3 Accounting for degrees of transparency between levels
A number of properties between the Interpersonal and Representational levels, the Representational and Morphosyntactic levels and the Morphosyntactic and Phonological levels are expected for a language to be considered transparent.
3.1 At the Interpersonal-Representational levelsnext section
The following properties between the Interpersonal-Representational levels are expected for Quechua to be considered a transparent language:
3.1.1 No cross-reference
According to Hengeveld and Mackenzie (2008:350), cross-reference occurs when person marking on the verb is capable of referring by itself. Cross-referencing pronominal argument affixes can therefore be treated as the bound expression of Referential Subacts (Interpersonal Level). The lack of transparency, i.e. the absence of an ideally one-to-one mapping from level to level, arises due to the possibility for cross-referential markers to be expanded further by a lexically realized argument, leading to the occurrence of two Referential Subacts at the Interpersonal Level referring to a single argument at the Representational Level. Quechua obligatorily cross-references the arguments that are assigned the semantic function of Actor or the syntactic function of Subject. It may also cross-reference the Undergoer argument. As in (4), arguments may be further expanded by means of proper names, which are introduced at the Interpersonal Level:
(Weber, 1989: 10)
As pointed out by Hengeveld and Mackenzie (2008:350), the overt expression of arguments by means of lexical/pragmatic units apart from cross-referential markers on the verb should be considered as cases of apposition. Within the context of FDG, apposition is a phenomenon that violates the transparent one-to-one relation between pragmatic and semantic units. Thus, Quechua can be said to lack of transparency in this respect.
3.1.2 No apposition
Apart from the type of apposition mentioned above in relation to cross-reference, Quechua does bear a more structural type of apposition in which case the constituents involved correspond to Referential Subacts. When two or more Referential Subacts correlate with a single entity at the Representational Level, they are said to be in apposition to each other. As the following examples from Quechua show, the noun and modifier marked with an agreement suffix, -ta, correspond to two Referential Subacts in apposition which will be analysed at the Morphosyntactic Level accordingly, that is as two Noun Phrases. The linking between them is done at the Representational Level, where they are represented as one semantic unit bearing a particular semantic function (Undergoer).
(Weber, 1989: 250)
Appositions need not to be yuxtaposed to one another, even if it overrides the semantic integrity of the unit denoting the entity at the Representational Level (see non-discontinuity in 3.2 below for further discussion), as the following examples show (Weber 1989:231):
The lack of a one-to-one correspondence between pragmatic and semantic units explains why appositions are not expected in transparent languages. Quechua is then not transparent in this respect.
3.1.3 No limitations on which semantic units can be chosen as predicates
In languages like English, there are certain restrictions concerning the semantic units that can be chosen as main predicates in a predication, which in this case must necessarily correspond to a verbal lexeme. As explained by Hengeveld and Mackenzie (2008: 392), in the absence of a verbal element at the Interpersonal Level or Representational Level in the realization of an Ascriptive Subact, English requires the insertion of a support-verb, such as the verbal copula ‘be’, at the Morphosyntactic Level, which will serve as the carrier of TMA distinctions strictly assigned to verbs in this language. The insertion of a support-verb at the Mophosyntactic Level to mitigate the mismatches at the higher levels of representation goes against the notion of transparency, understood in terms of a one-to-one mapping between units at all levels.
Following Weber (1989), Quechua characterizes for having two major open lexical classes, namely verbs and nouns, and a large open class of noun-adjectives. As in English, verbal lexemes are generally used as main predicates within a predication frame. When other classes of lexemes rather than verbs are used predicatively then the insertion of a verbal copula at the Morphosyntactic Level is necessary, as the following examples show (Weber 1989: 24):
However, the insertion of the copula does not occur in cases in which an adjective/noun is used as a predicate in the present tense and third person singular, and further TMA specifications are unnecessary. In these cases then, nouns and adjectives can directly be used as main predicates without the insertion of further operations at the Morphosyntactic Level, as shown in (11) and (12) below:
(Weber, 1989: 14)
Except for these cases, Quechua cannot be said to be fully transparent with respect to the type of semantic units that can be chosen as predicates, as these are generally restricted to verbal lexemes.
3.2 At the Representational-Morphosyntactic levels
The following properties between the Representational-Morphosyntactic levels are expected for Quechua to be considered a transparent language:
3.2.1 No grammatical relations
Grammatical relations are formalized in FDG as syntactic functions that pertain to the Morphosyntactic Level. Syntactic functions are relevant in those cases in which the formal properties of linguistic units cannot be reduced to the pragmatic and semantic categories and functions underlying them. As one can see, this neutralization of semantic and pragmatic distinctions goes against a transparent one-to-one mapping between units at all levels, or at least between units from a higher level (either pragmatic or semantic) to units from the lower levels (mophosyntactic and phonological), which would be expected considering the hierarchical ordering followed in a functional grammar for the formulation and encoding of linguistic forms. Grammatical relations are then not expected in transparent languages, in other words, alignment, i.e. the way in which non-hierarchically related pragmatic and semantic units map onto morphosyntactic ones, is expected to be sensitive to either pragmatic or semantic aspects, but not morphosyntactic ones, in a transparent grammar.
As for morphosyntactic alignment, it is sensitive to the syntactic functions assigned to morphosyntactic constituents, which, depending on their neutralized behaviour, may be either Subject or Object, and/or to the complexity of such constituents.
Quechua does bear morphosyntactic alignment, in which case the grammatical function Subject is relevant to account for the neutralized behaviour of morphosyntactic constituents. As explained by Hengeveld and Mackenzie (2008:325), the syntactic function Subject is relevant when there is neutralization between the Actor/Undergoer argument of one-place predication frames and, as far as Nominative-Accusative languages concern, the Actor argument of two-place predication frames, as illustrated in the following examples (Weber, 1989:176):
The relevance of the grammatical relation Subject and its neutralizing behaviour gets manifested in Quechua by means of case marking on free noun phrases, which, as shown in the examples above, corresponds to a (zero) nominative case marker. Note, however, that another case marker, -ta, serves to mark the Undergoer argument in two-place predication frames such as (6), reflecting then a direct semantic opposition that pertains to the Representational Level. The neutralizing behaviour of the grammatical function of Subject gets also manifested in person marking on verbs, in which case the Actor and Undergoer arguments are cross-referenced on the verb by means of cross-referential markers. In one-place predication frames, the cross-referencing pronominal markers are not affected by the semantic function of the argument involved, as can be seen in (13) and (14) above, but, in two-place predication frames, there is a distinctive form to mark the Undergoer argument, as the examples below show (Weber 1989:176,180), which can be left out in cases where it is conveyed by lexical or pragmatic units, as in (15) above.
As pointed out by Hengeveld and Mackenzie (2008:325), the relevance of the syntactic function Subject does not only follow from its neutralizing behaviour, but it may also be manifested by special operations leading to differential assignment of the Subject function to arguments with different semantic functions, as is the case with passivization in Nominative-Accusative languages, in which the non-Actor argument of a transitive predicate is made into the Subject. According to Weber (1989), Quechua does bear passivization, in which case a non-Actor argument is given the syntactic function of Subject receiving properties relevant to this function, such as zero nominative case marker, and triggering further special operations on the verb, as illustrated in the next examples:
As shown in (20), the differential treatment that characterizes passive clauses in Quechua is accomplished either analytically, by the addition of a participle marker, -sha, on the main verb followed by the verbal copula ka- ‘be’, or morphologically, by means of a passive marker, -ka, or durative marker, -ra, on the verb, as shown below (Weber 1989:178) :
A question that still remains to be answered is if the grammar of this language is sensitive too to the syntactic function Object, that is if it is necessary to postulate an Object function for Quechua. As explained by Hengeveld and Mackenzie (2008:326), the grammatical function of Object becomes relevant when the opposition between the Undergoer argument in a two-place predication and the Recipient argument in a two or three-place predication is neutralized, in which case they may receive the same case marker and/or trigger the same agreement patterns on the verb. On the basis of the following example (Weber 1989:180), there seems to be no need to account for an Object function in Quechua:
There is no neutralization involved in (25), that is the Undergoer, karta ‘letter’, and the Recipient, warmi ‘wife’, behave differently bearing each of them a distinctive case marker. However, it is possible to assign the Object function to the Recipient argument, with the result kown as ‘dative shift’ shown in (26):
(Weber, 1989: 11)
The neutralizing behaviour of the Object function manifests itself in the use of the same case marking for both the Undergoer and Recipient argument, but also by triggering the same person marking on verbs, as illustrated in the examples below (Weber 1989: 180):
As these examples show, the person marker ma is assigned the Object function, cross-referencing either with an Undergoer (27) or Recipient argument (28).
The phenomenon of ‘dative shift’ in Quechua justifies the relevance of the Object function in its grammar. In this sense, Quechua is comparable to English, which also allows for dative shift in three-place predication frames. In Quechua, however, the neutralizing behaviour of the Object function may also occur in two-place predication frames between the Undergoer argument and a Locative argument (understood in its broader sense).This neutralization gets also manifested by non-differential case marking, as in (30) and (32) below (Weber, 1989:182,197,190):
It is important to mention that the neutralizing behaviour of the Object function applies, according to Weber (1989), to the central varieties of Quechua, where the assignment of the syntactic function Object to Locative arguments, including further distinctions such as Recipient, spatial Goal and Benefactive, appears to be an outcome of grammaticalization that can be subjected to a scale, as the process has not (yet) led to a fully-neutralized state between Undergoer and Locative oppositions, which are still relevant and rich in the southern varieties of the Quechuan family.
On the basis of the previous explanation and examples, we can conclude that Quechua is a language in which the syntactic function Subject becomes relevant to describe the neutralized behaviour of morphosyntactic constituents that cannot be reduced to the pragmatic or semantic functions underlying them. Quechua does also bear a Directive-Indirective alignment (Hengeveld and Mackenzie, 2008:327), which means that Undergoers and Recipients behave distinctively, leading to the direct manifestation of semantic oppositions onto morphosyntactic ones. However, the grammar also allows for the assignment of Object function to Recipient arguments in certain contexts, which appear to be increasing, as far as the central varieties of Quechua concern, as a result of grammaticalization. The relevance of the grammatical relation Subject and, though to a lesser extent, Object in the grammar of this language accounts, as mentioned before, for its lack of transparency in this respect.
3.2.2 No discontinuity
In FDG, the notion of continuity is directly linked to the idea of domain integrity, which, together with the notions of iconicity and functional stability, constitutes an important principle that governs the relation between the Morphosyntactic Level and the two input levels, that is the Interpersonal and Representational Level. As explained by Hengeveld and Mackenzie (2008:285), domain integrity refers to the crosslinguistic preference for the units that belong together at the Interpersonal Level and at the Representational Level also to be juxtaposed to one another at the Mophosyntactic Level. In other words, there is a preference for one-to one relation between the hierarchical structure of the input levels and that of the Morphosyntactic Level. This principle goes then in accordance with the notion of transparency, which guarantees the easy interpretability of linguistic structures. The violation of domain integrity, which leads to discontinuity, is therefore not expected in transparent languages. However, as Hengeveld and Mackenzie (2008) point out, many languages show instances where domain integrity is overridden by other communicative strategies. For instance, the syntactic domain integrity of the Verb Phrase (Vp) is generally violated in English to signal an Interrogative Illocution. Other languages allow relative massive violations of domain integrity, relying on morphological agreement and government to signal interpersonal and representational connectedness. Quechua appears no to be an exception in this respect, as the following examples show (Weber 1989:250):
In Quechua, modifiers are placed next to their heads, the former generally preceding the latter (33), following, in that way, a principle of integrity, in which case the semantic integrity of heads and modifiers is reflected at the Morphosyntactic Level and translated into syntactic domain integrity. However, morphological agreement may allow the infringement of domain integrity, as shown in (34), where the integrity of the Noun Phrase, (Np), hatun runa ‘a/the big man’ (33), has been overridden by assigning each of its constituents accusative marking which allows to place them at any position in the Clause and makes it possible for the Addressee to understand them as one semantic unit, bearing the semantic function Undergoer. Nevertheless, the violation of integrity in (34) should be better understood as an infringement to the general cross-linguistic preference for juxtaposing, morphosyntactically speaking, elements that belong to the same semantic unit to one another, without this necessarily meaning a violation of syntactic integrity. As Hengeveld and Mackenzie (2008:298) point out, there is morphosyntactically no reason to assume that discontinuous constituents, e.g. runa-ta ‘man-ACC’ and hatun-ta ‘big-ACC’ in (34) above, belong to one particular syntactic unit, e.g. the Np hatun runa ‘a/the big man’. As a matter of fact, even when juxtaposition is to be expected at the Morphosyntactic Level due to the organization of the elements involved at the Representational Level, the choice of word order may be determined by other factors pertaining to the Interpersonal Level. In that case, the constituents in (34) carrying accusative marking can be analysed at the Interpersonal Level as two independent Referential Subacts evoking runa ‘the man’, hatun ‘the big one’, as suggested in the second translation, which are quite directly reflected at the Morphosyntactic Level as two independent Noun Phrases, though corresponding to one semantic unit at the Representational Level bearing a particular semantic function, Undergoer.
On the basis of the analysis above, we cannot arrive at a categorical conclusion with respect to the transparency of Quechua in relation to the feature in question. One possibility would be to assume that Quechua does present discontinuity, that is, a violation of domain integrity, semantically speaking due to pragmatic reasons, which does not necessarily imply syntactic discontinuity.
3.2.3 Lexeme functions and derivational processes not sensitive to nature of input
As explained by Hengeveld and Mackenzie (2008:225), languages may have specialized classes of lexemes to fulfil every single functional specification that result from distinctions made at the higher levels of representation: Interpersonal Level (Referential Subacts versus Ascriptive Subacts) and Representational Level (heads versus modifiers). For instance, to fulfil the function of head within an Ascriptive Subact, Dutch has available four different classes of lexemes, namely verbs, nouns, adjectives and adverbs (Hengeveld and Mackenzie 2008: 220-221). However, whereas only verbs can directly be used predicatively, the other three classes of lexeme will require the insertion of a verbal copula at the Morphosyntactic level to be suitable in the grammatical environment in which they occur. From this, we can conclude that in Dutch, the function of a lexeme is sensitive to nature of input, in the sense that it has specialized lexeme classes to fulfil a certain function; otherwise further operations would be required at the Morphosyntatic level to adapt a lexeme inserted into an underlying representational slot it was not meant to occupy. The opposite case is found in languages with a flexible lexeme class such as Warao (Romero-Figueroa 1997; cited in Hengeveld and Mackenzie 2008: 225), in which the same lexical item may be used as the head within a Referential Subact, as a modifier within a Referential Subact or as a modifier within an Ascriptive Subact. In this language, only verbal lexemes are distinguished from the rest as they are specialized to function as head within an Ascriptive Subact. From this, we can also conclude that the more flexible the language is with respect to the class of lexemes that can be used to fulfil a certain function, that is the less sensitive lexeme functions are to the nature of input, the more transparent it is, as there would be no need for further adaptations at the Morphosyntactic Level that would affect, after all, the ideally one-to-one correspondence between units at all levels of representation.
Quechua bears certain degree of flexibility with respect to the lexical item that can be chosen to fulfil a relevant function, though, as will be shown, such flexibility is restricted. As mentioned in 3.1.3, for a lexical item to fulfil the function of head within an Ascriptive Subact, that is for a lexeme to be used as main predicate, it should belong to the class of verbal lexemes, otherwise copula insertion will be required at the Morphosyntactic Level. Here is precisely where the restricted flexibility of this language arises. As may be recalled, in contexts where nouns and adjectives are used predicatively in the third person and present tense, no further adaptations are needed at the Morphosyntactic Level, so then they can be used as main verbal predicates. The specialization of verbal lexemes to fulfil this relevant function without further morphosyntactic adaptations is evidence to distinguish them as a particular lexical class in Quechua. In the same way, it is relevant to account for a class of nouns in this language due to their distinguishing function as heads within Referential Subacts. For verbal lexemes to fulfil this function (heads of Referential Subacts), further adaptations at the Morphosyntactic Level would be required, namely those related to derivational processes. This explains why lexical functions are not expected to be sensitive to nature of input in transparent languages. If they were, then that would imply the introduction of further specifications/adaptations at the Morphosyntactic Level with no interpersonal or representational counterparts.
A relevant question to ask at this point is if apart from verbs and nouns it is justifiable to account for adjectives and adverbs as specialized lexical classes in Quechua, that is classes with a distinguishing lexical function. According to Weber (1989), neither adjectives nor adverbs can be considered as separate lexical classes in this language. He explains that adjectives may behave as nouns, morphosyntactically speaking, as they can be used as heads of a nominal phrase and carry, accordingly, markers that pertain to nominal lexemes, namely case marking (36). Besides, there is no need to make a distinction between nouns and adjectives, as there are a number of nouns that may behave as modifiers without undergoing any morphosyntactic adaptation (37). The following examples illustrate what have just been said:
Weber classifies those modifiers that semantically behave like adjectives but morphosyntactically like nouns as a subclass of noun/adjective lexemes within the major class of nouns, his main criterion for this classification being a morpohosyntactic rather than a semantic one. However, if we take into consideration both the pragmatic and semantic parameters suggested in FDG to account for functions of lexemes and identify first the relevant functions and then study the way in which lexemes are distributed across these functions, we will be able to identify lexemes classes more consistently, not only on the basis of formal criteria, but also on the basis of criteria that correspond to the levels of representations lexical items pertain, that is the input levels: Interpersonal and Representational Level.
Bearing this in mind, I consider relevant to account for a class of adjectives in Quechua as they generally fulfil the function of modifiers within Referential Subacts, without requiring any further morphosyntactic adaptations. What is more, the semantic integration between adjectives, as modifiers, and their corresponding heads, i.e. nouns, gets expressed by means of syntactic integration too in the sense that modifiers generally precede their heads, when present, as explained in section 3.2.2. In (36) above, the adjective is not functioning as the head of a Referential Subact, a function that is generally assigned to nouns, but as a modifier whose head is absent. As Weber himself acknowledges, a construction like this is good provided that the speaker assumes the hearer knows who/what is being talked about (1998:249). From the perspective of FDG, there would be no reason to assume that the modifier in (36) is being used in that context as a lexical head; in other words, the nominal use of adjectives should not be taken as evidence to consider them as a subclass of nominal lexemes rather than a distinct lexical class.
There is, however, certain flexibility with respect to the class of lexemes that can fulfil the function of modifiers within a Referential Subact. Apart from adjectives, nouns may also be used as modifiers, as in (37) above, without this necessarily meaning that a distinction between nouns and adjectives is irrelevant, especially if we take into account that only a particular subset of nouns may be subjected to such flexibility, namely those whose semantic properties can be somehow ascribed to the entity that is being referred. It must be also noted that even in cases where nouns are used as modifiers within a Referential Subact, there is a syntactic constraint that apply to these nouns so that they can appropriately fulfil this function, that is they must precede the head they modify, otherwise it can result into an wrong reading in which the noun, instead of being interpreted as a modifier, is understood as the head of the Referential Subact. Such restriction, introduced at the Morphosyntactic Level, does not apply to lexical adjectives, which, even though they are generally used in pronominal position, can also follow their heads without this leading to ambiguity.
As for what adverbs concern, I agree with Weber that there is no reason to account for them as a separate class of lexemes in Quechua. To fulfil the function of modifiers within an Ascriptive Subact, i.e. Manner adverbs, verbs, nouns and adjectives must generally undergone further adaptations at the Morphosyntactic Level, namely by derivation, in order to be suitable in the grammatical environment in which they occur. Interestingly, some adverbs, such as sumaq ‘very’, ‘well’, and fiyupa ‘very’, ‘hard’, appear to have undergone grammaticalization and so can be regarded as adverbs in their own right, though they are part of a very limited group of lexicalised forms.
In conclusion, Quechua cannot be said to be transparent with respect to this feature. Despite its flexibility in certain contexts with respect to the kind of lexical unit that can be chosen to fulfil a given function, there are clearly specialized lexemes that bear a distinguishing function, which demand further morphological adaptations in cases where lexical items, generally belonging to other classes, are used to fulfil this function. As mentioned before, such adaptations can, for instance, be fulfilled by means of derivational processes.
3.2.4 Function marking not sensitive to nature of input
Function marking is basically used here to refer to the way nucleus-dependent relations, that is those characterizing the relationship between a predicate and its argument(s), whether that predicate be verbal, nominal or adpositional, get expressed at the Morphosyntactic Level. The expression of nucleus-dependent relations at the Morphosyntactic Level may appear on the nucleus, on its dependents, on both or on neither. According to Nichols (1986), languages tend to use the first or second type of marking, that is they are predominantly either head-marking or dependent-marking.
In a transparent grammar, function marking, whether predominantly of the nucleus or dependent type, is not expected to be sensitive to the nature of the linguistic unit that serves as input, as it would mean the introduction of further specifications/operations at the Morphosyntactic Level that would affect the ideally one-to-one relation between units at all levels of representation.
According to Nichols (1986:72), Huallaga Quechua can be regarded as a double-marking language as it marks several of its constructions twice, on both the head and the dependent, as shown in the following examples:
(Weber, 1989: 180, 57)
(Nichols, 1986: 72)
In cases like (39), there is no need for the arguments cross-referenced on the verbal predicate to be further expanded by means of pronouns, but in cases in which arguments need to be expressed they will always be subjected to dependent-marking processes, namely by means of case markers as can be seen in (39), which should be better referred to as clitics due to their behaviour as bound morphemes, phonologically attached to a host, which are not sensitive to the nature of the linguistic unit they mark, which may well be a single form (39), a phrase or a whole clause, as illustrated in the examples below:
These examples clearly show the transparency of Quechua with respect to function marking. However, double-marking in itself as reflected in (39), (40) and (41) above could be argued to lack of transparency in the sense that the semantic function which expresses the relation between the nucleus and its dependents at the Representational Level is morphosyntactically marked twice. As mentioned before, there is no need for the dependent arguments in (39) to be further expanded by means of pronouns, therefore function marking on the verbal nucleus by means of cross-referential pronominal markers may be sufficient. However, in nominal phrases such as (40), function marking needs to be expressed on both the head noun and its dependents, which leads in this case to double possessive marking, expressed by means of a genitive case marker and a possessive person marker both alluding to the relation between possessee and possessor, whether the possessee is alienable or inalienable. The same holds for adpositional phrases such as (41), in which case both the head and the dependent are marked, leading to double possessive marking, as long as the dependent argument is human, otherwise only the head is marked, as shown below (Nichols 1986: 73):
Whether or not the relation between nucleus and dependents is morphosyntactically marked twice, and whether or not function marking is sensitive to semantic aspects such as that of animacy, cases such as those in (40), (41) and (44) cannot be regarded as lacking of transparency in terms of function marking as it clearly obeys specifications made at the Representational Level. Function marking in Quechua is semantic in nature and so is not sensitive to the (morphosyntactic) nature of the linguistic units that serve as input. Thus, this language can be considered as transparent with respect to function marking.
3.3 At the Morphosyntactic-Phonological levels
The following properties between the Morphosyntactic-Phonological levels are expected for a language to be considered as transparent:
3.3.1 Phonological phrasing and morphosyntactic phrasing run in parallel
In keeping with the one-to-one correspondence between units at all levels of representation that characterizes language transparency, it is expected that the morphosyntactic encoding of the input formulated at the higher levels runs in parallel with its corresponding phonological encoding, which provides a representation that serves as input to the Output Component, responsible for articulation. However, as stated by Hengeveld and Mackenzie (2008:427), the relation between these two encoding levels is only of partial parallelism, especially if we consider that the phonological representation of a given utterance may be influenced by many other factors, rather than purely morphosyntactic ones.
Unfortunately, the descriptions provided in grammars of the different varieties of Quechua, including the variety in question on this paper, do not provide with an exhaustive characterization of its prosodic phonological features. As stated by Cerron-Palomino (1987: 128), suprasegmental features in special, such as those related to intonation, accentuation and rhyme, have been the least studied phenomena in the phonology of Quechua and so they definitely deserve further investigation. The lack of relevant data in this respect does not allow for an exhaustive analysis of phonological phrasing, and its corresponding parallelism with morphosyntactic phrasing, in this language. As most accounts on Quechua phonology are related to segmental features, especially to stress patterns, which are regarded as the criterial property of Phonological Words, I will refer to them and attempt to account for their relevance with respect to the parallelism between morphosyntactic and phonological encoding.
As explained by Hengeveld and Mackenzie (2008: 443), the correlation between morphosyntactic and phonological words is best in isolating, agglutinating and fusional languages. As an agglutinating language, Quechua does present a good correlation between morphosyntactic and phonological words, with primary stress being placed on the penultimate syllable of the morphosyntactic word, as shown in the following example (Weber 1989:55), in which primary stress falls on the syllable /ku/:
According to Weber (1989:457), this stress pattern is not always followed, especially in cases in which stress is used for emphasis, in which case it usually falls in the last syllable. The violation of this rule appears, however, to have a pragmatic trigger, therefore it is justifiable to account for the Phonological Word as a relevant category in which stress assignment takes place.
Hengeveld and Mackenzie (2008:446) also explain that the correlation between morphosyntactic and phonological words becomes problematic when clitics are at stake. Clitics are morphemes generally treated as words at the Morphosyntactic Level and are represented as such according to the templates relevant to that level (for further explanation see 4.1.5 below). At the Phonological Level, however, a clitic is treated as part of the Phonological Word where it operates, which inevitably affects the parallelism between morphosyntactic and phonological phrasing. In the grammar of Quechua, the use of Clause and Phrase-layer clitics is a prominent feature, and so lack of correspondence between morphosyntactic and phonological phrasing, where clitics are involved, is expected, as shown in the following example (Weber 1989:470):
This structure constitutes an Adposition Phrase (Adp), integrated by morphosyntactic words, namely a Nominal Word (Nw) and a Grammatical Word (Gw), and is represented as such at the Morphosyntactic Level:
The locative marker chaw is represented as a Grammatical Word (Gw) within the Adposition Phrase (Adp). At the Phonological Level, however, this belongs to a single Phonological Word (pw), which is, in accordance to a general rule for Phonological Words in Quechua, stressed on the penultimate syllable, as shown in the following phonological representation:
The lack of correspondence between morphosyntactic and phonological phrasing in this example shows a partial parallelism between these two encoding levels. Thus, Quechua can be regarded as partly transparent in this respect.
3.3.2 Phonological weight does not influence morphosyntactic placement
In an ideal transparent language, neither morphosyntactic nor phonological factors are expected to determine the ordering of clause constituents. On the contrary, alignment is expected to reflect the organization of the higher levels: the Interpersonal and the Representational Level, which would go in accordance with the ideally one-to-one correspondence between processes of formulation and encoding operations.
In Quechua, the ordering of constituents may obey pragmatic factors, though in neutral contexts, alignment is generally the result of semantic and syntactic function assignment. As mentioned before, Quechua has a syntactic function Subject, which tends to occupy a clause-initial position. The placement of U and R arguments appear not to be attributable to syntactic factors, but semantic ones, as this is generally the Undergoer the argument that precedes the predicate, whereas the Recipient tends to precede the Undergoer. Thus, whether determined by syntactic or semantic factors, the placement of constituents in Quechua normally corresponds to that of a predicate-final language. However, when constituents are phonologically heavy, they can be placed after the predicate, as is commonly the case of sensory verb complements such as the one in (49) which can be shifted to the right as illustrated in (50):
(Weber, 1989: 290)
In the same way, nominal modifiers, which are normally placed before the nominal head, may be placed Clause-finally due to their being phonologically heavy. This is what generally occurs with relative clauses in complex constructions such as that in (51) which appears in (52) in clause-final position:
(Weber, 1989: 282)
Quechua cannot be regarded as transparent with respect to the morphosyntactic organization of Clause constituents, as this may not only be determined by pragmatic or semantic factors, but also by morphosyntactic or phonological ones.