The Emergence of Axial Parts Peter Svenonius CASTL, University of Troms0 Abstract Many languages have specialized locative words or morphemes translating roughly into words like 'front,' 'back,' 'top,' 'bottom,' 'side,' and so on. Often, these words are used instead of more specialized adpositions to express spatial meanings corresponding to 'behind,' 'above,' and so on. I argue, on the basis of a cross-linguistic survey of such expressions, that in many cases they motivate a syntactic category which is distinct from both N and P, which I call AxPart for 'Axial Part'; I show how the category relates to the words which instantiate it, and how the meaning of the construction is derived from the combination of P[lace] elements, AxParts, and the lexical material which expresses them. 1. Introduction Many adpositions include parts which are historically nominal, for example English beside, from Old English be sidan, 'by side.' The historical source in such cases is a relational noun ('side' in this case) referring to a concrete part of an object, used with a genitive dependent expressing the whole, in something like 'by the side of the house.' The locative sense would then have been entirely from the locative preposition; however, over time, the relational noun has come to be reanalyzed as a locative expression, referring not to a part of the object, but to a space defined with reference to that part (I return to etymological issues in §7). In the case of the English expression in front of, the decomposition seems transparent as in is a preposition and the word front still has a use as a relational noun. Consider the pair in (1). (1) a. There was a kangaroo in the front of the car. b. There was a kangaroo on the front of the car. The front part of a car is conventionally understood either as the part which enters an area first when the car is moving in its usual direction (typically the bumper, grill, front wheels, engine, and hood or bonnet) or else the * Thanks to the participants in the Moving Right Along seminar held in Fall 2005 and Spring 2006, whence these ideas emerged, and especially to Minjeong Son, Marina Pantcheva, and Gillian Ramchand for ongoing collaboration throughout that period. Thanks also to audiences in Venice (Cartography conference), Amsterdam (LOT Winter School), Utrecht (P conference), Troms0 (NORMS workshop), Montreal (McGill colloquium), and Paris (EALing at ENS) in 2005-6. © 2006 Peter Svenonius. Nordlyd: Troms0 Working Papers in Linguistics, 33.1, special issue on Adpositions, ed. Peter Svenonius and Marina Pantcheva, pp. 49-77. CASTL, Troms0. http://www.ub.uit.no/munin/nordlyd/ The Emergence of Axial Parts compartment including the driver's seat and any seat beside it. Example (la) is most naturally understood as stating that a kangaroo is in one of the two front seats of a typical car, but it could also refer to a kangaroo being in a cargo space under the hood or bonnet in a rear-engined car, for example. In both situations, in would be appropriate because the kangaroo would be contained within an enclosed space. In (lb), the kangaroo would be in contact with the top surface of the front part of the car, for example on its hood or bonnet. Now consider (2), which differs from (1) only in lacking the determiner the in front of front. (2) a. There was a kangaroo in front of the car. b. *There was a kangaroo on front of the car. Example (2a) refers to something quite different from (la), namely a kangaroo located in a space projected forward from the car. Example (2b) is ungrammatical. In this paper I argue that the word front in (2a) is not a noun, but lexicalizes a functional projection which I call AxPart, based on Jackendoff's (1996) and Marr's (1982) discussion of axial structures in spatial cognition.1 I argue that AxPart is a category like aspect or modality, realized in many languages. I will henceforth refer to the 'part' sense of front as its N use, and to the the spatial sense as its AxPart use. Another test which distinguishes the two senses of front is the plural; the N use of front can be pluralized, but not the AxPart use. Here the definite article is retained to indicate which example is the N and which is the AxPart. (3) a. There were kangaroos in the fronts of the cars, b. * There were kangaroos in fronts of the cars. Another distinction between the N use and the AxPart use of front is adjectival modification, which is only acceptable with N. (4) a. There was a kangaroo in the smashed-up front of the car. b. *There was a kangaroo in smashed-up front of the car. Measure phrases are acceptable with many locative expressions, including the expression in front of which refers to a vaguely bounded space. Measure phrases are not acceptable with usual uses of in, however, and it can be seen that when front is an N, then in is a non-measurable P (more specifically, it is category Place). 1 Jackendoff (1996:14): "The "axial parts" of an object—its top, bottom, front, back, sides, and ends—behave grammatically like parts of the object, but, unlike standard parts such as a handle or a leg, they have no distinctive shape. Rather, they are regions of the object (or its boundary) determined by their relation to the object's axes. The up-down axis determines top and bottom, the front-back axis determines front and back, and a complex set of criteria distinguishing horizontal axes determines sides and ends." 50 Peter Svenonius (5) a. *There was a kangaroo sixty feet in the front of the car. b. There was a kangaroo sixty feet in front of the car. Another difference between the AxPart use of front and the N use is that only the N use admits replacement of its projections by pro-forms; thus, it can refer to the front of the car in (6a) but cannot refer to front of the car in (6b); nor can one refer to front of the car in (6c). (6) a. The kangaroo was in [the front of the car];, but the koala wasn't in it;. b. The kangaroo was in [front of the car];, but the koala wasn't in it*;. c. The kangaroo was in [front of a car];, and the koala was in one*; too. Finally, DP can move away from a preposition, but AxPart cannot. (7) a. It was the front of the car that the kangaroo was in. b. *It was front of the car that the kangaroo was in. Though the exact diagnostics differ from language to language, the behavior of English front is fairly typical of AxPart elements. Schematically, a relational noun like the N use of front might appear in a structure like that in (8); the N appears with a genitive complement, labeled K[ase] here. The Place head in takes any DP complement, for example in in the car, the Place head in combines directly with the DP the car (or there is a null AxPart, and/or a null K). (8) Place the N front K of^^DP the car As depicted in (9), AxPart might similarly appear as a complement to N, and with a genitive complement, but lacking the functional structure associated with N, for example D. 51 The Emergence of Axial Parts (9) Place in AxPart front K of DP the car I suggest below specific semantic contributions for the various categories, extending the model outlined in Svenonius (2004b). Briefly, genitive K acts as a kind of type-shifter, lifting DPs to predicates over some projection of N;2 in languages in which AxPart combines with genitive K, AxPart must then either be of the same semantic type as N, or else be a predicate over elements of type K. The semantic function of AxPart is to identify a region (a set of points in space, cf. Nam 1995, Kracht 2002) based on the Ground element (the complement DP; see Svenonius 2004a for discussion of the Ground interpretation of P complements). Typical AxParts refer to the front, back, top, bottom, sides, and middle of an object, though other regions can be defined as well. Regions are unstructured, like sets of points, and therefore cannot be measured. Examples like (5b) show that PPs containing AxParts can be measured, however. I assume that this extra structure is contributed by elements of the category Place. The semantic contribution of Place is to specify how space is projected from a region; I will assume a modelling of space in terms of vectors along the lines proposed by Zwarts (1997), Zwarts and Winter (2000). Vectors are one-dimensional objects with direction and length which define points in a space when they are drawn from a region. Measures and directions can be defined over sets of vectors, so that expressions like (5b) and diagonally over the doorway can be translated. The exact conditions on measure and directional phrases are complex (see Zwarts 1997, Zwarts and Winter 2000, Winter 2001); I will assume that even non-measurable locative expressions include a component Place. There are minimal pairs like in back of and behind, which mean nearly the same thing. I assume that in back of involves an AxPart back, and the Place head (lexicalized by in) projects vectors away from the Ground from that axial part ('away from Ground' being a common direction for vectors in P semantics). With behind, on the other hand, I will assume that there is a default mapping from Ground to a region (what Wunderlich 1991 calls 2 The typical relation which results from genitive predication is a vague one, suitable not only for part-whole relations but also various other associations. See Asbury (2006) for further discussion of the categorial status of genitive and its close relation to the D system. 52 Peter Svenonius its eigenspace), and behind is a Place head specifying that vectors point backward. Some of the subtle differences between the two expressions are hoped to follow from these two different ways of identifying essentially the same space, but the details are still murky at this point. Important members of the category Place are on and in; on specifies that there is contact, which implies that the vectors used to define the space have length zero, and in specifies that the space is bounded. The fact that these two combine with AxParts, while other putative Place heads do not, suggests that they may be importantly different. For the purposes of this paper, however, I will treat them as Place heads. 2. Cross-linguistic examples of AxParts In this section I review the AxPart inventories of several languages. In each of the languages discussed, a set of somewhere around a dozen words has a distinctive syntactic and semantic pattern; the words can be identified as referring to regions or directions, but are used to denote spaces, usually in conjunction with some more generalized locative morphemes. For practical reasons, I divide the languages up into prepositional and postpositional, grouping local case systems in the latter group, and then discuss the range of variation observed. 2.1. Prepositions For example, Pantcheva (2006a) examines a series of elements in Persian, which she calls Class 2 prepositions, and concludes that they are AxParts in the sense used here. A partial list appears in (10). (10) Persian Class 2 Prepositions (Pantcheva 2006a) word noun gloss prepositional translation tu inside 'in' ru face 'on' zir lower part 'under' dour cycle 'around' bala upper part 'above' TMEzdik vicinity 'near' daxel interior 'inside' tceh bottom 'on the bottom of posht back 'behind' paiin below 'below' xarej exterior 'outside' birun exterior 'outside' mian center 'between' seer head 'on top of 53 The Emergence of Axial Parts Like English front, some of these elements have nominal senses (especially the first few in the list), but have developed a specialized use in locative expressions; their locative use is illustrated in (11). Note the 'ezafe' linker element or case-marker (Samiian 1994) glossed here as ez (following Pantcheva 2006a, whence the examples). (11) a. Raeft posht-e xane. went back-EZ house 'She/He went behind the house' b. Sib oftad ru(-ye) miz. apple fell face-EZ table 'The apple fell on the table' In the example in (11a), the word posht 'back' is clearly used as an AxPart, as the sentence does not assert that the subject went to a part of the house itself, but to the space behind the house; thus the truth conditions are calculated on the basis of a space projected from the back of the house. In (lib), the ezafe marker is omissible; ru meaning 'on' can (optionally) assign case directly. Similarly, Levinson (1994) discusses a class of nouns in Tzeltal (a Mayan language) which he calls 'relational nouns,' listed in (12) (with possessive prefixes). Levinson carefully distinguishes these from a distinct class of nouns referring to specific topological shapes, which have a slightly different syntactico-semantic behavior (cf. Jackendoff's 1996 comment, quoted in n. 1, that axial parts have no distinctive shape). (12) Tzeltal Relational nouns (Levinson 1994:802,807) Word noun gloss s-ba 'top surface' y-ajk'ol 'uphill region' y-alan 'downhill region' y-anil 'underneath' y-util 'inside' y-ejtal 'bottom surface' s-tojol 'straight ahead' y-olil 'middle' s-tz 'eel 'side, edge' A couple of illustrations of the use of these relational nouns are shown in (13). (13) a. ta s-ba mexa at POSS-top table 'on top of the table' (Tzeltal, Levinson 1994:801) b. Tek'el ta y-ajk'ol karo te winik-e. standing at POSS-uphill car the man-CL 'The man is standing uphill of the car' (Tzeltal, Brown 1994:751) 54 Peter Svenonius In the first example, the locative expression ta sba mexa 'on top of the table' could involve a simple part-noun 'top,' since the location is one of contact with an actual part of the table. In the second example however, the locative expression ta yajk'ol karo 'uphill of the car' does not locate the man directly at a part of the car, but at a region projected from a part of the car. Since the free P element ta 'at, on' is the same in both cases, the difference must come from the second element, (y-)ajk'ol 'uphill region.' Thus, this element clearly has the semantics of an AxPart: a space is projected from the Ground element (the car in this case). Although the semantics of the first example, with (s-)ba 'top' is more equivocal, I conclude on the basis of general syntactic similarity (discussed in detail in the references cited) that all of the relational nouns in (12) are in fact AxParts. Muriungi (2006) has also shown that Kiitharaka, a Bantu language spoken in Kenya, has a class of AxParts, which he calls Class B adpositions. Most of them have nominal class marker prefixes, as shown in (14). (14) Kiitharaka Class B adpositions (Muriungi 2006) Word nominal gloss prepositional gloss ru-ungu 'space below' 'under' ru-teere 'side' 'beside' ga-tigati 'center' 'between' ka-thengengani 'edge' 'on the edge' í-gůru 'top' 'above' mbere 'front' 'in front' nyuma 'behind' 'behind' nja 'outside' 'outside' ndeni 'inside' 'inside' nkoma 'bottom' 'at the bottom' karibů 'vicinity' 'near, next to' Some illustrations of typical uses of these AxParts are given below (also from Muriungi 2006). Nouns typically have class markers affixed, though in class 9 the segmentation is not clear (the prefix might be a nasal feature); the verb bears a prefix agreeing with the subject in noun class, and the associative marker a 'as' agrees with the AxPart which introduces it. (15) a. Gi-kombe ki-ri ga-tigati k-a metha. 7-cup SM7-be 12-center 12-AS 9.table 'The cup is at the center of the table' b. Gi-kombe ki-ri ndene j-a i-kumbi. 7-cup SM7-be 9.inside 9-AS 5-granary 'The cup is inside the granary' As with previous examples, the case marking on the Ground DP is essentially genitive. The phrase projected by the AxPart can be the complement 55 The Emergence of Axial Parts to a Path P like 'to' or 'from.' The AxPart has nominal characteristics, such as occurring with a noun class marker, but when it has its projective, spatial meaning, it cannot occur with the full range of nominal modifiers, such as quantifiers and adjectives. 2.2. Postpositions and local case suffixes In each of the examples of AxParts presented above, the AxPart preceded its DP Ground complement. There are (unsurprisingly) also many languages in which AxParts may follow the Ground, in a postpositional pattern. I review some such cases here, along with local case systems.3 Takamine (2006) discusses Japanese AxParts in a contribution to this volume. Here I present the closely related system of Korean.4 Korean has a set of what I analyze as AxParts, as illustrated in the table in (16). (16) Korean Place noun Word nominal gloss prepositional gloss alay 'lower part' 'under' müh 'bottom' 'under' wi 'upper part' 'on' aph 'front' 'in front of an 'interior' 'in' pakk 'exterior' 'out' kawuntey 'center' 'between' yep 'side' 'beside' twi 'back/rear' 'behind' kunche 'vicinity' 'near' cwuwi 'vicinity' 'around' These elements usually appear with a locative case suffix, for example the generalized locative -ey as in (17a) or the directional -ulo as in (17b). (17) a. Inho-ka cip pakk-ey se-e iss-ta. Inho-NOM house exterior-LOC stand-E be-DC 'Inho is standing outside the house' b. Inho-ka kakey ap-ulo ttwi-e ka-ss-ta. Inho-NOM store front-BlB. run-E go-PAST-DC 'Inho ran to the front of the store' Although Korean has overt case markers (the genitive suffix is -uy), Korean 3In the literature on case, 'local case' lias become the standard term for cases which distinguish locational meanings, as opposed to 'locative,' which is usually the label for a single vaguely locational case. See Blake (2001) for an excellent overview. 4I am here especially indebted to Minjeong Son, who presented much of this material in the Moving Right Along seminar in Troms0 in 2006, as documented on the website at www.hum.uit.no/mra/, see especially the February 28th summary. 56 Peter Svenonius is unlike Japanese in not usually expressing genitive case on the Ground element with an AxPart; the Ground DPs are normally bare, as in (17). Elaborate local case systems can usually be decomposed into two or more parts, with an inner part expressing basic Place distinctions (such as 'on' and 'in') and an outer part expressing Path distinctions (goal or 'to,' source or 'from,' and basic location, here glossed as 'loc'); see Hegediis (2006) for discussion of the Hungarian system. The related Finnish system can be decomposed as follows, where -C is an underspecified consonant (assimilating to the preceding consonant), -A an underspecified low vowel (participating in vowel harmony): (18) Finnish (Karlsson 1977) Place: in on Path -S -I loc -CA inessive adessive from -tA elative ablative to -Ce illative allative Thus, for example, the inessive of talo 'house' is talo-ssa meaning 'in a house' (or 'in the house') and the elative is talo-sta meaning 'from in a house' (but the illative is not morphologically decomposable). It is unclear to what extent these suffixes correspond directly to semantically in-terpretable heads, for example the same suffixes appear on nominal dependents in case agreement (see Svenonius to appear for some discussion). But it is clear that functionally, local cases are the equivalent of adpositions. It can be assumed, therefore, that a locally case-inflected noun like talossa 'in a house' is a PP, more precisely at least a PlaceP and possibly a PathP (see Svenonius 2004b on the issue of whether locative PPs project higher than Place; see Fong 1997, Kracht 2002 on differences between the Finnish outer element and English Path heads to and from). When a language with a local case system develops AxParts from nouns, the AxParts will at least initially appear with the local cases. However, since there are semantic restrictions on the kind of Ground that can combine with Place heads like 'in' and 'on,' a given AxPart may not combine with the full range of cases. For example, Finnish ylapuole 'upper side' combines with the 'on' series (adessive, ablative, and allative) while viere 'side' combines with the 'in' series (inessive, elative, illative).5 This is illustrated in (19) and (20) (as noted above, the illative is not decomposable, and is here simply glossed ill). (19) a. poyda-n yla-puole-l-la table-GEN upper-side-ON-LOC '(at) above the table' 5 Thanks to Marina Pantcheva for pointing this out to me. Thanks also to 0ystein Alexander Vangsnes and Eugenia Romanova for discussion. 57 The Emergence of Axial Parts b. poyda-n yla-puole-l-ta table-GEN upper-side-ON-FKOM 'from above the table' c. poyda-n yla-puole-l-le table-GEN upper-side-ON-TO '(to) above the table' (20) a. poyda-n viere-s-sa table-GEN sirfe-iN-loc '(at) beside the table' b. poyda-n viere-s-ta table-GEN sirfe-iN-from 'from beside the table' c. poyda-n viere-en table-GEN side-iLL '(to) beside the table' A similar example can be seen in Kham, a Tibeto-Burman language spoken in Nepal. Watters (2002) identifies a set of what he calls relator nouns, each of which combines with a specific local case suffix. The relator nouns are given with their case suffixes in (21), from Watters (2002:137). (21) Kham Relator nouns Kham literal gloss translation ji:-h inside-in 'inside of du:h-h beneath-in 'underneath' khar-ld center-in 'between' bd:-h bottom-in 'at the bottom of ijah-kd front-at 'in front of chv.-ks behind-at 'behind' lap-ka side-at 'beside' khi-ka foot-at 'at the foot of chyo:-ga edge-around 'at the edge of leo-ijd whereabouts-around 'at the place of tar-to top-on 'on top of sora:-td line-on 'above, in line with' I suggest that the relator nouns, like the Tzeltal relational nouns and so on, are AxParts. As with most of the other languages examined so far, the Ground element is marked in the same way as a possessor: the first person possessor prefix in (22a) is identical to the first person marker in the AxPart construction in (22b). 58 Peter Svenonius (22) a. rja-zihm lPOSSR-Ziowse 'my house' b. rja-rjah-ka lPOSSR-front-at 'in front of me' (Kham, Watters 2002) If each AxPart combines with a limited series of local cases, it is easy to see how the combination of AxPart plus case might become lexicalized and opaque to analysis. This has arguably occurred in Northern Sami, as discussed in Svenonius (2004a). 3. AxPart in the functional sequence Suppose, then, that something about the way that humans construct linguistic representations of space leads to the categorial hierarchy Path-Place- AxPart-K-DP, where DP is the Ground, and there may be other categories as well (cf. Svenonius 2004b on Dir dominating Path, and Svenonius 2003 on a Figure-introducing projection p, probably located above Place but below Path). In English, to is a lexical realization of the category Path, as schematized in (23). (23) Path to Place in AxPart front K of^^DP the car More specifically, we might assume that the categories Path, Place, AxPart, and so on can have different values, and that the words to, in, front and so on express some of these values. This could be schematized as follows, where syntactic categories (Path, Place, etc.) are in normal roman type, semantically interpretable features are written in small caps (e.g. bd for 'bounded'), and elements corresponding to conceptual and phonological information is given in italics. 59 The Emergence of Axial Parts (24) Path Path [goal] I to Place Place [BD] I in AxPart AxPart [facet] I front K K [gen] I of DP the A less cluttered representation simply eliminates the redundant category labels from the heads. (25) Path front gen I of DP the car A marked value like facet would have syntactico-semantic consequences, while the choice of a piece of encyclopedic lexical material such as front or back to fill it might not, the difference between front and back being strictly conceptual (like the difference between camel and reindeer). Suppose, furthermore, that each category has an unmarked interpretation which it can have when not filled. Then the goal PP to the car can be schematized as in (26) (depending on exactly how prepositional case in English is to be analyzed). 60 Peter Svenonius (26) DP the car Under certain conditions, a marked value might be null, for example the Path head might have a goal value without being overt in The bottle will float under the bridge (Carter 1988).6 (27) Path goal Place —BD AxPart I I under K I DP the bridge In general, though, a category which is not overtly lexicalized can only have its default interpretation, for the simple reason that null expressions of marked values are difficult to learn. 4. Similarities between AxPart and N There are many striking similarities between AxParts and Ns, in many languages. It is therefore not surprising that previous researchers have often assumed that AxPart are Ns. Here I review some similarities, moving on in §5 to discuss the differences. For some other similarities, see Takamine (2006), who notes for example that when AxParts are coordinated, they take the same coordinator (to 'and') as nouns do. 4.1. Gender Only nouns are have inherent gender (adjectives, demonstratives, and other elements may agree with a noun in gender, but are not inherently specified 6Here I represent the preposition under as a spell-out of a marked value of Place, with a default interpretation of AxPart; cf. the discussion in §1 of behind. 61 The Emergence of Axial Parts for it). In many languages, some AxParts are consistently associated with gender values. For example, ten of the eleven Kíítharaka AxParts in (14) (from Muriungi 2006) belong to particular noun classes, as signalled by agreement on the associative marker (cf. (15)); Bantu noun classes are essentially like Indo-European genders (Greenberg 1978, Corbett 1991). Similarly, if Roy (2006) is correct that French téte 'head' has an AxPart use, then at least that AxPart appears to have feminine gender, as indicated by the form of the article which appears with it: la téte du lit, 'the head of.the bed' (the form of the article appearing with most AxParts is le, arguably the default). 4.2. Plural Occasionally, AxParts can appear with plural morphology. One such case comes from Persian, as pointed out by Pantcheva (2006a). (28) Shekaer rixt in zir-ha-ye miz. sugar spilled this under-PL-EZ table 'The sugar spilled here all over under the table' Another case is provided by Muriungi (2006) for Kíítharaka. (29) rungu ~ ndungů 11.under 10.under 'under' — 'unders' This simple diagnostic seems to demonstrate that these words are nouns. 4.3. Case AxParts in languages like Finnish regularly inflect with local cases (cf. (19)-(20)). However, if these local cases are analyzed as manifestations of Path and Place heads (as I suggested in §2.2), then they are not indicative of nominal status; that is, Place can quite regularly take an AxPart complement (cf. (26)-(27)). However, sometimes even apparent structural cases are observed on AxParts, for example in Russian where several AxParts alternate between accusative and locative case forms depending on what Path head they combine with; a goal Path head ('to') gives accusative, and source ('from') or pure locative senses give locative case (this was pointed out to me by Eugenia Romanova, who also provided the following examples). (30) a. v-perěd mašiny ~ v-peredi mašiny in-front.ACC car.GEN in-front.LOC car.GEN '(to) in front of the car' — '(at) in front of the car' b. na-verx doma ~ na-verxu doma on-top.ACC house.GEN on-top.LOC house.GEN '(to) on top of the house' — '(at) on top of the house' 62 Peter Svenonius If the prefix in these cases is the Place head, then the locative case suffix is not a manifestation of Place but rather cooccurs with it. 4.4. Articles, Quantifiers, and Demonstratives 4.4.1. Articles In English, the AxPart constructions in front of and on top of lack definite articles, distinguishing them from similar constructions involving nominals (as pointed out in §1). However, there are other elements such as vicinity, left, and east which behave in many ways like AxParts but which tend to appear with the definite article. (31) a. There was a policeman in the vicinity of the house. b. A hooded monk stood to the left of the candelabra. c. The mountains to the east shone purply. In French, too, many AxParts appear with the definite article le, often incorporated into the preposition d (d + le — au), as exemplified in (32) (based on Starke 1993, and parsing autour into au-tour). (32) a. II y a du paprika a l'interieur de la casserole. it there has of.the paprika at the.interior of the pot 'There is paprika in the inside of the pot' b. II tourne toujours au-tour de la maison. he turns always at.the-turn of the house 'He always makes a trip around the house' See Roy (2006) for additional examples and discussion. 4.4.2. Quantifiers There are examples of what I call AxParts appearing with quantifiers, such as the Kiitharaka one here discovered by Muriungi (2006), in which a quantifier onthe 'all' can quantify over the AxPart ndungu 'under.' (33) Maria n-a-ciat-ir-e ndungu ci-onthe. 1.Maria F-SMl-sweep-PERF-FV 10.under 10-all 'Maria swept all the [spaces] under' Since quantifiers normally combine with nouns, this would seem to confirm the nominal status of ndungu. 4.4.3. Demonstratives Demonstratives appear in AxPart constructions in many languages, including Kiitharaka, Persian, Korean and Japanese. For example, in the Persian example in (34) (from Pantcheva 2006a), a demonstrative in 'this' appears with the AxPart tu 'in.' 63 The Emergence of Axial Parts (34) Ketab in tu-ye keshu bud. book this in-EZ drawer was 'The book was here in the drawer' The Korean example in (35), from Son (2006), shows the difference between a demonstrative with an AxPart (in this case mit, 'bottom') and a demonstrative with the Ground DP (in this case oscang, 'chest'). (35) a. Ku sangca-nun oscang ce mit-ey twu-ess-ta. the box-TOP chest that bottom-LOC place-PAST-DC 'I put the box over there under the chest' b. Ku sangca-nun ce oscang mit-ey twu-ess-ta. the box-TOP that chest bottom-LOC place-PAST-BC 'I put the box under that chest' Takamine (2006) provides the following example, showing a demonstrative in Japanese (which regularly appears with genitive case: ko-no ie 'this-gen house') appearing with an AxPart (Takamine notes that there is also an elliptical interpretation of (36), i.e. 'park the car in front of this'). (36) Kuruma-o ko-no mae-ni tomete! car-ACC this-GEN front-LOC park 'Park the car here in front!' Finally, an example from Kiitharaka from Muriungi (2006) shows the same effect: a demonstrative on an AxPart contributes a proximal or distal meaning, rather than picking out a specific discourse referent. For comparison, a noun phrase with a demonstrative and some other modifiers is shown in (37b), showing that demonstratives follow the noun they modify. (37) a. Nkaari n-i-ri mbere i-no kana? 9. vehicle f-SM9-&e 9.front 9-this q 'Is the vehicle here in front?' b. nteere ino i-nya ntune ci-onthe 10. side 10.this 10-four 10.red 10-all 'all these four red sides' (N-Dem-Num-Adj-Quant) I return to the structures for these examples in §6. 4.5. Head-marking and dependent-marking of the AxPart:Ground relationship In most of the languages reviewed, the Ground DP is morphologically related to the AxPart in the same way that possessors are related to their possessees, which is often the same way that DP dependents on NP are expressed, for example by genitive case. Overt markers for genitive case are seen for example in Japanese, French, and Kannada AxPart constructions, illustrated here with Japanese. 64 Peter Svenonius (38) a. Taroo-no ie-ga Taroo-GEN house-NOM 'Taroo's house' b. ie-no mae-de house-GEN front-LOC 'in front of the house' In Finnish, the genitive has syncretized with the accusative, and the Ground DP is in that case. In Kiitharaka, there is an associative marker for DP dependents of N, and this is used with AxParts, illustrated below (example from Muriungi 2006), and in French the marker de also functions both to mark possessors and Grounds. (39) a. ga-tigati k-a metha 12-center 12-of 9. table 'the center of the table' b. gi-kombe ki-a Maria 7-cup 7-of I.Maria 'Maria's cup' In Chinese and Persian, the marker used to link Ground DPs with their AxParts is a generalized marker for nominal dependents (including possessors and thematic complements but also adjectives and other modifiers). In Tzeltal and Kham, morphology also signals a possessor or genitive-like relation between the AxPart and the Ground, illustrated for Kham in (22) above, and again (with different examples) below. (40) a. u-du:h-la 3POSSK-beneath-in 'underneath it' b. u-kwi:-ye ja-ke-o 3POSSR-/ianrf-iNSTR mafce-PFV-3s 'He made it with his hands' (Kham, Watters 2002:57, 132) Thus it can be said that it is a typical property of AxParts that the Ground DP is marked with genitive case, or otherwise as a nominal dependent, as if the AxPart were a noun. 5. Differences between AxPart and N Given all the similarities between AxParts and Ns noted in §4, it may not be surprising that AxParts are often analyzed as nouns. In this section, I note various differences between N and AxPart which suggest that they should be distinguished. To say that AxParts are a subcase of N is essentially to say that there are some sufficient diagnostics for N-hood, for example gender, and the AxParts have it. Here I present three different types of argument against this. The 65 The Emergence of Axial Parts first is that there are elements in each language which do not pass these diagnostics, e.g. there is one AxPart in Kiitharaka which does not belong to any gender class, therefore there must be a category AxPart, distinct from N. The second type of argument is that even when morphology appears which is from a nominal paradigm, it does not have the same semantic content as when it appears with a noun. The third argument is that many of the items which I have identified as AxParts may have N uses as well as AxPart uses; I argue that these uses show that even a word which has a gender or other typical nominal characteristics can be used as an AxPart (as identified by the special characteristics noted above) rather than as a noun. The relationship between N and AxPart in such cases may be either historical or derivational. 5.1. AxParts which do not have gender Muriungi (2006) identifies a single AxPart in Kiitharaka which does not have a noun class prefix, namely karibu 'near' (cf. English vicinity). Compare karibu 'near' with the noun class 11 AxPart ru-teere 'side' (examples from Peter Muriungi, personal communication). (41) a. Mbea iri karibu na i-buku. 9.mouse is near with 5-book 'The mouse is near the book' b. Mbea iri ru-teere ru-a i-buku. 9.mouse is 11-side 11-of 5-book 'The mouse is beside the book' Of course, it is possible that these two elements are not of the same category; more careful investigation will be necessary to be sure. But assuming that they are, the example of karibu 'near' shows that not all AxParts have noun class features in Kiitharaka. Assuming that all nouns do have noun class features, AxParts will then have to be distinguished from nouns in general. 5.2. Different interpretation of plural Generally, AxParts do not pluralize, as noted for English in §1 and for other languages in the various articles cited above. Although an example was given above in (29) of a plural AxPart from Kiitharaka, that example was exceptional; normally, AxParts in Kiitharaka do not pluralize, or if they appear in the plural it is only with their nominal meaning, as illustrated here, from Muriungi (2006). (42) i-guru ~ ma-guru 5-top 6-top 'on top' ~ 'legs' 66 Peter Svenonius Thus, (29) can be said to be exceptional at best. Possibly, (29) can also be understood as a nominal use of ndungu, rather than an AxPart, though this needs to be investigated further. The Persian case discussed by Pantcheva (2006a) is very interesting in that the interpretation of the apparently plural form of AxParts is distinct from a nominal plural interpretation. Her example from (28) above is repeated here. (43) Shekaer rixt in zir-ha-ye miz. sugar spilled this under-PL-EZ table 'The sugar spilled here all over under the table' Note that the plural marker does not have its ordinary interpretation here— the plural of zir 'under' does not refer to plural spaces underneath, nor to plural Grounds—but gets a kind of distributive reading, translated as 'here and there' (see Pantcheva 2006b for discussion). The general pattern is that plural is absent from AxParts, surprising if they are nouns. Even more surprisingly for a nominal analysis of AxParts, when plural morphology does appear it seems not to have its usual meaning. 5.3. Articles, quantifiers, and demonstratives 5.3.1. Idiosyncratic Determiners As noted above, determiners appear with some AxParts. If all AxParts required articles in the same contexts that nouns do, then clearly a generalization would be missed if AxParts were claimed to be distinct from nouns; but the distribution of articles with AxParts is highly idiosyncratic; even in French, where articles are ordinarily obligatory with nouns, there are many AxParts which lack them. Roy (2006) notes that for example á côté de 'at side of and á travers de 'at traverse of ('across from') allow the omission of the article, while others do not. Furthermore, some AxParts which are not nouns appear with the article nonetheless (Roy notes au long de, 'along,' and au delá de, 'beyond'). The definite articles in these AxPart constructions do not seem to have the semantic contribution of definite articles with nouns; for example they do not presuppose familiar discourse referents. 5.3.2. Idiosyncratic Quantifiers Quantificational determiners are even more restricted. A Persian example from Pantcheva (2006a) illustrates the general tendency. (44) (*haer) ru-ye eshkaf every on-EZ wardrobe '(*every) on the wardrobe' 67 The Emergence of Axial Parts An example from Kiitharaka was given in (33) of a quantifier with an apparent AxPart. However, that case was like the French and English ones involving the definite article in that it is highly lexically idiosyncratic. As Muriungi notes, only certain AxParts admit the modification of only certain nominal modifiers. If the patterns must be lexically listed in any case, then no generalization is gained by calling all AxParts nouns. Consider, in this light, the distribution of all in English, in combinations like all over the floor, all around the house, all along the river. The fact that all is primarily a quantifier in the D-system has not led investigators into English to call over, around, and along nouns. I maintain that examples like (33) provide no greater reason to call ndungu a noun. 5.3.3. Phrasal Demonstratives As noted above, demonstratives are normally part of the D-system. However, there is reason to suspect that they may in many cases be phrase-like rather than head-like; for example, the order N-Dem correlates statistically with VO and prepositions, while Dem-N correlates with OV and postpositions.7 If the category demonstrative is a phrasal dependent of a projection of D, then it is possible that the same category can also be merged as a phrasal dependent of some projection of P (much as PP is standardly assumed to be a possible adjunct both to NP and to VP). 5.4. Adjectival modification AxParts, even those with nominal etymology or morphology, generally do not accept adjectival modification, as illustrated for English in (4a). I illustrate here with Korean. First, consider the Korean pair in (45), which corresponds to the English difference between in the front of and in front of; the noun in Korean is a compound ap-pwupwun, 'front-part,' and its complement appears in the genitive case; the AxPart in Korean is simply ap(h) 'front' and does not take the genitive, as indicated. (45) a. Kay-han mali-ka cha-uy ap-pwupwun-ey anc-a iss-ta. dog-one CL-nom car-gen front-part-LOC sit-CONN be-DC 'A dog is sitting on the front of the car' b. Kay-han mali-ka cha ap-ey anc-a iss-ta. dog-one CL-nom car front-LOC sit-CONN be-DC 'A dog is sitting in front of the car' 7For example, looking at the 872 languages that WALS (Haspelmath et al. 2005) lists as either VO or OV and as either Dem-N and N-Dem (i.e. excluding mixed cases, free word order, etc.), about 70% of VO languages are N-Dem, and about 70% of OV languages are Dem-N; similarly, of the 725 languages listed both as either prepositional or postpositional and as either Dem-N or N-Dem, about 70% of prepositional languages are N-Dem and about 70% of postpositional languages are Dem-N. See Svenonius (2006) for more discussion of the ordering of elements in the DP. 68 Peter Svenonius Compare the English minimal pairs in (l)-(2) in §1. The pair in (46) shows that the noun, but not the AxPart, can be modified by a suitable adjective such as kunulcin 'shaded, shady.' (46) a. Kay-han mali-ka cha-uy (kunulcin) ap-pwupwun-ey dog-one CL-nom car-gen shady front-part-LOC anc-a iss-ta. sit-CONN be-DC 'A dog is sitting on the (shady) front of the car' b. Kay-han mali-ka cha (*kunulcin) ap-ey anc-a iss-ta. dog-one CL-nom car shady front-LOC sit-CONN be-DC 'A dog is sitting in (*shady) front of the car' This generalization is fairly robust; for most of the languages examined in the survey here it holds absolutely; see Pantcheva (2006a) regarding Persian, Muriungi (2006) regarding Kiitharaka, Takamine (2006) for Japanese, and Roy (2006) for French. An exception appears to be the English adjective immediate, which appears with certain spatial words which otherwise show the syntax of AxParts, for example vicinity or left, as shown in (47). (47) a. There was a policeman in the (immediate) vicinity of the house. b. The towels are to the (immediate) left of the sink. As noted above, these words also deviate from the syntax of in front of and on top of in taking a definite article. Interestingly, the possibility of the definite article and of the adjective appear to be linked. (48) a. The towels are left of the sink. b. *The towels are (immediate) left of the sink. Nonetheless, despite the apparent counterexample of immediate, the robust ungrammaticality of adjectives with AxParts suggests that AxParts are not nouns, or not fully nouns, or not ordinary nouns. 5.5. The underspecifled meanings of head-marking and dependent-marking In §4.5 I discussed the fact that Ground dependents of AxParts are often related to them morphologically in the same way that possessors are related to possessees, for example in the case-marking on the Ground and possessor, or on agreement marking on the possessee and AxPart. A full understanding of this situation would require a more detailed analysis of the possessor-possessee relationship than will be possible here; I will simply comment on the outline of the solution. 69 The Emergence of Axial Parts Generally, the morphological cases available to a language underdeter-mine the grammatical relations that the cases are used to express; thus, for example in Icelandic the dative is used to mark benefactive indirect objects, experiencer subjects, and direct objects which are themes of ballistic motion, for example (see e.g. Maling 2001). Thus, that the genitive case is used in some languages for both Grounds and possessors does not automatically motivate a collapse of the two. Of course, the systematic patterning of Grounds with possessors bears explanation. But the point of this section is to argue against the notion that AxParts are identical to nouns, and to that end it is important to point out that the nominal-like head or dependent-marking is often present even when the AxPart is not derived from a noun; some of the French examples, for example, have de even when they are not etymologically nouns (e.g. Roy 2006 notes au delá de 'at beyond of ('beyond') and au long de 'at long of ('along')); and Icelandic also shows cases of genitive with what are plausibly AxParts derived from directional particles.8 I mentioned above that Korean shows genitive case on possessors but not on Grounds appearing with AxParts (cf. (17)). Kham is similar in this respect. A full DP Ground in Kham appears to the left of the AxPart with no overt case morphology (as illustrated in (49)). (49) a. hä: khí:-ka cliff foot-at 'at the foot of the cliff b. zihm lap-ka house side-AT 'beside the house' c. juhr dú:h-la boulder beneath-m 'under the boulder' (Waiters 2002:138) A full DP possessor, on the other hand, is marked with genitive case and the possessed N is marked with a possessor prefix.9 (50) a. mi:-ra-e ya-sun-ra ya-cä:-di-ra person-PL-GEN 3p-gold-PL 3p-silver-PL 'the people's gold and silver' b. la:-ra-e ya-raja leopard-PL-GEN 2>p-king 'the leopards' king' (Waiters 2002:435-436) 8This is an argument that I am developing in ongoing work, presented among other places at the Workshop on Comparative Germanic Syntax in Santa Cruz, California in April 2006. ^Examples are also found without one or the other, e.g. 'Manlaľs blanket' p. 423 with no genitive case, 'leopards' meeting-place' p. 436 with no possessor prefix. I have not been able to determine the exact conditions. 70 Peter Svenonius Thus, a nominal analysis of AxParts in Kham would have to stipulate that these nominals are special in allowing DP dependents with no overt case marker. 6. Refining AxParts I am assuming what Borer (2005) calls 'Neo-Constructivism,' the hypothesis that certain aspects of meaning are contributed by syntactic structure, while other aspects of meaning come from the lexical content of the words inserted into the syntax. A lexical item like English front is polysemous. Inserted under an N node in a tree, it expresses a noun, and will combine with plural and determiners and so on. It has other features as well, for example it is count and not mass, and this will restrict what sort of N structure it is inserted in. Inserted under an AxPart node, it will express an Axial Part, and combine with Place and Path and so on. Here, too, it has certain properties, for example it requires the Place head to be in. In those languages which allow demonstratives with AxParts (discussed in §4.4.3), the lexical semantic contribution of the demonstrative could be characterized as consisting entirely of its proximal or distal deictic content; when inserted in a projection of D, it is interpreted as a nominal demonstrative (this or that); and when inserted into a projection of P, it is interpreted as locative (here or there). The Korean example in (51) would have the universal hierarchical structure as depicted in (52), presented with the heads on the right though obviously it could be translated into a roll-up structure. Note, though, that if the demonstrative is hierarchically higher than the AxPart, as suggested by the interpretation, then the Ground (or the KP containing the DP Ground) must have moved across it, as suggested in the following diagram.10 Ku sangca-nun oscang ce mit-ey twu-ess-ta. the box-TOP chest that bottom-LOC place-PAST-DC 'I put the box over there under the chest' (51) (52) AxPart facet t; I mit Now, the distal element ce 'that' which is also merged in DPs is identified oscang 'chest' 10Takamine (2006) suggests for a similar configuration in Japanese that the case head originates higher than AxPart, rather than being moved there as I have depicted here. 71 The Emergence of Axial Parts as the lexicalization of a distal feature somewhere above AxPart but below the landing site of the leftward-moved KP. Tortora (2005) argues, on the basis of data from Italian and Spanish, for an aspectual element in the extended projection of P, which can be used to express a notion of boundedness. She argues that the a in (53a) represents that element, and that the preposition dentro 'inside' has moved across it. (53) a. Vai dentro a-lla stanza. go inside at-the room 'Go inside the room' (bounded) b. Vai dentro la stanza. go inside the room 'Go inside the room' (unbounded) (Italian, Tortora 2005:314) Arguably, the 'distributed' reading of the Persian plural marker, when appearing with AxParts, is of the same general type as Tortora's bounded marker. Suppose, then, that these are both examples of Place, as illustrated in (55), for the Persian example in (54) (repeated from (28) above). The marked value for Place here is dist for 'distributed.' (54) Shekaer rixt in zir-ha-ye miz. sugar spilled this under-PL-EZ table 'The sugar spilled here all over under the table' (55) Place miz 'table' The surface order of the AxPart and Place heads is reversed; this can be modelled in terms of head-movement or in other ways. In Russian, the unexplained nominal affix was a case marker. The two options are sketched for an example from (30a) in §4.3 above, repeated in (56) below. The case value itself is determined by the higher context (a Path head with the value to controls Accusative case). Thus, the semantic contribution of the case is made at a higher level, and the spell-out of locative or accusative on the AxPart represents agreement. 72 Peter Svenonius (56) v-perěd mašiny ~ v-peredi mašiny in-front.ACC car.GEN in-front.LOC car.GEN '(to) in front of the car' — '(at) in front of the car' An interesting difference between these examples and the previous ones is the presence of an overt preposition in addition to the case head. Possibly, the head v in these cases is a relational head introducing the Figure of the prepositional phrase. I label it p, in the tentative representation in (57), and give it the value CONTD for 'contained.' (57) p CONTD Place pered- maši 'car' The French and Kiitharaka examples can be treated similarly, with the noun class prefixes and definite articles being dummy expressions of a head above AxPart, possibly Place. Lexical items inserted in the tree which happen to also be nouns bear gender features, and this can control agreement, but they are not inserted under N nodes and therefore there is no NP or DP to host adjectives and quantifiers and to establish reference and so on. To sum up this section, I suggest that a Neo-Constructivist approach to PP structure can accommodate the many nominal elements that appear in the PP. The reasons for the various polysemies needed for such an analysis would be likely to be historical and external to the system of UG itself. For an interesting alternative involving a null nominal element PLACE, see Terzi (2004), Pantcheva (2006b) and references there. 7. The historical emergence of AxParts There is a seemingly simple historical explanation for many cases in which an adpositional element has nominal characteristics, namely that a noun has been reanalyzed as an adposition. A case I mentioned in §1 is beside, originally from be sidan, 'by side,' where sidan is a dative form of side, 'side.'11 However, this explanation turns out to have serious limitations. Other Old English adpositions with this same basic form include be+nidan 'beneath,' be+foran 'before, in front of,' and be+tweonum 'between'; but apparently none of these involves a noun. The words nidan, a form of nider 11 Etymologies here are mainly based on the Oxford English Dictionary, the American Heritage College Dictionary, 3rd ed., and the Bosworth-Toller Anglo-Saxon Dictionary. 73 The Emergence of Axial Parts 'down' and foran, a form of 'forth,' are directional particles or adverbs and tweonum 'two each' seems to be some sort of quantifier (related to two, 'two,' cf. Watkins 1985:16). Thus, it cannot be said that the be- series of adpositions in English comes from a reanalysis of a preposition plus a noun. I speculate that instead, be served in these examples as a Place head, and different lexical elements were recruited to serve as AxParts. These elements might be recruited from the ranks of nouns, adverbial directional particles, or quantifiers. Another illustrative series is the one beginning with a-:12 about, above, along, among, across, around, atop, and so on. Historically, some of them can be traced back to an origin starting with a form of the preposition on 'on, in.' Again, however, it would be wrong to think that they systematically consisted of a preposition plus a noun, which became reanalyzed as an adposition. Their history is more complex. About and above are formed from butan 'outside' and bufan 'over,' which themselves include be plus the directional particles utan related to 'out' and ufan related to 'up,' just as with beneath and before. Among comes from on gemang 'in crowd,' with a collective prefix ge- on a root which is apparently verbal (cf. mengan 'mix,' Watkins 1985:38). The forms across, around, atop appeared much later, along with forms that have never became complement-taking prepositions, such as abed, afoot, asleep, ashore, afield (cf. atop the dresser but * ashore the island). In some cases it is not clear that the element after a(n)- was ever a noun; consider for example afloat, which appears to be formed on a verb, or the now-defunct alow, which appears to be formed from the adjective low. Again, if a(n) was a Place head, then it seems that material from different categories was recruited to serve as AxParts. Clearly, a more careful historical investigation into these elements is warranted. The tentative conclusion, however, is that there are other sources for complex prepositions than P+N. This is expected if UG provides a functional lattice on which to build spatial constructions, and various lexical material is recruited to fill this lattice. 8. Conclusion I have suggested that a striking cross-linguistic pattern of prepositional syntax be provided with a uniform analysis in terms of a category AxPart, which translates semantically as a region on the basis of which a vector space is constructed. I have argued against the idea that AxParts are a subclass of category N, though N is undeniably an important source for AxParts diachronically. On this Neo-constructivist analysis, lexical material is inserted into a syntactico-semantic structure. Polysemous lexical material can be inserted under nodes of different categories, retaining its conceptual 12 Thanks to Ans van Kemenade for helpful discussion. 74 Peter Svenonius and phonological content, but showing different syntactic behavior in the different syntactic environments. References Asbury, Anna. 2006. Towards a typology of morphological case. In Nord-lyd, Troms0 Working Papers in Language & Linguistics: 33.1, Special Issue on Adpositions, edited by Peter Svenonius and Marina Pantcheva, pp. 120-148. University of Troms0, Troms0. Available at http://www.ub.uit.no/baser/nordlyd/. Blake, Barry J. 2001. Case. Cambridge University Press, Cambridge. Borer, Hagit. 2005. In Name Only (Structuring Sense, vol. 1). Oxford University Press, Oxford. Brown, Penelope. 1994. The INs and ONs of Tzeltal locative expressions: The semantics of static descriptions of location. Linguistics 32: 743-790. Carter, Richard. 1988. Compositionality and polysemy. In On Linking: Papers by Richard Carter, edited by Beth Levin and Carol Tenny, pp. 167-204. Center for Cognitive Science, MIT, Cambridge, Ma. Corbett, Greville G. 1991. Gender. Cambridge University Press, Cambridge. Fong, Vivienne. 1997. The Order of Things: What Directional Locatives Denote. Ph.D. thesis, Stanford. Greenberg, Joseph H. 1978. How does a language acquire gender markers? In Universals of Human Language, edited by Joseph H. Greenberg, Charles A. Ferguson, and Edith A. Moravcsik, vol. Ill, Word Structure, pp. 47-82. Stanford University Press, Stanford, Ca. Haspelmath, Martin, Matthew S. Dryer, David Gil, and Bernard Comrie (eds.) . 2005. The World Atlas of Language Structures. Oxford University Press, Oxford. Hegediis, Veronika. 2006. Hungarian spatial PPs. In Nordlyd, Troms0 Working Papers in Language & Linguistics: 33.2, Special Issue on Adpositions, edited by Peter Svenonius, pp. 220-233. University of Troms0, Troms0. Available at http://www.ub.uit.no/baser/nordlyd/. Jackendoff, Ray. 1996. The architecture of the linguistic-spatial interface. In Language and Space, edited by Paul Bloom, Mary A. Peterson, Lynn Nadel, and Merrill F. Garrett, pp. 1-30. MIT Press, Cambridge, Ma. Karlsson, Fred. 1977. Finsk grammatik. Suomalaisen Kirjallisuuden Seura, Helsinki. Kracht, Marcus. 2002. On the semantics of locatives. Linguistics and Philosophy 25: 157-232. Levinson, Stephen C. 1994. Vision, shape, and linguistic description: Tzeltal body-part terminology and object description. Linguistics 32: 791-855. 75 The Emergence of Axial Parts Maling, Joan. 2001. Dative: The heterogeneity of mapping among morphological case, grammatical functions, and thematic roles. Lingua 111: 419-464. Marr, David. 1982. Vision: A Computational Investigation into the Human Representationand Processing of Visual Information. W.H. Freeman, New York. Muriungi, Peter. 2006. Categorizing adpositions in Kiitharaka. In Nord-lyd, Troms0 Working Papers in Language & Linguistics: 33.1, Special Issue on Adpositions, edited by Peter Svenonius and Marina Pantcheva, pp. 26-48. University of Troms0, Troms0. Available at http: / / www. ub .uit .no/baser / nordlyd /. Nam, Seungho. 1995. The Semantics of Locative PPs in English. Ph.D. thesis, UCLA. Pantcheva, Marina. 2006a. Persian preposition classes. In Nordlyd, Troms0 Working Papers in Language & Linguistics: 33.1, Special Issue on Adpositions, edited by Peter Svenonius and Marina Pantcheva, pp. 1-25. University of Troms0, Troms0. Available at http: / / www. ub .uit .no/baser /nordlyd/. Pantcheva, Marina. 2006b. The place of PLACE in Persian. Ms. CASTL, University of Troms0. Roy, Isabelle. 2006. Body part nouns in expressions of location in French. In Nordlyd, Troms0 Working Papers in Language & Linguistics: 33.1, Special Issue on Adpositions, edited by Peter Svenonius and Marina Pantcheva, pp. 98-119. University of Troms0, Troms0. Available at http: / / www. ub .uit .no/baser /nordlyd/. Samiian, Vida. 1994. The Ezafe Construction: Some implications for the theory of X-bar Syntax. In Persian Studies in North America: Studies in Honor of Mohammed Ali Jazayery, edited by Mehdi Marashi, pp. 17-41. Iranbook, Bethesda, Md. Son, Minjeong. 2006. Syntactic status of spatial terms in Korean locative expressions. Handout from Moving Right Along seminar in Troms0. Starke, Michal. 1993. Notes on prepositions and clause-structure. Memoire de Diplome, University of Geneva. Svenonius, Peter. 2003. Limits on P: filling in holes vs. falling in holes. Nordlyd, Troms0 Working Papers on Language and Linguistics. Proceedings of the 19th Scandinavian Conference of Linguistics 31 2: 431-445. Svenonius, Peter. 2004a. Adpositions, particles, and the arguments they introduce. Ms. University of Troms0; to appear in a volume on Argument Structure (John Benjamins); available at http://ling.auf.net/lingBuzz/000042. Svenonius, Peter. 2004b. Spatial prepositions in English. Ms. University of Troms0; available at http://ling.auf.net/lingBuzz/000001. Svenonius, Peter. 2006. The position of adjectives and other phrasal modifiers in the decomposition of DP. Ms. University of Troms0, available 76 Peter Svenonius at http://ling.auf.net/lingBuzz/000329. Svenonius, Peter, to appear. Interpreting uninterpretable features. Linguistic Analysis . Takamine, Kaori. 2006. Categorical properties of space expressions in Japanese. In Nordlyd, Troms0 Working Papers in Language & Linguistics: 33.1, Special Lssue on Adpositions, edited by Peter Svenonius and Marina Pantcheva, pp. 78-97. University of Troms0, Troms0. Available at http://www.ub.uit.no/baser/nordlyd/. Terzi, Arhonto. 2004. Locative prepositions and possession. Ms., Technological Educational Institute of Patras. Tortora, Christina. 2005. The preposition's preposition in Italian: Evidence for the boundedness of space. In Theoretical and Experimental Approaches to Romance Linguistics, edited by Randall S. Gess and Edward J. Rubin, pp. 307-327. John Benjamins, Amsterdam. Watkins, Calvert (ed.) . 1985. The American Heritage Dictionary of Lndo-European Roots. Houghton Mifflin, Boston. Watters, David E. 2002. A Grammar of Kham. Cambridge University Press, Cambridge. Winter, Yoad. 2001. Measure phrase modification in vector space semantics. In WCCFL 20 Proceedings, edited by K. Megerdoomian and L.A. Bar-el, pp. 607-620. Cascadilla, Somerville, Ma. Wunderlich, Dieter. 1991. How do prepositional phrases fit into compositional syntax and semantics? Linguistics 29: 591-621. Zwarts, Joost. 1997. Vectors as relative positions: A compositional semantics of modified PPs. Journal of Semantics 14: 57-86. Zwarts, Joost and Yoad Winter. 2000. Vector space semantics: A model-theoretic analysis of locative prepositions. Journal of Logic, Language, and Lnformation 9: 169-211. 77