Abstract
I argue for the reconstruction within Proto-Indo-European morpho-syntax of the phenomenon of “Locative Alternation”, i.e. the alternation for the same verb between a locatum-as-object construction (e.g., English spray paint onto the wall) and a location-as-object construction (e.g., English spray the wall with paint). Drawing inspiration from Iwata’s (2008) lexical-constructional approach to the phenomenon, I discuss its possible relevance to the reconstruction of the morpho-syntax and semantics of PIE verbal formations belonging to three PIE polysemous roots, namely *leu̯g- ‘bend, make bent, twisted, not straight; enclose’, *u̯el- ‘roll round, coil; enclose, enfold’, and *pleh1- ‘be abundant, much; be full, filled’.
1 Introduction
As is well known, Proto-Indo-European (PIE) roots are lexical elements that may be reconstructed by comparing several formations occurring in the lexicon of two or (usually) more Indo-European (IE) languages. From a semantic perspective, however, the reflexes of such PIE roots often attest meanings that are, on the one hand, clearly related, but, on the other, also quite distinct, allowing for the reconstruction of polysemous roots, which must be separated from homophonous roots (i.e., formally identical roots with completely unrelated meanings). In this paper, I propose to explain the polysemy of (at least) some of these roots by taking into account the phenomenon of “Locative Alternation” (LA).
English has a class of so-called “locative alternation verbs”, which may “realize either a locatum (paint or cartons) or a location (the wall or the truck) in direct object position, as in (1)and (2)” (quote and examples (1)–(2) from Iwata 2008: 1; small caps added).
(1) a. Jack sprayed paint onto the wall. (locatum-as-object variant)
b. Jack sprayed the wall with paint. (location-as-object variant)
(2) a. Bill loaded cartons onto the truck.
b. Bill loaded the truck with cartons.
Far from being a peculiar feature of contemporary English, LA is actually a widespread phenomenon already attested in Old English (Sówka-Pietraszewska 2019) and in several other ancient IE languages, such as Ancient Greek (de Boel 2017) and Vedic Sanskrit (Kulikov 2017: 383 with further literature), as well as in other language families, e.g., in Japanese (Iwata 2008: 169–205).
Even though an impressive number of studies have dealt with the LA phenomenon, most of them have approached it from a synchronic perspective, without exploring its relevance for diachronic linguistic analysis and for historical semantics. A notable exception is Haudry’s (1978: 175–177) theory of the “two models”, set forth in his book on the use of cases in Vedic Sanskrit: Haudry interprets this phenomenon as the synchronic manifestation of a gradual diachronic development from a more ancient “modèle 1” (i.e., the location-as-object construction) with the accusative of the “prospective object” (reflecting the goal function, which Haudry regards as the original function of the PIE accusative case) to a more recent “modèle 2” (i.e., the locatum-as-object construction) with the accusative of the “object in contact” (which, in Haudry’s view, was originally expressed with the instrumental case in PIE). While this reconstruction may be plausible for some formations, not all of them easily allow for the assumption of such a diachronic change. In the present study, I rather argue that a reconstructed polysemy may sometimes be explained as the reflex of a synchronic morphosyntactic alternation within PIE; in those cases for which a diachronic evolution may indeed be assumed, the opposite development seems actually more likely to have taken place (i.e., an earlier locatum-as-object construction must be reconstructed beside a younger location-as-object construction).
The analysis set forth in this paper is inspired by Iwata’s (2008) “lexical-constructional” approach to LA in English and Japanese. Previous accounts of LA, such as Pinker’s (1989) lexical rule approach and Goldberg’s (1995) constructional approach, tend to assume that the key to the LA phenomenon is to be found either at the level of the lexicon (Pinker) or at the level of quite abstract “constructions” (i.e., “form and meaning pairings”, as per Goldberg 2006: 3), two assumptions that lead to theoretical and empirical issues (see the critical review in Iwata 2008, especially 1–25 & 119–131). Iwata manages to overcome several of these issues by taking into account constructions both at the “verb-specific” (i.e., lexeme-specific) level and at the verb-class-specific level (i.e., constructions shared by semantic verb classes), an approach that, among other things, has the advantage of being more compatible with current theories of language acquisition (e.g., Tomasello 2003), which show, on the one hand, that lexical meanings are not stored independently from their constructional context and, on the other hand, that more abstract, higher-level constructions are stored beside (and are actually generalized on the basis of) lower-level, verb-specific constructions. I refer to Iwata’s volume for a more detailed discussion of the theoretical framework; in order to schematically introduce it for the purposes of the present paper, examples (3)–(4) represent its application to LA verbs like Engl. spray (3) and load (4); the subject X, being irrelevant for LA, will be ignored here and in the remainder of the paper.
On the one hand, Iwata argues that the (a) locatum-as-object constructions of each of these verbs (“verb-specific constructions”), in, e.g., spray paint onto the wall and load bricks onto the truck, can always be analyzed as particular instances of the more general caused-motion construction generally associated in English with verbs like put and move in, e.g., put a box on the desk (Iwata’s “put-class-specific construction”, hereafter labeled “put-class construction”). Iwata shows how the use of this “verb-class-specific construction” seems to reflect an underlying put-like conceptualization of the events spray and load as caused-motion events.
On the other hand, Iwata demonstrates that the (b) location-as-object verb-specific construction of Engl. spray in, e.g., spray the wall with paint instantiates the same verb-class-specific construction of Engl. cover in, e.g., cover the floor with a rug (Iwata’s “cover-class-specific construction”, hereafter labeled “cover-class construction”), reflecting an underlying cover-like conceptualization of the event spray. In contrast, the (b) location-as-object verb-specific construction of Engl. load in, e.g., load the truck with bricks realizes the same verb-class-specific construction of Engl. fill in, e.g., fill the glass with wine (Iwata’s “fill-class-specific construction”, hereafter labeled “fill-class construction”), correspondingly reflecting an underlying fill-like conceptualization of the event load.
(3) Engl. spray (as per Iwata 2008: 39)
a. locatum-as-object construction
Occurrence:
spray
painty
onto the wallz
Spray-specific construction:
Form:
[spray
NPy
directional-PPz]
Meaning:
“spray Y causing its motion into/onto Z”
put-class construction:
Form:
[Vput
NPy
directional-PPz]
Meaning:
“move Y into/onto Z”
Cf. put:
put
the boxy
on the deskz
b. location-as-object construction
Occurrence:
spray
the wally
with paintz
Spray-specific construction:
Form:
[spray
NPy
with-PPz]
Meaning:
“make Y covered by spraying Z over it”
cover-class construction:
Form:
[Vcover
NPy
with-PPz]
Meaning:
“cause Y to have a layer of Z over it”
Cf. cover:
cover
the floory
with a rugz
(4) Engl. load (as per Iwata 2008: 160–161)
a. locatum-as-object construction
Occurrence:
load
bricksy
onto the truckz
Load-specific construction:
Form:
[load
NPy
directional-PPz]
Meaning:
“transfer object Y onto container Z”
put-class construction:
Form:
[Vput
NPy
directional-PPz]
Meaning:
“move Y into/onto Z”
Cf. put:
put
the boxy
on the deskz
b. location-as-object construction
Occurrence:
load
the trucky
with bricksz
Load-specific construction:
Form:
[load
NPy
with-PPz]
Meaning:
“fill container Y with object Z”
fill-class construction:
Form:
[Vfill
NPy
with-PPz]
Meaning:
“fill Y with Z”
Cf. fill:
fill
the glassy
with winez
As schematically represented in examples (5)–(6), Iwata goes on to demonstrate that his approach has cross-linguistic validity by applying it to Japanese LA verbs like nuru ‘smear’ (5) and tsumeru ‘stuff’ (6). Similarly to their English counterparts spray and load, these verbs’s (a) locatum-as-object verb-specific constructions can always be analyzed as particular instances of the more general caused-motion put-class construction associated in Japanese with verbs like tsukeru ‘attach, put’, whose use thus reflects an underlying put-like conceptualization of the events smear and stuff as caused-motion events.
Correspondingly, the (b) location-as-object verb-specific construction of Japanese nuru ‘smear’ instantiates the same cover-class construction of Jap. oou ‘cover’, reflecting an underlying cover-like conceptualization of the event smear, whereas the (b) location-as-object verb-specific construction of tsumeru ‘stuff’ realizes the same fill-class construction of Jap. ippai-ni suru ‘fill, make full’, reflecting an underlying fill-like conceptualization of the event stuff.
(5) Jap. nuru ‘smear’ (as per Iwata 2008: 178–179)
a. locatum-as-object construction
Occurrence:
kabez-ni
penkiy-o
nuru
wallz-loc
painty-acc
smear
“smear painty on the wallz”
Nuru-specific construction:
Form:
[NPz-ni
NPy-o
nuru]
Meaning:
“smear Y causing its motion into/onto Z”
put-class construction:
Form:
[NPz-ni
NPy-o
Vput]
Meaning:
“move Y into/onto Z”
Cf. tsukeru ‘attach, put’:
kabez-ni
penkiy-o
tsukeru
wallz-loc
painty-acc
put
“put painty on the wallz”
b. location-as-object construction
Occurrence:
kabey-o
penkiz-de
nuru
wally-acc
paintz-with
smear
“smear the wally with paintz”
Nuru-specific construction:
Form:
[NPy-o
NPz-de
nuru]
Meaning:
“make Y covered by smearing Z over it”
cover-class construction:
Form:
[NPy-o
NPz-de
Vcover]
Meaning:
“cause Y to have a layer of Z over it”
Cf. oou ‘cover’:
teeburuy-o
nunoz-de
oou
tabley-acc
clothz-with
cover
“cover the tabley with a clothz”
(6) Jap. tsumeru ‘stuff’ (as per Iwata 2008: 184–185)
a. locatum-as-object construction
Occurrence:
hakoz-ni
ichigoy-o
tsumeru
boxz-loc
strawberryy-acc
stuff
“stuff strawberriesy into the boxz”
Tsumeru-specific construction:
Form:
[NPz-ni
NPy-o
tsumeru]
Meaning:
“stuff Y into/onto Z”
put-class construction:
Form:
[NPz-ni
NPy-o
Vput]
Meaning:
“move Y into/onto Z”
Cf. tsukeru ‘attach, put’:
kabez-ni
penkiy-o
tsukeru
wallz-loc
painty-acc
put
“put painty on the wallz”
b. location-as-object construction
Occurrence:
hakoy-o
ichigoz-de
tsumeru
boxy-acc
strawberryz-with
stuff
“stuff the boxy with strawberriesz”
Tsumeru-specific construction:
Form:
[NPy-o
NPz-de
tsumeru]
Meaning:
“cause Y to be stuffed with Z”
fill-class construction:
Form:
[NPy-o
NPz-de
Vfill]
Meaning:
“cause Y to be full of Z”
Cf. ippai-ni suru ‘fill’:
gurasuy-o
mizuz-de
ippai-ni suru
glassy-acc
waterz-with
full make
“fill the glassy with waterz”
As clear from the examples above and as extensively argued by Iwata, LA verbs in both English and Japanese may always be classified as either cover-class verbs or as fill-class verbs, depending on which conceptualization underlies their location-as-object construction.
The aim of this paper is to explore the consequences of Iwata’s findings to make a case for the reconstruction of LA in Proto-Indo-European and to argue for the relevance of this phenomenon to the historical semantics of some reconstructed roots that notably display polysemy. The methodology adopted here combines the well-established comparative-phraseological approach to the reconstruction of IE morpho-syntax and semantics (e.g., García Ramón 1999; Kölligan 2007; Covini 2017) with the constructional approach to historical linguistics that has proven very successful in recent years (e.g., Dahl & Fedriani 2012; Barðdal & Eythórsson 2012; Fedriani 2014; Luraghi 2021).
To this aim, I will first focus on the possibility of reconstructing LA for cover-class verbs in PIE (§ 2). After briefly revising the current formal and functional reconstruction of the PIE root *leu̯g- ‘be bent, twisted, not straight’ (§ 2.1), I will propose an analysis of some related three-place verbs in Celtic and Germanic as reflexes of the same, originally two-place, factitive verb, PIE *lu-né-g- ‘make bent, twisted, not straight’ (§ 2.2–3). Drawing a parallel with two semantically similar LA verbs in English and Japanese, i.e., wind and maku ‘wind’, respectively (§ 2.4), I will advance a reconstruction of PIE *lu-né-g- as a LA verb (§ 2.5) of the cover class, i.e., a LA verb whose location-as-object construction reflected a cover-like conceptualization. A similar development will then be reconstructed for the reflexes of the polysemous PIE root *u̯el- ‘roll round, coil; enclose, enfold’ (§ 2.6).
In the second part, I will turn to the possibility of reconstructing LA for fill-class verbs in PIE (§ 3). After briefly reviewing the current reconstruction of the PIE root *pleh1- ‘be abundant, much; be full, filled’ and of its factitive verb *pl̥-né-h1- ‘make *pleh1-, fill’ (§ 3.1), I will draw a parallel with semantically similar LA verbs in English and German, i.e., fill and füllen ‘id.’, respectively (§ 3.2), and discuss an even closer correspondence with the Japanese LA verb mitasu ‘accumulate; fill’ (§ 3.3). By analyzing the distribution of its Vedic reflex pr̥nā́-ti ‘fill’, I will advance a reconstruction of PIE *pl̥-né-h1- as a LA verb (§ 3.4). I will then argue that PIE *pl̥-né-h1- may have undergone a development similar to that of Lat. cumulāre and of its French reflex combler: from an etymological locatum-oriented meaning ‘heap up, pile up, accumulate’ to a secondary—but eventually longer-lived—location-oriented meaning ‘fill’ (§ 3.5).
Finally, I will summarize the results of the study and discuss their possible implications for the methodology of IE reconstruction in general (§ 4).
2 Locative alternation with cover-class verbs in PIE
In the present section, I will argue that it is possible to reconstruct LA for PIE verbs that were semantically close to the cover-class LA verbs of English and Japanese.
2.1 PIE *leu̯g- as a Caland root with adjectival meaning ‘be bent, twisted, not straight’
The following reconstructed verbal formations and their corresponding reflexes in attested IE languages are mentioned in LIV2 (416) as the main grounds for the reconstruction of a verbal root “*leu̯(ǵ)- ‘biegen’ ”, i.e., ‘bend’:
-
a nasal-infix present *lu-né-(ǵ)-/lu-n-(ǵ)-´ with reflexes in Celtic, namely OIr. fo-loing ‘(s)he bears, endures’ and in-loing ‘(s)he connects, puts together’, and MW ellyng- ‘set free’ (from *eks-lung-o/e- ‘bend out’); these formations are outcomes of a Proto-Celtic (thematicized) formation *lung-o/e- (Matasović 2009: 249);
-
a R(ø)-ó/é- present *lu(ǵ)-ó/é- with reflexes in several Germanic verbs of the second strong class, e.g., Goth. ga-lūkan ‘(en)close’, us-lūkan ‘open’; these verbs have been traced back to a Proto-Germanic verb *lūk-a- (Seebold 1970: 338–339) in which the originally short u was lengthened by analogy with the long i of the strong verbs of the first class (e.g., *bīt-a- ‘bite’ and *bīd-a- ‘wait’);
-
a less-safely reconstructable perfect *le-lóu̯(ǵ)/lu(ǵ)-, with reflexes in OIr. in-lolaig ‘(s)he tied; imposed’ and possibly in Germanic, e.g., in Goth. ga-lauk ‘(s)he (en)closed’, us-lauk ‘(s)he opened’;
-
a less-safely reconstructable desiderative *li-lu(ǵ)-só/é- with a reflex in OIr. fo-lil ‘(s)he will bear, endure’.
Most formal aspects of this reconstruction can easily be agreed upon (but see below on the root’s final consonant and on the origin of the Germanic presents). As for the functional analysis, however, it is easy to notice that none of these verbal formations seem to attest the root’s alleged primary meaning ‘bend’.
In fact, this semantic reconstruction is rather based on the meaning of nominal or denominative formations of this root (LIV2: 4161), such as Hom. Gk. lúgos ‘flexible twig’ and Lat. luctāre ‘wrestle’ (on which see below). According to the analysis in LIV2, the meaning of the Germanic verbs must be regarded as the result of a semantic shift from ‘bend together’ (a meaning that may have arisen in combination with preverbs like *ga- or *bi-; LIV2: 4161)1 to ‘(en)close’, whereas the meaning of the Celtic verbs is seen as the result of a development from ‘bend in a direction’ to ‘put’ (LIV2: 4162). While these semantic changes are theoretically possible, a unitary explanation, supported by parallels in other languages, of the semantic history of the Celtic and Germanic verbs would certainly be preferable.
As a premise to my analysis, I must first propose a new phonological, morphological, and semantic reconstruction of this root as *leu̯g- ‘be bent, twisted, not straight’, a so-called “Caland root”:
-
As for the root’s phonology, the final consonant must be reconstructed as a pure velar *g, as shown by Lith. lùgnas ‘flexible’, which cannot reflect a root with palatal *ǵ. Even though Lith. lùgnas is etymologically identical to PGmc. *lukka- ‘lock’ (both clearly reflect PIE *lug-nó-; Lühr 1988: 197 & 347), it is often regarded as a dubious member of this lexical family because it does not show any reflexes of Winter’s Law. The latter, however, most probably did not occur in the context *-Vgn- (Rasmussen 1992: 537; cf. Dybo 2002: 498).
-
The root’s primary meaning must be reconstructed as intransitive (adjectival) ‘be bent, twisted, not straight’, rather than transitive ‘bend’. As shown in Table 1, not only most of the primary formations reconstructed for PIE *leu̯g- attest or imply adjectival semantics for this root, but they also display a defined subset of PIE morphological types—belonging to the so-called “Caland system” (Nussbaum 1976; Rau 2009)—which is often attested by a subset of PIE roots (correspondingly dubbed “Caland roots”) with “property-concept” semantics (Dixon 1982: 1–62), i.e., prototypically adjectival meanings, e.g., *h1reudh- ‘be red’, *pleh1- ‘be much, full’ (discussed below, § 3), and *ǵelh3- ‘be yellow’.
Table 1
Morphology and semantics of the reflexes of PIE *leu̯g- and other Caland roots
|
Caland morphology |
PIE *leu̯g- |
PIE *h1reudh- and other Caland roots |
|---|---|---|
|
R(ø)-ó- adjective |
*lug-ó- ‘bent, twisted’ (Hom. Gk. lúgos ‘flexible twig’ with substantivization; cf. various Gk. formations)2 |
*h1rudh-ó- ‘red’ (Latv. ruds ‘reddish’) |
|
R(ø)-nó- adjective |
*lug-nó- ‘bent, twisted’ (Lith. lùgnas ‘flexible’; PGmc. *lukka- ‘lock’ with substantivization) |
†h1rudh-nó- ‘red’ (no known reflexes); cf. *pl̥h1-nó- ‘full’ (PGmc. *fulla-, Ved. pūrṇá- ‘id.’) |
|
R(ø)-ró- adjective |
*lug-ró- ‘bent, twisted’ (Gk. lugrós ‘baneful, mournful’ with metaphorical meaning)3 |
*h1rudh-ró- ‘red’ (Gk. eruthrós, Lat. ruber ‘id.’) |
|
R(ø)-s-ó- adjective derived from R(é)-os-/-es- abstract substantive4 |
*lug-s-ó- ‘that has bending, twisting’ (Lat. luxus ‘dislocated, sprained’) from *léu̯g-os-/-es- ‘bending, twisting’ (no known reflexes) |
*h1rudh-s-ó- ‘that has red color’ (Lat. russus ‘red’) from *h1réu̯dh-os-/-es- ‘red color’ (Gk. eréuthos) |
|
R(ø)-tó- adjective |
*lug-tó- ‘bent, twisted’ (denominative verb: Lat. luctāre ‘wrestle’, if original meaning ‘to twist [oneself or others]’) |
†h1rudh-tó- ‘red’ (no known reflexes); cf. *ǵl̥h3-tó- ‘yellow’ (PGmc. *gulþa- ‘gold’ with substantivization) |
|
Nasal-infix factitive present |
*lu-né-g-/lu-n-g-´ ‘make bent, twisted’ (PCelt. *lung-o/e- ‘connect, put together’; PGmc. *lūka- ‘[en]close’ with remodeling of the nasal-infixed form, see below) |
*h1ru-né-dh-/h1ru-n-dh-´ ‘make red’ (OIr. rondid, °roind ‘redden, color’) |
An adjectival meaning ‘be bent, twisted’ for *leu̯g- would be compatible with the property-concept semantics expected for a Caland root: the opposition straight vs. crooked (sub-class physical property; Rau 2009: 79) is attested, e.g., by Caland formations such as Ved. r̥j-ú- ‘straight, upright’ vs. vaṅk-ú- ‘crooked’ (Rau 2009: 91–92; 9147).
2.2 PCelt. *lung-o/e- ‘connect, put together’ and PGmc. *lūk-a- ‘(en)close’ as reflexes of the PIE nasal-infix factitive *lu-né-g-/lu-n-g-´ ‘make bent, twisted, not straight’
If the analysis of *leu̯g- as a Caland root with adjectival semantics is sound, the transitive ‘bend’ generally reconstructed by LIV2 as the primary meaning of this root on the basis of the Celtic and Germanic verbs must rather be re-formulated as a secondary, stem-specific meaning ‘make bent, twisted’, expected for factitive formations of this root.
Indeed, as shown in the final row of Table 1, the nasal-infix present *lu-né-g-/lu-n-g-´ that underlies PCelt. *lung-o/e- ‘connect, put together’ (and its Celtic congeners) may be analyzed as such a factitive formation with precisely this meaning, with several parallels in similar formations attested by other Caland roots, e.g., the nasal-infix factitive present *h1ru-né-dh-/h1ru-n-dh-´ ‘make red’ (Caland root *h1reudh- ‘red’) that underlies PCelt. *rund-o/e- ‘redden, color’ (OIr. rondid, °roind ‘id.’).
PGmc. *lūk-a- ‘(en)close’ and its reflexes in the Germanic languages—Goth. ga-lūkan ‘(en)close’, OE ge-lūcan ‘(en)close, intertwine’, ON lúka ‘(en)close’, among others—may be traced back to the same PIE nasal-infix present as well. As argued by Mottausch (2013: 85 & 157–161), PIE *lu-né-g-/lu-n-g-´ and the majority of inherited nasal-infix presents were systematically remodeled in Germanic as thematic formations based on the aorist or perfect stems. The PGmc. present stem *luk-a- (later analogically lengthened to *lūk-a-) may thus be analyzed as an innovative Germanic remodeling of the inherited nasal-infix present (based on the weak preterite stem *luk-, the regular outcome of the PIE weak perfect stem *[le-]lug-), and not as the reflex of an inherited R(ø)-ó/é- present (which would have no parallels in other IE languages).
Further support for this view comes from the fact that R(ø)-ó/é- presents are a morphological type that is not often attested for Caland roots, and, even if it were, one would rather expect intransitive semantics for the corresponding verb (e.g., ‘be/become bent, twisted’), as R(ø)-ó/é- presents were not factitive formations: the transitive meaning of PGmc. *lūk-a- ‘(en)close’ must thus reflect that of the PIE factitive nasal-infix present.
Correspondingly, the meaning of the PIE perfect stem *le-lóu̯g/lug- reconstructed on the basis of its Celtic and Germanic reflexes must originally have been intransitive as well: as argued by Rau (2013: 267), a perfect built on a Caland root usually functions either as a perfecto-present or as a “nacto-static” perfect (“which indicates that the subject is in the state denoted by the root as the result of a change of state”). The transitive semantics of, e.g., OIr. in-lolaig ‘(s)he joined; imposed’ and Goth. ga-lauk ‘(s)he (en)closed’ must thus have analogically spread from the present as well.
To sum up briefly: the transitive meanings of the Celtic and Germanic verbal stems of PIE *leu̯g- ‘be bent, twisted’ must be traced back to the originally stem-specific meaning of the nasal-infix factitive present *lu-né-g-/lu-n-g-´ ‘make bent, twisted’, underlying both the reflexes of PCelt. *lung-o/e- ‘connect, put together’ (with thematicization) and the reflexes of PGmc *lūka- ‘(en)close’ (with heavy remodeling of the present stem after the weak preterite stem). The transitive semantics (and the corresponding argument structure construction, on which see the following section) of the PIE present stem *lu-né-g-/lu-n-g-´ ‘make bent, twisted’ must have analogically spread to the other non-factitive stems (e.g., the perfect stem) either already in PIE or in later linguistic stages (a trivial diachronic development within ancient IE languages).
2.3 From one PIE two-place verb to several IE three-place verbs
The reconstruction above accounts for the fact that the Celtic and Germanic transitive verbs must all be traced back to a PIE root with primary intransitive (adjectival) semantics. Further reasoning is needed, however, to explain how the two apparently very different basic meanings attested in Celtic and Germanic, namely ‘connect, put together’ and ‘(en)close’, may have diverged from a single one. The similarities and differences between these meanings lie at the interface between syntax and semantics:
-
the Celtic verbs meaning ‘connect, put together’ entail a subject X who moves an object Y (a locatum) onto a location Z;
-
the Germanic verbs meaning ‘(en)close’ entail a subject X who surrounds an object Y (a location) by means of a (curved) locatum Z placed around it.
Each of these verbs thus requires three participants, whereas the original meaning of the PIE factitive verb, namely ‘make bent, twisted’, only entailed two, a subject and a single object which is made curved: in order to understand this development, we must try to imagine how the PIE verb may have acquired a third argument.
A possible clue may be provided by the contexts in which the related term Hom. Gk. lúgos ‘flexible twig’ (reflex of the substantivization of an inherited adjective *lug-ó- ‘bent, twisted’, see above) occurs: these are exclusively binding contexts, see, e.g., (7)–(8).
(7)
hṓ
pot
᾽ Achilleùs
/ […]
dídē
móschoisi
lúgoisi
rel.acc.du5
once
Achilles.nom
bind.impf.3sg
twig.dat.pl
flexible_twig.dat.pl
“These two had Achilles once bound with flexible twigs” (Il. 11.104–106)
(8)
toùs
akéōn
synéergon
eustrephéessi
lúgoisin
det.acc.pl
silently
bind_together.impf.3sg
well_twisted.dat.pl
flexible_twig.dat.pl
“These (rams) I silently bound together with well-twisted twigs” (Od. 9.427)
Hom. Gk. lúgos was regularly employed to specifically refer to ‘twigs bent, twisted’ around other items for the purpose of binding them, e.g., around the arms of Achilles’s prisoners in (7).6 It is thus conceivable that the related PIE factitive formation *lu-né-g-/lu-n-g-´ ‘to make bent, twisted’, originally a two-place verb exclusively entailing a subject X and an object Y, became a three-place verb by being frequently used in the sense ‘make something bent, twisted around something else as to enclose it’, to refer to an event in which a twig-like or rope-like item Y was bent or twisted by an agent X around an item Z, resulting in the binding or wrapping of the latter.
2.4 Engl. wind and Jap. maku ‘id.’ as LA verbs
This possible usage of PIE *lu-né-g-/lu-n-g-´ closely resembles the usage of the English verb wind in its transitive sense “[t]o turn or pass (something) around something else so as to encircle or enclose it and be in contact with it; to twine, twist, fold, or wrap (something) about, round, or upon something else” (OED, s.v. wind, 14.a). The latter may be referred to as wind’s locatum-as-object construction, occurring, e.g., in (9) (a Middle English quote from year 1303) and (10) (from year 1916; both examples are from OED).
Engl. wind, however, may also be used with a symmetrically opposite construction, in the sense “[t]o encircle with or enclose in something passed round and in contact; †to wrap up; †to embrace, enfold in the arms; now, in ordinary prose use, only of binding a thing round with tape, wire, or the like” (OED, s.v. wind, 16.a). In other words, the verb wind attests LA (pace Pinker 1989: 126), given that it occurs in a location-as-object construction, e.g., in (11) (a Middle English quote from year 1300) and (12) (from year 1885; both examples from OED).
(9)
Aboute
þe
body
a
rope
þey
wonde
Around
the
body
a
rope
they
wind.pret.3pl
“They wound a rope around the body” (R. Mannyng, Handlyng Synne 8055)
(10) [He] began to wind about his neck a dark blue muffler. (J.J. Bell, Little Grey Ships 18)
(11)
And
wind
it
siþen
well
wit
wand.
and
wind.impt
it
after
well
with
flexible_twigs
“And, after, wind it well with flexible twigs.” (Cursor Mundi 1672)
(12) And wind the front of youth with flowers. (Ld. Tennyson Anc. Sage 97)
Iwata (2008: 72) discusses the LA phenomenon with respect to Engl. wind and its Japanese counterpart maku ‘wind’ (2008: 183–184). As schematically represented in (13)–(14), Iwata regards the (a) locatum-as-object constructions instantiated by Engl. wind a bandage around the arm and Jap. udeni houtaio maku “id.” as reflexes of the same caused-motion constructions generally employed with verbs meaning put, such as Engl. put and Jap. tsukeru ‘attach, put’. In contrast, Iwata analyzes the (b) location-as-object constructions instantiated by Engl. wind the arm with a bandage and Jap. udeo houtaide maku “id.” as reflexes of the same constructions generally employed with verbs meaning cover, such as Engl. cover and Jap. oou ‘id.’.
(13) Engl. wind (as per Iwata 2008: 72)
a. locatum-as-object construction
Occurrence:
wind
a bandagey
around the armz
Wind-specific construction:
Form:
[wind
NPy
directional-PPz]
Meaning:
“to turn Y around Z so as to enclose it and
be in contact with it”
put-class construction:
Form:
[Vput
NPy
directional-PPz]
Meaning:
“move Y into/onto Z”
Cf. put:
see (3) above.
b. location-as-object construction
Occurrence:
wind
the army
with a bandagez
Wind-specific construction:
Form:
[wind
NPy
with-PPz]
Meaning:
“to enclose Y in Z passed round and in
contact”
cover-class construction:
Form:
[Vcover
NPy
with-PPz]
Meaning:
“cause Y to have a layer of Z over it”
Cf. cover:
see (3) above.
(14) Jap. maku ‘wind’ (as per Iwata 2008: 183–184)
a. locatum-as-object construction
Occurrence:
udez-ni
houtaiy-o
maku
armz-loc
bandagey-acc
wind
“wind a bandageY around the armZ”
Maku-specific construction:
Form:
[NPz-ni
NPy-o
maku]
Meaning:
“to turn Y around Z so as to enclose it and
be in contact with it”
put-class construction:
Form:
[NPz-ni
NPy-o
Vput]
Meaning:
“move Y into/onto Z”
Cf. tsukeru ‘attach, put’:
see (5) above.
b. location-as-object construction
Occurrence:
udey-o
houtaiz-de
maku
army-acc
bandagez-with
wind
“wind the army with a bandagez”
Maku-specific construction:
Form:
[NPy-o
NPz-de
maku]
Meaning:
“to enclose Y in Z passed round and in
contact”
cover-class construction:
Form:
[NPy-o
NPz-de
Vcover]
Meaning:
“cause Y to have a layer of Z over it”
Cf. oou ‘cover’:
see (5) above.
2.5 Reconstructing LA for PIE *lu-né-g-/lu-n-g-´ ‘make bent, twisted’
PIE *lu-né-g- ‘make bent, twisted’ may have been used as a three-place verb in the same sense as Engl. wind and Jap. maku, i.e., ‘make something bent, twisted around something else as to enclose it’. If this were the case, the polysemy attested by its reflexes in Celtic and Germanic may be traced back to the phenomenon of LA. As shown in (15), the semantics ‘put, connect’ of the Celtic verbs (e.g., OIr. in-loing) may have arisen within the (a) locatum-as-object construction of PIE *lu-né-g-, i.e., the caused-motion construction regularly employed with verbs meaning put (e.g., PIE *dhé-dheh1- : Ved. dádhā-ti, Gk. títhēmi, LIV2: 136). In contrast, the semantics ‘(en)close’ of the Germanic verbs (e.g., ON lúka) may have arisen within the (b) location-as-object construction of PIE *lu-né-g-, the same construction usually employed with the semantic class of verbs meaning cover (e.g., PIE *tég-o/e- : Lat. teg-ō, LIV2: 589; PIE *sḱod-éi̯o/e- : Ved. chād-áya-ti ‘id.’, LIV2: 546).
(15) IE reflexes of PIE *lu-né-g-
a. OIr. in-loing (pret. ellacht) as reflex of the locatum-as-object construction
Occurrence:
ellacht
imhuaimy
im rithz
put/connect.pret.3sg
harmony.accy
around course.accz
“He put/connected the harmonyy around the coursez” (RC xx,256,§ 61)
OIr. in-loing construction:
Form:
[in-loing
ACCy
im ACCz]
Meaning:
“move Y to fit into/onto Z”
PIE put-class construction of *lu-né-g-:
Form:
[*lu-né-g-
ACCy
LOC/ACCz]
Meaning:
“put Y into/onto Z by bending/twisting it”
Cf. Ved. dádhā-ti ‘put’ and Gk. títhēmi ‘id.’ (PIE *dhé-dheh1-):
suastíy […]
yóniṣuz […]
dadhātana
well-being.accy
womb.loc.plz
put.imp.2pl
“Put well-beingy into wombsz” (RV 10.63.15cd)
khrḗmatay […]
mukhôiz […]
theímen
good.acc.ply
recess.datz7
put.opt.aor.1pl
“Let us put goodsy in the innermost recessz” (Od. 13.363–364)
b. ON lúka (pret. lauc) as reflex of the location-as-object construction
Occurrence:
Lauc hannx
micy
scioldomz
lock.pret.3sg 3sg.nomx
1sg.accy
shield.dat.plz
“Hex locked mey with shieldsz” (Hlr. 91–94)
ON lúka construction:
Form:
[lúka
ACCy
DATz]8
Meaning:
“cause Y to have a layer of Z around it”
PIE cover-class construction of *lu-né-g-:
Form:
[*lu-né-g-
ACCy
INSTRz]
Meaning:
“make Y covered by bending/twisting Z around it”
Cf. Lat. teg-ō, -ere ‘cover’ (PIE *tég-o/e-) and Ved. chād-áya-ti ‘id.’ (*sḱod-éi̯o/e-):
corpusy […]
texit […]
pallioz
body.accy
cover.pf.3sg
mantle.ablz
“(She) covered the bodyy with a mantlez” (Cic. Div. 2.69.143)
mármāṇiy […]
vármaṇāz
chādayāmi
vulnerable_place.acc.ply
armor.insz
cover.1sg
“I cover the vulnerable placesy with armorz” (RV 6.75.18a)
In each branch, the eventual selection of one of the two constructions as the verb’s main argument structure construction (at the expenses of the other one) led to a semantic narrowing of its lexical meaning, which came to be restricted to only ‘put, connect’ in Celtic and only ‘(en)close’ in Germanic.
The data discussed in this section point to the reconstruction of LA in PIE morpho-syntax, or at the very least for the morpho-syntax of the PIE verb *lu-né-g- ‘make bent, twisted’; as detailed in the next and following sections, however, the same phenomenon may be reconstructed for other verbs as well.
2.6 Reconstructing LA for PIE *u̯el- ‘roll round, coil; enclose, enfold’
The reconstruction discussed here for PIE *leu̯g- may have a further possible parallel in the root listed in LIV2 as “1.*u̯el-” ‘enclose, cover’ (LIV2: 674) and “2.*u̯el-” ‘turn, wind’ (LIV2: 675), two roots that were probably originally identical, as long noted (cf. LIV2: 6751). Haudry (1978: 264) already explained the polysemy of the reflexes of this root with his theory of the “two models” (on which see above, § 1), assuming, as expected within his framework, that the “modèle 1” (the location-as-object construction) is older than the “modèle 2” (the locatum-as-object construction).
By analyzing the contexts of use of the Homeric reflexes of this root (16), however, we may alternatively reconstruct a development opposite to Haudry’s theory and similar to the one proposed above for PIE *leu̯g-. More precisely, the verbal formations of PIE *u̯el- may originally have meant ‘turn, wind’ (a locatum Y, e.g., around a location Z) and may have correspondingly had a default (a) locatum-as-object construction. Later, by the phenomenon of LA, the verbs may have acquired a secondary meaning ‘cover’ (a location Z by means of a locatum Y) when used in their (b) location-as-object construction.
(16) Homeric Greek reflexes of PIE root *u̯el- ‘roll round, coil; enclose, enfold’
a. Hom. Gk. aor. pass. elustheìs ‘rolled round/coiled’ as reflex of locatum-as-object construction
Passivized occurrence:
Y
hupò gastér’z
elustheìs
(Iy)
beneath belly.accz
roll_round.ptc.aor.pass
“(Iy, who was) rolled round/coiled beneath the bellyz” (Od. 9.433)
Corresponding Gk. active construction:
Form:
[ACCy
(hupo) ACCz
elúō]
Meaning:
“roll round, coil Y to fit into/onto (under) Z”
PIE put-class construction of *u̯el-verbs:
Form:
[ACCy
LOC/ACCz
VERB (root *u̯el-)]
Meaning:
“put Y into/onto Z by rolling it round, coiling it”
Cf. Ved. dádhā-ti ‘put’ and Gk. títhēmi ‘id.’:
see (15) above.
b. Hom. Gk. eilúō ‘enclose, enfold, wrap’ as reflex of location-as-object construction
Occurrence:
miny […]
eilúsō
psamáthoisinz
3sg.accy
wrap.fut.1sg
sand.dat.plz
“I will enfold himy in sandsz” (Il. 21.318–319)
Gk. construction:
Form:
[ACCy
eilúō
DATz]9
Meaning:
“cause Y to have a layer of Z around it”
PIE cover-class construction of *u̯el-verbs:
Form:
[ACCy
VERB (root *u̯el-)
INSTRz]
Meaning:
“make Y covered by rolling round, coiling Z around it”
Cf. Lat. teg-ō ‘cover’ and Ved. chād-áya-ti ‘id.’:
see (15) above.
2.7 Section summary: cover-class LA verbs in PIE
Let us recapitulate the results of this section. The PIE root currently reconstructed as *leu̯(ǵ)- ‘to bend’ in LIV2 (416) should rather be reconstructed as a Caland root *leu̯g- with an adjectival meaning ‘be bent, twisted, not straight’. The factitive semantics ‘bend’ of some of its reflexes must have analogically spread from the expected meaning of the root’s nasal-infix verb *lu-né-g- ‘make bent, twisted’.
The latter was originally a two-place verb (only entailing a subject X and an object Y), which could also be used as a three-place verb in a locatum-as-object construction [*lu-né-g- ACCy LOC/ACCz], meaning “put Y into/onto Z by bending/twisting it” and reflecting the PIE construction of put-class verbs (e.g., *dhé-dheh1- ‘put’). By the phenomenon of LA, however, the verb also developed a location-as-object construction [*lu-né-g- ACCy INSTRz] meaning “make Y covered by bending/twisting Z around it”, reflecting the usual PIE construction of cover-class verbs (e.g., *tég-o/e- ‘cover’).
In each branch, only one construction was eventually selected as the verb’s main argument structure construction, at the expense of the other one: the locatum-as-object construction associated with put semantics in Celtic and the location-as-object construction associated with cover semantics in Germanic. The corresponding semantic narrowing led to the attested meanings of the reflexes of PIE *lu-né-g- ‘make bent, twisted’, namely ‘put, connect’ in Celtic (e.g., OIr. in-loing) and ‘(en)close’ in Germanic (e.g., ON lúka).
A similar development may have taken place in the history of the PIE root currently listed in LIV2 as two homophonous roots, “1.*u̯el-” ‘enclose, cover’ (LIV2: 674) and “2.*u̯el-” ‘turn, wind’ (LIV2: 675). The root’s verbal formations may have had an original primary meaning ‘turn, wind’, which allowed for a locatum-as-object construction [VERB (*u̯el-) ACCy LOC/ACCz], meaning “put Y into/onto Z by rolling it round, coiling it” and reflecting the PIE construction of put-class verbs. The root’s verbal formations may have then entered LA and acquired a secondary meaning ‘enclose, cover’ when used in the location-as-object construction [VERB (*u̯el-) ACCy INSTRz], meaning “make Y covered by rolling round, coiling Z around it” and reflecting the PIE construction of cover-class verbs.
At this point, a consideration is in order. PIE *leug- ‘be bent, twisted, not straight’ may be reconstructed as a Caland (prototypically adjectival) root, whose corresponding factitive verb *lu-né-g- is expected to have meant ‘make bent, twisted, not straight’, which allows us to identify the location-oriented sense ‘enclose’ as the secondary one. The same is not possible for the PIE root *u̯el-, however, as the latter is not a Caland root and is thus not expected to have had adjectival primary semantics. In this case, the safest option would thus be to limit ourselves to reconstruct *u̯el- as a PIE root whose verbal formations synchronically displayed LA and could thus be used in both a locatum-as-object construction and a location-as-object one. It must be noted, however, that LA has been widely recognized to not occur with verbs that have primary cover semantics, like Engl. cover or Jap. oou ‘cover’ (see Iwata 2008: 81–82 & 176). If the same restriction was valid for LA in PIE as well (which is not necessarily the case, but certainly possible), it would thus be far more likely for the cover semantics of PIE *u̯el- to have developed from ‘turn, wind’ than the other way around.
3 Locative alternation with fill-class verbs in PIE
In the present section, I argue that PIE had at least one LA verb which may be compared to the alternating fill-class verbs of English and Japanese.
3.1 PIE *pleh1- as a Caland root with adjectival semantics ‘be abundant, much; be full, filled’
PIE *pleh1- is a root whose meaning is currently reconstructed as ‘become full’ in LIV2 (482–483; “sich füllen, voll werden”), where several PIE verbal formations of this root are reconstructed. The formation that is most relevant to the topic of this paper is:
-
the nasal-infix present *pl̥-né-h1-/pl-n-h1-´, with reflexes in, e.g., Ved. pr̥nā́ti ‘fills’, Av. pərənā ‘fill!’, Arm. lnowm ‘fill’, Alb. m-blon ‘fills’, Lat. polleō, -ēre ‘be strong, be capable of’, and OIr. do-lín ‘(over)flow’.10
As is well known, this root is also a very prominent Caland root (cf., e.g., Rau 2009: 174): it is attested by several formations, mostly nominal or denominative, that correspond to some of the most important morphological types of the Caland system, a selection of which is listed in Table 2.
Table 2
Morphology and semantics of the reflexes of PIE *pleh1- and other Caland roots
|
Caland morphology |
Root *pleh1- |
Other Caland roots |
|---|---|---|
|
R(ø)-u- non-resultative adjective |
*pl̥h1-ú- ‘much, abundant’ (Ved. purú- ‘id.’, Av. pouru- ‘id.’, et al.; PGmc. *felu-/fulu- ‘id.’) |
*gu̯r̥h2-ú- ‘heavy’ (Ved. gurú- ‘id.’, Gk. barús ‘id.’) |
|
R(ø)-ró- non-resultative adjective |
* pl̥h1-ró- ‘much, abundant’ (Lat. plērusque ‘most, a majority’) |
*h1rudh-ró- ‘red’ (Gk. eruthrós, Lat. ruber ‘id.’) |
|
R(e/ø)-nó- patientive/resultative adjective |
*pl̥h1-nó- ‘full, filled’ (Ved. pūrṇá- ‘id.’; PGmc. *fulla- ‘id.’) |
*tep-nó- ‘hot, heated’ (cf. OIr. tene, -ed ‘fire’, YAv. tafnah- ‘heat’) |
|
Nasal-infix factitive present |
*pl̥-né-h1-/pl̥-n-h1-´ ‘make full’ (Ved. pr̥nā́-ti ‘fills’, Av. pərənā ‘fill!’, …) |
*h1ru-né-dh-/h1ru-n-dh-´ ‘make red’ (OIr. rondid, °roind ‘redden, color’) |
As mentioned above, PIE roots whose derivatives attest the morphological types of the Caland system usually have “property concept” semantics, i.e., prototypically adjectival meanings; correspondingly, Rau (2009: 84) classifies the reflexes of PIE *pleh1- within the property-concept category “Dimension/Physical Property (thick/abundant)”. However, as can be observed in Table 2, this root was apparently polysemous: it occurred both in derivatives implying a primary semantics ‘(be/become) much, abundant’ and in formations implying ‘(be/become) full, filled’. Among the formations listed above, it is actually possible to notice a clear-cut distribution:
-
the ‘much, abundant’ meaning seems to be mainly attested by non-resultative formations, such as the -u- and -ró- adjectives;
-
the ‘full, filled’ meaning seems to be the exclusively attested by the factitive verbal stem *pl̥-né-h1-/pl-n-h1-´ ‘make full’ and by the -nó- adjective *pl̥h1-nó- ‘full, filled’; the latter was a morphological type that prominently derived patientive/resultative adjectives opposed to non-resultative ones and corresponding to factitive verbal stems (Rau 2009: 174–175).
In the remainder of this section, I argue that the phenomenon of LA may be reconstructed for the PIE nasal-infix stem *pl̥-né-h1-, and that this reconstruction may be relevant to understand how the root came to be polysemous. More precisely, I propose that the connection between the two primary meanings of PIE *pleh1-, ‘be/become much, abundant’ and ‘be/become full, filled’—the former being mostly attested by non-resultative adjectives and the latter by factitive and resultative formations—can be accounted for by reconstructing LA for the PIE factitive stem *pl̥-né-h1-, which thus meant both ‘make much, abundant (i.e., accumulate, heap up)’ when occurring in the locatum-as-object construction and ‘make full, filled (i.e., to fill)’ when occurring in the location-as-object construction.11
An important premise to my proposal is that verbs meaning ‘fill’ seem to attest LA in several languages (both IE and non-IE), as detailed in the following sections.
3.2 Engl. fill and Germ. füllen ‘id.’ as LA verbs
The first close parallel for the analysis of PIE *pl̥-né-h1- ‘make full, fill’ as a LA verb is the English verb fill itself, for which an (obsolete and archaic) locatum-as-object construction is attested in the sense “[t]o put (wine, etc.) into a vessel with the view of filling it; hence, [t]o pour out” (OED, s.v. fill, 13.), e.g., in (17) (from year 1615; from OED). As noted in OED, an analogous usage is also well-attested in German, where the verb füllen ‘id.’ occurs in a similar construction in, e.g., (18) (from year 1551; from DWB, s.v. füllen, 4).
Of course, the most frequent current usage of Engl. fill is rather within a location-as-object construction as in example (19) (year 1886; from OED), in the primary sense “[t]o supply with as much as can be held or contained; to put or pour something into (a receptacle) till no more can be received” (OED, s.v. fill, 1.). The same location-as-object construction also obviously occurs in German, see, e.g., (20) (year 1754; from DWB, s.v. füllen, 1).
(17) Having filled it [Milk] into a clean vessel. (G. Markham Eng. House-wife ii. i. 12)
(18)
dasz
du
vil
speisz
darein
mögst
füllen
so_that
you
much
food
in_there
can
fill.inf
“so that you can fill much food in there” (Scheit grobianus C ija)
(19) The broken … gentleman … filling his pockets with fairy bank-notes. (D.C. Murray Cynic Fortune vi)
(20)
vier
taschen
füllt
der
held
mit
seinen
tabaks
dosen
four
bags
fill.3sg
art
hero
with
his.dat.pl
tobacco
tin.dat.pl
“The hero fills four bags with his tobacco tins” (Dusch verm. werke 163)
Iwata (2008: 84–85) acknowledges that Engl. fill marginally attests LA and analyzes the locatum-as-object construction as the result of the verb being used loosely with a meaning “to put a large amount of substance into something” (i.e., without actually making the location ‘full’ of the locatum). Since the locatum-as-object construction is the reflex of an underlying put-like semantic conceptualization, “[i]t comes as no surprise, then, that fill may be found in the locatum-as-object frame”, when it is used in this sense (Iwata 2008: 84).
Even though the put-class construction of Engl. fill and Germ. füllen is an interesting parallel to our proposal, it should be noted right away that it is surely a secondary development from their etymological fill semantics: both these verbs must be traced back to PGmc. *full-ija- ‘make full, fill’ (also the source of Goth. fulljan, ON fylla, among others), a factitive denominative verb of the adjective *full-a- ‘full’, the source of Engl. full and Germ. voll (as well as Goth. fulls, ON fullr, OS full, among others).12
3.3 Jap. mitasu ‘fill’ as a LA verb
As further argued by Iwata (2008: 196–197) and schematically represented in (21), a much more prominent case of LA is actually attested by the Japanese counterpart of Engl. fill. Jap. mitasu ‘fill’ may occur both in (a) the same locatum-as-object construction generally employed with verbs meaning put, as well as in (b) the location-as-object construction expected for a verb meaning fill.
(21) Jap. mitasu ‘fill’ (as per Iwata 2008: 196–197)
a. locatum-as-object construction
Occurrence:
gurasuz-ni
mizuy-o
mitasu
glassz-loc
watery-acc
fill
“fill watery into the glassz”
Mitasu-specific construction:
Form:
[NPz-ni
NPy-o
mitasu]
Meaning:
“to put Y into Z causing Y to be abundant”
put-class construction:
Form:
[NPz-ni
NPy-o
Vput]
Meaning:
“move Y into/onto Z”
Cf. tsukeru ‘attach, put’:
see (6) above.
b. location-as-object construction
Occurrence:
gurasuy-o
mizuz-de
mitasu
glassy-acc
waterz-with
fill
“fill the glassy with waterz”
Mitasu-specific construction:
Form:
[NPy-o
NPz-de
mitasu]
Meaning:
“cause Y to be full of Z”
fill-class construction:
Form:
[NPy-o
NPz-de
Vfill]
Meaning:
“cause Y to be full of Z”
Cf. ippai-ni suru ‘fill’:
see (6) above.
While LA is a marginal phenomenon for Engl. fill, it is not at all marginal for Jap. mitasu: Iwata (2008) explains this asymmetry with the fact that these verbs are slightly different in respect to their derivation and semantics.
-
Engl. fill is the causative corresponding to full (see also above), an adjective which may only be predicated of containers: fill thus means ‘make something full’ (usually of containers).
-
Jap. mitasu is the transitive counterpart to an intransitive verb mitiru ‘be abundant, full’, which can be predicated both of abundant liquids and full containers: mitasu ‘make something mitiru’ can thus mean both ‘make something abundant’ (of liquids) and ‘make something full’ (of containers).
The same is true for the synonym ippai-ni suru ‘fill’ (counterpart to the intransitive verb ippai-ni naru ‘become full’) and for the semantically close verb tsumarasu ‘clog’ (matching the intransitive verb tsumaru ‘get clogged’), both of which display LA as well.
3.4 PIE *pl̥-né-h1- ‘make abundant, accumulate; make full, fill’ as a LA verb
As detailed above, the primary semantics of the PIE root *pleh1- were—as for all Caland roots—of the property-concept kind, and thus inherently intransitive, and may be reconstructed as alternating between ‘be much, abundant’ and ‘be full, filled’. This polysemy mirrors that of Jap. mitiru ‘be abundant; be full’ and of the other Japanese intransitive verbs mentioned in the previous section. Correspondingly, in the same way as Jap. mitiru’s transitive counterpart mitasu ‘fill’ and the other Japanese transitive verbs mentioned above display LA, the main factitive formation belonging to PIE *pleh1- may have been a LA verb as well.
Indeed, there is evidence that the PIE nasal-infix stem *pl̥-né-h1- ‘make something *pleh1-’ occurred both in a locatum-as-object construction, in the sense ‘make abundant, accumulate’ (e.g., liquids in a container), as well as in a location-as-object construction, in the sense ‘make full; fill’ (e.g., containers with liquids). More precisely, this reconstruction of PIE *pl̥-né-h1- finds support in the distribution of its Vedic reflex, the verb pr̥nā́-ti. The latter usually means ‘fill’ and is correspondingly construed with the location as object in the accusative case and the locatum in the instrumental or genitive. However, as long noted (cf. Grassmann 1873: 775–777), the same verb may also occur in the sense ‘fully grant’ with the locatum as object in the accusative case, with or without a beneficiary in the dative case.
As shown in (22), within our framework the meaning ‘fully grant’ may reflect (a) the locatum-as-object construction of PIE *pl̥-né-h1- in its sense ‘make abundant, accumulate’, matching the caused-motion construction widely employed with verbs meaning put (e.g., PIE *dhédheh1- : Ved. dádhā-ti, Gk. títhēmi); the dative of the beneficiary may also be understood locationally as expressing direction, reflecting the probably original allative function of the PIE dative.13 In contrast, the more frequent construction with the well-known fill-semantics may reflect (b) the location-as-object construction of PIE *pl̥-né-h1-, in its sense ‘make full, fill’.
(22) Vedic reflexes of PIE *pl̥-né-h1- ‘make much, full’
a. Ved. pr̥nā́-ti ‘fully grant, accumulate’ as reflex of the locatum-as-object construction
Occurrence:
pr̥ṇītá
bheṣajáṁy […]
tanúvez máma
fill.imp.2pl
healing.accy
body.datz 1sg.gen
“(O Waters,) fully grant healingy for/onto my bodyz” (RV 1.23.21ab)
Vedic construction:
Form:
[pr̥nā́-ti
ACCy
DATz]
Meaning:
“move a great quantity of Y for/onto Z”
PIE put-class construction of *pl̥-né-h1-:
Form:
[*pl̥-né-h1-
ACCy
DATz]
Meaning:
“put Y onto Z by accumulating it”
Cf. Ved. dádhā-ti and Gk. títhemi (PIE *dhé-dheh1-):
gíray índrāya […]
dhehí
tanúvez
song.acc.ply Indra.dat
put.imp.2sg
body.datz
“Put songsy for Indra for/onto his bodyz” (RV 8.96.10cd)
koleôiz […]
áory
théo
sheath.datz
sword.accy
put.imp.aor.2sg.mid
“Put up the swordy in the sheathz” (Od. 10.333)
b. Ved. pr̥nā́-ti ‘make full, fill’ as reflex of the location-as-object construction
Occurrence:
sómebhirz […]
pr̥ṇatā […]
índramy
soma_juice.ins.pl
fill.imp.2pl
Indra.acc
“fill Indray with soma juicesz” (RV 2.14.10b)
Vedic construction:
Form:
[INSTRz
pr̥nā́-ti
ACCy]
Meaning:
“cause Y to be full of Z”
PIE fill-class construction of *pl̥-né-h1-:
Form:
[INSTRz
*pl̥-né-h1-
ACCy]
Meaning:
“make Y full by accumulating Z inside it”
Cf. Ved. pr̥-ná-k-ti (PIE *pr̥-né-k-) and Gk. epistéphō ‘fill to the brim’ (PIE *stegwh-o/e-):
támy ít
pr̥ṇakṣi […]
rāyā́z
3sg.acc part
fill.2sg
wealth.ins
“Just himy do you fill with wealthz” (RV 6.15.11d)
krētêrasy
epestépsanto
potoîoz
bowl.acc.pl
fill_to_the_brim.aor.3pl.mid
drink.gen14
“they filled the bowlsy brim full of drinkz” (Il. 9.175)
The polysemy of Ved. pr̥nā́-ti was actually explained differently by Kuiper (1938: 314–315; accepted by LIV2: 4754), namely as the result of the conflation in Vedic of two originally different PIE formations, *pl̥-né-h1- ‘fill’ and *pr̥-né-h3- ‘procure, give’, the latter being otherwise attested exclusively by OIr. -ern* ‘bestow’ (LIV2: 474).
This is certainly possible, but unnecessary: as argued by Haudry (1978: 235–236), the locatum-as-object construction (“modèle 2” in Haudry’s terminology) of Ved. pr̥nā́-ti has parallels in Vedic itself and matches, e.g., that of another IE reflex of the root *pleh1-, namely the Lithuanian verb pìlti ‘fill, pour, pile up’ (reflex of an innovative post-PIE formation *pl̥h1-ó/é-), which may also be construed with the locatum as object in the accusative case and the dative of the beneficiary. Further etymological parallels mentioned by Haudry (1978: 238) include the Avestan epithet of Mithra frat̰-āp- ‘water-filling’ (Yt. 10.61), a compound of the same type as Ved. bharád-vāja- ‘strength-bearing’, which implies a locatum-as-object construction of the Avestan root par- ‘fill’ as well.
3.5 Lat. cumulāre as a LA verb and the semantic development of PIE *pleh1-
As detailed in the previous sections, LA is a phenomenon that is synchronically well attested for verbs meaning ‘fill’ not only in IE languages like English and German, but also in non-IE languages like Japanese. From a diachronic perspective, however, it is still possible in some cases to tell which of the two constructions is etymologically prior and which one is instead the result of an innovation—as in the case of Engl. fill and Germ. füllen, whose locatum-as-object construction is clearly a later innovation compared to the etymological location-as-object construction (see above).
Despite reconstructing an original meaning ‘pour’ for PIE *pl̥-né-h1- (which would rather support the antiquity of the locatum-as-object construction), Haudry (1978: 234–238), as mentioned above, tends to regard the location-as-object construction (“modèle 1” in his terminology) as the more ancient one, following his theory of the “two models”, according to which this alternation would be an epiphenomenon of the on-going diachronic development from the archaic “modèle 1” (location-as-object) to the innovative “modèle 2” (locatum-as-object).
I propose instead to identify the sense ‘make much, abundant; accumulate, pile up’ occurring within the locatum-as-object construction of PIE *pl̥-né-h1- as the more ancient one, given that it matches the semantics ‘much, abundant’ of non-resultative adjectives like *pl̥h1-ú- and *pl̥h1-ró-, which are more likely to reflect the primary semantics of the root (compared to the factitive and resultative formations). The relative infrequency of the locatum-as-object construction (associated with put semantics) within all occurrences of Ved. pr̥nā́-ti, at least compared to the location-as-object construction (associated with fill semantics), may thus be explained as a dying archaism, rather than as a failed innovation.
PIE *pl̥-né-h1- may have undergone a diachronic development similar to that of Lat. cumulō, -āre, a denominative verb derived from the Latin noun cumulus ‘heap, pile, mass’, which clearly displays LA, as shown in (23). On the one hand, Lat. cumulō is attested in its primary etymological sense ‘heap up, pile up, accumulate’ (a locatum Y into a location Z) within its expected (a) locatum-as-object construction. On the other hand, the verb eventually came to occur in a (b) location-as-object construction and acquired a secondary sense ‘fill, make full’ (a location Y with a locatum Z). The latter has now become the only meaning of its French reflex combler ‘fill, make full’, as shown in example (24).
(23) Lat. cumulō, -āre as a LA verb
a. Lat. cumulō ‘heap up’ as reflex of the locatum-as-object construction
Occurrence:
tantumy […]
in mez
cumulasti
so_much.accy
part 1sg.accz
heap_up.pf.2sg
“You have heaped upon mez so muchy (of honour and wealth)” (Tac. Ann. 14.53)
Lat. cumulō construction:
Form:
[ACCy
in ACCz
cumulō]
Meaning:
“move Y to fit into Z as to accumulate it”
Latin put-class construction:
Form:
[ACCy
in ACCz
VERBput]
Meaning:
“put Y into/onto Z”
Cf. Lat. ponō, -ere ‘put’:
omniay
pone […]
in ignesz
all.acc.pl
put.imp.2sg
part fire.acc.pl
“Consign them ally to the flamesz” (Ov. R. Am. 719)
b. Lat. cumulō ‘fill’ (> Fr. combler ‘id.’) as reflex of the location-as-object construction
Occurrence:
cumulat=que
altariay
donisz
heap_up.3sg=conj
altar.acc.pl
gift.abl.pl
“and he is filling the altars with gifts” (V. A. 11.50)
Lat. cumulō construction:
Form:
[cumulō
ACCy
ABLz]15
Meaning:
“cause Y to be full of accumulated Z”
Latin fill-class construction:
Form:
[VERBfill
ACCy
ABLz]
Meaning:
“cause Y to be full of Z”
Cf. Lat. impleō, -ēre ‘fill’:
implessem=que
forosy
flammisz
fill.subj.ppf.1sg=conj
deck.acc.pl
flame.abl.pl
“I should have filled the decks with flame” (V. A. 4.605)
(24)
il
la
combla
de
meubles
3sg
obj
fill.pst.3sg
with
furniture_item.pl
“he filled it with furniture” (Huysmans, Les Sœurs Vatard, 1879, p. 241; from TLFi, s.v. combler)
It must be noted that Lat. cumulō behaves just like its counterparts in English and Japanese, all of which are clearly LA verbs, namely Engl. heap, pile, and stack (Iwata 2008: 12 & 33–34) and Jap. yamamori-ni suru ‘heap up’ and yamazumi-ni suru ‘pile up’ (2008: 185–186).
This parallel is highly relevant to the present study: if the primary meaning of the PIE root *pleh1- must be reconstructed as ‘be much, abundant’, the factitive verb *pl̥-né-h1- would originally have meant ‘make much, abundant’, i.e., in a more concrete sense ‘accumulate, heap up, pile up’ (a locatum Y in a location Z), and its etymologically expected construction would thus have been the locatum-as-object one. Just like Lat. cumulō and its English counterpart heap, PIE *pl̥-né-h1- may have later come to be used within a location-as-object construction and developed a secondary sense ‘make full, fill’ (a location Y with a locatum Z).
The latter may have become the verb’s most prominent sense already in PIE (given that it is already so in Vedic), possibly spreading to other reflexes of the root *pleh1- and becoming the meaning that is best attested in the daughter languages: the same happened for Fr. combler, which exclusively means ‘fill’ and not ‘accumulate’, even though the latter was the etymological meaning of its Latin source cumulō. The verbal origin of the ‘full, fill’ semantics of PIE *pleh1- and their link to the phenomenon of LA is still signaled by the fact that it is mainly attested by reflexes of the factitive verb and of the resultative (quasi-participial) adjective, whereas the non-resultative adjectives seem to mainly attest the original meaning ‘much, abundant’.
4 Conclusion: Locative alternation verbs in Proto-Indo-European and the lexical-constructional approach to historical morpho-syntax and semantics
The results of this study may be summarized as in the following points:
-
A historical-comparative analysis of the morphology, syntax, and semantics of various IE verbal formations allows for the reconstruction of the phenomenon of locative alternation in Proto-Indo-European, and for its identification as the primary cause for the apparent polysemy of (at least some) specific PIE roots (*leu̯g-, *u̯el-, *pleh1-).
-
Locative alternation with PIE verbs of the cover semantic class was the main cause behind the polysemy attested by reflexes of the PIE root *leu̯g- ‘be bent, twisted’: more precisely, by the Celtic and Germanic reflexes of the same PIE nasal-infix factitive *lu-né-g-/lu-n-g-´ ‘make bent, twisted, not straight’, namely PCelt. *lung-o/e- ‘connect, put together’ and PGmc. *lūk-a- ‘(en)close’.
-
PIE *lu-né-g-/lu-n-g-´ ‘make bent, twisted, not straight’, just like its counterparts Engl. wind and Jap. maku ‘id.’ (three-place verbs with similar semantics), occurred as a three-place verb within (at least) two different constructions: in a locatum-as-object construction [*lu-né-g- ACCy LOC/ACCz], meaning “put Y into/onto Z by bending/twisting it” and reflecting the PIE construction of put-class verbs (e.g., *dhé-dheh1- ‘put’); and in a location-as-object construction [*lu-né-g- ACCy INSTRz], meaning “make Y covered by bending/twisting Z around it” and reflecting the PIE construction of cover-class verbs (e.g., *tég-o/e- ‘to cover’).
-
The locatum-as-object construction developed into the main construction of PCelt. *lung-o/e- ‘to connect, put together’ (a verb with put semantics, as expected), whereas the location-as-object construction became the main construction of PGmc. *lūk-a- ‘(en)close’ (a verb with cover semantics, as expected).
-
A similar development may also have taken place in the history of the polysemous PIE root *u̯el-, whose divergent meanings ‘turn, wind’ vs. ‘enclose, cover’ may be traced back originally to two distinct constructions of its verbal formations: a locatum-as-object construction [VERB (*u̯el-) ACCy LOC/ACCz], meaning “put Y into/onto Z by rolling it round, coiling it” (i.e., with put semantics), and a location-as-object construction [VERB (*u̯el-) ACCy INSTRz], meaning “make Y covered by rolling round, coiling Z around it” (i.e., with cover semantics), respectively.
-
-
Locative alternation with PIE verbs of the fill semantic class was the cause behind the polysemy attested by the IE reflexes of the PIE root *pleh1- ‘be much, abundant; be full, filled’.
-
The primary semantics of PIE *pleh1- must have been ‘much, abundant’, being better attested by non-resultative formations like PIE *pl̥h1-ú- ‘id.’ (e.g., Ved. purú- ‘id.’), whereas the ‘full, filled’ semantics rather seems to be mainly attested by resultative or factitive formations like PIE *pl̥-né-h1-/pl̥-n-h1-´ ‘make full’ (Ved. pr̥nā́-ti ‘id.’).
-
PIE *pl̥-né-h1-/pl̥-n-h1-´ was a LA verb that meant both ‘to make much, abundant (i.e., accumulate)’, when used within a locatum-as-object construction, and ‘to make full (i.e., fill)’, when used within a location-as-object construction, an alternation exactly paralleled by its Japanese counterpart mitasu (and with a looser parallel in Engl. fill and Germ. füllen as well). This reconstruction is grounded in the attestation of its Vedic reflex pr̥nā́-ti both within a locatum-as-object construction [pr̥nā́-ti ACCy DATz], meaning “move a great quantity of Y for/onto Z” and also occurring with put verbs (like Ved. dá-dhā-ti), and within a location-as-object construction [pr̥nā́-ti ACCy INSTRz], meaning “cause Y to be full of Z” and also occurring with fill verbs (like Ved. pr̥-ná-k-ti).
-
The development reconstructed here for PIE *pl̥-né-h1-/pl̥-n-h1-´ may have resembled that of the Latin LA verb cumulō ‘heap, pile, accumulate; fill’, the source of Fr. combler ‘fill’. Even though the Latin verb’s etymology (derived from the noun cumulus ‘heap’) clearly allows us to identify the sense ‘to accumulate’ occurring within the locatum-as-object construction as the more archaic one, the secondary sense ‘fill’ that developed within the location-as-object construction eventually came to be more prominent, and even became the only meaning of the verb’s French reflex combler.
-
-
The three case studies presented in this paper support the reconstruction of LA for PIE morpho-syntax. It must be stressed that LA is a widespread synchronic phenomenon in a number of languages and it is thus unnecessary to explain all such alternations as manifestations of a diachronic development from one morphosyntactic pattern to another (as proposed instead by Haudry 1978). By taking into account further aspects of PIE grammar and lexicon (e.g., the Caland system), it has been possible to identify the locatum-as-object construction as the older pattern for at least two of the polysemous roots discussed here (PIE *leu̯g- and *pleh1-), but this was not necessarily the case for all LA verbs in PIE.
From a methodological perspective, I hope to have corroborated the (by no means novel) idea that an integrated approach combining the historical morpho-syntax and the historical semantics of the IE languages may provide us with valuable insight on both. The etymological reconstruction of the PIE lexicon and the historical semantics of the IE languages will benefit greatly from approaches that take into account the interplay between, on the one hand, the morpho-syntax and phraseology within which IE lexemes are historically attested and, on the other hand, the meanings that these terms evoke in the respective contexts. Rather than exclusively dealing with lexical meanings extracted from dictionary entries, the semantic side of etymology should involve the study of the constructions within which words occur, and their analysis both at the lexeme-specific level and at the more general and abstract level of semantic classes.
A final consideration is in order. The term Proto-Indo-European has been used here to refer to the reconstructed ancestor of the Indo-European language family, including the languages of the Anatolian branch (such as Hittite and Luwian). As the reader may have noticed, however, the three reconstructed roots on which the present study focuses (*leu̯g-, *u̯el-, *pleh1-) seem to lack obvious verbal reflexes in Anatolian (at least according to LIV2). This very limited dataset makes it impossible to speculate if the locative alternation phenomena reconstructed here arose only after the branching of Anatolian, i.e., in the prehistoric ancestor of the so-called Core Indo-European languages (non-Anatolian Indo-European languages). Providing a definitive answer to this question will require a systematic investigation of all these languages, which lies beyond the scope of the present study; future research along the lines of the methodology proposed here may attempt to address this point, however, and to clarify if the phenomenon of locative alternation should indeed be reconstructed already for Proto-Indo-European, or rather for Core Indo-European.
Acknowledgments
For encouragement, criticism and suggestions, I am grateful to Erica Biagetti, Andrea Lorenzo Covini, Chiara Fedriani, José Luis García Ramón, and Daniel Kölligan, as well as to the anonymous reviewers.
In LIV2, the Germanic non-prefixed strong verbs are thus said to be secondary and to occur only in OE, but this is not accurate, cf. Seebold (1970: 338).
Gk. lug-íz-omai/-ō ‘turn, wind, bend (something [act.] or oneself [med.])’ (first attested in Hippocrates) likely reflects a denominative verb (with causative meaning ‘to make bent, twisted’) derived from the adjective *lug-ó- ‘bent, twisted’; in the case of a derivation from substantivized Hom. Gk. lúgos ‘flexible twig’, the semantic path would be much less clear (cf. LSJ, s.v.). Gk. lugóō ‘tie fast’, first attested in Hellenistic poetry (AP 9.150.5), may be derived from either the adjective or the substantive. Gk. lugaîon* ‘bracelet’, attested in the Hesychian gloss lúgaia: tà perì taîs khersì pséllia ‘the armlets around the hands’, may reflect the substantivization of an adjective *lugaîos ‘of the bending, twisting’, in turn derived from an abstract *lugā́ ‘bending, twisting’, a further possible reflex of PIE *lug-ó-.
As suggested to me by Daniel Kölligan (p.c.), the Greek adjective may alternatively be traced back to *(s)lug-ró- ‘sobbing, mournful’, a derivative of PIE *(s)leu̯g- ‘swallow; sob’ (LIV2: 567–568), a root which may also underlie Lat. lūgeō, -ēre ‘mourn, grieve’ (Kölligan 2005).
A formation of the so-called russus-type; cf. Höfler (2015: 35–36).
For reasons of space, in the present paper glosses of nominals only specify the plural or dual number (singular is not indicated), whereas gender is ignored; glosses of verbs do not mark the indicative mood, the present tense, and the active voice (other moods, tenses, and voices are specified).
This connection finds further support in a Hesychian gloss in which lugízetai (3sg.mid of the denominative verb lug-íz-ō) is glossed as sundédetai ‘has been bound together’ (pf.3sg.mid of the denominative verb sundé-ō ‘bind together’).
On the Greek dative case as a reflex of the PIE locative case, see Luraghi (2003: 51–52).
On the Old Norse dative case as a reflex of the PIE instrumental case, cf., e.g., Faarlund (2004: 20, 142, & passim).
On the Greek dative case as a reflex of the PIE instrumental case, see Luraghi 2003: 51–52.
On the quite divergent semantics of the Latin and Old Irish formations, cf. LIV2: 48311–12.
Even though in more vague terms, the possibility of a similar connection was evoked en passant by de Boel (2017: 673): “After all, the continuity we see here between a meaning which presupposes a container and a meaning which does not finds an exact parallel in the usage present day Romance and Germanic languages make of words meaning ‘full’. The etymological link of these adjectives with verbs meaning ‘to fill’ is obvious. But this link does not prevent a word like full to apply, e.g. to ‘a table full of good food’. In present day spoken French, the word plein is even a synonym of beaucoup ‘much’, in a way which recalls the Gk. polús, which is related to pímplēmi, much as Dutch veel, Germ. viel ‘many’ are related to vullen, füllen ‘fill’, and vol(l) ‘full’ ”.
On PGmc. *full-a- and its factitive derivative *full-ija-, see Heidermanns (1993: 220) and Ringe (2017: 283).
On which see Luraghi (2003: 39; 51–52; 33737).
On the alternation between the partitive genitive and the instrumental dative in Ancient Greek, see Luraghi (2003: 60–61).
On the Latin ablative case as a reflex of the PIE instrumental case, cf. Weiss (2009: 202 & 213).
References
Barðdal, Jóhanna, & Thórhallur Eythórsson. 2012a. Reconstructing syntax. Construction Grammar and the Comparative Method. In Sign-based Construction Grammar, ed. Hans C. Boas & Ivan A. Sag, 257–308. Stanford: CSLI Publications.
Covini, Andrea Lorenzo. 2017. Formazioni causative nelle lingue indoeuropee di più antica attestazione. Ph.D. dissertation, Università per Stranieri di Siena and University of Cologne.
Dahl, Eystein, & Chiara Fedriani. 2012. The argument structure of experience. Experiential construction in early Vedic, Homeric Greek and Old Latin. Transactions of the Philological Society 110: 342–362.
De Boel, Gunnar. 2017. Locative alternation as lexical derivation. The examples of νάττω and βάλλω. In Ancient Greek linguistics, ed Felicia Logozzo & Paolo Poccetti, 663–678. Berlin: De Gruyter.
Dixon, Robert M.W. 1982. Where have all the adjectives gone? And other essays in semantics and syntax. Berlin: Mouton.
Dybo, Vladimir A. 2002. Balto-Slavic accentology and Winter’s Law. Studia linguarum 3: 295–515.
Faarlund, Jan T. 2004. The syntax of Old Norse. Oxford: Oxford University Press.
Fedriani, Chiara. 2014. Experiential constructions in Latin. Leiden: Brill.
García Ramón, José L. 1999. Zur Bedeutung indogermanischer Wurzeln: *h2nek̂- ‘erreichen, reichen bis’, *h1nek̂- ‘erhalten, (weg)nehmen’. In Gering und doch von Herzen. 25 indogermanistische Beiträge Bernhard Forssman zum 65. Geburtstag, ed. Jürgen Habisreitinger, Robert Plath, & Sabine Ziegler, 47–80. Wiesbaden: Reichert.
Goldberg, Adele E. 1995. Constructions. A Construction Grammar approach to argument structure. Chicago: University of Chicago Press.
Goldberg, Adele E. 2006. Constructions at work. The nature of generalization in language. Oxford: Oxford University Press.
Grassmann, Hermann. 1873. Wörterbuch zum Rig-Veda. Leipzig: F.A. Brockhaus.
Haudry, Jean. 1978. L’emploi des cas en védique. Introduction a l’étude des cas en indoeuropéen. Lyon: L’Hermès.
Heidermanns, Frank. 1993. Etymologisches Wörterbuch der germanischen Primäradjektive. Berlin: De Gruyter.
Höfler, Stefan. 2015. Dark matter: The root *√k̑u̯el- ‘dark, black’. Indo-European linguistics 3: 24–41.
Iwata, Seizi. 2008. Locative alternation. Amsterdam: John Benjamins.
Kölligan, Daniel. 2005. Lat. lūgēre ‘trauern’. International journal of diachronic linguistics and linguistic reconstruction 2: 169–175.
Kölligan, Daniel. 2007. Suppletion und Defektivität im griechischen Verbum. Bremen: Hempen.
Kuiper, Franciscus B.J. 1938. Indoiranica. Acta orientalia 16: 295–326.
Kulikov, Leonid. 2017. The syntax of Indic. In Handbook of comparative and historical Indo-European linguistics, ed. Jared Klein, Brian Joseph, & Matthias Fritz, i 377–409. Berlin: De Gruyter.
LIV2 = Helmut Rix et al. 2001. Lexikon der indogermanischen Verben. Wiesbaden: Reichert.
Luraghi, Silvia. 2003. On the meaning of cases and prepositions. Amsterdam: Benjamins.
Luraghi, Silvia. 2021. Experiential verbs in Homeric Greek. A constructional approach. Boston: Brill.
Lühr, Rosemarie. 1988. Expressivität und Lautgesetz im Germanischen. Heidelberg: Winter.
Mottausch, Karl-Heinz. 2013. Untersuchungen zur Vorgeschichte des germanischen starken Verbs. Hamburg: Kovač.
Nussbaum, Alan J. 1976. ‘Caland’s Law’ and the Caland system. Ph.D. dissertation, Harvard University.
Pinker, Steven. 1989. Learnability and cognition. The acquisition of argument structure. Cambridge, MA: MIT Press.
Rasmussen, Jens E. 1992. Winter’s Law of Balto-Slavic lengthening—an unnatural fact? Copenhagen working papers in linguistics 2: 63–77.
Rau, Jeremy. 2009. Indo-European nominal morphology. The decads and the Caland system. Innsbruck: Institut für Sprachwissenschaft der Universität Innsbruck.
Rau, Jeremy. 2013. Notes on state-oriented verbal roots, the Caland system, and primary verb morphology in Indo-Iranian and Indo-European. In Multi nominis grammaticus. A Festschrift for Alan J. Nussbaum, ed. Adam I. Cooper, Jeremy Rau, & Michael Weiss, 255–273. Ann Arbor: Beech Stave Press.
Ringe, Don. 2017. A linguistic history of English I, From Proto-Indo-European to Proto-Germanic2. Oxford: Oxford University Press.
Sówka-Pietraszewska, Katarzyna. 2019. The locative alternation with spray/load verbs in Old English. In: Historical linguistics 2015. Selected papers from the 22nd International Conference on Historical Linguistics, Naples, 27–31 July 2015, ed. Michela Cennamo & Claudia Fabrizio, 446–457. Amsterdam: John Benjamins.
