4.3 Intermediate do
4.3.8 Comparison to other analyses
This account is consistent with an explanation which sees the origin ofdo-support in the Middle English causative verbdo. The debate about the origin of thedo-support construction goes back very far; Ellegård defended a version of the causative-origin hypothesis against a variety of other theories including most prominently the suggestion thatdo-support is a syntactic borrowing from Welsh.
Denison (1985) extended this hypothesis, proposing four stages of evolution: 1. dois one among many causatives
2. docausatives spread at the expense of others 3. dobecomes an auxiliary
4. doacquires its modern distribution
Denison’s description raises the question of whether and how the intermediatedoaccount has been proposed in previous literature. Denison proposes an account which comes close to the present one. For him, early (that is, ME) uses ofdoare “factitive,” and vague between a causative interpretation and a non-causative (directly agentive) one. That is,dois not a causative in ME on Denison’s account. This construction is then pressed into service as a marker of non-perfective Aktionsart (a development prompted by the decay of the OE system of Aktionsart marking through verbal prefixes), before being discarded in that function. Denison’s account nonetheless differs from the presently proposed one in several ways. If Denison’s account
26The examples are:
• “But you haue heard already how he and his brother haue deuised so with the Turke, that hemight oppresse sodainly vs
only and our fellowes.” (1560)
• “yf they do thus theyshall so worshyp veryly me” (1581)
• “And Escobar with others having no regard at all to this distinction,will condemn absolutely themboth of mortal sins” (1670)
• “so that no Passion can rise or mutiny within, but itmust betray presently itself without” (1700)
Notably, only in the second example is the object pronoun bereft of following modifiers, making possible an analysis of the other three examples which treats them as instances of extraposition of heavy objects (of a type, to be sure, which is degraded in PDE).
is accepted as an explanation of the behavior ofdoin the EME period, then we must accept that there is a 300-year time gap between the first attestations of the “vaguedo” (in the 1200s) and its rise in affirmative declaratives (beginning ca. 1500). On its own, this explanation is not satisfactory, and some other change in the language must be adduced to explain why the rise of affirmative declarativedooccurs when it does, and not in any of the earlier centuries that seem available to it.27 Secondly, the data from the corpora do
not support the Aktionsart-marking hypothesis. Figure 4.30 shows that agentive non-perfectives such as “believe” and “desire” have high levels ofdo-support while non-agentive ones such as “please” and “seem” are much lower. Figure 4.31 shows the same graph for perfective verbs. Though (the commonest) perfective verbs are generally more agentive, the non-agentivedieshows a low trajectory (as doestell, anomalously). Indeed, the association of perfectivity with agentivity gives the impression that Denison’s proposal thatdo
is associated with non-perfectives in the earliest period of the change is precisley backwards. However, the data support my arguments above that argument structure (specifically related to the presence and role of the external argument) is a more important determinant ofdousage behavior than Aktionsart. It may be the case that Denison’s account can shed light on the earliest stages of the emergence ofdo-support, but it cannot explain the behavior ofdoin EME. Denison himself admits this fact, identifying his final stage four with the period after 1400 (“mainly [the] fifteenth and sixteenth centuries”, p. 55). Thus, the account I have sketched above does not directly compete with Denison’s account, but rather addresses a different part of the trajectory ofdo-support through the language.
27The loss of V-to-T raising is a noticeable change that takes place at the right time to be a candidate to influence the development of
do-support. However, as I have extensively argued above,dousage in affirmative declaratives is not related to the otherdo-support contexts, and thus a separate explanation is needed under Denison’s account for its apparent 300 year lag.
Figure 4.30:A graph of the evolution of affirmative declarativedo-support with the most common non- perfective verbs in the parsed corpora, as diagnosed by the “for/in” test. The black line is the (weighted) average trajectory for the class.
Figure 4.31:A graph of the evolution of affirmative declarativedo-support with the most common perfective verbs in the parsed corpora, as diagnosed by the “for/in” test. The black line is the (weighted) average trajectory for the class.
Roberts (1993) also claims an analysis with an intermediate do: “Sixteenth centurydo, then (until after 1575), was intermediate between the ME main verb and the [ModE] ‘supporting’ auxiliary in terms of its distribution.” (p. 296) However, this is not an intermediate analysis in the same terms as that proposed here. Roberts proposes a reanalysis ofdofrom a main verb to a modal in one fell swoop, occurring in the early 1500s (he proposes 1530). He concludes that after this date,dois “semantically empty” – which is clearly not
the case, given the data on argument structure presented above. Roberts’ intermediate stage just marks the time between the loss of the MEdoand the emergence ofdo-support, and is not attributed any properties of its own. Though it shares the name “intermediate,” this account is very different in spirit from the one I propose.
The account presented here accords with the causative-origin theory, advocated by Ellegård and (in a modified form) Denison and Roberts. It is natural to suppose thatdomight be reinterpreted as a marker of agentivity or external argument presence, both those features being present in causatives. The data presented do not, however, resolve the question of whether it is truly agentivity or the presence of a (not necessarily agentive) external argument which drives the insertion ofdoin the intermediate grammar. The data from affirmative sentences (declaratives and questions) contradict that from negative declaratives; in the former sentence types experiencer-subject verbs pattern with the agentive verb types, whereas in the latter they behave like unaccusatives. The root of this difference could lie in the interaction of the lexical semantics of these predicate types with negation, especially ifdois agentive. On the other hand, an account within the framework of Grammaticalization (Hopper and Traugott 1993), where semantic bleaching is an important element of the notion of language change, would prefer to see the intermediate grammar as marking external arguments, since Agent is not a bleaching of Causer but rather a strengthening. That is, forX to have the thematic role Causer with respect to an eventemeansXis the cause ofe. ForX to be the agent ofe,Xmust both be the cause ofeandX must act volitionally.