4. Antecedents i estat de la qüestió
4.2. L‟estudi de les sequeres a l‟àmbit geogràfic de les Illes Balears
5.1 Overview
As discussed in Chapter Two, many grammaticalization specialists insist that grammaticalization is unidirectional, in spite of numerous counterexamples; many of them even structure their theories so that de-affixation is completely impossible. This is somewhat understandable, given the theoretical assumptions these specialists tend to hold: if affix- genesis is considered a “natural process”, de-affixation, or (as we might also call it) affix- exodus,164 is somehow exceptional, outside the bounds of nature. One way to account for its
exceptionality is to label it a by-product of other types of linguistic change: if the grammar has undergone radical changes, perhaps the usual rules may be temporarily suspended.
The synchronic theory of morphosyntax assumed here, however, has no such restrictions and does not automatically rule out affix-exodus. Morphological words are themselves complex syntactic structures, and they are built up via a series of syntactic derivations. Therefore, in principle, the theory could allow for the derivation to halt at any stage of the process. In diachronic terms, rather than one generation adding a new step to the derivation, as in affix-genesis, in the case of affix-exodus speakers are reducing the number of steps in a syntactic derivation to exclude nodes that were previously included. However, this is not likely to happen sporadically, and the result might even be unpronounceable. Halting a derivation prematurely requires a great deal of motivation. Thus, the theory can account for affix-exodus as readily as affix-genesis; nevertheless, it neither predicts nor implies that affix- exodus should happen as frequently as affix-genesis.
Other consequences follow if the notion that affix-genesis is somehow a privileged, speaker-independent process is set aside. I argued in Chapter Two that grammaticalization was essentially a by-product of various semantic, phonological, and ultimately syntactic changes, rather than a type of change of its own. This means that affix-genesis is, in fact, itself a by-product of changes elsewhere in the grammar, and, therefore, that there would be nothing remarkable about affix-exodus resulting from other linguistic changes either. In fact, we will see that while affix-exodus sometimes seems to have arisen in the context of other changes, sometimes it seems to happen more or less spontaneously, with no obvious motivations. While the existence of affix-exodus generally supports the view that linguistic change is in no way deterministic, this particular property of it – that it sometimes appears in contexts where no obvious grammatical changes have occurred – makes the case for the non- deterministic nature of linguistic change particularly compelling.
What is a necessary precondition for both affix-genesis and affix-exodus is structural ambiguity from the perspective of the learner (the ambiguity may not be obvious to the adult speaker). The data must support both an analysis in which Morpheme X is an affix and an analysis in which it is not. Learners seem to be biased in favour of the former, but this apparent bias must be overridden in cases of affix-exodus, and why this should be the case will be the focus of this section of my dissertation.165
164 I introduced the term affix-genesis because the more usual affixation and grammaticalization are both too polysemous for clarity. Since de-affixation is unambiguous, no other terminology is necessary; but this results in an unpleasant lack of parallelism when one speaks of affix-genesis and de- affixation, which to my ear is utterly abhorrent, as it makes the two sound completely unrelated. Because I care far too much about these things, when I am contrasting the two phenomena, I reserve the right to use affix-exodus in lieu of de-affixation, as it is nicely parallel to affix-genesis.
165 Kiparsky (2005b) has also attempted to account for affix-genesis and affix-exodus within the same morphological system, but with a very different theoretical framework; cf. Chapter Eight.
165 Most previous discussions of instances of affix-exodus have neglected to consider the grammatical context in which these changes were situated, or to examine fully what feature or features in the grammar allowed an unusual kind of change to occur. In 5.2, I discuss six case studies with as much attention to structural details as allowed by the available data; in some instances, unfortunately, the structural details are obscure.
In Chapter Three, I developed a structural typology of types of affix-genesis, repeated here as (5.1). One of the obvious questions that arises with regard to affix-exodus is whether a similar typology can be formulated on the basis of known examples.
(5.1a) Type I: Acquisition of an additional movement operation;
(5.1b) Type II: Reanalysis of a terminal as an exponent of a different structural position;
(5.1c) Type III: Extension of an M-word boundary to include a former clitic; (5.1d) Type IV: Reanalysis of a compounded element as an affix.
Since affix-genesis and affix-exodus are mirror images of each other, the logical place to start is simply by reversing the components of (5.1). This produces something like the typology in (5.2); note that (5.1b) does not need to be revised, which will become important in the course of the chapter.
(5.2a) Type A: Loss of a movement operation;
(5.2b) Type B: Reanalysis of a terminal as an exponent of a different structural position;
(5.2c) Type C: Retraction of an M-word boundary to exclude a peripheral affix; (5.2d) Type D: Reanalysis of an affix or compound element as an adverbial.
It is intuitively clear what examples of Types A–D would look like. Suppose that P- speakers in Language Q have verb-raising to T, so that T is part of the verbal M-word, but that this rule is lost by innovators, so that T in the innovative grammar is a freestanding exponent of T: this would be an example of Type A. Furthermore, Type B, which is minimally different from Type II affix-genesis, would be exemplified by a case affix in Language R developing into an independent adposition. But while Types A and B are both conceptually and structurally clear (regardless of whether they are attested), Types C and D are markedly less so, in part because it is difficult to think of an illustration of one of them that could not be better described as an example of either Type A or Type B.
Moreover, the actual attested examples do not fit neatly into categories like those in (5.2). As we will see, the case studies discussed in this chapter have a miscellaneous feel to them. Type B is clearly attested, but some examples that might be examples of the other logical categories could also be taken as examples of Type B. Type A, despite being intuitively clear, is not obviously attested. Although it is clear that all of the relevant action occurs on the periphery of an M-word, this does not take us very far in delineating the logical possibilities. 5.3 discusses this problem in the context of the typology from Chapter Three and demonstrates that it fails to provide precise structural categories.
Previous discussions have also failed to take advantage of a valuable opportunity to consider deeper questions about the learnability of grammatical structures and the factors which drive morphosyntactic change; this is the subject of 5.4.
5.2 Case Studies
In this section, I briefly discuss several examples of affix-exodus, and demonstrate that changes in other areas of the language are implicated in only some of the examples.
166 Since affix-exodus is not necessarily a by-product of other linguistic changes, it must be acknowledged that there is nothing “unnatural” about this type of change. The order in which the case studies are presented has been carefully chosen to highlight this.
The first three case studies did all arise in the context of other linguistic changes; these are Estonian C-particles (5.2.1), which developed as a result of phonological changes; the abessive postposition in Northern Saami (5.2.2), originally a case suffix, which was also affected by phonological changes; and the independent Irish first plural pronoun muid (5.2.3), originally a verbal ending, which first appeared during a time of massive changes to the verbal and pronominal systems in Middle Irish. Traditionally, the fourth case study – the group genitive in Germanic languages (5.2.4) – is believed to have arisen as the nominal case system was breaking down; however, as we will see, the matter is less certain than often believed. This prepares the ground for the fifth case study, concerning the new English adverb ish, which is still transparently related to a fully-productive adjectival and nominal suffix (5.2.5) and which cannot be connected to any other linguistic changes whatsoever. Last comes the mysterious recent development of a quantifier from a former suffix in Frisian, Dutch, and German (5.2.6), the sketchiest of the case studies from this chapter. This last case is discussed in less detail than the others, because few details are known; however, it is worth including because, like English ish, there is no obvious connection between it and other linguistic changes in these languages.
5.2.1 Estonian
The Estonian particles es and ep have long been cited as one of the core cases of counter-directional grammaticalization. According to the standard analysis, es, an archaic question particle, and ep, an emphatic/affirmative particle, were originally clitics, but have since become independent words located in C. However, of the case studies discussed in this chapter, this is easily the most speculative. Most previous work on these particles, e.g. Campbell (1991) and Nevis (1986a), were rooted in traditional etymologies (Ariste 1973, Alvre 1976). This analysis is based on comparative evidence, but it is only as strong as its etymology, and the etymology is not without contention: recently, Metslang, Pajusalu, and Habicht (2008) have put forth an alternative proposal for the origin of these clitics; unfortunately, their work appears to be available only in Estonian, and I am unable to evaluate it.166
Despite this controversy, I am choosing to retain this example, with the appropriate caveats. Even were the etymology entirely secure, the Estonian case is less than ideal because it involves clitics rather than bona fide affixes, and because it is not clear what happened structurally, since the changes occurred in the remote past. My reason for including it is simply to demonstrate how language-learners might be influenced by phonology when making their decisions about the location of M-word boundaries, and as an illustration of this kind, the Estonian particles still suffice. Since this is essentially a glorified “toy” example, however, I will not speculate as to the nature of the structural changes involved. As the most basic references on both particles are in Finnish,167 the description of the data relies on the
summaries of Campbell (1991) and Norde (2009).
Traditional philological work traces the Estonian affirmative particle ep back to the Balto-Finnic emphatic clitic *-pa. Although Estonian underwent a process of final vowel apocope, the presence of this clitic was sufficient to bleed the process:
166 The nature of etymologies of this sort makes it unlikely that there will ever be a truly definitive answer in favour of Metslang et al.’s analysis over the traditional, or vice versa. Metslang et. al’s motives are also unclear to me; if they are of the Haspelmath school of thought, they may well have chosen to formulate an alternative analysis in order to remove a counterexample from the list of “counterdirectional” changes.
167
(5.3a) *päällä > *pääll
‘on top of’
(5.3b) *päällä-pä > *päällä-p
Once the process of apocope had ceased to be a surface phonetic rule, and the final vowels were absent from the underlying form, the vowel that had been “protected” by a following clitic was reinterpreted as belonging to the clitic. Various other sound changes occurred, including the loss of vowel harmony, ultimately yielding the form peall-ep. Since vowel harmony had been lost, there was no further evidence for ep as dependent on the host form, and it was cut loose. Ariste (1973:37) notes:
After this suffix was lexicalised, the word ep could change its syntactic position and precede the affirmed words: see ep → ep see.168
By the standard account, the history of the interrogative particle es is similar but more convoluted. It was not originally an interrogative clitic; it can be traced back to the pan- Balto-Finnic clitic -s, said to be a marker of “informal speech”, added to the original Balto- Finnic interrogative clitic *-ko. The vowel of this suffix was variably lost in some of the Balto-Finnic languages (cf. colloquial Finnish -ks, Campbell 1991:291-2), and the suffix itself had become optional in Estonian, so that *-s had become one of the possible interrogatives. *-s became es in the same way that *-pä became ep, and was then relocated to the C-domain, as in the following examples from Older Estonian.169
(5.4a) Nüüd es tee uSSute. now Q 2nd.pl. believe.2nd.pl.
‘Now do you believe?’
(5.4b) MiSt es minna Seddä peä tundma. it.from Q 1st.sg.NOM. it should know.INF
‘How should I know that?’
Even setting aside structural considerations, there are several important pieces to this picture. The first is to note that, in this case, one of the essential ingredients was a change in the location of a Sub-word boundary. P-speakers segmented the sound sequence peallep as
pealle-p, innovators as peall-ep, and this decision figured crucially in their subsequent analyses. Knowing that p was not a legitimate phonological word in their language, P- speakers could not treat peallep as two separate phonological words once they had arrived at the segmentation pealle-p. But innovators could: for them, the emphatic Vocabulary Item was an entire syllable, ep, which met the qualifications for distinct phonological word.
Note, in addition, that previous phonological changes were essential in allowing this segmentation as a possibility. Estonian had lost final vowels. Prior to apocope, a child learning Estonian knew to segment *päälläpa as *päälla-pa because s/he would have learned that *päällä was a sequence occurring elsewhere with the same semantics as the form *päälläpa, minus the emphatic component. This would guide him to make the same choice his parents had made. Moreover, even when the surface form had become *pääll, there was likely to be a period in which the final vowel was still present at some stage of the phonological derivation, and this, too, would enable a child to match the Sub-word segmentation of the earlier grammar, *päälla-p(a).170 However, when apocope was well and
168 Translation of Campbell (1991:291).
169 Examples from Alvre (1976:345). Glosses and translations are those of Campbell (1991:290). 170 Don Ringe (p.c.) has pointed out that the final vowel of the clitic may have been lost in the underlying forms while the final vowel of *päälla was still underlyingly present, since the former would never have been protected by a following clitic.
168 truly final, and final vowels had vanished from even the underlying forms of the grammar, this was no longer possible, because the non-emphatic form had now become *pääll.
It is possible that the final stage of loss of underlying final vowels and the resegmentation occurred in a single generation. This is only one possibility, however; another would be that speakers took the e in peallep as a connecting vowel between a consonant-final host and consonantal clitic. Another phonological change becomes relevant at this juncture: the loss of vowel harmony. As long as Estonian retained vowel harmony, speakers would be more likely to take the vowel as part of the host, since there was indication for such a segmentation. Loss of vowel harmony meant a loss of another analytical cue, and a more uncertain status for the ambiguous vowel.
And so we return to the innovators, whose analytic decisions did not end with their choice of segmentation. The segmentation itself could easily have had no syntactic consequences whatsoever; in Chapter Six we will see a number of such cases. One could envision a scenario in which both segmentations of peallep are current in the same community, even by members of the same generation and family, and quite possibly this variation would escape notice. Ultimately, however, there were syntactic consequences, because the particles gained independence and relocated. Even though we are not concerned with structure in this sub-section, there is still much to be said about this change.
A speaker who segments peallep as pealle-p has little choice about where the clitic will appear, because the clitic is not capable of phonological independence. But for a speaker who segments peall-ep, the distribution of the M-word ep is no longer restricted for phonological reasons. Here another analytical choice must be made. If the speaker’s input data is primarily given him by P-speakers, never will -p/ep occur other than adjacent to a host. The learner could very well mark this required adjacency as crucial, and no syntactic innovation would occur. Yet another conclusion is certainly possible. The difference between emphatic and affirmative is rather slight; the English I did return the DVD last night
has both emphatic and affirmative connotations. Therefore, rather than taking ep to be a word-level emphatic suffix, a learner might conclude that ep has a more general usage and can apply to entire sentences, and fix it in the C-domain accordingly. After one such innovation, any children learning the language who encounter this speaker are likely to be biased in favour of acquiring the innovation themselves, because the status of ep is already ambiguous in the speech of more conservative speakers.
One interesting question to ask at this junction is how many stages are necessary in changes of this kind. A scenario in which speakers with the innovative segmentation nevertheless persist in analysing ep as having obligatory narrow scope for several generations before someone acquires still-more-innovative wide-scope ep is plausible, but there is nothing in principle keeping innovators from leaping to the second innovation once they have acquired the first. Put another way, the three stages listed in (5.5) are obligatory, as is their ordering, but there is nothing at all preventing (5.5b) and (5.5c) from occurring in the same generation.
(5.5a) pealle-p
(5.5b) peall-ep, narrow scope only (5.5c) peall ep, wide scope possible
Here we again encounter the black box of actuation. When multiple analyses are