On analysing fragments: the case of No?

4.1 A Celtic-contact explanation?

In thinking about a possible source for the INOPO pattern in English, one candidate that may come to mind is language contact. Although the general role of this factor in shaping the syntax of English is controversial (compare, for example, the optimistic assessment in Emonds and Faarlund [2014] with the more pessimistic views in Bech and Walkden [2016]; Fischer et al. [2017: 51–76]; Stenbrenden [2016]), there is no a priori reason to rule out an effect of this type in the present case. It turns out that there is indeed a potential source construction for INOPO in Welsh/Brythonic Celtic, a language known to have been in close contact with English for many centuries. In (27), we give an example of a pattern in modern Welsh (taken from Jones [1999: 139]) which corresponds rather closely with INOPO in (i) being interrogative; (ii) being a response to an other-speaker negative statement; and (iii) being composed of a negative element followed by an auxiliary/operator.

Jones (1999) makes clear that this pattern is not restricted to interrogative use but is widespread in Welsh answers to questions and in other response utterances. A simple question-answer example, taken from Jones (1999: 150, ex. 7), is given in (28).

From the description in Rottet and Sprouse (2008), it appears that the same pattern existed in Early Modern Welsh and Evans (1972: 177) provides examples of it as used in responses to questions in Middle Welsh (though in that period, the pattern allowed use of lexical verbs as well as auxiliaries).

Given these facts, it is conceivable that the English INOPO pattern was calqued from Welsh or its Celtic ancestor, on the basis of utterances similar to the B-part of (27). Vennemann (2009) indeed suggests that the English use of short answers in the shape of He did/They were/We won’t instead of bare yes/no is due to Celtic influence. He argues that English is alone among the Germanic languages in having them and attributes their origin to contact with Brythonic during the Old English period. Some researchers have suggested such an origin for various other features of English (see Benskin 2011; Filppula et al. 2008; Schrijver 2009), which provides some circumstantial evidence for the idea that INOPO may have originated in a situation of contact with Celtic.

However, the data in Table 1 do not provide any great support for the idea that the pattern originated during the Old English period: all of the examples that we have been able to find are from the 16th and 17th centuries. Indeed, tag questions in general are hardly attested in English before 1500 (Tottie and Hoffmann 2009: 133). As a solution to this problem of chronology, Vennemann (2009: 325–26) suggests that tags may have been a substratal feature in speech for several centuries before surfacing in the written superstratum in Early Modern English. Such an appeal to non-attested data obviously is not the most attractive idea in the world. Moreover, as Vennemann (2009: 326) makes clear, there is some scholarly disagreement about whether the use of this kind of response pattern already existed in Old Welsh (i.e. before the 12th century), or whether it came into existence only in Middle Welsh (12th–15th centuries). Altogether, therefore, the idea that INOPO originated in Old English under Celtic influence faces clear problems.^[8]

If we want to avoid throwing out the Celtic baby with the bathwater, we might hypothesise that transfer of the pattern took place not in Old English but in the Late Middle English/Early Modern period itself, perhaps as a result of more contact, raised status and greater acceptance of the Welsh language after the accession to the throne in 1485 of Henry VII, who had a partly Welsh background.^[9] However, such an approach is also not without problems. As Chrimes (1999: 3–4) makes clear, there is actually no evidence of use of Welsh or any special orientation towards Wales on Henry’s part after he ascended the throne. Further data for the examples of INOPO that we have found also do not support the idea. For each of the twenty-five authors that used INOPO, we have tried to establish where they were born and/or grew up, using biographical information from the Oxford Dictionary of National Biography ODNB. The results are shown in Table 2.

Clearly, there is no predominance of writers from Wales. Among the authors born in England, there are a few from the West Midlands but also several from London, East Anglia and the North, without any apparent Welsh family connection. The three most prolific users of INOPO, John Foxe, John Philpot and John Bridges were from Lincolnshire, Hampshire and Devon respectively and do not appear to have had any significant ties with Wales. Thus, the geographical distribution of the attested data does not point strongly at a Celtic origin for the INOPO pattern.

Furthermore, if nevertheless INOPO entered English through contact with Welsh, it would need to be explained why the pattern in English was restricted to interrogative contexts. As we saw above, Welsh uses (and used) not only an interrogative Neg+V pattern, as illustrated in (27), but also a declarative pattern, as in (28). An impression of the approximate frequency of the two patterns in Welsh speech can be gained from the figures given in Jones (1999: 287). Although those figures are based on recordings made of a very specific group of speakers (children aged three to seven, interacting together in play areas at school) and are contemporary (so any extrapolation to earlier Welsh must be tentative), the difference is stark: the material collected by Jones featured 635 negative responses that were declarative and only 14 that were interrogative. If anything like this proportion was also characteristic of the Welsh spoken in the 15th and 16th centuries, then we might expect borrowing to have taken place of the much more frequent declarative pattern. However, responses of that type are not found in Early Modern English: INOPO only occurs in interrogatives.

In summary, it does not look as though English INOPO has arisen through contact with Welsh or early Celtic. Our reasons for rejecting this account are that (i) INOPO arises long after the time which might be held to be most favourable to Celtic-English grammatical borrowing, i.e. long after the Old English period; (ii) the writers using INOPO, in the 16th and 17th centuries, are predominantly not of Welsh origin; and (iii) positing Welsh as the source for INOPO does not explain why INOPO is restricted to interrogative contexts in Early Modern English.

4.2 Language-internal origins: a role for thou?

Having rejected a language-contact-based explanation, we will now propose and evaluate two potential accounts in which INOPO arises as the result of language-internal developments in the Early Modern English period. Such accounts of INOPO need to explain why no is followed by an operator; why the negative word is no rather than the more typical clausal negator not; and why the INOPO construction lacks an overt subject, as Early Modern English is not a generalised pro-drop language.

The first possibility that we will explore is that INOPO originated in full yes/no questions starting with whether or no, a pattern that is attested in the 16th and 17th centuries. Examples are given in (29) and (30).

In both examples, whether or no is the first constituent of an overt clause. Inversion of the finite verb and subject suggests that – like many interrogatives – these clauses have a focused constituent (whether or no) in initial (SpecCP) position accompanied by fronting of the verb (T-to-C movement). There does not appear to have been any scholarly work on this pattern, perhaps because it can be regarded as being a minor variant of the construction with only the word whether heading a main-clause question, as in (31)–(32).

Such direct yes/no questions introduced by whether already existed in Old English and continued into the Early Modern period (see e.g. Allen [1980] on their syntax in Old English and Walkden [2014: 144–155] on their diachronic development).

It is the variant in (29)–(30), with negative interrogative no coordinated with the neutral interrogative particle whether, that may have led to the emergence of INOPO, through a number of different steps. Firstly, the coordinated use of no in (29)–(30) could lead to uncoordinated use of no, since the existence of ‘A or B’ in some specific function generally implies the possibility of using uncoordinated or bare ‘B’ in the same function. In fact, examples like (31)–(32) could provide a model for this, since they have uncoordinated whether as compared with the coordinated whether or no seen in (29)–(30). Once uncoordination of no was possible, sentences like (33) would result. The change would cause the word no to stand alone in the clause-initial SpecCP position of a fully spelled-out interrogative clause.

In sentences like this, VP deletion could result in short questions introduced by no, of the form in (34).

Several examples of the type illustrated in (34), with no + operator + subject, are in fact attested in our 16th–17th-century EEBO data. Two are given in (35), both with the subject pronoun you. (36) gives two further examples, with the subject pronoun thou.

Having taken these two hypothetical steps, involving uncoordination of no and subsequent VP-ellipsis, we are of course getting closer to the INOPO pattern. However, in contrast to the examples given in (35) and (36), which have overt subjects, the INOPO pattern, as in (37), repeated from (16), is distinguished by the absence of an overt subject.

Therefore, this account still needs some further step(s). An observation made by Visser (1966: 5) and Barber (1997: 200) is helpful here: they note that Early Modern English has a restricted form of pro-drop in interrogative clauses with second person singular verbs, where we might expect the overt subject pronoun thou. The questions in (38) illustrate this phenomenon of thou-drop.

We assume that the structure of an example like (38a) is as outlined in (39), with a pro-subject licensed by V-to-C movement.

Assuming that thou-drop could also affect negative interrogatives like No wilt thou? in (36a), its application would yield a sentence consisting of just No plus an auxiliary, as the derivation in (40) illustrates. Observe that the construction in (40b) with thou-drop has the same surface form as the INOPO construction illustrated in (37).

Consequently, it is possible that examples with the subject pronoun thou such as those in (36) could be the forebears of the subjectless INOPO pattern. To get from the former to the latter, we can use the independently attested mechanism of thou-drop. Some support for this account comes from Table 3, which shows the numbers of tokens of thou-drop in eligible contexts in two parsed corpora covering the period 1430–1710. It can be seen that thou-drop is most prevalent during the 16th century, which is exactly the period when most of our INOPO examples are found (cf. Table 1).

This then is how instances of INOPO interpreted as having thou as a subject may have arisen. Cases where a different subject is understood may have come from the following further development: clauses with thou-drop plus VP-ellipsis, with a structure as shown in (41a), might be open to reanalysis as cases of TP-ellipsis, with a structure given in (41b).

While (41a) was only possible with second person singular subjects, (41b) generalised to subjects of all persons and numbers, in effect yielding the complete INOPO pattern, as shown in (42).

To summarise, the explanation for the emergence of INOPO that we are now considering takes as its starting point fully spelled-out yes/no questions introduced by whether or no; the entire development involves six successive stages, given and labelled in (43a–f).

In this way, (43) provides a series of syntactic mechanisms to derive the INOPO construction, each of the steps being due to the operation of processes that can be attested independently of this particular development.

However, we may wonder: how plausible is the account outlined in (43) when viewed from an empirical perspective? The answer is that, unfortunately, it does not square exactly with the data provided by 15th–17th-century texts. Specifically, we find no examples – in the corpora examined or elsewhere – representing the hypothesised stages (43b) (No wilt thou lend me any?) and (43d/e) (No wilt?). This may be due to accidental gaps in what remains rather sparse data but their absence makes it difficult to assign a pivotal role to such examples as bridging contexts between examples like (43a) and the INOPO pattern in (43f). In addition, examples like (43a), the hypothesised starting point of the entire development, are rather infrequent and most of them are somewhat later in date than many of our INOPO examples, casting doubt on their role as the ultimate source for INOPO.^[10]

Furthermore, the stages in (43) might lead us to expect a similar development for bare interrogative whether, yielding other-speaker question tags of the form whether + operator, paralleling INOPO (e.g. a presumed Whether will?, parallel to attested No will?). However, such tags with whether are not attested in our Early Modern English corpora. All the examples that we have found of yes/no questions introduced by uncoordinated whether contain fully spelled-out clauses similar to (31) and (32), so without TP- or VP-ellipsis. These various empirical problems make it rather difficult to maintain the account outlined in (43).

4.3 Language-internal origins: a role for echoing?

We now turn to a third account for the origins of INOPO, which has echoic interrogative no as its starting point. By way of introduction to the relevant phenomenon, it is useful first to consider present-day examples of it like (44) and (45), both taken from the British National Corpus BNC.

In both of these cases, the word no asks for confirmation of the negative polarity of the immediately preceding clause. From a functional point of view, this usage, which we shall label ‘interrogative no’ (INO), is equivalent to an intonation question, i.e. a question without operator-to-C, with interrogative force marked solely by intonation (Bailey et al. 2010), as in the B-part of (46).

In Early Modern English too, both interrogative no and intonation questions existed. Examples are given in (47) and (48) respectively.

Since interrogative no is rather similar in form and function to the INOPO pattern, it is worth examining its distribution over time in some more detail. In Table 4, we compare its frequency during the 16th and 17th centuries with that of the INOPO pattern.

The earliest example of INO in the corpora is from Malory’s Morte D’Arthur (ca. 1470). The Middle English Dictionary MED gives one clear instance from the 13th century and one from Chaucer’s works (MED s.v. nō interj. 1[c]) but it looks as if it became more usual only after 1500. In the Early Modern parts of the corpora, the frequency of INO is similar to that of INOPO until 1600 but while the latter declines after that year, the INO pattern clearly remains in use and still exists today.^[11]

In INO the interrogative nature of the utterance is only signalled by intonation, just as in intonation questions like the B-part of (44). If we adopt an ellipsis approach à la Holmberg to INO and assume that its syntax includes a full (intonation) question which is elided, the structure of an example like (49) could be as in (50).

We can follow Holmberg (2016: 212) in assuming that no is focused and occupies SpecFocP only when it responds to an interrogative, as in (13).^[12] Hence we take no in (50B) to be the specifier of an unexpanded CP. All the examples of INO that we have found respond to (negative) declaratives, so we can say that the no in them lacks focus.

While (50B) has a fully elided clause without verb movement, it may be noted that intonation questions generally have an alternative form, where the interrogative nature of the utterance is signalled syntactically through fronting of the operator. For the intonation question He didn’t? in (46B), for example, the alternative is the regular syntactic question in (51B), which would have the structure shown in (52), with fronting of the operator (and also VP deletion, just like [46B]).

If we now consider the structure of an utterance like (50B), Noo, these be no suche tydynges , but add to it an application of operator fronting (of the type seen in regular yes/no questions), the result will be a configuration where TP is still elided but the operator survives because it has been extracted from TP prior to ellipsis. This would yield a structure as in (53), which in effect is the INOPO pattern, with no followed by the fronted operator and ellipsis of the following clause.

It therefore appears that INOPO can be derived from INO through one simple step: while INO consists of interrogative no followed by a clause that is elided in its entirety, in INOPO the operator is preserved because it has moved out of its clause prior to ellipsis. The motivation for the movement can be taken to be the same as in regular yes/no questions: the operator is attracted by the feature [+Q] on the relevant functional head (here: C⁰). Under this account of the emergence of INOPO, it consists in the formation of a regular yes/no question based on a construction that previously only occurred as an intonation question. In present-day English, yes/no questions with fronting of an operator are much more frequent than intonation questions, in speech as well as writing (see Biber et al. [1999: 212] for data) and it is likely that the same pattern obtained in Early Modern English. The change from INO to INOPO can therefore be thought of as an extension of the process of operator fronting to a clause type that originally did not have it but that did meet the conditions for its application, in being a main-clause polar interrogative. Conceivably, there may also have been specific influence coming from the category of tag questions, such as are you?, can they? and doth she not?, which developed in the course of the 16th century (see Tottie and Hoffmann [2009] for data and analysis).

This account for the emergence of the INOPO construction can also accommodate the cases that feature a sequence of operator + not/n’t, as in (25), repeated here as (54).

As in PDE, can’t represented a unitary form, which would be eligible for operator fronting as one single constituent. For the cases of INOPO featuring forms like would not, we can assume – following Rissanen (1999a) – that, before spellings like wouldn’t became acceptable, two-word spellings were sometimes used to represent what was in effect a contracted form, i.e. <would not> could represent [wudnt], in regular declaratives but also in instances of fronting, such as INOPO. The low number of attested INOPO cases with operator + not/n’t may reflect a certain unwillingness on the part of writers to use a weakened or contracted form that was still in the process of establishing itself in speech. This interpretation fits well with the 16th-century data in Tottie and Hoffmann (2009: 138–139), who found 92 tag questions of the type is it not? and do you not?, as against one with the order Operator + not + Subject Pronoun (in the example I told you, did not I?; John Jefferes, The Bugbears, 1563). We will come back to the INOPO cases with not/n’t in Section 5.3, exploring a slightly different analysis.

4.4 Analysing INOPO: a comparison

From an empirical point of view, the account presented in Section 4.3 has several advantages compared to the two other accounts that we have examined. It avoids the uncertainties associated with the Celtic-origin hypothesis discussed in Section 4.1, which requires belief in the existence of unattested INOPO sentences already in Old and Middle English or – alternatively – borrowing from Welsh by speakers/writers not likely to have had much contact with the language. On the technical side, the account sketched in 4.3 provides a ready explanation for the absence of declarative NEG+AUX answers to questions, which under the Celtic scenario remains a puzzle. Under the account where INOPO derives from INO, the explanation is that the presence of an overt auxiliary in INOPO is a result of operator fronting, which takes place regularly in interrogatives but not in (declarative) answers.

Similarly, deriving INOPO from INO avoids several problems raised by the account incorporating thou-drop as a crucial element, discussed in 4.2. As we saw, some stages in that presumed development turned out to be unattested in data from the period. Especially striking is the absence of instances of INOPO featuring a silent thou and the verb form appropriate to it. If the development was indeed most directly based on sentences containing thou, we would expect instances of INOPO with silent thou to predominate, certainly early in the development. But what we find is no instances with silent thou at all. Since the INO-based scenario of Section 4.3 is not tied to examples specifically with (silent) thou, their absence – though notable in itself – does not diminish the plausibility of the account. Furthermore, while the thou-drop account predicts that there could easily be a parallel development leading to short positive interrogative responses headed by whether (yielding examples like Whether will? or Whether can?, as sketched at the end of Section 4.2), the INO-based account makes no such prediction. Indeed, since interrogative responses with bare whether are not attested (i.e. there are no cases of Whether? uttered in response to an affirmative statement), the INO-based account of INOPO predicts that no interrogative responses with whether + operator could have arisen. And indeed, no such responses are attested.

The existence or non-existence of a potential (short) source expression also helps explain the patterning of affirmative counterparts to negative INOPO. As we saw at the end of Section 3, there are some examples consisting of interrogative yea + operator as in (26), repeated here as (55), but none featuring yes + operator.

For IYEAPO cases like (55), the expectation would be that they originated from interrogative use of bare yea, in basically the same way that INOPO originated from bare interrogative no. Data from our corpora shows that bare yea could indeed be used as a positive short question. Examples are given in (56)–(58).

Continuing to adopt an ellipsis approach to short utterances, we can say that in examples like this, bare yea is part of an intonation question in which the entire TP is ellipted, leaving yea as the only overt element. For example (58), this means the structure is as shown in (59a). Opting to apply in such sentences the operator fronting that was usual in interrogatives would yield a structure where the operator has moved out of the TP before it is deleted. This is shown in (59b), which is exactly the IYEAPO pattern of (55).

This account for yea + operator is therefore entirely parallel to that given above for the emergence of INOPO. The absence of examples with yes + operator then leads to the expectation that there was no source expression for it, i.e. no bare yes functioning as an interrogative response. This expectation is indeed fulfilled. We have found no examples of interrogative yes in our corpora and the citations for it in the OED (s.v. yes, adv., n., and int., sense A5) suggest that such cases did not come into existence until the mid-19th century.

A further piece of empirical data mentioned above consists of responses containing no + operator + subject, as in (60), repeated from (35a).

In the thou-based account of Section 4.2, such examples originated from a presumed full interrogative clause. For (60B) this would have been No have you not wronged me?, with subsequent VP-ellipsis yielding No have you?. However, we saw that this type of presumed full-clause source is actually not attested. For these cases too, the INO-based account of INOPO yields a more successful explanation. Rather than viewing the construction in (60B) as coming from a full clause, it can be seen as an expanded form of a reduced clause, i.e. of INOPO. While INOPO has ellipsis of its entire TP, sentences like (60B) have ellipsis of VP only. The resulting structure for such sentences is shown in (61).

In this way, we can account not only for the emergence of INOPO but also of INOPO + subject: the latter is structurally identical to the former, but it involves spelling out more of the clause. Viewed like this, the development consisted of a widening of options. The starting point was bare interrogative no, which was similar to an intonation question in not having T-to-C and in which ellipsis of TP thus meant that only no was spelled out. Applying T-to-C movement to such a structure enabled the operator to escape from the TP-ellipsis site, resulting in INOPO. Choosing to elide not TP but VP resulted in INOPO + subject. A further hypothetical step would have been to elide nothing at all, which would have resulted in a fully spelled-out clause, e.g. No have you not wronged me?. As pointed out above, this option is not attested but this may simply be because all the interrogative no utterances are responses. Elision of nothing at all would mean that the response would repeat the complete relevant clause of the immediately preceding utterance, resulting in a hypothetical exchange like (62), based on the actual exchange in (60), which has VP-ellipsis in the response.

As Sadock and Zwicky (1985: 191) point out in their discussion of cross-linguistic patterns in answers to polar questions, such systems often show a “prohibition against undue prolixity”. This may be the reason why non-elliptical responses as in (62B), consisting of interrogative no + full clause, are not attested.

An additional point speaking in favour of the idea that INOPO represented an extension of the INO pattern, rather than borrowing from Celtic or reduction of a fuller clause, can be made on the basis of chronology. We saw in Section 4.3 that INO, while attested a few times in Middle English, seems to have become frequent only after the year 1500. If this pattern in the textual data is a more-or-less accurate reflection of INO’s frequency in actual speech of the period, it could mean that INOPO’s emergence in the early years of the 16th century was a consequence of the greater use of its source construction in that period. What may also have contributed is the appearance of regular tag questions, which started becoming established at the same time (see Tottie and Hoffmann [2009] for this development). The result would be a chronologically rather tight explanation, without a need to postulate the existence in speech of constructions absent from the written record – which both the Celtic-origin and clause-reduction accounts require.

Having established, on the grounds of these various considerations, that the emergence of the INOPO pattern is best explained as being based in clauses with interrogative no, we can now return to the question posed in Section 2: should the response items yes/no be analysed as being part of a full clausal structure (some or all of which can be elided) or are yes/no radically bare? Of course we have couched our account of the development from INO to INOPO in a full-clause framework. But there is a good reason for this: that account will not work in a bare-phrase framework.

Thus, consider how such a framework would deal with the emergence of an example like (63), repeated from (19).

Before sentences like (63B), i.e. INOPO, existed, a negative interrogative response to (63A) could take the form of bare no, i.e. INO, as in (64).

If this no is analysed as a bare phrase which is anaphoric to a preceding proposition, then no in (64) would express the (interrogative) meaning ‘You did not know me and you did not hear of me?/Did you not know me and not hear of me?’. That is possible. But it is not clear then what would at some point have triggered or enabled the appearance of the word did following anaphoric no. It is true that, after no, some of its meaning (‘you did not know me/hear of me?’) might conceivably be spelled out overtly but then we would expect something like You did not? or, with T-to-C, Did you not?, rather than the single word did. Thus the bare-phrase analysis of no seems to provide no insight into the origins of INOPO. Moreover, if no in (63B) is indeed radically bare, that must mean that did forms a separate and equally bare clause/speech act/non-sentential fragment on its own. However, English utterances consisting of a single operator are only plausible if there is some kind of contrastive focus on the operator. Under a bare-no approach to INOPO, we would therefore have to say that a bare utterance with no can exceptionally license a following bare-operator utterance without contrastive focus. It is hard to see what mechanism could bring this about. Similar observations can be made about the IYEAPO pattern of (55).

Under an ellipsis account of INOPO/IYEAPO responses as in (63B)/(55), no such problems arise. No/yea and the operator are part of the same interrogative clause and both elements behave as expected: no/yea occupies initial position in CP, as usual, while the operator undergoes the fronting that is standardly found in interrogative clauses. Moreover, as we argued above, such an analysis allows for a successful account of the origins of INOPO/IYEAPO, deriving it from interrogative no/yea in a rather simple manner. Our conclusion is therefore that, on present evidence, there are good reasons for preferring an ellipsis account of no/yea over a bare-phrase account, at least for the interrogative response construction at hand.

This conclusion also has implications for an issue that we have glossed over so far, i.e. the question of whether all utterances starting with response yes/no have the same syntax. Holmberg (2016: 210–215) in fact argues that, since yes/no uttered in response to a statement rather than a question is not involved in polarity assignment or agreement, it is more similar to a confirmatory expression like (that’s) (not) true or that’s (not) right. He also points out that in some languages the yes/no used in answers to questions cannot be used in responses to statements. These facts lead him to suggest that yes/no after a statement may be of the bare anaphor type proposed in Krifka (2013). Holmberg does not consider interrogative responses to statements but it is clear that the emergence of INOPO, as analysed above, entails that interrogative response no/yea in Late Middle English/Early Modern English was the initial element of an otherwise silent clause, and not a bare anaphor.

5.1 Clausal ellipsis and the distribution of early English no/yes

The analysis of early English no as the first element of an elided clause, as sketched in Section 4.3 above, raises some further questions about its nature and function, which we discuss here. Obviously, in all the uses that we have considered so far, no is a negative response item. However, there are certain restrictions on the distribution and function of this no in early English that suggest a more precise characterisation is needed. The main fact is the following: from its earliest attestation until the 16th century, no as a response item was virtually exclusively used after negative utterances. Marsh (1864: 578–584) seems to have been the first to note the possible existence of such a pattern, though the scarcity of edited early texts at the time meant that he was unable to provide systematic quantitative data to confirm his impressions. The OED (s.v. nay, adv.1 and n.) also hints at such a restriction on the early use of no, though their wording suggests it may have been a matter of propriety of usage rather than principle of grammar. Culpeper and Van Olmen (2018) examined data from Shakespeare’s plays and found that negative answers to negative questions always take the form of no but that, in addition, no is also sometimes used in responding to non-negative questions. They interpret this as the result of a partial breakdown of an earlier system, where no was restricted to answers to negative questions. Inspection of corpus data for Old and Middle English confirms that there was indeed such an earlier system. EEBO yields 17 examples of response item no in the period 1473–1499. The PPCME2 contains eight examples of response no from the period 1150–1500 and the YCOE yields one example from 850 to 1150. All 26 instances occur in contexts such as (65), in which no confirms the negative proposition introduced by the question.

By contrast, in the period from the 12th to the 16th century, negative responses to neutral questions typically take the form nay (n = 65/66 in the PPCME2 corpus), as (66) illustrates.

This restriction on the distribution of no follows if bare no in this period was a concordant negative item standing as the initial element of a clause with TP-ellipsis. In terms of the analysis in Holmberg (2016: 162–165), no in cases like (65) does not express negation on its own (technically speaking, it has an uninterpretable neg-feature). Instead, as shown in (67a), it agrees with the (interpretable) negator not that is part of the TP copied from the antecedent question. Subsequently, this TP undergoes ellipsis at PF under identity with the negative TP in the antecedent, leaving no spelled out in SpecCP, as shown in (67b).^[13]

(67)

a.

Syntactic agreement (negative concord):

… knowyst þu not me?

[CP No sir [TP [ADVP forsothe] [TP I know you not]]]

b.

TP-ellipsis:

… knowyst þu not me?

[CP No sir [TP [ADVP forsothe] [TP I know you not]]]

For response item no in present-day English, Holmberg (2016) proposes an analysis where no can be either interpretable (and therefore able to value the polarity feature of a following clause, as in Has he left? No he hasn’t left ) or uninterpretable (and therefore requiring the presence of an interpretable negative item that it can agree with, in the manner shown in [67a]). As we have just seen, no in English before 1500 worked differently: it always had an uninterpretable negative feature, hence was restricted to occurring as a response to negative utterances. For responses to neutral or positive utterances (i.e. non-negative yes/no questions or positive statements), the response item nay, with an interpretable negative feature, was used.

A similar dual system existed in early English for positive response items. As shown in Wallage and van der Wurff (2013) for Old English, Wallage (2018) for Middle English and Culpeper (2018) for Early Modern English, yes was largely restricted to occurring in responses to negative questions or declaratives, where it served to provide an emphatic statement of the positive polarity proposition, as in exchanges like (68a) and (68b).

Given this restriction, yes was naturally not a suitable response to a positive statement. For this, the response item yea was used.

When INOPO/IYEAPO arose in the course of the 16th century, this division of labour between no and nay and yes and yea was still largely in place. Since INOPO responds to a negative utterance and IYEAPO to a positive utterance, the use of no and yea (rather than nay and yes) in this pattern thus fully fitted with the grammar of responding that existed at the time. Within this grammar, response no needed to agree with another negative element. Since negative agreement (aka negative concord) is typically regarded as a clause-bound syntactic phenomenon (see, for example, Zeijlstra [2004] on negative concord in Present-day European languages, and Wallage [2017] on early English negative concord), the facts provide further support for the ellipsis approach to no. We noted above (see Section 4.4) that a bare-phrase analysis of no in INOPO leads to the undesirable conclusion that early English allowed a single operator without special focus to occur as a sentence fragment. We can now add a further undesirable consequence of the bare-phrase analysis: it does not make available a mechanism that will enable response no to agree with an interpretable negative feature in its clause.

5.2 Ellipsis, sluicing and head movement in interrogatives

Our syntactic analysis of INOPO/IYEAPO distinguishes it from other short response utterances in one crucial respect, raising further questions about the analysis of interrogative no. In the INOPO construction both the SpecCP and C⁰ positions are overt. However, in other interrogative responses, such as (69B), where the SpecCP position is overtly filled, C⁰ cannot be lexicalised, as shown by the ungrammaticality of (69B′).

These facts were first noted by Ross (1969), who referred to constructions like (69B) as examples of sluicing. The issue was later taken up by Merchant (2001: 61, 2006b), who attributes the restriction seen in (69B′) to the ‘sluicing-COMP generalisation’, formulated as (70).^[14]

It has been proposed that the contrast in examples like (69) is due to a wider principle named MaxElide (Merchant 2008; Takahishi and Fox 2005), which in essence states that, given the choice of which projection to elide, the largest possible projection must be chosen. However, arguments have been presented against MaxElide (see e.g. Griffiths 2019; Messick and Thoms 2016) and there is currently no consensus about the best approach to contrasts as between (69B) and (69B′).

Whatever its ultimate explanation, it is clear that something like the sluicing-COMP generalisation has held for English throughout its history: the pattern in (69B′) is unattested at any stage of the language. If SpecCP is occupied (in [69B/B′], by a wh-item), a process of elision must also target the element in C⁰. Yet, in the analysis of INOPO that we presented in Section 4.3 and argued to be superior to alternatives in Section 4.4, there is a configuration in which SpecCP is filled by the response particle no and C⁰ by an auxiliary. What is there about INOPO that allows it to evade the effects of the sluicing-COMP generalisation?

Interestingly, the INOPO facts would follow from a specific analysis of verb movement that is proposed, for independent reasons, in Holmberg (2016). He notes that Finnish has verb fronting to C in polar questions but not in wh-questions (Holmberg 2016: 37).^[15] On that basis, he proposes that there is a difference between the verb movement in yes/no questions and in wh-questions. He argues that movement to C in yes/no questions occurs in narrow syntax, since it has both interpretative effects at LF and linearisation effects at PF, whilst movement to C in wh-questions occurs at PF since it only has a linearisation effect and there are no LF-interpretative effects associated with it.

To see what interpretative effects are at stake, we need to consider in some more detail the analysis of questions adopted by Holmberg (2016: 32–38). To begin with, Holmberg (2016: 32) follows Hamblin’s (1958) proposal that the meaning of a question is the disjunction of a mutually exclusive set of possible answers. In the case of neutral yes/no questions, that set (the question set) comprises an affirmative proposition and a negative proposition, as shown for a simple question in (71).

The answer to a yes/no question like this selects one of the propositions from the disjunction. Holmberg argues that the disjunction is encoded syntactically by means of an underspecified polarity head Pol⁰ above TP, with the two possible values [+/-pol]. The feature [+/-pol] has to have sentential scope to ensure that the question is interpreted as the disjunction of two complete propositions. Holmberg (2016: 38) argues that this is achieved by Pol⁰ moving to C⁰. As a result, [+/-pol] scopes over the two propositions and becomes the Centre of Attention under Q-force – in other words, it is this [+/-pol] disjunction that is in question.

Negative questions have the same syntax and semantics, except that the starting proposition in the disjunction is negative. The question set in negative questions is therefore negative ¬p or its negation ¬(¬p), as shown in (72).

Here too, the syntax includes a [+/-pol] feature on Pol⁰ which moves to C⁰ to achieve sentential scope. Crucially, the negation in the proposition they will not put it right does not value the Pol⁰ head as negative, because this negation scopes under the [+/-pol] feature in C⁰. The truth of the proposition is still at issue in the negative question and Pol⁰ therefore remains unvalued. As Holmberg (2016: 40) notes, it is impossible to have a question where the [+pol]/[-pol] feature is the Centre of Attention if Pol⁰ is actually valued.

In a wh-question, the situation is different. The variable in such a question is not the polarity of the clause but a wh-constituent. Its sentential scope is marked by wh-movement (Ā-movement) into SpecCP. There is also movement of T⁰ to C⁰ but this does not have an interpretative effect at LF, hence Holmberg (2016: 37) hypothesises that it is derived by PF linearisation rather than syntactic movement. This means that, while questions in English always have finite verb movement to C, polar questions and wh-questions involve different types of movement. In yes/no questions, Pol⁰ moves to C⁰ as part of the syntactic derivation and it is interpreted in C⁰ as marking the polar interrogative nature of the sentence. In wh-questions, on the other hand, it is wh-movement (A′-movement of a wh-constituent to SpecCP) that marks the sentence as a question. The movement of T⁰ to C⁰ does not contribute to interpretation and is best thought of as a PF phenomenon.

The 16th–17th-century English INOPO construction provides empirical support for these ideas. Recall that the structure that we have proposed for INOPO is as in (73), which appears to violate the sluicing-COMP generalisation since both SpecCP and C⁰ are filled.

This contrasts with cases of sluicing as in (74), where violation of the sluicing-COMP generalisation results in ungrammaticality.

Once we make a formal distinction between narrow syntax Pol-to-C in polar questions and PF T-to-C in wh-questions, the difference between (73) and (74) follows. When both ellipsis and head movement occur at PF, as they do in wh-questions, ellipsis bleeds head movement, so no operator fronting can take place.^[16] However, if Pol-to-C movement occurs during narrow syntax, as it does in polar questions (including INOPO), the operator in C⁰ will escape any ellipsis that takes place later at PF.

5.3 From beginning to end: some thoughts on INOPO’s demise

Our proposals for the structure and origin of INOPO have enabled us to shed light on several pieces of empirical data in Early Modern English. These include various kinds of utterances that are attested (and we may assume were grammatical) and various that are not attested (which we have taken to be ungrammatical). We list some of them in (75)–(87), providing constructed examples but attempting to be faithful to the actual syntactic patterns found in Early Modern English texts. For the attested cases, we use the simple labels introduced and explained in earlier sections. For the unattested patterns, we give a short-hand identification of the source of the ungrammaticality according to the analysis argued for above, with an initial asterisk expressing ‘ungrammatical due to’.

Each of the B-variants above has a certain plausibility given the inventory of grammatical elements and their uses in Early Modern English. Yet only some of them are attested and the question therefore arises why the pattern of attestation is the way it is. We have tried to show that some success can be achieved with an answer hypothesising that INOPO originated from an intonation question consisting of response particle no followed by an elided TP which was at some point used as a model for the formation of a corresponding syntactic question, in which the operator moved to C and therefore escaped the ellipsis of the TP. We considered two alternative approaches but these are not entirely successful in accounting for the data. Further approaches might be taken, where no is analysed not as a response item but as a (negative) question particle or perhaps as a variant of the clausal negator not, but it is not clear to us that these would achieve a better fit with the data.

One pertinent fact remains to be addressed: INOPO, as exemplified in (76B), disappeared in the second half of the 17th century and is clearly ungrammatical in present-day English. We leave it to further work to provide an explanation for the loss of this option from the language but, as a start on this, we will here point at some possible directions of thought on this matter, also briefly commenting on their initial plausibility. They are all based on the working hypothesis that the loss of INOPO is related to other changes taking place in English in the relevant period. Since INOPO is a negative interrogative response type, it is natural to look for possible connections between INOPO’s demise and changes in negation, question types and responses at the time.^[17]

As for changes in negation that may be linked to the demise of INOPO, one important development taking place in the relevant period is the shift that the negator not undergoes from phrase to head status. As is well-known (see Ingham [2013] for a description), not/nought originated in the expression no wiht ‘no thing/no bit/not at all’, which gradually developed into the single word nought/not and later, when following an operator, became an enclitic (or even inflection, if we accept Zwicky and Pullum’s [1983] proposals) in the form n’t. Kroch (1989) and others interpret that final stage as involving a shift in the category of not, from phrasal specifier of NegP (or perhaps adjunct to VP) to head of NegP. They date this development to the second half of the 17th century, which is exactly when INOPO stopped being used. This means there is temporal alignment of the two developments and, since INOPO always contains an operator and the negator not (though this not is usually deleted, along with the rest of the TP), there is some prima facie plausibility to the idea that the two changes are connected. But then the question arises: why should the emergence of forms like can’t, won’t, and don’t lead to the loss of responses like No can?, No will? and No do?, i.e. to the loss of INOPO? Such a connection looks all the more difficult because, as we saw in Section 3, there are examples of the type No can’t? and No would not? (the latter being interpreted as representing contracted forms like [wudnt], as Rissanen [1999a] proposes). It therefore does not look as if the two constructions are in complementary distribution, which means that the prospects of deriving the loss of INOPO from the development of contracted negative operators may appear to be somewhat dim.

However, in the spirit of setting out directions of thought, let’s consider a possible line of approach that could conceivably be pursued. Even if there are obstacles, examining them will help throw into relief several facts and considerations to be taken into account in further work on INOPO. The point of departure could be an observation relating to example (21), the relevant part of which we repeat in (88).

In his edition of Sanderson’s works, Jacobson (1854: 360) notes the following about this example:

“Editions subsequent to the second have a comma inserted: ‘No, is?’”

The second edition of the relevant set of sermons was published in 1660 (Jacobsen 1854: vi) and, as we have seen, this is the time at which INOPO stops being attested. It could therefore be surmised that insertion of this comma in later editions (e.g. in 1671, 1673, 1686 – the latter also included in EEBO) reflected uncertainty about the acceptability of INOPO on the part of the editors/typesetters. Specifically, they may have felt that No is as one intonational phrase was uninterpretable and have tried to remedy this defect by means of the comma, suggesting there were two intonational phrases. This would assimilate the use of no in (88) to its well-established use as a bare interrogative response item, i.e. the INO discussed in Section 4.3. It would still leave is as a rather odd unit on its own but at least the repair action would free up the word no, allowing it to fulfil its familiar function and mark the utterance as a negative interrogative response.

Interpreting the insertion of a comma between no and the operator in this way, as an attempt at rescuing a sentence that no longer made sense once INOPO had fallen out of use, it is striking to see that, among the 65 attested examples of INOPO, four cases in fact have such a comma.^[18] Even more strikingly, all four of them feature an added not after the operator, as in (24), repeated in (89).

The three further examples like this are from 1616, 1647 and 1654. Although there are also two examples of a negated operator without a comma, given in (22) and (25), the patterning might be taken to suggest that, by the first half of the 17th century, cases with no + operator + not/n’t were in the process of becoming less than fully acceptable, leading some editors/typesetters to take remedial action by inserting a comma.^[19] This would suggest that fronting of operator + not/n’t (the first of which we provisionally continue to interpret as also representing contraction of the operator and the negative marker, following Rissanen [1999a]) was somehow blocked in INOPO. To see what the offending feature might be, consider the structure in (90), which shows the fronting of the contracted form consisting of operator + negator which moves it to C⁰ and enables it to escape from the TP before that is deleted.

We have seen that response no in this period would usually take part in negative concord (negative agreement) with an element in its following clause, in the manner shown in (68a). In (90), this agreement would be between no in SpecCP and the contracted operator in C^o. A clue to what might be wrong in this structure could come from a comparison with the (modern) Spanish negative concord pattern seen in (91), from Batchelor and San José (2010: 417).

Sentence (91a) illustrates the usual pattern in Spanish, where clausal negation is expressed by the preverbal marker no. The presence of the negative adjunct nunca ‘never’ later in the clause does not make a difference to this. However, if nunca is in the preverbal field, as in (91b), use of the clausal negative marker no would yield an ungrammatical sentence. An interesting analysis of this contrast is developed in Tubau Muntañá (2008: 224–231), based on the notion of Spell-out domains. The guiding idea is that no and n-words like nunca ‘never’ should not be in the same domain. She proposes that (91a) has no and nunca in separate domains, with nunca being in vP and no in a higher domain, but that in (91b), with fronted nunca, they would both be in the same (higher) domain and this state of affairs is blocked. Hence, no cannot surface.

As pointed out in Ortega-Santos (2018: 320–321), Tubau Muntañá’s (2008) proposal fits in with other phenomena where elements of a similar nature are barred from being too close to each other, which can be captured under the general notion of anti-locality. A further case of this type may be represented by English (90), with the relevant elements being the response item no and the fronted negative operator. Their relation is of the head-specifier type, which is clearly very local. Even if it turns out that further functional projections need to be added to this admittedly very simple structure, it would be plausible to maintain that the two elements are still in the same local domain, which means that the necessary agreement relation is blocked by some form of anti-locality. As a result, INOPO would not allow the use of contracted negative operators and it could be argued that this was the reason it went into decline and was lost. Note that this line of reasoning would commit us to the assumption that, in the highly colloquial and interactive varieties of spoken English that were the typical habitat of INOPO, fronting of operator + n’t was starting to become the norm in elliptical negative interrogatives after around 1650.^[20] It would have been natural for INOPO to share in this development, yielding no + operator-n’t. But syntactic principles, specifically anti-locality, made this option ungrammatical. If this turns out to be the correct approach, it could be said that, indirectly, INOPO fell victim to contraction.

An important issue to consider in the further assessment of this hypothesis that is still needed comes from data presented in Nakamura (2018). He shows that contracted forms with two final consonants (as in can’t and won’t) gradually start appearing in the historical record from around the year 1600, while forms with three consonants (e.g. isn’t and didn’t) first make an appearance only 100 or more years later. He attributes the difference to phonological factors, with CC clusters being less marked than CCC clusters. But if contracted forms were already in regular use in the 16th century, as Rissanen (1999a) argues, we would not expect the gradual adoption of contracted spellings one century or more later to be regulated by phonetic factors. Nakamura’s findings also conflict with the idea that, after 1650, use of contracted forms became the rule in some types of negative questions. If forms like isn’t and didn’t did not yet exist at that time, their use can obviously not have become compulsory in any context. However, use of the INOPO construction might already have been avoided when there were as yet only a few contracted auxiliaries in existence that caused problems for it, since they made the entire construction syntactically unpredictable. It may also be possible to reconcile the proposals of Rissanen (1999a) and Nakamura (2018) by assuming that operator-negative contraction was a two-step process, whereby operator + not first developed into a single unit composed of operator-[nǝt], with reduction of the vowel (perhaps represented by unitary spellings like didnot, nynnot [i.e. ne-will-not] and ninnat, cited by Rissanen [1999a: 196]), with a subsequent second step then leading to the modern forms with complete loss of the vowel, as in [dɪdnt] etcetera. If the reduced-vowel variants represented a unitary syntactic element, they would be eligible for operator fronting in interrogatives. But if anti-locality blocked this in INOPO, this could have led language users to refrain from using this construction. Such avoidance did not cause problems of ineffability: the use of bare interrogative no was and remained possible and another option had become available in the other-speaker use of the tag questions that had arisen in the 16th century (see Tottie and Hoffmann 2009). Thus, while (92B), with a structure as in (90), may not have been a viable option, both (93B) and (94B) were.

The above account, then, represents one possible line of approach to the loss of INOPO. It is rooted in the assumption that this loss is tied up with independent changes taking place in the system of negation in English at the time. This idea is attractive enough in itself but some explanatory acrobatics are needed to get the story to work and fuller examination of the idea will be required to establish its plausibility. One specific issue to be addressed arises from the following question: if insertion of a comma in examples like (89) represented editorial or compositorial action, what was the original in the author’s copy text? It might be thought that that would have been the sequence No + operator + not, without a comma after no. But, under the account sketched above, such a sequence would be an example of the structure in (90), which should be blocked by anti-locality. A way out might be to slightly adjust our reasoning above and hypothesise that cases like (89) resulted from action by editors/compositors whose internalised grammar did not allow INOPO at all anymore. When faced with an example of INOPO like No will? in a copy text, they might feel that not was missing. Insertion of this word would yield No will not? (or No won’t?), which was interpretable as a case of interrogative no (with a deleted TP, as in [50B]), followed by a clause with VP-ellipsis and also (somewhat exceptionally) subject ellipsis. This could be what produced examples like (22) and (25). A further remedial step may have been the insertion of a comma after no to explicitly mark the presence of an intonational boundary, leading to No, will not?, which can be seen in (23) and (24). Although the number of relevant examples is small, these are some of the questions that arise and answers that could be tried out for this particular issue. Further issues that would require attention in the negation-based account of INOPO’s loss include the precise nature of the anti-locality that might be at work in structures like (90) and the diachrony of the use of contracted operators in elliptical questions in general.

A different approach could try to relate INOPO’s demise to the loss of direct polar questions introduced by whether, as in (95), repeated from (32).

After having existed since Old English times, this construction becomes less and less frequent in the Early Modern period and “is gone completely by the eighteenth century” (van Gelderen 2013: 9). We argued in Section 4.2 that sentences like (95) or variants of it played no direct role in the emergence of INOPO. But once INOPO existed, there were certainly connections between the two constructions. Both were main-clause polar questions introduced by an element signalling polarity: neutral for whether, negative for no. It is therefore plausible to assume some degree of shared structure and it is not unnatural to think that it was a change affecting an element of that shared structure that led to the loss of both constructions. The precise nature of that element would need to be identified in further work, but we will here offer some initial observations, focusing on the exact type of whether clause involved, its properties and its diachronic development.

A main point to note is that the relevant whether construction involves polar questions. There was a different main-clause use, attested in earlier centuries but also obsolete today, in which whether introduced an alternative question, as in (96) and also example (iii) given in footnote 10.

This quotation comes from the table of contents at the start of Twyne’s book. The title of each of its 362 sections takes the form of a main-clause interrogative. Eight of them are alternative questions and all of these are introduced by whether, (96) being one of the cases. Walkden (2014: 144–155) makes clear that the alternative-question use must have preceded the polar use of whether and was in fact a first step in the development of whether from its original meaning ‘which of the two’ to its later use as an interrogative marker (see also OED, s.v. whether, pron., adj. [and n.], and conj.). Since none of the examples of INOPO has the syntax or semantics of an alternative question, any connection with or influence from whether would have to lie in its polar interrogative use.

An objection that might be raised to the idea of a shared decline of main-clause polar whether and no questions could come from the difference between them with regard to possibilities for ellipsis, as mentioned in our discussion above. In (97) we summarise the patterns in which each is attested.

From this array, it looks as if the two constructions were in complementary distribution, suggesting some fundamental structural difference and therefore a lessened likelihood of their decline springing from a shared cause. However, there are reasons to think the arrangement in (97) reflects patterns of usage rather than patterns of grammar. We noted above that the restriction of interrogative no to use in elliptical utterances could be because non-ellipsis would result in somewhat unnatural repetitiveness from an interactional perspective. Main-clause interrogative whether, on the other hand, may have resisted ellipsis because of the specific rhetorical function that it seems to have had. Eckardt and Walkden (2022) carried out semantic-pragmatic analysis of the Old English examples of the construction and come to the conclusion that it was used for ‘pedagogical questions’, in ways that they specify in detail. Similar – though more impressionistic – comments about examples in Middle and Early Modern English are made in Rissanen (1999b: 275) and references given there. It is also difficult to see what kind of syntactic principle would have ruled out the elliptical uses as in (97b,c) for whether and the non-elliptical use of (97d) for no. If this is correct, it could be concluded that the two forms were indeed grammatically similar, in their position and general function. We also saw in (29)–(30) that, in some cases, the two items could occur in coordination with each other, which reinforces the idea of grammatical similarity and potentially the idea that the loss of INOPO was part of a somewhat wider change, affecting main-clause polar interrogative whether as well. The facts in (97) would not pose an insurmountable obstacle to such an approach.

Another potential problem that could be raised for assuming a link between the decline of INOPO and of main-clause polar whether questions has to do with their relative diachrony.^[21] The last example of INOPO that we have been able to find, given in (21), dates from 1656. Yet van Gelderen (2014: 20) gives three apparent instances of main-clause whether dating from the first half of the 18th century which might be taken to cast doubt on the idea of a no-whether link. However, closer examination of these examples, reproduced in (98)–(100), suggests none of them are direct polar whether questions.

The sentence in (98) starts with three coordinated interrogative clauses, the first two headed by whether and the third one by how. If the whether clauses are interpreted as direct questions, this would also have to be done with the how clause. But the absence of inversion in that clause suggests it is an indirect question and the continuation of the sentence shows that – though the construction is complex – the how clause is to be interpreted not as a main clause but as a dependent of the noun inquiry (which perhaps needs to be reconstructed in front of the how clause). Given the clear parallel between the three interrogative clauses, this is then also how the two whether clauses in (98) can be read. For the cases in (99) and (100), it is important to note that they are alternative rather than polar questions, which makes them unlike the (polar) INOPO construction. The OED (s.v. whether, pron., adj. [and n.], and conj.) gives examples for the alternative-question use of whether (sense II 4a) up to the early 19th century, which suggests that such cases are not linked to INOPO chronologically – but as we noted above, their different syntax and semantics make such a link unlikely anyway. For direct polar questions with whether (sense II 5), the last example given in the OED is from 1588 but our corpora contain examples up to the 1650s (see footnote 10). Thus, our data explorations, though by no means exhaustive for the whether cases, have yielded attestations of both INOPO and direct polar whether up to the second half of the 1600s but not beyond, which means there may be enough chronological correspondence to consider the possibility of shared causation of their loss. We have also suggested that there is enough functional similarity to posit the existence of a link and this then is another route that could be pursued in further work attempting to pin down the reason(s) for INOPO’s demise.

A final route that could be explored is the idea that INOPO’s demise might be related to changes in the use of the response item no itself. Given no’s crucial role in the INOPO construction, this would in fact be the most natural hypothesis to examine. In Section 5.1 above, we saw that there was indeed a specific change that affected response no in the course of the Early Modern period: it switched from exclusive use after negative utterances to general use after both negative and affirmative utterances, which is the way no is still used today. Given the undoubted existence of this change in the use of no, one might try to capitalise on it in trying to account for the loss of INOPO. In such work, the main issue would be to pinpoint what exactly changed in the syntax of no. We saw in Section 5.1 that current syntactic thinking, as embodied in Holmberg (2016), attributes the modern system to no having dual status, such that it can be involved in either negative agreement (when used in response to negative utterances) or negative valuation (after affirmative utterances). Under that view, the emergence of this duality would involve an addition but no restructuring or loss: when the new derivational option became available in which no assigned negative value to the polarity of the associated clause, the already-existing option of no undergoing negative agreement continued being part of the system, without structural change. However, the loss of INOPO, which represented a case of agreement between no and another negative element, might be taken to suggest that some adjustment to this reasoning is needed.

A first step in developing such an approach would have to be the gathering of more detailed empirical data for the use of yes/no and their variants in the Early Modern period. There has in recent years been some focused interest in the form and use of responses in present-day English (see e.g. Goodhue and Wagner 2018; Meijer 2022; Tian and Ginzburg 2017; Wiltschko 2017) and this has led to greater understanding of several previously unexplored patterns. Using Early Modern data, there are of course limitations on what can and cannot be examined but, as we hope our description of the INOPO and IYEAPO constructions demonstrates, in this case as in others, the more carefully one looks, the more one can expect to see. The relative lack of data means that, currently, it would be hard to advance any detailed suggestions as to how the demise of INOPO could be made to follow from the changes to no as a general response item. Nevertheless, it is plausible to assume a link of some kind and, if nothing else, the INOPO facts that we have presented add to the data that will have to be taken into account when it comes to developing a model for the wider changes that yes/no have undergone.

Altogether, while the demise of INOPO does not affect our central findings about the structure of elliptical utterances with response no, it is certainly a topic that deserves attention in its own right. As pointed out in work like Hundt (2014: 169–171), Leung and van der Wurff (2018: 163–164, 176–177) and Kranich and Breban (2021), the loss of syntactic options is a field that has been somewhat neglected over the years, which means that tools and theoretical approaches for exploring it are as yet rather underdeveloped. Above, we have mentioned a few possible avenues that could be taken in further work trying to explain the loss of INOPO, also making some initial observations about their prima facie plausibility. But these should be regarded as tentative pointers, not fixed signposts.