Recent years have seen a variety of projects utilizing theoretically-driven syntactic questions to build dialect atlases (Atlante Sintattico d’Italia (ASIt); Syntactic Atlas of the Dutch Dialects (SAND); Explaining syntactic variation: Basque and Beyond (Bas&Be); Atlas Sintáctico del Español (ASinEs); Scandinavian Dialect Syntax (ScanDiaSyn), etc.), and the results from these projects have already made important contributions to syntactic theory (Benincà 1994; Poletto 2000; van Koppen 2005; Barbiers et al. 2005; 2008; 2016; van Craenenbroeck 2010; Etxepare & Oyharçabal 2013; Fábregas & Gallego 2014; Fernández & Ortiz de Urbina 2016). In this paper, we examine a variety of understudied dative constructions in American English, and show how the variation we find across dialects contributes to a finer-grained understanding of argument-introduction theory (in the sense of Pylkkänen 2002). We show how large-scale data collection, with appropriate mapping techniques, can reveal the existence of syntactic patterns that might not have been visible otherwise.

Beginning with foundational works such as Kratzer (1996) and Pylkkänen (2002), a long line of productive research has investigated the extent to which argument structure alternations can be reduced to the syntactic and semantic properties of special argument-introducing heads like Voice and Appl (see e.g. Cuervo 2003; Schäfer 2008). This line of work has revealed broad, cross-linguistic generalizations that need to be derived, as well as finer-grained distinctions that need to be captured. This is particularly so in the case of applied dative arguments, the analysis of which has enjoyed no broad consensus within the field. The analysis of applied datives is particularly important, because they can be selected or nonselected, and have shared morphosyntactic properties across a variety of thematic interpretations. In the typology of Wood & Marantz (2017), Appl heads are the most structurally complex form of argument-introducing head.

In this paper, we will discuss some constructions from Southern American English that are generally not brought to bear on the theory of applicatives, and argue that low applicative phrases in these dialects have a wider distribution than usually thought, similar to that of DPs. We will suggest that this is because they are DP-like categorially (cf. Johnson 1991: 614ff.). Along the way, we will show how geographical maps of acceptability judgments can help us decide between competing structural analyses.

The paper is organized as follows. Section 2 provides an overview of our methodology for data collection and analysis. Section 3 focuses on Personal Datives, discussing previous literature and arguing in favor of a low applicative analysis of them. We also present our survey results showing them to have a broadly Southern geographic distribution. Section 4 turns to Southern Dative Presentatives, which have an even sharper Southern distribution, and argue that these cases involve a low ApplP as the subject of a small clause. Section 5 introduces Extended Benefactives, which we argue involve a low ApplP as a complement of a preposition. We also show examples of Personal Datives that seem to have this same structure. Finally, in section 6, we present our syntactic analysis that takes ApplP in Southern American English to be a “big DP,” explaining why it appears in the variety of positions it does. This hypothesis also explains why prima facie distinct constructions have nearly identical geographic distributions. Section 7 concludes.

In this section, we present a brief overview of the methodology used in the data collection and analysis. The reader may consult Wood et al. (2015), Wood et al. (2018), Zanuttini et al. (2018), and Tyler & Wood (to appear) and Wood (to appear) for more detailed discussion of the survey methodology and the hot spots and interpolation techniques.

In our work, we have been constructing surveys and administering them on Amazon Mechanical Turk (Sprouse 2011; Gibson et al. 2011; Erlewine & Kotek 2016). Our surveys typically contain 45 sentences, 30 of which are fillers. About half of the fillers are used to “pilot” possibly interesting constructions. The other half are control sentences (designed to be acceptable or unacceptable for everyone), which are used to ensure that we only analyze surveys from participants who seem to understand the task in the way that we want them to understand it. Participants rate sentences on a scale of 1–5, 5 being fully acceptable and 1 being completely unacceptable. We also ask for various kinds of demographic information, including where they currently live and where their primary childhood residence was. We then process the surveys, removing any from participants who failed the controls or reported living in their primary childhood residence for fewer than 8 years. Once we process the survey results, we plot them on maps and subject them to various statistical analyses.

The maps in this paper give three basic kinds of information. First, the locations of the dots show where survey participants grew up; the color of the dots indicate whether they judged a sentence to be relatively acceptable (4–5 on a scale of 1–5), in which case the dots are green, or unacceptable (1–2 on a scale of 1–5), in which case the dots are black. Participants who judged a sentence as a 3 are not shown on the map, but all 3s were included in all quantitative analyses, including the hot spots and interpolation analyses. Second, the red borders around the dots indicate hot spot regions. Essentially, the area around each dot inside a hot spot region has a higher concentration of high judgments than would be expected by chance, in comparison with the full dataset. Third, the shades of blue come from an interpolation function, such that the darker shades appear in areas where the judgments are higher and the lighter shades occur in areas where the judgments are lower.

Personal Datives (PDs) are traditionally found in Southern dialects of American English (Wolfram & Christian 1976: 177–181; Christian 1991), but are widely familiar and may be spreading (see discussion below). They are exemplified by the sentences in (1):

The most striking two identifying properties of PDs are (a) that the dative is obligatorily co-referential with the subject, and (b) that despite this, it has no reflexive morphology.

The presence of PDs has been noted in numerous studies of particular communities in the South, such as Appalachian English and Ozark English. However, Christian (1991: 18) already suggested, “If anything, its use is expanding as part of a general Southern-based variety,” and Gerwin (2014: 206) notes that “the regional and social distribution of this construction has not been scrutinized so far.”1 Horn (2008: 176ff.) presents numerous examples showing that its use has spread into the general popular consciousness of American English speakers. Part of the reason, he proposes, has to do with what he calls the “Braxton effect.” The idea is that pop songs, and specifically the Toni Braxton song “I Love Me Some Him,” may have played a role in spreading the use or awareness of Personal Datives beyond the South.

Our own survey data support both the prevalence of PDs in the South, and the claim that they have spread beyond the South. First, consider the maps in Figure 1. There, we see that for both sentences, there are statistically significant hot spots in the South, which are outlined in red. We see darker blue interpolation in the South (or at least parts of the South) than in other areas, indicating a higher concentration of higher judgments and fewer lower judgments. Both of these considerations strengthen the view that PDs are characteristic of English in the South, to a greater degree than English in other areas of the country. We can also observe that, in general, a sentence with a 1st person subject is rated higher than one with a 3rd person subject, corroborating an observation by Christian (1991: 13), who noted that they “can be found much more frequently with first- and second-person subjects than with third-person.” However, note that PDs are also widely accepted outside the south. The maps show many green dots (indicating judgments of 4 or 5) in other areas of the country. This is particularly the case for the 1st person example (the top map). But even the 3rd person example (bottom map) is accepted by many speakers outside of the South, including especially the North and Northeast, and even the Pacific Northwest. We will see below that this is different from other dative constructions that we will discuss.

The syntactic properties of PDs have been the subject of a number of studies, including Webelhuth & Dannenberg (2006), Conroy (2007), Horn (2008), Haddad (2011), Teomiro García (2013), Bosse (2014), Hutchinson & Armstrong (2014) and Lee (2016). Hutchinson & Armstrong (2014) argue that the dative pronoun is introduced by a low Appl head, which relates a DP complement to its DP specifier. This is similar to the Appl head we find in double object constructions (Pylkkänen 2002), but with a special flavor: Appl_SAT (for satisfactive) introduces a relation of “satisfaction” between the subject and the event denoted by the predicate. Note that the material after the “ : ” is argued to be on a non-truth-conditional tier, where it conveys a conventional implicature. Their analysis is presented in (2) and (3).

As Bosse (2014) notes, the material on the truth-conditional tier (before the colon) is essentially an identity function on the verb it combines with semantically. That means, informally, that it restates the semantic information contained in the verb, and provides no new truth-conditional semantics. This captures Horn’s (2008; 2013) observation that the PD contributes only a conventional implicature, which is a non-truth-conditional, “not-at-issue” meaning.

The low applicative analysis is supported by a number of considerations. First, as pointed out by Hutchinson & Armstrong (2014), it captures the essential property noted by Christian (1991: 17) that PD constructions are structurally like double object constructions. The PD argument occupies the same syntactic “slot” as the first argument of a double object construction and is therefore in complementary distribution with that argument.

The example in (4b) is ungrammatical because his son and him compete for the same position: there cannot be more than one low Appl in the same structure.3

Second, the low applicative analysis accounts for the fact that PD constructions are built on transitive verbs, and cannot occur with unergative or unaccusative verbs.

An exception noted by Horn (2008) involves sentences like I lay me down; Bosse (2014) takes these to be the rule, rather than the exception, writing that her analysis assumes the “broadest distribution of the PD […including] all persons and referents as well as both transitive and unergative verbs” (Bosse 2014: 99) and does not “limit the direct object in PD sentences in any way” (Bosse 2014: 102). She thus argues for a high applicative analysis. However, we agree with Hutchinson & Armstrong (2014) that these exceptions are part of a different trajectory through the history of English, and that the overwhelming generalization in English is that unergative configurations are incompatible with the PD construction.

Third, the low applicative analysis allows for a direct relationship between the object and the PD (which is of course bound by the subject) to be stated. Hutchinson & Armstrong (2014) define a MATTERS-TO relation such that the direct object “matters to” the subject.5 We will see how this view offers a natural account of the “Extended Personal Dative” constructions analyzed below, one which would be challenging for a high applicative analysis.

Beyond this, the datives of PDs have been argued to have the syntactic behavior of SE anaphors (Conroy 2007; Haddad 2011; Bosse 2014; Hutchinson & Armstrong 2014; Lee 2016) (though see Horn 2008; 2013 for a different suggestion, based on Reinhart & Reuland’s 1993 co-argument-based formulation of Condition B). This, too, is supported by a number of considerations. First, cross-linguistically, constructions that resemble the PD construction almost always involve SE anaphors, if the language has them, rather than ordinary pronouns with the same form as those subject to Condition B (Horn 2008: 184–188; Campanini & Schäfer 2011). Second, in many languages with SE anaphors, the form is only distinct from an ordinary pronoun in the third person. In the 1st and 2nd person, SE anaphors look exactly like weak unbound 1st and 2nd person pronouns. Thus, it seems reasonable to imagine that this formal identity could carry over to 3rd person as well. Third, some varieties of West Germanic, including Frisian (Reuland 2011: 268–273) and Old English (van Gelderen 2000; Keenan 2002), use even 3rd person pronouns as SE anaphors. Thus, we know that this hypothetical option is in fact available in closely related languages.

To sum up this section, PDs have a Southern-based geographic distribution, but they are not only found in the South; they are also accepted to varying degrees in other parts of the country. Syntactically, we adopt the analysis in Hutchinson & Armstrong (2014) that they involve a low applicative structure, where the dative introduced in SpecApplP is a SE anaphor (in the morphosyntactic shape of a pronoun). Before we move forward, we would like to emphasize that at this point we are taking the low ApplP to be a constituent containing an Appl head, a DP specifier and a DP complement. In the present case, we see this constituent form the complement of the lexical verb, which is the canonical placement for a low ApplP going back to Pylkkänen (2002; 2008). As we proceed, we will propose that this low ApplP can also appear in much less familiar positions.

We turn now to a construction that has been much less studied, which we call the Southern Dative Presentative (SDP), following Wood et al. (2015). A presentative is a sentence whose function is that of pointing the addressee’s attention to the presence of a certain entity (or set of entities) in the context of utterance:6

Cross-linguistically, presentatives tend to consist of a deictic element along with a noun phrase denoting the entity in question. The nature of the deictic element may vary. It can be a locative, such as English here, a form that resembles demonstratives, as in Serbian evo, or a dedicated particle that is not used outside of presentative constructions, such as Italian ecco. This element can also be derived, at least historically, from a verbal source, as in the the French voilà (which comes from voi ‘see’ and là ‘there’).

In some languages/varieties, a dative pronoun may be present in a presentative construction. Some examples taken from Wood et al. (2015: 295–296) are given in (7):

We should emphasize here that we do not necessarily assume all of these languages to have the same structure and function for the dative. In fact, there are reasons to think that in some of these and other languages the dative functions in a distinct manner. The point here is that the applicative system (assuming that it is responsible for the introduction of dative arguments), however it works in a given language, can in principle be incorporated into presentative syntax.

It turns out that certain varieties of English also allow a dative pronoun in their presentative constructions, and this is what we call the SDP. The earliest mention of them that we know of comes from Dudley (1946) (in the context of his observations about dialect forms, mostly lexical items, encountered in Kentucky).

Such cases have occasionally been mentioned in passing (Montgomery & Hall 2004; Liberman 2009; Horn 2014), though Wood et al. (2015) is the first detailed study of them as far as we know.7

We have found a generally Southern distribution among people who find them acceptable. The map in Figure 2 shows the distribution of judgments of the sentence Here’s you a piece of pizza. As we can see there, the sentence is widely accepted in the South, but overwhelmingly rejected in the North and the West. Notice that the regional pattern here is much sharper than what we saw earlier for Personal Datives. We return to this point below.

We find quite a bit of syntactic flexibility in the SDP. This is illustrated by the sentences in (9), which are drawn from the studies in Wood et al. (2015) and Wood et al. (2018). Here can be replaced by where (9a) or there (9b, g). The copula can be at least ’s (9a, b, d, f, g, i, j) or are (9c), and for some speakers is (9e) or even (noncopula) comes (9h). The dative itself can be 1st person (9a, f, g), 2nd person (9b–e, h), or 3rd person (9i, j).

As discussed in detail in Wood et al. (2018), some of these sentences are more marked than others, meaning that they are more geographically restricted and accepted by fewer speakers in the regions where they are found, but all are well attested in the South, in roughly the region indicated in the map in Figure 2. See Wood et al. (2018) for maps of the sentences in (9b–j) and Wood et al. (2015) for a map of the sentence in (9a) (along with other examples of where and here sentences).

We now turn to the question of the structure of presentatives, and how the dative fits into that structure. There has in fact been very little syntactic work on the structure of presentatives. Focusing on Italian, Zanuttini (2017) proposes that they are derived from an underlying small clause structure consisting of a locative predicate and its subject.8

Both of these constituents undergo leftward movement, such that the locative “qui PLACE” is higher than its subject.

What is important in the present case is the base structure: the small clause predication. We need a place in this structure to merge the dative. There are two possible options. First, the dative could be introduced with the subject, as a low Applicative, such that the low ApplP would serve as the specifier of the small clause. Second, the Appl could take the entire SC as a complement, in what is essentially a high applicative structure. These two options are illustrated in (11).

Note that we consider the analysis on the right, the high Appl analysis, to cover any analysis where the complement of Appl is some kind of predicate, i.e., anything other than a DP.9 We make no claims here about the internal structure of the small clause, such as whether there is a functional head introducing the embedded subject as in Den Dikken (2006) and many others. The point is that in (11a) Appl takes a syntactic DP (semantically, an entity) and in (11b) Appl takes some kind of small clause (semantically, a state/eventuality).10

The high Appl analysis has at least two considerations working against it. First, there seems to be a direct semantic relationship between the subject and the dative. This follows from the low ApplP structure in (11a), but is not expressed in the high Appl structure in (11b). Second, we will see other structures, which cannot be a high Appl, that are better characterized in terms of the distribution of a low ApplP. These constructions co-vary geographically with SDPs. Given their existence, (11a) falls into place, but (11b) is an anomaly.

In fact, we hasten to add that the proposed structure in (11a) bears a close resemblance to a structure proposed by Pylkkänen (2008) (building on Cuervo 2003) for Spanish. The structure proposed by Pylkkänen (2008: 45) for the complement of consideró ‘considered’ in (12) is shown in (13).

If this is correct, then the question is not if ApplP can form the subject of a small clause—syntactically, it is possible—the question is when this is possible, and what constrains it. Pylkkänen (2008) suggests one factor: the Spanish Appl head in question has a stative semantics that makes it compatible with the small clause structure, whereas the more familiar low Appl from double object constructions has a dynamic semantics that makes it incompatible with the small clause structure. We return to the question of semantics below.

Before proceeding, however, we would like to address one possible concern raised by a reviewer, which we will not be able to address in this paper. Cuervo (2003: 120) argues against having an ApplP in a specifier position, partly on theory-internal grounds, and partly on the grounds that ApplPs do not seem to occur as external arguments, i.e., specifiers of VoiceP. Assuming this is true, this asymmetry needs to be explained. We would like to add, though, that it is not obvious whether this is a semantic/thematic restriction or a structural one. A key question is whether there is a difference between an embedded external argument (such as in a causative) and a matrix external argument. To get a sense for the difference, consider the sentences in (i).

(14a) is attested in Kentucky English, where the string his wife a bracelet could conceivably be an ApplP in the the embedded (derived) subject position. If so, the question is whether this would be possible when the final landing site is a matrix subject position, as in (14b) or (14c). These two examples are almost certainly ungrammatical independently, since causative have does not passivize. But the potential contrast should be clear, and should be investigated in future research. We note in addition that the same issue arises for “Big DP” analyses of clitic doubling (see Uriagereka 1995; Anagnostopoulou 2006; Arregi & Nevins 2012), which shows that the fact that a “Big DP” can be constructed does not entail that it can be licensed in just any thematic position.

If our proposal is on the right track, then what presentative datives show us relates to the external distribution of ApplP. In particular, we see that ApplP can be the subject of a small clause. We will see further evidence for this option, based on other constructions, in the next section.

Another kind of sentence, also originally reported by Dudley (1946: 271), is presented in (15). We will refer to this kind of construction as an “Extended Benefactive,” since the dative in these cases can be paraphrased with benefactive for-PPs. (15d) is reported to be said “of a spider spinning a web”:

We tested the sentences in (16). The sentence in (16a) is adapted from the sentences Dudley (1946) found, and the attested sentence in (16b) was originally heard by Matt Tyler (p.c.) in a Mississippi diner.

Notice that unlike Personal Datives, and like SDPs, the dative here is not coreferential with anything in the sentence. Moreover, in this case, the ApplP seems to be contained within a PP (as a complement of P).

The map in Figure 3 shows the distribution of the sentences in (16). Notice that the geographic distribution of these sentences is extremely similar to that of SDPs seen earlier. Of course, just because two constructions have a Southern geographic distribution does not mean that they are necessarily syntactically related in some deep way. Nevertheless, we would like to emphasize how really very similar they are.

We can show this by comparing them to other constructions, which have a Southern distribution without necessarily having exactly this shape. One example was given earlier, where the maps of Personal Datives were Southern, but looked quite different in character from the SDP and Extended Benefactive maps. Another example comes from for-to infinitives (Henry 1992). Consider the map in Figure 4, which shows the distribution of He turned off his phone for to avoid his girlfriend. This also has a Southern distribution, which is consistent with what is reported in the literature on Southern dialects (Elgin & Haden 1991; Montgomery & Hall 2004). But it is clearly quite different in character from both the SDP/Extended Benefactive pattern, on the one hand, and Personal Datives, on the other. This illustrates just some of the wide variety of geographical patterns we find, even among constructions that are primarily found in the South.

The similarity between the SDP pattern and the Extended Benefactive pattern is, then, not to be taken lightly. All else being equal, it would be a good thing if this followed from the structures available; the similar geographic distribution makes sense if they are structurally similar in a syntactically meaningful way.

Our proposal is that they are related. ApplP, in this dialect, distributes like a DP: unlike the canonical low ApplP, which is usually analyzed as only occurring as a complement to a verb or verb root, ApplP here can be the complement of a preposition or the subject of a small clause:

Returning now to Personal Datives, we find examples where they, too, appear to be in the complement of a preposition. One example was attribed to comedian Sheryl Underwood by the Chicago Tribune (9/11/16).12 Given that the dative is coreferent with the subject, this appears to be a PP-internal Personal Dative:

Sroda & Mishoe (1995) provide a number of similar examples, with selected prepositions, which are presented in (19):

The existence of such sentences supports the view that ApplP has a wider distribution in Southern dialects than previously thought. For Personal Datives, too, ApplP can be the complement of a preposition.14

We can also find the ApplP of Personal Datives in the subject of a small clause. The following attested examples are due to Michael Montgomery (p.c.), which he gathered in preparation for a new edition of the Dictionary of Smoky Mountain and Southern Appalachian English:

If we assume a kind of small clause structure of ditransitive PP and particle verb structures, as in (22a–b), then the ApplP could be the subject of the small clause, as in (22c–d).

An interesting consequence of the analysis involves the semantics. Recall the denotation of the satisfactive Appl head from Hutchinson & Armstrong (2014), given earlier in (3), repeated here as (23):

When ApplP is the subject of a small clause, the matters-to predicate still applies the theme, a block or a red ear, to the dative. So we still get the implicature, corresponding to speaker intuitions, that there is a direct relationship between the dative and the theme. (The referent of the theme matters to the referent of the dative.) However, the satisfied-by predicate applies not to the verb, but to the result state denoted by the PP. In I’d saw off me a block, the block matters to me, and I am satisfied not by the sawing event per se, but by the resultant state—that the block is off. Similarly, in I’d stick me a red ear in my pocket, the red ear of corn matters to me, and I am satisfied not by the ‘sticking’ event per se, but by the resultant state—that the ear (of corn) is in my pocket.

A high Appl analysis of the sort advocated in Haddad (2011), Teomiro García (2013), and Bosse (2014) would struggle to capture the relationship between the dative and the resultant state, and simultaneously between the dative and the theme. The Appl head would simply be too high—well above the causing subevent that the verb denotes—to have semantic access to the resultant state. The resultant state semantics would be already computed and buried under the verb by the time the dative came into the interpretation. The fact that the semantics works out supports the view that PDs are small clause internal low ApplPs, not high ApplPs.

Researchers have held a variety of positions regarding what kind of functional category Appl is, where Appl is broadly construed as the category that mediates the relationship between the two DPs of a double object construction. Some take it to be a kind of preposition (Harley 1995; Pesetsky 1995; Legate 2002), while others take it to be more like a DP (Johnson 1991; Wood & Marantz 2017; see also Kayne 1993 for related ideas). In fact, in Harley (2012) and Harley & Jung (2015: 707–708), the discussion of Harley’s P_HAVE brings it in line with that of Kayne (1993), where the construction containing the possessor and possessee is a big DP. Parallelling these conclusions, Wood (2015: 214) notes that “the possessive relation introduced by low Appl is on par with the relations introduced by DP-internal possessive morphology in English.” Our proposal will build on Wood & Marantz (2017), since the latter is an explicit attempt to reduce argument-introducing heads like Appl to the syntactic features they consist of (though see also Myler 2014; 2016).

The leading idea in Wood & Marantz (2017) is that Appl is not a primitive functional category at all, and that argument introducing heads generally do not have their own category. Rather, all argument-introducing heads—Voice, little p, P, Appl—are derived from one functional category, which they label as i*. Abstracting away from some of the details, the way this works is that this head, in many cases, inherits the grammatical category of its complement, and then selects the kind of specifier it wants to combine with. If i* combines with a vP, it projects a v*P and takes a DP specifier. If it combines with a PP, it projects a P*P and takes a DP specifier. If it combines with a DP, it projects a D*P and takes a DP specifier. This last case is what defines a canonical low applicative, structurally.

In effect, what this means is that a canonical low ApplP is a subtype of DP, one that contains two DPs. The “*” notation indicates that it has some feature that distinguishes it from ordinary DPs, so that it can be selected for; it need not distribute exactly like a DP. But the fact that it has the category D means that it can distribute like a DP. That is, it shares features with ordinary DPs. So in the basic case, all else being equal, we might expect that ApplPs—that is, D*Ps—may occur in positions that DPs occur in. The structures given above, with ApplP as the subject of a small clause, exemplify this possibility.

For now, let us propose that the one thing characterizing the range of dative constructions we find in Southern American English is that ApplPs distribute a lot more like DPs than in standard English. This would lead us to ask why standard English is so restrictive. One possibility would be to posit the following difference between standard and Southern American English ApplPs:

The * notation for standard English is removed for Southern American English, indicating that whatever feature of standard English distinguishes ApplP from DP, it is not there in Southern American English.15 Assuming that the ability of a verb to occur in the double object construction involves the ability to select ApplP—D*P here—this would also predict that the Southern type of ApplP would not exhibit the verb class restrictions that Standard English ApplPs are subject to, so that verbs which do not allow the double object construction in Standard English would allow PDs and related constructions in Southern American English. Since we do not have data that bear on the question, we must leave it for another occasion.

Changing gears, another aspect of these constructions has to do with the properties of the specifier. As discussed above, in the Personal Dative, the specifier is a SE reflexive (which in English—as well as in the 1st and 2nd person in Germanic and Romance—looks just like an ordinary pronoun). It must be coreferent with the external argument subject. Here we follow the intuition that SE reflexives are structurally smaller than DPs: they are φPs. This, then, is the second dimension of variation in the featural content of an Appl head: it may take a DP specifier, or a φP specifier. We propose that in general, non-reflexive pronouns are in fact DPs, and not φPs, so selecting for a φP effectively means selecting for a SE reflexive.16 In principle, this property is independent of the D/D* distinction discussed above; therefore, we might expect to find personal dative ApplPs with the restricted distribution of D*Ps—that is, they must be the complement of V—and also personal dative ApplPs with the broader distribution of DPs.

In fact, this is what we have seen. The broad geographic distribution of the Personal Dative indicates that speakers are able to acquire the basic case without taking the SDP and Extended Benefactive constructions along with it. This follows from the proposed typology of low ApplPs. All it takes to go from the standard low Appl to the Personal Dative is an Appl head that selects for a φP specifier; the other features—in particular the * feature that distinguishes it from an ordinary DP—may stay the same. Such speakers have essentially just made a minor adjustment to the double object construction, encoded in the featural makeup of one head.

However, we also find constructions where ApplP seems to have a wider syntactic distribution, one that is much more similar to ordinary DPs. In these cases, which are more restricted geographically, we have the structure in (24a), with a φP in the specifier instead of a DP. This forces the specifier to be a SE reflexive, and it is this kind of head that gets the interpretation associated with Personal Datives (e.g. the satisfactive meaning discussed above).17

Thus, what we find in the South is a more generalized, productive system of dative constructions. This system allows two kinds of ApplP specifiers (φP or DP) and an ApplP that has a wider distribution, due to its being categorially just like a DP. When the Personal Dative construction spread outside of the South, speakers encountering it did not acquire this full system. Rather, they made a minor modification to their existing ApplP systems to allow for the restricted sort of Personal Dative—it must be the complement of a verb, but like the Southern ApplP, can take a φP or DP specifier.

In this paper, we have discussed a wide range of dative constructions that vary across dialects of American English. We have seen how geographic maps of acceptability judgments can help distinguish between hypotheses regarding the underlying structures of those constructions. In particular, we have seen a strikingly similar geographic pattern for two constructions—SDPs and Extended Benefactives—which, superficially, appear to be quite different from each other. However, we have observed that one of the competing analyses of SDPs, the one which took ApplP to be the subject of a small clause, offers a ready explanation for the existence of Extended Benefactives, where ApplP seems to be the complement of a preposition. The explanation is based on the idea that an ApplP is categorially just like a DP, so its distribution is like that of a DP. The maps have also shown that the Personal Dative has a wider geographic distribution than the SDP and Extended Benefactive constructions. This supports treating Personal Datives along a distinct syntactic dimension from the other two constructions. We have proposed that it has to do with the kind of specifier ApplP can take, and that this factor is distinct from the external distribution it may have. The overall results of this study show how syntactic theory can inform and be informed by the study of dialect variation, and how geographical variation can serve as a useful source of information on the relationship between distinct constructions.