Substantivization of adjectives


 The process of deriving substantives from adjectives in the classical Indo-European languages can be accomplished in two fundamentally different ways. The first possibility is a derivational one, i.e. the adjective is substantivized by a word-formation process that typically consists of an overt morphological or morphonological operation such as suffixation, accent shift, introduction of new ablaut grades, or a combination thereof. The second process, on the other hand, is a gradual one: an adjective can be substantivized through the ellipsis of a head noun that this adjective was originally paired with. In this paper, I intend to outline the differences and similarities between these two mechanisms and discuss their role in the interpretation of Proto-Indo-European stems in *-(e-)h2-.


Introduction
When talking about substantivization (or nominalization) of adjectives, most Indo-Europeanists will probably first think of substantivizing suffixes such as *-(o)n-, *-(o/e)t-, *-i-, *-(e)h2-, etc.,1 and formations like Lat. Catō, -ōnis m., a name (literally 'clever one'), from catus 'clever' , Gk. Στράβων, -ωνος m., a name (literally 'squint-eyed one'), from στραβός 'squint-eyed' , or Goth. blinda m. 'a blind one' (as in Lk 18.39: blinda sums sat faur wig 'a certain blind man sat by höfler Indo-European Linguistics 8 (2020) 181-204 the way'), from blinds 'blind' , and so on.2 Substantivizing suffixes are known in modern languages, too. Compare English -y (as in smart → smarty 'knowit-all' , or white → whitey 'a silvery fish'), German -ling (as in schwach 'weak' → Schwächling 'wimp, weakling' , weiß 'white' → Weißling 'white butterfly'), or French -et (as in rouge 'red' → rouget 'red mullet (a fish)' , bleu 'blue' → bleuet 'cornflower'). These suffixes are, however, quite marginal and largely unproductive. The substantivization of adjectives in languages like English, German, and French is productively achieved by a combination of syntactic, morphological and/or lexical methods: to refer to something that is green we can say a green one or the green one in English, ein Grüner/eine Grüne/ein Grün(es) or der Grüne/die Grüne/das Grün(e) in German, and un vert/une verte or le vert/la verte in French; to refer to the abstract concept of 'green-ness' , we can use a/the green, ein/das Grün, and un/le vert, respectively (as in the green of the dress, das Grün des Kleids, le vert de la robe). What can be seen from these examples is that the article (± morphological or lexical material) fulfills the role of a determiner to substantivize an adjective. In functional terms, we could therefore think of the PIE suffixes *-(o)n-, *-(o/e)t-, *-i-, *-(e)h2-, etc., as analogous to our modern-day articles, though this anachronistic comparison should not be overstrained.
Yet there is another, lesser-known process that leads to the substantivization of an adjective, both in PIE and the IE daughter languages. This process is ellipsis, and it is-as we shall see-fundamentally different from the mechanisms outlined above. An ellipsis is the omission of a substantive that an adjective was originally paired with, so that the adjective alone remains in substantivized meaning. A case in point is the Latin adjective calidus 'hot' in the syntagma aqua calida 'hot water' (Cato Agr. 156.5 [brassicam] coicito in aquam calidam 'throw [the cabbage] into hot water') and (with ellipsis of aqua) in the substantivization calida f. 'hot water' (Cato Agr. 156.3 qui poturus erit, lauet calida 'before drinking from it, one should bathe in hot water'). In this paper, I intend to outline and discuss the basic similarities, but also the fundamental differences between these two descriptive types of substantivization of adjectives.

Ellipsis
Substantivization via ellipsis3 is a gradual process. It starts out as a syntactic phenomenon, and at this stage is known as contextual ellipsis or antecedentbased ellipsis (cf. Günther 2018: 77-78 with references). An example from Ancient Greek with the adjective κύνειος 'belonging to a dog' can be seen in the following passage from (Archigenes apud) Galen Comp. Med. Sec. Loc. 6 (12.954).
'Canine excrement of dogs which were fed bones is by far the best and helps a great deal, and furthermore causes no nauseousness to the patient. (…) Musa was the first one to use canine (sc. excrement), he always took the white one.' Since a κυνεία κόπρος is mentioned at the beginning of the passage, the context allows the substantive κόπρος 'excrement' to be ellipted in the second reference (τῇ κυνείᾳ) without this being detrimental to the overall comprehensibility of the sentence. Note that the adjective λευκήν in the clause immediately following τῇ κυνείᾳ is likewise a case of ellipsis, yet an even more subtle one since the antecedent is again the noun κόπρος (or, conceivably, κυνεία), but without there being an earlier reference to a syntagma λευκὴ κόπρος (or, for that matter, λευκὴ κυνεία). In both cases, the context provides the pragmatic background that enables our understanding of the passage. If the context is, however, ambiguous, the omission of the word κόπρος would not lead to a meaningful sentence. The same can be said about an example from Modern English. Billy Joel's 1977 song Scenes from an Italian Restaurant concludes after almost 8 minutes with the following verses: A bottle of red, a bottle of white / Whatever kind of mood you're in tonight/ I'll meet you any time you want / In our Italian Restaurant. Any listener immediately understands that the singer refers to a bottle of red or white wine that he offers to drink with the object of his desire even höfler Indo-European Linguistics 8 (2020) 181-204 though the word wine is never explicitly uttered in the entire song-it is just not very likely that someone would like to drink red or white dye, red or white pasta sauce, and perhaps not even red or white non-alcoholic grape juice. Without context, however, sentences like Here is a bargain organic white or I'd have preferred red are not immediately understandable as referring to white wine or red wine, respectively. Compare the same sentences, but with antecedent: Languedoc is better known for its red wine, but here is a bargain organic white from Château de Lascaux and He gave me a glass of white wine but I'd have preferred red.4 Just for the sake of the argument, however, the sentences, in a different context, could also read: Our company is better known for red pasta sauces, but here is a bargain organic white (one) with bacon and gorgonzola and He gave me a glass of white grape juice but I'd have preferred red.
Only over the course of time has such a contextual ellipsis the potential to become visible and relevant as an actual and descriptive word-formation process.5 The final step of this development is the lexicalization of the adjective as a substantive. Whether or not the process reaches this point (and the majority of contextual ellipses do not) is undoubtedly facilitated by a gradual obsolescence of the adjective in adjectival meaning in other contexts than with the soon-to-be ellipted substantive (which is arguably the reason why red and white cannot be interpreted as lexicalized substantivizations in the meaning 'red wine' or 'white wine').6 A telling example comes from Vedic, where we find a substantivization Ved. mahiṣá-m. 'buffalo' (RV+). The underlying adjective mahiṣá-started out as an ordinary adjective meaning 'tremendous' (from *megh2-s-ó-'having (considerable) size'), but already in the Rigveda is only marginally used to refer to things other than the buffalo, as in 10.66.10 mahiṣásya tanyatós 'of tremendous thunder' . In the other adjectival attestations it is used in combination with mr̥ gá-m. 'animal' , denoting 'the tremendous animal' (for example mahiṣáṃ mr̥ gám RV 8.69.15), i.e. the 'buffalo' , which explains the lexicalization of mahiṣá-in its substantivized meaning. The sub-Indo-European Linguistics 8 (2020) 181-204 stantive mahiṣá-, unlike the adjective mahiṣá-, lives on in the Modern Indic languages (cf. Konkani mhos 'buffalo').7 Cases of Greek ellipses on their way to become fully lexicalized before our eyes include ἡ δεξιά / δεξιτερά (with or without χείρ) 'the right hand' (in Homer already always without χείρ), ἡ γυμναστική (with or without τέχνη) 'gymnastics' , ἡ μουσική (always without τέχνη) 'any of the Muses' arts' , and many more. As can be seen from all these examples, the gender of the substantivized adjective depends on the gender of the ellipted word.8 Examples with textual evidence for the adjective, the syntagma of adjective and substantive, and the substantivized adjective-as with Ved. mahiṣá-'tremendous' , mahiṣá-mr̥ gá-'tremendous animal' , and mahiṣá-m. 'buffalo' , or similarly Ved. pṛthvi-'broad' (the feminine of pṛthú-), kṣam … pr̥ thvim 'broad earth' (RV 10.31.9; cf. also YAv. ząm pər ə θβīm Y.10.4), and pr̥ thivi-, pr̥ thvi-f. '(Mother) E/earth'-are actually not too common. In many cases we are not lucky enough to have actual evidence from texts that authorize a secure interpretation as an ellipsis. In some cases, the adjective is attested but the underlying syntagma is not, as in Gk. ψωλός 'with the foreskin drawn back' and ψωλή f. 'membrum virile praeputio retracto' (Ar.), which is from an unattested *ψωλὴ πόσθη with πόσθη f. 'penis' , and in yet other instances, not even the adjective is attested. Compare (V)Lat. *formaticus 'shaped' , the source of French fromage, Italian formaggio, Catalan formatge 'cheese' , where the underlying unattested syntagma is *caseus formaticus 'shaped cheese' (as opposed to unshaped fresh cheese), containing an otherwise unknown adjective *formāticus 'shaped' (derived from forma 'form, mold' like umbrāticus 'belonging to the shade' from umbra 'shadow, shade' and viāticus 'pertaining to traveling' from via 'road'), or Gk. *κρατήριος 'related to the κρατήρ "mixing bowl", κρατήρ-like' , the source of κρατηρία f. 'bowl' (Dsc.), which is from an unattested syntagma *κρατηρία λεκάνη 'κρατήρ-like dish' .9 In these cases, it is often difficult to identify a substantivized adjective as an ellipsis-based substantivization. But since an ellipsis-sourced substantivization is morphologically still just an adjective, it is conceivable that the substantivization process did not have any 'morphological effect' (i.e. no suffixation, no accent shift, no change in ablaut10), which 7 Cf. EWAia (II: 340 s.v. mahiṣá-). 8 Compare also Lat. caprea, -ae f., the generic term for 'roe-deer' , whose feminine gender is explained very elegantly by Adams (1985b: 278) as originating in a syntagma of the old word for 'roe-deer' *i̯ órk-f. (Gk. ζόρξ and δόρξ f.) and an adjective *capreus 'goat-like' with subsequent ellipsis. 9 In the passage Dsc. 4.155, κρατηρία and λεκάνη refer to the same vessel used to collect the juice of some sort of wild cucumber. 10 The term 'effect' is used here in purely descriptive terms. In connection with accent shift höfler Indo-European Linguistics 8 (2020) 181-204 separates them from substantivizations that are created by means of the substantivizing suffixes mentioned above.

'Direct' substantivization
In order to draw a clear distinction between the two types of substantivizations, I intend to name the non-ellipsis type 'direct' substantivization, in the sense that the adjective is converted 'directly' into a substantive without the prior existence of a syntagma of substantive plus adjective and subsequent ellipsis. Unlike ellipsis that evolves gradually, a 'direct' substantivization is a spontaneous and deliberate word formation process, which is especially clear in the many cases where suffixation or another 'morphological effect' is involved (on which more below). The result of this process11 can be either: (A) a concrete individualization ('X' ADJ → 'a person/animal/thing that is X' SUBST ), or (B) an abstract ('X' ADJ → 'the state/quality/condition of being X' SUBST ) In many cases, the abstract can simultaneously have a concrete meaning. This is nothing remarkable and can be observed in many modern-day examples of abstract-forming suffixes. Compare the English noun whiteness as in 'the quality or condition of being white in color' (the whiteness of the snow on the mountains), which can also be used in a concrete sense 'a white substance or part of something' or 'snow' (I tried to imagine the movie-set roofs piled with fluffy whiteness).12 This can also be observed in the classical languages. The substantivizing suffixes mentioned at the outset typically create substantivizations that have both meanings.13 For derived i-stems, compare, for example, the Latin nouns febris, -is f. 'fever' < *dhegwhri-'heat' (from *dhegwh-ro-'hot'; cf. Gk. τέφρα f. 'ashes') and rauis, -is f. 'hoarseness' (from rauus 'hoarse') as abstracts, and cīuis, -is m. 'citizen' < *kȇi̯ u̯ i-'member of the society' (from *k(e)i̯ u̯ o-'socially close'; cf. Ved. śéva-, śivá-'auspicious, dear, friendly' , PGmc. *heiwa-'household' , Latv. siẽva f. 'wife') or imber, -bris m. 'rain' (cf. Osc. dat. pl. anafríss 'imbribus') < *n̥ bhri-m. 'wet stuff' (from *n̥ bh-ró-'wet'; cf. Ved. abhrá-n. 'rain-cloud' , Gk. ἀφρός and change in ablaut it would be more precise to speak of morphological and/or morphonological 'effect' . For the sake of brevity, however, I will subsume all of the mentioned 'effects' under 'morphological effects' (with quotation marks). 11 Cf. Nussbaum (2014: 304-06). 12 Cf. OED (s.v. whiteness, n. meaning 1. a. and 1. b.) where the bracketed sample sentences are taken from. 13 Cf. Nussbaum (2014: 304-05).
Indo-European Linguistics 8 (2020) 181-204 m. 'foam').14 It is of no significance for now whether the abstract meaning was original and the concrete meaning developed from it, if it was the other way round, or whether both meanings were inherent in such formations simultaneously and the specialization of individual suffixes to create one or the other was a secondary development. At least from a theoretical perspective, the latter option seems to be the most plausible one if we interpret the PIE suffixes *-(o)n-, *-(o/e)t-, *-i-, *-(e)h2-, etc. as analogous to determiners.15 When it comes to the formal side, the use of suffixes is actually only one of several ways to create a 'direct' substantivization. Apart from suffixation or suffix substitution (see the i-stem examples above), we find for example accent shift16 as in Ved. kr̥ ṣṇá-'black' → kr̥ ṣṇa-m. 'black antelope' , Gk. κνηκός 'pale yellow, tawny' → κνῆκος f. 'safflower' (Myc. ka-na-ko), etc. We can also observe a systematic change in ablaut, which for athematic adjectives is called 'internal derivation'17 as in *pum-és-'having a beard' (>> Lat. pūbēs, -eris 'pubescent; downy') → *p(é)um-ōs m. 'the bearded one'   The two phrases mentioned above containing whiteness in an abstract and a concrete meaning, respectively, can easily be rewritten using a determiner (the, some) plus the substantivized adjective (white) as the white of the snow on the mountains and I tried to imagine the movie-set roofs piled with some fluffy white. 16 Cf. In the 'zero' case, the formal overlap with ellipsis-sourced substantivizations is obvious. Or, in other words: it is not quite clear from the outset whether a word like, e.g., Ved. vatsá-m. 'calf' continues a direct substantivization of the adjective *u̯ et-s-ó-'one year old'23 or whether it goes back to a syntagma of this adjective with an ellipted word for 'calf' , 'steer' , or '(young) animal' . This question might not seem overly relevant at first sight but it is, I think, of fundamental importance for our understanding of substantivization as a word-formation process.

4
How to distinguish the two types?
Unfortunately, there is no easy way to distinguish ellipsis-based substantivizations from direct substantivizations with no overt 'morphological effect' . Sometimes, however, there is circumstantial evidence from the semantic side to help decide whether a deadjectival formation belongs to the former or the latter group.

4.1
Adjectival abstracts Adjectival abstracts, the type *dhegwh-ro-'hot' → *dhegwhri-'heat' (Lat. febris, -is f. 'fever'), are always the result of a direct substantivization. Ellipses virtually never create such abstracts. Exceptions to this rule are only apparent: even if words like γυμναστική 'gymnastics' , μουσική 'any of the Muses' arts' are abstracts in the sense that they do not refer to concrete entities, they can hardly be interpreted as adjectival abstracts of a semantic structure 'X' ADJ → 'the state/quality/condition of being X' SUBST . 22 The label 'change in ablaut' for both internal derivation and substantivization with a new accented e-or o-grade does not necessarily imply that these processes and their respective origin have to be analyzed in the same way. Another caveat is that it is not clear how (or, indeed, if) the three types of substantivization of thematic adjectives-viz. plain accent retraction, new accented e-grade, new accented o-grade-are interrelated. 23 For the formal analysis of this adjective (i.e. possessive adjective 'having a year' or degenitival hypostasis 'of the year') see Vine (2009: 216-17; arguing for the latter).

Concrete referents or 'individualizations'
The most prominent case of deadjectival substantivization as 'individualization' is in names. Deadjectival names are always 'direct' substantivizations, which is especially obvious in personal names like the already mentioned Catō and Στράβων types; an ellipsis of uir or ἄνθρωπος (or the like) is totally out of the question in these and any other cases of anthropological name-giving. This rule also extends to other names, like pet names (ἀργός 'swift' → Ἄργος, Odysseus's dog), ship names (see Ἀργώ above), or also many place names (ἑλικός 'eddying' → Ἑλίκη, a town in Achaea), etc.
Another difference is that while ellipsis-based substantivizations are always resolvable as 'a/the Y SUBST that is X ADJ ' or '(a/the) X ADJ Y SUBST ' (i.e. a syntagma that-by definition-is the ultimate source of this substantivization), a specific 'Y SUBST ' is often not needed in paraphrasing a direct substantivization; it is enough to think of 'the/a X ADJ one' , as is evident, for example, from the names listed above (e.g. Ἀργώ < 'swift one').24 In some direct substantivizations, 'X ADJ ' is applicable for all 'Y 's' and 'X SUBST ' is a synonym or paraphrase of 'Y ' . Compare the adjective *rot-h2-ó-'having wheels, wheeled' (derived from *rot-eh2-'wheel' , attested in Lat. rota f.) and its direct substantivization (with accent retraction) *róth2o-'the wheeled one' 24 In terms of underlying structure, one could interpret 'X SUBST ' more generally as 'the/a X ADJ one' and think of 'Y ' as a silent empty noun (cf. Günther 2018: 99) whose underlying semantic representation can be rendered as 'person' , 'animal' , 'thing' , or just 'one' , depending on pragmatic context. höfler Indo-European Linguistics 8 (2020) 181-204 (attested in Ved. rátha-m., Av. raθa-m. 'chariot'). Since all carts or chariots arguably possessed wheels, the substantive *róth2o-'the wheeled one' is a synonym or paraphrase of 'chariot' . No ellipsis of a word for 'cart' is, therefore, needed here. Similarly, the direct substantivization (with accent retraction) *kȓ̥ h2seh2-'the horned one' (on the basis of *kȓ̥ h2s-ó-'horned' , derived from *kér-h2-s n. 'horn' , attested in Gk. κέρας) is the source of PGk. *kárahā f. 'the horned one' > Cret. κάρᾱ f. 'tame goat' (Hsch.).25 Here, 'the horned one' is likewise a synonym or paraphrase of 'goat' since all goats (both male and female) have horns.26 In yet other cases, the direct substantivization is explicitly not resolvable as '(a/the) X ADJ Y SUBST ' , namely when the adjective 'X ADJ ' does not necessarily apply to the semantic head as a whole, but rather only refers to one prominent feature of it. The term Lat. pānsa, -ae m., also used as a cognomen Pānsa (from pānsus 'spread out'), for instance, does not refer to a spread-out person but rather to someone with feet that are spread-out or splay.27 A similar case is scaeuus 'left' and scaeua m. 'left-handed person, lefty' . Another field where this feature can be observed nicely is in fish names. The English whitling 'a sea trout' (from white), and the Greek λεῦκος 'a fish' (from λευκός 'white') do not have to be 'white fish' . The name can refer to a prominent white stripe or mark, white flesh, or a relative whiteness in relation to a darker or different-colored but otherwise similar species. The Greek μύξος fish, a kind of mullet (Ath.), derives from *muk-s-ó-'slimy' ,28 but does not necessarily designate a 'slimy fish' . The name probably refers to the adipose eyelids that the family Mugilidae (or mullets) is known for. Lastly, the English fish name roughy (as in orange roughy) does not designate a 'rough fish' , but rather a fish with large rough-edged scales.29 A special case of concrete referents is the formation of collectives and 'delibatives' . As convincingly argued by Nussbaum (2014: 277-88), the substantivization of possessive or genitival adjectives derived from (i) countable or 25 Cf. Nussbaum (1986: 156-57). 26 In contrast to goats, only the males of red deer have 'horns' or antlers, which is why derivatives of *kó/éru-'horn' such as *kȇru̯ o-(Lat. ceruus), *kȓ̥ u̯ o-(Welsh carw), *kȇrud-(OHG hiruz, OE heorot), etc., originally only referred to the 'hart' or 'male cervid' (< 'the antlered one'), as argued by Adams (1985b: 274-75), the generic term for 'red deer' being *h1el(h1)en-(plus variants). Within Tocharian, the adjective *kȓ̥ h2s-ó-'horned' is continued as TB karse 'hart, deer' . 27 Cf. Walde & Hofmann (II: 248 s.v. pānsa). 28 Cf. *méu̯ k-ōs > Lat. mūcor m. 'mold; mucous substance'; *múkseh2->> Gk. μύξα f. 'mucus, slime' . On the remodeling of inherited ā-stems to ă-stems in Greek see Solmsen (1909: 238). 29 Cf. OED (s.v. roughy, n. meaning 2).
Indo-European Linguistics 8 (2020) 181-204 count nouns and (ii) uncountable or mass nouns can lead to the semantics of (i) a collective, i.e. 'a syntactic [and/or] … morphological singular denoting a group of "identical" things' ,30 and (ii) a delibative, i.e. 'a finite sample or … a particular variety of the stuff in question' ,31 respectively. For the collective type, compare Lat. arbōs 'tree' → arbustum n. 'copse, plantation' (thematic neuter in *-to-), and for the delibative type Lat. strāmen 'straw' → strāmentum n. 'bundle of straw' (thematic neuter in *-to-).32 These collectives/delibatives are also based on 'direct' substantivization processes, as can be seen from examples showing a 'morphological effect' such as accent retraction; compare Gk. ἀήρ 'mist, haze; air' (*h2usḗr 'morning thing') → *h2usr-ó-(cf. Ved. usrá-'matutinal') → Gk. αὔρα, Ion. αὔρη f. 'a cool breeze, the fresh air of morning, morning mist' (quasi *h2úsreh2-in a delibative sense 'a bit of mist/air'),33 or *méu̯ k-os n. 'slime' → *muk-s-ó-→ *múkseh2->> Gk. μύξα f. '(a bit of) mucus, slime' (see also note 28), a 'delibative' with plural forms attested since Hesiod. These semantic deliberations should be seen as general tendencies rather than rigid rules. In many cases, of course, the distinction can be blurry.34 For the above-mentioned example Ved. vatsá-m. 'calf' (from *u̯ et-s-ó-'one year old'), for instance, no clear verdict can be decided. Semantically speaking, a syntagma 'one-year-old calf' with subsequent ellipsis is just as plausible as 'the oneyear-old one' as a direct substantivization. Any identification should, therefore, rely-if possible-on (a combination with) formal arguments rather than on semantic considerations alone. 30 Nussbaum (2014: 285). 31 Nussbaum (2014: 278). 32 Examples taken from Nussbaum (2014: 290-92). 33 Cf. Nussbaum (2014: 297). 34 In some cases, the boundaries seem particularly fluid: there is a subtype of deadjectival 'direct' substantivizations that shares a specific feature with ellipsis-based substantivizations in that the gender of the substantivization is dependent on the gender of another word. But unlike in the ellipsis-case where the gender of the substantivized adjective depends on the ellipted word, the gender of this type depends on the gender of a semantically associated generic term or synonymous model word, or on a lexically determined gender restriction. Examples of the former type might be the already mentioned fish names λεῦκος and μύξος, masculines just like the generic term ἰχθύς m. 'fish' , and the word for 'palm (of the hand)' (see below in section 5) Lat. palma f., Gk. παλάμη f., etc., whichas most words for 'hand' in the IE languages-is a feminine; an example of the lexically determined gender restriction type is Gk. λεύκη f. 'white poplar' that presumably owes its gender to the fact that most dendronyms in Greek are feminine (even though the generic term for 'tree' , δένδρ(ε)ον, is neuter). höfler Indo-European Linguistics 8 (2020) 181-204

The formal side: substantivizations with *-(e-)h2-
In case of suffixation with substantivizing suffixes such as *-(o)n-, *-(o/e)t-, *-i-, or *-k-(cf. *séno-'old' → *sene-k-m. 'old guy' > Lat. senex), it is obvious that we are dealing with a 'direct' substantivization, i.e. a substantivization that was intentionally created by a derivational process that consisted of adding a suffix to the stem of the adjective. The same holds true for other 'morphological effects' accompanying the substantivization, such as change of ablaut (internal derivation, substantivizations with a new accented e-grade, substantivizations with a new accented o-grade) and change of accentuation (accent shift), or a combination of them. This means that, by implication, these morphological processes are excluded from appearing in ellipsis-based substantivizations, which themselves-as already mentioned above-look still like adjectives on the surface. In other words, in ellipsis-based substantivizations there is no other formal process to mark or create the substantivization. One group of words, in which these premises play a significant role, are formations in PIE *-(e-)h2-, which therefore deserve a more detailed treatment. For a substantive that can be traced back to a formation in which *-(e-)h2is seemingly attached to the stem of a thematic adjective, can be one of two things: (A) the feminine of this thematic adjective, substantivized via ellipsis. (B) a 'direct' substantivization of this adjective via the suffix *-(e-)h2-. In (A), *-(e-)h2-is just an agreement marker added to the thematic adjective to express gender concord. In (B), however, the formant *-h2-is an independent substantivizing suffix that attaches to the thematic vowel *-e-to create 'direct' substantivizations, i.e. either concrete individualizations ('X' ADJ → 'a person/animal/thing that is X' SUBST ) or abstracts ('X' ADJ → 'the state/quality/condition of being X' SUBST ). As a substantivizing suffix, *-h2-was primordially not associated with feminine gender, which can be deduced not only from the rich evidence of Anatolian,35 but also from the masculines in Gk.
A clearer picture emerges from some 'doublets' , i.e. from instances where it is possible to identify both a 'direct' and an ellipsis-based substantivization of one and the same adjective in the individual languages. One case in point is the pair Ved. varṣa-f. 'rainy season' (AV+) and Gk. ἐέρση f. 'dew' (ep. poet. Il.+). Both nouns can be seen as a substantivization of the adjective *h2u̯ ersó-'rainy, wet' (cf. Ved. varṣá-n. 'a rain' RV+).64 However, the seemingly parallel formations do not only differ in meaning, but also in the position of the accent.65 For Greek, the preservation of *-rs-as -ρσ-after the accented vowel66 points towards an immediate preform *au̯ érsā < *h2u̯ érseh2-, while Ved. varṣa-seems to continue an oxytone *h2u̯ erséh2-, at least without having to invoke additional hypotheses. In combination with the semantic mismatch of the two terms, we could be tempted to identify *h2u̯ érseh2-as a 'direct' substantivization 'moisture' (Gk. ἐέρση 'dew') on the one hand, and Ved. varṣa-'rainy season' (from virtual *h2u̯ erséh2-) as an ellipsis-based substantivization on the other, namely from a syntagma with a feminine word for 'season, time of the year' , like for example (the precursor of) Ved. sámā-f. 'year, half-year, season' (RV+), originally *'summer' (cf. YAv. ham-'id.').

Conclusion
The aim of this paper was to establish a distinction between two basic processes of substantivization in PIE and the IE languages, namely between 'direct' substantivizations and substantivizations via ellipsis, and to outline the basic similarities and the fundamental differences between these two mechanisms. I argued that a 'morphological effect' (i.e. suffixation or suffix substitution, change in ablaut, accent shift, or a combination of them) in the substantivization vis-à-vis the underlying adjective is indicative of the former process, while in ellipsis-based substantivizations the substantivization looks just like the adjective of an erstwhile noun phrase, in which this adjective agreed with a subsequently ellipted substantive. This broad rule is only somewhat tarnished by a subgroup of 'direct' substantivizations that show no 'morphological effect' 67 For the philological details cf. Schaffner (2001: 418-20 with notes). 68 I refrain from invoking a specifically 'feminine' agreement here. See also below in the Conclusion. 69 Cf. in nuce Dieu (2016: 152) for Greek.
(like, for example, Gk. ἀγαθός 'good' → Gk. ἀγαθόν n. 'good, blessing, benefit' and Gk. ἀγαθός m. 'a noble man'); however, the semantic considerations outlined in section 4 might be able to serve as a guideline for their categorization.
Regarding stems in *-e-h2-that can be seen as either 'direct' substantivizations of thematic adjectives or as feminine adjectives substantivized by ellipsis, I argued that the same rules apply: if the deadjectival stem in *-e-h2-is barytone, it is of the former, if oxytone, of the latter type. A couple of handbook examples of secondary oxytone *-e-h2-stems (as, for example, the 'wool' word) are reconstructible for PIE including Anatolian, which means that this hypothesis might have repercussions on our understanding of adjectival agreement in the proto-language. However, this will have to await another occasion.