Enclitic subjects and agreement inflection in Viskadalian Swedish

In this article, I describe the subject clitics of the traditional Swedish dialect in Viskadalen; I refer to this variety as “Viskadalian”. The Viskadalian subject clitcs are prosodically dependent on a verbal host to their left; thus, they occur only with VS word order. In most cases, it is the clitic that conveys the reference of the subject (kan=ik, ‘can.PRS.SG=I’, kan=et, ‘can.PRS.SG=it’). However, when the verb is inflected for both number and person, Viskadalian uses the seemingly meaningless clitic ä: fing-em=ä, ‘get.PST.PL-1PL=ä’, så-st=ä, ‘see.PST-2SG=ä’. I argue that the clitic ä is not a pure expletive but in fact represents a partially reduced phi-feature bundle: it is a pronominal version, as it were, of such partially reduced verbal agreement endings in VS contexts that have been attested in many Germanic varieties. Petzell Enclitic subjects and agreement NALS Journal


Introduction
This article deals with finite verbs and their pronominal subjects in the Swedish dialect of Viskadalen, 'the valley of the river Viskan'.The river Viskan, which is the northernmost of the rivers of Halland (the others are Ätran, Nissan and Lagan), actually starts in the province of Westrogothia (Sw.Västergötland) before it meets the sea of Kattegat just north of the town of Varberg; see Map 1.I will refer to the traditional dialect of the area around the southern reaches of the river Viskan as Viskadalian.
Map 1: Viskadalen (southern reaches of the river Viskan in blue, southwestern border between Westrogothia and Halland in black).
Intriguingly, when verbal inflection is distinctive (i.e. is only used for that grammatical person), the enclitic pronoun has the same form, irrespective of what person the subject is.Thus, both when the verb is inflected for first person plural (as in (2) below) and when it is inflected for second person singular (as in (3)), the post-verbal subject is enclitic ä; see (2a, 3a).With subject initial order, on the other hand, the subjects are different in the two cases, vi and du respectively, as I show in (2b, 3b).Thus, Viskadalian is like any historic variety; in order to describe it, we are dependent on an array of different sources, each of which contributes pieces to the larger puzzle.At the same time, Viskadalian is indeed alive (in some sense of the word), since we can still listen to it in audio recordings from the middle of the 20 th century.

Background
As far as I know, no dialectologist has (hitherto) taken an interest in Viskadalen.This lapse is hardly surprising, given the strong focus on the province (Sw.landskap) as the natural area of inquiry within traditional dialectology.Being situated on the border between Halland and Westrogothia, Viskadalen, as a whole, simply falls out of the scope of such studies.As a rule, the dialect of each province has been given a separate description.From such a perspective, both the Hallandian and the Westrogothian part of Viskadalen emerge as peripheral areas within their respective provinces. 2When, on occasion, some linguistic parallel between southwestern Westrogothia and northern Halland comes up, it is merely noted as a parenthetical piece of information.For instance, Lundell (1909Lundell ( :1110) ) points to the unity between northern Halland and the hundred of Mark (roughly, the part of Westrogothia closest to the river Viskan on Map 1) regarding 1PL and 2SG morphology (cf. examples (2) and (3) above); his remark is made in passing and the link is never explored any further. 2In the most comprehensive description of the dialect of Westrogothia, Götlind & Landtmansson (1940-1947) and Landtmanson (1950), the language of the southwestern part of the province is indeed identified as one of five sub-varieties of Westrogothian.On the other hand, in the most influential work on the dialect of Halland, Wigforss (1913-18) directs his main interest towards the southern areas of the province.Consequently, the Hallandian part of Viskadalen, situated in the far north of the province, receives very little attention.
The border between Westrogothia and Halland, cutting through Viskadalen, is, in fact, more than just a provincial border; before 1645, it was the border between Denmark and Sweden.However, as stressed by Grill (1954), this was a political border, which did not coincide with the economic border.For Westrogothians living along the river Viskan, the place to sell their agricultural produce was always the Danish town of Varberg; its hinterland reached the parish of Örby, located where the highlighted part of the river Viskan ends on Map 1 (Grill 1954:679; see also Linge 1969:75-76).In other words, Viskadalen has been an economic region for centuries.As such, it forms a natural dialect area.

Viskadalian data
Most of the Viskadalian data that I use in this article were collected around the middle of the 20 th century by local dialect archives in Lund (linguistic data from Halland) and Uppsala (data from Westrogothia).All these data have now been incorporated in the collections of the national Institute for Language and Folklore.The earliest written archive records are from the end of the 19 th century, the first audio recordings from the mid-1930s. 3My empirical focus is on recorded speech.More specifically, for the mapping of the Viskadalian dialect (in section 2.1 and 2.3.1),I consider only spoken data.
There are at least two reasons for basing the maps on recorded speech only.First, the (more or less) spontaneous language use in the audio recordings is indeed primary linguistic data and, as such, is comparable across the area; it is homogeneous enough to translate well into dots on a linguistic map.The contents of the written records, on the other hand, are filtered by the collector in the field.To be more precise, the written records contain what the collector wanted to document; in some cases, the purpose is clearly ethnological rather than linguistic.Thus, they are not only secondary data but are also quite diverse both qualitatively and quantitatively when compared across a larger area. 4In addition to the archive data, there are other written sources of interest to this study.For instance, there are linguistic descriptions of Viskadalian from before 1800, and there are numerous collections of dialect data that have been compiled by local amateurs who are interested in the traditional language of their home parish.It would only be misleading to try to convert all these various sorts of secondary data to (seemingly) comparable dots on a map.
Nevertheless, the written records often have the virtue of containing forms and types of examples that are hard to come by in the audio recordings.Generally speaking, reference to second person is very scarce in the spoken material, which is simply a consequence of the documentation situation and the questions asked.Typically, the interviewer asked questions about local life in the past: "What was it like when you grew up?", "What did people do?", etc. Naturally, this triggers narratives in first and third person.So for second person examples, I have turned to such written material that aims at documenting other phenomena than pure narratives.For instance, many of the second person examples are from collections of children's games and funny stories.
How much of the traditional dialect still lives on is, at present, largely unknown.
During the spring of 2016, I conducted a couple of interviews with people from the parish 4 Furthermore, differences between the archives in Lund (collecting Hallandian data) and Uppsala (collecting data from Westrogothia) regarding documentation policies, principles of transcription, prioritized areas and phenomena of course adds to the relative heterogeneity of the written records (see Skott, 2008:32-33, on the relations between different Swedish archives at the beginning of the 20 th century).
of Istorp in the heart of Viskadalen, just east of Horred on Map 1.These informants still used some of the traditional morphology under investigation in this article, at least in certain contexts.In the following, I will relate these recent data to the traditional data.5However, I should make it clear that my main object of study in this article is the traditional dialect.A more comprehensive description of modern Viskadalian would, of course, be of great value.Hopefully, it will be possible for me to take on that task in the not too distant future.

Results
In this section, I first present the Viskadalian agreement inflection on finite verbs (2.1).This is followed by a description of the form and syntactic distribution of the Viskadalian subject pronouns (2.2).Finally, I address the spread of the enclitic pronouns across the dialect area (2.3).

Person and number agreement on finite verbs
The Viskadalian verbal agreement affixes are given in (4) below.This paradigm is more or less identical to what we can reconstruct from Old Swedish sources (see Wessén 1958:116-118).An important difference, however, regards the 2SG affix: it combines with a much larger group of verbs in Viskadalian than it did in Old Swedish.In mediaeval texts (both from Sweden and Denmark), the (s)t-ending for 2SG was (as it still is in Icelandic) quite restricted: it occurred only with strong preterites (including preterite-present verbs in the present tense): 6 kan-t, bröt-st, 'can.PRS.SG-2SG, break.PST-2SG' (Belfrage 1871:13, Wessén 1958:133, Skautrup 1944:276).In Viskadalian, this 2SG affix is more versatile.In fact, in this variety, all verbs except those ending in present tense r occur with (s)t.The most apparent effect thereof is that Viskadalian contains an abundance of weak preterites inflected for 2SG, forms never found in the mediaeval language: ha-dde-st, bo-dde-st, vess-te-st, 'have-PST-2SG, live-PST-2SG, know.PST-PST-2SG'. 7 Turning now to the plural, all persons retain the vowel e from the past tense affix of regular verbs (which is te/de), the 3PL form then coinciding with 1/3SG: tal-te-m, tal-te-n, 'speak-PST-1PL, speak-PST-2PL', tal-te, 'speak-PST'.The vowel is also the same with monosyllabic verbs in the present tense, the vowel coinciding with the stem vowel, and the 3PL form coinciding with the infinitive: ha-m, ha-n, 'have-1PL, have-2PL', ha, Disyllabic verbs in the present tense, on the other hand, trigger different vowels in all cases: e for 2PL (kunn-en, 'can-2PL'), a for 3PL, this form again coinciding with the infinitive (kunn-a, 'can-PRS.3PL/can-INF'),and u (or o) for 1PL (kunn-um, 'can-1PL').
Furthermore, strong verbs in the past tense come with an e in 2PL and 3PL (åt-en, åt-e, 'eat.PST-2PL, eat.PST-PL'), and with o/u in 1PL (åt-om, 'eat.PST-1PL'). 8The parentheses 6 Originally, the s occurred only when the verb stem ended in t/d (bad-st, 'pray.PST.SG-2SG'), but during the early modern era, most notably in the Bible from 1541, the s was used more generally, that is also with other stem endings (gaf-st, tok-st, 'give.PST.SG-2SG, take.PST-2SG').With stems ending in l or n, the s is never part of the affix, neither in historical texts nor in Viskadalian.In the latter case, the stem-final consonant is sometimes suppressed in these contexts, for instance skal-t → ska-t, 'shall.PRS.SG-2SG'. 7Even when we expect a present tense r, the st-ending sometimes appears in its stead: ha-st, si-st, 'have-2SG, see.PRS-2SG' (Fag:95, 81).However, the affix never co-occurs with r (*ha-r-st). 8The usage of u/o in 1PL varies quite a bit.In some recordings, it is more generally employed (ha-d-om, 'have-PST-1PL' (Dags:2911B), bo-dd-om, 'live-PST-1PL' (Grime:1934A)); in others, around 2PL n indicate that it is sometimes absent, sometimes present.Naturally, when n is absent, the forms for 2PL and 3PL usually coincide, both ending in e (e.g.I/da åt-e, 'You.PL/they eat.PST-PL').Without the n, it is only with disyllabic verbs in the present tense that there is a difference in the paradigm between 2PL and 3PL (kunnekunna), the latter form, however (as noted), coinciding with the infinitive.
(4) 9  1/3SG -- 2PLe(-n) 3PLe/a In the dialects surrounding Stockholm, the person distinction in the plural seems to have been lost before 1500 (Neuman 1925), and number during the 17 th century (Larsson 1988).In what is present-day Denmark, verbal agreement was probably lost even earlier (Skautrup 1944:273-274). 10However, the dialects in the southern provinces of Sweden, which once (before the middle of the 17 th century) belonged to Denmark (these provinces include Halland, Skåne and Blekinge), people continued to inflect their verbs for person and number long into the 20 th century.As pointed out in the introduction, Viskadalen is part of this larger southwestern (from a Swedish perspective) dialect area, featuring the o/u is replaced by e in almost all contexts (druck-em, 'drink.PST.PL-1PL' (Käll:350B), åt-em, 'eat.PST-1PL' (Öxn1:6804B)). 9Here, I disregard present tense r, although it functions as a marker of both tense and grammatical number: in Viskadalian (unlike Standard Swedish), the r is restricted to the singular.By contrast, the past tense affix te/de is used across the board. 10In the case of Norwegian, there are hardly any written sources of the language between 1500 and 1800, due to Danish rule and linguistic domination (see Indrebø 2001:177-192).
agreement morphology on finite verbs.How much of the paradigm in ( 4) is preserved across this area varies considerably; see Horn af Åminne (2015) for details.
As stated above, the audio recordings contain instances of reference to second person only sporadically.Those inflectional endings in (4) that appear in recorded speech are thus predominantly 1PL and 3PL.I have listened to recordings from 76 parishes in Halland, southwestern Westrogothia and western Småland, paying special attention to verbal inflection and pronominal subjects; see the appendix for more details about these recordings.
On Map 2 below, all 76 parishes are represented by coloured dots (located where the parish churches are).The colour of the dot indicates the extent to which I have come across person and/or number inflection of finite verbs in spontaneous usage.White means that the recorded informants do not inflect their verbs for person/number.Light grey indicates that verbs are inflected for number, with e as a generalized plural ending.Black means that there is inflection both for number and person, and that the 1PL forms end in m.In some parishes further to the south in Halland, both the 1PL and the 2PL end in n; this is marked with dark grey.There are 20 black dots on Map 2. This means that I have come across spontaneous usage of the m-ending for 1PL along with verbs agreeing with 3PL subjects in 20 of the 76 parishes that I am considering.There are sporadic occurrences of inflection for second person as well, but these have no bearing on the mapping of inflection (for reasons given in 1.3 above).Some of the informants represented by the black dots inflect their verbs in all possible contexts; others switch, on occasion, to a more standard-like system (with no agreement morphology whatsoever).The informants of the former type are all from the parishes immediately surrounding the southern reaches of the river Viskan.

Pronominal subjects in Viskadalian
The complete Viskadalian paradigm of pre-and post-verbal subject pronouns is given in (5) below.Here, I use a disyllabic modal verb like kunn-a, 'can-INF', in the present tense, since all the inflectional distinctions in (4) are then put to use.Note that the pre-verbal pronouns do not have to come before the verb: they are free pronouns that can occur both before and after the verb (just like any subject).In contrast, the post-verbal pronouns always need to cliticize to the right of the finite verb.Note also that 2SG and 2PL differ with respect to the usage of the enclitic ä: the ä occurs in 2SG even without person agreement morphology on the verb (see (5bʹ) with komm-er, 'come-PRS.SG'); in 2PL, on the other hand, the presence of n in the affix is a prerequisite for ä (cf.5g to 5gʹ).
( A clear case of a special clitic is 1SG ik.Historically, the form ik predates ja; the latter is derived from the former by so-called breaking, which is presumably some sort of assimilation process (see Dyvik 1978).Synchronically, however, the forms are not phonologically related.The same goes for hu and na.Originally, na is from the oblique paradigm, corresponding to Old Swedish hana, 'her.ACC' (Wessén 1958:188, Howe 1996:295; cf. also footnote 12 above).
The di could possibly be a special clitic as well, given that there is no synchronic phonological rule in Viskadalian that converts a (in da) into i.Still, the original diphthongized pronoun dai, commonly occurring further down in Halland (as already mentioned), was once more widespread in Westrogothia too (Landtmanson 1950:96); dai is clearly in a simpler relationship with the di-form than da is.Also, the status of di as a clitic is quite recent: in a not too distant past, the di was an allomorph of the free pronoun (see also 2.3.1 below).
Moving on to et, it could belong to either category.One possibility is that it is a synchronically distinct form that is etymologically related to Gothic 3SG.N ita (as suggested by Noreen 1904: §508), in which case we should treat it as a special clitic.
Alternatively, it is simply a reduced form of det (as proposed by Reinhammar 1975:177), which makes it a simple clitic (given that the free pronoun de is also a reduced version of det).
As for en, it was most likely derived in the same fashion as na, namely from the oblique paradigm.As an enclitic object, en was commonly used in all sorts of Swedish texts up until the 18 th century; however, as a subject, it is only sporadically attested (see Wessén 1958:188-189 for examples).Whether en is synchronically related to han is hard to say; unlike the feminine, the masculine clitic shares with the free pronoun the vowel+n structure (cf.fem.hu vs. (n)a).However, deletion of initial h is hardly a general process in the dialect, although it tends to occur elsewhere in the pronominal system (see Landtmanson 1952:46).
Turning now to ä, it is clearly a special clitic, at least in 2SG and 1PL.Both du and vi are impossible to convert to ä in the synchronic phonological system.Historically, however, precisely this development appears to be derivable in both contexts.In 2SG, the initial dental of the pronoun is vulnerable to reduction, since the affix ends in a dental too: (s)t + du.In our earliest record of Viskadalian (Öxn3), which is a wedding poem from 1746 (appearing in Lundell & Hesselman 1937), the dental of the pronoun is intact, whereas the vowel is more or less identical with the vowel of the clitic (spelled e rather than ä).More specifically, we can detect the post-verbal dental in VS sequences containing a verb form that ends in present tense r.Since we expect no (s)t affix in such cases (see the beginning of 2.1 above), the dental sound must stem from the pronoun.
Thus, the spelling with hl, indicating a retroflex r [ɽ] in hahle höht, 'have you heard' (Öxn3:311) only makes sense if the r-ending of the verb (har) is followed by dä.
(Merging [r] and [d] into [ɽ] is still a common process in many Scandinavian varieties.)If the subject were ä, we would, instead, expect hare.As we saw above, precisely such forms occur in the more recent Viskadalian sources (see (6bʹ)).Consequently, here, the subject is clearly ä rather than dä (as before).
In 1PL contexts, on the other hand, it is often impossible to determine whether the enclitic pronoun is ä or mä.We know that the 1PL pronoun has indeed become mä (or some similar form) in many Norwegian dialects (see Tylden 1944), as well as in some Swedish dialects close to the Norwegian border; for instance, in the traditional dialect of the province of Härjedalen, the 1PL pronoun is mö (Pamp 1978:118).In Viskadalian, I have found no clear cases of an m-initial 1PL pronoun.On the contrary, there are examples in the written records that indicate that the pronoun is indeed ä: dä hôpp-um-s=ä, 'that hope-1PL-DEP=ä' (Öxn2:01).In this example, the affix and the clitic are separated by the deponent marker s; if the clitic were mä, we would expect *hôppums=mä.Presumably, the change in vowel quality (i  ä) occurred later than the reduction of v.We thus find examples such as ha-de-m-i, 'have-PST-1PL=we' (Öxa:5187), indicating a dropped v but no other difference from the free pronoun (i.e. the i of vi is intact). 13The parish of Öxabäck, where this example is recorded, is located on the edge of the Viskadalen area, where the last step (i.e.ä) is not attested.
Moving on to 2PL, I have not found any examples that unambiguously indicate that the clitic is indeed ä, rather than nä. 14Certainly, I have shown that the m-ending was never reanalyzed as part of the 1PL pronoun in Viskadalian (as in other Scandinavian varieties further to the northwest).However, the mere lack of m-pronouns in 1PL does not automatically exclude the possibility that there might be n-pronouns in 2PL.
Consequently, the ä in 2PL is analogically stipulated: since the ä follows the distinctive m in 1PL, I assume that it follows the distinctive n in 2PL too.

The clitic status of post-verbal pronominal subjects
Although there are certainly differences between definitions of clitics in the literature (see Spencer & Luis 2012:12 for an overview), the following criteria, at least, are generally included: clitics play a role in syntax (as independent words but unlike affixes) but they are prosodically dependent on some other word.In both these respects, the Viskadalian post-verbal pronouns are clitics: they need a host, namely a finite verb; at the same time, they are syntactic subjects, filling the inverted subject position in the syntactic structure.
13 V-reduction of this sort has also been attested in some varieties of Dalecarlian, for instance wel-um=i, 'want-1PL=we' (in the parish of Våmhus, quoted by Neuman 1925:189), where only the i remains of the 1PL pronoun. 14As mentioned above, the 2PL pronoun in standard Swedish begins with an n (ni).Presumably this n was part of the verbal affix to begin with (Wessén 1958:189-190; see also Ahlgren 1978 for the distribution of ni in the dialects).
This means that the clitic is syntactically real insofar as it cannot combine with another subject, as I show in (7a) below.Furthermore, since the clitic needs to keep close to its verbal host, other elements that can also follow the verb must never precede the clitic.
Thus the clitic pronouns are never found post-adverbially, nor in a position after shifted objects; see (7b-c).In these contexts, the free pronoun is the only option.
( The subject positions in (7b-c) are typically associated with some sort of prosodic prominence (often conveying contrastive focus).Certainly, post-verbal subjects that are indeed adjacent to the verb (unlike the subjects in (7b-c) may also be prosodically prominent; if so, they are realized as free pronouns, as shown (8a) below.Here, there is a clear contrast between the two possible agents 'you' and 'I'; hence, the post-verbal subjects are du and ja (rather than ä and ik).Furthermore, since the clitics are so tightly bound to the finite verb, they can never be part of a larger DP; only the free pronouns are possible in such contexts; see (8b where the verb precedes the subject. In sum, VS order is a necessary (but not sufficient) condition for the Viskadalian subject clitics.In addition, the VS sequence needs to be inside the same prosodic phrase.
In other words, there can be no intervening material between V and S (as in (7b-c)), nor can S itself be prosodically prominent (as in (8a)) or part of another phrase (as in (8b)).I will return to the significance of prosody for a deeper understanding of the Viskadalian clitics (especially the versatile ä) in section 3.

Mapping the enclitic pronouns
In what follows, I take the reader through the paradigm in (5-6) person by person.As stated above, only first and third person morphology occur in audio recordings to such an extent that making maps of their spread is meaningful.I therefore start with first and third person clitics in 2.3.1,where I present dialect maps based on the actual attestation of these features in spontaneous speech across the area.In 2.3.2,I discuss the written documentation of second person morphology in Viskadalian, trying to reconstruct the reality behind the records.

First and third personoccurrence in actual usage
Starting with 3SG, these three pairs of free pronouns and clitics (see (5c-e) and (6c-e) above) are not only spread across Viskadalen but are also found in neighbouring areas (see Landtmanson 1950:92-93, 96 (1923:137).Although geographically restricted, ik appears to be quite stable where it is attested.In the Viskadalian wedding poem from 1746 mentioned above (Öxn3), the enclitic 1SG pronoun is there: så bli-r-ek, 'so become.PRS.SG=I' (Öxn3).Also, among present-day people in the area who still preserve features from the traditional dialect, ik is one of the traits that live on (dä kan=ik inte, 'that can.PRS.SG=I not' (Ist2)).
A more widespread trait is the usage of the clitic ä in 1PL: ha-de-m=ä, 'have-PST-1PL=ä' (Gsjö:1036B).This is found in recordings from 11 parishes in the Viskadalian heartland; see Map 4. The m-inflection alone is spread over a slightly larger area than 1PL ä, both further to the north-east and to the south (cf.Map 2).In these areas, post-verbal 1PL reference is realized in a variety of ways.We encounter post-verbal vi (i.e. the free pronoun) on a more regular basis (tok-om vi då, 'take.PST-1PL we then' (Tores:255A_m24)).In some recordings, the 1PL affix is missing in precisely these contexts (dä ha vi, 'that have.PL we' (Örb2:6207)), although it is present with SV order (vi ha-m aldri, 'we have-1PL never' (Örb2:ibid.)).A third possibility is that the postverbal subject is fully deleted (så ha-d-om fyra, 'so have-PST-1PL four', Dags:2911B).By contrast, in the ä-varieties, the affix is hardly ever missing, and the pronoun itself may be omitted only if it is directly followed by a vowel, indicating that in those cases we are dealing with a purely phonetic reduction: kall-te-m(=ä)=et, 'call-PST-1PL(=ä)=it', where there is clearly no subject, although the following word does not begin with a vowel (cf. the similar example from Dags above).
Moving on to 3PL, the distinction between a free pronoun da and an enclitic di (see (6h-h')) is not mentioned in the previous literature.Both Landtmanson (1950) and Kalén (1923) note that there are certainly several forms of the 3PL pronoun, but neither identifies di as a clitic.However, in 11 of the audio recordings that I have listened to, the distinction between free da and enclitic di is as clear as the other distinctions between free pronouns and clitics in the paradigm; see Map 5.
and the free morpheme da hardly at all.Today, the standard-like dum (cf.standard dom) is often used across the board. 15

Second person in secondary sources
The occurrence of second person reference in the audio recordings is, as stressed before, only sporadic.More specifically, there is no single informant on tape who uses all four types shown in (5b-b') and (5g-g'), although each type is attested somewhere in the collection of spoken material.Turning instead to the written records, and first taking 2SG into consideration, the largest collection of examples representing the two types in (5b-b') comes from the Hallandian parish of Fagered, just on the border with Westrogothia.
We have no reason to believe that Fagered was special in any way.This parish happened to be targeted by one of the most influential recorders of folklore of his time, namely Johan Kalén, since this was the parish where he started his professional career (as a teacher).In addition, Kalén was particularly keen on documenting authentic dialogue, which naturally contains much second person reference, especially in the singular.
Nevertheless, it is probably significant that the area where Fagered is situated is quite far from the coast and very close to Westrogothia.In the literature, inflection for 2SG clearly emerges as more of a Westrogothian than a Hallandian trait.Landtmansson states (1950:135) that 2SG inflection occurs both east and north of Mark (the hundred furthest to the southwest).Even earlier, Hof (1993Hof ( [1773]]:42) noted that 2SG inflection was common among country folk all over Westrogothia.Consequently, in the Viskadalian wedding 15 The standard-like pronoun has been winning ground over traditional forms in many parts of Sweden.How this proceeds, more exactly, varies considerably.For instance, some speakers of the urban dialect of Eskilstuna (a town situated ca. 10 km east of Stockholm), use the standardlike dåmm in post-verbal position and the traditional form di in pre-verbal position (Grass 1970:39-40).
Whereas the 2SG affix is something of an eastern (or Westrogothian) trait, the usage of the full 2PL affix (including n) followed by enclitic ä (see (5g, 6g)) is clearly a western In other languages, for instance Old French, pronominal subjects can be fully reduced (i.e.deleted) when they are in the same prosodic phrase as an inflected verb containing the same phi-features as the pronoun: {corur-ent Ø}, 'ran-3PL (they)' (example from Adams 1987).Ackema & Neeleman (2003:723-725)  verb?In fact, I believe that the Viskadalian ä fills precisely this empirical lapse, being indeed an example of a partially reduced pronominal phi-feature bundle.Recall, first, that inverted 2SG, 1PL and 2PL pronouns are all spelled out as ä when the phi-features of these persons are present on the verb.Thus, it seems reasonable to conclude that the ä at least carries less referential information than any of the free pronouns.However, it cannot be that all of its phi-features are reduced and that the ä is an expletive.Such an analysis predicts that ä should be suitable as a non-referential subject, which is contrary to fact: *nu kan=ä regna, 'now can.PRS.SG=ä rain'.Instead, the ä could be a realization of exactly the one phi-feature that 2SG, 1PL and 2PL have in common, that is, what is left of the pronominal phi-feature bundle after the partial reduction of redundancy has taken place.Clearly, this has to be some sort of participant feature, but I will not go into any details here.
Still, I want to point out that any analysis of ä that treats it as an effect of redundancy deletion is unable to capture its entire repertoire.Certainly, in the plural this analysis works fine: the ä is there only when there is an inflectional ending conveying the referential meaning of the subject (m or n).But in 2SG, the ä occurs even without inflection, that is when the pronominal bundle appears to be non-redundant ({få-r=ä}, 'get-PRS.SG=ä' (Fag:95)).In order for the reduction analysis to work here, we would need to assume that inflection is underlyingly present in these contexts and that, for some reason (possibly related to the r-ending), the (s)t-affix is suppressed, but only after it has triggered the reduction of du  ä.A more straightforward (but not very elegant) account could be to analyse the ä as a clitic with 2SG meaning in these particular contexts (i.e. there would be two separate äs).
Another complicating factor with regard to the morpheme (s)t is that it is more common in VS contexts (as I mentioned in 2.3.2 above).In fact, in present-day Viskadalen, informants judge the (s)t-ending as ungrammatical with SV order, although they often use it with inverted word order: ?du kan-t, 'you.SG can.PRS.SG-2PS', Wa blai-st=ä å?, 'what become.PST-2SG=ä of' (Ist2).This strict limitation to VS order suggests that (s)t is not an affix anymore; instead, it has been reanalysed as part of the enclitic pronoun ä.That such a reanalysis has indeed taken place can be deduced from the sources quite early on, at least in the parishes towards the coast.What we find there are examples such as töck-er=tä, 'think-PRS.SG=tä' (Himl:73) and behöv-er=stä, 'need-PRS.SG=stä' (Dags:2917B), along with uninflected verbs in SV contexts (du ha-de, 'you have-PST' (Dags:ibid.)).As we know, with present tense r, as in these two cases, there was never any 2SG inflection to begin with (cf.tôck-er=ä, 'think-PRS.SG=ä' (Fag:107)).
However, if (s)t=ä is perceived as one morpheme, that is as a 2SG clitic (i.e.(s)tä), it makes perfect sense that the limitations associated with the affix ("do not attach to present tense r") have no significance anymore.
However, the affix legacy is clearly visible in the high degree of formal variation that the 2SG clitic displays.Exactly how the inherited allomorphy from its affix days is accommodated in systems without inflection for 2SG is not clear to me.The only conclusion I dare to draw is that the reanalysed clitic contains elements of the original affix, but that the distribution of these elements is not guided by the same principles as it was when (s)t was an affix.Apart from the stä and tä after the present tense r mentioned just above, we find, on the one hand, s-less forms where the affix originally contained the s (ha=tä, 'have=you.SG' (Träsl); cf.ha-st have-2SG' (Fag:135)), while on the other hand, there are also s-forms where the inflectional affix never contained any s (kan=stä, 'can.PRS.SG=you.SG.'(Faur:54); cf.kan-t 'can.PRS.SG-2SG' (Fag:64)).In addition, the ä sometimes occurs with uninflected verbs that were always inflected in the traditional system (e=ä, 'be.PRS=you.SG' (Mark:3); cf.ä-st 'be.PRS-2SG' (Fag:73)).In order to fully understand this variation in clitic form, we need to investigate more thoroughly the entire clitic system of present-day Viskadalian.

Summary
In this article, I direct my attention to the subject pronouns of the previously neglected Swedish dialect of Viskadalen ("Viskadalian").The pronominal subjects of this variety (in its traditional form) come in two guises: full pronouns, which are possible in all contexts, and enclitic forms, which are prosodically dependent on the finite verb.I show that some clitics are more widespread than others.For instance, the 1SG clitic ik is restricted to a relatively small geographical area around the lower reaches of the river Viskan, whereas the 3PL clitic di, although having acquired its clitic status quite recently, occurs in a larger area.The Viskadalian clitics are semantically equivalent to the full pronouns in all cases but one: the fascinating exception is ä.This clitic combines with all finite verbs that are distinctively inflected for person and number.I argue that ä represents a reduced phi-feature bundle, and that the reduction is triggered by the semantic richness of the verbal affix.  1Below, I give the accession numbers only for recordings containing inflection (at least for number).In other words, there are indeed recordings in the archives of ISOF from the "white" parishes, but I have found no agreement inflection in them.

Map 2 :
Agreement inflection on finite verbs in spontaneous speech (white = no inflection; grey = only number; black = m for 1PL; dark grey = n for 1PL).

(
or Hallandian) trait.The only examples of the enclitic ä in 2PL that I have come across are from the parishes surrounding the outlet of the river Viskan: vör-en=ä, 'be.PST.PL-2PL=ä' (Strå:224B); well-en=ä, 'want-2PL=ä' (Värö2:14); sko-n=ä, 'shall.PRS.PL-2PL=ä'(Himl:72); förstå-n=ä, 'understand-2PP=ä' (Ås:4227B).The n of the affix is not as robust further up the river: in the Öxnevalla record from 1927 mentioned above, some verbs are reported as having the n (få-n, 'get.PRS-2PL', vess-te-n, 'know.PST-PST-2PL'(Öxn2:01)), others as not having it (ät-e, 'eat.PRS-PL', sto-e, 'stand.PST-PL' (ibid.)).The only example with an inverted 2PL pronoun in this particular collection lacks the n on the verb; the subject is thus the free pronoun I: Nu skräpp-e I, 'now boast-PL you.PL' (ibid.; cf.(5gʹ, 6gʹ) above).Further down in Halland, the inflection with n is much more general, but in this area, there is no enclitic ä in any context: Wa vell-en I?, 'what want-2PL you.PL' Among the Viskadalian clitics described in this article, one stands out: ä.The other clitics, ik, en, na, et and di, are obviously clitic versions of the full phi-bundle: they mean the same thing as their free counterparts, but they are prosodically dependent on the verb, which, in turn, restricts their syntactic distribution.By contrast, ä occurs whenever the referential meaning of the subject is expressed in the preceding verbal ending (by (s)t, m, or n).Below, I present an analysis of the Viskadalian clitics that relates both their restricted syntax and the versatile nature of ä to the spell-out of prosodic phrases.In their 2003 paper, Ackema & Neeleman discuss prosodically conditioned morphological reduction in a wide selection of languages (both in the past and the present).For instance, the authors show that pronominal subjects are reduced to clitics in Middle Dutch whenever they are in the same prosodic phrase (indicated by {…}) as another head: {datt=en}, 'that=he'(Ackema & Neeleman 2003:705).As I show in 2.2.2, it is also true in Viskadalian that subject clitics need to be in the same prosodic phrase as a head, but not just any head.Unlike Middle Dutch clitics, Viskadalian clitics demand a verbal host: ?*{att=ik}, 'that=I; {kan=ik}, 'can.PRS.SG=I' (Ist2).
for examples from Westrogothia).In other words, although clearly part of the clitic system under scrutiny, they are hardly distinctively Viskadalian traits.In addition, as enclitic objects, the forms en, na, and et are even more widely spread (SAG 2:271).With 1SG ik, quite the contrary is true: it is only attested in recordings from 6 parishes in the central part of Viskadalen; see Map 3.This geographical limitation is confirmed by Kalén in his work on Fagered, a parish in Halland just on the border of Westrogothia (between Ullared and Älvsered on Map 1).Kalén notes that ik does occur in Fagered, but that all those using it have moved there from neighbouring parishes in Westrogothia analyse this as a reduction of feature redundancy within the prosodic phrase.As I mention in 2.3.1, similar deletion is attested on the edges of the Viskadalen area: {kör-d-om Ø}, 'drive-PST-1PL' (we) What Ackema & Neeleman claim as being unattested, however, is the partial reduction of a pronoun(2003:730).From a theoretical viewpoint, however, such a difference is unexpected.Why should the phonological component (where reduction takes place) care, as it were, if the partially reduced bundle is tied to a pronoun or to a = Veddige (par), ISOF nr.: 2064A.[audio; recorded in 1958] Värö1= Värö (par), ISOF nr.:245A.[audio: recorded in 1967] Värö2 = Värö (par), ISOF nr.:7827.[written; recorded in 1962] Värö3 = Värö (par), ISOF nr.: 8362.[written; recorded in 1964] Öxn3 = Öxnevalla (par), in: Lundell, Johan August & Bengt Hesselman, Bengt (eds.), 1937.Bröllopsdikter på dialekt och några andra dialektdikter från 1600-och 1700-talen.Stockholm: Geber.