The grammaticalization and grammatical meaning of a ‘do’ support verb in the northern Italian Camuno dialect

1 Introduction

An auxiliary verb usually has a main verb counterpart from which it was derived in the course of language change. Typically, in the first stages of the grammaticalization process there are two main verbs: the auxiliary-to-be, which is a support verb with lexical content, and its supported verb (e.g. Heine 1993).^[1] At this point, the meaning of the construction is the semantic sum of two recognizable and separate predicates. When grammaticalized, the construction takes on a meaning as a whole. At that final stage, if the construction is ‘dissected’ for analysis, the support verb is a functional remnant and its formerly unique, lexical content is gone.

This kind of grammaticalization pathway is epitomized in the history of the use of ‘go’ as an adaptive support verb then future auxiliary, a pathway common in both Germanic and Romance languages, among others. In the modern languages, both the more ‘primitive’ and ‘advanced’ stages coexist, as illustrated here in (1), (2), pairs of English and French sentences with identical meanings.

There are numerous other instances of the same source verb adopted by different languages ending up performing the same grammatical function, as amply demonstrated in the compilation of Bybee et al. (1994). These authors show a general, cross-linguistic tendency for a support verb to lose its lexical and concrete components, leaving behind the functional, abstract and relational parts. In most instances, the functional remnant in the grammaticalized support verb provides an aspectual or modal function to the resulting construction.

The support verb ‘do’ in standard English is then something of an enigma, as although it carries inflections for tense, person and number, it does not directly express aspect or modality. A connection to habituality may be largely by default, as non-habitual change uses the progressive (with auxiliary ‘be’).^[2] In fact, it is not clear what, in a functional sense, do actually ‘does’.^[3]

In generative syntax, the function attributed to do-support (DS) is that it is purely part of the mechanics of sentence construction, required to prevent the derivation from ‘crashing’. English, it is said, requires insertion in T of a ‘dummy’ do verb (to use Chomsky’s famous 1957 terminology) to pick up tense affixes so that the lexical verb can remain in the verb phrase. This contrasts with the situation in French and other Romance languages where the lexical verb raises (Pollock 1989). The support verb do can then be understood as little more than the ‘spellout’ of the T syntactic projection. A consequence of using an auxiliary is that the lexical verb and complement remain adjacent in all syntactic contexts: interrogative and negative declarative, as well as the affirmative declarative. This is characteristic feature of the modern English language.

The concensus is that the do of present-day English DS is vacuous and without lexical content as it is used without semantic restrictions on the type of supported verb. This is shown for the affirmative declarative in (3) with the stative verb ‘know’. Yet in this example, where do is syntactically optional, and being used as an intensifier to counteract an expectation of the opposite, the word itself also seems to be contributing meaning to the construction. Similarly in (4), where there is no such prior expectation, do is used emphatically to indicate that the degree of ‘liking’ is above normal.

Much has been written about the history of English DS: its origin, and its gradual rise in all syntactic contexts between 1400 and 1700, and subsequent fall for affirmative DS post-1575 (Ellegård 1953); there are many disagreements, and perhaps some false trails. Despite much high quality scholarly work, doubts remain about the use, and therefore meaning, of the early construction, both as a whole, and (by inference) of the support verb do, and how these changed through time. Yet if the grammaticalization of an auxiliary verb is indeed so predictable (at least in its general principles) as Bybee et al. suggest, it is worth searching for a pathway for ‘do’ in an alternative, living language. It is this that the Camuno dialect of northern Italy can provide us.

Camuno is an example from the Romance family where ‘do’-support with the verb fa (a reflex of Latin FACERE) is used to make the interrogative, in a phenomenon referred to here as fa-support, or FS (5). The support verb fa inflects for tense, person and number and bears a subject clitic that also references the subject person, number and gender. The following supported verb is in infinitival form.

As in Camuno (and Romance in general) the lexical verb raises past adverbs out of the verb phrase (see examples (6) to (9) below), unlike in English, the function of ‘do’ cannot be to keep verb and object adjacent.

Camuno is spoken in Val Camonica, a relatively isolated valley that leads from the Alpine foothills down to the Po Plain. In dialects higher in the valley, such as the Monno dialect described by Benincà and Poletto (2004) (B&P), FS use is obligatory with almost all verbs and it serves all question functions. However, in dialects lower in the valley (as demonstrated by this author: Swinburne 2021), FS is optional and co-exists with a form based on the main verb only. In these ‘optional FS’ dialects, FS use is most common with activity verbs, there being little use with stative verbs. Furthermore, when FS is chosen in preference to the ‘default’ question form, the speaker is making a ‘special’ type of question, with specific linguistic and social meanings; the alternative is used purely to request information.

It will be suggested here that represented within the valley dialects from south to north is a grammaticalization cline, with speakers’ understanding of ‘do’ transitioning from one of a lexical ‘do’ to a dummy, meaningless ‘do’. Furthermore, this synchronic change is also representative of the pathway of diachronous change.

The general questions about ‘do’-support to be asked, and then answered for Camuno, are these:

1. What is a possible pathway for a DS construction that begins with a fully lexical ‘do’ verb and ends up with a largely meaningless auxiliary?

2. How is the lexical content lost and how does that affect the meaning of the construction?

3. What grammatical meaning can be attributed to ‘do’ in the final construction?

This paper must necessarily focus on only certain aspects of the phenomenon. The account here concentrates on the lexical content of fa as deduced from the probability of FS use with different types of support verbs, in one of three experimental phases of investigation. A fuller discussion of all three phases (P3, P4, P5) with a greater semantic (and syntactic) variety of verbs is presented in Swinburne (2024). Information on the special FS question meaning is summarized here but discussed in detail in Swinburne (in prep. a, b). In addition, the relationship between the degree of grammaticalization of FS and the valley geography must also be left to another work (Swinburne in prep. c). However, a preliminary account of all these aspects (based only on the P3, P4 fieldwork) can be found in Swinburne (2021). There is also no space here for any detailed comparison to English, which requires a separate treatment.

The following account begins in Section 2 with a brief description of Val Camonica, outline of the syntax of FS in the Camuno dialect and summary of the special meaning attributed to the FS question. Section 3 provides a description of the elicitation experiment used to measure the probability of FS use with different semantic verbal classes. Section 4 describes how the overall outcome of the experiment was that, in dialects where FS use is optional, the likelihood of its use reflects the degree to which fa is compatible with the subject thematic role of that verb in the order: agent > theme > experiencer. Section 5 characterizes FS use by speaker and groups speakers with a similar degree of semantic selection into those with a low-use ‘pragmatic’ FS, medium-use ‘agentive’ FS, and very high-use ‘generalized’ FS. This aggregation of speakers allows for sufficient tokens for FS use to be charted by verb. In Section 6, sequences are aligned in an order that is considered to represent the direction of grammaticalization (from ‘pragmatic’ to ‘agentive’ to ‘generalized’), and what is revealed is a pattern of extension of FS use across different semantic types of verbs. Section 7 connects the lexical extension of the construction to the loss of its special meaning. It also considers and rejects an alternative interpretation that what is portrayed is a degrammaticalization sequence. Section 8 sums up the importance of Camuno fa ‘do’-support in providing us with a sequence of the changes to ‘do’ that occur during the grammaticalization of ‘do’-support.

2 fa-support in Val Camonica

Val Camonica is the valley of the River Oglio and it extends from the Alpine foothills southwestwards to Lago d’Iseo, and out onto the Po Plain between Bergamo and Brescia. The valley dialect is Camuno, which is a subdialect of Bresciano. Within Camuno there is a small amount of variety between the dialects of the individual communities in phonology and lexis and in the presence and use of FS, but otherwise little relevant morpho-syntactic variation. The area in which FS is found today starts in the south at the small town of Esine within the so-called Middle Valley (MV), and reaches into the Upper Valley (UV) as far north as Monno and east to Vezza d’Oglio, as outlined on Figure 1.^[4]

Figure 1:

Map of the distribution of fa-support in Camuno dialects, 2017–2022 (as two parts, a on left, b on right).

Camuno is similar to standard Italian in most respects, such as in the complexity of morphology on the finite verb and the position of the verb with respect to aspectual and temporal adverbs (B&P; Swinburne 2021: Chs 3, 4). As with most northern Italian dialects, Camuno differs from standard Italian in the presence of subject clitics.^[5] An interrogative is usually formed by inverting the order of main verb and subject clitic (subject-clitic inversion – SCI) compared to the order in the corresponding declarative. Thus the declarative (6), (8), has an order of scl-main verb, and the interrogative (7a), one of main verb-scl. In an interrogative, the lexical subject is usually sentence-final; or alternatively sentence-initial, where it is interpreted as the topic.

FS, which is an exceptional form with a restricted geographical distribution, employs an ‘additional’ support verb (in the sense that it is not present in the corresponding declarative) to bear the subject enclitic (7b), (9). The resulting construction, FS, has the order fa- scl, thus resembling the English interrogative with support verb ‘do’ and its order of do-subject. In the following examples, (6), (7) with supported verb ‘drink’, are from Esine, where both interrogative forms are available, and (8), (9), from Monno, where only FS is possible.

Syntactic observations made by B&P in the 1990s on FS in the Monno dialect were verified by this author throughout the valley, the only significant exception being that this author found FS could be used with both available modals. Importantly, there are no relevant differences in the (overt) syntactic structure of FS between dialects with optional FS, such as the Esine dialect, and those with (essentially) obligatory FS, such as the Monno dialect. The only constituents that can intervene between the two verbs are various aspectual adverbs and pragmatic particles: there is never a complementizer.

The most natural interpretation (and that of B&P) is that obligatory FS, and by extension also optional FS, is monoclausal – as is English DS. Despite that, the preferred structural interpretation of Swinburne (2021, in prep. a) is that at least optional FS has a biclausal control structure, a hypothesis based soley on the question meaning. As argued there, a biclausal structure explains how, in using FS, the questioner is asking for an opinion about a preconceived idea, the piece of tested information being syntactically embedded in a separate clause. Some examples of relevant question meanings are provided below. Whether or not obligatory FS should also be viewed as biclausal is much less clear. What can be concluded though, is that the part of the grammaticalization sequence reported here – which is the loss of lexical content to fa – is not correlated with any detectable syntactic changes. These structural considerations are not, however, relevant to the main arguments of this paper.

FS is used in both yes/no and wh-questions and it is available in the interrogative for all tenses which would otherwise lack an auxiliary: the present, future, conditional, and imperfect. The present tense, in Camuno as in Italian, can be used to refer to either an event happening ‘now’ – provided that is a short interval of time, otherwise a progressive which employs a ‘be’ auxiliary is used – or habitually, as in these examples. Therefore FS cannot be said to be associated with habitual aspect, nor imperfectivity in general, as its use was found less likely with the present tense, an imperfective use, than with the future tense, which (when used to describe a single event) is a perfective use (Swinburne 2021: 197–207).

A key point to note about Camuno FS compared to English DS, is that it is only present in the interrogative. All declarative instances of a support verb fa are of a different, causative verb (10). In a causative construction (in Camuno as in northern regional Italian), a causee is usually present, either lexically (10a) (Tunì), or with a clitic (10b) (la), or in some Camuno instances, both. The clitic in (10b) is accusative, as curì ‘run’ is an intransitive verb (and has switched to la to avoid duplication of masculine al). With a transitive verb, the causee would be dative.

The examples below contrast non-causative interrogatives (11) with causative interrogatives (12). In Camuno dialects, a causative interrogative can be made in two ways: either by encliticizing the subject clitic to the causative verb (SCI) (12a′), (12b′), or by adding an interrogative verb (FS) (12a), (12b), (12c), the latter resulting in two fa morphemes: fa ‘do’ followed by fa ‘cause’.

The reader familiar with the literature on historical English (beginning with Ellegård 1953) may be wondering if, in the optional FS dialects, there is ever confusion between an interpretation of fa as ‘cause’ or ‘do’.^[6] This could potentially occur in a one-fa interrogative which could be either non-causative with FS (11a) or causative with SCI (12a′).

The answer is that, in Camuno, this rarely ever happens. Even if that distinction was not clear from the prosody or context, the forms are distinguished syntactically by either the presence of a named causee, and/or the position of a non-subject clitic: with fa ‘cause’ the clitics raise to the causative verb; with interrogative fa ‘do’ they are borne on the following infinitival verb. In instances where a clitic would not necessarily be present it can be introduced by doubling the accusative or dative causee; and where a causee is optional (such as in a rare, faire-par type causative), by doubling an object and extraposing it. In addition, in some dialects, such as the Bienno dialect (11b), (12b), (12b′) the two support verbs are also morpho-phonologically different, interrogative ‘do’ having an aspirated pronunciation as ha, but ‘cause’ (and the main verb ‘do’) with the labiodental pronunciation of fa. (See additional discussion and examples in Swinburne 2021: Ch. 3.)

Referred to above were the special meanings associated with the FS question, described in detail in other works. Many speakers who could, in their dialect, make a question with either FS or SCI, noted certain systematic differences in the meanings of the two variants, SCI being the ‘default’ form and FS having additional semantic and pragmatic meanings. One of these special meanings, that of presupposition of the answer, is illustrated here in (13) to (17) (for brevity with SCI then FS written as alternatives within the same example).

As the examples demonstrate, in describing the particular connotations of FS, speakers used terms such as ‘doubt’, ‘surprise’, or ‘seeking confirmation’, to indicate they had an expectation of a certain answer. In (14) and (15) this presupposition is negative (which is the more common direction), and in (13), (16) and (17) it is positive. Other special functions of FS (not illustrated here) are that the speaker is seeking the interlocutor’s opinion rather than factual information, or expressing their own emotions about the topic.

The examples above demonstrate that the same explanations are a) produced from speakers throughout the valley (Esine (15), Cividate (13), Berzo Inferiore (14), all in the MV; Megno di Lombro (16), Monno (17), both in the UV); and b) for all types of supported verb (agentive-subject, activity verb (13); theme-subject, change-of-state verb (14), (15); experiencer-subject, stative verb (16), (17)). This establishes: a) that FS is one and the same valley phenomenon; and b) that the addition of fa ‘do’ to the supported verb is not simply coercing the underlying verb into an agentive reading. The latter is evident in the speakers’ descriptions: in no instance do they indicate any intentionality or volitionality associated with FS but not SCI. (For instance, in (14), in both cases the leaves are just falling on their own; they are not actively falling.) Instead speakers are focused on describing what they are thinking when they utter FS.

One further syntactic/semantic trait is relevant here (Swinburne 2021: 56–60, in prep. a), the ungrammaticality of an FS question with a subject with arbitrary reference in an optional FS dialect (18) but not an obligatory FS dialect (19).^[7] In being able to ask, or answer, this particular question, there is no requirement that either party actually knows a 100-year old woman, or that one even exists in their community; it is therefore a hypothetical question. The significance of this for the relative meaning of fa ‘do’ in optional and obligatory FS will be returned to in the conclusions.

Besides the linguistic meanings described briefly above, FS is also a device to show interest and engage the addressee in conversation. This links to its social meaning. FS is a polite form, used to show respect to other members of the community who are not close family. It is a way for the speaker to make a tentative guess about what the addressee is thinking, while still being able to be wrong without losing face. If the arguments that FS has a biclausal structure and is therefore an embedded question are correct, this connection would be readily understandable, as indirect questions are naturally gentler and less confrontational than direct questions. It would also explain why FS, with its function of promoting closer relationships, would tend to be most favoured in the more tightly knit communities.

3 Elicitation experiment methodology

In investigating the syntax of FS in the MV dialects (originally by asking participants for spontaneous translations from Italian), it became apparent that FS was far more likely to be used with a verb that described an activity and had an agentive subject (e.g. curì ‘run’ or maià/mangià ‘eat’^[8]), than with a stative verb with experiencer subject (e.g. penhà ‘think’, or piadì ‘please, like’). This pointed to the existence of some lexical content within interrogative fa ‘do’. Speakers indicated that the dispreference for use with a stative verb was not because the combination was ungrammatical, but because it was hardly ever heard. The likelihood with which a speaker would use FS with a particular verb and verbal class, was therefore measured quantitatively using an elicitation experiment.

A summary of the method is presented here. There were three research phases that used this elicitation technique (P3, P4, P5) each with its own selection of verbs and investigating different areas of the valley (although there was substantial overlap between phases). Results reported here are from P5. In the experiment, participants listened to an oral questionnaire recorded in various local dialects. This consisted of a series of descriptions of situations, or contexts, each of which ended with a request for the participant to ask a question of an imagined interlocutor. In this way they were given the stimulus verb in declarative form and asked to rephrase the sentence as a yes/no interrogative. Example (20) provides an English translation of one of these contexts, with question request and elicited question also in the original MV dialect.

Elicited questions that measured the relative production of FS by verb were all: a) in the present tense (as it would have been too awkward to have consistently used the future, and the only available past, the passato prossimo, does not use FS because it has its own ‘have/be’ auxiliary); b) in a habitual sense for non-stative verbs (or speakers are tempted to use a progressive tense with ‘be’ auxiliary), c) were yes/no rather than wh-questions (or speakers are likely to cleft the wh-item, again using ‘be’ in the main clause); and d) used the 2nd person for all verbs with human subjects (as question requests based on the 3rd person are much more difficult to understand).

In the P5 set of questions that produced the results described here, the result for each verb is an average of four contexts/question-requests.^[9] This ‘averages out’ the effect of the context as FS is more likely to be produced according to what the speaker knows (factually), or thinks (emotionally) about the situation presented. (In fact all of the contexts are in some way presuppositional or encourage emotional involvement in order to increase the likelihood FS is produced.) In the resulting oral questionnaire one token of each verb was allocated to a block, with the order randomized within the block. This way if the participant tired, it would not disproportionately affect any particular verb.

The verbs investigated are divided broadly into four main categories, grouped here according to the verb’s subject thematic role, as previous phases of the experiment indicated that this was the primary determinant of FS use (Swinburne 2024).^[10] Verbs classified as having theme rather than agentive subjects are those which are syntactically unaccusative in Camuno (as in Italian) and take the auxiliary ‘be’ rather than ‘have’ in the passato prossimo tense.

• Agentive-subject verbs: mangià/maià ‘eat’, laà-do ‘wash’, hcrìer ‘write’, lidì ‘read’, cumprà ‘buy’, vinhì ‘win’, mitì ‘put’, hercà ‘search’, troà ‘find’, fà ‘do’

• Causer-subject verbs: dervì ‘open (trans)’, dà ‘give’, fà (anim) ‘make’

• Theme-subject verbs: rüà ‘arrive’, htà (a cà) ‘live, stay (at home)’, gnì-ho ‘grow’, dervìh ‘open (intrans)’

• Experiencer-subject verbs: penhà ‘think’, piadì ‘please, like’, cugnohì ‘be-acquainted-with’, pudì (ability and request) ‘can, could’

4 Relationship between FS probability and verbal semantic class

In all three phases of the elicitation experiment, results from speakers with optionality in FS use, always produced the same pattern (Swinburne 2024). Described using the (normal) subject thematic role of the supported verb, the likelihood with which a verb would be used with FS was:

agent > theme > experiencer

Causer-subject verbs behaved in two main ways. Verbs which include a causing action in their semantics (so the subject is a causer-agent), such as dervì ‘open (trans)’ or rumpì ‘break (trans)’ (a P4 verb, so not listed here), and which have an anti-causative, reflexive, theme-subject counterpart, dervì-h, rumpì-h, patterned among the agentive-subject verbs. Verbs without such an obvious causing action, such as gnì-ho ‘grow’ and rüà ‘arrive’, patterned with the theme-subject verbs, or at the top of the agentive-subject verbs, the precise position depending on how much the verb conveyed the notion of activity (particularly of motion) pragmatically.

As main verb ‘do’ is an agentive-subject, activity verb, the inevitable conclusion was reached that the probability that a particular verb would be used with FS was then a reflection of the degree to which it resembled ‘do’ in its internal lexical semantics.

5 Characterization of FS use by different groups of speakers

Within dialects where FS is an optional dialect trait, there are varying degrees of optionality. Going south to north up the main Oglio valley, FS becomes more commonly used: it starts as a rare pragmatic effect, becomes the main interrogative method, and ends as (with most verbs) the only grammatical option available.

It is contended that this synchronic, geographic progression of dialects reflects their diachronic development: it is the pathway of extension of FS across the different types of supported verbs, and of grammaticalization of the construction. Thus in a community such as Esine at the southernmost extreme of the FS area, FS is representative of the primitive stages of the construction encountered in the north at Monno. (An alternative vision of the sequence, with degrammaticalization from Monno to Esine, is entertained and rejected, in Section 7 below.)

To demonstrate the pattern of extension of FS across the different verbs, results from speakers who have a similar concept of FS must be aggregated. For this, a speaker’s use of FS is classified according to the degree to which, on average, they use FS with agentive-subject verbs and with experiencer-subject verbs (a classification made using results from all three phases of the experiment).^[11] Three labels then characterize FS use: ‘pragmatic’, ‘agentive’, or ‘generalized’. The labels refer to individual speakers, but are also characteristic of the speech of their communities.

The ‘pragmatic’ speakers mostly restrict their use of FS to the agentive-subject verbs, and with these there is considerable optionality. The degree of optionality is measured according to the percentage of instances with which they used FS with verbs of a certain class during the experiment, and this is taken to be representative of its use in normal conversation. For ‘pragmatic’ speakers, FS use with agentive-subject verbs is generally in the 40–60 % range, but could be less (although only results for speakers with >20 % use were analyzed), and by definition it is below 80 % use.^[12] In most cases there is almost no use, i.e. <10 %, with experiencer-subject verbs, and by definition it is less than 25 %. For the ‘pragmatic’ speakers, choice of form depends more on the pragmatics of the situation: what they are trying to express, and to a lesser extent who they are speaking to.

The ‘agentive’ speakers have very high use with agentive-subject verbs, usually around 90 %, implying that their choice is primarily based on the verb semantics, specifically subject type, and that there is no ‘room’ for the pragmatics with these, the most compatible verbs. Use with experiencer-subject verbs most commonly varies from almost none to up to 20 % (but for a couple of speakers, with the P5 selection of verbs, this reached 45 %). Question pragmatics still influences choice of FS when SCI is still a likely alternative.

The ‘generalized’ speakers have effectively categorial use with most verbs: agentive-subject verbs (>95 %); and most experiencer-subject verbs (the percentage varying with the selection of verbs used in the phase of the experiment). Apart from the categorical exclusion with (v)ei ‘have’ and èher/eser ‘be’ (Monno/Esine forms), optionality remains for all ‘generalized’ speakers with (v)ulì/olé ‘want’ and haì/saì ‘know’ (verbs included in the P3 and P4 selection) (with FS most strongly disfavored with the 2nd person forms) and sometimes pudì/possé, at least in its epistemic meaning, ‘is-it-possible-that, could’. Zone 4 speakers have, in addition, optionality with all forms of pudì/possé ‘can/be-able-to (ability), could (request)’ (and SCI may be strongly favoured), and limited optionality with others, e.g. penhà/pensà ‘think’.

In Swinburne (in prep. c.) the simpler three-fold division is further split into five: ‘pragmatic’ (Zone 1), ‘pragmatic-agentive’ (Zone 2), ‘agentive’ (Zone 3) and ‘generalized’ uses (Zones 4 & 5). Zones 1–3 are representations of the degree of optionality in ‘optional FS’; and Zones 4–5 of the degree of obligatoriness in ‘obligatory FS’ (a term which is obviously an approximation). The finer separation of Zones 4 and 5 requires the full suite of verbs from the combined three phases of the experiment. The fivefold division of dialects is then plotted on a map, emphasizing that it has a geographic reality. The progression of zones broadly follows the valley topography, going from south to north up the Oglio valley, as well as up the Grigna valley and down the Ogliolo valley, two tributary valleys. Zone 5, the most generalized form of FS is found in the UV at Monno, and in the MV at Prestine, which is at the top of the Grigna close to Bienno (see Figure 1).

The simpler, three-fold division in types of FS relevant to this paper is summarized in Table 1.

6 Grammaticalization of FS across verbs from optional to obligatory

Figure 2 presents three sequences showing the relative order of FS use for the 21 verbs investigated in P5 with the percentage of FS use decreasing upwards. The sequence on the left represents the ‘pragmatic’ speakers, in the middle are the ‘agentive’ speakers, and on the right the ‘generalized’ speakers. In each sequence the verbs are colored: agentive-subject verbs (red); causer-subject verbs (light and dark purple); theme-subject verbs (blue); experiencer-subject verbs (green).^[13]

Figure 2:

Grammaticalization of FS.

Lines joining the sequences match corresponding degrees in frequency of FS use. So while the ‘pragmatic’ speakers in the left hand column use none of the verbs over 75 % of the time; the ‘agentive’ speakers in the middle column are using approximately half of them this much (everything from blue rüà ‘arrive’ and below); and the ‘generalized’ speakers in the right hand column have almost categorical use (most with only 1 token of non-FS) for almost all verbs except two with experiencer subjects, penhà ‘think’, and pudì ‘can, could’.

This demonstrates how, for the ‘pragmatic’ speakers, with all eligible verbs there is a real choice of question form that is dependent on the precise nature of the question and the context in which it is made; but for the ‘agentive’ speakers, it is normal to use FS with agentive-subject verbs, so there is no longer a choice based on the pragmatics, but this choice remains with the theme-subject or experiencer-subject verbs; and for the ‘generalized’ speakers, with almost all verbs, there is no choice at all.

The most striking feature of the ‘pragmatic’ and ‘agentive’ speaker sequences in Figure 2 is that the order of verbal categories as shown by the colored blocks is almost identical. There is the same basic separation into a lower, red agentive-subject verb block and an upper, green experiencer-subject verb block. Within this there are also correlations between the purple causer-subject and blue theme-subject verbs.

Focusing first on the lower half of the sequences and on the verbs most likely to use FS, the ‘pragmatic’ speakers’ sequence is dominated by the red agentive-subject verbs but intercalated within this from bottom to top are the dark purple causative verb with causer-agent subject dervì ‘open (trans)’ (68 %), then the two light purple causative verbs with causer-but-not-so-obviously-agent subjects fà (anim = animate subject) ‘make’ (45 %) and dà ‘give’ (34 %). The same order of categories (although not of verbs within those categories) pertains for the ‘agentive’ speakers, although all are used at a higher frequency: lowest is dark purple dervì ‘open (trans)’ (88 %), and above this come the two light purple verbs dà ‘give’ (80 %), and fà (anim) ‘make’ (73 %).

Slightly higher in the sequence, note the correspondence between the two pairs of light blue, theme-subject verbs in the ‘pragmatic’ and ‘agentive’ sequences. Although the theme-subject attribution of all four verbs has been made on the syntactic basis that they all are unaccusative (taking a ‘be’ auxiliary in the passato prossimo), there are two semantic types. The lower pair of htà (a ca) ‘live’ and rüà ‘arrive’ are verbs where the subject could in some ways also be pragmatically interpreted as an agent carrying out an activity (‘live’ invoking the daily activities of life; ‘arrive’ as the motion required to achieve the arrival). The upper pair of gnì-ho ‘grow’ and dervìh ‘open (intrans)’, are verbs where the subject is a typical undergoer and is in no way responsible for any change. For both the ‘pragmatic’ and ‘agentive’ speakers, htà/rüà ‘arrive’ pattern lower and intercalated with the red agentive-subject verbs, while gnì-ho/dervìh are above all the red agentive-subject verbs and below (or slightly intercalated with) the green experiencer-subject verbs.

Comparing the ‘agentive’ speakers with the ‘generalized’ speakers, there is a coherent order within the different green experiencer-subject verbs (an effect not visible with the ‘pragmatic’ speakers where there was effectively no use, so no discrimination, between these verbs). For the ‘agentive’ speakers, cugnohì ‘be-acquainted-with’ is the first of these verbs to be used a significant amount (63 %) with FS, next comes piadì ‘please, like’ (24 %) then penhà ‘think’ (5 %). Within the ‘generalized’ speakers’ usage, the first two of these verbs, cugnohì and piadì have become largely categorical for FS use (94/95 %, one non-FS token remaining). The last verb, penhà (53 %) is used more by these ‘generalized’ speakers (who all represent a Zone 4) but its use is not yet categorical (although it is used categorically in communities investigated in earlier phases of the research, such as Monno and Prestine, both Zone 5). Similarly, the verb pudì ‘can, could’ is, among this selection of verbs and speakers, the last to be used with FS, and these ‘generalized’ speakers from Zone 4 rarely used FS with this verb. However, in Zone 5 areas, FS with ‘can, could’ is largely obligatory for at least the ability and request uses of the verb, although less so those of possibility (Swinburne 2021: 228, Fig. 8.6).

7 A ‘do’-support grammaticalization pathway

The part of the grammaticalization process attested in the valley dialects is that of lexical extension, or semantic generalization (Bybee and Pagliuca 1985). It is preceded by a stage where FS use increases within the core group of activity verbs until it is effectively categorial. At that point, any further increase must necessarily involve extension to other groups of verbs, firstly change-of-state verbs, then stative verbs. As shown on Figure 2 (38 % > 70 % > 91 %), the process results in an overall increase in frequency of FS use.

Along with an increase in its use, the special meaning of the FS question becomes less apparent. Thus for the ‘pragmatic’ speakers, who use FS relatively infrequently, the special properties of presupposition, speaker emotional involvement, opinion seeking, etc. are clearly noticeable – although in addition, the FS question is also being used to request information. With the ‘agentive’ group, in using FS, although the speaker may be intending a special meaning, they may just be asking an informational question. For the ‘generalized’ speakers, who have no choice but to use FS with almost all verbs, a unique FS question meaning is rarely apparent. Thus as FS becomes the new ‘normal’ so its particular meaning becomes less and less ‘special’.

As FS is a device to build relationships, show respect, and make courteous inquiries, it is particularly advantageous in the more isolated, smaller communities. Being also the socially favoured form, helps drive the increase in FS use and thus its grammaticalization.

This description is typical of accounts of the grammaticalization process in the literature (e.g. Bybee 2017). Grammaticalization produces an increase in frequency of use, and is itself likely driven by increasing frequency. Initially the new form is preferred because it is more ‘expressive’ (Hopper and Traugott 1993: 65; Lehmann 2015 [1982]), or ‘extravagant’ (Haspelmath 1999). This results in an extension in use of the new form to semantic contexts which are increasingly distant from that of its starting point (as with the progression of the be-going-to future from motion to non-motion contexts). With the overall increase in frequency of the new form, so it becomes over-used and fails to ‘stand out’ as effectively. The account here also adds a missing ingredient: the connection between the linguistic and social meaning of the form, and hence the reasons behind the frequency increase.

Before concluding, it is worth stepping back and considering an alternative interpretation of the sequence presented in Figure 2. It has been alleged in the literature that the unidirectionality of a grammaticalization process is often assumed without independent evidence (see Newmeyer 2001 and discussion in Haspelmath 2004). After all, in the absence of historical information to show which form was first present (the largely functional fa, or the lexically rich form), the direction of change in FS is inferred and not directly demonstrated. It is also conceivable that Figure 2, instead of representing a grammaticalization process running left to right, is in fact a degrammaticalization sequence from right to left.

An apparent problem with the grammaticalization version is that it seems to be at odds with what is known about the preservation of archaic dialects. Under the grammaticalization scenario, the ‘pragmatic’ form, such as found at Esine (in the south of the MV), is the more primitive in terms of its degree of grammaticalization, and the ‘generalized’ form at Monno (in the north of the UV), is the more advanced. This is despite the fact that Monno is the more isolated location, so it would be expected to be the more conservative dialect (which was no doubt why B&P originally chose it for their research). Under a degrammaticalization scenario, a ‘dummy’ ‘do’ verb used for syntactic purposes (even though it is not obvious why it would be needed), was adopted in all valley dialects and later re-lexicalized. Under this scenario, the Esine dialect would have a form ‘contaminated’ by the recent influx of speakers of non-FS dialects that occurred with valley industrialization in the post-World War II years.

The linguistic argument that grammaticalization rather than degrammaticalization is shown by Figure 2 runs as follows. If speakers from Esine and other central MV communities had re-acquired an SCI form in conversing with speakers of non-FS dialects, their FS would be still be of a ‘generalized’ type equally available for stative and non-stative verbs. The preference for use with activity verbs (which attests to the strong sense of ‘do’ in fa), and special meaning to the question, would then be a secondary development. Had this happened, it is very unlikely that the ‘rebuilt’ verb would have the same content as the original fa ‘do’ (as ‘what’s gone is gone’). Likewise, if the particular special meanings of the FS question were due to reintroduction of lexical content to fa, there is no reason why the same special meanings should be found in all communities (both south and west of Monno, in optional FS areas separated by ones with obligatory FS). Furthermore, there is no likely linguistic mechanism for re-introduction of semantic content to fa.

Reasons for rejecting the degrammaticalization hypothesis independently of the above considerations have to do with sociolinguistic meaning of the form and patterns of communication between different valley communities. Three reasons are relevant here: a) There is no need for speakers of optional FS dialects such as the Esine dialect, to accommodate their speech by using SCI to converse with speakers of SCI-only dialects – and they claim that they do not: speakers report that Camuno dialects are all mutually intelligible and FS is understood by those for whom it is not a trait of their dialect. b) Both FS and SCI are used by Esine speakers who grew up in the 1920s before the main phase of industrialization took place (although Esine was always a zone of through traffic). This demonstrates that the FS/SCI opposition in the Esine dialect is an old one. Lastly: c) In some communities there are gender differences, FS having been generalized by the women but not the men. Accepting that women are most likely to lead in language change (Labov 1990a), this suggests that the recent direction of change is towards grammaticalization of the form.

It is probably best to think of both the Monno and the Esine FS form as ‘old’, but the less grammaticalized form found in the contact zone with other dialects, and the more grammaticalized form in a remote enclave.

8 Changes to ‘do’ during grammaticalization

Although the descriptive terminology used here has been to portray the progression as an expansion in the range of subject thematic roles used with fa, this must ultimately be related to a change in the semantics of fa. ^[14] There are two ways to explain how fa can combine with a non-activity verb. The first, that the meaning of the supported verb is somehow coerced was discussed briefly in Section 2 and dismissed because that is not what is reflected in the speakers’ descriptions. The second, and preferred solution, is that instead the meaning of fa broadens to accommodate this ‘unnatural’ subject type. As certain components of meaning of ‘do’ are lost in this process, the changes to fa can also be called semantic reduction, alternatively bleaching (Givón 1975), or erosion (Heine and Reh 1984; Lehmann 2015 [1982]). At the point reached in the most ‘generalized’ version of FS, such as in the Monno dialect, the concept of fa is so ‘fuzzy’, or poorly delimited, that little of the original lexical meaning remains. Note however that, even in that dialect, FS is still not normally used with the classic epistemic verb ‘know’. This is best explained semantically by Monno fa still having a lingering sense of ‘do’.

Envisaging the changes as due to loss of the specificities of meaning (which Bybee et al., p. 24 note is often confused with metaphorical extension), the following cognitive progression is proposed for an extension of the form (each one adding to the previous), as represented by a set of binary features.

1. fa describes an activity, which is a durative event involving ongoing change [+ activity] [+ non-stative]

2. Loss of the durative aspect leaves the notion of the occurrence of an event involving change. [+ non-stative]

3. Loss of the notion of change opens up the additional possibility of a stable situation.

An alternative way in which the changes might be represented cognitively would be through conventionalization of the implicature, or inference (e.g. Traugott 1989, 2012; Heine et al. 1991, among others). Using this, there would be the following pattern of extension:

1. fa means actual ‘doing’ or activity;

2. ‘doing’ usually implies a changed situation (or that something is ‘done’);

3. The result of change is often a stable situation.

The reader might accept that either of the two above rationales for the extension in application of FS is plausible. Yet to know how this part of the grammaticalization process happens does not address the grammatical purpose, if any, of the obligatory construction. Thus the investigation into Camuno fa ‘do’-support stalls at the same place as does our understanding of English do-support discussed briefly in the introduction: what is the ‘point’ of ‘do’-support? To address this, it is worth noting the advice given in Bybee et al. (1994: 3) that ‘the cognitive and communicative factors which underlie grammatical meaning are often more closely revealed as change occurs, or, generally in variable as opposed to static situations’. We must therefore look ‘backwards’: the solution to the grammatical meaning of the bleached fa/do lies within the fully lexical ‘do’-support form.

Bybee et al. also remind us that a language must already be able to express tense, aspect, and modality, by some means before the new grammatical form is invented. They quote Labov’s (1990b: 45) perspective on the reason for existence of grammar: ‘On the whole grammar is not a tool of logical analysis; grammar is busy with emphasis, focus, down-shifting and up-grading; it is a way of organizing information and taking alternative points of view.’ Guided by this perspective then, it is not even the meaning of the support verb ‘do’, as deduced from likely combinations, which is the primary determinant of the grammatical meaning, but the meaning of the lexical construction as a whole.

The reader is asked to look again at examples (13) to (17) above that demonstrate the presuppositional meaning of FS in Camuno. These show how one function of optional fa-support is emphasis in the question as required to counter, or affirm, the speaker’s preexisting expectation. This can be compared to the intensifying, or emphatic function provided by optional do-support in English in the statement (3), (4). An emphatic function is then observable in both ‘do’-support in its incipient stages of grammaticalization in Camuno, and its much more advanced stages as in present-day English.

Accepting that emphasis is inherent in use of ‘do’ as a support verb and can be preserved during grammaticalization, at issue then is what part of the grammatical meaning is lost. For this the reader is referred back to examples 18, 19, which demonstrate the incompatibility of FS and a subject that has arbitrary reference in an optional FS dialect, but the compatibility in an obligatory FS dialect. This indicates that in optional FS, fa ‘do’ has an additional characteristic, that of requiring a realis situation, which is dropped at a certain point in the grammaticalization process as FS becomes obligatory.^[15]

By including this additional information it becomes possible to construct a grammaticalization pathway based on the relevant semantic contribution that fa ‘do’ makes to the construction. For comparison, consider first the following sequence of extension of meaning by incorporation of the inference as proposed by Bybee (2017) for English ‘can’ (21).

The verb ‘do’ instead refers to an agentive subject and its ‘doings’, and begins its trajectory describing a reality for which there is actual physical evidence. A sequence of meaning loss from a construction containing a fully lexical fa/do ‘do’ transitioning to one with a bleached form, would then run as follows (22) (and for simplicity it is an affirmative rather than interrogative one):

By incorporation of the inference, the final outcome of grammaticalization of ‘do’-support construction requires that ‘do’ becomes an epistemic modal.