The morphological and constructional evolution of ohg huuanta and Dutch want

Under a formal and functional reconstruction, the form and semantics of Old High German huuanta and Dutch want receive an explanation for the first time. Both conjunctions, together with Latin unde and Tocharian b ente , a äntā(ne) , descend from pie interrogative-relative * kwo-m-dheh1 , * kwo-m-dhoh1 , * kwo-m-dhah2 ‘whence, where’, whose semantics may be compositionally analyzed as ablatival-instrumental * kwo-m plus locatival-directional *- dho(h1) , *- dha(h2) . The novel equation of Old High German huuanta , Dutch want with Latin unde and Tocharian b ente , a äntā(ne) sheds light on a number of phonological and syntactic questions, including the merger of pie *- nd and *- ndh - in Latin and Tocharian (§2.1) and the non-affrication of *- nd - in Tocharian (§3.1.2). Another consequence is that a case can be made for clause-initial aphaeresis which triggered the loss of the labiovelar onsets in unde and ente/äntā(ne) , thus point-ing to the existence of wh -movement and clause-initial wh -words in both Latin and Tocharian (§3.1.1).


The etymology of ohg huuanta, present-day Dutch want
Etymologizing conjunctions is challenging for at least two reasons. First, conjunctions tend to preserve frozen nominal and especially pronominal morphology. Second, the functional reconstruction of conjunctions and their morphemes cannot be carried out successfully on the basis of the morphophonemic material alone. It is necessary to distinguish between the etymology of a construction marker and the history of a construction. The two need not be identical, and inferring the history and historical function of the construction from the synchronic or historical function of the construction marker is problematic if the construction has undergone functional changes. An example is the English causal conjunction for, which etymologically is to be equated with the homophonous purposive or benefactive preposition for as in for this purpose and for me, while the causal speech-act construction marked by for (as in Come … take your inheritance, the kingdom …! For I was hungry and you fed me) derives neither from a purposive nor from a beneficiary construction. The functional breadth of a conjunction thus cannot be explained on the basis of its construction marker alone (cf. Harris and Campbell 1995:284 on the markerstructure fallacy). But the etymology of a construction marker certainly provides an indispensable starting point and basis for syntactic reconstruction, which has to augmented by the formal and functional evolution of the construction. An instructive example is the etymology of the ohg interrogative adverb and conjunction huuanta 'why?; for; because' . huuanta did not survive into Modern German except in South Bavarian dialects like Cimbrian in northern Italy, where it remained at least until the 19th century (Old Cimbrian bant [Schmeller 1855:109]; bánda, ban 'for; because' [Schweizer 2008:917, 918]), but is still in living use in present-day Dutch want 'for; because' . Both the formal reconstruction of ohg huuanta and its constructional history pose problems, but as I hope to show in the following pages, it is possible to make some headway in the search for external etymological matches. Methodologically, the discussion must begin with the formal reconstruction, on which also hinges the identification of Indo-European cognates outside Germanic. Schmidt (1962:28ff.) equated ohg huuanta 'why, etc.' , os hwanda, hwande, OFri. hwande, hwant with ocs kǫdě 'whence?' , arriving at a purely phonological an anonymous reviewer for their help in correcting a preliminary version of this article. All remaining flaws are entirely my own responsibility.

Phonological and morphological reconstruction
Indo-European Linguistics 5 (2017) 1-30 reconstruction of ohg huuanta as Proto-Germanic *hwandẽ, contracted from pie "*kwu-n-dhe-e" or "*kwo-n-dhe-e". According to Schmidt, the reconstructed form is to be segmented as interrogative morpheme *kwo-followed by a nasal morpheme -n-and adverbial morpheme -dhe.1 For the development of wordfinal pie *-eh1 in Germanic, see Boutkan (1995:379). But the functional identification of the latter two morphemes remained elusive at Schmidt's time, and the same holds for the final long *ē of the form. Regarding the pie reconstruction of ohg huuanta and of the related Modern Dutch conjunction want, the etymological dictionaries of German and Dutch adhere to Streitberg's 1920:241 andSchmidt's 1962 etymological identifications. The most recent example is Philippa's (2009) account, which traces pd Dutch and Frisian want back to Old Saxon hwanda, Middle Dutch wande, wende, want, went, Old Frisian hwande, hwende, ohg huuanta and Proto-Germanic *hwandē from a pie interrogative *kwo-m with assimilation *m > *n before the dental morpheme *-dhe-.2 In sum, the form and function of the two morphemes and the origin of the vowel length in *-dhe have been left unexplained, and ohg huuanta as well as Dutch want have remained etymological orphans, lacking exact etymological matches outside Germanic. Proto-Germanic *hwandē, as posited by Schmidt, is composed of a total of four morphemes, to wit *kwo-, *-m, *-dhe, and *-(e)h1. I will begin with the formal identification of these four morphemes and then turn to a compositional account of the function of Proto-Germanic *hwandē.
a. The morphological base is the interrogative stem *kwo-. b. *kwo-is followed by a suffix *-m; this morphological parse has recently been proposed by Dunkel (2014; (Dunkel 2014b: 367); pie *an-dhe 'below' > Toch. a āñc 'downwards' (Dunkel 2014b:41-44;cf. Klingenschmitt 1987:175 fn.15 "A āñc < uridg. *h2en-dhe"), cf. āñc tā-'to place below' in e.g. āñc tāluneyo wälts akṣaräṣ pikäṣ 'by putting (a line) beneath, he is writing a thousand akṣaras' (Pinault 2013:211). d. Finally, locatival -dhe may be followed by an instrumental case morpheme *-(e)h1 conveying perlative-directional function, hence pie *dhoh1/dheh1 'all over' . Cf. the pairs Gk. ἀνά: ἄνω 'up' , reflecting pie *h2en-eh1 (> Goth., os, ohg ana 'at, in, upon';cf. Boutkan 1995:377), and κατά: κάτω 'towards below' , πρός: πρόσσω 'forward' (García-Ramón 1997, Zeilfelder 2001. Another possibility is *-dho in directional *-o, on which cf. Dunkel 2014a: 154-161. 1.1.1 Metonymic source-goal shifts Turning to a compositional account of ohg huuanta, the question arises how to derive the functional breadth of ohg huuanta 'for, because' from the four aforementioned morphemes *kwo-m-dhe-h1 or *kwo-m-dh(e)-oh1. Can the ablatival meaning of *kwo-m-dheh1 or *kwo-m-dhoh1 be reconciled with the locativaldirectional suffix *dhoh1/dheh1, and does the combination of two morphemes of opposite meaning (ablatival versus locatival-directional) make sense? In fact, modern and ancient languages offer parallels for just this kind of complex source-goal expression. While it is may be taken for granted that every complete motion-to contains both a starting and an end point, it may be communicationally less important to focus equally on both ends of the event of motion and more salient and hence natural to focus either on the motion's source or on its target. Crosslinguistically, languages show either complex source-goal expressions with source prominence, cf. e.g. In addition to complex source-goal expressions like the foregoing, it is also possible for simple morphemes to acquire source-goal meaning with prominence placed on either the source or the goal. Since projective motion always combines motion away from a starting point and motion towards somewhere, the two aspects are two sides of the same coin, i.e., ablatival and directional motion entail each other. Hence shifts may occur metonymically and the focus may shift a) either from ablatival to locatival and directional by sourceplace/goal metonymy, or less frequently b) from directional to ablatival by goal-source metonymy. Examples of directional to ablatival by goal-source metonymy are comparatively less frequent, cf. e.g.
1.1.2 Paths of functional ramification in source-goal expressions To conclude, ablatival and directional meaning are not contradictory, but mutually inclusive and hence compatible. This and the mechanisms of shifting between ablatival and directional-locatival open a door towards explaining the functional ramifications of ohg huuanta, Dutch want as well as Latin unde and Tocharian b ente, a äntā(ne). The starting point is The indications are that pie interrogative-relative *kwo-m-dheh1, *kwo-m-dhoh1 diversified its functions along the following four paths:

1.2
The functional range of ohg huuanta and ModDutch want The functional domains of ohg huuanta include its deployment as: a) a causal interrogative, b) a (speech-act) causal conjunction (cf. Handschuh 1964:164, Schrodt 2004, and c) a temporal-conditional conjunction. In its inherited and oldest use, ohg huuanta occurs as a causal interrogative. ohg uuanta: quare 'why?' is attested in glosses (ii,70,64; iii,12,63; v,519,5), see Starck and Wells 1983:695, and  In order to reconcile the functional difference between ohg huuanta 'why?' and the etymologically related interrogatives ocs kǫdě, Latin unde 'whence?' , the hypothesis that the ablatival-local function is primary and reason represents a secondary, derived concept turns out to be much likelier than the converse assumption of why → whence. Viewing reason as the primary concept is a priori unlikely, since Indo-European lacks an inherited, cross-linguistically Indo-European Linguistics 5 (2017) 1-30 The observed lexical instability of reason interrogatives is the result of their cognitive complexity. The more abstract a functional concept is, the less homogeneous appears its linguistic expression across related languages. This supposition is borne out both by the interrogatives of pie and by the Indo-European lexicon. Just as the Indo-European languages diverge tremendously in the expression of reason interrogatives, so too on the lexical level the Indo-European languages offer a variety of etymologically heterogeneous expressions to denote the concepts of reason; cf. Buck 1949: 1242 f. (reason, cause). Whereas interrogatives are more lexically stable in the functional domains of person, thing and place, they tend to be considerably less stable in the functional domain of reason. Consequently, the linguistic means of expressing interrogative adverbials for reason exhibit a greater measure of lexical variability (synchronically) and lexical renewal (diachronically).
Among the lexical sources of reason interrogatives, the path of source → reason turns out be quite common, cf. Table 1.
Therefore, ohg huannta 'why' can straightforwardly be derived from pie *kwo-m-dheh1, *kwo-m-dhoh1 'whence' . Ba → Bb: reason interrogative why? → causal speech-act linker and causal conjunction 'for; because' . Already from the beginning of the ohg transmission, the inherited use of ohg huuanta as 'why?' appears as recessive, and is outnumbered by its innovative use as a coordinating causal conjunction, or more precisely as a speech-act linking conjunction. The indications are that ohg huuanta 'why?' and wanne 'why not?' were on the wane as interrogatives already in the (pre-)ohg period, as Indo-European Linguistics 5 (2017) 1-30 they were gradually superseded by analytic expressions like fone wiu, bi wiu 'why?' . Eventually, mhg want was replaced by war umbe, the source of Modern Germ. warum. But ohg huuanta and wanne survived in functional specialization, being relegated to noninterrogative conjunctional uses. ohg huuanta 'for; because' derives from the the well-known shift from why to because; for the interrogative-to-conjunction shift, see the examples and references in § 1. The use of ohg huuanta as an interrogative linker and as the marker of a paratactic speech-act construction turns out to be remarkably robust, persisting in mhg and older Bavarian dialects of German until at least the 19th century as well as from Old Low Franconian (Old Dutch) to present-day Dutch, cf. e.g. In what follows I will show that ohg huuanta and ModDutch want are not isolated formations, as etymological matches can be identified in the ancient Italic and the Tocharian branches of Indo-European. The next two sections examine accordingly the formal and constructional history of Latin unde 'whence' and Tocharian b ente, a äntā(ne), which turn out to differ in interesting respects from the Germanic forms.

2
The etymology of Latin unde

2.1
Phonological reconstruction It may be hypothesized that Latin unde 'whence' is a cognate of ohg huuanta and derives like the latter from pie *kwo-m-dheh1, *kwo-m-dhoh1, including the generalized sandhi-forms with loss of the final laryngeal in pausa. An additional possibility is *kwo-m-dho in directional *-o, for which cf. Dunkel 2014a: 154-161. Based on forms like alicunde and nēcunde, it can safely be assumed that unde had a velar onset originally.

2.2
The functional range of Latin unde The functional range of Latin unde overlaps with that of ohg huuanta. give:perf.3sg 'on whose side was the law, to him it gave the victory. ' (Liv. 21,10,9) Cf. Hofmann and Szantyr (1972:209f.)  In short, Lat. unde 'whence' is formally and functionally reconcilable with ohg huuanta and Dutch want.

3.1
Phonological reconstruction: aphaeresis, dental development and vowel weakening Like ohg huuanta and Latin unde, both Toch. b ente and Toch. a äntā appear as place interrogatives and locative and temporal-conditional conjunctions, thus meeting the minimal conditions for being functionally equatable. The previous proposal, advanced by Adams (2013), sought to derive b ente from PToch. *en-te, where *en-and weakened *än-, in-are from the demonstrative stem *h1eno-, *h1ono- (Adams 2013:69) and "*-te must be from pie ablative *tōd" (Adams 2013: 91). But it is difficult to imagine where the interrogative function of Toch. b ente 'where?' , a äntā 'where?' could come from if the underlying stem is a demonstrative pronoun. An alternative and simpler proposal is to derive Toch. b ente from pie *kwo-m-dho(h1) or *kwo-m-do(h1), and Toch. a äntā(ne) from *kwo-m-dha(h2) or *kwo-m-da(h2) (suffixed with allative *-ah2). This account presupposes the operation of velar aphaeresis (as in Latin), a phonological development pie *-nd-> Toch. -nt-, and vowel weakening in the initial syllable in Tocharian a. Let us address each of these issues in turn.

3.1.1
Aphaeresis in clause-initial interrogatives and wh-movement in Latin, Armenian and Tocharian A case can be made for aphaeresis in clause-initial interrogatives and for whmovement in Latin, Armenian and Tocharian. In general, aphaeresis describes the tendency to eliminate segments in unstressed initial syllables, e.g., Engl. alone → lone. More precisely, however, the evidence demonstrates that aphaeresis tends to target utterance-initial (= sentence-initial) and clause-initial phrasal onsets more often than the onsets of subphrasal onsets that more likely occur utterance-and phrase-internally. This raises the question whether there is a connection between utterance-initial placement and aphaeresis, and what its explanation might be. To begin with, utterance-initial segments and clauseinitial onsets are prone to show weak articulation, as was already observed by Jespersen (1917:6), who referred to the process as prosiopesis 'silencing' and used it to explain the sentence-initial elision of negations. But the deeper mechanism behind utterance initial reduction may be sought in a lag in the articulatory planning and synchronicization of utterance-initial segments.8 Evidence can be adduced that points in this direction. There are contrasting treatments with and without aphaeresis, such as the following.

Aphaeresis in personal pronouns:
Across Indo-European, there is "a cross-linguistic tendency for personal and demonstrative pronouns to aphaeretize" (see Katz 1998:102), which likewise originates from their tendency to occur utterance-initially.

Aphaeresis in interrogative pronouns:
a. In Latin, aphaeresis affects *cunde sentence-and clause-initially, when used as an interrogative and relative, but not phrase-internally necunde. b. Classical Armenian inčc 'something' from *[kw]im=kwid shows the phraseinitial operation of aphaeresis while leaving the onset of the phraseinternal constituent =kwid intact. Another example is the Classical Armenian complementizer etce, univerbated from e and tce pie *[kw]e(h1) te(h1) 'how so?' (cf. Goth. ƕe 'how?'; and oe. þe demonstr. instr. 'by which' and relative particle, Homeric Greek τῆ 'here'). Alternative explanations for the loss of the interrogatives onsets in Italic and Armenian have been proposed, but all of these suffer from drawbacks. In the case of Italic, scholars have invoked resegmentation (indefinite ne|cubi → nec|ubi >> interrogative ubi), but the spread of the resegmented ubi from the negated indefinite to the interrogative would reverse the more natural and expected directionality of analogical extension from basic to derived. In the case of Armenian, it has been proposed to attribute aphaeresis in Arm. interrogatives to a dissimilatory loss *k…k… → Ø…k…, e.g. * HLuw. kwita 'where' would thus support the explanation of Toch. a tā as an aphaeresized *kutā. In Tocharian, aphaeresis shows all the earmarks of a vernacular phonostylistic phenomenon with the possibility of the generalization of phonostylistically reduced variants, as would be the case with Toch. a tā under the present explanation.

Synthesis
The reconstruction of ohg huuanta as an inherited interrogative conforms to a morphological and semantic template that recurs in Latin and Tocharian. The morphological template consists in an ablatival-directional expression, i.e. an ablatival interrogative *kwo-m plus suffixed locatival-directional morpheme, which for all three branches can be posited as *-dhe/o-, Latin and Tocharian additionally allowing for *-de/o-, cf. the overview in Table 2. The three interrogatives and conjunctions, ohg huuanta, Latin unde and Toch. b ente, a äntā(ne), exhibit the same array of functions. Their (original) core function is ablatival-directional. The semantic development involves a possible shift from ablatival to causal meaning, as in the case of ohg huuanta Indo-European Linguistics 5 (2017) 1-30 (whence > why) and Lat. unde, or from ablatival-directional to locatival-directional meaning (whence > where), as in the case of Lat. unde and Toch. b ente, a äntāne. The shift from whence to where accords with the typologically observed preferential focus on goal in motion events. The data presented in this paper thus confirm that in a projective source-goal motion, goal is the unmarked, cognitively more salient concept, cf. Fillmore 1997, Stefanowitsch and Rohde 2004, Zwarts 2010