Albanian presents in -oj and -uaj

Milan Lopuhaä1

1 Introduction

One of the more prominent present classes in modern Albanian is that of the nasal
present in -oj, which goes back to Old Albanian /-o /. This conjugational type must have
been productive early, as it is the regular class into which Romance verbs are borrowed, e.g.
pëshkoj to fish ← Lat. .2 As is described by Genesin [2005b], the most probable origin
of this conjugation is from a combination of a causative or factitive nasal suffix and the
factitive or denominative suffix *-eh2-, so that we may reconstruct pseudo-PIE3 *-eh2-n-ie/o-.
This would then develop into Proto-Albanian4 *-ånje/a-, where å denotes the product of late
PIE and that developed into Albanian o.
There is a sound law stating that before a resonant stressed Proto-Albanian *å in a
final syllable gets lenghtened and subsequently diphtongised to OAlb. uo, Geg ue, Tosk ua. A
good example for this is duaj sheaf , cognate to Skt. grain , dh yá grain and Lith.
dúona bread , for which we may reconstruct PIE *dheh2n-ió-. This diphthongisation did not
take place in the verbs in -oj, where one would have expected -uaj. Genesin [2005b] writes
that the absence of this sound law in this conjugational class does not need to be explained,
because the ending does not reflect an archaic situation, but is the result of restructuring in
the course of the history of the language. This argument is methodologically unsound: the
conjugational class must have come to being somewhere in the history of Albanian, and
from that moment on it will undergo all sound laws for which the phonological
requirements are satisfied. Since this conjugational class predates the earliest period of
borrowing from Romance, and the sound law must have occured after the loss of word-final
unstressed syllables that also took place in Romance loans, an explanation must be given as
to why this lengthening did not take place in verbs in -oj.
Although there is a productive class of verbs in -oj, there is also set of about fifteen
verbs ending in -uaj, such as shkruaj to write . There are now two problems to be solved: we
need to explain why the productive class of verbs show the ending -oj instead of **-uaj, and
we need to find an etymology for the verbs in -uaj that explains the occurence of a diphthong

1 This paper is based on a BA thesis [Lopuhaä, 2012] written at Leiden University under the supervision of De
Vaan, who has provided valuable input.
2 In this article the symbol ← will be used to denote loanwords.

3 The term pseudo-PIE´ is used here because it is not clear whether this concatenation of suffixes was used as a

present formation in PIE already. For a possible cognate in Vedic verbs in -anyá- and Hittite verbs in - -, see
[Oettinger, 1992].
4 Because the exact point of divergence of the different Albanian dialects is difficult to establish, the language

stage in the prehistory of Albanian that is used as Proto-Albanian in this article is the stage of which the
reconstruction sheds the most light on the prehistory of the verbs ending in -uaj and -yej. In concrete terms I have
chosen the stage directly after the development of the inherited difference in vowel quantity (*a vs ) to a
difference in quality (*a vs *å, which develops into the opposition of a vs o). Since some Latin loanwords share this
development, e.g. mollë apple ← Lat. malum id. , while others do not, e.g. shtratë bed ← spread , my
Proto-Albanian would be dated around the Roman era. It would then be placed slightly after Early Proto-
Albanian as reconstructed by Orel [2000], and around Late Proto-Albanian as reconstructed by Schumacher and
Matzinger [2013].
in these verbs.
There are two more conjugational classes that are relevant for this discussion. First, there
is a class of verbs ending in -ej, which goes back to pseudo-PIE suffix concatenation *-i-eh2-n-
ie/o-. In these verbs one would expect a diphthongisation *e > **ye; furthermore, there is a
limited class of verbs ending in -yej. The relation between the classes ending in -ej and -yej is
then exactly parallel to the relation between the classes ending in -oj and -uaj.

2 Verbs in -oj and -ej

In this section I aim to explain the absence of the lengthening in the verbs terminating
in -oj and -ej . A possible explanation is given in [Genesin, 2005b, pp. 179], namely that the
ending in the 2/3sg., that would reflect quasi-PIE *-eh2-n-i-s/t , would not have a syllable-final
nasal, because the final syllable would not have been completely lost; hence no lengthening
would have taken place. This, however, seems improbable, as the development PAlb. *dånja >
duaj sheaf shows that the lengthening of stressed vowels before resonants took place after
the loss of word-final vowels. Although it is possible that the loss of final unstressed vowel
was a process that happened in multiple stages, so that there might be a stage at which the
final vowel *a was lost whereas the final *i was retained, there is a priori no reason to
suppose such a staged development. It is preferable to give an explanation which only relies
on the Proto-Albanian reconstruction of the present paradigm. Such a reconstruction is given
by Orel [2000] for the consonant stems as follows:

*-a *-ame
*-e(s) *-es ny/jy
*-et *-anti

We can reconstruct approximately the same endings for verbs showing the present
suffix -nj- > -j-. The 2/3sg. in -on, however, shows that the PIE *i in -eh2-n-ie/o- cannot have
been consonantal, for we then would have expected a palatal rather than a dental nasal. One
way to solve this is to assume a development - (s) < - (s). If we assume such a
development we can reconstruct the following:

-oj < - -ojmë < -

-on < - (s) -oni < - /jy
-on < - (t) -onjë < -

After the reduction of final syllables, but before the lengthening of stressed *å, *ö to * ,
(which develop into ua, ye) before resonants, this would yield the following paradigm:

- -
- - -
- -
 Since the nasal is syllable-final in the singular, and not in the plural, we expect
lengthening in the singular but not in the plural, so that an opposition - vs - come
into being. This opposition created asymmetry in the sys-tem of present conjugations: the
nasal presents in vowels other than * or *ö did not have any difference in suffix vowel length
between singular and plural. It then seems probable that Albanian generalised one of the two
vowel lengths in order to maintain symmetry in the system.
The question that now remains is why the short vowel from the plural is generalised
instead of the long vowel from the singular. The reason is that the short vowel * was also
present in other parts of the verbal paradigm that lack the present suffix *-nj-, such as the
impt. 2sg. -o, aor. sg. -ova5. This caused the short vowel to be considered more characteristic
of the verbal paradigm, so that it re-placed the long vowel of the singular rather than the
other way around. Similarly, the verbal conjugation with 1sg -ej, 1pl -ejmë goes back to an
earlier opposition *- ń vs *-ö . Here too the short vowel was more prominent in the verbal
con-jugation and replaced the long vowel in the singular.
Two problems with the 2/3sg. ending -on remain to be discussed. The first is the
absence of the shift of intervocalic *n to r characteristic of Tosk. This also took place in loans,
such as kërp hemp Lat. cannabis (Gh. kanp), so it must have happened before the reduction of
final syllables; hence would expect the 2/3sg. of the verbs in -oj to be **-or < - (s)/t < - (s)/t
in Tosk. The actual forms can, however, be explained by a simple analogy. Apart from the
verbs in *-n-j-, Proto-Albanian had various j-presents, like mas to measure < *matja. At this
stage, the intervocalic *n in the verbs in *-n-j- was restored to keep its structure as a j-present,
of which an example is given in the following table.

*matja -
*mati(s) - (s) ≪6 *- (s) < *- (s)
*matit - ≪ *- < *-

This analogy has taken place in all nasal presents, such as bëj to make , / sg. bën (Gh. bâj,
ban) < PAlb. *banj-.
It still needs to be explained why there is a nasal consonant in the ending -on, as we
expect final n to drop after a stressed vowel, as in the word To. gjë, Gh. thing , which
reflects the PIE participle *h1s-ont-s > PAlb. *sants. Another example is the Latin loan
Latin , borrowed as PAlb. *latina, which is reflected as To. lëti, Gh. Italian, Latin,
Catholic . Based on these examples we would expect an ending **-o rather than -on. Again
this can be explained by analogy. The class of verbs ending in unstressed -ënj, discussed in
more detail in the next section, originally had an unstressed vowel in the ending -ën. Here the
nasal was retained, as it was only lossed after stressed vowels. From here this nasal was
restored in the 2/3sg. ending of the other nasal verbs, probably because these would
otherwise stand out in the Pre-Old Albanian paradigm for being the only endings without
any nasal consonant.

5 We find ua in the aorist plural, for example aor. 1pl. -uam, but this seems to go back to - -m, probably via an
intermediate stage - .
6 I will use the symbol ≪ to denote analogical restoration.
 I conclude this section by giving a complete overview of the developments of the
verbs in -oj and -ej from pseudo-PIE to modern Albanian. To keep it all on one page I only
give the new, restored form for all analogical developments, and not the old form that was
removed from the system by analogy.

Alb. PAlb. Pseudo-PIE

-oj ≪ - < - < *-eh2-n-io-m
-on ≪ - < - (s) ≪ *-eh2-n-ie-s
-on ≪ - < - ≪ *-eh2-n-ie-t
-ojmë < *- < - < - < *-eh2-n-io-me
-oni < - < - ≪ *-eh2-n-i(e)-te
-onjë < *- < - < - < *-eh2-n-io-nti
-ej ≪ *-ö < - < *-i-eh2-n-io-m
-en ≪ *-ön < - (s) ≪ *-i-eh2-n-ie-s
-en ≪ *-ön < - ≪ *-i-eh2-n-ie-t
-ejmë < *- < *-ö < - < *-i-eh2-n-io-me
-eni < *-öni < - (s) ny ≪ *-i-eh2-n-i(e)-te
-ejnë < *- < *-ö < - < *-i-eh2-n-io-nti

3 Verbs in -uaj and -yej

In this section the origin of the verbs in -uaj and -yej will be discussed. It is important
to note that the diphthongs in these verbs cannot be the result of the lengthening of the
vowels and *ö in front of word-final resonants, because the results of this lengthening were
again removed by analogy, as was seen in the preceding section. The rise of these
diphthongs then requires a different explanation. Let us first look at all verbs that need to be
taken into consideration. In Albanian the following verbs ending in -uaj and -yej exist:

bluaj to grind fyej to offend

druaj to dread gërryej to scratch
gatuaj to prepare kryej to reach
gjuaj to hunt lyej to smear
kruaj to scratch ngjyej to paint
luaj to play ryej to suffer
paguaj to pay rrëfej, Buz. rëfyenj to tell
pruaj to bring shqyej to break
quaj to call thyej to break
ruaj to guard vyej to be worth
ruaj to plunder yej to wake
rruaj to shave zhyej to smear
shkruaj to write
shuaj to extinguish
truaj to dedicate
vuaj to suffer
 This list is rather limited, and all words will be discussed. Let us start with one of the
words with the clearest etymologies, namely shkruaj to write , which is a loan from Latin
id. [Meyer, ]. This would yield Proto-Albanian -, as Romance loans were
incorporated in the Proto-Albanian conjugation in - nj- [Genesin, 2005b]. Here I reconstruct
an approximant *w rather than a stop *b as the reflex of Latin b, since Demiraj [2001] has
shown that inherited (possibly aspirated) voiced stops are not lost intervocalically whereas
borrowed Latin b, d, but not g, are lost in this position. Therefore it is best to assume that
shkruaj was actually borrowed from a variety of Romance in which the intervocalic stop was
already lenited into a fricative β, which we may represent in Proto-Albanian with *w.
Given the form - we can now explain the diphthong in shkruaj as a contraction
> > ua. More explicitely, we would have a development -> -> ->
shkruaj. Now we need to know what the chrono-logical relation is between the development
, > and the analogical restoration of - in the verbs in -oj. There are two possibilities:

1. The contraction , > took place before the analogical restoration. In this case we
must assume that the analogy did not take place in verbs in -uaj, because they had a
long vowel in the entire present paradigm, so there was no basis for the analogical
restoration.

2. The contraction took place after the analogical restoration. In this case the contraction
takes place in all forms of the present, which explains the presence of ua in the entire
paradigm.

I strongly favour the second explanation. In the first scenario there would have been a
class of nasal presents having the vowel in all forms in the present paradigm. If that were
the case, there would be little reason why the restructuring of verbs with an alternation
between *- and - would result in a short vowel, since this required the creation of a new
type of present, as opposed to generalising the long vowel. Taking the first explanation
would then undermine our explanation for the origin of the verbs in -oj, whereas the second
explanation poses no such problems, so it is the most straightforward to just assume this one.
Based upon these developments we can posit the following relative chronology of sound
laws:

1. Replacement of quantity opposition by quality opposition in vowels PIE > PAlb.

, PIE > PAlb. *ö.
2. Reduction (and loss) of final syllables.
3. Lengthening of stressed ,*ö before syllable final resonants.
4. Analogical restoration of the short vowels and *ö in nasal presents.
5. Contraction of and to * , and of *ëö and *öë to .
6. Diphthongisation of to ua and to ye; shifts > o and *ö > e.

3.1 Verbs in -uaj

What we now have found is that the diphthongs ua and ye can be the result of a
contraction caused by the loss of intervocalic consonants. As noted before, Latin b and d are
lost intervocalically. The Proto-Albanian approximant *w (< PIE ) is also lost in this position
[Orel, 2000, p. 18], as is *j (< PIE ) after front vowels Kortlandt [1996]. It is also possible that
PIE *s is lost intervocalically, but this is uncertain: [Orel, 2000, pp. 61-62] gives a number of
examples that point to a development to h/∅ normally, with a RUKI variant sh. The number
of examples, however, is rather limited, and counterexamples to this development exist as
well [Demiraj, 1990]. Using these contractions we can now discuss the etymologies of the
various verbs in -uaj; the verbs in -yej will be discussed later, along with possible origins for
the palatalisation in these verbs.

3.1.1 Latin loans

Except for shkruaj there are several more latin loans which end up as verbs in -uaj in
modern Albanian. The following etymologies are given by Orel [1998]:

Alb. PAlb. Lat.

shkruaj to write < < *skriw nj- ← id.
vuaj to suffer < *w < *w - ← to write
truaj to dedicate < *t < *trad - ← tradere to hand over
rruaj to shave < *r < *rrad - ← id.

Some comments need to be made about this table. First of all, the verbs shkruaj and
vuaj show that a pretonic vowel *i must have been reduced to *ë before the contraction,
because we would otherwise expect an i-mutated result like > **shkryej. There must also
have been some kind of semantical development in vuaj, probably from experiencing
[pain/disease] to suffering [pain/disease] . Finally we must add that Orel rejects the
etymology as a Latin loan, due to Meyer [1891], and prefers to reconstruct Proto-Albanian
-nj- for rruaj, giving Greek ῤήγνυμι to tear as a cognate, and positing a development
-> . Since in Albanian PIE *-g- is not lost intervocalically, this etymology is to be
rejected. Furthermore, there is no reason to regard the etymology as a Latin loan.
There are two more words for which we can find a Latin etymology. The first of these
is shuaj to extinguish . Orel considers this to be a loan from Lat. to subjugate .
However, this is impossible, as we do not expect the intervocalic Latin g to be lost in
Albanian. Other suggested etymologies are borrowings from Greek σβέννυμι to quench
[Camarda, 1864], Latin exungere to anoint [Meyer, 1891] and Greek ψάω to rub [Mann,
1950], but all of these have phonological problems. I propose as an alternative a borrowing
from Latin to calm , which could be used in a construction fl to
extinguish fire . Furthermore, when borrowed into Albanian it would yield precisely shuaj.
Another possibility is that the Albanian word is onomatopoeic. This, however, does not
exclude a Latin borrowing; it is possible that there were multiple ways to express I
extinguish the flame in an earlier stage of Albanian, and that became the most
prominent one because the sound matches the action.
The word pruaj has two meanings to guard and to plunder . These contradictory
meanings are difficult to explain from a single meaning, so it is best to assume we are
dealing with two homophones with unrelated etymologies. The verb pruaj to plunder
admits a straightforward derivation from Latin deponent verb to plunder , possibly
via an intermediate Romance form . This would be borrowed into Proto-Albanian
- which would yield pruaj. The word pruaj to defend is more problematic. Meyer
[1891] is unsure whether it is a derivation of ruaj to guard or a loan from Lat. to
prevent . Neither solution is very attractive we would either be left with an unexplained
suffix p- or a semantic development from to prevent to to defend . A better solution is
given by Orel [1998], who cites this word as proj and considers this to be a loan from Latin
to make ready, to defend . We then have to explain how the word pruaj to plunder
came into being. The answer to this lies in the aorist form prova of proj. To this aorist a present
form pruaj to defend was created on the basis of verbs as rruaj to shave , aor. rrova. The
creation of this form was a contamination of the already existing pruaj to plunder and the
aorist proj. The new verb pruaj to defend subsequently took over the aorist formation pruajta
from pruaj to plunder .

3.1.2 gatuaj to prepare

According to Orel [1998] this word is to be connected with Slavic *gotovati, which has the
same meaning.7 [Orel, 2000, p. 38] states that an unstressed Slavic *o is regularly reflected as a
in Albanian, as in matukë hoe ← Slav. *motyka and karrup fish trap ← Slav. ъ. After the
loss of the intervocalic *-v- the vowels would contract to give the expected diphthong; the
same development can be seen in patkua horseshoe ← Slav. ъ ъ.
An explanation needs to be given, though, as to why this verb has become a nasal
present in Albanian. Slavic verbs are mostly borrowed as verbs in *-tj-, like buças to thunder
← Slav. and grabis to steal ← Slav. *grabiti; we would then expect **gatuas, 2/3sg.
**gatuat. According to Orel this is because the borrowing is old; another example of this is
pëgëj to smear Slav. *poganiti. However, this is not a direct parallel, as this word had a nasal
in Slavic already. It is possible that *gatuas entered the class of verbs in -uaj after the
contraction because of the diphthong in the final syllable. Another possibility, given by Jokl
[1931], is that the verb is a denominative from Slavic ъ ready . The Proto-Albanian verb
would then be - from PAlb. *gatowa-. However, this word is not attested as such in
Albanian. There is gat ready , which Orel sees as a secondary form derived from gatuaj. At
any rate, a Proto-Albanian reconstruction - seems probable, which indeed develops
regularly into gatuaj.

3.1.3 paguaj to pay

For this word a borrowing from Italian pagare seems probable, but this regularly
yields pagoj, which in fact occurs as an alternative form. It is unclear where the diphthong
comes from in this word. One possibility is that the form paguaj is derived from the Old
Italian 3sg. preterite in -áo (Sicilian -au), so that the Proto-Albanian form would be -.
However, it is unclear why this present would be derived from the Old Italian preterite.

7 Camarda [1864] connects this word to Greek αγαθός good . However, a Proto-Albanian form - would yiel
**gatoj.
Another explanation is to assume that pagoj was the original form, and that paguaj was
created analogically from the aorist pagova. The reason why this form was created, however,
is still unclear.

3.1.4 gjuaj to hunt

Within Albanian this word is connected to the noun gjah hunt . The etymology of
both words is uncertain. It seems attractive to connect the word to OHG to hunt ,
whose etymology outside Germanic is unknown as well. Here the problem is that an
intervocalic *gh is not lost in Albanian, so we cannot simply reconstruct PIE *iegh- on the basis
of these words. Demiraj [1997] sees a connection with Lat. to perceive keenly and
Goth. to seek and reconstructs *seh2g-, but the this etymology has the same problem. A
reconstruction based on Slav. , ъ to drive away as done by Meyer [ ] would
probably yield **gjëj < PIE *gien-i-, so this has to be rejected as well.
The best solution is to regard gjah as coming from PIE *sh2g-so-, a thematisation of a
heterodynamic s-stem, and to regard gjuaj as a denominalisation of this, from earlier *gjahoj.
The regular outcome of PIE *ks in Albanian is h, as can be seen from, for example, hirrë
whey , connected to Skt milk -. One would then expect an earlier *gs to give the same
outcome. Alternatively, one may re-construct *iogh-so- if one prefers a connection to .
Yet another reconstruction is possible if we follow Camarda [1864] in connecting the
word to Gr. δίζημαι to seek , for which Beekes [ ] reconstructs PIE *ieh2- to pursue we
could then reconstruct the PIE predecessor of gjah as 2-so- or 2- -. The vocalisation might
seem awkward, but the consonantal reflex of the *i might have spread from other forms in
the paradigm with a full grade in the root.

3.1.5 kruaj to scratch

Mann [1950] connects this to Gr. κνάω to scrape , but this would be the only example
for a sound law *kn- > kr-, so it is preferable to look for alternatives. According to Orel [1998]
this word reflects PIE *grebh- to scratch , with an irregular devoicing in anlaut again the loss
of intervocalic *bh is unexpected. By contrast, Genesin [2005b] sees it as a fientive from PIE
*(s)ker- to scratch , so *kr-eh1- > , which would then have gotten the suffix -ënj. This seems
improbable, as the function of the fientive was taken over by the Proto-Albanian suffix - -,
so we would expect this fientive to be restructured as - > **kroj. Alternatively, one could
see it as a borrowing from Germanic, connected with OHG to scratch . This would
be borrowed as PAlb. -, which develops into kruaj. However, it is very well possible
that the cluster kr- arose through onomatopoesis, so that this verb would not be
etymologisable. This would also explain the variation between kruaj and the forms gërruaj and
kruj mentioned by Jokl [1916].

3.1.6 luaj to play

This word is more complicated. Next to this word there is also the word loz with the
same meaning. According to Orel [1998] these words originate in, respectively, PAlb. -
and dj-, and he as cognate Slav. ъ order, peace , so we would have to reconstruct PIE
(h)
*leh2d -. However, we again cannot account for the loss of intervocalic *d(h). For these words
we would also have to assume that they are two different denominatives based upon a word
- that has no reflexes within the attested Albanian (there is a word lodër toy . According
to Meyer [1891] we must regard luaj as a loan from Latin to play , which is
phonologically and semantically perfect. We then still have to account for loz. For this we can
cite Demiraj [2001], who connects loz to lodhem to become tired < *leh1d-, cf. Goth. letan to let ,
Gr. ληδεῖν to become tired , so we can reconstruct loz < *leh1d-ie/o-, and we must assume that
the meaning to play is secondary. These two verbs thus have ultimately unrelated
etymologies.

3.1.7 quaj to call

This word is attested in Buzuku in 3pl. cluognene /kluo ënë/ 8 , which implies a
connection with PIE - to listen , cf. Skt. to hear , Gr. κλέομαι to become famous ,
Lat. to be famous . We would then do best to reconstruct j-, which we can
regard as a denominative from the noun *klewa- < léuos fame , cf. Gr. κλέος, Skt. -, OIr.
clú. Note that both the depalatalisation in the cluster and the adaptation of a PIE s-stem
into a Proto-Albanian a-stem are regular. Demiraj [1997] reconstruct PAlb. - < PIE
-eh1-, but this presumes a laryngeal in the root, for which no other language shows any
sign; therefore I prefer the reconstruction as a denominative. Note that this PIE root is also
reflected in quhem to be called < PIE *klu-ske/o-.

3.1.8 bluaj to grind , druaj to dread

According to Meyer [1891], bluaj is a loan from Latin molere to grind . However, we
would expect this to be borrowed as -, which would yield **bloj. Klingenschmitt [1982]
points to OGh. blo, which he derives from *mlh2-ie-t from the root *melh2- to grind , cf. ON
melja, Lith. málti, OIr. meilid. The 1sg is bluo < *mlh2ioH, on the basis of which the form bluonj was
created, which Klingenschmitt compares to to spread , Buz. perdaa 1sg. / ë /< *dh2ioH
, 3sg. perdah /përda/ < *dh2iet. There are two possible ways how this may have come into being:
either the verb was incorporated into the by then already existing class of verbs in -uonj in
order to lose the otherwise unknown vowel alternation uo/o, or the new conjugation came
into being as the result of a contraction bluon < *blo-ën, where -ën is a present suffix that in
standard Albanian and Tosk is only found in the three verbs eci to come , hipi to mount and
iki to go away . In Old Albanian and in Gheg, however, the suffix is more prevalent; see
[Genesin, 2005a] or [Lopuhaä, 2014] for a more detailed discription of the use and history of
this suffix. The addition of this suffix to vowel stems can be seen in the Arbëresh dialect of
San Costatino Albanese, where the verb mba to have has a suffixed counterpart mbanj to
carry [Breu, ]. Here, we can be sure that the latter verb is the result of suffixation rather
than incorporation into a different verbal paradigm, since its semantics show the
imperfective meaning that is one of the meanings of the suffix -ën in this partic-ular dialect. It

8 This phonological interpretation is due to Fiedler [2004].

is difficult to establish what happened between Old Albanian bluo and modern bluaj. Either
way, a diphthong was present in the Old Albanian 1sg already, where it is the result of a
contraction bluo < *mlh2ioH after the loss of intervocalic PIE *i. In the other forms of the
paradigm, the diphthong either was the result of a contraction of the form bluan < *blo-ën, or it
was introduced from the 1sg.
The word druaj probably underwent a similar development. In Buzuku we find 1sg.
druo /druo/, 3sg. dro /dro/, 2pl. droni /droni/, 3pl. druone(h) /druonë/. Orel [1998] connects this
word to Latv. druvas fear and Lith. ũ to scare off , so that we can reconstruct this word
as PAlb. - < PIE *dreu-. Probably also connected to this root is drojë fear , for which we
then would have to reconstruct PAlb. < pseudo-PIE 2.

3.1.9 ruaj to guard to wait

The final word in this category is ruaj. Various etymologies have been given for this
word. According to Meyer [1891] it is a loan from Slav. *xorniti to keep , but then we would
probably find no diphthongisation. Jokl [1911] connects the word to Gr. ὁράω to see , so that
we would have to reconstruct PAlb. - < PIE *ur-eh2-n-ie/o-. This would give us either
**rroj or *roj, depending on the development of the initial cluster, but not a diphthong.
Genesin [2005b] mentions a i-present to PIE *reh1dh-, cf. OCS raditi to worry , OI imraidi to
think . This would develop as *reh1dhiom > *roz, *reh1dhit > *ro, which would then receive a
present suffix -ënj. This cannot be true as we do not expect intervocalic *dh to drop. Orel
[1998] connects it to ON to care and Gr. ἀρήγω to defend , but reconstructing on this
basis would give the same problem.
I suggest a connection with PIE *h1roh1-u-eh2 rest , cf. Gr. ἐρωή, PGm. > OHD
ruowa, OE row, ON ró [Beekes, 2010]. This would develop into PAlb. , from which a
denominative - to rest . We expect this to develop into ruaj; we would then have to
assume a semantic development from to rest to to wait for and finally to to guard .

3.2 Verbs in -yej

When looking for etymologies of verbs ending in -yej, one might be tempted to
assume that these verbs are the result of a contraction of *-VCönj-. However, the verbs in *-önj-
come from an earlier suffix - -. This means that in a con-struction *-VCönj- < - - the
consonant is not intervocalic, and usually does not drop as a result. For example, Latin
to take a siesta is borrowed as Proto-Albanian - which develops into
Albanian mërzej to rest in the shade rather than **bryej. Also, the example shkruaj < -
shows that a front vowel present in the contraction does not trigger the palatalisation of to
* . Still, there are some specific developments that can lead to a verb in -yej. One of these is a
sound law, stated implicitely in Kortlandt [1996], that states that an intervocalic Proto-
Albanian cluster *lj develops into - - > *-j-. In constructions of the form - - this *j is lost,
yielding the present ending -yej. An example of this is the verb fyej to insult , considered by
Orel to be a denominative of faj guilt, sin . This we can consider to be a loan from Rom.
*fallia, which is to be connected with Lat. fallere to deceive . On the basis of this we can
reconstruct PAlb. *falja- and a denominative -, which then would yield ö > ö >
fyej.
A related word is Buz. fëjenj to sin . This too seems a denominative of faj, but a
semantical develpoment from fëjenj to fyej is unlikely. Also, as can be seen from the Buzuku
verb luegn /lye / to smear , the verbal conjugation in -yej was present in Old Albanian
already. For these reasons we must regard fyej and fëjenj as two similar, but independent,
derivations of the noun faj. In this regard fëjenj must be the later one, that was made after the
loss of intervocalic *j.
The same development took place in vyej to be worth , which we can regard as a loan
from Latin to be worth . This would have been borrowed as PAlb. -, which
indeed develops into vyej. There is also a variant vlej, which we can explain by assuming that
the pretonic *a was lost here before the development *lj > *j; so here we would expect -
> - > ö > vlej.
A different kind of example is to be found in yej to wake . Orel [ ] considers this a
denominative from yll star , but this is semantically not ideal. Rather, I would consider it a
borrowing from a Romance form from Latin 9
to wake . In Romance, the
Latin cluster gl would yield a palatalised , so that we would get a palatalised o after the
loss of intervocalic consonants and contractions.

3.2.1 lyej to smear

Following Orel [1998] this word is best to be considered a loan from Gr. ἔλαιον olive
oil , Dor. ἔλαιϝον. 10 We can assume that this was borrowed as *elajwa-, from which a
denominative - was created; the palatalisation of the diphthong is then due to the *-j-.
Note that the *j was lost relatively late, since an intervocalic *j does not cause palatalisation,
as in bluo < *mlh2ioH.

3.2.2 zhyej to smear

This appears to be a denominative of zhul dirt . This word has no certain etymology,
so we cannot be sure why the intervocalic *l was lost; perhaps it is a de-nominative based
upon a plural *zhuje nonexistent in modern Albanian, so that we could reconstruct u -.
However, this is very speculative if we have no in-formation about the etymology of this
word.

3.2.3 kryej to reach

This seems to be connected to krye head . If we want to reconstruct this to PIE rh2-es-
head 11 (cf. Gr. κρᾱνίον skull , Skt. - head , a sensible reconstruction would be PAlb.
*kröna- < PIE rh2-os-no-, which is comparable to Gr. κάρᾱ head , pl. κάρηνα < rh2-es-n-h2. If

9 Note that Latin to wake would yield an initial **v- when borrowed into Albanian.
10 The alternatives, considering it as a loan from Lat. to anoint [Meyer, ] or from Gr. ἀλείφω to anoint the
skin with oil [Camarda, 6 ], have various phonological problems.
11 see [Beekes, 2010] for more information about this reconstruction.
 this is indeed the correct etymology, the denominative would have to be late, because
the final *-n must have been lost by the time this derivation was made. It is unclear how we
would expect a formation like *krye- to contract, but it is also possible that it was formed by
analogy to other verbs in -yej.

3.2.4 ngjyej to paint and trushkyej to steal from a holy place

The word ngjyej is considered by Orel [1998] to be a loan from Lat. unguere to smear .12
However, a direct loan would yield PAlb. -, which would explain the diphthong nor
the palatalisation. We can solve both of these problems if we assume a Romance derivation
from unguere. A possibility would be to assume a Romance form , which would be
borrowed as PAlb. *ungwalja-, from which a de-nominative - would have been
formed, which we indeed would expect to develop into ngjyej. Unfortunately, we do not have
any Romance sources point-ing to the existence of such a word.
We face similar problems when trying to find an etymology for trushkyej. Orel [1998]
considers this a loan from Rom. (based upon Latin triumphare), but this would
yield **trushkoj in Albanian. Here too we would have to assume a Romance form
,which is undesirably complicated, but I do not see a better solution.

3.2.5 rrëfej to tell

This word is very problematic. It is treated in this section because the form attested in
Buzuku is refuegn /rëfyenj/. Orel [1998] considers this a loan from Lat. referre to tell .
However, we would expect PAlb. *r - based on the Latin 1sg. refero, which would yield
Alb. **rëfroj or something similar; in other words, we would not expect the intervocalic *r to
drop. Çabej [1976], on the other hand, sees it as a loan from Gr. φαίνω to show , after the
monophthongisation of αι to [ɛ]; this, however, leaves the diphthongisation unexplained.
An alternative explanation is given by Genesin [2005b], who reconstructs -θö-ën,
with the root coming from PIE - to point to , cf. Ved. to instruct , OAv. to
teach . The development > f is different from the usual development to th, but this is an
alternation we see more often: the word /uthullë/ vinegar in Buzuku is reflected as /ufullë/ in
Bogdani. There is also the word fle to sleep , which Klingenschmitt [ ] reconstructs as PIE
lei-, cf. Toch. A klis-, B klänts- to sleep . However, the conditions under which *k develops
13

into f are unknown, so the reliability of this etymology can only be ascertained when this has
been sorted out.
Equally speculative would be to assume that this was formed after the forms of Latin
referre without thematic vowel, e.g. 2sg. refers, 3sg. refert. Regularly, a cluster *-rnj- is
simplified to a cluster *-nj- with a lengthening of the preceding vowel. This is also the process
that Orel assumes in ziej to boil , which he reconstructs as - < PIE *gwher-, cf. Arm.

12 Camarda [1864] connects this word to Gr. χέω to smear , but Beekes [ ] reconstructs this as
ǵ eu- on the basis of Skt. hutá- sacrificed , from which we would expect a dh in the Albanian form.
13 According to Rix [ ], the meaning of this root is to lean and many other cognates are given, such as Av.
sraiiata, Gr. κλίνω to lean , Lat. re to bend , OHD to lean , Skt. . The meaning in Tocharian and
Albanian, if this etymology is correct, would then be secondary.
 warmth , Gr. θέρομαι to become warm, to burn , OI fogeir to heat , Skt. háras- glow , where
we should assume a development > > *ziej. For refyenj we should then assume a
development *refernja- > > *rëf , where the rounding of the might have been caused
by the preceding *f. However, this seems unlikely, as bie to bring has a similar development
bie < 14
< *bera- < *bher-, cf. Gr. φέρω to carry , Skt. bhárati, Lat. ferre, but here there is no
rounding of the . It would require more research to determine whether this etymology is
plausible, or one of the other two is to be preferred.

3.2.6 gërryej to scratch

Genesin [2005b] states that the etymology of this word is unclear. Orel [1998]
considers it to be a denominative of gërresë scraper , for which he reconstructs PAlb. .
We would then expect a verb -. However, the -s- < *-tj- does not drop intervocalically,
so another etymology is needed. Genesin notes that there are many similar verbs, all having
the meaning to scratch , like gërvisht, gerdhuc, gërzhit, gërruc, grith and gris, and that these verbs
are prone to onomatopoesis and phonoesthetics, so it might be impossible to trace this word
back to Proto-Albanian, as was the case with kruaj.15

3.2.7 thyej, shqyej to break

The etymology of thyej is unclear as well. Orel [1998] reconstructs -, based on

PIE - to sharpen , cf. Skt. , Arm. sowr sword , Alb. thikë knife . The problem is that we
expect the PIE diphthong to be monophthongised within Proto-Albanian already, probably
to i, since this is the reflex of PIE *ei,16 so this would yield **thij. Genesin [2005b] mentions a
reconstruction PIE - to cut , cf. Skt. , Lat. to miss , but here we would not
expect a palatalisation of the diphthong.
Equally problematic is shqyej. Orel reconstructs PIE *sken- but the oldest attestation is
skluej. For this reason Genesin regards this word as a reflex of PIE *kelh2- to hit , cf. Gr. κλάω
to destroy ,17 Lat. percellere to hit , Lit. kalù to hit , from which we would get a development
analogous to bluaj; however, this would then yield **shquaj, as there would be no reason for
palatalisation.
As an alternative to both I propose to reconstruct * -18 for *kelh2-. This is possible, as
in Balto-Slavic the cluster * l before a non-front vowel; we can then assume the depalatalised
form was generalised in Lithuanian. This depalatalisation also happened in Albanian, as can
be seen from gju knee < PAlb. *gluna < ǵ -no, cf. Gr. γόνυ, Lat. genu, Skt. -. We can derive
both Albanian forms from this root as follows. We may assume that in Albanian, the
alternation between * elH- and * lH- was resolved by a paradigmatic split, yielding two Proto-

14 Klingenschmitt [1982] regards bie as /bje/ and considers the ie a regular development of PALb. *e, so he asumes a
development bie < *be < *ber. However, bie reflects a phonological /bíe/, so it cannot be the result of the sound law je
< *e; see [Orel, 2000, p. 3-7] for more information about this sound law.
15 Jokl [1911] reconstructs PIE *gred-, but we do not expect the intervocalic *d to drop. Mann [1950] connects it to

Gr. χραίνω to stain , but this leaves the Albanian diphthong unexplained.
16 As in dimër winter < ǵheimeno-, cf. Gr. χειμών, Lit. .
17 Beekes [2010] concludes that this root cannot account for the Greeks forms and rejects this etymology.

18 If the Greek evidence is to be rejected we have no evidence for the nature of the laryngeal.
Albanian roots - and *klel-, where the latter is a contamination of *kl- and el-. These were
both made into *j-presents19 - and *klalja-. These were then given the suffix -ënj, which
yielded *thejënj, *klejënj, which then developed into thyej and *klyej; the latter was then given
the prefix sh- < PIE *h1 ǵh(s) out , cf. Gr. ἐκ, ἐξ, Lat. ex.

3.2.8 ryej to suffer

According to Orel [1998], this word reflects PIE *Hreud-, cf. Skt. ródati to lament , Lat.
rudere to roar , Lit. raudà crying , Slav. *rydati to cry . This etymology cannot be correct as the
PIE *d is not expected to drop. The reconstruction *uren- by Çabej [1976] also cannot explain
the Albanian diphthong. I am not able to find a better explanation for this word.

3.3 Overview

I conclude by giving a tabel with reconstructed Proto-Albanian forms for all the
treated verbs, as well as their source, be they inherited words from PIE or loans from other
languages. I use ⊲ to denote derivations from, for example, nouns.

Albanian Proto-Albanian Source

bluaj to grind ⊲ - < PIE *mel2-ie/o-
gjuaj to hunt < *sax nj- ⊲ PIE *sh2g-so-
kryej to fulfill ⊲ *kröna- < PIE *krh2os-no-
quaj to call < *klew nj- ⊲ PIE -os-
ruaj to guard < *r w nj- ⊲ PIE *h1reh1-u-eh2-
shqyej to break < - ≪ PIE *h1 ǵh - -ie/o-
thyej to break ⊲ *drawa- ? < PIE *dreu- ?
< -θö-ën- ? < PIE -?
rrëfej to tell < *wrefenj- ? ← Gr. φαίνω ?
≪ *refernj- ? ← Lat. referre ?
luaj to play < - ← Lat.
pruaj to plunder < - ← Lat.
pruaj to guard < *par - ← Lat.
rruaj to shave < *rad - ← Lat.
shkruaj to write < *skriw - ← Lat.
shuaj to extinguish < *sed - ← Lat.
truaj to dedicate < *trad - ← Lat. tradere
vuaj to suffer < *wiw - ← Lat.
vyej to be worth < *walj - ← Lat.
yej to wake < *ewilj - ← Rom.
fyej to offend < *falj - ← Rom. *fallia
ngjyej to paint < *ungwalj - ? ← Rom. ?

19 This is parallel to OCS klati, 1sg. .

 trushkyej to steal from a holy place < *triumpiskalj -? ← Rom. ?
paguaj to pay < - ← OIt. pagare
gatuaj to prepare < *gatow - ← Slav. *gotovati
lyej to smear < *elajw - ← Gr. ἔλαιϝον
kruaj to scratch < *kraw - ← OHD ?
zhyej to smear ⊲ ? < ?
gërryej to smear < ? < ?
ryej to suffer < ? < ?

4 Conclusion

The aim of this article was to regularise the development of the verbs in -oj and -uaj
(and -ej and -yej). This was done by explaining the absence of diphthongisation in the present
singular of verbs in -oj and -ej from an analogy from the present plural and other parts of the
verbal paradigm. Furthermore, the diphthongs in verbs in -uaj and -yej can almost all be
shown to be the result of either a contraction of two vowels due to the loss of an intervocalic
consonant, such as shkruaj < -, or due to the creation of nasal present verbs from
words end-ing in a diphthong, such as kryej ⊲ krye. This has several consequences. First, new
etymologies for the words shuaj, pruaj to plunder , gjuaj, ruaj, vyej, yej, thyej and shqyej were
found. Another consequence is that we now have a new instrument for finding etymologies
for verbs in -oj, -ej, -uaj and -yej; verbs of the latter two conjugations are the result of either a
contraction or the secondary creation of a nasal present verbs. Conversely, vowel
contractions due to the loss of intervocalic consonants always yield a diphthong ua or ye,
which limits possible etymologies for verbs in -oj and -ej. Some caution needs to be applied
here, though, as we have already seen that verbs ending in -oj might secondarily receive the
stem vowel ua, e.g. pruaj to defend ≪ proj ← Lat. . The reverse development might also
be possible; for example, Orel [2000] regards arnoj to repair as a borrowing from Latin
to renew . This would regularly yield *arnuaj, which might have been reformed to
arnoj on the basis of the aorist arnova. More research is needed to find out which factors
played a role in this change of conjugation that has taken placein some verbs in -uaj. The
question still remains whether the reverse development is a possibility as well.

References

Beekes, Robert Stephen Paul.

2010 Etymological dictionary of Greek. Brill, Leiden. With the assistance of Lucien van Beek.
Breu, Walter.
2008 Aspetto verbale ed aspettualità nel dialetto italoalbanese di San Costantino Albanese. In Omaggio a
Girolamo de Rada: Atti del V Seminario Inter-nazionale di Studi Italo-Albanese (2–5 ottobre 2003), page 93–112,
Rende, Università della Calabria.
Camarda, Demetrio
1864 Saggio di grammatologia comparata sulla lingua albanese. successore di Egisto Vignozzi e C., Livorno.
Eqrem Çabej
1976 Studime gjuhësore, volume 2. Rilindja, Pristina.
Demiraj, Bardhyl
1990 Rreth zhvillimit të spirantes ie. */s/ në gjuhën shqipe. Studime Filologjike, 2:195-202.
1997 Albanische Etymologien: Untersuchungen zum albanischen Erbwortschatz. Rodopi, Amsterdam.
2001 Das Meyersche Gesetz über den Schwund der intervokalischen Media im Albanischen. Münchener
Studien zur Sprachwissenschaft, 61:59-72.
Fiedler, Wilfried
2004 Sistemi foljor i shqipes në veprën e Gjon Buzukut (1555) : u miratua për botim në mbledhjen e Seksionit të
Gjuhësisë dhe të Letërsisë mbajtur më 24.09.2004, në bazë të recensioneve të akademik Rexhep Ismajlit dhe
akademik Idriz Ajetit = Das albanische Verbalsystem in der Sprache des Gjon Buzuku (1555) : approved at the
meeting of the Section of Language and Literature, held on 24.09.2004, on the grounds of the reports of
academician Rexhep Ismajli and academician Idriz Ajeti. Akademia e shkencave dhe e arteve e Kosovës,
Tirana.
Genesin, Monica
2005a Una classe verbale „speciale . In Albanologische und balkanologische Studien: Festschrift für Wilfried Fiedler,
page 31–44, Hamburg, Verlag Dr. Kovac.
2005b Studio sulle formazioni di presente e aoristo del verbo albanese. Centro Editoriale e Librario Università della
Calabria, Rende.
Jokl, Norbert
1911 Studien zur albanesischen Etymologie und Wortbildung. Hölder in Komm., Vienna.
1916 Beiträge zur albanischen Grammatik. Indogermanische Forschungen, 37:90-121.
1931 Zur geschichte des albanischen diphtongs -ua-, -ue-. Indogermanische Forschungen, 49:274--300.
Klingenschmitt, Gert
1982 Das Altarmenische Verbum. Dr. Ludwig Reichter Verlag, Wiesbaden.
Kortlandt, Frederik
1996 Pie *j in Albanian. Dutch Studies in South Slavic and Balkan Linguistics, 23:173-176.
Lopuhaä, Milan
2012 Nasaalpresentia in het Albanees. Unpublished BA thesis in Comparative Indo-European Linguistics,
available at http://media.leidenuniv.nl/legacy/lopuhaa-milan-ba-scriptie.pdf.
2014 Form, function and history of the present suffix -i/-ën in Albanian and its dialects. Unpublished ResMA thesis in
Linguistics, available at https://openaccess.leidenuniv.nl/handle/1887/28648.
Mann, Stuart E.
1950 The Indo-European vowels in Albanian. Language, 26:379-388.
Meyer, Gustav
1891 Etymologisches Wörterbuch der albanesischen Sprache. Verlag Karl J. Trübner, Strassburg.
Oettinger, Norbert
1992 Zu den Verben auf vedisch -anyá- und hethitisch -annie-. Münchener Studien zur Sprachwissenschaft, 53:133-
154.
Orel, Vladimir
1998 Albanian Etymological Dictionary. Brill, Leiden.
2000 A Concise Historical Grammar of the Albanian Language. Brill, Leiden.
Rix, Helmut
2002 Lexikon der indogermanischen Verben: Die Wurzeln und ihre Primärstammbildungen. Dr. Ludwig Reichter
Verlag, Wiesbaden.
Schumacher, Stefan and Matzinger, Joachim
2013 Die Verben des Altalbanischen. Harrassowitz Verlag, Wiesbaden.