1 48. Person Marking on Adpositions
3 Dik Bakker
5 1. Defining the values
7 This map gives a survey of the distribution of languages with
8 person markers on their adpositions. An example of a person-
9 marked preposition is found in (1) from Maybrat (West Papuan;
10 Papua, Indonesia).
12 (1) Maybrat (Dol 1999: 88)
13 T-ai m-kah ara.
14 1SG-hit 3SG.N-with stick
15 ‘I hit with a stick.’
17 The 379 languages in the map have been characterized in terms
18 of the four values in the table below.
@ 1. No adpositions 63
@ 2. Adpositions without person 209
@ 3. Person marking for pronouns only 83
@ 4. Person marking for pronouns and 23
21 In order to determine whether a language has person
22 marking on its adpositions, we first need to establish whether it
23 possesses the category adposition at all. In general, the major
24 function of an adposition is to relate its object, i.e. the noun
25 phrase with which it forms a constituent, to another nominal or
26 a verbal constituent on the basis of a more or less specific
27 semantic relationship, such as location, time, property,
28 instrument or possession. Languages may use several strategies
29 to fulfill this function, either lexical or morphological. In
30 example (2) below, from Barbareño Chumash (California), we
31 have a verbal construction fulfilling such a function. What is
32 expressed by an adverb and an adposition in the English
33 translation, across from, is rendered by a fully inflected verb in
34 this language.
36 (2) Barbareño Chumash (Wash 2001: 75)
37 Kʰ-ili-ʔete .su s hi lwí.sa hik -ken
38 1-HAB-be.across.from DEP Luisa DEP-1.sit
39 hi-h - a e.sa.
41 ‘I used t sit acr ss fr Luisa at the tab e.’
43 Example (3), from Tauya (Madang; Papua New Guinea), shows a
44 nominal strategy to express the locality relationship rendered by
45 the preposition beside in the English version.
47 (3) Tauya (MacDonald 1990: 283)
48 ya nai-sa
49 1SG rib-LOC
50 ‘beside e’
52 Although diachronically forms such as the verb ʰi iʔete .su s
53 in (2) and the noun naisa in (3) may eventually give rise to
54 adpositions, they will not be considered as such here. In general
55 a specific element in a language will be assumed to be an
56 adposition only if it is morphologically independent and displays
57 morphosyntactic behaviour distinct from more clearcut verbal,
58 nominal or adverbial elements in that language.
59 In addition to languages such as Chumash and Tauya
60 which encode adpositional meanings via verbs or nouns, there
61 are languages in which the relevant relations are coded by
62 morphological means, typically by a case suffix. The example
63 from Arabana (Pama-Nyungan; South Australia) in (4) illustrates
64 this strategy. The ab ative case is used t express ‘away fr ’.
66 (4) Arabana (Hercus 1994: 71)
67 Maka-ru kilta-rnda.
68 Fire-ABL pull-PRES
69 ‘He pu s it ut f the fire.’
71 Another type of morphologically dependent expression of
72 nominal relations is via adverbial nominal affixes, as in example
73 (5) from Jaqaru (Aymaran; Peru).
75 (5) Jaqaru (Hardman 2000: 21)
78 ‘fr within the h use’
80 Arguably a borderline case between morphological and syntactic
81 expression are clitics which attach to the noun phrase rather
82 than to nouns, as exemplified in (6) from Ngankikurungkurr
83 (Daly; Northern Territory, Australia).
85 (6) Ngankikurungkurr (Hoddinott and Kofod 1988: 72)
86 Kalla ngayi yedi tye jeningkisyi yaga=nide.
87 Mother 1SG went PST canoe DEM=LOC
88 ‘My ther ca e in that can e.’
90 In many instances, case affixes as in (4) and clitics as in (6) can
91 be analyzed diachronically as the result of the affixation or
92 cliticization of adpositions. However, taking a rather
93 conservative, strictly syntactic position on this matter, I will not
94 consider such bound forms as true adpositions. A further
95 argument for not doing so may be that since such forms are
96 (syntactic) dependents rather than heads, they are not potential
97 targets for person marking but, to the contrary, may attach
98 themselves to pronominal forms. As a result of this, all
99 languages which resort exclusively to strategies as exemplified
100 under (2)-(6) are coded as having no adpositions (value 1) on
101 the map. They are by definition irrelevant for the type of person
102 marking under consideration here.
103 Having established the criteria for what constitutes an
104 adposition, we must now determine what qualifies as person
105 marking on adpositions. In order for a potential marker to be
106 considered a person marker it should fulfill two requirements.
107 First, there should be enough differentiation between the
108 relevant forms that a distinction exists either between all three
109 persons or between any combination of first, second or third
110 person. One of the forms may be zero. And secondly, the forms
111 should be affixes rather than clitics.
112 Languages with only bare adpositions, and those for
113 which markers on their adpositions do not meet the criteria
114 mentioned above irrespective of the nature of the adpositional
115 object, are assigned value 2 (adpositions without person
116 marking). English is a case in point, and so is Polish, as shown
117 in example (7a-b).
119 (7) Polish (Anna Siewierska p.c.)
120 a. Idę do Kasi.
121 go.FUT.1SG to Kasia.GEN
122 ‘I g t asia.’
123 b. Idę do niej.
124 go.FUT.1SG to 3SG.F.GEN
125 ‘I g t her.’
127 When languages do show person marking on adpositions
128 for pronominal objects but not for nominal objects, they are
129 assigned value 3 (person marking for pronouns only). Paamese
130 (Oceanic; Vanuatu) has this kind of adpositional marking, as
131 shown in (8a-b) below.
133 (8) Paamese (Crowley 1982: 182)
134 a. Mail Ham sān ēta min-nau
135 Mail Ham 3SG.send letter to-1SG
136 ranaut Vī a.
137 from.place Vila
138 ‘Mai Ha sent e a etter fr Vi a.’
139 b. Kai se ūs min tāta ven mane onak.
140 3SG 3SG.speak to father about money POSS.1SG
141 ‘He sp ke t father ab ut y ney.’
143 Typically, in these languages the adposition has no separate
144 independent pronominal object apart from the person marker
145 itself, although there may be one under certain conditions, such
146 as contrastive stress. Hence, more often than not, the person
147 marker on the adposition is the only element representing the
148 pronominal object. In this group are also included languages for
149 which only a subset of the adpositions have person marking. In
150 Burushaski (isolate; Pakistan or India) the postposition pači
151 ‘with’ in (9a) ccurs with a prefixed pers n arker; the
152 postposition gΛnε ‘f r’ in (9b) d es n t.
154 (9) Burushaski (Lorimer 1935: 96, 97)
155 a. Ja aˑ-pʌči huruˑt.
156 I 1SG-with stay
157 ‘Stay with e.’
158 b. Xʊdaˑ-ɛ gʌnɛ.
159 God-GEN for
160 ‘F r G d’s sake!’
162 The occurrence of person marking may also be subject to other
163 constraints. For instance, in Kiribati (Oceanic) there is person
164 marking only when the referent is animate, as shown in (10a-b).
166 (10) Kiribati (Groves et al. 1985: 65)
167 a. nako-ia mooa
168 to-3PL chickens
169 ‘t the chickens’
170 b. nako taian nii
171 to ART coconut.trees
172 ‘t the c c nut trees’
174 However, borderline cases such as Spanish, which has bound
175 pronominal forms for just one preposition, con 'with', and only
176 for the first and second person singular forms, were coded as
177 showing value 2.
178 Finally, there are languages which have adpositional marking
179 with both pronominal and nominal objects. Abkhaz (Northwest
180 Caucasian; Georgia) is a case in point, as shown in (11).
182 (11) Abkhaz (Hewitt 1979: 103)
183 a. a-j yas a-q’n
184 DEF-river 3SG-at
185 ‘at the river’
186 b. sar s-q’ nt˚’
187 I 1SG-from
188 ‘fr e’
190 Such languages are assigned value 4. Like value 3, this value
191 includes languages which do not normally use a full pronoun in
192 constructions such as (11b).
193 In my sample there were no languages of the other
194 logically possible type, which would have person marking on
195 adpositions for nominal but not for pronominal objects.
197 2. Geographical distribution
199 The phenomenon of person marking on adpositions is not
200 distributed evenly over the world. First, the languages of North
201 America and Australia lack adpositions much more often than
202 the other areas. Of the North American languages in the sample,
203 48% are adpositionless; for Australia the percentage is as high
204 as 73%. The overall figure for adpositionlessness is around 17%
205 in the current sample.
206 If we look only at languages with adpositions, then the
207 following may be observed. Person marking on adpositions is
208 non-existent in Australia, and rare in Southeast Asia (3 out of
209 30 languages, all belonging to the Austronesian family). For
210 Southeast Asia this does not come as a surprise, given the
211 general lack of person marking in this area. In Australia, on the
212 other hand, person agreement on the verb is found in the
213 majority of the languages, which may be an indication that the
214 two types of person marking are not necessarily closely related
215 to each other. In Mesoamerica and the Pacific, person marking
216 on adpositions is clearly abundant. These are the only areas
217 where a majority of the languages have this kind of marking:
218 over 60%.
219 Other striking figures are the relatively frequent
220 occurrence of languages of type 4 (person marking for
221 pronouns and nouns) in Mesoamerica. It is above all the Mayan
222 and Uto-Aztecan languages which have this property. In Africa,
223 although this area has a relatively high amount of type 3
224 languages, none of the languages show type 4. Eurasia, New
225 Guinea and South America are all close to the overall
226 distribution. Of the 20 Indo-European languages in the sample,
227 only the two Celtic languages Irish and Welsh and the two
228 Iranian languages Persian and Kurdish have any marking at all;
229 they all show type 3.
231 3. Theoretical issues
233 Due to the fact that linguists differ in the criteria that they use in
234 determining the existence of adpositions in a language, there
235 may be considerable discrepancies in the percentage of
236 languages which are seen to lack adpositions altogether. My
237 figure of 17% is close to that of Hawkins (1983), and also
238 Tsunoda et al. (1995). By contrast, Matthew Dryer (chapter 85 of
239 this atlas) is more liberal in his interpretation of what
240 constitutes an adposition and treats in some cases as
241 adpositions what I have considered to be case clitics.
242 A second point that deserves a word of comment is that it
243 is not always possible on the basis of the information presented
244 in grammars to decide what the precise grammatical status of a
245 specific person marker is, i.e. whether it is indeed an affix
246 attached to the adposition, or a clitic, or even a more or less
247 independent pronominal form. Apart from morphosyntactic
248 information, the phonological form of the person marker may
249 be an indication, as well as the amount of (dis)similarity to other
250 person markers, notably independent and possessive pronouns.
251 A final issue which needs to be briefly mentioned is the
252 diachronic status of the markers. They may be relatively recent,
253 and introduced after adpositions were formed in the language.
254 In these cases they are often phonologically related to the actual
255 personal pronouns of the language, typically in their object or
256 oblique form. This is the case in about 12% of the languages
257 that display some form of person marking on their adpositions.
258 Or the markers may be relatively old, and already present on the
259 verbal or nominal precursor of the adposition. If there is
260 similarity to any other person marker in the language at all,
261 similarity to subject or object agreement markers on the verb
262 may (but need not) be an indication of the verbal origin of the
263 adposition. This seems to be the case in about 13% of the
264 relevant languages. Similarity to possessive markers, which was
265 attested in about 24% of the relevant languages, may point to a
266 nominal origin. In around 41% of the cases, person markers on
267 adpositions bear a similarity to both verb agreement and
268 possessive markers in the language concerned. In the remaining
269 11% no clear similarities could be detected.