I am of the opinion that every generation or so we should reexamine the evidential support for phenomena that have come to be accepted, indeed cherished, items in the Indo-Europeanist repertoire—items about which often the only questions now raised are their prehistory and their formal analysis, with the same examples repeated in publication after publication, their accuracy and correct philological analysis assumed. One of these phenomena is the so-called ‘dāti-vāra type’ in Vedic—that is, verbal governing compounds with a first member ending in -ti- supposedly governing the second, as in the standard gloss for dā́ti-vāra- ‘granting wishes’. Although the identification of this type goes back well over a century, it remains a very live item in the perennially popular debate about Indo-European compounding—as witnessed by the posthumous publication of a 1988 handout of Jochem Schindler’s, edited and annotated by Martin Peters, in the 2022 Festschrift for Mark Hale (Schindler & Peters 2022). I will show that almost all of the supposed examples of this type in Vedic not only allow, but require, different interpretations, and the last one standing, dā́ti-vāra- itself, is also seriously tainted and even more mysterious than before. I have been anticipated in some of this by Benedicte Nielsen Whitehead in her 2012 Leiden University dissertation (40–48), summarized by Olga Tribulato in her 2015 book, Ancient Greek verb-initial compounds (177–179). Nonetheless I think I can offer a richer and often more accurate philological account and rather different analyses of the material. I would also like to emphasize at the outset that I am strictly staying away from both comparative and diachronic questions: I will have nothing to say about the type of Gk. βωτι-άνειρα ‘men-feeding,’ nor will I directly confront the issue of the prehistory of the compound type—if, indeed, it existed in prehistory. I’ll just muck around in my own little patch, the text of the Rig Veda, which has been curiously neglected in recent Indo-European treatments of this supposed compound type.¹

The dossier of examples in more or less its current form was, to my knowledge, first assembled by Wackernagel in the volume on nominal compounds of the Altindische Grammatik (Wackernagel 1905). Though his account occupies only about a page (320–321), he set the parameters and identified most of the class members for all subsequent discussions. His list consists of six items. I give them in his order with his glosses, as well as the number of Rigvedic attestations. (No new examples appear after the Rig Veda.²)

• dā́ti-vāra- (3×) ‘Schätze gebend’

• púṣṭi-gu- (1×) NP, eig. ‘Kühe aufziehend’

• rīty-ā̀p- (2×) ‘Wasser strömend lassend’

• vītí-rādhas- (1×) ‘die Spende geniessend’

• vītí-hotra- (5×) ‘das Opfer geniessend’:

• vṛṣṭí-dyu-³ (2×) ‘den Himmel regnen lassend’

The only regular additions⁴ to this list in later discussions are śrúṣṭi-gu- (1×), rhyming with púṣṭi-gu- and found adjacent to it, and, less commonly, the PN vṛṣṭi-hávya-, which Wackernagel excluded on the grounds of its second-member accent (1905: 320), but Debrunner (1957: 88–89 [Nachtr. to Wackernagel 1905: 321 & 12]) adduces, following Renou (1953: 232 with n. 10; see also Schindler & Peters 2022: 333). Debrunner’s glosses for these two additional examples are ‘die Kühe gehorchend machen’ and ‘die Opfergabe regnend lassen’, respectively. The same examples (Wackernagel’s plus Debrunner’s), with the same glosses, are continually repeated in the scholarly literature (e.g., Pinault 2018: 332–335).

To support the transitive interpretation he imposes on the first members, Wackernagel cites parallel Rigvedic syntagms with transitive verbal forms to the appropriate roots and accusative objects, for most of them—but, notably, not for dā́ti-vāra-. In order these are, with my translations:

• púṣṭi-gu-: III.45.3 krátuṃ puṣyasi gā́ iva ‘you foster your resolve like cows’

• rīty-ā̀p-: VIII.7.28 =X.138.10 (etc.) riṇánn apáḥ ‘They let flow the waters’

• vītí-hotra-: IV.48.1 vihí hótrāḥ ‘pursue the oblations’

• vṛṣṭí-dyu-: V.63.3, 6; IX.96.3 dyā́m varṣaya- ‘make heaven rain’

These transitive syntagms are simply requoted in later discussions, including Schindler’s 1988 handout (Schindler & Peters 2022: 332–334). First note that there is no stable relationship between the transitive verb stem and the -ti- compound member—a ya-present, a 9th-class present, a root present, and even a causative. Further, I must now reluctantly accuse Wackernagel of cherry-picking evidence—indeed, worse, of suppressing evidence. For intransitive syntagms involving the same elements are also found.

• púṣṭi-gu-: X.19.3 asmín puṣyantu gópatau ‘Let (the cows) [= the dedicands of the hymn] prosper under this herdsman.’

• rīty-ā̀p-: no *ā́po rīyante, but cf. X.40.9 rīyante … síndhavaḥ ‘the rivers flow’

• vṛṣṭí-dyu-: no *dyaur varṣati, but cf. V.84.3 divó várṣanti vṛṣṭáyaḥ ‘The rains from/of heaven rain.’

At the very least, Wackernagel should have acknowledged these competing syntagms: it would have saved us all a lot of trouble by generating some skepticism about the transitive analysis. As it is, subsequent scholarship seems to have accepted, on the strength of Wackernagel’s prestige, his implicit assertion that a transitive interpretation is the only one supported by the data, without checking the text independently.⁵

The interpretational problem has become acute as the morphological analysis of the first member has shifted. For Wackernagel (1905: 320–321) the first member must be an old imperative in -ti. Where exactly we find these old -ti imperatives he does not tell us (simply citing Brugmann on this point, but without examples), but such an analysis would easily account for the transitive verb–object function he ascribes to this compound type. Similarly some of the alternative views summarized by Wackernagel—that it is an imperatively used infinitive (Brugmann) or a 3. sg. finite form (Jacobi)—facilitate a transitive interpretation. However, the pendulum has definitely swung: the standard view today is that the first member is a nominal form, a -ti-abstract⁶—a view that Wackernagel himself explicitly rejects. He does so on eminently reasonable grounds: the function of such nouns as nomina actionis, which makes it difficult to derive additively the transitive sense he ascribes to this compound class. And this is the dilemma that all subsequent attempts to account for these compounds have struggled with: what makes sense morphologically (a nominal -ti-stem first member) makes it difficult to motivate the supposed transitive sense of these compounds; what makes sense functionally (a verbal first member, imperative vel sim.) cannot be justified morphologically. Choosing the path of morphological austerity has led to the fantastically convoluted, multi-step functional derivations associated especially with Schindler, as embodied in Schindler & Peters (2022). But this impasse can be resolved, and rather easily, by asking a simple question: do these compounds really mean what they’re said to mean?

Answering this merely requires revisiting the data in context, and in fact on various grounds we can eliminate almost all these forms. If we accept Wackernagel’s list as expanded by Debrunner, we have eight items, but in fact several of them are twinned and require only a single explanation:

1. The rhyming hapax personal names púṣṭi-gu- and śrúṣṭi-gu- in a single pāda.

2. The two vītí- compounds, of which vītí-rādhas- is only attested once.

3. rīty-ā̀p- and vṛṣṭí-dyu-: both are technically attested twice, but both times they are attested together in the same pāda—and the pādas are essentially identical, so we’re actually dealing with hapaxes.

4. If we accept vṛṣṭi-hávya- as a member of this class despite its accent, it presumably pairs with vṛṣṭí-dyu-.

In other words, we have only four data points:

• púṣṭi-gu-/śrúṣṭi-gu-

• vītí-hotra-/vītí-rādhas-

• rīty-ā̀p-/vṛṣṭí-dyu- and, along for the ride, vṛṣṭi-hávya-

• dā́ti-vāra-

which I will treat in turn in what follows.

Before considering the examples individually, we should first note that independent -ti-stems are attested beside all the first members. This pervasive matching certainly favors the current view that the first members are -ti-stem nominals, against the older view that the first members are verb forms of some sort or other.

• -dāti-: only attested as second member of havyá-dāti- (quite common) ‘bestowal of oblations’⁷

• puṣṭí-: very common in a variety of cases and numbers

• śruṣṭí-: pretty common, especially in instr. śruṣṭī́

Note the accents of these two, which differ from that of their respective compounds púṣṭi-gu / śrúṣṭi-gu-:

• rītí-: 5×

• vrṣṭí-: very common, but has been concretized to ‘rain’; no clear examples of abstract ‘raininess, raining’ vel sim.

• vītí-: quite common, especially as dative infinitive vītáye

Now to the actual data. The easiest to dispense with are the rhyming personal names púṣṭi-gu- and śrúṣṭi-gu- found in a single pāda in a Vālakhilya hymn, VIII.51.1,⁸ in a sequence of other names referring to people among whom Indra drank soma. Since these are names, there is no contextual clue to their literal sense, and as was noted above, both transitive and intransitive syntagms (‘make cows thrive’ / ‘cows thrive’) can be cited as parallels for púṣṭi-gu-. What śrúṣṭi-gu- might mean is even less clear. One might construct ‘obeying cows,’ but this is, alas, pragmatically excluded; ‘making cows obey’ is the preferable governing interpretation, but in the absence of a robust verbal root √śruṣ this remains a hypothetical construction. (This may be why Wackernagel does not include it.) It should be noted that Schindler (Schindler & Peters 2022: 333) considers both compounds ‘stative’, with púṣṭi- and śrúṣṭi- functionally corresponding to the possessive derivatives puṣṭi-mánt- and śruṣṭi-mánt-. Although he provides no glosses, presumably by his analysis these compounds mean ‘having thriving cattle’ < ‘having cattle that have thriving’ and ‘having obedient cattle’ < ‘having cattle that have obedience’. Alternatively one could construct an interpretation with an instrumental first member: ‘having cows with their thriving / with their obedience’. Either seems preferable to a transitive, governing interpretation like Wackernagel’s.⁹

There is also an alternative possible analysis. In the standard interpretation of these compounds, the second member -gu- is a form of ‘cow’, which is found in other compounds, like bhūri-gu- ‘having many cattle’ (1×, voc.), saptá-gu- ‘having seven cattle’. But there is another -gú-, found in vanar-gú- ‘wandering in the woods’ (2×), built to the root √gā ‘go,’ as analyzed by, e.g., Debrunner (1954: 471–472) and Scarlata (1999: 103, 108). If we assume this -gu- in the two compounds at issue, the meanings would be ‘going to prosperity’, ‘going to obedience/attentive hearing’. For the former see the syntagms puṣṭím … agman (I.122.7), puṣṭíṃ yāti (I.77.5) with different synonymous roots ‘go’. True, the accent would be wrong for a pseudo-root-noun compound (expect *puṣṭi-gú-, like vanar-gú-), but the accent is wrong anyway: for the governing analysis we should have *puṣṭí-gu- to match the -ti-stem puṣṭí- (like vītí-hotra- beside vītí-). Personal name compounds seem often to undergo accent shift (as, in fact, Mayrhofer [2003: 59] claims for púṣṭi-gu- itself); see vṛṣṭi-hávya-, if this form belongs here, as well as the underlying bahuvrīhi vadhry-aśvá- PN < ‘having gelded horses,’ beside vádhri-vāc- ‘having gelded speech’. I don’t particularly champion this alternative analysis, but it adds one more question mark to these two hapaxes, especially since, as names, nothing in their context favors one semantic interpretation over another. I think it can be generally agreed that púṣṭi-gu- and śrúṣṭi-gu-, hapax personal names with unclear semantics in a late and poorly transmitted¹⁰ part of the Rig Veda, cannot be used as evidence for the function of this putative compound type.

The next pair to be considered are rīty-ā̀p- and vṛṣṭí-dyu-, supposedly meaning ‘making waters flow’ and ‘making heaven rain’, respectively. Recall that Wackernagel cites transitive syntagms for the governing meaning(s), but that I also cited matching intransitive syntagms. Recall also that the independent stem vṛṣṭí- has only the concrete sense ‘rain’, not abstract ‘raining, raininess’. Thus, if vṛṣṭí-dyu- really means ‘causing heaven to rain’, the -ti-stem in the compound will have to have preserved the abstract verbal action-noun sense only there, not in the abundant attestations of the stem elsewhere, which only refer to the concrete substance (rainwater, raindrops) and could not have a verbal relationship, whether intransitive (*‘heaven raindrops’) or transitive (*‘causes heaven to raindrop’), with the second member—thus adding yet another distancing step to the analysis. Recall even further that these two stems, each found twice in the Rig Veda, are only found together, adjacent to each other in almost identical pādas: V.68.5 vṛṣṭídyāvā rītyā̀pā [dual, of Mitra and Varuṇa], IX.106.9 vṛṣṭídyāvo rītyāpaḥ [voc. pl., of soma drops], so that the two occurrences are really only one. Whatever these compounds are doing, they’re doing it together. And there’s more. Another coincidence of pairing: as noted above, the -tí-stem rītí- occurs independently 5×; two of those occurrences are construed with the gen. apā́m, with the same elements as our compound, and those two occurrences are found in the same pāda as another, more common, genitival syntagm, divó vṛṣṭí-: VI.13.1 divó vṛṣṭíḥ … rītír apā́m; IX.108.10 vṛṣṭíṃ diváḥ … rītím apā́m both meaning ‘the rain of heaven, the streaming of waters’. This to me is the smoking gun. The two compounds appear together; the two syntagms appear together. And the independent phrases are not transitive, as the compounds are supposed to be—they mean ‘the rain of heaven … the flowing of waters’—not *‘the causing of heaven to rain, the causing of waters to flow’. In VI.13.1 these are good things that Agni arranges for us to have; in IX.108.10 (very near the compounds in IX.106.9, notice) they either are the good liquids that Soma purifies himself into or that he attracts through his purification. The two compounds and the two syntagms make a formulaic magic square:

vṛṣṭí- diváḥ	rītí- apā́m
vṛṣṭí-dyu-	rīty-ā̀p-

The two pairs have to be interpreted together because they pattern together, and this to me excludes a transitive / governing interpretation of those compounds.

What then do they mean and how should their first members be construed? Here a scenario rather like Schindler’s for púṣṭi-gu-—a ‘stative’ double-possessive bahuvrīhi—comes to mind, perhaps mediated through an instrumental first member: ‘having waters that have flowing’ → ‘having waters with their flowing’; ‘having heaven that has rain’ → ‘having heaven with its rain’.¹¹ In the case of Mitra and Varuṇa the latter would be pretty literal: as gods they do in some sense possess heaven; for the soma drops, metaphorical, but soma has cosmic properties in the Rig Veda. For a possible mediating instrumental, see V.53.5 vṛṣṭī́ dyā́vo yatī́r iva ‘like the heavens coming with their rain’.

Let us briefly treat the personal name vṛṣṭi-hávya-, which, as we saw, Wackernagel excluded from his list on the grounds of accent, but which figures in subsequent literature, including Schindler’s handout (Schindler & Peters 2022: 333). It is found only once in a late hymn (X.115.9) and as a name has no contextual semantics; the referent is the father of the poet(s) of the hymn in question. Pace Wackernagel, its accent doesn’t disqualify it from consideration, since, as we saw above, personal names can undergo accent shift. Although Schindler groups it with his ‘stative’ type (like púṣṭi-gu-)—hence, ‘having oblations that have rain’? (Schindler & Peters 2022: 333)—others have attributed transitive value to it: Mayrhofer (2003: 88) ‘*die Opfertränke regnen lassend’ or ‘*dessen Opferguss Regen bewirkt’, Scarlata (1999: 526) ‘dessen Opferguss Regen hat/bewirkt’. Once again I would point out that vṛṣṭí- is only concrete, so that an interpretation based on verbal transitivity has to assume a usage of vṛṣṭí- that predates this functional shift. The simplest interpretation, and the one that best conforms to the Rigvedic usage of vṛṣṭí-, is to take the compound as an equational metaphorical bahuvrīhi: ‘whose X is (like) Y’—hence ‘whose oblations are (like) rain’. In the IXth Maṇḍala soma is regularly described as flowing like the rain of/from heaven: e.g., IX.89.1 divó ná vṛṣṭíḥ pávamāno akṣāḥ ‘Like the rain from heaven, the self-purifying (soma) has flowed’. The image is of gushing liquid abundance.

We now turn to the last set of paired items on our list: vītí-hotra- (5×) and vītí-rādhas- (1×). One of the issues relevant to these compounds is that the meaning of the root √vī has been reassessed since the compound type was identified, and so the compounds and the root from which they’re derived have been glossed in numerous, not-very-principled ways. For the now-standard view of the root’s semantics, namely ‘pursue, follow’, see EWAia s.v. vayⁱ with further literature. Older assessments of the root’s meaning often contain the notion ‘enjoy’; Grassmann (1872–1875) glosses vītí- as ‘Geniessen / Genuss’ (enjoyment) and concretely ‘Opfermahl’. Wackernagel (see above) glosses the compound vītí-hotra- as ‘das Opfer geniessend’, with vītí- as transitive ‘enjoying’.¹² However, re-evaluation of the uses of the root √vī and a stricter imposition of semantic boundaries have unified the various uses of √vī as ‘pursue, follow’, with the further implication that what one pursues is something one desires. The -tí-stem vītí- ‘pursuit, pursuing’ is quite common, especially in the dat. vītáye, which as a purpose infinitive means ‘for pursuing, to pursue’, often with an accusative object. As noted above, Wackernagel cites a transitive syntagm vihí hótrāḥ (IV.48.1) as support for his compound gloss ‘das Opfer geniessend’. I freely admit that both finite and non-finite forms of √vī, especially including the infinitive vītáye, take synonyms of hótrā-¹³ as object, especially havyá- ‘oblation’: see hávyāni (…) vītáye (I.74.6, 135.3–4, 142.13; II.2.6; VIII.20.10, 16, 101.7).

But it is absolutely crucial to note who is the subject of these forms of √vī. Both gods and mortals can serve as subject of √vī, but their objects differ. When a mortal is the subject, a god is usually the object, as in VIII.4.17 vémi tvā pūṣan ‘I pursue you, Pūṣan’. The mortals are actively seeking the gods to bring them to the sacrifice. But in the far more numerous cases in which gods are the subjects, they are pursuing what we mortals and ritualists have to offer them, often oblations—as in III.53.1 vītáṃ havyā́ni adhvaréṣu devā ‘Pursue the oblations at (our) ceremonies, o you two gods [=Indra and Parvata].’ But—crucially—the compound that has the same apparent structure as these verb phrases, vītí-hotra-, which should mean, by the ‘governing’ hypothesis, ‘pursuing the oblations’, does not modify the gods who are doing the pursuing, but generally the mortals who are providing the objects being pursued.¹⁴ Cf. I.84.18, whose subject is the energetic ritualist:

(1) I.84.18 kó agním īṭṭe havíṣā ghṛténa, srucā́ yajatā ṛtúbhir dhruvébhiḥ

…, kó maṃsate vītíhotraḥ sudeváḥ

Who reverently invokes Agni with oblation and ghee? (Who) will perform sacrifice with a ladle according to the fixed ritual sequences?

… Who might think himself (a man) whose oblations are worth pursuing, who has gods well on his side?

In other words, the pragmatic ritual context excludes a straight transitive-type reading, ‘pursuing the oblations’, such as was assumed by Wackernagel, because that should be appropriate only to gods. The referent of the compound is not pursuing the oblations, but possesses the oblations for others to pursue. I take the first member, vītí-, as standing functionally for the dative infinitive vītáye, in an otherwise standard bahuvrīhi. In the relationship between its parts vītí-hotra- is no different from its etymological and structural close kin, vītá-havya- (3×), with a past participle as first member. In all cases this latter compound qualifies a mortal and should mean ‘whose oblations are pursued’. The difference between the -tí-stem and the -tá- stem in these two compounds may be that of ‘potential’ (to be X-ed) versus ‘achieved’ (X-ed). Unfortunately I do not know of any other pairs of this type.

I have now demonstrated that none of the compounds just discussed shows a transitive governing relationship between first member in -ti- and second member, though they do all have reliable relationships with independently attested -ti-stems. Leaving aside the two names púṣti-gu- and śrúṣṭi-gu-, whose meaning and structure cannot properly be assessed, I consider all the others just treated to be bahuvrīhis of a more or less standard sort: ‘whose Y is (with) X’ or, in the case of the vītí- compound, ‘whose Y is for Xing’.¹⁵ Crucially, however, none of these needs to undergo a second step to produce a Schindlerian factitive double bahuvrīhi. I do not see why it is necessary, or how it is even possible, to get from rīty-ā̀p- ‘having waters provided with flowing’ to ‘having waters flowing’ to ‘making waters flow’, as in the—to me, highly implausible—semantic derivation provided by Peters (Schindler & Peters 2022: 335; see also Pinault 2018: 335) for the hapax OAv. jə̄nar- (= jan-nar-): ‘having the men provided with killing’ → ‘having the men killed’ → ‘killing the men’.¹⁶ In this example I don’t see why his interpretation, especially of the middle step, would not lead to the hatá-putra- (‘having slain/dead sons’) bahuvrīhi type, with no agentive governing sense, before moving on to the verbal-governing stage. Since this is clearly not the sense, I am quite dubious about the sequence of derivational steps; I think it easier to get to ‘killing men’ directly.¹⁷ It is the tyranny of the bahuvrīhi that complicates the analysis and requires the positing of a functionally implausible derivational chain.

This leaves us with dā́ti-vāra-,¹⁸ which seems to present us with an entirely different situation. But before delving into it, I’d like to point out that what the previous discussion has done is to show that there’s no ‘dā́ti-vāra- type’; the other standard examples have been eliminated, and there is just dā̇ti-vāra-. Whatever historical and comparative conclusions one wishes to draw will have to depend solely on this form.

This compound occurs three times in the Rig Veda (I.167.8, III.51.9, V.58.2). It always occurs at the end of a Triṣṭubh line, and in all three occurrences it modifies the Maruts.¹⁹ I do not see any way out of taking it as a transitive construction ‘granting wishes, giving choice things,’²⁰ and so I concur with the usual interpretation. E.g.,

(2) V.58.2 tveṣáṃ gaṇáṃ távasaṃ khā́dihastaṃ, dhúnivratam māyínaṃ dā́tivāram

… the turbulent (Marut) flock, powerful, with bangles on their hands, of boisterous commandment, masters of artifice, granting wishes …

Like most of the other putative compound examples we examined above, it participates in a formulaic web—but, significantly, not with a -ti-stem nominal, as they all did. Recall that though -dāti- is found in the well-attested compound havyá-dāti- ‘bestowal of oblations,’ it is not attested independently. Instead dā́ti-vāra- exists parallel to a number of syntagms containing almost the same elements, but with a finite verb. The verb is 3. sg. dāti, seemingly belonging to the root aorist, but with a primary ending—making it, of course, strikingly anomalous (though see below). As object we find vā́rⁱya-, a variant of vā́ra- whose suffix allows distraction and therefore a Jagatī cadence (or a dimeter iambic cadence).

(3) VII.15.12 dítiś ca dāti vā́rⁱyam#

And Diti gives something to be desired.

(4) V.48.5 yáto bhágaḥ savitā́ dā́ti vā́rⁱyam#

… from which Bhaga (and) Savitar give something to be desired.

The phrase is also found pāda-internal:

(5) VII.42.4 sá … dāti vā́rⁱyam …

He gives something to be desired.

And instead of dāti, the imperative is once substituted—but it’s entirely parasitic on dāti, which follows in the next verse (see VII.15.12 cited above).

(6) VII.15.11 bhágaś ca dātu vā́rⁱyam#

And let Bhaga give something to be desired.

We also find dāti with semantically similar objects:²¹

(7) VI.24.2 vājī́ … dā́ti vā́jam#

The prize-winner gives a prize.

Once non-adjacent:

(8) IV.8.3 dā́ti priyā́ṇi cid vásu

He gives especially dear goods.

All of these have Verb Object order and most are adjacent and pāda-final, just like the compound dā́ti-vāra-. Outside of this little clutch of passages, vā́ra- / vā́rya- are seldom objects of √dā,²² so this phenomenon is not a byproduct of a larger formulaic system, but localized in the dā́ti-vāra- realm.

I find these data baffling—and I also find it baffling that with all the ink spilled on the compound dā́ti-vāra- these syntagms barely figure. Among other things these are the only five occurrences of the anomalous finite dā́ti with primary ending, and the fact that it’s only attested in this formulaic sphere, closely tied with the anomalous compound dā́ti-vāra-, should give us pause.

But to keep our focus on the compound, one thing we can say for certain is that synchronically, to Vedic poets, dā́ti-vāra- contains a finite verb, not a -ti-stem noun, and in this it is quite distinct from the other forms supposedly belonging to this ‘type’, whose formulaic affinities to -ti-stems were explored above. I would like to emphasize the word ‘synchronically’. As for their history, two diametrically opposed diachronic scenarios present themselves to account for the relationship between dā́ti-vāra- and dāti vā́rⁱyam (etc.)—scenarios that are almost equally balanced—and therefore almost equally flawed—in a classic chicken-or-egg dilemma. On the one hand, the compound dā́ti-vāra-, the lone example of the precious PIE verbal-governing type in *-ti- to be found in Vedic, was inherited, but, quite possibly because of its isolation, lost any transparency it once had. It was then reanalyzed as containing a finite verb, and a limited set of formulaic verb phrases were then generated from this misanalysis: compound → verb phrase.

Alternatively, the finite verb dāti in the formulaic phrase dāti vā́ram could have lost its transparency, since it should belong to a root aorist stem, where indicative forms should not have primary endings. The best way to account for such a finite form is as a root aorist subjunctive (*dā-a-ti).²³ There is no positive evidence for this analysis unfortunately; a distracted root syllable would be highly desirable but is not found. However, distraction is not a universal requirement in these forms, and in fact all of the finite occurrences of dā́ti / dāti and dhā́ti / dhāti in the RV are compatible with a subjunctive interpretation.²⁴ Nonetheless, since non-distracted dāti appears to be a present indicative because its root vowel and mood sign coalesced, it sits uncomfortably in the well-attested root aorist of √dā; unlike other root aorist subjunctives like karati, the root syllable and the mood sign cannot be kept separate and so the morphological analysis is not clear: the default analysis would be as root present indicative 3. sg. active. According to this scenario, because of this apparent anomaly the verb phrase became univerbated to the dā́ti-vāra- nominal compound that we have: verb phrase → compound.

For parallels to this phrasal univerbation I can point to other examples of compounds based on phrases:²⁵ already in the RV we have forms like mama-sat-yá- (X.42.4) ‘[struggle for] what’s mine’ (< ‘[this] being mine’), khada-prī́- (I.38.1, VIII.7.31) / kadha-priya (voc. I.30.20) ‘when?-friend’ (= fair-weather friend), in the AV (Ś+P) aham-uttará- / aham-uttara-tvá- ‘(contest for) preeminence’ (lit. ‘I (am) higher’); verbal syntagms are found in forms like AV (Ś+P) mām-paśya- ‘look at me’ (name of a plant used in a love charm) and the name of the literary genre itihāsá- < íti ha+āsa ‘thus indeed it was; it was just so’ already AV (Ś+P) and well attested throughout the epic and Classical periods. However, these phrasal compounds generally refer to an entity external to the phrase, e.g., the literary genre Itihāsa, whereas the compound dā́ti-vāra- is semantically and functionally indistinguishable from the free syntagm.

Still, on the whole, this latter scenario, verb phrase → compound, seems somewhat more likely to me, primarily because under the compound → verb phrase scenario it is hard to envision how this archaic inherited compound made its way from PIE to Vedic: a single precious relic surviving the perilous centuries, indeed millennia, leaving no trace in Iranian (a significant absence), to fetch up three times in the Rig Veda and then die.

But this does not mean that the alternative scenario, verb phrase → compound, is without problems. First, even if the aorist subjunctive dāti isn’t a particularly favored form, what could possibly motivate the univerbation to an unprecedented type of compound—since, as I have just shown, the scattered -ti-stem compounds usually grouped with dā́ti-vāra- are functionally and formally distinct and can’t provide a model? Moreover, the syntagms involving dāti vā́rⁱyam (etc.) show that dāti was capable of remaining an independent finite verb. Why did the vā́ram forms univerbate, but not vā́rⁱyam?²⁶ As I pointed out above, those formulae are mere metrical variants, with -vāram suiting a Triṣṭubh cadence and vā́rⁱyam a Jagatī cadence. What could trigger such an extreme syntactic cleavage? I do not have answers for these serious questions (and do not want to force them), but I still weakly favor the verb phrase → compound trajectory over compound → verb phrase, simply because the time depth required to bring the isolated compound safely from PIE to Vedic seems excessive and its absence from Iranian concerning.

But, either way, the fact that dā́ti-vāra- ‘presents’ as a verb phrase precludes the production of new such compounds within Vedic on its model, with nominal -ti-stems and transitive semantics: dā́ti-vāra- stands apart.

To sum up briefly, for well over a century a small group of Vedic compounds with first members ending in -ti- have been interpreted as transitive verbal-governing compounds comparable to supposedly similar formations in other IE branches and supposedly bolstered by transitive syntagms in Vedic itself. Both the roster of examples and their interpretation have become frozen, deployed as an unexamined class in publication after publication, with the accuracy of the interpretations assumed and theoretical superstructures erected upon them. By reexamining the evidence, I have shown that all but one of the canonical Vedic examples have to be interpreted in other ways—leaving only one standing, dā́ti-vāra- ‘giving choice things’. But since this compound is closely linked to verbal syntagms with the same or similar elements, containing dāti as a finite verb with accusative object, the history and status of this compound are muddied: is it a recent univerbation of a verb phrase, or were the associated verb phrases generated from the compound? If the former is the case, Vedic has no evidence for the supposed PIE verbal governing compounds to -ti- stems. If the latter, dā́ti-vāra- may be inherited, even from PIE, but as an isolate, since, as I’ve just demonstrated, none of the other compounds that have been associated with dā́ti-vāra- since Wackernagel actually belongs to such a class. In this case those who wish to claim deep antiquity for the supposed type have only a single Vedic form, not a class, to play with.