Volume 14 Issue 1 (2016) DOI:10.1349/PS1.1537-0852.A.469 Note: Linguistic Discovery uses Unicode characters to represent phonetic symbols. Please see Optimizing Display for requirements to accurately reproduce this page. Relative clauses in Upper Necaxa Totonac: Local, comparative, and diachronic perspectives[1] David Beck University of Alberta Relativization strategies
in the Totonacan family are largely undescribed, but detailed examination of
one of the languages in the group, Upper Necaxa Totonac, reveals the presence
of both externally- and internally-headed relative constructions. Also of note
is the presence of relativizers that mark the animacy (human/non-human) of the
head of the relative construction. This paper will show that, while
phylogenetic evidence clearly demonstrates the relativizers to be descended
diachronically from interrogative pronouns, they are best treated
synchronically as complementizers, an analysis that follows directly from the
presence of internally-headed relative constructions. Totonacan languages are spoken by approximately 240,000 people (INEGI 2010) living in an area of east-central Mexico centred on northern Puebla State and including adjacent parts of Hidalgo and Veracruz (see Figure 1; languages dealt with directly in this paper are shown in red). The family is generally considered an isolate; however, recent work has suggested links to Mixe-Zoque (Brown et al. 2011) and Chitimacha (Brown et al. 2014). Although the family has only recently become the object of serious investigation and description, the focus has been largely on its (admittedly spectacular) morphology; little has been written about syntax, and even less about the structure of complex clauses. Relative clauses in particular seem to have been given shortshrift—which is surprising, given that from what we do know about them they seem to have some unusual properties. Consider the example in (1) from Upper Necaxa Totonac, the language for which we currently have the most data on relativization:[2] Figure 1: Location of Totonacan Languages
Here we see a typical relative construction with an overt nominal, ʔawa̰čá̰n ‘boys’, being modified by a subordinate clause. The modified noun, or head of the relative clause, is external to the relative clause itself and is linked to it by a relativizer—in this case, the animate (human) relativizer, tiː. The head of the relative clause corresponds to the subject of the embedded verb, which is elided or gapped, and the relative construction as a whole can be considered subject-centred (i.e., the head of the clause is co-referential with the subject of the subordinated verb). Upper Necaxa also allows headless relative clauses (a.k.a. free relatives), as in (2)
In a construction like this, the head of the relative clause is elided. The subject in (2) is also gapped, but the relative remains subject-centred and the referent of the construction as a whole is the subject of the embedded clause. Such constructions in Upper Necaxa function quite happily as arguments of verbs, like any other noun phrase, and are frequent in text. In addition to the constructions in (1) and (2), Upper Necaxa also allows a third possibility, shown in (3):
In (3), there is no overt nominal head external to the subordinate clause. Instead, what appears to be the head, ʔawa̰čá̰n ‘boys’, is found inside the embedded clause, in this case immediately following the verb. This type of construction is what is referred to as an internally-headed relative clause. Cross-linguistically, internally-headed relative clauses are comparatively rare, and they are even rarer in languages where, as in Upper Necaxa, they are an infrequent, as opposed to a dominant, type of relative clause (Dryer 2013). Another notable feature of the constructions in (1)–(3) is the element tiː, which is used exclusively in relative clauses with animate (human) heads and which has a counterpart, tuː, used with inanimates (non-humans). While these are typically labeled “relativizers” in grammatical descriptions of Totonacan languages, this term is actually (and probably deliberately) vague in that it does not make it clear whether the relativizers are to be understood as subordinators or complementizers—linking elements in the clause signalling/licensing subordination—or as relative pronouns, pronominal elements standing in for the relativized argument of the embedded verb. The fact that they encode animacy, a characteristic of nouns, suggests that they might in fact be pronouns, as does the fact that in several of the languages in the family, like Upper Necaxa, they are homophonous with the interrogative pronouns, tiː ‘who?’ and tuː ‘what?’. On the other hand, the relativizers are not, as canonical relative pronouns are, inflected for case, nor is animacy a morphological or inflectional category that manifests itself in other kinds of pronouns or in any other area of the grammar. So, not only does Upper Necaxa have the cross-linguistically unusual internally-headed type of relative clause, there is also an analytical question at play here as to the exact nature of the relativizers that appear in the relative constructions. As this paper will argue, these two phenomena are linked, and the presence of the internally-headed construction in (3) can be used to show that the relativizers are best not analyzed as relative pronouns in the usual sense, but should instead be described as complementizers that, perhaps unusually, agree with a semantic feature of the head of the relative constructions that they introduce. This agreement, naturally, has a diachronic explanation which is revealed by a comparative examination of relativization strategies across the family. The remainder of this paper will proceed as follows. Section 1 offers a brief description of Upper Necaxa Totonac, focusing on the structure of the simple clause, while Section 2 gives a detailed account of relativization in that language. In Section 3, I survey what is known about relativization in the rest of the family. The data available is rather sketchy and there are only one or two dedicated descriptions of relative clauses in other Totonacan languages; nevertheless, a reasonably coherent picture of the family, and how the different branches of Totonacan compare to each other and to Upper Necaxa, can be arrived at by culling through texts and other sources. Section 4 will assess the implications of the comparative picture for the diachronic development of the different relativization strategies, and will propose an analysis of the relativizers that supports the claim that they are best described as complementizers, at least in Upper Necaxa and the other languages in the family that allow internally-headed relative clauses. 1. Upper Necaxa Totonac
Upper Necaxa Totonac is a member of the Northern Totonac branch of the Totonacan language family, spoken by 3,293 people (INEGI 2010) in the Necaxa River Valley in northern Puebla State, Mexico. Like all Totonacan languages, Upper Necaxa can be characterized as polysynthetic and primarily head-marking in the sense of Nichols (1986). Verbs agree in person and number with their syntactic subjects; transitive verbs also agree with their objects:
In (4) we see the verb paːɬ’aː ‘X cuts Y’s belly deeply’ bearing the first-person singular subject prefix ḭk- and the plural object prefix, kaː-. The object-noun itself, pá̰šnḭ ‘pig’ is unmarked for number, which is an optional category for nouns, the preferred locus for number-marking in the clause being on the verb. The categories of person and number of object are marked separately, as in (5):
Here, the person of the object is encoded by the suffix, -n ‘second-person object’, while the plural number of the object is indicated by the same prefix, kaː- ‘plural object’, that we saw in (4). Third-person singular subjects, third-person objects, and the singular of objects are morphological zeros (and in this paper will not be included in the glossing to avoid cluttering the presentation). In multivalent clauses, verbs can also agree in person with a second object, if that object is first- or second-person:
This example shows the trivalent verb sta̰ːmaškíː ‘X sells Y to Z’, used in this context to refer to the exchange of a girl for a dowry, agreeing with both a second-person and a first-person object. Upper Necaxa is a primary-object language in the sense of Dryer (1986), making the second-person recipient in this clause the primary object and the first-person theme the secondary object.[3] Copular clauses can have either nominal (7) or adjectival (8) predicate complements:
The copula is based on the verb wan ‘be’ and is immediately preceded by its complement, as it is in other Totonacan languages. In the past tense, the copula is invariably inflected in the past perfect, and in the future it takes the future imperfective form nawán. It is generally omitted in the present tense, as here in (8), but may optionally be realized in the present imperfective form, wan. Also of note in (8) is the head-marked possessive construction ḭšnaːná̰ Xiwán ‘Juan’s grandmother’ in which the possessed NP is inflected for the person of the possessor. The plurality of the possessor is optionally indicated by the suffix -ka̰n (e.g., ḭšnaːna̰ká̰n Xiwán ʔeː Tséntṵ ‘Juan and Rosendo’s grandmother’). As in the rest of the family, constituent order in Upper Necaxa clauses is extremely flexible and is governed primarily by information/communicative structure (Vallduví 1992; Mel’čuk 2001). The communicatively unmarked order is verb-initial, placing Upper Necaxa in Dryer’s (1997) VS/VO category; however, essentially any order of verb and arguments is possible for a given sentence in the appropriate context, as shown in (9):
Although utterances in which all arguments are realized as NPs are relatively uncommon, sentences with all possible orders can be found in the corpus and are accepted during elicitation. Copular clauses show flexible ordering of subject and predicate phrase, but the copula, when present, is preceded immediately by its complement in all but a few examples:
The final order shown here occurs occasionally, most often in the context of elicitation-by-translation of Spanish sentences, but it is usually rejected or corrected by speakers when offered. It seems likely to be an effect of calquing from Spanish (cf. era bruja esa mujer). Like predicate complements, most types of adverbial are also required to precede the verb:
Ideophones as well as dynamic, configurational, descriptive, and manner adverbs obligatorily precede their verbal heads, while temporal and locative adverbials like naḭščík ‘in her house’ in (11) may be either pre-verbal or post-verbal (Beck 2008). Constituent ordering is also relatively more fixed in information questions (a.k.a. wh-questions), which require that the interrogative pronoun be the first element in the clause:
Questions can be formed on any clausal constituent. In addition to tiː ‘who?’ and tuː ‘what?’, Upper Necaxa has the interrogatives xa̰ː ‘where?’ and xá̰ːkšnḭ ‘when?’, the former but not the later also being homophonous with the corresponding relativizer. 2. Relative clauses in Upper Necaxa Totonac
Relative clauses in Upper Necaxa Totonac are first described in Beck (2004), where they are characterized as being externally-headed or headless, with a gapped argument inside the embedded clause. Relatives are introduced by what is characterized as a relativizer, either tuː or tiː, depending on the animacy of the head of the construction. Example (14) illustrates a relative clause with an inanimate nominal head:
The head of the relative clause is škaːn ‘water’ and corresponds to the gapped object of the intransitive verb wanḭ́ ‘X calls Y Z’. The clause follows its head and is introduced by the non-human relativizer, tuː. The relativizer appears on the left edge of the clause and in most cases immediately follows the modified noun, although there are a few examples from texts in which the relative clause is separated from its head, as in (15):
The relative clause in (15), tu: tsa̰x šmaːn nataliːwá púɬkḭ ‘which they would use to drink pulque’, modifies the noun ḭštumiːnká̰n ‘their money’, the secondary object of the embedded verb, but is separated from it by the NP ḭšpuskaːtká̰n ‘their wives’. Example (16) shows a relative clause introduced by the human relativizer, tiː:
In this example, the head of the relative clause, kɾistiánṵ ‘person’, corresponds to an object of the embedded verb leːnḭ́ ‘X takes Y away from Z’. The form of the relativizer indicates that the head of the clause is human, but gives no indication of its number. Relatives can also be formed from copular clauses, as in (17):
As in the previous examples, the relative clause immediately follows the noun it modifies and is introduced by a relativizer, in this case tiː, which encodes the animacy of the head. Headless relative clauses are illustrated in (18) and (19):
In both of these sentences, the matrix verb is wiːɬ, which literally means ‘be sitting’ but is also used as a general existential locative (≈ Eng. ‘be there’). The headless relatives in both examples take the role of matrix subject. The relative clause in (18) is a subject-centred clause, introduced by tiː, with an animate referent, while that in (19) is an object-centred clause, introduced by tuː, with an inanimate referent. Headless relatives formed on copular clauses are also attested, both with and without an overt copula:
Such clauses are necessarily subject-centred, as relative-clauses centred on a predicate-complement seem impossible, or at least difficult to conceive of.[4] Research subsequent to 2004 has shown that, in addition to externally-headed and headless relatives, Upper Necaxa also has internally-headed relative clauses like those in (22):
In (22), the head of the relative clause, tsṵma̰xáːn ‘girls’, corresponds to the plural subject of the antipassivized verb, s’o̰ʔanán ‘X gives hugs’, and occurs inside what is otherwise identical to a headless relative construction such as that in (18). That these constructions are indeed internally-headed, rather than right-headed, can be seen in (23):
Here the head of the construction is ʔawa̰čá̰n ‘boys’. It functions as the subject of the embedded verb, li:tatsḛ́ʔ- ‘X hides behind Y’, which is inflected for a third-person plural subject, and appears followed by the verb’s object, ḭstsḭːká̰n ‘their mother(s)’. Such constructions, while not abundant, occur naturally in spontaneous speech and are accepted readily in elicitation. Thus, it seems that Upper Necaxa allows for three types of relative clause—externally-headed, headless, and internally-headed. Other than the presence of the relativizer, Upper Necaxa relative clauses seem virtually identical to matrix clauses and show the same flexibility with respect to the ordering of arguments—although there is a dispreference for arguments of the embedded clause in pre-verbal position. Such constructions are difficult to elicit, but examples such as (24) are found in texts:
In this example, the primary object, ḭštéːkṵ škaːn ‘the spirit of the water’, precedes rather than follows the verb. Fronting of arguments in general seems to be associated with focalization (Mel’čuk 2001), and in the examples where an element inside a relative clause is fronted there is a clear emphatic or focal attention on that element. In (24), this is reflected by the presence of the intensifying clitic =tṵnká̰ and the Spanish gloss given by the consultant, el mero dueno del agua ‘the very/the one and only spirit of the water’. Similarly, adverbials also maintain the pre-verbal position found in matrix clauses:
As
shown in (25), the manner adverb obligatorily precedes the verb, while the
locative is permitted in either pre-verbal or post-verbal position. These rules
of constituent order apply both when the clause is headless and when it is
internally, rather than externally, headed. In terms of accessibility to relativization, Upper Necaxa allows for the relativization of elements of all applicable ranks on the Accessibility Hierarchy (Keenan and Comrie 1977). Examples of relatives centred on subjects, primary objects, secondary objects and possessors can be found in texts.[5] The sentence in (26) shows a subject-centred relative clause:
Here, the head of the relative clause lakstín ‘children’ is coreferential with the subject of the embedded verb, taá̰n ‘they go’. A headless subject-centred relative clause is shown in (27):
The headless relative here is based on the verb laní ‘X happens affecting Y’, which is formed by adding the benefactive applicative -nḭ́ to the light verb la ‘X happens’. The gapped argument corresponds to the subject of the verb, the event that happens and affects Y. Internally-headed subject-centred relative clauses are also attested as in (28):
Relatives formed on copular clauses can also be internally-headed:
Not unexpectedly, then, it appears that all subjects are accessible to relativization. In terms of object relations, both primary and secondary objects are accessible. In (30), we see a relative clause centred on the sole (primary) object of the transitive verb la̰ʔatíː ‘X likes Y’:
(31) also shows a headless primary-object centred relative clause:
The reference of the relative clause in (31) is to the object of the verb ča̰n ‘X plants Y’. In this example, the verb ča̰n is inflected for the indefinite voice, which suppresses the expression of the subject but does not promote a first- or third-person object to subject position (Beck 2004, 2016). Verbs in this voice have either an indefinite subject (≈ Eng. indefinite they or you) or a reflexive reading. An internally-headed primary-object centred relative clause is shown in (32):
The head of the relative clause is the primary object of ɬawa ‘X does Y’, kintaskuxút ‘my job’, and it appears in this sentence inside the embedded clause. An externally-headed secondary-object centred relative clause is shown in (33):
In (33), the head of the relative clause is the applied, secondary object of the verb liːlakaɬtaŋtayá ‘X pulls Y taut with Z’, formed from the verb lakaɬtaŋtayá ‘X pulls Y taut’ with the instrumental applicative prefix liː-. Note that in this example the head of the relative clause is a numeral classifier construction used anaphorically to mean ‘something’, although unlike the English something in the gloss, the expression is not necessarily indefinite (that is, in the right context the relative might have been translated as ‘the thing that they could use to pull it tight’). A headless secondary-object centred relative clause is shown in (34):
The referent of the clause here is the animal being named, which corresponds to the applied object of the verb waní ‘X says Y to Z’, the translation equivalent of English call/name, which is formed from the verb wan ‘X says Y’ by adding the benefactive applicative suffix -nḭ. The same verb can be seen in an internally-headed secondary-object centred relative clause in (35):
In (35) the head of the relative is the object being named, ḭšliːčḭkiːtawíːɬ ‘their town’, which is realized inside the embedded clause, to the left of the primary object, the name Patla. The only type of construction that is clearly a relative clause centred on an adjunct is the locative-centred relative. Locative-centred relatives are introduced by xa̰ː ‘where’, as in (36):
Perhaps unsurprisingly, locative-centred relatives are more frequently headless, as in (37):
While locative relatives seem to share some of the properties of relative clauses introduced by tiː and tuː, little more will be said about them in the remainder of this paper. Likewise, there are temporal constructions introduced by a̰kšní ‘when’ that look very similar to relative clauses, but are not clearly attested modifying nouns. These will also be left aside in this discussion.[6] Moving further down the Accessibility Hierarchy, a look through the corpus finds possessor-centred relative clauses, like those in (38) and (39):
Headless and internally-headed versions of relatives of either the possessor- or comparative-centred types have not been found or elicited, though it seems entirely possible this is simply a gap in the data rather than a specific grammatical restriction, like that in (40):
3. Relative clauses in Totonacan
As we’ve seen in the preceding section, Upper Necaxa takes a rather free-wheeling approach to relativization, allowing externally-headed, headless, and internally-headed constructions, and permitting relativization along nearly the full length of the Accessibility Hierarchy. Constituent order in relative clauses continues to be flexible, though there seem to be discourse conditions on “fronting” NPs to a position between the verb and the relativizer. The relativizers themselves make a human/non-human animacy distinction and, in this respect (showing some sort of agreement with the head of the relative construction), resemble pronouns. Taken together, this is an unusual typological profile, particularly with respect to the presence of internally-headed relative clauses. According to Dryer (2013), internally-headed relative clauses are rare, found in only 63 languages in his 824-language sample (7.6%) and occurring as a non-dominant (less frequent) type in only 10 (1%). The fact that internally-headed relatives in Upper Necaxa are not in any sense nominalized contradicts a universalist claim made in de Vries (2005: 18) that such structures are always nominalized, and are only found in languages that have a similar type nominalized non-relative clauses (which Upper Necaxa lacks). Likewise, de Vries (2005: 19) claims that internally-headed relatives should only be found in languages with RelN and NDet order, neither of which is true of Upper Necaxa, although a loose correlation of RelN order with the presence of internally-headed relatives clauses appears to be borne out by the data in Dryer (2013). In the (surprising) absence of large-scale, quantitative typological studies of the Accessibility Hierarchy, it is not possible to determine how unusual access to such a wide range of elements on the Hierarchy is, though presumably it is relatively infrequent given the number of other possible language types foreseen by the Hierarchy. Given the interesting profile of Upper Necaxa relative clauses, it seems worthwhile at this point to turn to other languages in the family, with an eye towards seeing what, if any, of these features are shared by other languages in the group, and if the familial pattern sheds any light on the nature of the Upper Necaxa relativizers. 3.1 The Totonacan language familyThe internal structure of the Totonacan family is still not well understood, though, as shown in Figure 2, it is generally agreed that Totonacan languages can be divided at the highest level into two branches, Tepehua and Totonac. Tepehua is considered to consist of three languages—Tlachichilco, Pisaflores, and Huehuetla, while Totonac is more highly ramified and contains an as-yet-unknown number of languages. The most basal division in the Totonac branch of the family is between the geographical outlier Misantla Totonac and the remaining Central Totonac languages. Central Totonac has traditionally been held to consist of three sub-groupings—Northern, Sierra (a.k.a. Highland), and Lowland (Papantla). Beyond this, the relations become murkier, and there have been various proposals for grouping together Northern and Sierra against Lowland (García Rojas 1978), Lowland and Northern against Sierra (MacKay and Trechsel 2014, 2015), and Northern against Lowland-Sierra (Ichon 1969; Davletshin 2008; Brown et al. 2011). Presently, the weight of the evidence, particularly the lexical evidence, seems to favour the last of these. There are, in addition, further uncertainties, and particularly problematic are the affiliations of Cerro Xinolatépetl and Filomeno Mata. While Cerro Xinolatépetl is not spoken in an area contiguous with Lowland-Sierra languages (see Figure 1 above), lexical evidence suggests its affinity is with these rather than with the adjacent Northern group. Filomeno Mata is grouped by MacKay and Trechsel (2014) in the Northern branch, based on shared morphological characteristics, and it does seem to be the case that this language also shares a few lexical forms with the Northern languages; however, the bulk of the lexical isoglosses, as well as statistical measures of lexical similarity, seem to point to a closer affiliation with Lowland-Sierra.
In terms of available descriptions of relativization strategies in Totonacan languages, the pickings are rather slim. For Tepehua, Smythe-Kung (2007) offers a sketch of relativization in Huehuetla, and some mention of relative clauses in Tlachichilco is made at various points in Watters (1988). In Northern Totonac, relative clauses are mentioned in passing in the pedagogical grammar of Apapantilla prepared by Reid (1991) and there are numerous unanalyzed examples in a lexical database for the language prepared by Reid et al. (n.d.). Relative clauses in Sierra languages are dealt with indirectly for Huehuetla Totonac in Troiani (2004) and for Coatepec in McQuown (1990). E. Aschmann (1984) presents a much more thorough and detailed description of all types of relatives in Zapotitlán Totonac, and additional examples from this language (again, unanalyzed) can be found in the lexical database prepared by H. Aschmann (n.d.a). Likewise, a large number of unanalyzed sentences that contain relative clauses can be found in the examples in H. Aschmann’s (n.d.b) lexical database for the Sierra Totonac language Coyutla. Unanalyzed translations of Spanish sentences containing relative clauses can be found for Misantla Totonac in MacKay and Trechsel (2005), and one or two examples of relatives can be found in analyzed texts in MacKay (1999) and MacKay and Trechsel (2012b). Beyond this, information on a few other languages can be gleaned from the interlinearized texts in Levy and Beck (2012) which contain examples of relative clauses from the Totonac languages Cerro Xinolatépetl, Filomeno Mata, Olintla, Ozelonacaxtla, and Cerro del Carbón, and for all three Tepehua languages. In the sections that follow, I will summarize what can be extracted from this fragmentary data, beginning with Tepehua in section 3.2 and then moving on to the Totonac group in section 3.3. 3.2 TepehuaThe most detailed description of relative clauses in Tepehua languages is found in Smythe-Kung’s (2007) doctoral dissertation on Huehuetla Tepehua. Smythe-Kung gives examples of both externally-headed post-nominal (41) and headless (42) relative constructions:
In both cases, these are subject-centred relative clauses introduced by a relativizer, huː, which is homophonous with the article that is found introducing noun phrases such as the head of the relative clause in (41), huː papaːnin ‘the men’. The relativizer, like the article, is invariant and shows no agreement for number or animacy. Although Smythe-Kung makes no explicit reference to constituent order within the clause, none of her examples have fronted arguments and nothing precedes verbs except adverbial elements such as particles and the locative phrase seen in (41). The issue of noun-phrase accessibility in Huehuetla is complicated by the fact that the terminology used in the description of this language does not map directly onto the categories traditionally used in the discussion of the Acessiblity Hierarchy. According to Smythe Kung (2007: 592), externally-headed relative clauses can be formed on subjects, direct objects, indirect objects, oblique objects, and locative adjuncts. For Smythe Kung, direct objects are the single objects of transitive verbs, and indirect objects are the recipients in ditransitive verbs such as that in (43):
The embedded verb in (43) is štaqni ‘X give Y to Z’ which contains a fossilized instance of the dative applicative, -ni. Other applicatives such as t’aː- ‘comitative’ add what Smythe Kung calls “oblique” objects, as in (44):
However, objects like these are oblique only in the semantic sense that they are not part of the basic valency of the verb; syntactically, it is not clear that they are oblique objects in the usual meaning of the term and, based on descriptions of applied objects in Tlachichilco (Watters 1989), they would probably not be considered “oblique” by Keenan and Comrie (1977), although they would still rank below direct objects and above locative adjuncts. As in Upper Necaxa, locative relatives in Huehuetla Totonac make use of a separate element meaning ‘where’ to introduce the subordinate clause:
In Huehuetla, however, huntaː ‘where’, is not homophonous with the interrogative word for questioning locations, ta̰nč ‘where?’ (Smythe Kung 2007: 567), and in this it differs from Upper Necaxa Totonac. There is less information available about headless relatives in Huehuetla. Although Smythe Kung (2007: 592) writes that these are confined to subject-centred constructions, there is some evidence that other types of headless relatives are possible, as shown by the following example from the text in Smythe Kung (2012):[7]
Here, the referent of the headless relative clause corresponds to the object of the verb ʔulaː ‘X places Y’. It may be that the constraint against headless relatives formed on non-subjects noted by Smythe Kung is more of a dispreference than an absolute prohibition, and additional examples may surface in future investigations. Relative clauses in Tlachichilco Tepehua have not been described in detail, although a few examples and some structural observations about them can be found in Watters (1988: 120, 461–2, 467–72). According to Watters, both externally-headed (47) and headless (48) relative clauses can be formed from any “direct argument” of the verb:[8]
In both of these examples, the target of relativization is an applied object added to the valency of the verb by an applicative—puː- ‘means, path’ in (47) and t’aː- ‘comitative’ in (48). Watters (1988) also presents examples of subject- (p. 472) and direct-object centred (p. 462) relative clauses, but does not mention the possibility of forming relative clauses on locative expressions. As in Huehuetla, in Tlachichilco relative clauses are introduced by an element, yuː, that is homophonous with a determiner also used to introduce noun phrases—although, unlike Huehuetla, Tlachichilco has other determiners as well, and the one used with relative clauses is textually less-frequent (Watters 1988: 466). This also seems to be true of Pisaflores Tepehua, judging by the text in MacKay and Trechsel (2012a), where the cognate element, yuu, is glossed as ‘that’ or ‘the one that’ when introducing relative clauses, both externally-headed (49) and headless (50):
Both of these relatives, one subject-centred (50) and one objected-centred (49), are introduced by yuu (glossed here as a relativizer to facilitate comparison), which (50) shows to be invariant for number. Judging from some unanalyzed examples in MacKay and Trechsel (2010), Pisaflores may also have relative clauses introduced by the determiner ʔan, though this remains to be confirmed by further investigation. In summary, then, Tepehua languages use determiners or elements cognate with determiners to introduce both externally-headed and headless relative clauses which can be centred on any type of object; the presence of locative-centred relative clauses has only been substantiated for Huehuetla Tepehua. There is no evidence for internally-headed constructions in any of these languages, and all attested examples of relatives thus far follow predicate-initial constituent order in the embedded clause. 3.3 TotonacThe Totonac branch of the family is somewhat larger and more ramified than the Tepehua branch. In the sections below I will begin with the most divergent Totonac language, Misantla, and then move on to the Central group, divided up into Northern, Cerro Xinolatépetl, Filomeno Mata, Lowland, and Sierra subgroups. Misantla TotonacRelatively little is known about relative clauses in Misantla. While there is a very good grammar of Misantla Totonac (MacKay 1999), this work is focused primarily on the phonology and morphology of the language and does not touch at all on relativization. There are, however, some unanalyzed examples in MacKay and Trechsel (2005) that are given as translations of sentences that contain relative clauses in Spanish. These are all externally-headed constructions and most are introduced by an element glossed as a determiner in the texts in MacKay and Trechsel (2012b), as in the example shown in (51):[9]
The relative clause in (51) is centred on the applied object of the verb maqniini ‘X kills Y affecting Z’ licensed by the dative applicative -ni. It not clear exactly what rank on the Accessibility Hierarchy to assign this object, but MacKay and Trechsel (2008) argue that objects in Misantla are symmetrical in the sense of Bresnan and Moshi (1990). Thus, presumably, if one type of object can be relativized then they all can, and accessibility to relativization extends at least as far down the hierarchy as the lowest-ranked object. The determiner in this construction, hun, is an obvious cognate of the Huehuetla Tepehua article huː. There are also two examples of translations of Spanish sentences with externally-headed relative clauses where the corresponding elements in Misantla are not introduced by a determiner. If these are relative clauses, one—shown in (52)—would be subject-centred, and the other object-centred:
There is also a sentence in the text in MacKay and Trechsel (2012b: 140–141, line 90) that could be a determiner-less subject-centred relative clause, though other interpretations of the structure are possible. Another possibility is that structures like that in (52) are in fact internally-headed relative clauses with a fronted argument (cf. the Zapotitlán example in 93 below), a hypothesis which merits further investigation. It appears from a single example in the text at the end of MacKay (1999) that it may be possible to form headless relatives introduced by the determiner as well:
However, what appear to be headless relative clauses formed by a somewhat different strategy are also attested in the text in MacKay and Trechsel (2012b):
If these are indeed relative clauses, we have a headless subject-centred relative clause in (54) and an (instrumental) object-centred construction in (55). Both are introduced by tuː(t) ‘what’, the cognate of the Upper Necaxa non-human relativizer. Thus, it seems possible that Misantla uses different relativizers for externally-headed constructions (and, potentially, internally-headed constructions if that is the correct interpretation of (52) above). However, given that both of these examples here involve a matrix verb of speaking (maːsiyṵ ‘X recounts Y’ and wan ‘X says Y’), another possibility is that we are looking at sentential complements of verbs in the form of “embedded questions”—subordinate clauses introduced by interrogative words subcategorized for by a certain class of verb. These would not be relative clauses in the traditional sense in that they are not adnominal modifiers (a role filled in Misantla by the determiner-headed constructions seen in (51) above), but it seems like a very small step, both semantically and syntactically, between the use of constructions like these in the more restrictive context (complement of a specifc type of verb) to a less restricted use as an argument of verbs in general, making them the functional equivalent of headless relative clauses in sentences like the Upper Necaxa example in (18) above. Where exactly on this cline the Misantla tuː-constructions are is still uncertain. Even so, it does seem to be the case that Misantla occupies an intermediate position between Tepehua, which makes exclusive use of a determiner in relativization, and Upper Necaxa (and other Totonac languages, as we’ll see below), which has taken an additional step and extended the use of tiː/tuː to adnominal relative constructions. We will return to this issue in section 4. Northern TotonacThe only member of the Northern group of the Central Totonac branch of the family that has any substantial amount of documentation yet, other than Upper Necaxa, is Apapantilla.[10] Not unsurprisingly, relative clauses in this language closely resemble those in Upper Necaxa, although the Apapantilla relativizers are a̰ntiː and a̰ntuː. The examples in (56) and (57) show externally-headed relative clauses with animate heads introduced by a̰ntiː:
The example in (56) is subject-centred, while that in (57) is object-centred. (58) shows a subject-centred headless relative clause introduced by a̰ntiː, while (59) shows an object-centred headless relative clause with an inanimate referent, introduced by a̰ntuː:
The relativizers here appear to be composed, at least etymologically, of tiː/tuː and what was historically a deictic element, *a̰n (cf. Upper Necaxa a̰n ‘medial non-demonstrative determiner’). Fronting of an argument within the relative clause is mentioned as a possibility in Reid et al. (1968), where the following example is given:
The authors note that this type of fronting within a dependent clause is possible when the fronted element is “emphasized” or the fronted element contains a quantifier, as in (60) above. This seems in line with the Upper Necaxa data, where fronting in the relative clause correlates with focalization. While nothing is said explicitly in my sources about accessibility to relativization, examples culled from the lexical database compiled by missionaries from the Summer Institute of Linguistics (Reid et al., n.d.) include locative-centred relative clauses introduced by a̰nɬaː ‘where’ (composed of *a̰n and ɬaː ‘where?’):
There are also attestations of possessor-centred relative clauses:
So it would seem that Apapantilla resembles Upper Necaxa in covering most of the Accessibility Hierarchy. Cerro Xinolatépetl TotonacThe remainder of the Totonac languages fall into the South Central–Lowland division, which is comprised by two large branches encompassing an undetermined number of variants, Lowland and Sierra, and two individual languages, Cerro Xinolatépetl and Filomeno Mata, which appear to be peripheral to either of these branches. The most divergent of the two, Cerro Xinolatépetl, is virtually undescribed and what information we have about relatives in this language comes from the text in Andersen (2012), which contains two examples of relative clauses, one with an animate external head (63), and the other a headless relative with an inanimate referent (64):
These examples are, respectively, subject- (64) and object-centred (63), and make use of tiːn and tuːn relativizers. Filomeno Mata TotonacFilomeno Mata phonology and morphology are described in McFarland (2009), but this work does not address relativization; however, several examples of relative clauses do appear in the text in McFarland (2012). The example in (65) is an externally-headed object-centred relative introduced by the non-human relativizer, tuu= (analyzed by McFarland as a clitic):
In (66) we see a headless subject-centred relative clause with an animate referent:
(67) shows a headless object-centred relative clause with an inanimate referent:
There are also some examples of locative-centred relative clauses such as that in (68):
The relative clause here is introduced by the locative relativizer ɬaa=, likely cognate with the relativizing element xa̰ː used in the locative-centred relative clauses in Upper Necaxa in (36) and (37)above. In total, there are 31 instances of relative clauses in the text in McFarland (2012) and while this is a very small sample on which to make generalizations about constituent order, in all but one of the examples the relativizing clitic attaches to a verbal or non-verbal predicate, and in one case (p. 274, line 41) it attaches to an adverbial element glossed as ‘now’ preceding the verb, suggesting that there is at least a strong preference for relatives clauses to be predicate-initial. Lowland TotonacFor the Lowland group, we have only information from Cerro del Carbón (a.k.a. Papantla Totonac), once again gleaned from texts (Levy 2012). In this language, we appear to find a structural distinction between externally-headed and headless relative clauses. An externally-headed object-centred relative clause is illustrated in (69):
The head of the relative clause here is sáqat ‘tall grass’, an inanimate noun; in (70) we see a subject-centred relative clause with a plural animate head:
In both examples, the relative clause is introduced by níːma, which varies neither with the animacy nor the number of the head of the relative construction. The texts contain 14 examples of externally-headed relative clauses, all of which are subject- or object-centred, and in all of which the relativizer immediately precedes the verb. Headless relatives, on the other hand, use tuː and tiː, as in (71) and (72):[12]
Both the examples here are object-centred, but a search through the 19 examples in the texts reveals that there are subject-centred clauses as well as clauses centred on what Levy (2002) analyses as secondary objects. All but one of the examples, shown in (73), has the embedded verb in absolute clause-initial position:
In (73), the verb kiɬwama ‘say something’ is preceded by a demonstrative amáː ‘that’; however, it isn’t entirely clear what the role of the demonstrative is in this sentence. One possibility is that it expresses the object (what the rabbit is saying), in which case this is an example of an internally-headed relative clause. Another possibility is that tuː amáː functions as a unit, forming a demonstrative relativizer. This is an interesting example and structures like these clearly merit further investigation. Sierra TotonacFor the Sierra group, there are a few examples of relative clauses from Olintla Totonac found in the text in Tino (2012). On the whole, these resemble the Northern Totonac pattern found in Upper Necaxa and Apapantilla, in which both externally-headed and headless relative clauses are introduced by a relativizer that varies according to animacy. Externally-headed constructions are illustrated in (74) and (75):[13]
These examples depart slightly from patterns we’ve seen previously in that the head noun is separated from the relativizer, tu or ti, by elements glossed as “particles”; however, there are other examples in the text where the head noun is immediately adjacent to the relativizer (e.g., p. 309, line 42; p. 310, line 63). Note that we have both subject- (75) and object-centred (74) relative clauses here in these examples. Headless relatives introduced by tu and ti are also attested:
Both of these examples are object-centred. Example (77) is of note in that it shows the embedded verb preceded by an adverbial, tlaːn ‘well’, which is the source of the ‘could’ in the translation. This indicates that Olintla, like Huehuetla Tepehua and Upper Necaxa, preserves the pre-verbal positioning of adverbial elements inside relative clauses. The nearby language of Huehuetla Totonac is described in Troiani (2004), which does not address relativization directly but provides a few examples in texts. Only headless relative clauses are attested at all for this language, and these make use of the tu and ti relativizers. Examples (78) illustrates a headless relative with an inanimate referent:[14]
Of note in example (78) is the combination of the relativizer with the suffix (most likely a clitic in morphosyntactic terms) -ku ‘still’ (Fr. ‘encore’). The corresponding element in Upper Necaxa, =kus, is not attested in combination with the relativizers, though it combines with a wide range of other elements. While most of the examples in these texts show the relativizer combining with -ku, examples like (79) show that this is not obligatory:
The relative clause here has an inanimate referent, the knowledge of the Actor in the matrix clause. The example in (80) illustrates a headless relative clause with an animate referent:
The second clause at the end of this example, kawása tɬawakaníːt, is glossed in the original as a relative clause ‘the boy that has been made’ but is set off from the rest of the utterance by a prosodic boundary (“//”) which I’ve represented in the transcription line as a comma; however, there is no relativizer in the Totonac and the indefinite actors in the first clause (the unspecified group that is coming to see the boy) and the final clause (the unspecified actor that made the boy) are not the same, whereas identity of unspecified actors would be expected within the confines of a single sentence. The possibility remains that this is indeed a paratactic relative construction of the type seen in Misantla in (52), although to date no further evidence that this structure might exist in Huehuetla Totonac, or any other Sierra or Central Totonac language, has been found. Turning to the Totonac spoken in Ozelonacaxtla, we find half a dozen examples of relative clauses in the text in Román Lobato (2012). These show the familiar Sierra pattern of externally-headed and headless relatives introduced by relativizers that distinguish the animacy of the head of the relative clause, although the relativizers have a slightly different form. The inanimate relativizer (81) is given as tuku (cf. the Huehuetla form in (78) above), while the animate relativizer is titʃi (82):
There is currently not enough data to determine if these are actually unanalyzed combinations of a relativizer and some other element, as we saw in Huehuetla Totonac, or if these are in fact fixed forms derived diachronically from such sources. The examples in (81) and (82) are object- and subject-centred, respectively. It also appears from an example given in Román Lobato (2008) that possessor-centred relatives are possible:
This is an interesting example because the target of relativization is the possessor of the complement of the copular verb rather than of its argument. Of particular note in the Ozelonacaxtla data is the following example, in which it appears that one of the arguments of the embedded clause intervenes between the verb and the relativizer:
In (84) we see that the noun wa̰ʝa̰ ‘hawk’, the object of the verb tla̰wa ‘X makes Y into Z’,[15] immediately follows the relativizer titʃi, separating it from the verb. This would appear to be an example of argument fronting inside the relative clause, although another possibility is that wa̰ʝa̰ ‘hawk’ here is not an ordinary object but some kind of predicate complement occupying the pre-verbal slot normally taken by secondary predicates. This will have to remain an open question, pending further investigation. Relative clauses in Zapotitlán Totonac (sometimes referred to in the literature as “Sierra” or “Highland Totonac”) are described in an article by E. Aschmann (1984), although the scope of that paper and the range of constructions discussed under the heading of “relative clause” is somewhat broader than ours is here. Drawing on the descriptions in this paper and on unanalyzed examples contained in the lexical database compiled by H. Aschmann (n.d.a), it can be seen that Zapotitlán strongly resembles the other Sierra languages in most respects. Externally-headed relatives are introduced by tiː and tuː relativizers that distinguish animacy, and may be subject- (85) or object-centred (86):
Of particular note in these constructions is the presence of the element wa̰ː intervening between the head noun and the relativizer. This is an extremely frequent feature of relative constructions in Zapotitlán, so much so that H. Aschmann (n.d.a) analyzes the relativizers as wa̰ːntiː and wa̰ːntuː, respectively; however, there are examples (see (92) below) where the relativizer appears without wa̰ː, and in E. Aschmann’s (1984) article wa̰ː is treated as a separate focus particle.[16] Comparison with probable cognates in other Totonacan languages shows that this element is likely derived from a demonstrative element of some kind. E. Aschmann’s article also provides examples of locative-centred (87) and possessor-centred | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||