Spanish prepositions and silent place

In this work I investigate the syntax and interpretation of two groups of spatial prepositions in Spanish: “small” Ps (e.g. bajo ‘under’) and “big” Ps (e.g. debajo ‘de.under’). I show that small versus big Ps display a series of asymmetries, such as the (in)ability to take bare nominals and the (un)availability of coordination and adverb intervention. I propose that these contrasts can be derived under the hypothesis that small Ps select a single nominal complement, whereas big Ps select a more complex Possessor-Possessum structure, where the nominal occupies a specifier position, acting as the possessor of a silent PLACE element. Furthermore, I argue that unifying big Ps (but crucially not small Ps) with possessive structures accounts for the delimited versus non-delimited interpretation of these Ps.


Introduction
This paper addresses the syntax of two groups of spatial prepositions in Spanish. These prepositions have the property that they come in pairs, and both members of the pair are very often reported by native speakers to be equivalent in meaning. (1) below shows an example with the bajo 'under '/debajo 'de.under' pair, (2) with tras 'behind'/detrás ' de. behind' and (3) with ante 'front'/delante 'de.l.front '. 1 (1) a. El libro está bajo la mesa. the book is under the (2) a. Juan estaba escondido tras la columna. Juan was hidden behind the column 'Juan was hidden behind the column.' b. Juan estaba escondido detrás de la columna. Juan was hidden de.behind of the column 'Juan was hidden behind the column.' structural complexity of big Ps put forth in this work. In section 4, I address the differences in interpretation between small and big Ps. I show that small Ps may have locative or non-locative interpretations. In contrast, big Ps necessarily have a locative interpretation. I claim that this asymmetry follows from the presence of a silent PLACE element in the structure of big Ps, but not in the structure of small Ps. I also present evidence that the nominal in the complement of a small P may have a delimited or a non-delimited interpretation (a notion I present and discuss in section 4), whereas the nominal in the complement of a big P has to be necessarily interpreted as delimited. I argue that this is a consequence of the fact that the nominal in the structure of a big P is a possessor (in particular, the possessor of PLACE), whereas the nominal in the structure of a small P is not. Finally, in section 5 I review all the evidence gathered and show how each of the findings lends support to the structures proposed. In this last section I also raise some open questions and conclude.

Small and big Ps
In this section I analyze the behavior of small and big Ps with respect to bare nominals. First, I review some general observations made about the distribution of bare nominals in Spanish inside the clause. I show how these facts are best captured if we assume, following Cuervo (2003), that "an unmodified common noun cannot be an A-specifier." 4 The relevance of this observation is that the distribution of bare nominals can now serve as a diagnostic to test further structures: if a bare nominal is not allowed in a certain structure, it is because it is occupying an A-specifier position and, conversely, if a bare nominal is allowed in a certain structure, then it follows that it is not occupying an A-specifier position. Then, I show how small and big Ps pattern differently with respect to the possibility of allowing a bare nominal. This will be taken as evidence that the nominal is a complement in the case of small Ps (bajo tierra 'under earth') but an A-specifier in the case of big Ps (*debajo de tierra 'de.under of earth').

Unmodified bare nominals in Spanish
It has been widely observed in the literature (Suñer 1982;Contreras 1985;Bosque 1990; among others) that unmodified bare nominals in Spanish have a restricted distribution. Some examples of these restrictions are given in (4), (5) and (6) below (examples (4) and (5) are taken from Cuervo 2003, and example (6) is adapted from Bosque 1990).
oil.nom was found in Jamaica 'Oil was found in Jamaica.' The contrast between (4a) and (4b) shows that the restriction on bare nominals applies to subjects (4b) and not to objects (4a). The examples with unaccusatives and passives, in (5) and (6) respectively, show, furthermore, that the restriction applies to a subset of subjects, namely, preverbal ones. As we can see from the facts in (5) and (6), the sentences are grammatical if the bare nominal subjects amigos 'friends' and petróleo 'oil' appear postverbally, but they are ungrammatical if they appear preverbally. Naturally, if the nominals in (4)-(6) are not bare, the (b) sentences become grammatical.
The observation that bare nominals cannot appear as preverbal subjects is captured by Suñer (1982) as The Naked Noun Phrase Constraint, whose formulation is given in (7).

(7)
The Naked Noun Phrase Constraint An unmodified common noun in preverbal position cannot be the surface subject of a sentence under conditions of normal stress and intonation. (Suñer 1982: 209) However, as Cuervo (2003) points out, the constraint in (7) needs to be refined. Although it accurately describes the asymmetries in (4)-(6), it fails to capture two important sets of facts: the impossibility of bare nominals as the subjects of unergative verbs and the impossibility of bare nominals as the subjects of small clauses. In (8) below we see an example where the bare nominal chicos 'kids' appears as the subject of the unergative verb festejaron 'celebrated' (examples from Cuervo 2003). In (8a) chicos 'kids' is a preverbal subject and the sentence is predicted to be ruled out by (7), which bans, precisely, bare nominals in preverbal subject position. In (8b), on the other hand, the subject chicos 'kids' appears postverbally and, according to (7), is expected to be grammatical (the formulation states that bare nominals are banned when they appear as preverbal subjects, and this is not the case in (8b), where chicos 'kids' appears postverbally). In other words, the formulation in (7) accounts for the facts in (4)-(6) but makes the wrong predictions for (8).
Apart from this, there is a set of facts that is simply not covered by (7). This is the case of bare nominals as the subject of small clauses, an example of which is given in (9) (example from Cuervo 2003).
(Cf. las películas) your friend considered interesting movies.acc 'Your friend used to consider movies to be interesting.' In (9) we can see that a bare nominal like películas 'movies' cannot be the subject of a small clause (in this particular case, it cannot be the subject of interesantes 'interesting'). In order for the ungrammaticality of this sentence to receive the same explanation as the ungrammaticality of the examples mentioned before, the formulation in (7) needs to be revised to include not only the subjects of verbs, but other subjects as well.
Cuervo proposes the following formulation to explain the bare nominal facts. (10) The Naked Noun Phrase Constraint Revised An unmodified common noun cannot be an A-specifier. (Cuervo 2003: 115) Let's now briefly review how the new formulation in (10) captures the bare nominal facts presented in this section. The bare nominals vino 'wine', amigos 'friends' and petróleo 'oil' are in complement position in (4a), (5a) and (6a), and they are therefore licit. In contrast, they are A-specifiers in (4b), (5b) and (6b) and they are consequently ruled out. The subject of an unergative verb is always merged as an A-specifier (standardly, of little v). This is why a bare nominal such as chicos 'kids' is never possible as the subject of an unergative verb, independently of whether it appears postverbally (8b) or preverbally (8a). Finally, the bare nominal películas 'movies' in (9) is merged as the A-specifier of the AP interestantes 'interesting' (or, alternatively, some higher functional projection). This is disallowed by (10), and its ungrammaticality follows. A fundamental advantage of the formulation in (10) over (7) is that it makes it possible to account for the ungrammaticality of bare nominals not only as the subjects of verbs or adjectives but also in other domains of the grammar. As a matter of fact, the extension of the formulation in (10) to account for the (un)grammaticality of bare nominals in the domain of spatial expressions is, precisely, one of the contributions made by this work. 5 In the next section I discuss the asymmetric behavior of small and big Ps in their ability to allow bare nominals. Following (10), I argue that this asymmetry is explained if the nominal that follows a small P (e.g. bajo 'under') is its complement, whereas the apparent complement of a big P (e.g. debajo 'de.under') is really in an A-specifier position.

Small and big Ps and bare nominals
As mentioned in section 1, it is a well-known fact that most varieties of Spanish display pairs of locative prepositions like the ones shown in the (a) and (b) Cuervo (2003) to propose the more general formulation in (10) over one that simply bans a bare nominal from being the subject of a predicate is that (10) also makes it possible to account for the ungrammaticality of bare nouns as dative arguments since, under Cuervo's analysis, datives are always licensed as A-specifiers. The reader is referred to Cuervo (2003) for examples and discussion. It is interesting to note that an unmodified bare nominal such as tierra 'earth' can appear as the complement of a small P like bajo 'under', as shown in (14).

(14)
El pirata escondió el tesoro bajo tierra. the pirate hid the treasure under earth 'The pirate hid the treasure underground.' However, this is not the case for all Ps. A bare nominal cannot be the complement of a big P like debajo 'de.under' (15). 6 (15) *El pirata escondió el tesoro debajo de tierra. 6 the pirate hid the treasure de.under of earth 'The pirate hid the treasure underground.' It appears to be the case, furthermore, that the difference in the availability vs. unavailability of bare nominals is not unique to the bajo 'under'/debajo 'de.under' pair, but is, rather, a contrast exhibited by small and big Ps in general, as shown below. In (16) we can see one more example with bajo 'under' and debajo 'de.under', (17) and (18)  What the data in (14)-(18) show is that whereas a bare nominal is possible as the complement of a small P like bajo 'under', tras 'behind' and sobre 'on', it is not possible as the complement of a big P like debajo 'de.under',detrás 'de.behind' and encima 'on.top'. In (16a) bajo 'under' is followed by the bare mass noun nieve 'snow' and the sentence is grammatical. On the other hand, in (16b) debajo 'de.under' is followed by the same nominal (introduced by de) and the sentence is ungrammatical. In a similar fashion, tras 'behind' can be followed by the bare plural muros 'walls', but a sentence where detrás 'de. behind' is followed by muros 'walls' (again, introduced by de) is marginal. 8 Finally, the P sobre 'on' can be followed by a bare mass noun like madera 'wood'. However, this option is unavailable for encima 'on.top'. 9 I propose that this contrast between small and big Ps follows directly from Cuervo's revised version of the Naked Noun Phrase Constraint, repeated below.

(19)
The Naked Noun Phrase Constraint Revised An unmodified common noun cannot be an A-specifier. (Cuervo 2003: 115) Specifically, I propose that bare nominals are banned in the Spanish PPs in the (b) examples in (16) In the following section I motivate the structures in (20) and (21). Building on work by Terzi (2005;2008; and others, I argue that (21) contains a silent noun PLACE and that what is standardly taken to be the complement of the locative P, (de) la mesa '(of) the table' in (21), is the possessor of silent PLACE. In contrast, I argue that the P in (20) does not select a Possessor-Possessum structure with silent PLACE, but rather a simple DP. This is going to set the stage for section 3, where we see that the numerous predictions made by these structures are borne out. 10

Big Ps and silent PLACE
Authors like Terzi (2005;2008; and Noonan (2010), among others, have argued that an unpronounced noun PLACE (henceforth, "silent PLACE") forms part of the structure of locative Ps. They have proposed, furthermore, that what is standardly taken to be the complement of the locative P (e.g. la mesa 'the table' in debajo de la mesa 'de. under of the table') is the possessor of silent PLACE. In this section, I very briefly go over the arguments put forward by Terzi (2010) to support her analysis. I propose that while locative Ps can select a Possessor-Possessum structure containing a silent PLACE element, this is not the case for all Ps. While big Ps like debajo 'de.under' do select this structure, small Ps like bajo 'under' select a single DP complement. I argue that by proposing these different structures for big and small Ps, we can arrive at a straightforward explanation for a series of syntactic and interpretative asymmetries exhibited by these groups of Ps. These asymmetries are the object of discussion of section 3. One of the main pieces of evidence that leads Terzi (2010) to propose that there is a silent noun PLACE in the structure of locative Ps comes from genitive clitics in Modern Greek. In Modern Greek, when a clitic appears as the complement of a locative P it must necessarily appear in the genitive case, as shown in (22) and (23) below (examples taken from Terzi 2010).

(22)
Stathika piso tis. stood-1sg behind she-cl-gen 'I stood behind her.' (23) Kathomun epano tu. was.sitting-1sg on he-cl-gen 'I was sitting on him.' As genitive case appears primarily on the complement of nominals in Modern Greek, this seems to constitute strong evidence that a nominal element of some kind is involved in these structures.
Once having established the existence of a nominal in the structure of locative Ps, Terzi points out that there is a strong parallelism between adjectives and locative Ps. This leads her to propose that locative Ps are not themselves nouns, as has been argued for in the literature (Marácz 1984;Bresnan 1994;Collins 2004;Aboh 2010), but that they are rather modifiers, modifiers of a noun, specifically, they are modifiers of silent PLACE.
Terzi then extends her analysis of Greek locative Ps to Spanish locative Ps, and suggests that locative Ps in Spanish can also be argued to contain an unpronounced PLACE noun. The presence of this unpronounced noun is responsible for the "nominal flavor" of these Ps, an observation that has very often been reported in the literature, but that has never received a satisfactory explanation. Specifically, Terzi points out that the presence of silent PLACE might explain, among other things, why a locative PP headed by debajo 'de. under' in Spanish can follow a P like desde 'from' as shown in (24) below, a distribution that is typical of nouns. It might also provide a straightforward explanation for why a locative P like detrás 'de.behind' can be followed by a possessive, as shown in (25)  The analysis that Terzi arrives at is the one shown in simplified form in (26) below. 11 11 An anonymous reviewer draws attention to the fact that when the nominal that is the complement of the P is replaced by a possessive pronoun, the possessive pronoun always surfaces in the masculine singular, independently of the gender and number of the nominal it replaces.
(i As the reviewer points out, these facts, also discussed in Fábregas (2007), support the view that there is a silent PLACE element used as a target for agreement. My analysis of small and big Ps in Spanish will considerably build on Terzi's analysis of Greek and Spanish locative Ps. My proposal will, however, differ from hers in certain important respects. I will assume with Terzi that it is necessary to postulate a silent noun PLACE in the structure of locative Ps. However, although Terzi suggests that silent PLACE is present in the complement of all locative Ps, I will propose that it is present in the complement of big Ps (e.g. debajo 'de.under') but it is absent from the complement of small Ps (e.g. bajo 'under'). For big Ps I will then propose an analysis à la Terzi, where what is traditionally taken to be the complement of the locative (e.g. la mesa 'the of the locative P is just a DP. 14 Interestingly, if we go back to the arguments put forth by Terzi to argue in favor of the presence of a silent noun PLACE in locatives in Spanish, we can see that the arguments presented to account for the 'nominal flavor' of Ps hold good of big Ps, as seen in (24) and (25) above, but not of small Ps, as shown below. (27) is considered marginal by most speakers, and (28) is rejected as ungrammatical.
(27) ?El gato me espiaba desde bajo la mesa. the cat cl acc .1sg was.spying from under the table 'The cat was spying on me from under the table.' (28) *Venía un hombre tras mío. was.coming a man behind mine 'A man was coming behind me.' I will therefore propose that there is a silent noun PLACE in the complement of big Ps but not in the complement of small Ps.
Another important point in which my analysis will differ from Terzi's is that I will not assume that the locative is a modifier of silent PLACE. I will propose rather that it is a P head, the head of PP Loc in (25)/(26), selecting either a Possessor-Possessum structure (in the case of big Ps) or a single DP (in the case of small Ps). The reason for this is that although a locative-as-modifier analysis may be justified for Modern Greek, there is in principle no evidence that points in this direction for Spanish. As adjectives are the only words that can modify a noun directly in Spanish, I assume that under a locative-asmodifier analysis this is what locative Ps would need to be considered. However, small 12 Under Terzi's analysis the [ QP/NP PLACE detrás] constituent then moves to the specifier of DP triggering the appearance of de in D, yielding detrás de la pastelería 'de.behind of the pastry shop' behind the pastry shop. 13 Inside the QP/NP PLACE detrás, the adjective detrás would be a postmodifier of PLACE, just as blanca 'white' is a postmodifier of flor 'flower' in la flor blanca, literally 'the flower white'. 14 It is important to mention that the idea that some locative Ps select a complement with silent PLACE whereas others do not is entertained but later on dismissed by Terzi (Terzi 2010 (29) a. Pusimos los estantes sobre la mesada. 15 put.past.1pl the shelves over the counter 'We put the shelves above the counter.' b. *Pusimos los estantes muy sobre la mesada put.past.1pl the shelves very over the counter 'We put the shelves too high up above the counter.' c. *Pusimos un estante más sobre la mesada que el otro. put.past.1pl a shelf more above the counter than the other 'We put one shelf higher above the counter than the other one.' a. Juan se paró detrás de la columna. Juan se stood de.behind of the column 'Juan stood behind the column.' b. *Juan se paró muy detrás de la columna. Juan se stood very de.behind of the column 'Juan stood a considerable distance behind the column.' c. *Juan se paró más detrás de la columna que Pedro. 16 Juan se stood more de.behind of the column than Pedro. 'Juan stood further behind the column than Pedro.' In short, I have shown that at least in Spanish locative Ps do not exhibit adjectival properties. 17 If it is only adjectives that can modify a noun directly in Spanish, and locatives are not adjectives, then, nothing else said, the null hypothesis is that they are not modifiers of a noun. In other words, they are not modifiers of silent PLACE. Although small and big Ps may diachronically derive from adjectives as seen in, for instance, the case of the locative P bajo (bajo la mesa, literally 'low the table', under the table) and its adjectival counterpart bajo/a (una montaña baja 'a mountain low', a low mountain), synchronically they do not show any trace of adjectival behavior. I will thus simply take them to be Ps. 15 Mesada is a word used in Rioplatense Spanish to refer to the kitchen counter. 16 This sentence has a grammatical reading under which the speaker is not comparing how far behind the column Pedro and Juan are, but is comparing, rather, whose position can be better described as 'being behind the column.' In other words, the interpretation would correspond to something like 'Juan stood more properly behind the column than Pedro.' Under this reading, más 'more' is not comparing distances but, as suggested by Christina Tortora (p.c), seems to be acting as a modifier of (perhaps) a silent PROPERLY adverb. For this reason, I will not take this example to imply that locative Ps in Spanish admit comparatives. Note that even nominals, which clearly admit no comparatives, allow this comparative 'properly' reading, as in, for instance This is more a magazine than a book, where the interpretation seems to be roughly 'This object should more properly be referred to as a magazine than a book.' 17 An anonymous reviewer mentions that, in the context of a marathon, it would be felicitous to utter sentences like the those below.
(i) Juan se paró muy detrás del pelotón de entrada. Juan se stood very de.behind of.the group of entry 'Juan stood very much behind the group at the entrance.'

Other syntactic asymmetries between small and big Ps
In this section I analyze two other syntactic asymmetries exhibited by small and big Ps: namely, the (un)availability of coordination and adverb intervenetion facts. I argue that these otherwise puzzling phenomena can receive a natural explanation if we assume the structures proposed in this work. In section 4 I discuss the interpretative contrasts: in particular, the availability of (non-)locative meanings and the (non-)delimited interpretation of the nominal.

Coordination
One surprising contrast between small and big Ps is that whereas coordination of the complement of a big P is perfectly grammatical, coordination of the complement of a small P is considerably degraded, as illustrated in (33a) and (33b) The unavailability of coordination in (33b) remains, therefore, somewhat of a mystery.

Case assigners and coordinate structures
Interestingly, Demonte (1991) points out a context in which coordination of nominals is disallowed. Her example is given in (35) Demonte (1991) relates the contrast between (35) and (36) to Vergnaud's (1974) and Jaeggli's (1982) observation that Case-assigners cannot be omitted in coordinate structures. The unavailability of coordination in (35) follows, Demonte (1991) claims, if a is taken to be a Case-assigner in these examples.
With this observation in mind, I suggest that the ungrammaticality of (33b) receives the same explanation as the ungrammaticality of (35). In other words, I propose that in the complement of bajo 'under' there is actually more than meets the eye. Specifically, I claim that there is a Case-assigner, and that it is precisely the presence of this element in the structure that is responsible for the ungrammaticality of (33b). I outline the specifics of this analysis in the subsection to follow.

Small and big Ps and Case
As discussed in section 2, Ps like bajo 'under', ante 'front' and tras 'behind' do not seem to exhibit adjectival properties (for instance, they do not admit degree words or comparatives). However, it seems clear that they are diachronichally derived from adjectives. We might argue, then, that although they are unlike adjectives in most regards, they still share with them the impossibility to directly Case-mark their complement. It might be possible then (and most probably likely, if Case is only assigned by functional heads) to postulate that the complement of a small P receives case not from P Loc itself but from a Case-assigner that happens to be null. What this means is that positing a null Caseassigner in these structures may not only capture the facts, as we will see shortly, but might in fact, be theoretically welcome. 18 An anonymous reviewer points out that while a sentence like (35) is ungrammatical if it describes two events, it becomes grammatical if it describes a single event. In other words, (35) is possible if, for example, my sister and Aunt Enriqueta live together and a single visit is involved. However, the focus of the discussion in the text is the two-event-reading, the one that renders the omission of a ungrammatical. I make a brief comment on the single-event reading in the context of Ps in footnote 23.
To be concrete, what I propose is that a small PP like bajo la cama 'under the bed' has the structure schematized in (37), where Ø represents a null Case-assigner.
(37) Structure for bajo la cama 'under the bed' la cama I further propose that this null Case-assigner is affixal in nature and needs to attach to an appropriate host. I take P Loc to be such a host. In the example above, Ø undergoes headto-head movement from the head position of KP and attaches to P Loc bajo 'under'. This is shown in (38).
(38) Attachment of Ø to P Loc in bajo la cama 'under the bed' It is worth pointing out that the analysis of the null Case-assigner proposed here builds on Pesetsky's (1992) analysis of the null complementizer in English. Pesetsky (1992) proposes that the null complementizer is an affix that must undergo attachment to a lexical head. This affixation takes place through head movement of C to V. His account therefore explains, among other things, why the null complementizer is possible in (39)  In (39), null C incorporates into V. However, (40) is ruled out because in this structure null C would be moving out of an island (specifically, a subject). The parallelism between null K and null C will be taken up again at the end of this section. I now go on to show how, with these two assumptions in mind (namely, that there is a Case-assigner in the structure of Ps, and that this Case-assigner needs to attach to a host), the mystery of the ungrammaticality of (33b) suddenly dissolves. Example (33b) is repeated below as (41). As mentioned before, it has been observed (Vergnaud 1974;Jaeggli 1982;Demonte 1991), that Case-assigners cannot be omitted in coordinate structures. What this means is that the structure of bajo la cama y la mesa 'under the bed and the table' has to necessarily be the one sketched in (42) and, crucially, cannot be the one in (43) As the structure in (43) is not possible (because, as was mentioned above, Case-assigners1 9 cannot be omitted in coordinate structures) we are simply going to disregard it. 20 We will therefore just focus on (42). By looking at the structure in (42) we can now get a clearer idea as to why bajo la cama y la mesa 'under the bed and the table' is ungrammatical.
There are, as a matter of fact, several reasons why the structure in (42) fails. Let's review them. In the first place, it is clear that the first null Case-assigner (Ø 1 ) cannot extract without falling into a Left Branch Violation (KP is the specifier of &P). In the second place, although the second Case-assigner (Ø 2 ) is in a position out of which it could in principle extract, nothing else said, it would not be able to incorporate into bajo 'under' without violating locality (there is an intervening head, namely, "&"). As one of these Case-assigners (Ø 1 ) and maybe even both, fail to incorporate, the structure is ruled out by some form of the Stranded Affix Filter (in the sense of Lasnik 1981). Finally, it is worth mentioning that even if both Case-assigners could extract (that is to say, in the unlikely event that we should be able to propose an alternative coordinate structure in which Ø 1 and Ø 2 could extract and incorporate) we would still be faced with a situation in which two affixes of exactly the same kind incorporate into a single host, something that is, at best, suspect. 21 With the ungrammaticality of (41) now explained, let's take a brief look at (33a), the big P counterpart of (33b), repeated here as (44) Example (44) does not of itself tell us anything interesting about the structure of big Ps. Its grammaticality is fully expected, as coordination of elements of the same category (with the exception precisely of the nominals mentioned in this section) is assumed to be possible. However, notice that (45) is ungrammatical, on a par with (41) and (35) The ungrammaticality of (45) is more interesting because it seems to confirm the intuition that de 'of' in the complement of big Ps is also a Case-assigner. In other words, sentence (45) is ungrammatical on analogy with (41) (and Demonte's (35)). What this means is that both the nominal in the complement of small Ps and the nominal in the complement of big Ps are embedded in an outer KP layer. 22 One important difference between these KPs, though, is that the K in the complement of a small P is null, whereas the K in the complement of a big P is overt, a matter to which I return at the end of this section.
Finally, it is important to point out that the asymmetry in coordination facts exhibited by the bajo 'under'/debajo 'de.under' pair is not unique to these Ps but, as expected, extends to the class of small and big Ps as a whole. This is confirmed by the contrast between the (a) and (b)  In sum, the structures we end up with for small and big Ps (with the addition of the KP layer argued for here) are the ones schematized in (48) and (49) (48) and (49). I proposed earlier that both the nominal selected by a small P and the nominal selected by a big P were KPs. I postulated that in small PPs K was null and affixal and I pointed out that in big PPs K was overt (in this latter case, spelled out as de 'of'). However, I did not provide any explanation as to why K had to be null in one case and overt in the other. Interestingly note that in (48) KP is the complement of P Loc but in (49) KP is the specifier of the complement of P Loc . The nullness vs. overtness of K in one case and the other might very well follow from this fact. In (48) K is in a position out of which it can extract and subsequently attach to P Loc by means of head movement. 23 An anonymous reviewer points out that, just as was the case with (35) Visité a [mi hermana] y [la tía Enriqueta], sentences like (46a) and (47a) are grammatical if the situation in question is conceived of as a single event. In the case of (46), if the cathedral and the house of government are next to each other, it might in principle be possible to stand in front of them both. As to (47a), if the drawing and the card are next to each other or on top of each other, it might also be possible to place a bag on top of both objects at the same time. Under these specific scenarios, which involve a single event, (46a) and (47a) become available. It is therefore likely that the (?) judgment for (46a) and (47a) actually stems from the fact that these sentences are ungrammatical under one reading (the two-event reading) and grammatical under another one (the single-event reading). Under the grammatical single-event reading, both DPs are embedded under a single KP layer and extraction of the null K is expected to be possible. However, as the single-event reading may not be easily accessible to speakers, the sentence is perceived as marginal. In contrast, a single-event reading is not available for (41)  However, this is not the case in (49). In this latter structure K is embedded in a specifier and extraction of K would inevitably result in a Left Branch Violation. This might be the reason why overt non-affixal de 'of' must surface instead. It would be possible then to say that the structures for small and big Ps proposed have the side advantage of explaining why (50a) is possible but (50b) is not.
(50) a. El libro está debajo de la mesa. the book is de.under of the This brings us back again to the parallel between null K and null C mentioned earlier in this section. The claim made here is that the ill-formedness of (50b) receives the same explanation as the ill-formedness of (40) (repeated as (51) below): both involve movement out of an island. 24 (51) *He liked linguistics was widely believed.

Adverb intervention
I will now present data that shows that when an adverb intervenes between a small P and its complement the sentence is ungrammatical, but when an adverb intervenes between a big P and its complement the sentence is acceptable to some speakers and unacceptable to others. I argue, once again, that these facts constitute evidence in favor of the structures presented in this work. Consider the following: What we can see from the facts above is that an adverb like probablemente 'probably' cannot surface between bajo 'under' and its complement la cama 'the bed'. However, when probablemente 'probably' occurs between debajo 'de.under' and its complement (introduced by de 'of'), some speakers find it grammatical and some others do not (as shown by the "%" symbol). As expected, the same contrast is exhibited by the other small/big P pairs, as shown below.
(53) a. Juan se escondió tras (*probablemente) la columna. Juan se hid behind probably the column 'Juan probably hid behind the column.' 24 Note that we have an explanation for the grammaticality of (50a) and for the ungrammaticality of (50b).
We have also provided an explanation for the availability of El libro está bajo la mesa 'The book is under the table.' This sentence is grammatical because in this structure the null Case-assigner is in a position out of which it can extract and incorporate into its host. However, nothing said so far accounts for the ungrammaticality of the fourth member of the paradigm: El libro está bajo de la mesa 'The book is under of the What I suggest here is that these facts can be easily accommodated if we assume the structures for small and big Ps proposed so far, and a general restriction on adjunction, such as McCloskey's Adjunction Prohibition (McCloskey 2006).

McCloskey's (2006) Adjunction Prohibition
McCloskey (2006) discusses the adjunction possibilities of a group of adverbs that typically occupy a left peripheral position in TP. This group of adverbs includes, among others, temporal modifiers at the sentential level such as usually, in general, most of the time, yesterday, when he arrived. He observes that many of these adjuncts can appear on the left edge of TP (55a), and many of them can also be adjoined to VP (55b)

Small and big Ps and the Adjunction Prohibition
Let's go back to the data in (52a) and (52b), repeated below as (61a)  According to the structure we proposed for small Ps, in (61a) probablemente 'probably' would be adjoined to KP, as shown in (62) below.

(62)
Adjunction of probablemente 'probably' in bajo probablemente la cama 'under, probably, the bed' Here, probablemente 'probably' is adjoined to a phrase s-selected by a lexical head. Specifically, it is adjoined to the complement of the lexical head bajo 'under'. This is precisely what the Adjunction Prohibition bans, so its ungrammaticality is expected. 25 25 Note that I do not suggest that adjunction to DP is another possibility. In other words, I do not propose the following structure as another option for the attachment of the adverbial in (61a): bajo [ KP Ø [ DP probablemente [ DP la cama]]. If this site of attachment were indeed possible, the sentence would be incorrectly predicted to be grammatical under this second parsing, as the adverbial would be attached to a phrase selected by a functional head (K) (something deemed to be possible by the Adjunction Prohibition) and not to a phrase s-selected by a lexical head. However, adjunction to a constituent smaller than a KP appears to be banned. Note that a string like *la probablemente cama 'the probably bed' (with adjunction of the adverbial to NP) is impossible, independently of its distribution. Also, adjunction of probablemente 'probably' to DP is ungrammatical in a string like *debajo de probablemente la cama 'de.under of probably the bed. ' We can therefore conclude that adjunction of probablemente 'probably' to the DP in bajo [ KP Ø [ DP probablemente [ DP la cama]] is not possible.
In (61b), there should be two possible sites of attachment for probablemente 'probably', as schematized in (63) below.

(63)
Adjunction of probablemente 'probably' to de la cama in debajo probablemente de la cama 'de.under probably of the bed' a. Possibility 1 In (63a) probablemente 'probably' is adjoined to the KP de la cama 'of the bed'. However, this is not the only option. As we can see from (63b), another possibility is that the adverb should be adjoined to the whole XP de la cama PLACE 'of the bed PLACE'. It looks like now we have an explanation for why some speakers find (61b) acceptable and others do not. The adjunction in (63a) gives rise to grammaticality: the Adjunction Prohibition bans adjunction to a phrase which is s-selected by a lexical head, but as is clear from (63a) KP is not s-selected by debajo 'de.under' (it is the specifier of its complement), so adjunction to KP is predicted to be possible. On the other hand, the adjunction in (63b) gives rise to ungrammaticality. In this latter case, probablemente 'probably' is adjoined to XP, and XP is s-selected by debajo 'de.under;' it is its complement. This explains the mixed judgments.

Locative vs. non-locative readings
Another important respect in which small and big Ps differ is that small Ps may give rise to PPs with a non-locative reading, whereas this is hardly ever the case for big Ps.
(64) Estos chicos están *debajo de/bajo mi responsabilidad. these kids are de.under of/under my responsibility 'These kids are under my responsibility.'

(65)
Nicolás decidió seguir *detrás de/tras los pasos de su padre. Nicolás decided follow.inf de.behind of/behind the steps of his father 'Nicolás decided to follow in his father's footsteps.'

(66)
Nos quedamos sin palabras *delante de/ante su generosidad. cl.1pl remained without words de.l.front of/ front his generosity 'His generosity left us speechless.' This contrast can be explained if big PPs have in their structure a silent PLACE element. 26 As small PPs do not have such an element, the locative interpretation is not forced on them. 27

(Non-)delimitedness
There is another interesting respect in which small and big Ps differ. Although speakers typically report no difference in meaning between sentences like (67a) and (67b), they do report a contrast between the (a) and (b)  26 The following appears to be an exception to the claim that big Ps do not give rise to a non-locative interpretation: No hay que olvidarse que hay una familia detrás 'no have.pres forget.se that have.pres a family de.behind' ('One must not forget that he has a family'). The most salient reading of this sentence is purely non-locative. It means that one must not forget that one's negative actions or comments towards somebody will affect that person's family, too. Also worth noting (thanks to an anonymous reviewer) is the fact that the big P detrás 'de.behind' can refer to location in time (in addition to location in place). The reviewer provides the following example from European Colloquial Spanish: Nos dieron a comer una sopa, y detrás de eso, un filete 'cl dat .1pl gave.3pl to eat a soup and de.behind of that a steak.' It could be the case that, as the reviewer suggests, silent PLACE is actually a more general deictic or spatio-temporal element. This does not affect the general proposal made in this work, though. 27 If it is true that small PPs do not contain a silent PLACE element as part of their structure, and if it is also true that silent PLACE is responsible for the locative interpretation of PPs, then we still need to explain where the locative interpretation comes from when small PPs such as bajo la mesa 'under the table', do express location. In other words, if there is no silent PLACE element in bajo la mesa 'under the table', it would seem to follow that the interpretation should be non-locative, contrary to fact. Explaining how we get a locative interpretation in these cases (without silent PLACE) does not seem to be an easy task. However, those who postulate a silent PLACE element for both what I call small and big Ps (as for instance, Terzi 2010) have to face the other side of the coin: how to somehow "suppress" PLACE in the non-locative examples with small Ps in (64)-(66). This task does not seem to me to be easy either. I leave this question open. (68) and (69) are not ungrammatical but are semantically anomalous. The only contexts in which sentences like (68b) or (69b) would be acceptable would be, for instance, in a situation in which there is a fake sun and fake stars and the speaker decides to lie right under them (as in, for instance, the case of glow-in-the-dark stars and a glow-in-the-dark sun in a child's room). The intuition seems to be that for the (b) sentences to be felicitous, the Figure (the speaker's body in (68b)) and the Ground (the sun in (68b)) need to be somehow aligned. In other words, it looks like the big P examples necessarily give rise to a more 'punctual' reading of the Ground. The nominal has to be understood as having clearly defined boundaries; it has to be interpreted as delimited. In contrast, in the small P examples the Ground can receive a non-delimited interpretation. 28 Note that this effect is not exclusive to the bajo 'under'/debajo 'de.under' pair. The same effect is exhibited by other pairs of small and big Ps such as ante 'front'/delante 'de.l.front' and sobre 'on'/encima 'on.top', as illustrated below.

The (b) examples in
(70) a. Juan se paró ante el lindísimo paisaje. Juan se stood front the nice.intens landscape 'Juan stood before the beautiful landscape.' b. #Juan se paró delante del lindísimo paisaje Juan se stood de.l.front of.the nice.intens landscape 'Juan stood before the beautiful landscape.' (71) a. Hay escarcha sobre la pradera. have.pres frost on the prairie 'There's frost on the prairie.' b. #Hay escarcha encima de la pradera. have.pres frost on.top of the prairie 'There's frost on the prairie.' Sentence (70a) can normally describe a situation in which Juan stands before a beautiful landscape. (70b), on the other hand, requires a very specific context. It would be acceptable, for instance, if Juan is standing before a poster or a painting depicting a landscape, but would sound odd otherwise. In a similar fashion, whereas (71a) typically describes a situation in which there is frost on a prairie, (71b) seems to imply that the frost is on a prairie that is not real, such as the one on a drawing or a board game. In other words, what examples (70) and (71) appear to confirm is that in a big PP the nominal is necessarily interpreted as being "delimited" (i.e. as having clearly defined boundaries). Note that if the nominal already describes a surface that is delimited, no contrast arises between the pairs, as expected. This can be seen in the following example with the nominal la mesa 'the table'. As the surface of a table is intrinsically delimited, in other words, it has fixed boundaries (contrary to snow, water and others), there is no "clash" in interpretation between the big P and the nominal.  The question of course is why this contrast should arise. In other words, why is it that big Ps force a delimited reading of the nominal, while small Ps do not? I believe these facts can receive an explanation if we assume the structures for small and big Ps proposed in this work. What I suggest, specifically, is that the interpretative effect that we observe here is a direct consequence of the fact that the nominal is a possessor in the case of big Ps, but not in the case of small Ps. I believe taking a quick look at possessors in dative constructions may throw some light on this matter. In the next subsection I go on a brief excursus and make some independent observations about the behavior of possessors in dative constructions in Spanish. After the excursus, I come back and discuss the examples presented in this section.

An excursus: the possessor in dative constructions
As Cuervo (2003) points out, dative arguments in Spanish can appear with all types of verbs and can have different meanings. One of the meanings expressed by datives, as is standard in many languages, is that of goal or recipient, as exemplified in (73) and (74) below (examples from Cuervo 2003).

(73)
Pablo le mandó un diccionario a Gabi. Pablo cl dat sent a dictionary to Gabi 'Pablo sent Gabi a dictionary.'

(74)
Pablo le puso azúcar al mate. 29 Pablo cl dat put sugar to.the mate 'Pablo put sugar in the mate.' (Lit. 'Pablo put the mate sugar.') In this excursus I do not go into the specifics of Cuervo's analysis of datives. I simply bring to light an observation that she makes when discussing recipient datives like (74) which I believe is relevant to our understanding of the behavior of small and big Ps. 29 Cuervo suggests that the general condition for a dative goal is that it should be able to be interpreted as a recipient or intended possessor. This would explain why, although both (75) and (76) are perfectly grammatical, there is a contrast between (77) and (78) (example (74) is repeated as (77)).

(75)
Pablo puso azúcar en el mate. Pablo put sugar in the mate. 'Pablo put sugar in the mate.' (76) Pablo puso azúcar en la mesa. Pablo put sugar on the Although this observation seems to me to be correct, there is a contrast that still remains unexplained. Consider (81) and (82).
(81) Los pescadores tiraron petróleo en el mar. the fishermen threw oil in the sea 'The fishermen poured oil into the sea.' (82) #Los pescadores le tiraron petróleo al mar. the fishermen cl dat threw oil to.the sea 'The fishermen poured oil into the sea.' (Lit. 'The fishermen poured the sea oil.') In the situation described in (81) above, the oil does become part of the sea when it falls into it, in the same way that the sugar becomes part of the mate when it dissolves in it. It seems to me, then, that there is in principle no reason why example (82), with the clitic le, should not pattern with (77).
What appears to bring about the anomaly is that the sea is non-delimited, it is unbounded, and that for some reason unbounded entities do not make good possessors. That the nondelimited nature of the possessor is what seems to be at issue appears to be confirmed by the grammaticality of (84) below.

(84)
Juan le tiró cloro a la pileta. Juan cl dat threw chlorine to the swimming-pool 'Juan poured chlorine into the swimming pool.' (Lit. 'Juan poured the swimmingpool chlorine.') The situation described by (84) is almost identical to the one described in (82): in both cases an element dissolves in a liquid and becomes part of it. The only difference between (84) and (82), which both contain the clitic le, and the reason for the anomaly of (82), seems to be that, unlike the sea, the swimming-pool is a delimited recipient. Furthermore, the fact that el mar 'the sea' cannot act as a possessor seems to be confirmed by the anomalous tener 'have' sentence in (85).
the sea has oil 'There is oil in the sea.' In contrast, la pileta 'the swimming-pool' is felicitous as the subject of a tener 'have' sentence, as shown below.

Small and big Ps and possession
If the intuition that possessors need to be delimited is on the right track, then this may explain the examples (68)-(71) presented at the beginning of this subsection. I have repeated example (68) as (87)  A possible hypothesis is that the anomaly of (87b) (and (69b), (70b), and (71b), for that matter) might be attributed to some kind of semantic clash between the non-delimitedness expressed by the nominals in these examples (el sol 'the sun' in (87)) and their insertion in a possession structure, which appears to force their delimited reading. If what I have been claiming so far is correct, namely that the nominal in small PPs is a complement, whereas the nominal in big PPs is a specifier and, crucially, a possessor, then this is exactly what we would expect.
An interesting question which arises at this point is why the incompatibility between possessors and non-delimitedness (or unboundedness) should exist. It is likely that the answer to this question lies in the type of possession involved here. I would like to suggest that all of these instances of possession have the following in common: they express part-whole relationships, with the possessor being conceptualized as "the whole" and the possessum as one of its parts, in the sense of Hornstein, Rosen & Uriagereka (2002). Expanding on HR&U (2002), I propose that anything that is conceptualized as a whole must have, by nature, limits or boundaries. It is therefore possible to refer to an apple as the whole apple or to a swimming-pool as the whole swimming-pool; this is because an apple and a swimming-pool are delimited (or bounded) entities. However, it is not possible to refer to water or the sea as *the whole sea or *the whole water precisely because these entities are non-delimited (i.e. they are mass). If I am right, then the clash we find in the sentences with big Ps discussed in this section boils down to an incompatibility between the non-delimitedness (non-wholes) entailed by certain nominals such as el sol 'the sun' (meaning "sunshine") and their insertion in a structure where they are interpreted as wholes, and thus required to be delimited (bounded). 30 30 The reader may have noticed that if it is true that possessors indeed have to be delimited, then the ban on bare nominals as the complement of big Ps (e.g. *debajo de tierra 'de.under of earth'), discussed in section 2.2, may follow directly from this fact. In other words, PPs like *debajo de tierra 'de.under of earth' would be ruled out because the nominal, in this case tierra 'earth,' is a non-delimited (unbounded) possessor, just as el sol 'the sun' in #debajo del sol 'de.under of.the sun' is. However, there seems to be evidence that subsuming both phenomena under a single explanation might not be desirable. One reason for this is that strings like *debajo de tierra 'de.under of earth' are downright ungrammatical, whereas strings like #debajo del sol 'de.under of.the sun' are not ungrammatical but anomalous, and become possible given the right context. Furthermore, as mentioned in footnote 6, as is the case with bare nominals in general, PPs like *debajo de tierra 'de.under of earth' become grammatical when the nominal is modified or coordinated (i.e. is no longer bare), as in debajo de tierra muy húmeda 'de.under of earth very moist' and debajo de tierra y arena 'de.under of earth and sand'. In contrast, the same anomaly remains when sol 'sun' is modified: #debajo del sol de otoño 'de.under of.the sun of fall'. For these reasons, resort to both a syntactic constraint (a ban on bare nominals) and a semantic one (a ban on unbounded possessors) seems to be needed.

Concluding remarks
In this article, I introduced a series of syntactic and interpretative asymmetries between small and big Ps, and showed how each of these asymmetries supported a different analysis for small and big Ps. In this final section, I review what part of our analysis it is that each piece of evidence discussed in the previous sections supports and show how all the pieces of the jigsaw put together confirm the structures proposed.
In section 2, I provided evidence that the nominal in the structure of a small PP is a complement and that the nominal in the structure of a big PP is a specifier. Two pieces of evidence seem to provide support for this claim. The first piece of evidence comes from the distribution of bare nominals. I pointed out that Cuervo proposed that bare nominals were banned from A-specifier positions in the clause in Spanish. I argued that the fact that bare nominals were not possible in the structure of big Ps but were possible in the structure of small Ps was a direct consequence of this fact: the nominal in the structure of a small P was a complement but the nominal in the structure of a big Ps was an A-specifier. The coordination facts presented in section 3.1 provide support for this specifier versus complement status of the nominal. In this section, I proposed that the complement of a small P was a KP. This led to a unified analysis of both the nominal in small and big PPs as KPs, the difference being that in the case of small PPs, K is null, whereas in the case of big PPs, K is overt (and spelled out as de 'of'). With this enriched structure of the nominal in place, I then hypothesized that the nullness versus the overtness of K in one case and the other could be made to follow from the fact that the nominal is a complement in the case of small Ps, but the specifier of its complement in the case of big Ps.
I also provided evidence that shows that the structure of a big P is more complex than the structure of a small P. In section 3.2 I discussed adverb intervention facts. I pointed out that when an adverb such as probablemente 'probably' intervenes between a small P and its complement the sentence is ungrammatical, but when an adverb intervenes between a big P and its complement, it gives rise to mixed judgments. I attributed this contrast to the fact that small Ps have a simpler structure with only one possible site of attachment for the adverbial (resulting in ungrammaticality), whereas big Ps have a more complex structure with two possible sites of attachment for the adverbial (one leading to ungrammaticality, the other one to grammaticality).
In section 4.1 I discussed the presence versus absence of locative readings. I pointed out that big PPs are forced to have a locative interpretation, whereas small PPs can have locative or non-locative readings. I proposed that the obligatory locative reading was a direct consequence of the presence of silent PLACE: as silent PLACE is present in the structure of big PPs, these PPs necessarily have a locative interpretation. As small PPs do not have a silent PLACE element as part of their structure, the locative interpretation is not forced on them.
In section 4.2 I observed that the nominal in the complement of big Ps has to be understood as delimited. In contrast, the nominal in the complement of a small P can be interpreted as delimited, but it can also receive a non-delimited interpretation. I made some independent observations about possessors in dative constructions and drew attention to the fact that in these structures there is also a requirement that the possessor should be interpreted as delimited. I therefore attributed this delimitedness effect to the fact that the nominal in big PPs is a possessor, the possessor of silent PLACE, whereas there is no possessor in the structure of small Ps.
To sum up, the evidence gathered seems to show that: (A) the structure of big PPs is more complex than the structure of small PPs; (B) the nominal in the complement of big Ps is a specifier, whereas the nominal in the complement of small Ps is not; (C) the structure of big Ps, unlike the structure of small Ps, contains a silent PLACE element; (D) the nominal in the structure of big PPs is a possessor, whereas the nominal in the structure of small PPs is not. These are precisely the structures I proposed in this work.
There are of course many questions which remain. One important question to address is why big Ps (e.g. de + bajo) are always morphologically more complex than small Ps (e.g. bajo). In footnote 10 I mentioned that I consider a big P like debajo 'de.under' to be bimorphemic. It would be possible to think, then, that there is a single item bajo 'under' which can select either a single DP complement (e.g. bajo la mesa 'under the table') or a Possessor-Possessum structure (e.g. debajo de la mesa 'de.under of the table'). 31 In the case of debajo de la mesa, de la mesa would be the specifier of a projection whose complement is silent PLACE. The prefix that we find with big Ps (debajo 'de.under') could actually be the spell-out of this possessive head (labeled "X" in example (32)), which has undergone head movement from the head of the Possessor-Possessum structure to the head of PP, leading to the formation of a complex P (debajo 'de.under'). 32 As there is no Possessor-Possessum structure in the case of small Ps, there is no source for prefix de.
Another important question is the following: Why can the complement of big Ps be omitted Ponélo debajo (de la mesa) 'put.imp.cl acc de.under of the table', while the complement of small Ps never can Ponélo bajo *(la mesa) 'put.imp.cl acc under the table'? I believe this is ultimately related to a more general property of the grammar, whereby omission of a complement is possible when a head is supported by a clitic. For example, consider the case of a transitive verb such as ver 'see' in Spanish. This verb does not normally allow omission of its complement: Vi a Maria 'see.past.1sg a Maria'/*Vi 'see.past.1sg'. However, when the head verb appears with an accusative clitic, such as la, as in La vi a Maria 'cl acc .fem.sg see.past.1sg a Maria', omission of the complement becomes possible: La vi 'cl acc .fem.sg see.past.1sg'. In previous work (Fraga 2017), I propose that de in debajo is similar to the clitic la in this regard: it is a clitic that allows omission of the complement (in this case, the complement of the big P). This means that just as omission of the complement of vi 'see.past.1sg' is possible when la is present: La vi (a Maria) 'cl acc .fem.sg see.past.1sg a Maria', omission of the complement of P is possible when the clitic de is present (e.g. de + bajo = debajo 'de.under'): Ponélo debajo (de la mesa) 'put.imp.cl acc de.under of the table'. In contrast, omission of the complement of vi 'see. past.1sg' is not possible when the clitic la is absent *Vi 'see.past.1sg', and omission of the complement of P is not possible when the clitic de is absent (e.g. bajo 'under'): Ponélo bajo *(la mesa) 'put.imp.cl acc under the table'.
There are also two other interesting contrasts between small and big Ps not addressed in this work. One of them has to do with the fact that whereas an animate DP can be the complement of a big P, as in for instance, El sobre está debajo de la señora 'the envelope is de.under of the lady', an animate DP is not possible as the complement of a small P: *El sobre está bajo la señora 'the envelope is under the lady.' The second contrast is the following: whereas big Ps allow a personal pronoun in their complement: debajo de la mesa 'de. under of the table'/debajo de ella 'de.under of her', and debajo de mi mano 'de.under of my hand'/debajo de la mía 'de.under of the mine', 33 this is not the case for small Ps: bajo la mesa 'under the table' but *bajo ella 'under her', bajo mi mano 'under mi hand' but *bajo la mía 'under the mine'. I leave the study of these contrasts for future research.
Finally, I would like to point out that the contrasts exhibited by small and big Ps also seem to be exhibited by other pairs of Ps such as frente 'front'/enfrente 'in.front' and sobre 'on'/encima 'on.top'. These pairs differ morphologically from those that are the main focus