INFLECTIONAL MORPHOLOGY AND THE LOSS OF V TO T RAISING IN ENGLISH REBECCA TOLLAN University of York Abstract Whilst it is commonly recognised that languages with rich inflectional morphology also exhibit verb raising, the reason why such a correlation exists remains largely unexplained. In my paper I consider proposals by Rohrbacher (1994) and Alexiadou and Fanselow (2002) in an attempt to account for how lexical verb raising came to be lost from Early Modern English following erosion of inflectional morphology several centuries earlier. I conclude that, as far as English is concerned, the connection between inflectional morphology and verb raising is one which is governed by the nature of language use, first language acquisition and general syntactic change. The loss of inflectional morphology set in motion a series of changes, with loss of verb raising as the end result. The analysis presented here appeals to biological (UG) endowment of Roberts’ (1985) syntactic principle of V-Visibility; as such I also present an evolutionary explanation for this supposed UG status. 1. Introduction This paper addresses the causal relationship between the erosion of Middle English subjectverb agreement morphology following the Viking and Norman invasions of Britain (circa 865 and 1066 CE respectively) and the loss of lexical V to T raising in Early Modern English. Two potential models are considered: firstly, a re-setting of a verb raising parameter which resulted from a Universal Grammar (UG) constraint on the licensing of verb movement by the morphological richness of the subject-verb inflectional paradigm (Rohrbacher 1994) and secondly, a lengthy diachronic process in which loss of inflections sets in motion a series of changes, of which loss of verb raising is the end result (Alexiadou & Fanselow 2002). Following a corpus analysis I argue predominantly in favour of the second model. The latter sections of this paper are dedicated to determining exactly what kinds of changes ensued from loss of verb inflection and how these eventually led to loss of verb raising. Whilst an account involving a series of changes is more convincing in light of the data considered, it is noted nonetheless that there is no evidence which necessarily rules out Rohrbacher’s model. The paper is organised as follows: section 2 addresses problems for Rich Agreement Hypothesis as proposed by Rohrbacher and the counter-argument of Alexiadou & Fanselow, section 3 is dedicated to analyses from my own data search, and sections 4, 5 and 6 lay out a plausible sequence of changes in English which resulted in loss of verb raising. I argue that the loss of V-to-T raising was a consequence of several factors; these are summarized in section 7. York Papers in Linguistics Series 2 © The Author, 2012 ISSN 1758-0315 Issue 12a 81 Inflectional Morphology and The Loss of V to T Raising in English 1.1 Assumptions I assume a cyclic approach for head movement which is governed by the Head Movement Constraint, as worded in (1). (1) The head movement constraint Movement from one head position to another is only possible between a given head and the next highest head in the structure. (Radford 2004, cited by Leppänen 2010:4) It will therefore be assumed that any verb in [head, CP] must have transited though [head, TP]. Where a negation head is present, the verb must raise though this head in order to reach TP. Where reference is made to Universal Grammar (UG), I define UG as information which is both specifically linguistic and biologically endowed at birth. 1.2 General remarks on verb raising A verb may either remain in its base position (henceforth V in situ), or raise from this position to [head, TP] (henceforth V-to-T). In Romance and most Germanic languages, verbs raise from V to T and hence precede negation and any VP-adjoined adjuncts (figure 1). In other languages verbs remain in situ and must follow negation and/or VP adjuncts (figure 2). TP TP T’ T T + V T’ NEGP T VP Neg Adv Neg VP < V > Figure 1: V-to-T NEGP Adv VP VP V Figure 2: V in situ In a V in situ language, I assume that a ‘tense-lowering’ operation must take place, whereby any agreement affixes (generated in T) adjoin to the verb-stem at LF. This is in accordance with the Stray Affix Filter (Lasnik 1995) which prohibits a sentence from containing stray affixes at spell-out. Due to locality restrictions however, tense lowering applies only if the affix is lowered to the head directly beneath T, phrased by Santorini & Kroch (2007) as in (2), Rebecca Tollan 82 (2) When a head A lowers onto a head B, A and B must be in a local relation in the sense that no projection of a head distinct from A and B intervenes on the path of branches that connects A and B. (cited by Leppänen (2010:4)) In other words, the presence of NEG between V and T prohibits the tense lowering operation. Modern English uses ‘do-support’ as a last-resort strategy, whereby ‘do’ is the spell-out of T when T cannot be spelt out on the verb; this occurs in questions and in the presence of negation, cf. (3) and (4), (3) (4) Do you like sprouts? I do not like sprouts In English, only lexical verbs remain in situ; auxiliaries and the copula ‘be’ raise to T and can host inflection in questions and negative, such that do-support is not required in these instances. By contrast, all lexical verbs in Old and early forms of Middle English1 underwent raising to T. There is disagreement about exactly when the loss V-to-T was complete. As discussed in Vagra (2005), Kroch (1989) suggests the end of the sixteenth century, Lightfoot (1997) the end of the seventeenth century and Warner (1997) sometime later than that (during the eighteenth and nineteenth centuries). Very broadly speaking then, it can be assumed that generalisation of V in situ occurred during the Early Modern and beginning of the modern periods. 1.3 V-to-T and inflectional morphology: the Rich Agreement Hypothesis Rohrbacher’s (1994) Rich Agreement hypothesis (RAH) claims that the richness of morphological agreement in a language is reflected in its verb raising requirements. It is defined as in (5), and taken by Rohrbacher to be a UG principle. (5) The Rich Agreement Hypothesis A language has V to T raising if its regular verbs distinguish the forms for first and second person in at least one number of one tense from each other, as well as the forms for third person in that tense/number combination and from the infinitive. (Rohrbacher 1994:116-7) This definition implies that first language (L1) learners must have access to the notion of the ‘full’ inflectional paradigm in order to determine ‘richness’; as such, the loss of V-to-T is the result of a UG constraint, namely the loss of inflectional morphology. Before lexical V-to-T was lost in English, inflectional morphology had been eroded such that the once ‘rich’ agreement paradigm for lexical verbs (which now only distinctively mark third singular 1 I use the following dates to define relevant historical periods: Old English (OE) – c.500-1100 CE Middle English (ME) – c.1100-1500 CE Early Modern English (EModE) – c.1500-1700 CE Modern English (ModE) – c.1700 CE-present. 83 Inflectional Morphology and The Loss of V to T Raising in English present) no longer fulfilled the criteria in (5). Drawing upon Chomsky’s (1986) claim that learners distinguish between ‘core language’ (a system determined by the fixing of parametric values of UG) and ‘periphery’ (general cognitive processes such as observations of word order), Rohrbacher further claims that the acquired agreement paradigm is the only specifically linguistic determinant of the position of the verb available to L1 learners. He also suggests that, because direct observations of word order patterns influence usage of language by a new generation of speakers, ‘remnants of V to T raising [...will remain...] long after the morphological (core) UG trigger’ (p.160). 2. Problems with the RAH for explaining diachronic change The RAH is not adequate when it comes to explaining cross-linguistic differences. During roughly the same period of time (i.e. the thirteenth and fourteenth centuries), inflectional morphology was largely eroded both in English and in Swedish, such that both languages no longer fulfilled the richness criteria described in (5). Rohrbacher considers the subsequent loss of V-to-T in both English and in Swedish embedded clauses as evidence in favour his claim that raising is determined by ‘morphological richness’. However, a crucial question remains unanswered: why was V-to-T not lost in Swedish matrix clauses? Whilst modern Swedish embedded verbs (both lexical verbs and auxiliaries) remain in situ, matrix verbs still raise. Similarly the modern English auxiliary ‘have’, which distinctively marks only third singular present (‘has’) also still undergoes raising. I do not intend to address these questions in this study, but they are important in the sense that they show that the RAH alone cannot account for loss of V-to-T. Nonetheless, the observation that no languages with rich morphology (as yet known) exhibit V in situ does suggest that some form of relationship between inflectional richness and verb raising exists. Note also that the behaviour of Swedish matrix verbs and English ‘have’ does not refute the RAH, since the connection between rich morphology and verb raising is not assumed to be bi-conditional. The RAH claims only that V must raise to T in languages with rich agreement and does not predict that languages with poor inflectional morphology cannot exhibit V-to-T. 2.1 The time lag Whilst the erosion of agreement morphology in Middle English was complete by around 1400 (Leppänen 2010), lexical V-to-T did not completely disappear until the late EModE/ModE periods (about three centuries later). Rohrbacher predicts that remnants of Vto-T may remain ‘long after’ erosion of agreement inflections but does not specify any time period. Indeed, the time lag observed in English, coupled with the preservation of matrix Vto-T in Swedish despite loss of inflection strongly suggests that some kind of ‘intermediate operation’ took place in English but not in Swedish (or at least, not as far as matrix clauses were affected) which both resulted from loss of agreement morphology and resulted in loss of verb raising. I return to this idea in section 3.3. 2.2 Arguments against the RAH In their 2002 paper, Alexiadou & Fanselow (A&F) argue that the RAH is empirically inadequate as a UG principle and that V-to-T raising is independent of morphology. Their claims are based upon several observations about cross-linguistic syntactic movement. The key problems with the RAH as identified by Alexiadou & Fanselow are given in (6). Rebecca Tollan (6) 84 Empirical inadequacies of the RAH according to A&F a. There are languages in which V-to-T is optional b. Presence or absence of Verb-second (V2) does not predict if morphology is rich. c. An approach relying upon the notion of ‘full paradigm’ cannot explain the distinction between main verbs and auxiliaries (especially in English where only the latter may raise). d. If verb movement is triggered by morphology, it should be expected that other types of syntactic movement are dependent upon morphology. The presence or absence cross-linguistically of the types of movement in (d1-3) cannot be predicted by the richness of the relevant morphology: (d1) N movement within DP (d2) Weak versus strong EPP (d3) Types of A-bar movement (wh-movement and topic/focus fronting) According to A&F, The supposed connection between morphology and raising cannot therefore be a UG principle, otherwise such a principle would be reflected in all types of syntactic movement. Despite the remarks in (6), A&F concede, in light of the observation that languages with rich subject-verb agreement also exhibit V-to-T, that a connection between morphology and verb raising does exist. They claim however that verb movement is only indirectly influenced by morphology and occurs due to a natural diachronic process in which the string ‘verb+clitic’ is re-interpreted over a period of several generations as ‘verb+AGR’ (p.13-15), such that rich suffixal agreement does not arise in a language without verb movement. This analysis, which I will not discuss in any more detail here, says nothing about a process in which verb movement is lost. However, A&F’s claim in the light of their observations in (6) that the connection between rich morphology and verb movement is due to natural diachronic development (rather than a UG constraint) is certainly worth investigating insofar as it may relate to changes in the history of English. Regardless of whether V-to-T in English was lost as a result of UG or natural diachronic changes, the burning question is this: what happened in the centuries between the erosion of inflection and the loss of verb raising? Only after determining this can we have a clearer idea of whether a UG constraint, rich agreement, or laws of diachronic change are responsible. 3. Corpus analysis and discussion A logical first step in answering the aforementioned question is to verify empirically, in terms of individual texts (and their respective authors), whether presence or absence of subject-verb agreement had any immediate impact on verb raising. A corpus search was conducted for four East Midlands Middle English texts from the Penn-Helsinki Parsed Corpus of Middle English (second edition) (Kroch & Taylor 2000). 3.1 Methodology The aim of this analysis was to search for verbs which could be unambiguously determined as having raised. Based upon an inflectional paradigm for East Midlands ME (see (9)), the numbers of these verbs both with and without full subject-verb agreement were recorded. A second search was run to find verbs remaining in situ. Only matrix lexical verbs were 85 Inflectional Morphology and The Loss of V to T Raising in English considered since a) this study is concerned primarily with behaviour of lexical verbs and b) differing word order in ME embedded clauses may obscure analysis. The four texts selected are listed in (7). Assuming 1400 as an approximate date by which loss of inflectional morphology were complete, two texts pre-dating 1400 and two post-1400 (written toward the end of the ME period) were chosen to allow for maximal comparison. It is predicted that uninflected verbs should raise, particularly in the earlier two texts, because the changes in inflectional morphology were so recent (if complete) at this point in time. (7) Texts used in the data search The Earliest Complete English Prose Psalter (c.1350) (Bülbring 1891) The Parson’s Tale (c.1390) (Benson 1987) Claxton’s History of Reynard the Fox (c.1481) (Blake 1970) Richard Fitzjames’ Sermo die Lune (c.1495) (Jenkinson 1907) In order to identify raised verbs, I used Vikner’s (1995) diagnostic for V-to-T movement as given in (8). CP-V2 constructions were also counted as instances of raised verbs. (8) ‘A verb has raised to T if it precedes a medial adverb (taken to be adjoined to the VP) or a negation’. Verbs were classified as inflected or uninflected according to the paradigm in (9) for East Midlands ME, (9) Inflectional Paradigm for East Midlands Middle English (adapted from Burrow & Turville 2005) Present tense Past Tense Person 1 2 3 1 2 3 Singular -e -est -eth -e -est -e Plural -en -en -en -en -en -en 3.2 Results The results of the search for uninflected verbs are shown in the table in (10). All verbs considered are listed in the appendix. Rebecca Tollan (10) 86 Results for uninflected verbs Text The Earliest Complete English Prose Psalter The Parson’s Tale Claxton’s History of Reynard the Fox Richard Fitzjames’ Sermo die Lune Year Total uninflected verbs Uninflected verbs raised Uninflected verbs in situ c. 1350 55 n 55 % 100 n 0 % 0 c.1390 8 8 100 0 0 c. 1481 10 10 100 0 0 c.1495 5 1 20 4 80 In all texts except one, uninflected verbs are never in situ. Two in situ verbs were identified in the two earlier texts; ironically though, both were inflected (third plural roteden in ‘The Earliest Complete English Prose Psalter’ and third singular retorneth in ‘The Parson’s Tale’). 3.3 Conclusion and resulting considerations The corpus search confirms predictions that loss of inflection in Middle English had no immediate impact on verb raising. Only in ‘Sermo die Lune’ (c.1495) does a change appear visible, with the majority of uninflected verbs remaining in situ, although little can be concluded on the basis of five tokens alone. Having confirmed the presence of a time lag between loss of inflection and loss of V-to-T, I turn now to the idea that loss of inflection led to a chain of diachronic changes, or at the very least, some kind of ‘intermediate stage’ (see section 3.3) which resulted in loss of V-to-T. Following Lightfoot’s (1995) suggestion that the rise of do-support in English constituted evidence against verb raising, Lappänen (2010) hypothesises that rise in ‘do-support’ constructions constituted such an ‘intermediate’ stage in the transition of English from a verb-raising to a non-verb-raising language. If we now consider Ellegård’s (1953) graph (figure 3) showing frequency of do-support in different sentence constructions (based on a sample of over 10,000 tokens), it appears that do-support began to become productive during the sixteenth century (i.e. sometime between the erosion of inflections and loss of V-to-T). This fits with Lightfoot and Lappänen’s suggestions. 87 Inflectional Morphology and The Loss of V to T Raising in English Figure 3: Percentages of do-support use in different sentence types (Ellegård 1953, adapted by Hans & Kroch 2000:2) The patterns of do-support use found by Ellegård raise several further questions: firstly, why did do-support not develop at the same rate across all sentence types? In particular, why should it have developed later in negative declarative and imperatives than in affirmatives and negative questions? Secondly, was the use of do the cause or the effect of loss of V-to-T? Lightfoot seems to view do-support as the cause of the decline in verb raising. In the following sections, I will argue that do-support was both a cause and an effect of the decline in raising. Even so, whichever view is adopted, the most important question perhaps is this: how did loss of inflection lead to productivity in do-support (if at all)? Lightfoot’s and Leppänen’s analyses are too simplistic in this respect. Somehow, loss of inflections must have led to either a) something which triggered do-support (Lightfoot), b) evidence against V-to-T which subsequently triggered do-support as an alternative to verb raising, or c) something which resulted in some form of preliminary evidence against V-to-T, which led to adoption of do-support, which constituted yet more evidence in favour of V in situ. In sections 4 and 5 I present a model which is compatible with the latter option. 4. Consequences of loss of morphology: The Subjunctive, Modals and V-Visibility Roberts (1985) notes that the loss of agreement inflections had a direct impact on the use of the subjunctive; because loss of agreement meant that little more than the verb stem remained, distinctions could no longer be made between indicative and subjunctive forms. In order to compensate for this, periphrastic constructions consisting of ‘modal+infinitive’ were used to express subjunctive mood and gradually grew in frequency during the late ME and EModE periods. According to Roberts, it was possible for modals to functionally substitute for the subjunctive because they already expressed ‘general notions of mood’. This Rebecca Tollan 88 development led to an important change in the categorization of modal verbs as a result of what I argue is a fundamental UG constraint. I turn now to addressing the behaviour of modals in more detail. 4.1 Changes in the syntax of modals In modern English, modals are generated in T. They do not take arguments, assign thetaroles, or have an infinitival form. In Middle English however, modals subcategorised for VP. They appeared in a position akin to that that which is occupied by causative and perception verbs, often referred to in syntactic literature as ‘little vP’. The reason for this was that, in Middle English, contrary to Modern English, models had arguments, albeit non-referential arguments (Roberts 1985). Figures 4 and 5 show the positions of modal verbs in Middle and Modern English. TP T’ T vP MODAL AKA VP V Figure 4: modal verbs in Middle English (prior to re-analysis) TP T’ MODAL vP v VP V Figure 5: modal verbs in Modern English (post re-analysis) Even before the re-analysis, modals were somewhat set apart as marked subclass of verbs. Besides having non-referential arguments, they also had irregular agreement inflections, such as lacking third singular agreement in the present tense (Roberts 1985). These idiosyncrasies meant that modals were particularly prone to further change. 89 Inflectional Morphology and The Loss of V to T Raising in English 4.2 V-Visibility Why did modals change from subcategorising for VP to being base-generated in [head, TP]? The answer relates to their newly-developed use in late ME and EModE as periphrastic substitutes for the subjunctive. As they were more commonly used to express (subjunctive) mood, modal verbs lost their properties which had once allowed for them to be categorized alongside lexical verbs. Lightfoot (1979) recognises the differences between Middle and Modern English modals given in italics in (11), which I illustrate by comparing English modal sentences to those of German and Dutch, whose modal forms subcategorize for VP and are therefore syntactically akin to those of Middle English . (11) a. Modals have lost their ability to take a direct object. E.g. Dutch Hij kan het goed English *He can it well = He can do it well b. Modals cannot occur in clusters E.g. Dutch Ik zal moeten German Ich werde müssen English *I will must c. Modals can no longer take past tense morphology E.g. Dutch Ik heb gemoeten German Ich habe gemusst English *I have musted (Note that the Dutch and German perfect tense prefixes ge- are roughly equivalent to the English suffix -ed ) d. Modals cannot take infinitive marking. Whilst modal infinitives of Old English and modern western Germanic languages (aside from English) have an –en suffix, the current infinitive marker ‘to’ can never precede a modal verb in modern English. E.g. Dutch moeten German müssen English *to must These changes took place during the EModE period (Roberts). They are believed to be a consequence of modal verbs ceasing to be able to assign theta roles, (as best exemplified in (11a)) and hence no longer being able to subcategorise for VP. This change to base generation of modals in T was driven by the UG principle of V-Visibility, as worded by Roberts in (12). Like Roberts, I assume this principle to be a part of UG. (12) V-Visibility ‘V assigns theta-roles if and only if V is governed...If a verb has no theta-roles to assign then it must appear in an ungoverned position. Verbs with no theta roles to assign have a radically different distribution compared to other verbs’ (Roberts 1985) I will not discuss in any detail the rules of government, but for the purposes of the issue in hand, it suffices to say that V and little v are governed positions (governed by either little v or T), whilst T is an ungoverned position. As such, once modal verbs ceased to assign theta roles, V-Visibility would have forced them to appear in T, and hence a new subclass of verbs is born. As a result, the construction ‘modal+lexical verb’ (as both a substitute for the subjunctive and in other modality contexts) would obscure the evidence in favour of V-to-T by providing L1 learners with construction in which T is occupied (by the modal) and as such Rebecca Tollan 90 the verb does not raise. There would subsequently be fewer instances of constructions which require V-to-T available to L1 learners. Because V-Visibility is assumed here as given by UG, the change in modals from subcategorizing for V to being generated in T would have occurred as soon as they ceased to assign theta-roles: L1 learners, already endowed with the V-Visibility principle, would not have had to wait for any direct evidence as such in incoming linguistic data. 4.2.1 V-Visibility: why UG? Although Roberts also regards (12) as a principle of UG, no explanation is as yet offered as to why this should be the case. Since such a core part of my argumentation will be reliant upon biological endowment of V-Visibility, I will now offer a suggestion as to why (10) is given by UG, based upon an evolutionary model of syntax involving lexical protolanguage. In doing so, I draw upon claims made by the evolutionary linguist Derek Bickerton (1998) to argue that the different syntactic treatments of theta-assigning (lexical) verbs and non-theta assigning verbs (e.g. modern English modals) result from the two different ‘steps’ in the evolution of modern language to which these two verb classes belong. Forms of complex language are widely believed to have first come into being in the human species around 50,000 years ago and Bickerton regards ‘Protolanguage’ as an intermediate evolutionary step between primitive forms of communication (akin to other primates) and full-fledged modern language. This protolanguage would have constituted the earliest form of human communication which made use of some very basic form of syntax. This, according to Bickerton, would have consisted of propositional-only communication (typically ‘subject+verb+object’ with no formal functional elements whatsoever). Lexical protolanguage is considered as the ‘fall-back position’ for modern humans who are incapable of complex language, for instance aphasic patients and ‘feral’ children who have not been exposed to a first language until relatively late in life. Whilst it is not clear exactly what prompted the transition from protolanguage to modern language in our pre-history (possibly an increase in conceptual power), what is important as far as my argument here is concerned is that our knowledge of thematic roles and information contained within a vP can be claimed to have a deep-rooted ancestry. Here, it could be claimed, is where the V-Visibility constraint first arose: all verbs in syntactic utterances in the ‘protolanguage period’ were those which assigned theta roles to other elements in the sentence. As such, they can be analysed as vPs lacking any higher functional material. To quote Bickerton (p.348), ‘the most basic unit of syntax is the clause...the scope of a clause is determined by the thematic grid of the single verb that each clause obligatorily contains’. The most logical first step in the transition from protolanguage modern language (which probably took place due to a need for a more sophisticated level of communication) would have been the introduction of functional categories such as TP. Functional categories contain all non-theta-related information and as such are distinct from lexical categories. Some modern evidence for the lexical-functional distinction can be seen in ‘pidgin’ languages, developed as a ‘lingua franca’ by adult speakers of different native languages as a basic form of communication. Notably, pidgins have very few functional items, and contain mostly thematic, propositional information only. Functional items, such as inflectional morphology and auxiliaries, appear only when a new generation learns the pidgin as a first language, and as such a new creole is born. Given that ability in language acquisition has been shown to wane significantly with age, it is interesting to note that adult pidgin speakers are nonetheless able to develop a form of communication which contains lexical, thematic information (akin 91 Inflectional Morphology and The Loss of V to T Raising in English to that which is hypothesised to have been found in protolanguage), but very little of functional value. Instead, functional information can only be injected into the language by virtue of a new generation of speakers, who are endowed with full linguistic competence, as opposed to that which was available to their parents when the pidgin was first developed. This, we could reasonably claim, is modern evidence to reflect the phylogenetic development of syntax: thematic information has a greater and more robust inherent ancestry in our species than does functional information, and is as such more easily accessible to humans in times of linguistic hardship. What it is most fundamental about the evolutionary scenario I have presented here is that lexical (VP) elements and functional elements belong to radically different stages in the evolution of syntax, the former having evolved first. As a result, the type of information which can (and cannot) be contained within a VP is especially hard-wired in the human brain. By the same evolutionary biological knowledge which allows speaker of pidgins to form a language expressing propositional information, an EModE L1 learner, upon encountering a (newly re-analysed) modal which cannot assign theta-roles, ‘knows’ from their biological heritage that this verb cannot belong in VP and must instead constitute functional information which must belong in a functional projection (namely TP). As far as modals are concerned, once these types of verbs ceased to assign theta roles, they would be analysed by L1 learners as functional items rather than lexical items. 4.3 Summary A key consequence of the loss of subject-verb agreement morphology was the obscuring of the distinction between subjunctive and indicative forms, which led to use of modal verbs to express subjunctive mood. This new function of modal verbs caused them to lose properties which had formerly meant that they were included in the class of verbs generated in little v. Due to the V-Visibility constraint, models then came to be generated in T. Evidence that a lexical verb must raise out of VP was subsequently obscured by the presence of another element in T, which entailed instances of V-to-T. As section 6 will discuss, it is this development which led to increased use of do-support. 5. The rise and spread of do-support 5.1 Uses of periphrastic do Ellegård (1953) notes that periphrastic do was present in English before it came to be used as an alternative to verb raising. In particular, it was exploited by poets for rhythmical purposes, and at the beginning of the fifteenth century it served as a convenient rhyme-word which allowed for an infinitive to be placed at the end of a sentence (Jesperen 1938). Since do already had periphrastic use and carried no semantic content, it is easy to imagine that it could have been naturally adapted for use in V in situ constructions where the tense lowering operation fails. Following the rise in periphrastic modal constructions, it seems natural that speakers would seek to employ a ‘dummy modal’ (i.e. do) for use when no other modal is appropriate. Due to loss of inflection, modal and ‘do’ forms allowed for tense to be signalled more clearly (Roberts 1985). In particular, Jesperen (1938) notes that periphrastic constructions would have ‘served to make tense clear in verbs which were alike in preterite Rebecca Tollan 92 and present’ (cited by Roberts p.44). Do-support constructions would then themselves have provided further evidence against V-to-T. 5.2 The spread of do-support constructions As mentioned in section 4.3, do-support grew more rapidly in some types of construction than in others. This is illustrated by Ellegård’s graph in figure 3 (shown again below for convenience). Figure 3 (repeated): Percentages of do-support use in different sentence types (Ellegård 1953, adapted by Hans & Kroch 2000:2) Do constructions rose most rapidly at the beginning of the sixteenth century (i.e. during the period in which modals were being re-analysed). This is empirically consistent with Roberts’ and Jesperen’s views that the rise in do was driven by the changes in modals and the subsequent requirement for a dummy tense marker. And yet, however, it was not until well into the eighteenth century until do-support was adopted categorically. So why did such a development take over two centuries and, to return to my first question in section 3.3, why was the development not uniform across all constructions? One important point which arises is that the non-uniform development of do makes an analysis of parameter re-setting based on morphological change (à la Rohrbacher) even more unlikely. The RAH would surely be more compatible with data showing steady adoption of do-support, at the same rate in all constructions in which it is used today. In order to explain why this was not the case, we turn to Warner’s (1993) explanation of long-term development in terms of Rosch’s (1978) ‘Principle of Cognitive Economy’, as cited by Hudson (1997 : 10) in (13), 93 Inflectional Morphology and The Loss of V to T Raising in English (13) ‘The task of category systems is to provide maximum information with the least cognitive effort: this principle favours categorization systems where distinct characteristics are highly correlated’ According to Warner, the spread of do-support would have taken place due to changes in a small number of verbs in a particular construction, which, in order to minimize cognitive effort, was eventually generalised to all instances of that particular construction and across all speakers. It is perfectly reasonable to assume that this happened at different times and rates in different constructions. Furthermore, it is important not to forget that development of do would likely have collided with other, independent syntactic changes which could have either slowed down or accelerated the process. In particular (and in response of my second question in section 4.3), adoption of do-support in negative declarative constructions was most likely hindered by the parallel loss of the double negative ‘ne...nought’ construction (Hans & Kroch 2000). Somehow though, negative questions were not affected in the same way. In sum, the rise in do-support was not only a consequence of the changes in modals and an initial substitute for V-to-T but its adoption in turn also provided L1 learners with yet further evidence against V-to-T, and a legitimate alternative for it. Section 7 deals with an independent development which also decreased the amount of evidence for V-to-T available to L1 learner. 6. The loss of CP-V2 Lightfoot (1999) considers the loss of CP-V2 in English during the fourteenth and fifteenth centuries as a contributing factor to the loss of V-to-T. Given that CP-V2 constituted evidence of verb raising (though [head, TP] to [head, CP]), its loss would have meant the loss of a set of constructions which would have provided a cue to L1 learners to raise the verb. Lightfoot’s suggestion seems logical, but has no empirical foundation. I intend to provide this here by correlating loss of V2 with rise in do-support. The table in (14) shows percentages of CP-V2 inversion (based on a corpus study by Jacobsson 1951, cited in Fischer et al. 2000:133). (14) Percentages of sentences with CP-V2 inversion Date Late 14th century Mid-15th century Late 15th century 16th century 17th century Percentage of sentences with V2 inversion 85 – 95% 28-39% 10% or lower average 34% (ranging from 3%-85%) less than 10% If we plot these figures next to the Ellegård’s data for the development of do-support, an interesting pattern is revealed (see figure 6). Rebecca Tollan 94 % V2 Figure 6: Loss of CP-V2 (shown by plain line) and rise of do-support (Ellegård’s graph as adapted by Hans & Kroch 2000: 2 with loss of V2 imposed; plain (V2) line my addition) (NB. Mid-points in terms of centuries given in (12) were used when plotting figures from (12) onto the graph above) Under the assumption that the development of do-support was (at least in some part) a result of the loss of evidence for V-to-T, then the sharp decline in V2 directly prior to the rise in the use of do-support seems to favour Lightfoot’s suggestion that loss of V2 was indeed a contributing factor to the decline in V-to-T. Whether or not CP-V2 was lost as a result of erosion of inflectional morphology is beyond the scope of this paper to discuss, but the fact that it may have contributed to the loss of V-to-T is relevant to the analysis presented here. 7. Summary The series of changes which led from the loss of subject-verb agreement to the loss of V-to-T is summarised in the diagram in (15). 95 (15) Inflectional Morphology and The Loss of V to T Raising in English The loss of V-to-T Loss of subject-verb agreement ? Use of modals as a periphrastic alternative to the subjunctive Loss of CP-V2 Re-analysis of modals (inc. Loss of ability to assign theta-roles) Due to V-Visibility, modals can no longer subcategorize for VP and are base-generated in T Loss of much direct evidence for V-to-T Rise and spread of do-support Further loss of direct evidence for V-to-T and provision of a legitimate alternative Loss of lexical V-to-T, gradually spreading from some verb and construction types to others Rebecca Tollan 96 8. Conclusion The corpus analysis confirms that loss of subject-verb agreement morphology had no immediate impact on V-to-T. I have argued instead that it set in motion a series of changes, starting with the loss of subjunctive-indicative distinction, which led to a fundamental reanalysis in modal verbs. Due to V-Visibility, modals changed from subcategorizing for VP to being generated in T. This led to fewer instances in data for L1 learners where V-to-T raising was apparent. The resulting adoption of do-support as a substitute for tense led to yet more evidence in favour of V in situ. A decline in V2 during the late ME period also reduced the evidence in favour of verb raising. At the heart of the model I have presented is the reanalysis of modals and the combined effect of V-Visibility. The fact that modals were not reanalysed in Swedish suggests that this development was crucial in the process by which V-toT was lost. Exactly how and if at all the subjunctive mood was affected by loss of inflection in the history of Swedish remains to be seen. To return to the question of whether V-to-T was lost due to a UG constraint on agreement morphology (RAH; Rohrbacher) or a diachronic process (Alexiadou & Fanselow), it is safe to say that my analyses strongly favour the latter approach. Verb raising disappeared from English due to changes which can be traced back to the loss of agreement inflections, but not as a direct result of such a loss. Nonetheless, Rohrbacher’s prediction of a time lag under his RAH means that it is not possible to entirely rule out the idea that a weak agreement paradigm was somehow simultaneously contributing to the decline in V-to-T. If so, the burden of proof now lies with supporters of the RAH to show that this was the case. References ALEXIADOU, A. AND FANSELOW, G. 2002. ‘On the correlation between morphology and syntax: The case of V to I’ from Studies in comparative Germanic syntax, proceedings from the 15th workshop, available at http://www.ling.unipotsdam.de/~thiersch/CWGS/vtoi.pdf BICKERTON, D. 1998. Catastrophic evolution: The case for a single step from protolanguage to full human language. In Hurford, J. R., Studdert-Kennedy, M. and Knight C. (eds.) Approaches to the Evolution of Language: Social and Cognitive Bases. Cambridge: Cambridge University Press. BURROW, J.A. AND TURVILLE-PETRE, T. 1992. A Book of Middle English: Third Edition. Blackwell. CHOMSKY, N. 1986. Knowledge of language: its nature, origins and use. Praeger, New York Ellegård, A. (1953) The Auxiliary ‘do’: the Establishment and Regulation of its Use in English. Stockholm: Almqvist and Wiksell. FISCHER, O., VAN KEMENADE, A., KOOPMAN, W. AND VAN DER WURFF, W. 2000. The Syntax of Early English (Cambridge 2000). HANS, C. AND KROCH, A. 2000. ‘The Rise of Do-Support in English: Implications for Clause Structure’, Proceedings of the North East Linguistics Society (NELS) 30, 311-325 HUDSON, R. 1997. The rise of auxiliary DO: verb-non-raising or category-strengthening? (UCL). 97 Inflectional Morphology and The Loss of V to T Raising in English KROCH, A. 1989. ‘Reflexes of Grammar in Patterns of Language Change’. In Language Variation and Change 1, p.199-244. KROCH, A. AND SANTORINI, B. 2007. The Syntax of Natural Language: An Online Introduction. Available at http://www.ling.upenn.edu/~beatrice/syntax-textbook/ KROCH, A.S., AND TAYLOR, A. 2000. Penn-Helsinki Parsed Corpus of Middle English, second edition. Philadelphia: Department of Linguistics, University of Pennsylvania. JACOBSSON, B. 1951. Inversion in English: with Special Reference to the Early Modern English Period. Uppsala: Almqvist and Wiksell. JESPERSEN, O. 1938. The Growth and Structure of the English Language. Doubleday, Garden City, NY. LASNIK, H. 1995. ‘Verbal Morphology: Syntactic Structures Meets the Minimalist Program’. In Campos, H. and Kempchinsky, P. (eds.) Evolution and Revolution in Linguistic Theory: Essays in Honor of Carlos Otero, P.251-275. Georgetown University Press. LEPPÄNEN, E.R. 2010. The Relation of the decline of inflectional morphology to the Loss of verb raising in English (University of Iceland) LIGHTFOOT, D.W. 1979. Principles of Diachronic Syntax. Cambridge. LIGHTFOOT, D.W. 1995. ‘Why UG Needs a Learning Theory: Triggering Verb Movement’. In Battye, A. and Roberts, I. (eds.) Clause Structure and Language Change, p.31-52 (Oxford University Press) LIGHTFOOT, D.W. 1997. ‘Shifting Triggers and Diachronic Reanalysis’. In Van Kemenade, A. and Vincent, N. (eds.) Parameters of Morphosyntactic Change, p. 253-272 (Cambridge University Press) LIGHTFOOT, D.W. 1999. The Development of Language: Acquisition, Change and Evolution. Blackwell. RADFORD, A. 2004. English Syntax: An Introduction. Cambridge: University Press. ROBERTS, I. 1985. ‘Agreement Parameters and the Development of English Auxiliaries’. In Natural Language and Linguistics Theory 3, 321-358 ROHRBACHER, B.W. 1994. The Germanic VO Languages and the Full Paradigm (Massachusetts) ROSCH, E. 1978. ‘Principles of Categorization’ In Rosch, E. and Lloyd, B.B. (eds), Cognition and categorization p.27-48. Hillsdale, NJ: Lawrence Erlbaum. VAGRA, E. 2005. ‘Lexical V to I raising in Late Modern English’. In Generative Grammar in Geneva vol.4 p.261-281 VIKNER, S. 1995. ‘Finite verb movement in Scandinavian embedded clauses’. In Lightfoot, D.W, and Hornstein, N. (eds.) Verb Movement, p.117-148. Cambridge University Press: Cambridge. Rebecca Tollan 98 WARNER, A. 1993. English Auxiliaries: Structure and History. Cambridge: Cambridge University Press. WARNER, A. 1997. ‘The Structure of Parametric Change, and V-Movement in the History of English’. In Van Kemenade, A. and Vincent, N. (eds.) Parameters of Morphosyntactic Change P.380-393. Cambridge University Press, Cambridge. WRAY, A. 1998. ‘Protolanguage as a holistic system for social interaction’ in Language and Communication 18, p.47-67. Pergamon. Primary sources BENSON, L. D. 1987. The Riverside Chaucer. Boston: Houghton Mifflin. Third edition BLAKE, N. F. 1970. The History of Reynard the fox. Translated from the Dutch original by William Caxton. EETS O.S. 263. London: Oxford University Press. BÜLBRING, K. D. 1891. The earliest complete English prose psalter. EETS O.S. 97. London: K. Paul, Trench, Trübner & Co. JENKINSON, F. J.H. 1907. Sermo die lune in ebdomada Pasche, by Richard Fitz-James. Printed at Westminster by Wynkyn de Worde about the year 1495. Cambridge: Cambridge University Press. Facsimile edition. Appendix: Corpus verbs used in section 4 analyses The Earliest Complete English Prose Psalter c.1350 Raised verbs without inflection forsake (3sg) forgate (3sg)×2; (1sg)×4 cleped (3pl)×2 make (2sg) despised (3sg)×3 turned (3sg) stode (3pl) florissed (3sg) hated (3sg) oponed (3sg) forgate (1pl) departed (3sg) sett (3sg)×2; (3pl); (2sg) failed (3pl) seig (3pl) woned (3sg) turment (3pl) fond (3sg) bougt (3sg) sege (1p1) 99 Inflectional Morphology and The Loss of V to T Raising in English kept (3pl); (3sg)×2 trowed (3sg) turned (?) aliðgt (3sg) spared (?) depart (1pl) þtouzgt (3pl) enuironed (3pl) halp (2sg) dresced (3sg) suffred (3sg) anoied (3sg) energed (3pl) susteined (?) leued (3pl) sprad (3pl) litteled (3sg) bowed (1sg)×3 erred (1sg) lordship (3pl) souþgt (3sg) loued (1sg) seid (1sg) made (3sg) desired (3sg) (total=55) In situ roteden (3pl) (inflected) (total=1, inflected) The Parson’s Tale c.1390 Raised verbs without inflection purposed (1sg) sey (1sg) rede (1pl) suffred (3sg)×3 douted (3sg) stant (3pl) (total=8) In situ retorneth (3sg) (inflected) Rebecca Tollan (total=1, inflected) Claxton’s History of Reynard the Fox c.1481 Rasied without inflection answerd (3sg) cryed (3pl) cam (3pl) clucked (1sg) departed (3sg) awayted (3sg) brayed (3sg) hoped (?) hasted (3sg) lokyd (?) Richard Fitzjames’ Sermo die Lune c.1495 Raised without inflection came (3pl) (total=1) In situ sauyed (3sg) fulfylled (3sg) accomplysshed (3sg) entendyd (1sg) broughte (3sg) (inflected) (total=4 uninflected, 1 inflected) 100 101 Inflectional Morphology and The Loss of V to T Raising in English Rebecca Tollan Department of Language and Linguistic Science University of York Heslington York email: [email protected]
© Copyright 2026 Paperzz