopzig (1922) has shown, however, that these constructions occur in Old English almost exclusively in glosses, and that they must be considered foreign to the language. It is only when the passive form of the infinitive becomes the rule for retroactive passives (i.e. it is to do becoming it is to be done, thus making way for a new type of is + intransitive to-infinitive construction), that these constructions become tolerated in English, and more frequent (see also Fischer, 1991: 147ff., 174-175). As far as movement is concerned in these constructions (the reason why TK proposes the need for the absorption of accusative case), I think that there is another possible (even though not a standard one) analysis,2o namely that constructions like These things are to say developed out of original adjectival structures, These things are easy to say, where the NP was base-generated in subject position and the infini-

i9 The construction is to cwebenne occurs six times in the Toronto Corpus (the complete corpus of Old English texts); five of these are passival, i.e. of the type hwrer is to cwebenne ‘what is to/can be said’, the only active one is our (30). I owe this information to Bettelou Los. *’ I fully realise that much has been written on this-construction (e.g. van Biemsdijk, 1982; Bennis, 1990, etc.). However, this is not the place to go into a detailed discussion of this difficult problem here. At this stage, I only wish to show that another interpretation is possible, not involving NP movement, which seems to work quite well considering the putative diachronic development discussed in (35).

0. Fischer I Lingua 99 (1996) 107433

130

tive an adjunct to the adjective (see also Bock, 193 1: 206, note). In some cases, the adjective came to be more strongly associated with the verb, and thus interpreted as an adverb. The putative development would have been as follows (see Fischer, 1991: 156; NB : types A to D all occur regularly in Old English), (35) A That question is easy to understand B (It) is easy that question to understand C Easyfly) is that question to understand D That question is to understand

It is important to notice that the D-type construction is similar to A in that it occurs with similar infinitives and both contain an implicit PRO, element (it is easy for ‘people’ to understand). This is not the case with construction types superficially similar to A, such as, The house was pleasant to live in, which also occurs in Old English, but which does not appear in the B, C and D types, and where there is also no implicit PRO,,,. The stage from A to B would involve topicalisation of the adjective. From B to’ C a reanalysis could take place, whereby the adjective can come to be interpreted as an adverb. Callaway shows that some adjectival endings can be interpreted as adverbial (because the nominative adjectival inflection was often -e in the singular, and always -e in the plural, and the adverbial ending was also -e), and that indeed some adjectives were adverbs (see Callaway, 1913; 159, note 4). As soon as the adjective can be treated as an adverb, it can (easily!) be left out, as most adverbs can. Callaway’s list of predicative infinitives after beon (on p. 98ff.) also shows that these constructions often occur with adverbial phrases close in meaning to the ‘easy’-group, such as swi2Je, swydor, butan tweon, nearolice (p. lOl), or with adverbs corresponding to the adjectives brad, beter, rihtlic, mice& which are listed in his chapter on the infinitives after adjectives. This would mean that These things are (easily) to say are in fact rather atypical constructions in Old English. To my mind, they should be considered offshoots of an adjectival construction, where movement is not involved. For this reason it is perhaps not surprising that they disappear together with the other ‘passival’ infinitives in the course of the Middle English period (apart from some idiomatic expressions that still occur, such as He is to blame).

4. Conclusion:

The status of to

TK has shown in his study that to + infinitive has the appearance of a word, that to cannot be separated from the infinitive in surface structure, and that the to-infmitive occurs in coordination with other nominal PPs, so it must itself be a PP. However, a simple PP analysis is not adequate according to TK, because in coordinate constructions with a second bare infinitive, it cannot account for the symmetry of the two infinitives. TK concludes, therefore, that the two infinitives are both bare and that to is in a higher position and governs them. I have shown in section 2 that this is not correct, that this analysis does not do justice to the facts of Old English; in

0. Fischer I Lingua 99 (1996) 107433

131

fact, to governs only the first infinitive and not the second, because the two infinitives are not coordinated. In other words there is no need on these grounds for a higher (AGR) position of to, or for an analysis of to in the to-infinitive different from that of to in a nominal PP. The main reason why TK wants to to be positioned under the AGR node is because it would help to explain the simultaneous occurrence in Middle English of a number of constructions that Old English lacked. In section 3, however, I have shown that first of all the new constructions do not all occur at the same time, as TK claims, but also that there are various problems with the analysis of to that TK has posited for these constructions. It was found that the functioning of to as an AGR feature and as an external argument is not convincing when one considers additional facts of Old English. For all these reasons it seems fair to conclude - even though I have not gone into all of TK’s examples in detail, but the most important points have been discussed - that there is no real need for an AGRP yet in Old English. There remains of course the problem of how the infinitives have to be analysed in Old English. In Fischer (forthcoming b) I have tentatively suggested that the best solution may still be to see the infinitives as essentially nominal (so contra what we believed in Fischer and van der Leek, 1981), but with already some verbal features incorporated. I point there to a parallel with the late Modem English gerunds, which also developed from nominal to verbal. Van der Wurff (1993) analyses this gerund as a category that is both [+N] and [+V], and which gradually loses its [+N] feature. 21 The [+V] feature would then account for the fact that the infinitives could already take objects in the accusative case. It is also important to realise that the to-infinitive in Old English is always inflected (Callaway, 1913, only notes a few exceptions, which he sees as scribal errors), which again points to the infinitive being nominal rather than verbal. The analysis of to as a preposition and the infinitive as a nominal element, also explains why to + infinitive has the appearance of a word, and functions in combination with other nominal PPs. All in all, seeing to simply as a preposition, and the infinitive as nominal, certainly solves most, if perhaps not all, the problems noted above. 22 And I certainly believe that this analysis creates less problems than the one suggested by TK.

References Anderson, J., 1993. Parameters of syntactic change: A notional view. In: C. Jones (ed.), l-42. Bennis, H., 1990. A note on modal passives. In: J. Mascaro and M. Nespor (eds.), Grammar in progress, 3340. Dordrecht: Foris.

21 Anderson (1993: 10-16) has a similar proposal with respect to the changes taking place in the infinitive, i.e. that it involves “a minimal categorial decrement”, whereby the present-day English infinitive has acquired “a high[er] proportion of P” (P stands for ‘predicativeness’, i.e. verbalness) (p. 15). However, Anderson believes that Old English to- and bare infinitive were not nominal to the same degree, the bare infinitive showing already more verbal properties. 22 Najib Narad (personal communication) correctly pointed out to me that a nominal to-infinitive might have problems in accommodating Old English infinitival preposition stranding constructions. I would like to leave this question for further research.

132

0. Fischer I Lingua 99 (1996) 107433

Benson, L.D. (gen. ed.), 1988. The Riverside Chaucer (3rd ed.). Oxford: Oxford University Press. Beukema, F. and W. van der Wurff, 1993. Easy fo please in Old English and Modem English. Paper given at the LAGB spring meeting. Ms., English department, University of Leiden. Bock, H., 1931. Studien zum prapositionalen Infinitiv und Akkusativ mit dem TO-Infinitiv. Anglia 55, 114-249. Bolinger, D.L., 1968. Entailment and the meaning of structures. Glossa 2, 119-127. Callaway, M., 1913. The infinitive in Anglo-Saxon. Washington, DC: Carnegie Institution. Campbell, A., 1959. Old English grammar. Oxford: Clarendon Press. Chaucer Concordance. J.S.P. Tatlock, 1927. A concordance to the complete works of Geoffrey Chaucer and to the Romaunt of the Rose. Washington; DC: Carnegie Institution. Duffley, P.J., 1992. The English infinitive. London: Longman. Fischer, O.C.M., 1988. The rise of the for NP to V construction: An explanation. In: G. Nixon and J. Honey (eds.), An historic tongue. Studies in English linguistics in memory of Barbara Strang, 67-88. London: Routledge. Fischer, O.C.M., 1989. The origin and spread of the accusative and infinitive construction in English. Folia Linguistica Historica 8, 143-217. Fischer, O.C.M., 1991. The rise of the passive infinitive in English. In: D. Kastovsky (ed.), Historical English syntax, 141-188. Berlin: Mouton de Gruyter. Fischer, O.C.M., 1992. Syntactic change and borrowing: The case of the accusative-and-infinitive construction in English. In: M. Gerritsen and D. Stein (eds.), Internal and external factors in syntactic change, 17-88. Berlin: Mouton de Gruyter. Fischer, O.C.M., 1994. The fortunes of the Latin-type accusative and infinitive construction in Dutch and English compared. In: T. Swan et al. (eds.), Language change and language structure. Older Germanic languages in a comparative perspective, 91-133. Berlin: Mouton de Gruyter. Fischer, O.C.M., 1995. The distinction between bare and to-infinitival complements in late Middle English. Diachronica 12, l-30. Fischer, O.C.M., 1996. Verbal complementation in Early Middle English: How do the infinitives fit in? In: D. Britton (ed.), English historical linguistics 1994, 247-270. Amsterdam: Benjamins. Fischer, O.C.M., forthcoming a. Infinitive marking in late Middle English: Transitivity and changes in the English system of case. In: J. Fisiak (ed.), Studies in Middle English, Berlin: Mouton de Gruyter. Fischer, O.C.M., forthcoming b. The grammaticalisation of infinitival to in English compared with German and Dutch. In: R. Hickey and S. Puppel (eds.), Festschrift in honor of Jacek Fisiak’s 60th birthday. Berlin: Mouton de Gruyter. Fischer, O.C.M. and F.C. van der Leek, 1981. Optional vs radical reanalysis: Mechanisms of syntactic change. Lingua 55, 301-350. Gaaf, W. van der, 1934. The connection between verbs of rest (lie, sit and stand) and another verb, viewed historically. English Studies 16, 81-99. Gelderen, E. van, 1993. The rise of functional categories. Amsterdam: Benjamins. Gorrell, J.H., 1895. Indirect discourse in Anglo-Saxon. PMLA 10, 342-485. Haiman, J., 1983. Iconic and economic motivation. Language 59: 781-819. Haiman J., 1993. Life, the universe, and human language (a brief synopsis). Language Sciences 15, 293-322. Heyne, M., Br. Crome, H. Meyer and H. Seedorf, 1919. Deutsches Wiirterbuch von Jakob und Wilhelm Grimm. Vol. 10. Leipzig: S. Hirzel. Jespersen, O., 1940. A modem English grammar on historical principles. Part V. Copenhagen: Munksgaard. Jones, C. (ed.), 1992. Historical linguistics. Problems and perspectives. London: Longman. Joseph, B.D., 1992. Diachronic explanation: Putting speakers back into the picture. In: G.W. Davis and G.K. Iverson (eds.), Explanation in historical linguistics, 123-144. Amsterdam: Benjamins. Kageyama, T., 1992. AGR in Old English to-infinitives. Lingua 88, 91-128. Kliipzig, W., 1922. Der Ursprung der fo be to-Konstruktion. Englische Studien 56, 378-389. KytB, M., 1991. Manual to the diachronic part of the Helsinki Corpus of English texts. Coding conventions and lists of source texts. Department of English, University of Helsinki. Langacker, R.W., 1992. Prepositions as grammatical(izing) elements. Leuvense Bijdragen 81, 287-309.

0. Fischer I Lingua 99 (1996) 107433

133

Lightfoot, D.W., 1991. How to set parameters: Arguments from language change. Cambridge, MA: MIT Press. Mitchell, B., 1985. Old English syntax. Vols. I-II. Oxford: Clarendon Press. Mittwoch, A., 1990. On the distribution of bare infinitive complements in English. Journal of Linguistics 26, 103-131. Ohlander, U., 1941. A study on the use of the infinitive sign in Middle English. Studia Neophilologica 14, 58-66. Plummer, C., 1896. Venerabilis Baedae. Opera historica. Oxford: Oxford University Press. Pullum, G.K., 1982. Syncategorematicity and English infinitival to. Glossa 16, 181-215. Ramat, P., in press. Typological comparison and area1 linguistics. In: J. van der Auwera and D.P. G Baoill (eds.), Adverbial constructions in the languages of Europe. Berlin: Mouton de Gruyter. Riemsdijk, H. van, 1982. A note on case absorption, Wiener Linguistische Gazette 27/28: 71-82. Stoett, F.A., 1909 (2nd ed.). Middelnederlandsche spraakkunst. Syntaxis. The Hague: Nijhoff. Sweet, H., 1871. King Alfred’s West-Saxon version of Gregory’s Pastoral Care. EETS OS 45. Oxford: Oxford University Press. Toronto Concordance. A. DiPaolo and R.L. Venezky, 1980. A microfiche concordance to Old English. Toronto: Pontifical Institute of Medieval Studies. Visser, F.Th., 1963-73. An historical syntax of the English language. Vols. I-IIIb. Leiden: Brill. Vriend, de H.J., 1984. The Old English Herbarium and Medicina de Quadrupedibus. EETS OS 286. Oxford: Oxford University Press. Wurff, W. van der, 1993. Gerunds and their objects in the Modem English period. In: J. van Marle (ed.), Historical Linguistics 1991. Papers from the 10th International Conference on Historical Linguistics, 363-375. Amsterdam: Benjamins. Wyngaerd, G. vanden, 1988. Passive and the analysis of auxiliary verbs. In: P. Coopmans and A. Hulk (eds.), Linguistics in the Netherlands 1988, 159-168. Dordrecht: Foris. Zwart, J.W., 1993. Dutch syntax. A minimalist appraoch, Ph.D. dissertation, University of Groningen.