a study of hittite sentence connectors and old irish preverbal particles
TRANSCRIPT
1
A study of Hittite sentence connectors and Old Irish preverbal particles
Paulus van Sluis
January 2014
Introduction
The Proto-Indo-European word *nū has kept a similar meaning throughout many of its
daughter languages (e.g. English now, Latin nunc, Sanskrit nu are all cognates and all of them
mean the same thing) and they also seem to be commonly used as an adverb in all of these
languages. The more peripheral languages of Hittite and Old Irish, however, do not share this
adverbial usage, employing their respective descendant of *nū as clause-linking conjunctions
and semantically null preverbal particles, respectively. Similarly, Hittite ta and Old Irish to
(later do), and Hittite šu and Old Irish se belong to the same word class as *nū in Hittite and
they demonstratively have a common etymology1, with šu and ta descending from the
demonstrative pronoun *so-, *to- 2. (However, PIE *so- would regularly yield Hittite **ša.
Kloekhorst suggests influence from nu may have caused šu to secondarily take over its -u-3.)
The precise phonological details of each descendant will not be treated in this essay.
Instead, this essay will attempt to elucidate upon the precise usage and meaning of this word
category in Indo-European. While these words themselves demonstrably have a common
ancestors, it is still a pressing question whether the parallels in their usage within a clause,
their stress and respective semantic values may be equally attributed to a Proto-Indo-
European archaism. Moreover, what innovations must be assumed if we are to assume either
independent innovations or a common ground?
Linguistic considerations
While neither historical semantics nor historical syntax are as hard a science as, for example,
phonological reconstruction, some principles may still be employed in order to make such a
reconstruction credible. One of these ways is to estimate the degree to which a particular
Indo-European dialect was isolated from other dialects used in the comparison and attribute
parallels between strongly isolated languages to a shared archaism. For example, Slavic and
Germanic may be phylogenetically significantly different, but semantic or syntactic parallels
may potentially be analysed as a Sprachbund feature. It is much harder to make a similar case
1 Watkins (1963) 1-49.
2 Matasovic (2009) 281
3 Kloekhorst (2008) 772
2
for Celtic and Hittite. If we find a similar parallel between these two languages, citing shared
archaism as the cause may be more economical than explaining it as an areal feature. It is for
this reason that the practice of making a comparison between Old Irish and Hittite is justified.
Moreover, the practice of making a comparison between Old Irish and Hittite, and
specifically these languages, is justified in that their usage shows some remarkable parallels.
In both languages, for example, these words always come at the very beginning of a clause,
and in both languages they appear to be unstressed. In Old Irish, their semantic value appears
to be null, but no semantic reconstruction so far poses any semantic difficulties in connecting
them to Hittite.
The analysis of the descendants of these particles will be carried out in a two-stage
fashion, first providing a synchronic account as full as possible of their respective meanings,
usage and pronunciation. Then, a diachronic study of their usage will further illuminate in
what way exactly these reflexes developed and hopefully from where they came.
The usage of sentence connectors in Hittite
In Hittite, clauses may be linked in two ways. One is by asyndeton. That is: simple
juxtaposition. The other way is by means of the so-called ‘clause-linking conjunctions’: nu,
šu, ta, -a/-ya, and –a/-ma. The last two of these can serve other purposes than clause linkage:
they can also be used to link a series of words in enumeration, or to contrast one phrase within
a clause with another, respectively. These last two will not be treated further in this essay.
Clause-initial clause-linking conjunctions
The other three clause-linking conjunctions mentioned are the clause-initial clause-linking
conjunctions. Their initial position in a clause means that these conjunctions are usually found
at the beginning of a particle chain. These conjunctions are never used, however, in discourse-
initial position. That is: usage of these conjunctions is necessitated by the need to connect the
clause in which they are located with the preceding clause. Thus, where there is preceding
clause, there is no need to start a clause with a clause-linking conjunction. In general, they
vaguely have the meaning ‘now’, ‘then’, or ‘and’, but may often be left untranslated.
In Old Hittite, all three conjunctions are in use, while in in later Hittite šu and ta- lose
ground to nu. The conjunctions šu and ta appear to be in complementary distribution, with šu
being used with preterite verbs, and ta being used with verbs in the present-future,
imperatives, and analytic present-perfects with ḫar(k)-. Finding conditions for usage of nu-
3
and šu/ta has proven difficult. Any account illustrating the type of clauses in which these
conjunctions are used will inevitably show overlap in their respective functions. 4
The usage of nu
Already in Hittite texts written in Old Script, nu is used to connect clauses in which actions
are sequential. Sometimes the second or final clause express a result from its preceding
clause, necessitating a translation of nu with ‘so that’ or ‘therefore’. E.g. zik=ma=wa=za
DUMU-aš nu=wa UL kuitki šakti ‘But you are child, and [therefore] know nothing at all’. It is
also possible to use nu as a means to connect two subordinate clauses, although ta and -a/-ya
may also be used here. E.g. dUTU-ŠI=ma maḫḫan iyaḫḫat nu maḫḫan ANA
URUŠallapa [arḫu]n
‘When I, My Majesty, marched, and when I [reach]ed Šallapa…’
Already in OS, nu may be used to connect an independent clause to a preceding
subordinate temporal clause: mān URU
Tamar[mara] arer nu taršikanzi ‘When they arrived in
T., then they said’. In OH, however, šu may also be used in this environment. Again from OS
onward, nu may be used to connect a preposed relative clause to a main clause: nu uit
[(LU)]GAL-uš tezzi nu apāt iyami ‘And what the king says, that (then) I will do’. In OH, nu
may rarely be used to connect a conditional clause with a main clause: mān INA KUR URU
ḫatti
nu=za unattallan=pat arnuzzi ‘If it is in the land of Ḫatti, then he shall bring the merchant
himself’.5
The usage of šu
šu usually connects two independent clauses. These clauses generally refer to sequential
actions or final actions š=an dḫalmaš[uiz]
dšīuš=(š)miš parā paiš š=an išpandi nakkit dāḫḫun
‘And Ḫalmašuit, their god, handed it over, and (thus) I took it by force’. It can also be used to
connect a temporal clause to a succeeding main clause, but cannot be used to connect a
conditional clause since šu is only used in the preterite. Finally, šu may be used to connect a
relative clause to a main clause6.
The usage of ta
Connector ta generally appears to be used to mark the last sentence of a piece of discourse
dealing with one topic and may therefore often be translated with ‘thus’.7
4Hoffner, Melchert (2008) 390
5 Hoffner, Melchert (2008) 390-392
6 Hoffner, Melchert (2008) 392-393
7 Kloekhorst (2008) 926
4
ta, like nu may be used to connect two clauses in a series of chronologically sequential
actions: LUGAL-uš 3⸗ŠU aīš⸗šet ārri [(t⸗at)] ḫurtiya[(li)]ya lāḫui MUNUS.LUGAL-š⸗a 3⸗ŠU
aīš⸗šet [(ā)]rri n⸗at ḫurtiyaliya lāḫui ‘The king washes his mouth three times and pours it
(the water) into the basin. The queen also washes her mouth three times and pours it into the
basin’. ta may also introduce the last consequential clause in a conditional sentence e.g. takku
A.ŠÀ.ḪI.A NÍG BA LUGAL kuiški ḫarzi šaḫḫan luzzi nat[ta karpiezzi] † LUGAL-uš GIŠ
BANŠUR-az
NINDA-an dāi ta⸗šše pāi ‘If someone holds fields as a gift from the king, he shall not [render]
šaḫḫan (or) luzzi: the king shall take food from his table and give it to him’. Like nu, ta may
also be used to introduce a result clause: ḪUR.SAG-an tarmaemi t⸗ašta edi natta nēari
[(arun)]an tarmāmi nu āppanatta lāḫui ‘I will fix the mountain in place, so that it may not
move; I will fix the sea in place, so that it may not flow back’.
When connecting a preposed relative clause to a main clause, ta may also be used:
[kui]š⸗za LU
ḫippari ḫāppar iezzi n⸗aš⸗kan ḫāpparaz [šeme]nzi LUḫipparaš kuit ḫapparāit
t⸗a(t)⸗z āppa dāi ‘Whoever makes a purchase from a ḫippara-man shall forfeit his purchase
price, and the ḫippara-man shall take back what he sold’. In post-OH copies, ta may connect
a conditional clause to a following main clause: takku⸗wa⸗ššan kī ḫazzizi ta⸗wa DINGIR-LUM
takku⸗wa⸗ššan natta⸗ma ḫa[zzizi] ta⸗wa antuwaḫḫeš ta⸗wa […] zaḫḫuweni ‘If he guesses this
correctly, then it is a god; if he does not guess correctly, then it is a human being, and we will
fight […]8.
Stress of nu, ta, šu
The question whether or not these conjunctions carried stress has been subject to some debate.
It would make sense if they were stressed, since these clause-linking particles often take one
or more unstressed suffixed pronouns. These pronouns could then be argued to be suffixed to
these particles through Wackernagel’s law, which states that unstressed elements tend to fall
right after the first stressed element of a clause. Moreover, Kronasser theorizes that indeed the
syllable preceding a geminate consonant (i.e. doubly written) is stressed9. Melchert, citing a
law formulated by Bernabé, shows that enclitics starting with */n/ or */s/ succeeding stressed
*nu-/to- had their first consonant reduplicated as a result of immediately following a stressed
syllable: e.g. *nu/to + soy > nu=šše/ta=šše, *nu/to + nos > nu=nnaš10
. He states that this
gemination was later generalized to cases in which another syllable stands between the
8 Hoffner, Melchert (2008) 393-395
9 Kronasser (1956)
10 Bernabé (1973) 153, 154
5
stressed syllable and the enclitic e.g. nu=mu=ššan. In this way he demonstrated that these
conjunctions had a stressed pronunciation. However, Melchert is not fully convinced of their
stressed nature, citing the fact that these conjunctions themselves never underwent vowel
lengthening.11
Kloekhorst asserts that these conjunctions are unstressed both when used
independently and when they form the head of a particle chain. This is evidenced by their
consistent non-plene spellings and by the fact that unstressed conjunction -ma is often
attached to a word following nu (and presumably ta and šu as well)12
. According to
Wackernagel’s law, unstressed elements such as -ma universally tend to come right after the
first stressed word of a sentence, thus indicating that nu is unstressed. Research on stress
patterns in poetic metre by Durnford13
and Melchert14
moreover reveals that sentence-intitial
nu and all its attached enclitics are indeed unstressed.
A consequence of this wholly unstressed nature of the whole particle chain in Hittite is
that one may not employ Wackernagel’s law to account for the fact that these conjunctions
may take unstressed suffixes. Moreover, non-fixed conjunctive particles –a/-ya and –a/-ma
tend to be used as a suffix after the first stressed word of a clause when used to connect
clauses. These particles show that, even in Hittite, Wackernagel’s law operates on clitics, but
clause-initial sentence connectors elusively evade this rule.15
The Old Irish verbal complex
To properly understand the Old Irish usage of preverbal particle no and preverb to, a basic
understanding of its unique verbal system is necessary. In most tenses and moods verbal
personal endings have two forms: absolute and conjunct. Conjunct verbs occur in all
compound verbs consisting of a verb and a prefixed preposition (i.e. preverb), in simple verbs
after verbal particles ro and no, and in simple verbs after a series of conjunctions or conjunct
particles e.g. do-beir, ‘gives’, nom-beir ‘carries me’. The absolute flexion occurs only with
simple verbs in positions other than the above-mentioned e.g. beirid ‘carries’16
.
Conjunct verbs and absolute verbs may both take an unstressed object pronoun which
is then attached to the verb. In the case of absolute verbs, this object pronoun is suffixed to a
11
Melchert (1994) 151 12
Kloekhorst (2011) 157-176. 13
Durnford (1971) 69-75 14
Melchert (1998) 492 15
Hoffner, Melchert (2008) 395-400 16
Thurneysen (1946) § 542
6
verb form. e.g. beirthi ‘carries it’ from berid17
. In the case of conjunct verbs, a similar
unstressed object pronoun is infixed between the preverb or the preverbal particle and the
verb e.g. nos-mbeir ‘carries her’. These object pronouns are always the direct object, except
with the verb ‘to be’ where it is the indirect object in order to denote possession18
.
Similarly, absolute and conjunct verbs each have their own way of introducing a
relative clause with a unique verbal inflection. The absolute flexion has unique relative forms
with its own set of endings e.g. beires ‘(that) he carries’ as opposed to beirid ‘he carries’.
Introducing a relative clause using a verb with a conjunct ending requires usage of a relative
particle an which is only visible after some conjunctions e.g. ara-mbeir ‘in order that he
carries’. If no such conjunction is used, however, a series of separate strategies to denote a
relative clause may be used, such as a separate set of infixed pronouns (e.g. dodom-beir ‘that
gives me’). A very common strategy when a preposition, a negative particle, or preverbal
particle no or ro is used, is to let lenition or nasalization follow these preverbal forms.
Lenition is used when the antecedent is the subject or object of the relative clause e.g. is hed
in so no-chairigur ‘it is this that I reprimand’ while nasalization is used when the antecedent
designates the time, place, manner, or degree of the relative clause e.g. is sí méit in sin do-n-
indnagar in díthnad ‘that is the extent to which the consolation is bestowed’19
.
The usage of no
When a simple (i.e. not compounded with a preverbal particle) verb was used without
any preverbal conjunctions in a relative position or with an object pronoun, the system as
described above might imply that in such cases the special relative ending or a suffixed
pronoun was always used. This however, is not the case. An alternative to these methods was
to use the semantically null no as a preverbal conjunction, turning the verb in a conjunct form
with a conjunct ending and providing a slot for the infixed pronoun or a relative mutation, so
no-chara ‘that loves’ means the same as caras. Also cf. ní hed not-beir í nem ‘it is not this
that brings you to heaven’ with the b in beir pronounced as lenited /β/, thus containing an
object pronoun as well as a relative mutation. Also, some persons, tenses and moods do not
have an absolute form in the first place, making usage of a preverbal particle obligatory20
.
Meaning of no
17
Thurneysen (1946) § 429 18
Thurneysen (1946) § 409 19
Thurneysen (1946) §§ 493-500 20
Thurneysen (1946) § 538
7
While no is traditionally held to carry no semantic load by itself, it is noteworthy that the only
preverbal particles that are not a conjunction or a preverbal preposition are ro and no. These
two preverbal particles are mutually exclusive. That is: a verbal complex with ro in it cannot
carry no and vice versa. It is quite possible, then, that these two particles have a meaning
contrastive to each other. This hypothesis makes further sense when comparing etymologies:
ro descends from Proto-Indo-European preposition *pro- ‘in front of, before’ while no, as
stated above, descends from PROTO-INDO-EUROPEAN *nū ‘now’.
Old Irish verbal particle ro is prefixed to a verb in the preterite indicative and past
subjunctive to change its tense into the perfect, e.g. as-bert ‘says’ becomes as-rubart ‘has
said’. In this way the Old Irish verbal system allows to distinguish between a perfect and a
narrative tense21
. Moreover, the verb forms which always have secondary endings are
imperfect indicative, past subjunctive and secondary future: these forms must therefore take
no if no other preverb or particle is available22
. Given this opposition, it may be argued that no
actually is some sort of marker for an imperfect and/or narrative tense not unlike Hittite. And
even if this is not the case in our attested Old Irish corpus, then surely it must be from this
meaning that no as a preverbal particle developed. This suggestion was first made by
Watkins23
.
The usage of se
Old Irish se serves to infix –ch (< *kwe) ‘and’ to is ‘is’ to compound se-ch is ‘that is to say,
namely’e.g. sech is ardi son dombera dia do neuch ‘that is: a sign that God will give to
someone'. Sech may also have adversative force: sech ba foirbthe a ires sidi ‘although his
faith was perfect’. Similarly, no may be used: no-ch ‘and’24
. Notably, it is followed by the
absolute or deuterotonic form (that is: stress in a compound verb would stay on the second
syllable) of the following verb25
.
The usage of to
In Thurneysen’s A Grammar of Old Irish, a preposition to ‘towards’ is described, occurring
only in composition with a verb. Before nouns an pronouns do and du with the same meaning
were used instead. Myles Dillon, however, was the first to hypothesize a different history of
the preverb to, citing some examples that show that this preverb rather marked completed
21
Thurneysen (1946) § 530 22
Watkins (1963) 15 23
Watkins (1963) 13 24
Thurneysen (1946) § 35, 883; Watkins (1963) 14 25
DIL (1913-1976) s.v. ‘3 sech’
8
action: boingid ‘breaks’ vs. do-boing ‘breaks off, plucks’; claidid ‘digs’ vs. do-claid ‘digs up’.
In some other verbs the meaning seeds the same: e.g. caithid, do-caithi ‘spends’ cuirithir, do-
cuirethar ‘puts’. He then tentatively connects this preverb with the Hittite connective ta. One
objection to this is that to is a preverb and not a sentence particle or even a preverbal particle.
Dillon theorizes on the basis of this observation that to, must historically have functioned as
an old connective. This hypothesis is corroborated by the fact that it often the sole preverb
within a verbal complex and if it is not, it is always the first preverb within a verbal complex,
implying it was historically positioned at the very beginning of a clause, like Hittite ta. Its
resultative meaning is also reminiscent of Hittite ta, which in turn always seemed to denote
the upshot of a certain condition. Its development into a perfective preverb must be
secondary.26
This same to-preverb (or conjunction) is also found on a Gaulish-language
funerary inscription as to, where it is considered semantically null, although current
knowledge of Gaulish is not sufficient to be sure27
:
TETVMVS SEXTI DUGIAVA SAŚADIS (1A)
Tetamus-NOMSG, Sextus-GENSG Dugiaua-NOMSG Saśadis-GENsg
‘Tetamus, son of Sextus; Dugiaua, daughter of Saśadis (are buried here)’
to-med=ec-lai obalda natina (1B)
PV=PRO.1SG.ACC-V.Pret.3SG Obalda-NOMSG daughter-NOMSG
‘(and) Obalda, their daughter, set me (i.e. the monument) up.’
Stress and Old Irish preverbal particles
When a verb is compounded with a preverb, the stress falls on the first syllable of the verb
itself. In fact, the first preverb does not even form a close compound with the second element
and may be separated from it by another clitic, or in verse even by other stressed words.
Conjunct particles and verbal particles ro, no are also unstressed if they stand in the first
position. Notably, no always stands in the first position of the verbal complex and is thus
always unstressed. This is not the case for to, as preverbs are similarly unstressed, but may be
moved to a stressed position when a verbal or conjunct particle is added to a verbal complex
with a preverb: do-‘beir ‘gives’ vs. ní-‘tabair ‘does not give’28
.
26
Dillon (1962) 120-126 27
Eska (1994) 41 28
Thurneysen (1946) § 37-39
9
Diachronically, this pattern of “unstressed element + infixed object or relative +
stressed verb” may be explained in two ways: the first explanation is that these particles
became unstressed at some point, but were stressed in an earlier stage. In this earlier stage,
then, Wackernagel’s law could cause object clitics to be suffixed to these particles. If,
however, we are to assume that these particles were always unstressed, it is impossible to
explain the placement of object enclitics or relative mutations through Wackernagel’s law. A
second explanation will then be required instead.
Old Irish and Hittite
As can be seen from both Old Irish and Hittite, no and nu, and to and ta are always
unstressed. It would be simple to infer that this was also the case in their shared protoform.
Assuming either stressed or unstressed nature of this word has a deep impact on several
models proposed to explain Old Irish VSO word order. Evidence from Hittite stress and usage
of nu, ta may aid us in reconstructing the development from Proto-Indo-European to Old
Irish, and may therefore also serve to reconstruct a Proto-Indo-European usage of these
particles.
A diachronic account of Old Irish
However, Calvert Watkins, working under the impression that Hittite clause-linking
conjunctions were stressed, developed a theory on Old Irish syntax presupposing the stressed
nature of no29
. He stated that the Proto-Indo-European word order was SOV, and then pointed
towards an archaic Old Irish word order dubbed Bergin’s law:
ceso femmuin mbolgaig mbung
Although seaweed-ACC blistered-ACC reap-1SGPRESCONJ (2)
‘Although I reap blistered seaweed’
In this phrase the verb stands at the end of the clause, and takes the conjunct inflection. This
construction is generally found in Old Irish texts that tend to preserve more archaic language,
such as poetry and legal texts. In this syntactic system, then, enclitic object pronouns were
moved to the beginning of the sentene under pressure from Wackernagel’s law. Along with
this the stressed preverbal particle moved. This separation between preverbal particle and
enclitic pronoun on one hand, and the verb itself on the other hand is called tmesis:
29
Watkins (1963) 39-40
10
No-m Choimm-diu coíma
PREVPART-PRON1SG beloved-God-NOM cherish-3SGPRESCONJ (3)
‘Beloved God cherishes me’
The next step in Watkins’ reconstruction, then, is the introduction of Vendryes’s restriction.
This rule states that nothing should stand between a preverbal complex and a the stem of a
verb. Following this line of reasoning, this univerbation will then cause either a reversion to
OV word order or it will cause the verb to stand at the very beginning of a clause. This last
development is what happened to Old Irish30
.
This account has received a great deal of criticism and a different account has
consequently been proposed originally by Sims-Williams31
, and was later refined by Eska32
.
Curiously, their refinements of Watkins’s account are not in the stressed nature of preverbal
particles and conjunctions, but in the weight given to tmesis as a vehicle for syntactic change.
The problem with an account invoking tmesis, they argue, is that it is only attested as a
literary form of expression in not only Old Irish, but also languages such as Latin, Greek and
Sanskrit. It does not make sense to them to cite this stylistic aberration as a vehicle for as
fundamental a syntactic change as shift to a VSO word order. Evidence from Hittite, however,
might demonstrate that tmesis may not have been such a highly literary construction after all
in Proto-Indo-European. Rather, it could be a perfectly normal day-to-day way of saying
things. This is, after all, what we see in the Hittite language with all of its separations between
nu and its enclitics on one hand and its verb on another.
A renewed account of Old Irish syntax on the basis of Hittite
According to Sims-Williams and Eska, original pre-Old Irish syntax inherited an Indo-
European pattern in which a stressed connective (*’to is used in the examples below, but
others should be equally applicable) stood at the front of a clause and took an enclitic suffixed
pronoun. This is different from Watkins’s account, since Eska does not assume
Wackernagel’s law to account for the placement of connectives at the beginning of a clause,
but rather recognizes that their place was originally fixed to the beginning of a clause as it is
in Hittite. Indeed, no argument needs to be made for its position, as it is assumed to be the
30
Watkins (1963) 11-33. 31
Sims-Williams (1984) 138-201 32
Eska (1994) 39-62
11
archaic situation, so it is ‘just there’. Thus, he takes the following Indo-European structure as
a point of departure33
:
#[‘to=Objpron] (‘S) ‘V# (4)
Following this situation, stronger association of the enclitic object pronoun and the verb led to
a rebracketing causing a pronominal object clitic to become proclitic to the verb instead of
suffixed to the connective34
.
#‘to [Objpron=‘V]# (5)
Naturally, this rebracketing could only occur in null-subject clauses, as the subject would
otherwise stand between the object pronoun and the verb, maintaining the situation as
described in (4). The semantic and syntactic attraction between a verb and its complement
was nevertheless strong enough to warrant a grammatical reanalysis even in this case, causing
the verb to ‘jump over’ the subject, thus giving birth to the VSO word order:
#[‘to=Objpron] ‘S ‘V# → #‘to [Objpron=‘V] ‘S# (6)
It is after this stage that *to was reanalyzed as a preverb and Eska argues that this reanalysis,
in turn, was the immediate cause for it to become itself an unstressed clitic.
This insight was formulated before the analysis of metre and plene spelling provided a
strong case for Hittite’s clause-linking conjunctions’ unstressedness. Indeed, Eska goes at
length to argue that *nu and *to were stressed as they allegedly were in Hittite, citing, among
others, evidence from Celtiberian. In Celtiberian, word dividers are written and they
unambiguously indicate whether a morpheme was stressed. For instance, the disjunction =ue
was a clitic because it was not written independently, but the sequence es : uertai contains a
word divider showing the preverb was independently stressed. However. even if Celtiberian
preverbs were stressed and even if the stressed nature of preverbs in Celtiberian reflected the
state of affairs of an immediate ancestor of Old Irish, then a stressed preverb would be more
likely to be an innovation from an unstressed preverb than the other way around. The reason
for this is that the first aforementioned innovation would simply be an application of
Wackernagel’s law while the second would innovate into a problematic state. Moreover, the
Celtiberian stressed preverbs cited by Eska appear to be prepositions used as preverbs, not
33
Examples taken from Eska (1994) 59-60 34
This same step may be employed to account for the existence of suffixed object pronouns in Old
Irish absolute verbs.
12
clause-linking conjunctions used as preverbal particles: e.g. es : uertai, with es being related
to Old Irish a(s) ‘out of’. Therefore, Celtiberian preverbs cannot be cited as evidence to infer
that clause-initial conjunctions were stressed, since even Eska himself argues that reanalysis
of *to as a preverb is a development after the development of the Old Irish VSO syntax35
. A
theory must fit the facts and not the other way around: while it is easy for the sake of syntactic
reconstruction to assume stressed *nū *to,*so, direct evidence shows they were unstressed. It
would be preferable, then, to concoct a theory explaining the VSO-word order without
requiring stressed *nū *to,*so
35
Eska (1994) 56
13
Conclusion and discussion
The account above shows that if we take Hittite syntax as well as Hittite stress patterns as a
starting point for a diachronic account of Proto-Indo-European to Old Irish usage of sentence
connectors, less assumptions need to be made, making it an attractive starting point. Indeed,
the only shift we now need to account for in Early Celtic is the stronger association of what in
Hittite we call clause-linking conjunctions with their respective verbs. Hypothesizing this
closer semantic association is not a huge leap of faith when analyzing meanings of Old Irish
compound verbs e.g. beirid ‘carries’ and do-beir ‘gives’, but with the elements of the
compound verb meaning ‘finishes bringing’ historically. From then onwards, the only rule we
need for a VSO word order is Vendryes’s restriction. The only assumption we need to make
on the Hittite part is a lack of semantic or syntactic change on this account.
As for other branches of Indo-European than Celtic: when assuming a shift away from
the Hittite-style clause-initial clause-linking conjunctions, the situations we actually encounter
in other languages would be expected due to their erratic behavior in Hittite. An unstressed
element that is also clause-initial is necessarily unstable due to Wackernagel’s law which
states that unstressed elements tend to fall immediately after the first stressed element of a
clause. One would thus expect reflexes of Proto-Indo-European *nū *to,*so to either become
stressed (and be subsequently moved to another lexical category, such as adverb) or to be
moved to the second position in a sentence. This is indeed what we see happening, as can be
witnessed in the particle *nū becoming an adverb in, for example, Germanic and in the Greek
particle νυν falling in the second position of a clause.
In order to reconstruct a Proto-Indo-European system in the usage of nu, ta, and se,
taking Hittite as a starting point is the most economic point of departure precisely because the
Hittite language does not make sense when it comes to stress patterns, so it would be easier to
postulate innovations away from the Hittite system than it would be to postulate innovations
leading to the Hittite system. In conclusion, it turns out that neither the position of Hittite
clitics nor that of Old Irish clitics can be explained as a result of Wackernagel’s law, so it
stands to reason that this similar idiosyncrasy may be explained as a shared archaism as long
as no alternative mechanism can be postulated to explain these phenomena.
There is, however, an obvious drawback to this train of thought: assuming archaism
exactly as a result of lack of understanding is a form of surrender to what is not understood.
Instead, a researcher should try to understand what is not understood instead, since that is
14
what a researcher does. In this case, being unable to comprehend how the Hittite situation
came into being is not in itself an argument for its archaism. Although this paper’s conclusion
is that Hittite serves as a fine basis when accounting for some Old Irish morphological and
syntactic peculiarities, care must be taken that approaching the same issue from a different
angle with knowledge of different languages might lead to a wholly different outcome. This
thorny issue described above is also known within the field of historical syntax as Teeter’s
law, which was formulated by Calvert Watkins and states that ‘the language of the family you
know best always turns out to be the most archaic.’36
To counteract this law, a comparison of
these clause-linking conjunctions between two other and phylogenetically similarly distantly
removed Indo-European languages would be valuable.
36
Watkins (1976) 310
15
References
Bernabé, Alberto, (1973) “Geminación de s y sonantes en hetita.” In Revista española de
linguística
Dillon, Myles, (1962 ) “History of the Preverb to” In Éigse vol. X, part 2 120-126
Durnford, S.P.B. (1971) “Some evidence for Syntactic Stress in Hittite.” In Anatolian Studies
69-75
Eska, Joseph F., (1994) “On the Crossroads of Phonology and Syntax: remarks on the origin
of Vendryes's Restriction and related matters.” In Studia Celtica, XXVIII 39-62
Dictionary of the Irish Language based mainly on Old and Middle Irish materials. (1913-
1976) Dublin (abb. DIL)
Greene, David, (1976) ‘Archaic Irish’, Indogermanisch und Keltisch: Kolloquium der
Indogermanischen Gesellschaft am 16. und 17. Februar 1976 in Bonn: Vorträge unter
Mitwirkung von Rolf Ködderitzsch, ed. K. H. Schmidt. Wiesbaden 11-33
Hoffner, Harry A., Melchert, H. Craig, (2008) A Grammar of the Hittite Language. Part 1:
Reference Grammar. Languages of the Ancient Near East; 1. Winona Lake, IN
Kloekhorst, A., (2008) Etymological Dictionary of the Hittite Inherited Lexicon (= Leiden
Indo-European Etymological Dictionary Series 5), Leiden, Boston
Kloekhorst, A., (2011) “Accentuation and poetic meter in Hittite” In Hethitische Literatur.
Überlieferungsprozesse, Textstrukturen, Ausdrucksformen und Nachwirken. Akten des
Symposiums vom 18. bis 20. Februar 2010 in Bonn (edd. M. Hutter & S. Hutter-
Braunsar) (= Alter Orient und Altes Testament 391), Münster, 157-176.
Kronasser, H., (1956) Vergleichende Laut- und Formenlehre des Hethitischen, Heidelberg.
Matasovic R. (2009) Etymological Dictionary of Proto-Celtic. Leiden, Boston
Melchert, H. Craig., (1994) Anatolian Historical Phonology (= Leiden Studies in Indo-
European 3), Amsterdam, Atlanta
Melchert, H.Craig., (1998), Poetic meter and phrasal stress in Hittite, In Mír Curad, Studies in
Honor of Calvert Watkins (eds. J. Jasanoff, H.Craig. Melchert & L. Olivier),
Innsbruck
Sims-Williams, Patrick. (1984) “The Double system of verbal inflexion in Old Irish”, In TPhS
138-201
Thurneysen, Rudolf. (1946) A grammar of Old Irish, Revised and enlarged edition, translated
from the German by D. A. Binchy and Osborn Bergin. Dublin
Watkins, C., (1963) “Preliminaries to a historical and comparative analysis of the syntax of
the Old Irish verb” In Celtica 6, 1-49.
Watkins, C., (1976), "Towards Proto-Indo-European syntax: problems and pseudo-problems",
in Steever, Sanford B.; Walker, Carol A.; Mufwene, Salikoko S., In Papers from the
Parasession on Diachronic Syntax, Chicago. 305–326.