Professional Documents
Culture Documents
1 Introduction
From a standard English perspective, the word agus (is, a's, 's) in Gaelic, which so neatly
corresponds to and in most situations, takes some curious detours in its syntactic
distribution and semantic function. One of these is its ability to introduce non-finite clauses
in sentences such as the following:
Previous treatments of this phenomenon (Boyle 1973; Ó Siadhail 1984) have characterised
it as a form of subordination. However, as Mathieson and Thompson (1988) caution, the
term 'subordination' is rarely operationally defined in the literature and frequently employed
in a vague or inconsistent manner. Boyle and Ó Siadhail do not offer definitions of the
term in their papers, thus one cannot be sure of the rationale behind their use of it. Indeed,
if we consider a clear instance of subordination in Gaelic, adverbial clauses headed by
nuair, we find significant differences in the form and distribution of the non-matrix clauses
involved versus those headed by agus:
(1.4b) *Is ann agus mi a' falbh dhan Àird a thachair an Camshronach rium.
(1.5a) Nuair a bha mi a' falbh dhan Àird, thachair an Camshronach rium.
'When I was going to the Àird, the Cameron man happened upon me.'
1
Examples (1.4a) and (1.4b) are instances of clefting while those in (1.5) are examples of
preposing, both part and parcel of the flexibility of subordinate/ adverbial syntagms in
Scottish Gaelic. I have used the ? symbol before the relevant sentences in (1.5) in
deference to Boyle's article, which maintains that preposing is possible in sentences of this
type in Irish. However, there are clearly pragmatic issues at stake here, to which I will
return later. For the moment, I would like to take the position expressed by my Scottish
Gaelic informants, that these sentences are actually ungrammatical.
(Examples from Franklin 1971 cited in Van Valin and LaPolla 1997, p. 450)
2
Example (b) is an instance of what is called a switch-reference expression. Looking at the
verbal morphology, one sees the suffix rí which is glossed as 'simultaneous action (SIM),
same pivot (SP)'. Switch-reference languages keep track of participants by stating, on a
verb's morphology, whether or not a verb's main participant, as it were, is the same as the
one of the previous verb. What is of concern to us here is: (1) this marking is in
complementary distribution to the marking of tense—it occurs in the slot where tense
would normally be assigned; and (2) the tense (and illocutionary force) of a verb with
switch-reference marking is specified by the matrix verb. Here we have a dependency not
on a structural basis, but—as defined in RRG—an operational one, operators being such
things as mood, tense, aspect, and negation. Constructions such as these, which may share
both coordinate and subordinate features but are operationally dependent, are instances of
what Olson (1981, cited in Foley & Van Valin 1984) termed cosubordination.
In Boyle (1973), the agus-headed non-matrix clause is said to have the verb 'to be' (Ir. a
bheith/ ScG a bhith) deleted from its surface structure. Ó Siadhail similarly states that there
is a 'deletion of tá from the underlying coordinate which is, in a sense, superfluous as the
tense is implied by the preceding phrase' (p. 126). However, I think the main point is that
tense and illocutionary force are unspecified in clauses of this type. Tense and illocutionary
force in Gaelic are carried on the clause-initial verb, the position assumed by agus.
Because the non-finite clause is obligatorily dependent on matrix clause for tense and
illocutionary force, this is, like the switch-reference expression, an example of
cosubordination. Finally, there are two further similarities to switch-reference expressions:
(1) They are not able to move to a detached position (such as pre- or postposing); and (2)
rather than being modifiers of the matrix clause, they are seen as being added-on in
sequence. In narrative, especially, one is able to find fairly long runs of non-finite agus-
headed clauses:
(2.2) Bha Ùisdean a' Bhaile Sear... a's a' Bhaile Sear agus am Baile Sear aige uileag ach
na bh' aig dusan croiteir agus e anabarrach cothromach agus e anabarrach mi-
fhàbharach dha na croiteirean.
...bha iad ann a-shin a's a' mhòintich agus an crodh aca agus iad... teine mòr aca
agus iad air biadh a ghabhail a's a' bheinn, agus iad a'...
If we take the position that these agus-headed clauses are typified by unspecified tense and
illocutionary force, due to the occurrence of agus in the place of the marker upon which
these are usually coded, we are able to readily account for additional data: (1) negation, (2)
impersonal forms; (3) localised temporality; and (4) copula structures. Consider the
following:
2.3 Bha iad a' rànaich 's a' rànaich 's gun fhios aca cà' 'n robh
were they crying and crying and without knowledge at them where were
iad.
they.
'They were crying and crying without knowing where they were.'
2.5 Chì thu dealbh de Mhòraig an-seo 's i aig a' chladach an-dè.
see you picture of Morag here and her at the beach yesterday.
'You see a picture of Morag here when she was at the beach yesterday.'
2.6 Bha boireannach brèagha cuide ris 's esan an duine as gràinde air an
was woman beautiful with him and him the man most ugly on the
t-saoghal.
world.
'There was a beautiful woman with him and him the ugliest man in the world.'
Example 2.2 demonstrates the only option for negation in this construction as it cannot
occur, as it normally would at the clausal level, on a finite verb (cf., ...agus cha robh fhios
aca...). In example 2.3, we see a form that will seem out of place to many readers, the
morpheme -ar which I, for one, thought was a bound morpheme and always confined to
finite verbs. However its appearance is not inexplicable; agus is assuming the position
normally reserved for verbal elements and -ar is just tagged at the end, acting like a normal
free pronoun. The grammaticality of (2.4) is a consequence of the unspecification of tense
in the agus-headed clause; the temporal adverb an-dè, being more local than the tense
specification of the matrix clause, is able to override it by virtue of its proximity. And
finally, in (2.6), because copular verbs occur in the first position, which is unavailable, one
is able to obtain a copular interpretation if the arguments present are consistent with
copular form. (Interestingly, it seems that tense is not implicated in the copular clause. In
other words, the tense of the matrix clause is not necessary for the successful parsing of the
second.) Thus, by simply stating that the elements carried by finite verbs are left
unspecified in agus-headed non-finite clauses, one is able to account for a wide-array of
data.
To return to the topic of nexus, we are left with a three-way division. The main difference
between clausal cosubordination and subordination is the obligatory operator dependence
evinced in cosubordination. In subordination, which can manifest as either embedding
(e.g., acting as an nominalised argument of a predicate) or modification, operator
dependence is not significant. This tertiary division can be represented as follows (from
Van Valin & LaPolla 1997, p. 454):
4
NEXUS
Dependent Independent
COORDINATION
Structural Operator
dependence dependence
SUBORDINATION COSUBORDINATION
Role and Reference Grammar developed out of an attempt to account for, in the first
instance, languages such as Lakhota, Dyirbal, and Kewa rather than the range of Indo-
European languages currently dominating most linguistic theories. There isn't space
enough here to provide a full introduction, but, in order to understand the tree diagrams, we
will cover the basics of clause structure within this paradigm.
There are three major syntactic units to consider: the core, the nucleus, and the periphery.
The core consists of the semantic predicate plus its argument, the nucleus is equivalent to
the semantic predicate, and the periphery is where the non-arguments get consigned to.
The clause is formed by the combination of the core and its periphery. This schema is able
to account for universal distinctions between predicating elements and non-predicating
elements on the one hand and arguments and adjuncts on the other (Van Valin & LaPolla
1997). It can be represented graphically as follows:
CLAUSE
CORE PERIPHERY
Nucleus Arguments
Figure 1.1 The Layered Structure of the Clause
Respectively, it can be represented in a tree-diagram:
SENTENCE
CLAUSE
CORE PERIPHERY
PRED
V NP NP PP
6
As mentioned earlier, RRG posits a group of operators such as tense, illocutionary force,
aspect, modality, and negation. Different operators modify different levels of the clause, so
there will be nuclear operators, core operators, and clausal operators. We are discussing
clausal juncture and only two of the clausal operators will be invoked in our
representations: tense and illocutionary force. Here is the previous tree diagram with its
operator projection inserted:
SENTENCE
CLAUSE
CORE PERIPHERY
PRED
V NP NP PP
SENTENCE
Now, if we consider some data involving both an adverbial adjunct and a cosubordinate
clause, we will be able to determine their relative positions in the LSC.
(3.1) a. Bha iad ann nuair a chaidh mi seachad 's teine mòr aca cuideachd.
was they in it when went mi past and fire big at them also.
'They were there when I went by, and they had a big fire too.'
b. ?Bha iad ann 's teine mòr aca cuideachd nuair a chaidh mi seachad.
'They were there and they had a big fire too when I went by.'
(3.2) a. Chunnaic mi an-dè i 's i ag obair aig an arbhar.
saw me yesterday her and her working at the corn.
'I saw her yesterday while/ when she was working at the corn.'
The problem with (3.1b) and (3.2b) isn't so much that they are categorically ungrammatical,
but that they have a different meaning, or at least, emphasis from the (a) examples. If one
is intending for the adverbial adjunct to modify the matrix clause, it must occur adjacent to
it. If one puts it behind the cosubordinate clause, then it is that clause that gets modified.
In 3.1, it is the difference between 'They were there when I went by and they had a big fire'
or 'There were there and they had a big fire when I went by', implying that they might not
have had one at another time. The difference in meaning is slighter in 3.2b, but it is
dispreferred to 3.2a and, in the words of one informant, is a mistake common to young
children and learners of Gaelic. If we provide a tree diagram, this difference will become
clear.
SENTENCE
CLAUSE
PRED PRED
V V
NUC NUC
CLAUSE CLAUSE
CLAUSE
TNS
CLAUSE
IF
SENTENCE
8
Here is the representation of (3.2b):
SENTENCE
CLAUSE
PRED PRED
V V
NUC NUC
CLAUSE CLAUSE
TNS CLAUSE
IF CLAUSE
SENTENCE
So, if we compare figures 3.4 and 3.5, we see the crucial differences, represented in both
the constituent and operator projections. Comparing the constituent projections, we see that
when an-dè is adjacent to the matrix clause it modifies that clause's core. But, when it falls
after the cosubordinate clause, it instead becomes its periphery rather than that of the matrix
clause. Looking at the operator projections now, we see an echoes of the same distinction;
temporality is a core operator and is shown in this role in each operator projection. If we
inserted the subordinate clauses in (3.1) in the place of an-dè, we would obtain the same
tree diagrams.
So, non-finite clauses headed by agus are distinguished from adverbial subordinate clauses
in that they are not peripheral modifiers and they are joined with it at the clausal, rather
than the core level. One more example will make this difference even more clear:
Imaginary dialogue
(4.1) <1> Thachair an Camhsronach rium agus mi a' falbh dhan Àird.
<2> Cuine thachair an Camshronach riut?
<1> *Agus mi a' falbh dhan Àird.
Nuair a bha mi a' falbh dhan Àird. (Italics added for emphasis)
Notice that in order to answer <2>'s question, <1> would have to resort to a temporal
interpretation using an adverbial clause. There is no explicit coding of cotemporality in the
initial utterance; it is merely an obligatory condition of the semantics involved. Because in
10
Gaelic, the concept of cotemporality is not lexicalised as it is in English—there is no word
corresponding to English while—<1> cannot answer <2>'s question in a direct fashion.
So, mentioning semantics, how best should we conceptualise the machinery of these
constructions? I think the answer is much simpler than the elaborate taxonomy offered in Ó
Siadhail (1984): agus, even when it heads a non-finite clause, is still a conjunction. It is
crucial to differentiate between conjunction, which is a formal construction type, and
coordination, which is an abstract linkage relation wherein there is independence and
equivalence between the units of the juncture (Van Valin & LaPolla 1997). Conjunction
can be either coordinate or cosubordination. As Lakoff (1984) comments "The presence of
'and' is a virtual red flag, signalling to the hearer, 'these ideas are related somehow, guess
how" (p. 487). This is the bare-bones semantics involved in agus-cosubordination. I think
that Ó Siadhail's classification—distinguishing between temporal and non-temporal,
functional equivalents of relative clauses, concessions, and so on—is useful from a
functional, comparative standpoint with English, but is missing the forest for the trees on a
descriptive level. The fact is, the distinctions that he makes are a product of world
knowledge, or, in Lakoff's terms, the 'guessing how'; they are still all instances of
conjunction.
Temporal
a. Nach minic a bhuail fonn mé, agus mé faoi bhrothall te tirim na gréine ar na pampas ó
dheas... How often during the cloudless dog days of the Pampas had I yearned for...
(translations are from Ó Siadhail)
Concessive
d. Bhí clann agus clann a cluine curth' i gcrí agus í féin beó fós. Her children and her
children's children were done for though she herself was still alive.
Causal
e. Bhí an lá chomh gearr sin is tú ag déanamh 'chuile bheaintáil. The day was so short
when (= since, seeing that) you were gallivanting around.