Research & Projects


The Origin of the Feminine Gender in

The origin of the feminine gender, the rise of adjectival agreement, and the development of agreement forms for the feminine gender have played a major role in my research over the past 5 years. I think these questions are some of the most puzzling ones in Indo-European linguistics. 

Grammatical gender is one of the most interesting categories in the world’s languages. Within the prehistory of the Indo-European family, we can discern a shift from an earlier two-gender system to a three-gender system, distinguishing masculine, feminine, and neuter nouns. The precise origin of the feminine gender has been a hotly debated question in Indo-European linguistics for over a century. 

I have investigated several different aspects of this topic in the following talks:

2021 “Agree to not Agree. On the Presence and Absence of Derived Feminine Adjectives in Greek and Indo-European.” 32nd Annual UCLA Indo-European Conference, University of California, Los Angeles / Online, November 5‒7. [handout]

2021 “Gender in Indo-European. A Synopsis.” Power, Gender and Mobility. Features of Indo-European Society, March 26–27, 2021, Copenhagen / Online. [slides] 

2021 “How to pull a wagon in Indo-European: The derivational history of the word for ‘shaft (of a cart)’ and remarks on substantivization and adjectival agreement in PIE.” Invited Speaker at the Cornell Linguistics Colloquium, Cornell University, Ithaca, New York / Online, February 25, 2021.

2019 “The two types of “secondary” *-(e-)h₂ stems in PIE.” Arbeitstagung der Indogermanischen Gesellschaft. 100 Jahre Indogermanistik an der Universität Ljubljana, Ein “Ljubljäum”, June 4–7.  [handout] 

2019 Story of O. On a peculiar substantivization type in PIE.” 5th Indo-European Research Colloquium, Leiden, March 21–22.  [handout]

2018 The Caprice of O.... On a Proto-Indo-European substantivization type and its excesses in Ancient Greek.” 37th East Coast Indo-European Conference, University of Michigan, Ann Arbor, June 15‒17.  [handout]

2017 “Hittite ḫišša- c. ‘thill, shaft (of a cart)’ and the feminine gender in Proto-Indo-European.” The Split ‒ Reconstructing Early Indo-European Language and Culture, University of Copenhagen, September 13‒15, Copenhagen, Denmark. [handout]


In addition, the following articles are devoted to this topic:

2022 “Kongruenz und Motion. Die femininen Formen des thematischen Adjektivs im Altgriechischen und Indogermanischen.” In Florian Sommer, Karin Stüber, Paul Widmer, Yoko Yamazaki (eds.), Indogermanische Morphologie in erweiterter Sicht, Grenzfälle und Übergänge, Beiträge zu einer 2020 in Zürich geplanten, aber nicht stattgefundenen Fachtagung der Indogermanischen Gesellschaft, Innsbruck: Innsbrucker Beiträge zur Sprachwissenschaft,  29–63. [download]

in press “Gender in Indo-European. A Synopsis.” In Riccardo Ginevra, Stefan Höfler & Birgit A. Olsen (eds.), Power, Gender and Mobility. Features of Indo-European Society. Proceedings of the conference, March 26–27, 2021. Copenhagen Studies in Indo-European Vol. 10, Copenhagen: Museum Tusculanum Press.

Adjectivization of substantives

Greek adjectives are one of the most fascinating things I know. Not only does Greek have thematic adjectives of two endings that have one agreement form serving for the masculine and the feminine at the same time (e.g., ῥοδοδάκτυλος Ἠώς 'rose-fingered Eos'), which is evidently an archaism, but it also exhibits masculine adjectives of the first declension in -ης (-ᾱς).

Greek adjectives in -ης (-ᾱς) such as ὑβριστής ‘violent, wanton’ are generally considered a secondary type, originating in an adjectivization of masculine substantives that became predominantly used in apposition. 

Τυφάονα ...  | δεινόν θ’ ὑβριστήν τ’ ἄνομόν τε ‘Typhon, terrible, outrageous, lawless’ (Hes. Th. 306–7)

While this is certainly the preferred analysis for a former agent noun such as ὑβριστής (: ὑβρίζω ‘wax wanton, run riot’), there is a second type of adjectives in ‑ης that behave (in meaning and function) just like the thematic adjectives they are seemingly derived from. Very often they appear as attributes or epithets in noun phrases.

One case is ἀργεστής ‘bright’ evidently from *arges-tó- ‘having, causing brightness’ (the s-stem *ἄργος, -εος ‘brightness’ being the basis of ἀργεννός ‘white’ < *arges-nó- and ἐν-αργής, -ές ‘visible’), attested in ἀργεστᾶο Νότοιο ‘of the white South Wind’ (Il. 11.306, 21.334), and ἀργεστὴν Ζέφυρον ‘clear Zephyrus’ (Hes. Th. 379), ἀργεστέω Ζεφύροιο (Hes. Th. 870). 

Another case is the adjective ἐτήσιος ‘yearly, annual’: in a syntagma with ἄνεμοι it seemingly becomes an adjective of the first declension (viz. ἐτησίαι ἄνεμοι pl. ‘the Etesian winds’). Compare the following passage in which it is used twice: once as the predicate of βορέαι in the shape ἐτήσιοι, and once as an attribute of βορέαι in the shape ἐτησίαι: 

Διὰ τί βορέαι μὲν ἐτήσιοι γίνονται, νότοι δὲ οὔ; … ἔτι οἱ μὲν ἐτησίαι βορέαι καθεστηκότος τοῦ ἀέρος πνέουσι (θέρους ǁ γὰρ πνέουσιν)

‘Why are the Boreas winds Etesian, whereas the Notos winds are not? … Furthermore, the Etesian Boreas winds blow when the air is still (for they blow in summer)’ (Arist. Pr. 940a35) 

My feeling is that the core of Greek adjectives in -ης (-ᾱς) could go back to an inherited stock of thematic adjectives in *-e-h₂ that were used adnominally in definite noun phrases in which *-h₂ had determining function.

Support for the reconstruction of determined thematic adjectives ending in *-e-h₂- might come from Anatolian, where the Luwian adjectival suffixes -izza- and -azza- do not show i-mutation and therefore probably continue *-i(s)k̑eh₂- and *-e/otyeh₂-, respectively. Just like the Greek forms in -ης, -ᾱς, these adjectives appear predominantly in adnominal position: cf. nom. sg. c. URUTaurišizzaš wašḫazzaš DLAMMA-aš ‘the patron tutelary deity of Taurisa’, dūwazza- ‘broad (?)’ (acc. sg. c. dūwazzan tiyammin ‘broad (?) earth’), urazza- ‘great, big’ (nom. sg. c. urazzaš DUTU-az ‘great sun god’), ārrazza- ‘?’ (epithet of a sheep).

I published a paper on this topic in the proceedings of the 32nd Annual UCLA Indo-European Conference:

2022 “Greek adjectives in -ης (-ᾱς). An overlooked type?” In David M. Goldstein, Stephanie W. Jamison, and Brent Vine (eds.). 2022. Proceedings of the 32nd Annual UCLA Indo-European Conference. Hamburg: Buske. 125–42. [download]

Odysseus (middle), accompanied by Eurylochus and Perimedes, encounters Tiresias (as a head on the lower left) in the underworld; scene from the 11th book of the Odyssey.

The name Τειρεσίας is also a first declension masculine noun.

Venetus A, page 143 verso.

The phrase ἀργεστᾶο Νότοιο ‘of the white South Wind’ (Il. 11.306) in the Venetus A manuscript with the explanatory interlinear gloss τοῦ λευκονότου 'of the White-Notus'.

Venetus A, page 122 verso.

The notorious phrase χλούνην σῦν (Il. 9.539) that might also contain a first declension adjective χλούνης 'male (?), fierce (?)'.

While working on this topic, I realized that it is often *very* difficult to draw a clear line between substantives and adjectives. What makes an adjective an adjective? And at what point does a substantive lose its substantival status and become an adjective? I found it useful to adopt the following indicators for establishing the adjectival nature of a certain formations in Ancient Greek, illustrated by way of the former agent noun, now adjective ὑβριστής 'wanton'.

a)  adnominal use: [+adnom]

Adnominal use as attributes of substantives is typical for adjectives. To be sure, such usage is also possible for many substantives (viz. as ‘appositions’), as shown by, e.g., αἰπόλοι ἄνδρες ‘goatherds’ (literally ‘goatherd men’), but this practice is only common for certain substantives and therefore more limited than for the average adjective. In addition, any given substantive typically occurs more often in non-appositional than in appositional use. If an appositional use becomes predominant, this might be a sign of adjectivization. For ὑβριστής compare παῖδας ὑβριστάς ‘violent boys’ (Hdt. 3.32).


b) used in syntactic coordination with other adjectives: [+coord]

When a word is used in coordination with other prototypical adjectives, it becomes attractive on syntactic grounds to interpret this word as an adjective, too. Compare ἦ ῥ’ οἵ γ’ ὑβρισταί τε καὶ ἄγριοι οὐδὲ δίκαιοι; ‘Are they cruel, and wild, and unjust?’ (Od. 6.120 = 9.175 = 13.201).


c) used with collective or non-human head nouns: [+coll]/[+non-hum]

Since masculine nouns in ‑ης almost exclusively refer to individual male persons, the use of such a formation in a syntagma with a collective or a non-human head noun is an argument in favor of seeing it as an adjective. Compare στρατὸν ὑβριστὴν Μήδων ‘the aggressive Median army’ (Thgn. 775), or ὑβριστὴς οἶνος διὰ νεότητα ‘wine so new that it’s an insult’ (Ael. Ep. 8).


d) governs an accusativus graecus: [+acc-graec]

Only verbs, participles, and adjectives can govern an accusative of respect (accusativus graecus). Compare διαφέρει γυνὴ ἀνδρὸς τὴν φύσιν ‘a woman is by nature completely different from a man’ (P. R. 453b), or πόδας ὠκὺς Ἀχιλλεύς ‘swift-footed Achilles’ = ‘swift with respect to his feet’ (Il.). For ὑβριστής compare Πέρσαι φύσιν ἐόντες ὑβρισταὶ ‘The Persians are violent by nature’ (Hdt. 1.89).


e) specified by adverbs: [+adv]

While only substantives can govern genitives (e.g., τὸν πυρὸς κλέπτην ‘the fire-thief’ A. Pr. 946), it is a prerogative of adjectives to be specified by adverbs (e.g., ἄγαν βαρύς ‘very heavily’ A. Pers. 515). For our word compare ἀλλ᾿ ὅτι λίαν ὑβριστὴς καὶ βίαιος; ‘That I am grossly insolent and savage?’ (Lys. 24.35).


f) comparison: [+comp]

The formation of a comparative and a superlative is usually seen as a typical feature of adjectives. Our ὑβριστής forms a well-attested comparative ὑβριστότερος (Hdt., Xen., Plat.) and superlative ὑβριστότατος (Ar., Xen., Plat.). However, on the one hand not every genuine adjective has gradation forms (mostly for semantic reasons), while on the other hand, gradation forms are sometimes also attested for substantives, as evidenced by βασιλεύτερος ‘more kingly’ (Il., Od., Tyrt.), βασιλεύτατος ‘most kingly’ (Il.) from βασιλεύς ‘king’. The creation of  these forms certainly became possible by the fact that βασιλεύς was used frequently as an apposition (e.g., Ἀλεξάνδρῳ βασιλῆϊ ‘for king Alexander’ Il. 4.96).


g) gendered agreement forms: [+gend]

Adjectives typically agree with their head noun in case, number, and gender, which implies that they can appear in all three genders. However, this is not true for all adjectives, as many are not used in the neuter (again, mostly for semantic reasons). Words in -της were inherently masculine and were therefore qualified to serve as masculine adjectival forms only. Accordingly, one had to come up with different strategies if one wanted to use them adjectivally in the neuter or feminine. In rare cases, writers used the masculine form also for feminine noun (e.g., τῆς πατροφόντου μητρός ‘of the mother who has killed your father’ Soph. Trach. 1125). Usually, however, the feminine was derived by adding -ις , -ιδος to the (η-less) base. Compare ὕβριστις f. (EM 595.39), κλέπτις f. ‘thievish’ ([+adnom], [+non-hum] in τὴν κλέπτιν ἀλώπεκα ‘the thieving fox’ Alciphr. 3.22). Since nouns in -της are (for the most part) agent nouns, it is understandable that their use for neuter head nouns was not too common. However, we do find a neuter form ὕβριστον quite early (ὕβριστον ἔργον ‘an outrageous deed’ [Pherecr. 162 = 173 Storey]; 5th c. bce). A remaining option was to use a different derivative altogether, such as ὑβριστικός (Att., Arist., etc.), which is attested in all three genders.


h) oxytonesis: [+oxy]

In many cases, adjectives and substantives whose origin lies in a nominalization of an adjective differ in the position of the accent. While adjectives tend to be oxytone (e.g., λευκός ‘white’, κυφός ‘curved’), many substantives and substantivizations exhibit a nominalizing accent retraction (e.g., λεῦκος m. ‘a fish’, κύφων m. ‘crooked piece of wood’). This is especially true for substantivizations in -ης that are almost exclusively barytone (cf. Leukart 1994:132 and see below). Oxytone accentuation could therefore be indicative of a potential adjectivization. However, agent nouns in -τής are evidently influenced by oxytone agent nouns in -τήρ (see Fraenkel 1910–2 I:1–5), so the accentuation of a single item should not be given too much importance.


i) semantic breadth: [+sem]                                         

The adjectivization of a substantive typically goes hand in hand with an extension of its original meaning. This is evident from ὑβριστής, which no longer only means ‘violent person’ but can refer to armies (στρατὸν ὑβριστὴν from above), animals (ταῦροι δ᾿ ὑβρισταὶ ‘rogue bulls’ E. Ba. 743), and even wine that is intense in taste (see above). Conversely, a substantivization of an adjective often has a narrower meaning than the underlying adjective as it only retains one aspect of the original polysemy. Compare λευκός ‘white, light, clear, distinct, happy’ and λεῦκος ‘a fish’ (named after the color) or κυφός ‘bent forwards, stooping, hunchbacked, curved’ and κύφων ‘crooked piece of wood’.

Archaism and Innovation in the Tocharian Nominal System 

In December 2021, I started a new project as the recipient of an APART-GSK fellowship of the Austrian Academy of Sciences (ÖAW) at the Department of Linguistics of the University of Vienna. I will be looking into the Tocharian nominal system and investigate 5 features, based on which a reassessment of our understanding of Proto-Indo-European nominal morphology might be necessary.

Homeric κνίση 

 In June 2022, I attended my first physical conference since the beginning of the pandemic. How exciting! I traveled to Madrid for the International Colloquium of Ancient Greek Linguistics, Madrid, June 16–18, 2022.  My topic was the fascinating word κνίση. [handout]

The noun κνίση is an epic and poetic word that appears 17 times in the Iliad, the Odyssey, and the Homeric Hymn to Apollo. It is also used—mostly in its unepic variant κνῖσα—by Aeschylus, Aristophanes, and Aristotle, as well as by a number of late and Hellenistic authors.

The surprising thing about κνίση is its twofold meaning. In about the half of its Homeric and the majority of post-Homeric attestations, the word refers to the smell or savor of a burnt sacrifice, the steam and odor of fat that exhales from roasting meat, or the odor of savory meat in general.

     (1) ἕρδον δ᾿ Ἀπόλλωνι τεληέσσας ἑκατόμβας

ταύρων ἠδ᾿ αἰγῶν παρὰ θῖν᾿ ἁλὸς ἀτρυγέτοιο·

κνίση δ᾿ οὐρανὸν ἷκεν ἑλισσομένη περὶ καπνῷ.               (Il. 1.315–7)

And they offered to Apollo perfect hecatombs

of bulls and goats by the shore of the unresting sea;

and the savor of it went up to heaven, eddying amid the smoke.’

However, in roughly the other half of Homeric attestations and the odd post-Homeric instance, κνίση has a quite different meaning. In these cases, the noun refers to caul fat, also known as lace fat or fat netting, which is the thin membrane that surrounds the internal organs of cows, sheep, and pigs, and, by extension, to animal fat in general.

     (2) μηρούς τ᾿ ἐξέταμον κατά τε κνίσῃ ἐκάλυψαν

δίπτυχα ποιήσαντες, ἐπ᾿ αὐτῶν δ᾿ ὠμοθέτησαν.            (Od. 12.360–1)

They cut out the thigh bones and covered them with a double layer of fat

and laid the raw bits upon them.’

The key to the understanding of this double meaning is that this caul fat, in which the sacrificial bones would be wrapped prior to their burning, was largely responsible for the steam and odor of the burnt offering as the fat would melt in the blistering heat and trickle down into the fire in sizzling drops. Thus, κνίση is the word for both the savor of the smoldering sacrifice and for that, which causes it. This semantic dichotomy is not only quite astonishing; it also calls for an explanation.

This paper sets out to find a rationale for the two meanings of κνίση. Departing from an etymological and formal analysis of the word, whose connection with Lat. nīdor, -ōris m. ‘vapor, steam, smell (from anything boiled, roasted, burned)’ < *kníhₓd-ōs has long been recognized, it will be argued that κνίση preserves the twofold reading of a denominal adjective *knihₓd-s-ó-, viz. on the one hand in a literal meaning ‘full of steam, savor’ (cf. κνισός in this meaning in Athenaeus) and on the other hand in a causative meaning ‘causing steam, savor’ (leading to κνίση ‘caul fat’). This situation is nicely comparable with the double meaning of adjectives like δακρυόεις ‘full of tears, weeping’ (cf. δακρυόεσσα … κούρη Il. 21.506) and ‘causing tears’ (cf. πόλεμον … δακρυόεντα Il. 5.737) and fits well into a broader picture of the different possible readings of denominal ‘possessive’ adjectives. 

Of beasts and men – The Animals of the Proto-Indo-Europeans

From September 2019 until August 2021, I conducted a research project on Indo-European words for animals, their etymology and word formation, at the Department of Nordic Studies and Linguistics at the University of Copenhagen.

The project was funded as part of the European Union’s Horizon 2020 research and innovation program under the Marie Skłodowska-Curie grant agreement No. H2020-MSCA-IF-2018-835954.

Here is a presentation on the progress and the results of the project:

The Animals of the Indo-Europeans.pdf

My project also had a – quite successful – Twitter account where I posted (and still post) threads on various animal terms in Indo-European. Please follow me here:

Unmetrical Verses in Homer

Unmetrical verses in the epic tradition have been a main focus of my interest for many years now. I guess all historical linguists have a penchant for exceptions and irregularities, and I am no exception. The exceptionally interesting thing about unmetrical verses, however, is that they provide insight not only into the prehistory of the epic tradition, but also into the wondrous workings of oral poetry, and sometimes even into the transmission of the epic texts after they had been written down.

Given the overall number of lines in the Homeric poems, the amount of verses that are considered unmetrical (i.e., that cannot be read as a flawless hexameter) is very low. There are several different sources of unmetricality; some are rooted at very early stages of the poetic tradition, some, on the other hand, on their very end and even beyond. 

Textual Corruption

In some cases, the unmetrical nature of a line is simply due to a corruption of the text. This is quite common in many of the less trustworthy medieval codices, but in some cases—if happened early enough—such a corruption might have become the unequivocally transmitted reading in our Vulgate. One such case is Il. 18.458 υἱεῖ ἐμῷ ὠκυμόρῳ δόμεν ἀσπίδα καὶ τρυφάλειαν with a ludicrous sequence υἱεῖ ἐμῷ ὠκυμόρῳ. It is quite conceivable, though, that the original text, as proposed by Nauck, had υἷι μοι ὠκυμόρῳ and that a copyist, still mindful of the line initial υἱεῖ ἐμῷ δόμεναι just some 300 verses earlier in the same book (Il. 18.144), assimilated the phrase by analogy. 

Quandoque bonus dormitat Homerus

Even if "Homer" was an excellent poet, there are verses whose unmetricality needs to be ascribed to the composer(s) of the Iliad and/or the Odyssey himself. These unmetricalities are often a consequence of the nature of oral composition. They include cases in which established formulas are inflected and transposed into other case forms but kept in their metrical slot (e.g. Il. 18.288 |7 μέροπε̄ς ἄνθρωποι# with a brevis in longo based on the formulaic |7 μερόπων ἀνθρώπων# 7× Il.), or when two hemistichs or formulary templates are put together that do not fit seamlessly (e.g., the formulary template #–⏕ ἐς λιμένα |5 scanning perfectly at Od. 10.87 #ἔνθ’ ἐπεὶ ἐς λιμένα κλυτόν … but resulting in in a (syllaba) brevis in (elemento) longo at Od. 10.141 ναύλοχον ἐς λιμένα, καί τις θεὸς ἡγεμόνευεν). 

Language Change

A number of cases of unmetrical behavior can be explained by language change. A prime example is the consonantal effect of a no longer existing digamma. An unmetrical line end such as Il. 13.573 ὣς ὃ τυπεὶς ἤσπαιρε μίνυνθά περ, οὔ τι μάλα δήν# becomes a perfect hexameter once we restore the etymologically guaranteed digamma in δήν < *dwā́n, thus lengthening the second syllable of μάλα.

The study of unmetrical verses can therefore potentially open up a window not only into the fascinating prehistory of the hexameter and the epic tradition, but also into the written transmission of the texts. In addition, it can help understand the intriguing practices of oral poetry.

One of the most fascinating and perhaps the most unmetrical verse in the Iliad is 

Μηριόνης (τ’) ἀτάλαντος Ἐνῡαλίῳ ἀνδρειφόντῃ (Il. 2.651; 7.166; 8.264; 17.259)

'Meriones, the peer of Enyalius, slayer of men.'

with three heavy syllables in the 5th foot (or two heavy and two light syllables if ἀνδρειφόντῃ is read with -εϊ-).

In this paper, I argue that the form ἀνδρειφόντῃ arose only within the written transmission of the text, and that "Homer" probably sang ...Ἐνῡαλίῳ ἀνδριφόντῃ (a form that is attested as a varia lectio in several manuscripts and in an ancient papyrus).

Geneva, Bibliothèque publique, Ms. gr. 44 (Homer, Iliad with scholia and an interlinear paraphrase of Books I to XII) , page 294.

In the fourth line from above, the verse-final word is the epithet ἀνδρϊφόντη.

Escorial Υ 1.1 (293) page 103 recto.

This manuscript has ἀνδριφόντηι (Il. 8.264; a second hand changed the ι into ε and inserted an ι).

Another famous unmetrical verse is Il. 14.78:

…, εἰς ὅ κεν ἔλθ

νὺξ ἀβρότη, ἢν καὶ τῇ ἀπόσχωνται πολέμοιο
Τρῶες, … 

'…until the immortal Night comes down, if the
Trojans will give over fighting for Night’s sake…'

On the one hand, ἀβρότη scanning ⏑⏑– with a short first syllble, stands in contrast to all other instances of the adjectives ἄμβροτος and ἀμβρόσιος where the first syllable scans long. On the other hand, the feminine ending -η, too, is quite unexpected in a compound, with νὺξ ἄμβροτος (e.g. Od. 11.330) being precisely the form we would expect. Thirdly, the hiatus before ἤν is quite awkward, making this verse in its entirety a rather bumpy one.

You can read my take on the problem in this paper.