Vedic dāśvā́ṃs- ‘pious one’, Homeric ἀδηκότες ‘inattentive’, and the “long-vowel” perfects of Proto-Indo-European


 Although the morphological components of the Vedic noun dāśvā́ṃs- are, from the Indo-European point of view, relatively transparent (root */dek̑-/ ‘perceive’, perfect participle suffix */-u̯ós-/), the exact derivation of the form is disputed, insofar as its history is bound up with an understanding of Proto-Indo-European “long-vowel preterites” (Schumacher 2005, Jasanoff 2012). This article argues that a shallow synchronic derivation of dāśvā́ṃs- in Vedic Sanskrit encounters problems in both morphology and phonology that have been overlooked by proponents of such a derivation (Jasanoff 2012, LIV2: 110–111). The article then further proposes that a cognate of dāśvā́ṃs- is to be found in the isolated Homeric adjective, ἀδηκότες, previously without certain interpretation or etymology; here the gloss ‘inattentive, oblivious, unheeding’ is proposed. The etymological connection of dāśvā́ṃs- to Homeric (ἀ-)δηκότ(-ε/ας) thus supports the reconstruction of a Proto-(Nuclear)-Indo-European (PNIE) form *[dēk̑u̯ós-]; within the grammar of PNIE itself, such a form would be synchronically derived as a perfect participle /RED-dek̑-u̯ós-/, in which a “long-vowel” form surfaces in perfect stems whose zero-grade form is phonologically dispreferred and therefore repaired (cf. Schumacher 2005, Zukoff 2014, Sandell 2015a, Sandell 2015b: Ch. 8, Zukoff 2017a: Ch. 5, 7). The larger implication is at least some “long-vowel” preterites of PNIE can be explained as phonologically driven allomorphs of perfect weak stems.

Indo-European Linguistics 6 (2018) 117-151 a Among forms built to the stem dāśa-is the 1.pl.opt daśema (11×), one occurrence of which at RV 7.3.7a falls in the cadence of the line, which ought to scan as ⏑ -, and thus, according to Tichy (1976: 80), preserves the optative to a Class II present, *dăśīma < *[dékȋh1me]. b Lubotsky (1998: 674) takes this participle to belong to the thematic Class I stem dāśa-, though the justification for this classification is unclear to me; a participle stem daśant-* would have been expected in Vedic (cf. the present active participle tákṣat-to √takṣ 'fashion'). Most likely this classification is a mistake, perhaps driven by the fact that thematic present participles (even Class I) do show suffix ablaut in Classical Sanskrit (cf. Whitney 1889Whitney [1960: 164-165).

dāśvaṃs-in Vedic
At first blush, one might reasonably be tempted to suppose that dāśvámṣ-is a formation built internal to Vedic or its shallow prehistory, to the fairly wellrepresented verbal root dāś-, the verbal forms of which are given in Table 1.
Nevertheless, a combination of phonological, morphological, syntactic, pragmatic considerations suggest that dāśvaṃs-is an archaism.
First, the form's high token frequency itself in the RV, a function of its pragmatic usefulness in ritual texts of the Vedic type,3 suggests a well-established lexeme, not a novel creation, or a form liable to undergo morphological renewal. Note that the token frequency of dāśvaṃś-is greater than the combined token frequency of all verbal forms (66) built to the same root. As Hay (2003) and Hay and Baayen (2003) have shown, lexical items that have a higher token frequency than their base of derivation are often not psychologically recognized by speakers as related morphologically related to their bases. This fact alone throws into doubt whether dāśvaṃs-has a direct synchronic connection sandell Indo-European Linguistics 6 (2018)  to the wider averbo of √dāś, despite the frequent attribution of the long vowel ā in both dāśvaṃs-and the synchronic root that underlies the verbal forms in Table 1 to the same diachronic source (see discussion at the opening of 1.2 below).4 Syntactically, dāśvaṃs-normally functions as a noun, not an adjectival modifier of another noun, unlike most productively derived perfect participles. Among its adjectival usages, in fact, evidence for a fixed expression in the RV is available: the collocation dāśúṣe mártyāya 'for the pious mortal' , occurring 11× in the R.V., is the most frequent two-word sequence involving a form of dāśvaṃs-; given the relative frequencies of the forms dāśúṣe (114×) and mártyāya (24×), a likelihood-ratio test indicates a very high degree of collostructional dependency (cf. Stewanowitsch and Gries 2003) between dāśúṣe and mártyāya (p < 0.001 on d.f. = 1), meaning that the null hypothesis that dāśúṣe and mártyāya are completely independent of one another should be rejected).5 From a morphological point of view, to treat dāśvaṃs-as a direct adjectival or nominal derivative of √dāś through the suffix -vaṃs-is challenging, since -vaṃs-is almost exclusively attested to perfect stems, and is not normally (if ever) found built directly to verbal roots.6 Besides forms of vidvaṃs-'knowing' , where the perfect stem to √vid regularly lacks reduplication, other possible Vedic examples of formations similar to dāśvaṃs-recorded in von 4 Under the approach to be argued for in this article, dāśvaṃs-and the derivatives of √dāś in Table 1 should probably be regarded as independent of one another. Another possibility, however, supported by the frequency relations in the Vedic corpus and the reading given to the forms by Jamison and Brereton 2014, may be that a root √dāś has been extracted from dāśvaṃs-itself. In either case, the important point for present purposes is that the long vowel of dāśvaṃs-should not be seen as dependent upon a PIE long-vowel present, supposedly reflected in dāṣṭi (cf. fn. 10 below). 5 Specifically, −2 log(λ) = 127. 4, given two words occurring 114 and 24 times, respectively, and occurring in the sequence Word1 Word2 11 times, in a text of 159430 tokens. The value 127.4 can be interpreted under a χ2-distribution to find the p-value. The likelihood value can also be expressed in terms of odds: we can say that the bigram dāśúṣe mártyāya is 4.2 × 1026 times more likely to be found under the hypothesis that mártyāya is more likely to follow dāśúṣe (and vice-versa) than the base rates of occurrence (i.e., token frequency) of the two forms would suggest. λ is the ratio of P(word2|word1) = P(word2|¬word1) to P(word2|word1) ≠ P(word2|¬word1), where P(X|Y) is the probability of seeing X given Y. See Manning and Schütze 1999: 172-174 for details on the calculation and interpretation of likelihood ratio tests. 6 It is, of course, normal for roots whose initial segment is a long vowel, e.g., √āp 'obtain' to show a perfect stem, including the weak stem to which the active participle is built, that is identical to the root itself. Forms exhibiting perfect inflection without any trace of reduplication to short-vowel or consonant-initial roots are unusual and rare. On the special case of perfect weak stems of the form C1eC2-, see Sandell 2015b: Ch. 8, Sandell 2017, andZukoff 2017a: Ch. 5. Indo-European Linguistics 6 (2018)  Böhtlingk and Roth (1855) are: ávarjuṣīṇām 'avoiding (?)' (a + √vr̥ j 'twist'; AVS7 .50.2), bhakṣivaṃs-'enjoying' (√bhaj 'share'?; AVŚ6.79.3),7 mīḍhvaṃs-'generous' (RV 42×), vijānivaṃs-'discerning one' (RV 10.77.1), and sāhvaṃs-(√sah 'conquer'; 10× RV).8 The only forms here perhaps really comparable to dāśvaṃsare mīḍhvaṃs-and sāhvaṃs-(the latter probably very much so); the others are variously dubious.
The gen.sg. form vijānúṣah at RV 10.77.1 is interpreted by Jamison and Brereton (2014) as belonging to the root √ jñā 'know' , as indicated by their rendering 'of a discerning one' . This stem is a hapax in the RV, where a perfect active participle to √ jñā is otherwise unattested (the middle participle jajñāná-occurs 26×); one might note that an expected X vijajñúṣaḥ would be metrically equivalent to the transmitted form. Oldenberg (1909Oldenberg ( -1912 approvingly cites a suggestion of Bartholomae that the form reflects a contamination of jānatáh and jajñúṣaḥ, to which he adds "daß nicht vijānatáḥ gesagt [wird], hängt vielleicht mit dem in [Vers] a vorangehenden -úṣo [abhraprúṣo 'showering rain from a cloud'], -uṣā [pruṣā vásu 'showering goods'] zusammen". In turn, Whitney (1905: 420) regards the reading of avarjuṣīṇam at AVŚ7.50.2 (preceded immediately by viśam) as "very suspicious", given that viśam vavarjuṣīnam occurs at RV 1. 134.6; perhaps [v] in the AVŚpassage was lost in transmission due to the immediately preceding labial nasal, which could have hindered the perception of a following labial continuant. Finally, bhaktivaṃsaḥ at AVS6 .79.3 should probably be interpreted as containing the adjectival suffix /-ván-/, not /-vaṃs-/: parallel passages at KS 5.4 and TB 3.7.57, read bhaktivāno and bhakṣivaṇaḥ respectively. It is altogether uncertain what the correct reading of this form is (Whitney (1905: 340) pronounces all variants "irregular or anomalous"), for which reason no compelling trace of an unreduplicated perfect participle can be seen here.
sāhvaṃs-, meanwhile, clearly does serve as a perfect participle synchronically to √sah, and competes with an alternative perfect active participle sāsahvaṃs-(6× RV). sāhvaṃs-exhibits the same peculiar absence of reduplication and long vowel in the root, which cannot readily be explained by derivation from a root that exclusively appears with a long vowel, as one might with dāśvaṃs-. The ultimate point at this juncture is that the possibility of deriving an adjective or noun, rather than paradigmatic perfect participle, synchronically with the suffix /-vaṃs-/ in Vedic is virtually to be excluded.
Finally, a crucial phonological fact categorically precludes the interpretation that dāśvaṃs-is a productively derived perfect participle in Vedic. As Cooper (2013, 2015 has systematically demonstrated (cf. also examples and discussion in Kümmel 2000: 38, 42-43, 50-51), consonant-initial suffixes (including the active participle in its strong form /-vaṃs-/) attached to a perfect stem regularly condition epenthesis of a vowel [-i-], just in case the syllable preceding the suffix would be superheavy (i.e., a long vowel with at least one coda consonant, or a syllable with two or more coda consonants). See the examples in 1 below.
(1) Epenthesis of Linking [-i-]  As examples 1(a) and (b) show, this linking [-i-] is attested with both the perfect participle suffix /-vaṃs-/ and the root √dāś; one cannot therefore presume that either the root or suffix involved in dāśvaṃs-might fail to condition or block the insertion of linking [-i-]. Given these facts, one may safely assert that the productive synchronic derivation of a perfect active participle to √dāś, in its strong stem, would have appeared in Vedic as X dāśivaṃs-; as seen above in Table 1 above, the reduplicated perfect active participle is attested in the RV, but only in its weak stem dadāśúṣ-.9 9 Although the strong stem dāśivaṃs-is, in fact, one time transmitted in a Vedic text, at SV 1.

1.2
Towards a prehistory of dāśvaṃs-For all of the above reasons, to assume that "dāśvam̐ s-is probably the analogical replacement (with dāś-from the present) of *dakṣvam̐ s-< *de-dk-, with the regular treatment of a "thorn" cluster", as per Jasanoff 2012: 128, fn. 5, is not readily defensible. If a form of PIE or early PIIr. antiquity *[dedkȗ̯ ós-] had indeed existed, there is good reason to think that a X dakṣvaṃs-would have persisted into Vedic. LIV2:110-111 accounts for the long ā of dāśvaṃs-through just the same assumption of reformation of the form after dāś-in the present stem: "Vielleicht Ptz. Perf. Akt. *de-dk-u̯ ós-dissimiliert zu *dek-u̯ ós-> *dać-u̯ ās-, dann umgeformt nach Präs. *dāć-> ved. dāś-zu ved. dāś-vaṃs-/-uṣ-." Again, the need to accept renewed derivation of dāśvaṃs-from the basic root is simply incompatible with the broader linguistic facts of Vedic mustered above. 10 Kümmel 2000: 243-244, like the LIV2, posits a dissimilation of */d/ with compensatory lengthening of the preceding vowel as responsible for daśvaṃs-, and situates the development in Proto-Indo-Iranian. Among earlier treatments of dāśvaṃsin the literature, Kümmel's is the most compatible with the linguistic facts of Vedic, but suffers from two arguable drawbacks: 1) the assumed consonant dissimilation is essentially ad hoc, and not grounded in more general facts about PIE or PIIr. phonology; 2) if the argument advanced in Sandell 2014 is correctnamely, that compensatory lengthening applied to PIIr. short */ă/ in PIIr. or dāśvaṃsam,dāśvaṃsam,and dāśvaṃsaṃ at RV 4.2.8d,7.37.4c,7.92.3a,8.57.4d,and 8.71.4c, respectively. The SV's repair of an apparent metrical fault by creating the form dāśivaṃ might be explained through a combination of two factors: metrical distraction of consonant + ⟨v⟩ into consonant + [uv] is significantly less common in the RV saṃhitā than distraction of consonant + ⟨y⟩ into consonant + [iy] (by my reckoning, comparing the saṃhitā text to the metrically restored text, there are~3777 instances of distracted [iy], but only~1597 instances of distracted [uv]), while the occurrence of ⟨i⟩ before the perfect participle strong stem form is reasonably common (44 out of 159 instances of the sequence ⟨vāṃs⟩ in the RV saṃhitā (= 27.7%) are ⟨ivāṃs⟩). One can explicitly contrast SV 1.2.1.1.1a with SV 4.7.2.9.2b (dāśvāṃ aśnoti mártyaḥ), which precisely copies RV 3.11.7b, where the need for an additional syllable could be, in principle, satisfied by distraction of either dāśvaṃ or mártyaḥ-or by use of the form dāśivaṃ. That dāśivaṃ does not also appear here probably lies with the frequent, and thus presumably more familiar, trisyllabic scansion of mártya-. Whether the creation of dāśivaṃ ultimately lies with the oral or manuscript tradition of the SV cannot be established, but that the form is not linguistically real seems indisputable. 10 The reason that Jasanoff (2012) and LIV2 make explicit reference to "analogical replacement" or "Umformung"-which can be seen here as merely descriptive labels that do not properly motivate the presumed changes-from the present stem is due to the reconstruction of a long-vowel present 3.sg * later results in Vedic e-, then dissimilation and compensatory lengthening applied to a *[dadu̯ as-] should have resulted in Vedic X deśvaṃs-. The Avestan past participle spara-dāšta-'served' (Yašt 13, Karde 8.1c, Yašt 19, Karde 7.10b) suggests that a root */dāć-/ had largely displaced an Indo-European lexeme */dek-/ (> PIIr. */dać-/) with short vowel.11 Any newly derived perfect participle to */dāć-/ ought then to have clearly shown reduplication, thus PIIr. *[dadāću̯ as-] > Vedic X dadāśvaṃs-.12 Such a reduplicated perfect participle stem synchronically exists in Vedic, but is functionally completely distinct from our dāśvaṃs-. Since there is no linguistic basis to set up the derivation of a non-reduplicated stem with the suffix */-u̯ as-/ in PIIr., just as such a derivation cannot be justified within the synchrony of Vedic, we fall back to the originally reduplicated perfect participles set up by Jasanoff, LIV2, and Kümmel 2000; the presumed *[dedkȗ̯ ós-], however, obviously cannot phonologically result in dāśvámṣ-without the assumption of unjustifiable morphological changes or ad hoc phonological developments.13 What is to be done?

11
OAv. dasəma-'offering; ritual ceremony' (Yašt 28.9) seems, however, to preserve the original short vowel form of the root. As a nominal derivative, this form is more likely to represent an archaism, while the productive past participle shows the synchronic base of derivation. 12 In the event that PIIr. *[dadāću̯ as-] were directly inherited into Vedic, X dadāśvaṃs-would be the result. Only when the perfect participle is synchronically generated would iepenthesis be found (giving dadāśivaṃs-*), since Vedic i-epenthesis in the perfect does not result directly from any sound change where aniṭ-roots are concerned. 13 If one believes that the strong stem form of the Vedic perfect participle itself, /-vaṃs-/, is a product of some morphological changes in the prehistory of Vedic (as argued in, e.g., Jamison 1991), then one could potentially object that the instances of the strong stem dāśvaṃs-(9× RV) in fact must reflect a newly productively derived form, and in fact, all strong stem forms of the perfect participle, when not synchronically derived, must have been newly built in the recent history of Vedic. At minimum, however, the much more frequent weak stem dāśúṣ-could readily continue a PNIE form *[dēkús-] without the presumption of any productive morphological derivation; in that case, the strong stem in Vedic might be generated by kinds of "rules of referral" (Zwicky 1985) or "paradigm structure rules" (Stump 2001), depending upon one's assumptions and tastes in morphological theory. The assumption that the strong stem of the perfect participle reflects a morphological change is itself though not necessary. I find more credible the assumption of a sound change [aː] > [ãː] / ___sa, seen also in the primary comparative suffix -yāṃs-and a few isolated lexemes like púmāṃs-'man'; see Ohala 1981 ], might conceivably represent the output of some PIE phonological process applied to an underlying form */de-dk-u̯ ós-/.14 A virtual *[dēkȗ̯ ós-] would reflect a process of consonant deletion and compensatory lengthening; how a *[deHkȗ̯ ós-] should relate to */de-dk-u̯ ós-/ is unclear.15 Indeed, there exists ample evidence across many older Indo-European languages that points towards the possibility of consonant deletion and compensatory lengthening in precisely the context that a virtual */de-dk-u̯ ós-/ would deliver. Zukoff (2017a) has systematically demonstrated that the phonologically diverse patterns of reduplication attested across Sanskrit, Greek, Anatolian, and Germanic, especially when built to roots containing initial [s] + stop or laryngeal + stop cluster, or roots lacking any sonorants (especially */TeT-/; cf. a similar proposal in Schumacher 2005) can be attributed to the effects of a phonetically grounded phonological constraint No Poorly-Cued Repetition (*PCR), as defined in 2.
14 For the sake of simplicity and transparency, I will sometimes represent PIE URs with expected ablaut grades, although some ablaut under some conditions might be phonologically derived. Similarly, the reduplicant should, following McCarthy and Prince 1995, properly be represented as /RED/ (or /RED(e)/, with a fixed segment). 15 More specifically, a less likely option as the forebear of dāśvaṃs-is *[deh1kȗ̯ ós-], where the [h1] would reflect debuccalization of an underlying /d/. The possibility that some Indo-European stops (and more specifically */d/) might be subject to lenition to *[h1] is to be inferred from the interpretation of Greek ἑκατόν '100' proposed in Kortlandt (1982): */dkmtóm/ → *[h1km̥ tóm] > *[ekaton] >> (contamination with εἱς 'one') ἑκατόν. Lubotsky (1994: 204) has suggested that dāśvaṃs-might reflect the same lenition of "d to h1". I take the evidence for a specific process of */d/ → [h1] (under whatever conditions) to be uncertain, at best; the crucial analysis of Greek ἑκατόν requires reference to contamination, in any case, to account for the rough breathing in non-psilotic dialects (cf. Chantraine 1968Chantraine -1980Chantraine [2009: s.v. ἑκατόν), for which reason a basis in contamination, or another origin altogether, for the prothetic ε-itself is difficult to exclude. Worth noting in this connection are also forms of the numeral '50' in Greek and Vedic, πεντήκοντα and pañcāśát-, which contain long vowels in the second syllables that might also be attributed to a sequence *[eh1k] ← */edk/. Following Rau (2009: 17, fn. 15) and Kümmel (2012: 302), I prefer to see the long vowel in these forms as the result of direct deletion of */d/ with concomitant compensatory lengthening: */penkwe-dk(o)mt-/. The Vedic form may be directly derived from */penkwe-dkḿt-/ → *[peŋkwēkm̥ t-] by reference to the established deletion of coronals preceding a dorsal + nasal sequence (cf. Gk. χάμαι and καίνω < *[ghm̥ h2ai̯ ], *[kn̥ i̯ ōh2] ← */dhghmh2ei̯ /, */tkni̯ oh2/, respectively, with discussion in Rau (2009) Languages may set stricter conditions (in terms of acoustic/auditory cues) for the licensing of C~Ø contrasts (i.e. the presence of C) when that C would be the second member of a transvocalic consonant repetition (i.e. C 2 α in a C 1 α VC 2 α sequence) than in other contexts. Assign a violation mark * for each C 2 α (i.e. each C~Ø contrast where C is a C 2 α ) which is not cued to the level required by the language-specific repetition licensing conditions.
Zukoff's *PCR thus serves to penalize transvocalic consonant repetitions (C α VC α sequences) in particular contexts. The *PCR constraint circumscribes the context(s) of repetition avoidance in phonetic terms: namely, repeated consonants are especially avoided when they lack robust acoustic/auditory cues to their presence (i.e., the contrast between that consonant and Ø), where the most important cue is a rising intensity contour (cf. Parker 2002, 2008, Yun 2016 between the consonant and the following segment; other relevant cues may be a [-sonorant] to [+sonorant] transition or stop-release burst.
As an example, consider the behavior in reduplication of roots in Sanskrit beginning with an [s] + stop sequence versus stop + sonorant. A root such as √druh 'be hostile' exhibits C1-copy in reduplication: the /d/ at the left edge of the root appears in the reduplicant, and thus the perfect and desiderative built to this root have a reduplicant of the form du-. The perfect strong stem dudroh-contains a C 1 α VC 2 α sequence dud-here, in which the C 2 α d following u is permitted to surface because the phonetic cues to the presence of the stop preceding the sonorant /r/ are considered adequate in the phonological grammar of Sanskrit; the sequence [dr] shows an intensity rise, sonorant transition, and stop-release burst, and thus in absolute terms is acoustically well-cued. On the other hand, a root such as √sthā 'stand' exhibits C2-copy, thus building a perfect stem tasthā-, precisely because the alternative with C1-copy, X sasthā-, would have contained a C 1 α VC 2 α sas in which the C2 s preceding the stop th would be considered inadequately cued. As a repair, the second consonant from the left edge of the root is pressed into service in the reduplicant. See Zukoff 2017a: passim for detailed formalization and application.
Particularly relevant to the problem at hand is that Zukoff takes effects of *PCR to be responsible for the emergence of perfect weak stems of the form C1ēC2-in Sanskrit (cf. also Sandell 2015b: Ch. 8, Sandell 2017) and for the preterite plural/subjunctive stem in Class IV and Class V Strong Verbs in Germanic (e.g., Goth. 3.pl nēm-un 'took' and gēb-un 'gave'). The common factor in both the Class IV-V strong verbs and the Sanskrit perfect weak stems is that Indo-European Linguistics 6 (2018) 117-151 they involve roots of the form /C1eC2-/ (especially where C1 = stop), reduplication with *e grade (~PIIr. *a grade) in the reduplicant, and zero grade of the root, thus /C1e-C1C2-/. Depending upon the precise sequence of consonants /-C1C2-/ and the language-specific phonetic cues required for the licensing of of C~Ø contrast, a *PCR violation may be incurred, and some phonological repair required. Such a repair could conceivably be the deletion, with concomitant compensatory lengthening, of a /C1/ in an unacceptable /C1C2/ cluster.16 See Zukoff 2017a: 199-205 for a more precise formal representation of such a process.
In general, the role of PCR in shaping reduplicants in Indo-Iranian is clear: besides driving reduplication with a stop in roots with [s] + stop clusters synchronically in Sanskrit, and playing a role in the creation and synchronic productivity of the C1ēC2-pattern in Sanskrit perfects, a handful of other isolated matches between Vedic and Avestan support the Indo-Iranian antiquity of C1deletion and compensatory lengthening at the Proto-Indo-Iranian level. Most compelling here is the reflex of the thematic reduplicated present /si-sd-e/o-/ 'sit' . The perhaps "expected" PIIr. form would be a *[sízdá-] > [sížda-] (by RUKI), which would be expected to yield Vedic X siḍati (with retroflex ḍ) and Avestan X hiždaiti. The actual forms, Vedic sidati and Avestan hiδαiti belie a reconstruction with ž-the segment ought to have been maintained in Avestan, and ought to have yielded a retroflex ḍ in Vedic. Faced with this problem, already Klingenschmitt (1982: 129) (followed by LIV2:513-514) assumes a sporadic dissimilation of [s … z] prior to the emergence of RUKI as a phonological process in Proto-Indo-Iranian. Such a sporadic dissimilation can capture the same facts in the languages, but is obviously ad hoc, for which reason connection to more general phonological phenomena of the languages ought to be preferred.17 Besides 16 Attention to the precise root segments is indeed important, since, as a reviewer notes, true zero-grade forms (as shown by the sequence *[uR] < PIE *[R̥ ]) are preserved in the Germanic preterite-presents *[mun-] 'will' and *[skul-] 'shall' , which have Class IV root shapes. These must be archaisms, unsurprisingly, given the isolation and functional specialization of the preterite presents. Among inherited roots of the Class IV shape */CeR-/, some probably built "long-vowel" perfect weak stems, while others admitted perfects with true zero grade; in Germanic, paradigmatically integrated Class IV forms were all eventually given a "long-vowel" preterite plural/subjunctive stem (cf. In Avestan, we we find productively built perfect forms to √sad 'sit' and √stā 'stand' , which would appear to contradict the application of a PCR-driven consonant deletion in a cluster consisting of a coronal sibilant fricative and a coronal stop: 3.sg.opt. ha-zd-iiāt̰ (Y. 65.5) and 3.pl.perf. vi-ša-sta-rə. In both cases, however, the perfect stem is precisely productively generated, and the resulting C1VC1-sequence produces consonants with two different places of articulation. A *PCR violation would not be expected in such cases.
, or any other *PCR-driven alternative to C1-copy. Here one may note a difference in Indo-Iranian between the behavior of */si-sd-a-/ and */si-stH-a-/: the former evidently became frozen as *[sida-], while the latter must have still been productively generable in Indic, given Ved. tíṣṭhati. That Vedic tíṣṭhati attests to the productive application of reduplication and the relevant phonology in Indic suggests that the same productive generation may have been possible in Iranian, whence hištaiti. Exactly what the Proto-Indo-Iranian surface realization of */si-stH-a-/ then was, is uncertain-either a *[sístHa-] (*[sištHa-] with RUKI) or a *[sita-] (PCR) is possible, provided that the underlying form */si-stH-a-/ remained recoverable.18 Regardless of the exact situation in Indo-Iranian, the equation of Lat. sistō, and Gk. ἵστημι strongly suggests that [s] + stop clusters under conditions of reduplication were not subject to repair in PIE.
Although the precise history and development of PCR-effects in reduplication between PIE and its daughters is not yet wholly clear, the foregoing discussion opens the possibility that dāśvaṃs-directly continues a PIE or early PIIr. (i.e., prior to the merger of PIE */e/ and */o/) perfect participle*[dēkȗ̯ ós-], resulting from the application of PCR-driven deletion and compensatory lengthening from virtual underlying sequence /de-dk-u̯ ós-/. As Cowgill (1957) and Schumacher (2005: 600) have previously observed, the perfect active participle sāhvaṃs-to √sah 'conquer' (< */segh-/) looks like an attractive parallel 18 Indeed, even *[tištHa-], with the same reduplication pattern of C2-copy seen in Sanskrit, cannot be totally excluded. The problem is that, in roots beginning with an [s] + stop cluster, Sanskrit attests no reduplicated forms with C1-copy that could be regarded as archaisms, and Iranian attests no reduplicated forms with C2-copy. It is not straightforward to argue that PCR-driven C2-copy is more easily an innovation of Indic, since [s]+stop forms in Iranian will not be subject to PCR effects in any case (/s/ will surface as [h] wordinitially before a vowel), thereby allowing a default C1-copy to apply. Another possibility, helpfully pointed out by a reviewer, is that the difference between sāhvaṃs-and sehāná-could be purely chronological: the active participle is an older formation and reflects the application of PCR-driven deletion in PIE or early PIIr., while the middle participle is a more recent formation in later PIIr. or Indic; the phonological constraints driving deletion in both cases would be similar, but the outcome of a compensatorily lengthened vowel (*/e/ in PIE, */a/ in PIIr.) yields different results. In either case, the similar outcome of *[dēkȗ̯ ós-] makes for an attractive parallel to *[sēghu̯ ós-].
The evidence for a stem *[sisth2e/o-] then appears problematic-if triconsonantal sequences trigger PCR violations, why is X *[sīth2e/o-] not found instead? Given that the sequence *[th2] is not directly reflected in Greek and Latin, one might posit deletion of */h2/ rather than of */s/ to escape a fatal PCR violation. Byrd (2015) has consistently shown that deletion of laryngeal segments is often a "low-cost" phonological repair in PIE (i.e., Max-H ["don't delete underlying laryngeal consonants"] is relatively low-ranked in the phonological grammar of PIE), on which basis the reconstruction of *[siste/o-] ← 19 Neither sāh-nor seh-is attested in the Family Books, for which reason one cannot securely argue that either seh-or sāh-is older. 20 As While the full phonological grammar surrounding the PCR-driven consonant deletions proposed for PIE here remains to be worked out, at this juncture the reconstruction of *[dēkȗ̯ ós-] itself may be bolstered through the discovery of a cognate form. In the following section, I will argue that the Homeric adjectival stem ἀδηκότ-is best interpreted as meaning 'oblivious, inattentive, heedless' , and likewise reflects the perfect participle *[dēkȗ̯ ós-].

Homeric ἀδηκότ-: formal connection and basic claim
The Indo-European root */dek-/ 'take in, perceive' that uncontroversially underlies dāśvaṃs-is well-represented by numerous verbal formations across the daughter languages, and attests a noteworthy diversity of formations in Greek especially (cf. LIV2:109-112). The basic meaning of the root is fundamentally concerned with mental perception (keeping watch, awaiting, mentally registering information): formations running from a root aorist (Arm. etes 'saw'), to middle participle of a root present (Gk. δέγμενος 'keeping watch'), to a causative (Gk. δοκεῖ 'seem' , Lat. docēre 'teach') establish such a semantics (cf. Tichy 1976). In Greek in particular, a robust derivational paradigm to the lemma δέχομαι 'receive, take (mentally or physically, of an object)' is well established in Homer, where a few otherwise-unattested relic formations are found (e.g., the root aorist (ἐ)δέκτο 'received').21 The adjectival stem ἀδηκότ-, of uncertain meaning, limited in Greek to the language of Epic, and without plausible etymology in the literature (see discussion in 2.2 below), bears a passing formal resemblance to */dek-/. The form 21 The Indo-European Linguistics 6 (2018) 117-151 resembles nothing so much as a perfect participle, formed in Greek with the suffix /-ót-/, built to a root of the shape ἀδη(κ)-or δη(κ)-with a prefix ἀ-; the -κcould belong either to the root or be part of the perfect stem formation. From a purely formal point of view, the following transponat reconstructions (modulo reformation of the derivational suffix) might be possible ( (2010) that the "third compensatory lengthening" (3rd CL) from loss of *[u̯ ] occurred only when the preceding consonant was [+coronal] (Weiss' notation: {n,r,l,s2,D}). Weiss' formulation is compatible with the positive evidence for the 3rd CL, but I am not aware of any negative evidence that demonstrates that the 3rd CL was necessarily restricted in this way. Moreover, for compensatory lengthening to be restricted by the place of articulation of segments involved is unknown in the typology of compensatory lengthening (cf. Kavitskaya 2002, Yun 2013. For this reason, I hold to the formulation of Rix (1976Rix ( [1992: 56), that the 3rd CL occurs simply upon loss of *[u̯ ] after any consonant. 23 The *[u̯ ] in a. and b. here is given in parentheses because the presence or absence of a *[u̯ ] following the dorsal would result in different phonological outcomes in Greek. This issue is bound up with the history of the perfect participle in Greek, and will be discussed under 2.4 below. sandell Indo-European Linguistics 6 (2018) 117-151

Homeric ἀδηκότ-: attestation, philological remarks, and existing etymologies
The stem ἀδηκότ-is attested a total of six times in Greek: it occurs four times as the nom.pl ἀδηκότες in Book 10 of the Iliad (10.98, 312, 399, 471), once in the Odyssey (12.281) as an acc.pl ἀδηκότας, and once in the Homeric Hymn to Apollo (460), again as the nom.pl ἀδηκότες. In these six attestations, the stem is, moreover, always immediately preceded by the dative καμάτῳ 'with toil/weariness' . An established fixed syntagm (or "formula") καμάτῳ ἀδηκότ-appears to be at hand, and the high degree of collostructional dependency (cf. 1.1 above) between καμάτῳ and ἀδηκότ-suggests a plausible site for the maintenance of a lexical archaism.24 This syntagm is furthermore metrically localized: except at Il. 10.312 (and the exact repetition of four lines at Il. 10.396-399 = 10.309-312), καμάτῳ ἀδηκότ-consistently begins with the second half of the second foot. In all occurrences, the -ῳ of καμάτῳ always scans long (no epic correption, which would result in a tribrach), as does the ἀ-of ἀδηκότ-(short ᾰwould produce a cretic). Although this syntagm still exhibits some flexibility-ἀδηκότitself may take different case forms (at least the nom.pl. and acc.pl.), while καμάτῳ may be conjoined with another dative (ὕπνῳ 'with sleep'; Il. 10.98, Od. 281) or modified by an adjective (αἰνῷ 'terrible, very bad'; Il. 10.312, 399)-the absence of the stem ἀδηκότ-outside of the language of Epic supports the notion that a lexical archaism preserved in a fixed expression is at hand. ἀδηκότ-is therefore unlikely to be a derivative productively generated from some base within the synchronic young Epic diction, which agrees with Danek's (1988: 85) detailed analysis of the formula ("es ist absolut unwahrscheinlich, daß der Odysseedichter die Formel für den Zusammenhang von μ 281 original geprägt hat").
The manuscript tradition of Homer presents some further philological uncertainties concerning the transmission of the form, namely, whether the α-

24
A likelihood ratio test indicates a very high degree of collostructional dependency between κάματος and ἀδηκότ-(in that specific form): κάματος occurs 16× in the Illiad and Odyssey, while ἀδηκότ-occurs 5× (all five of which are preceded by κάματος). The dependency of κάματος and ἀδηκότ-is highly significant (p < 0.001 on d.f. = 1, meaning that the null hypothesis that κάματος and ἀδηκότ-occur independently of one another should be rejected). Specifically, −2 log(λ) = 96. 04369, given two words occurring 16 and 5 times, respectively, and occurring in the sequence Word1 Word2 five times, in a corpus of 199047 words. The value 96.04369 can be interpreted under a χ2-distribution to find a p-value. In other words, we can say that the bigram καμάτο-ἀδηκότ-is 7.17 × 1020 times more likely to be found under the hypothesis that ἀδηκότ-is more likely to follow καμάτο-(and viceversa) than the base rates of occurrence of the two lexemes would suggest. See also footnote 5 above. Thiel, von der Mühll) print ἀδηκότες and ἀδηκότας, following the majority of the manuscripts from different lines of descent. The philological consensus is thus that ἁδηκότες in the text of the Iliad is erroneous.25 Arguably, the rough breathing on this form is an error belonging to the family of manuscripts from which the Venetus A descends, introduced on account of a supposed etymological connection with ἅδην 'to satiety' , which goes back to late antiquity (Apollonius the Sophist (1-2 CE), Lexicon Homericum:9, 9-10; see further below).26 In short, ἀδηκότ-, with smooth breathing, may be safely assumed. Before examining the usage of ἀδηκότ-in context, I will review the renderings of the form offered in modern lexica and translations, as well as the existing etymological proposals. By and large, the lexica and translations offer glosses that, when based on context alone, are approximate but reasonable, but when based principally on etymological considerations, are senseless in context.27 Approximative renderings appear in the translations of Murray 1924 25 A notable exception to this consensus appears in the recent edition of the Odyssey of West 2017: 237, where ἁδηκότας, with rough breathing, is printed at 12.281 (Heubeck and Hoekstra (1989: 134) also represent the form with a rough breathing in their commentary on the passage, but do not remark on this issue specifically). Surprisingly, West's apparatus criticus for the line makes no mention of this editorial choice, which is all the more puzzling, because the rough breathing in this passage has no manuscript support whatsoever, at least to judge from the apparatus in van Thiel 1991: 151, which  ('worn out') and Cerri 1996 ('stremato' = 'exhausted'), while the lexica base their glosses on etymologies and thus offer the senseless 'sated' (per Liddell et al. 1925Liddell et al. -1940, the peculiar 'weighed down' (Montanari et al. 2015), or strained 'voll Unlust' (Snell 1979).28 The most detailed consideration of the contextual semantics of ἄδηκότ-is given by Danek (1988: 84-86), who does not commit to a precise gloss, but remarks that "vom Zusammenhang her ist also kein Unterschied zwischen καμάτῳ αδηκότες und καμάτῳ δεδμημένον bemerkbar;" on this basis a gloss 'subdued' (= δεδμημένον) could be imputed to ἀδηκότ-. Murray, Cerri, and Danek are close to the mark, though a more precise reading may be possible. For the etymology, Chantraine 1968Chantraine -1980Chantraine [2009: s.v. ἀδηκότες discusses two basic possibilities, both of which are already present in Apollonius the Sophist (Lexicon Homericum).29 On the one hand, ἀδηκότ-could be regarded as a kappatic perfect participle, thus making the form relatively young, built to an otherwise unattested verb *ἀδέω 'to satiate, to fill up' , which would supposedly underlie the adverb ἅδην 'to satiety' (Epic ἄδην through psilotic East Ionic) and noun ἅδος 'satiety' .30 The form is either strangely lacking in reduplication (/e-adɛː-/ ought to have given [ɛːdɛː-] by contraction in Ionic, if it were contracted at all), or could be regarded as a form from an Aeolic dialect, in which /e-adɛː-/ would yield [aːdɛː-]. Treating ἀδηκότ-as an Aeolic form has been argued for most systematically by Peters (1988: 237-238), but there is, however, no independent reason beyond the outcome of vowel contraction to view the form as Aeolic, and as shown below, the long α-need not be etymological. The presumed development from an original meaning 'sated' to the sensible reading 'exhausted, overcome, worn out' has struck some commentawe should probably read "ἀηδήσειεν" with Wackernagel (cf. H.G. p. 25). But this does not explain the present phrase. Nor is any satisfactory sense to be got from "ἄδην" … It is simpler to say that though the meaning of the word is obvious, its affinities are unknown." 28 The glosses of ἀδηκότ-given in Liddell et al. 1925-1940and Montanari et al. 2015 relate to the etymological proposals discussed by Chantraine 1968Chantraine -1980Chantraine [2009 discussed below. The gloss offered in Snell 1979 assumes a connection with the perfect stem ἅδηκ-'pleased' (ἅδηκε βουλή 'the plan was approved' in Hipponax frag. 132; see West 1971: 110-152), but this perfect certainly belongs to ἁνδάνω 'please' (so correctly Liddell et al. 1925Liddell et al. -1940. 29 Frisk (1960) adopts the original meaning 'gesättigt' and a connection with ἅδην, in a much more cursory entry than Chantraine's. Beekes (2009) adds nothing further to the discussion, simply referring the reader back to Chantraine 1968Chantraine -1980Chantraine [2009, but declines to give a gloss. 30 ἅδην has a semantically plausible root etymology in */seh2-/ 'satiate' , to which formations built with /-ti-/ are attested in Italic (Lat. satis 'enough'), Germanic (Goth. saþs 'satiated'), and Baltic (Lith. sótis 'satiety'). The Greek forms would evidently reflect a zero-grade of the root, but are otherwise of unclear derivation.
Indo-European Linguistics 6 (2018) 117-151 tors as prohibitively strained (so Leaf 1900). That ἀδηκότ-is construed with a dative καμάτῳ rather than a genitive (from which the reading 'sated with tiredness' could be extracted) is also peculiar, given that ἄω 'sate' and the adverb ἄδην in Epic are regularly construed with genitives (e.g., Il. 19.402 ἕωμεν πολέμοιο '(when) we have had our fill of war'). While the base ἅδη-is formally acceptable for etymologizing ἀδηκότ-, the semantics are dubious, and the necessarily young date of the formation for which a kappatic perfect would speak is out of sync with fact that the form is known exclusively in a seemingly established syntagm. The second option presented by Chantraine is a connection to a form ἀαδεῖν (Hsch. α 10 Latte), glossed by ὀχλεῖν 'disturb, importune' , and possibly related to ἀηδής 'distasteful, unpleasant' . The uncontracted sequence αα-in ἀαδεῖν would point to an original *αϝαδε-, in which case Chantraine finds the contraction to α-in Homer surprising.31 Yet as Chantraine (s.v. ἀαδα) notes, ἀαδεῖν may have been invented outright to provide an explanation for ἀδηκότες. Both etymological proposals found in Chantraine thus present considerable problems, and so room for alternative proposals is open.
Under the view that ἀδηκότες contains the negative prefix *[n̥ -], two considerations are important: 1) word-formation concerning *[n̥ -] and participles; 2) the consistent scansion of the first syllable as long. On the one hand, the occurrence of privative ἀ-on forms belonging to the verbal system in Ancient Greek is unusual. On the other hand, credible examples are not unknown, e.g., ἀπειθέω 'disobey' (largely attested as the sigmatic aorist ἀπίθησε in Epic; but see Peters 2007 for a different analysis), ἀτίει 'dishonor' , ἀέκοντες 'unwilling' . The view of Wackernagel (2009Wackernagel ( : 759 = 1928 was that the negation of participles with ἀ-rather than οὐκ was an archaism: "the old practice of negating the true participle with the privative prefix gradually declined, at the earliest date in Greek." While no other perfect participles containing privative ἀ-are known in Homer (cf. Risch 1974: 341-348), at least two R̥ gvedic perfect participles with privative a-are attested: ájaghnuṣī-'not striking' (8.67.15c) and ásaścuṣī-'not drying up ' (9.86.18c).32 The combination of Wackernagel's discussion and the Vedic examples suggests that the perfect participle with privative suffix *[n̥ -] might have been possible in PIE, and thus that the presence of negative ἀ-is compatible with the interpretation of ἀδηκότες as an archaism (see further 2.4 31 Here, Chaintraine notes, the form ἁδήσειεν (Od. 1.134) 'be disgusted by' , whose connection to ἀδηκότ-is proposed by Apollonius, could be a corresponding aorist. The formal problem with a connection to ἁδήσειεν is that the form might better be read as ἀηδήσειεν, on account of the meter; cf. fn. 27 above. 32 Thanks to a reviewer for reminding me of these Vedic comparanda.
sandell Indo-European Linguistics 6 (2018) 117-151 below). ἀέκοντες 'unwilling' , as the negated form of an old participle ἕκων 'willing' that synchronically lacks a base of derivation, suggests a further possibility for explaining the negation of ἀδηκότ-: an unattested *δηκότ-'aware, perceptive' was divorced from the paradigm of δέχομαι (to which only a perfect middle participle exists in 1st Millennium Greek), and thus regarded as a simple adjective; in this case, negation with ἀ-would be a trivial further formation. Regardless, the question of negation strategies for participles in older Indo-European languages remains a topic that requires closer investigation on a broader basis of evidence.
The fact that the ἀ-of ἀδηκότ-is always treated as long metrically poses as relatively minor issue. Of course, the phrase καμάτῳ ἀδηκότες, if scanned as ⏑ ⏑ -⏑ -⏑ ⏑, would not fit the hexameter, having one short syllable trapped between two longs (a cretic sequence). Thus, the ἀ-must be treated as long, giving a scansion ⏑ ⏑ ---⏑ ⏑.33 Taken on its own, the form ἀδηκότες or ἀδηκότας, having the shape ⏑ -⏑ ⏑ can be fit into a hexameter unproblematically after a sequence -⏑. It is, however, well-known that especially an underlying short /ă/, especially in the first syllable of a word, may be subject to metrical lengthening, whether to repair a tribrach sequence (e.g., in ἀθάνατοι 'immortals' or ἀκάματος 'tireless') or a short that would be trapped between two longs when the preceding word ends in a long or two shorts (Ἀπολλωνι, forms of ἀείδω 'sing'); cf. Chantraine 1958: 97-98. Clear instances of the latter, directly or roughly comparable to καμάτῳ ἀδηκότες, in which a word-initial /ă-/ scans long, are can be illustrated with the following lines: -Μουσαών θ' αἳ ἄειδον ἀμειβόμεναι ὀπὶ καλῇ (Il. 1.604).
-καρδίῃ ἄληκτον πολεμίζειν ἠδὲ μάχεσθαι (Il. 11.12 = 14.152). Examples of this sort can easily be multiplied. Note in particular that the form ἄληκτον 'unceasing' exhibits lengthening of an etymological privative prefix, just as would be supposed for ἀδηκότες. In sum, the lengthening of /ă/ within Epic diction to avoid the trapping of a short between two longs is philologically well-established.
Two other relevant observations may be made in connection with the metrics of καμάτῳ ἀδηκότες. One is that καμάτῳ in Homer is robustly localized in 33 A reviewer raises the question of whether the absence of epic correption here has any import for the chronology of this formula. Based on the recent work of Garner (2011) on correption (see especially pp. 133-140), who has shown that there are not significant dif-Indo-European Linguistics 6 (2018) 117-151 the second half of the second foot (10/14 instances, including 4/5 instances of καμάτῳ ἀδηκότες). Besides ἀδηκότες, the other attested options for filling the third and fourth foot following καμάτῳ are δὲ καὶ ἱδρῷ (Il. 17.385), τε καὶ ἱδρῷ (Il. 17.745), τε καὶ ἄλγεσι (Od. 9.75, 10.143), ἀρημένος (Od. 6.2), δεδμημένον (Od. 14.318), and θυμαλγέϊ (Od. 20.118). This leads to the second observation, namely, that the two perfect (middle) participles ἀρημένος 'distressed, βεβλαμμένος' δεδμημένον 'subdued' attested after καμάτῳ are of the metrical shape --⏑ ⏑. Taken together, καμάτῳ ἀδηκότες, καμάτῳ ἀρημένος, and καμάτῳ δεδμημένον may point to a general formulaic template (or Construction in the sense of Bozzone 2014) καμάτῳ [ --⏑ ⏑] Perf.Part. in the Epic language; the lengthening of /ă/ in ἀδηκότες would be necessary to satisfy this template. In short, there is no strong reason to prefer an etymology of ἀδηκότ-in which the long [ā-] is original. The combined features of Epic diction treated here show that the assumption of metrical lengthening is relatively unburdensome. I would contend that a metrical lengthening is small price to pay for a plausible morphological and etymological understanding of ἀδηκότες.

2.3
Homeric ἀδηκότ-: contextual interpretation In this section, I will examine five of the six occurrences of the stem ἀδηκότ-in context (the passage at Il. 10.396-399 exactly repeats Il. 10.309-312 and therefore will not be examined separately). In each case, I believe that the context virtually speaks for itself: a rendering of ἀδηκότ-as 'inattentive, unaware, heedless' (⇐ 'not having mentally taken in one's external circumstances') fits each passage not just unproblematically, but indeed well. Consistently translating ἀδηκότ-in this way also makes much better sense of the formulaic expression καμάτῳ ἀδηκότ-, where many freer translations (e.g., Lattimore 1951 of the Iliad and McCrorie 2004 of the Odyssey) essentially only pick up on the meaning of the known element, κάματος 'weary' , or produce a more or less redundant expression (like the "stremato dalla stanchezza" = "exhausted by tiredness" of Cerri (1996)). Under the reading 'having become inattentive/heedless due to weariness' , the form ἀδηκότ-makes a genuine, non-trivial semantic contribution to the passages in which it appears.
In the subsequent sections, each passage is given with English translation; translations from Il. 10 are based on Murray 1924; Od. 12.279-285 is based on Murray 1919; verses 456-461 of the Hymn to Apollo are adapted from Càssola 1975Càssola [1994 and West 2003. The particular translation given to the phrase sandell Indo-European Linguistics 6 (2018) 117-151 καμάτῳ ἀδηκότ-is bolded in the English in each case. The Greek text reproduced is that of Allen 1931 (Iliad), von der Mühll 1962 (Odyssey), and Allen et al. 1936 (Hymn to Apollo).34

2.3.1
Il. 10.98 The context: Il. 10.96-99: Agamemnon, taken with insomnia and worry, has just come to find Nestor, to consult with him about sending spies into the Trojan camp.
But if you wish to do something, since sleep does not come to you either, let us go down to look on the sentinels, in case, having become inattentive on account of weariness and sleepiness, they have gone to sleep, and entirely forgotten their watch.
Here Agamemnon, before making any decisions, wants to check that the nighttime sentinels (τοὺς φύλακας) of the Achaeans are in condition needed to carry out their principal duty (i.e., watch for intruders or danger). One may infer that ἀδηκότες here would seem to indicate precisely opposite of the mental state that is desirable for a sentinel, that is, sharpness and attentiveness.

2.3.2
Il. 10.305-312 The context: Just after the leaders of the Achaeans have met, and girded Diomedes and Odysseus to go out on a raid in the Trojan camp, the narrative shifts to the Trojans, where Hector, like Agamemnon at 10.98, is awake, and is attempting to rouse a volunteer to spy amongst the Achaeans.

34
The text of precisely these editions is reproduced because they are available through the Thesaurus Linguae Graecae, and thus easily accesible to many readers. I have noted no differences in other editions of these texts that are relevant to the present discussion, apart from the breathing on ἀδηκότες already discussed above.
For I will give him a chariot and two horses with high-arched necks, those that are best at the swift ships of the Achaeans, to the man who will dare-and for himself win glory-to go close to the swift-faring ships, and spy out whether the swift ships are guarded as before, or whether now, beaten at our hands, they [the Achaeans] are planning flight among themselves, and are not minded to keep watch through the night, being inattentive on account of terrible weariness.
Just as in 10.96-99, the focus here lies with the watchfulness of the military component. Likewise, just as 'guards' and the 'watch' made an appearance in the preceding passage, here two forms of φυλάσσω occur: once with respect to the ships, which ought to be subject to an attentive watch, and once with respect to the guard that should be kept through the night. Indeed, the phrase οὐδ' ἐθέλουσι νύκτα φυλασσέμεναι 'are not minded/willing to keep watch through the night' suggests that the act of φυλάσσω 'keeping watch' requires some deliberate mental concentration and effort. One may once again infer that it is the state of being ἀδηκότ-that would precisely inhibit one's capacity to keep watch effectively.

2.3.3
Il. 10.469-473 The context: Diomedes has just slain the unlucky Trojan spy Dolon, who, before his death, revealed to Diomedes and Odysseus where the camp of the Thracians would be found. The two Greeks have taken spoils from Dolon, hidden them, and continue through the Trojan camp, where they come upon the group of sleeping Thracians.
But they went forward through the weapons and the dark blood, and swiftly came in their course to the company of Thracian warriors. They sandell Indo-European Linguistics 6 (2018) 117-151 were sleeping, inattentive out of weariness, and their fair battle gear lay by them on the ground, all in good order, in three rows. And by each one was his yoke of horses.
Here the Thracians are simply asleep, obviously unable to take account of immediate dangers in their environment. Such inattentiveness, brought on by weariness and promoting sleep, then precisely allows Diomedes and Odysseus to come among the Thracians without meeting any resistance and slay twelve of the company, and for Odysseus to release their horses, before receiving any response.
You are stubborn, Odysseus; neither your mind nor your limbs ever grow weary. Indeed, you are wholly made of iron, as you do not allow your comrades, rendered inattentive with weariness and sleepiness, to go ashore, where on this land surrounded by water we could prepare once again a savory supper; instead you order us to wander on through the swift night, driven away from the island, over the misty sea.
Odysseus' rowers at this point must be suffering from attentional exhaustion: they are telling Odysseus that, in their present mental condition, and given the night and mist that would make the task of navigation more challenging, they cannot make productive progress, but will merely 'wander' (ἀλάλησθαι) if compelled to continue. The state and capacities of the crew are explicitly contrasted with those of Odysseus, whose limbs and mind both are said not to suffer from exhaustion. This contrast implies that ἀδηκότ-involves more than just the physical tiredness implied by κάματος.
Indo-European Linguistics 6 (2018) 117-151 2.3.5 Hymn to Apollo 456-461 The context: Phoibos Apollo addresses men on a ship arriving to Crisa, which he has diverted there from its original destination, Pylos, so that the men will be compelled to serve as his priests. This long and stressful journey is recounted in verses 413-439.
Why do you sit so afflicted with sadness, without going out onto the land, or putting away the ship's equipment? That is the custom of civilized men, when they reach land in their dark ship, rendered heedless by weariness, and their hearts are at once seized with appetite for sweet food.
The point of interest here is that ἀδηκότ-characterizes the state of sailors following a long (unwanted) sea journey; the usage here is precisely parallel to the preceding passage at Od. 12.279-285. In this particular case, the description 'rendered heedless by weariness' may resonate well with the condition of the specific sailors, who were compelled to endure a more difficult journey than anticipated, without, moreover, knowing the goal.

2.4
ἀδηκότ-as archaism The passages from the Epic Greek corpus give relatively clear indications that ἀδηκότ-describes a state of mind precisely opposite to the state of mind that would be desirable for activities where alertness and careful attention is warranted (e.g., keeping watch, steering a ship). The instances from Iliad 10 and Odyssey 12 support the reading of ἀδηκότ-as 'inattentive, heedless, unaware' nicely, while the instance in the Hymn to Apollo is entirely compatible with this interpretation, even if such a sense appears somewhat bleached in context.
Overall, this connection with attention and mental focus is entirely apposite to a derivation from the root */dek-/; ἀδηκότ-then may be seen as the negated form *[n̥ dēkȗ̯ ós-] 'inattentive' of the perfect participle *[dēkȗ̯ ós-] 'attentive' built to */dek-/. In 1st Millennium Greek, the likely paradigmatic isolation of ἀδηκότ-(in antiquity, any presumed relationship to the verb δέχομαι had evidently been lost, given that the earliest etymologies conceived [cf. under 2.1 above] fail to draw such a connection) and relative fixity in the formula καμάτῳ sandell Indo-European Linguistics 6 (2018) 117-151 ἀδηκότ-point unambiguously to an archaism, just as the characteristics of dāśvaṃs-in Vedic discussed above were likewise argued to be suggestive of an archaism.
The very fact that, morphophonologically speaking, ἀδηκότ-cannot be derived in from /dek-/ in 1st Millennium Greek also speaks in favor of an archaism. The systematic investigation into Greek reduplication in Zukoff 2017b shows that the generally preferred repairs to potential PCR problems (which virtual candidate *[dedkȗ̯ ós-] would have encountered) in reduplicated forms in Greek were either "Attic" reduplication (in the case of roots with an initial */HC-/ sequence) or omission of the consonant of the reduplicant (assuming that zero-grade ablaut would still have applied). A productively rebuilt perfect active participle with zero grade of the root in Greek might have resulted in a X ἐκτότ-, with "non-copying" reduplication (cf. perfects ἔκτονα [κτείνω 'kill'], ἔσταλκα [στέλλω 'prepare']) and thorn-cluster treatment of *[dk]; with full grade of the root, simply X δεδεκότ-(cf. the Homeric middle participle δεδεγμένος). Since δέχομαι/δέκομαι is synchronically a medium tantum in 1st Millennium Greek, any other perfect active participle is unattested. This fact, too, would likewise mean that ἀδηκότ-, if connected to IE */dek-/, is likely a relative archaism.
Nevertheless, ἀδηκότ-, unlike Vedic dāśvaṃs-, cannot simply be a direct inheritance of a form built in PIE, simply run through the expected sound changes. Namely, the fact that the perfect participle suffix has been reformed or replaced in the history of Greek, and that the sequence *[kȗ̯ ] would be expected to yield [p] in 1st Millennium Greek (falling together with inherited labiovelars), excludes the scenario of mechanical inheritance. Specifically, had an Indo-European nominative plural form *[n̥ dēkȗ̯ óses] been directly inherited into 1st Millennium Ionic Greek, a form X ἀδηπόες (or contracted to X ἀδηποῦς) might have been expected.35 Before the situation attested in 1st Millennium Attic-Ionic Greek, in which perfect participles are productively derived with a suffix /-ót-/, the only point of reference later than PIE itself is formed by a handful of neuter plural forms in Mycenaean, a-ra-ru-wo-a [araːrwóha] 'fitted' and te-tu-ko-wo-a2 [tetukhwóha] 'produced, built' (see the discussion in Szemerényi 1967), which continue the full grade of the Indo-European suffix */-u̯ ós-/. Since no forms in 1st Millennium Greek preserve the Indo-European full-grade */-u̯ ós-/, that ἀδηκότ-, has, in one way or another, 35 A simplex [p] rather than geminate [pː] as in ἵππος < */h1kȗ̯ o-/ is expected, since the vowel preceding the *[kȗ̯ ] sequence is long (cf. Kostopoulos 2014: 186-198). Thanks to a reviewer for raising this issue and for the relevant reference.
Indo-European Linguistics 6 (2018) 117-151 been given the productive suffix /-ót-/ requires no special explanation. Historically, as Szemerényi (1967: 23-24)  Finally, there is also a possible textual objection against viewing ἀδηκότ-as a relative archaism in Greek, namely, the fact that the collocation within which it exclusively occurs appears principally in the tenth book of the Iliad, which has long been regarded as a later interpolation into the text of the Iliad (see generally Danek 1988, Danek 2012. This type of objection, however, rests on the problematic assumption that archaisms should necessarily be first attested in older textual layers (see especially Hackstein 2002: 80-87 on this point with special reference to Homer). Without a complete accounting of a language at a particular point in time, that some relatively older forms might escape attestation until later is inevitable. In considering the case of the R̥ gveda, casual use of the LIV2 turns up many archaisms that are first attested in Books I and X, but not in the Family Books-not all forms that are first attested later are 36 Perhaps the best possible example of a metrical trace of [w] in Homer is the participle to ἁνδάνω, ἑαδότα 'pleased' (Il. 9.173 = Od. 18.422) < *[hwehwadwóta], discussed in Leumann 1955. The long ᾱin the form can easily be the result of metrical lengthening to avoid a sequence of four (!) shorts. Likewise, other forms of the same perfect stem in Epic (e.g., ἕαδεν at Ap. Rhod. Argonautica 1.867) also scan with a long ᾱthat can only be explained as a metrical lengthening. sandell Indo-European Linguistics 6 (2018) 117-151 themselves perforce younger. Although some scholars (of the Odyssey in particular) have asserted that καμάτῳ ἀδηκότες in Iliad X is dependent upon the occurrence in Book 12 of the Odyssey (so Laser 1958: 393-394, Heubeck andHoekstra 1989: 134), this view is not universally shared: contrast Danek 1988: 85 (quoted under 2.2 above) and Hainsworth 1993: 166 (re 10.98: "The similarity of the verse to Od. 12.281 has been used as a leading argument for the dependence of this Book on the Odyssey … but could equally be attributed to the random effects of formular composition"). Iliad 10 in particular also contains mention of one of the most striking material archaisms in the Iliad: the description of a boar-tusk helm (Il. 10.260-265), a type of object known in the Mycenaean culture of the 2nd Millennium BCE, but absent in 1st Millennium Greece (cf. Andersen and Haug 2012: 9)-the helm's appearance in a later book does not speak against its antiquity. That the expression καμάτῳ ἀδηκότες does not occur elsewhere in the Iliad might be yet another stylistic peculiarity of the poet of the Doloneia as against the larger Iliad. From the point of view of oral tradition of Greek Epic, it is licit to assume that καμάτῳ ἀδηκότες belonged to the traditional oral-formulaic repertory, and the capacity of the poet of the Doloneia to modify the formula (cf. 2.1 above) is more coherent with the assumption that he knew and understood this phrase independent of its occurrence in the Odyssey. One might suppose that the formula's use was dispreferred in the style adopted in much of the Iliad; its relatively high frequency in Book 10 may precisely be another instance, if one follows the thesis of Danek (2012: 108-116), in which the poet of the Doloneia deliberately constructs verses contrary to the linguistic and stylistic expectations established by the rest of the Iliad. At the same time, it may be the case that καμάτῳ ἀδηκότες is thematically inappropriate to much of the Iliad, since, to judge by its attestations, it belongs to scenes of nighttime rest (in Il. 10) or seafaring (in Od. 12 and the Hymn to Apollo); neither the sea nor the night (excepting Books 2 and 24, for the latter) make many appearances in the plot of the Iliad.

Conclusion: ἀδηκότ-, dāśvaṃs-, and the PCR
The first and second sections of this article have argued that Vedic Sanskrit and Homeric Greek each respectively preserve two archaisms: dāśvaṃs-'pious man' and ἀδηκότ-'inattentive, heedless, unaware' . When taken as archaisms, each of these forms is most readily explained as the ultimate continuation of an Indo-European "long-vowel" perfect participle built to the root */dek-/, *[dēkȗ̯ ós-] 'attentive, having given attention' . If the novel reading of ἀδηκότ-proposed is indeed correct, then the direct equation of two independent archaisms in Indo-European Linguistics 6 (2018) 117-151 Greek and Vedic implies that their last common ancestor possessed the form *[dēkȗ̯ ós-]-in short, that at least one such "long-vowel" perfect active participle can be projected back to (Core) Proto-Indo-European with some confidence. Vedic sāhvaṃs-'conquering, having conquered' is another credible such example, though a possible cognate form is as yet unknown. While the reconstruction of "long-vowel" perfect participle forms to account for dāśvaṃs-and sāhvaṃs-is not new (cf. Schumacher 2005: 640), the further support furnished by Greek ἀδηκότ-makes the rejection of their Indo-European antiquity (as per Jasanoff 2012: 128) harder to accept. Furthermore, given that a well-grounded phonological constraint (the No Poorly-Cued Repetitions Constraint of Zukoff 2017a) can motivate and predict the occurrence of such "long-vowel" perfect stems (see discussion under 1.2 above), I believe that one may justifiably view the occurrence of apparent long-vowel forms under morphological conditions where reduplication would otherwise be expected as a consequence of the phonology of Proto-Indo-European. Further details on the precise domain of application for a PCR constraint in PIE remain to be systematically investigated, but at least for now, the comparison of dāśvaṃsand (ἀ)δηκότ-admits of the tentative conclusion that the PCR was active when the second of two identical stops was part of a triconsonantal sequence.
Given then, that the PIE underlying form /RED-dek-u̯ ós-/ surfaced as *[dē-kȗ̯ ós-], at least in the last common ancestor of Greek and Indo-Iranian, I propose the following tentative reconstruction of PCR Effects in reduplication along the Indo-Iranian to Indic line of descent: The larger question concerning the morphology of PIE and its daughters then becomes how the category of "long-vowel" preterites (Cowgill 1957, Schu-sandell Indo-European Linguistics 6 (2018) 117-151 macher 2005, Jasanoff 2012) is to be handled. Can all such forms be ultimately explained as perfects of particular root-shapes and later innovations of the daughter languages? This question awaits a more thoroughgoing investigation.