1. Preliminaries

1HAVE GOT + DP_possessee as illustrated in (1) is one of the predicative possession markers in English. It has been argued that the meaning of ‘stative possession’ in HAVE GOT developed during the conventionalization and semanticization of the meaning component ‘possession’ in the present perfect expression HAVE got(ten) ‘have received’ as exemplified in (2) (Schulz 2012a: 104-132, Schulz 2012b).

2The spelling HAVE got(ten) is adopted here to represent present perfect have gotten, has gotten, have got, has got, ’ve gotten, ’s gotten, ’ve got and ’s got. While British English uses HAVE GOT to refer to both present perfect ‘have received’ and to stative possession, American English uses HAVE gotten for present perfect ‘have received’, with HAVE got reserved for stative possession (Quirk et al. 1985: 113; Biber et al. 1999: 467). Historically, both gotten and got were available in British English during the Early Middle English period and were used interchangeably in present perfect contexts (Crowell 1959: 285). Gotten, however, dropped out of use and got was subsequently used for both the present perfect and the possessive expression. In the following, the spelling HAVE got(ten) will be used to reflect both the American and the British spelling.

3The present paper discusses differences between HAVE got(ten) and HAVE GOT brought about by the pragmatic process of the conventionalization of ‘stative possession’ and addresses them within the frameworks of Minimalism and Distributed Morphology (Chomsky 1995; Halle and Marantz 1993).

4These differences between HAVE got(ten) and HAVE GOT are as follows: Temporal reference changes from pre-present to present. GET loses its eventive component v_become and starts to signal stative possession in the invariable form got, while the thematic role of the subject changes from recipient to possessor.

5Got(ten) loses its status as a past participle and a marker of ‘consequent state’ in the sense of Giorgi and Pianesi (1997), evidenced by the loss of -en in those varieties that mark the past participle overtly. Intriguingly, HAVE keeps its status as an auxiliary with respect to negation and question formation but is defective in its combinatorial properties. It does not occur in non-finite forms or in combinations with other auxiliaries and is commonly perceived to be excluded from past possession contexts (McIntyre 2010: 22).

6Sections 2 and 3 of the present paper briefly lay out the conven­tionalization and semanticization processes and discuss evidence for the status of HAVE GOT as a stative rather than a perfect construction. Sections 4.1 and 4.2 are concerned with decompositional and relational accounts of possessive and auxiliary uses of HAVE and the structure of the present perfect in English, drawing mainly on Harley (1998, 2002, 2004) and Giorgi and Pianesi (1997).

2. The conventionalization of conversational implicatures

8The conventionalization of conversational implicatures was first proposed by Grice (1975: 58), who argues that it is possible “for what starts life […] as a conversational implicature to become conventional”. The concept was then picked up by Traugott and König (1991) and promoted as one of the central pragmatic forces in the process of grammaticalization.

9With reference to the Informativeness Principle (Atlas and Levinson 1981), Traugott and König (1991) argue that conversational implicatures arising from information strengthening processes become conventionalized and develop into new meanings: “The approach taken here is that distinct new polysemies of a form are new conventional meanings” (Traugott and König 1991: 193).

10In a nutshell, conversational implicatures or invited inferences arise in so-called conducive environments and add an inferred meaning to the coded meaning of a particular expression in that environment. Over time the inferred meaning is conventionalized or semanticized and develops into a second coded meaning of the expression in question.

11Examples for the conventionalization of a conversational implicature include inferred causation, where the initially purely temporal marker since acquires causal meaning in contexts of temporal overlap between two events, or inferred concessivity, where markers of concomitance like English while or German zugleich ‘at the same time’ acquire concessive meaning (Traugott and König 1991).

12The conventionalization of a formerly merely implied meaning can have an impact on the syntactic make-up of an expression. A classical example is the development of the future marker BE going to. For BE going to in a sentence like (3a), for example, two structural options can be postulated.

13The original structure, a sequence of a full lexical verb followed by a purpose clause introduced by to as illustrated in (3b), is reanalyzed as an auxiliary verb in combination with a bare infinitive as illustrated in (3c), both modeled on the stages of development presented in Hopper and Traugott (2003: 69). Later during the development of the expression, instances can be found where only the reanalysed version of the underlying structure is viable, as illustrated in (4b)-(4c).

3. From onset of possession to ‘stative possession’

14It has repeatedly been argued that the predicative possession marker HAVE GOT started its linguistic life as an inference of ‘stative possession’ in the context of the present perfect expression HAVE got(ten) ‘have received’/‘onset of possession’ (Johnson 1773; Jespersen 1931; Visser 1973; Plank 1984; Gronemeyer 1999). The present paper argues that this inference can be identified as a conversational implicature in the sense of Levinson (2000) and became conventionalized into the coded meaning of ‘stative possession’.

Do not make your contribution more informative than is required. The underlying idea is of course, that one need not say what can be taken for granted… Brief and simple expressions thus encourage, by this heuristic, a tendency to select the best interpretation to the most stereotypical, most explanatory exemplification (Levinson 2000: 37).

16The inference from ‘onset of possession’ to ‘stative possession’ is an implicature motivated by the Informativeness Principle, as ‘stative possession’ can be shown to meet the criteria for an implicature of that kind set out in Levinson (2000). It is cancellable, nondetachable, reinforceable and calculable. Example (5) demonstrates that ‘stative possession’ is cancellable.

17‘Stative possession’ is also non-detachable in the sense that “an expression with the same coded content will tend to carry the same implicature” (Levinson 2000: 15), as illustrated in (6). Have received and have been given, like have got(ten), combine the coded content ‘onset of possession’ with the present perfect. They all yield the implicature ‘stative possession’. (7) illustrates that the implicature ‘stative possession’ is also reinforcable.

18The calculability of a conversational implicature is defined as “the more or less transparent derivation of the inference from premises that include the assumption of rational conversational activity” (Levinson 2000: 15). Calculability is given in the case of ‘stative possession’ with respect to the use of the present perfect, which commits the speaker to the current relevance of the utterance.

19In light of the current relevance of the utterance expressed by the present perfect, an implicature from ‘onset of possession’ to ‘stative possession’ can be argued to be “the most stereotypical, most explanatory exemplification” (Levinson 2000: 37). Put differently, if the onset of possession has current relevance at the time of utterance, it is more natural to assume that the state of possession has not ceased to hold at the time of the utterance than to assume that it no longer holds.

20The implicature ‘stative possession’ in its non-conventionalized form is frequently found in instances of present perfect HAVE got(ten) in the 17th century. Examples (8) and (9) taken from the LION database illustrate a situation typical for an invited inferencing process, as they exhibit different degrees of prominence of the coded and the inferred meaning:

21Example (8) allows both an ‘onset of possession’ and a ‘stative possession’ reading. Cassius has just learned of Orest’s fate and his line might be read as a comment on this particular recent development. On that reading ‘stative possession’ would have the status of a conversational implicature only. On the other hand, Cassius’s line might also be read as a comment on the present state of affairs, where he is not alone in his madness. On this reading ‘stative possession’ could be argued to be fully semanticized.

22A stronger case for semanticization can be made in (9). While shee hath got a foole admits both onset and stative readings, the inference of stative possession is very strong and reinforced by the context, where vnhappy state suggests a preoccupation with a present state rather than with how it came about.

23Full semanticization is marked by the attestation of contexts which no longer admit the formerly coded meaning but only the formerly implied meaning. For ‘stative possession’ this is the context of inalienable possession, which is no longer compatible with the formerly coded meaning ‘onset of possession’. HAVE GOT starts to be attested in contexts of inalienable possession during the second half of the 17th century, as illustrated in (10) from the 1659 The London Chaunticleres.

24Example (10) is a clear case of inalienable possession as there is no sense of receiving one’s individual body parts from somebody or something. Thus, the implicature ‘stative possession’ is fully semanticized here. Contexts of inalienable possession strongly support the case of HAVE GOT as a stative rather than a perfect marker. While present perfect HAVE got(ten) still exists with its meaning ‘onset of possession’ and its conversational implicature ‘stative possession’, the conventionalization of the implicature has led to the establishment of a new meaning, namely stative possession in the form of possessive HAVE GOT.

4. Decompositional approaches to HAVE and GET

4.1 Possessive HAVE

25Decompositional accounts of HAVE go back to work by Freeze (1992) and Kayne (1993, 2000), where HAVE is the spellout of an empty preposition D/P_e incorporating into a verbal head that contains copula BE as illustrated in (11).

26The concept of decompositional HAVE is extended and refined in Harley (1998, 2002, 2004) within the framework of Distributed Morphology. The abstract preposition D/P_e is recast as P_have, a regular vP is substituted for Kayne’s BeP.

27The meaning of P_have is argued to be completely relational, “establishing a connection between two arguments but otherwise not contributing anything to their interpretation” (Harley 1998: 197).

28The possessor-possessee relation is established via the structural configuration of Spec P_have and the complement of P_have. If the complement of P_have contains an overt element co-referent with Spec P_have, a locative reading is established. If the complement of P_have does not contain an overt element co-referent with Spec P_have, a possessor-possessee relation between the two is established, as illustrated in (12) and (13) (cf. Ritter and Rosen 1997 for a similar account).

29Thus, P_have selects for two DPs which are assigned the roles of possessor and possessee by virtue of their structural configuration as specifier and non-co-referent complement of P_have. The possessor then raises to Spec TP for EPP reasons (Harley 2004). The complete structure adapted from Harley (2004: 262) is provided in (14) above, incorporating a hybrid Agr/T1 head in the style of Giorgi and Pianesi (1997) which will be discussed in more detail in section 4.2.

30The relational account of the meaning contribution of P_have sketched out here will be used for the subsequent discussion of present perfect HAVE got(ten) and possessive HAVE GOT as it lends itself very well to a unified account of both auxiliary and non-auxiliary uses of HAVE, which is envisaged but not completely worked out in Kayne (1993, 2000).

4.2. Auxiliary HAVE

31For the auxiliary use of HAVE, Kayne’s D/P_e does not take a possessee DP complement but a verbal complement, resulting in a “particular kind of nominalization, that is, a verbal (participial) structure embedded in a DP that is akin to a CP” (Kayne 2000: 111). The participial status of the verb and the assignment of present perfect meaning is handled within the framework of Giorgi and Pianesi’s (1997) minimalist instantiation of the Split-Infl hypothesis, where “separate AGR_S and T categories are projected only when there is positive evidence in the morphosyntax” (Giorgi and Pianesi 1997: 69-70).

32Giorgi and Pianesi (1997) build on Reichenbach’s differentiation between speech time, event time and reference time and postulate two T-projections, T1 and T2, whose different values describe the relationship between reference time and speech time and reference time and event time. T-projections assign T-roles to event positions in the thematic grid of verbs. Every T has to have a verbal complement (Giorgi and Pianesi 1997: 27-30).

33For the English present, past and future tenses a hybrid Agr/T1 head is postulated, which checks person and number on the subject and assigns nominative case, assigns a T-role to the verb and expresses the relation of the eventive variable of the verb to the speech event, as illustrated for present tense possessive HAVE in (14) above (Giorgi and Pianesi 1997: 70-71).

34For the English present perfect two hybrid tense heads are postulated, Agr/T1 and T2/Asp. Agr/T1 checks nominative case, assigns a T-role to auxiliary HAVE in V_aux and ensures that the eventive variable of the auxiliary is coextensive with the speech event. T2/Asp assigns the aspectual value of ‘consequent state’ to the event denoted by vP and enters into a relationship with the eventive variable of V_aux (Giorgi and Pianesi 1997: 27-30, 38, 99-101). The notion of consequent state is conceptualized as a derivative rather than a primitive eventuality.

35A consequent state is defined as a set of eventualities whose left temporal boundary co-incides with the right temporal boundary of a culminated event. Participial morphology thus has a two-fold function. It turns the event denoted by the vP into a culminated, topologically closed event and introduces the notion of ‘consequent state’ (Giorgi and Pianesi 1997: 97-99).

36The relationship of T2/Asp with the eventive variable of V_aux then ensures that any event which follows the right temporal boundary of the V_aux variable will also follow the right temporal boundary of the consequent state variable (Giorgi and Pianesi 1997: 111-114). Within a framework that views the meaning contribution of P_have as relational only, it can now be argued that the structural configuration of P_have and the participle in T2/Asp assigns auxiliary verb status to HAVE. The full structure is provided in (15) above.

4.3. Decomposing GET

37The most influential decompositional accounts of GET use the relational account of the meaning of possession sketched out in section 4.1 and postulate an additional inchoative marker (Harley 2004; McIntyre 2005, 2010). The structural configuration of two DPs as specifier and non-co-referent complement of P_have posited for possessive HAVE is also part of GET and accounts for the ‘possession’ component.

38The inchoative or ‘onset’ component in GET is identified as an additional v_become in Harley (2004) and McIntyre (2005) and as v_cause[–AG] , a causer-less causative marker, in McIntyre (2010). As both analyses fit my purpose equally well, the former will be adopted for simplicity’s sake. The whole structure is illustrated in (16), adapted from Harley (2004: 9).

39The combination of this structure with the one outlined for the present perfect in section 4.2 yields (17) for present perfect HAVE got(ten) ‘have received’. Note that there are two incorporation sites now. The meaning complex ‘onset of possession’ is contributed by v_become in combination with the structural configuration of two DPs as specifier and non-coreferent complement of the lower P_have.

40The lower incorporation site has P_have incorporate into v_become and spell out as get, which subsequently moves to T2/Asp where it is assigned the status of a past participle and marks ‘consequent state’. In American English the participle ending spells out as -en, in British English it spells out as Ø (cf. section 1).

41The upper incorporation site has P_have incorporate into v_be and combine with T2/Asp, resulting in auxiliary HAVE in V_aux. Present perfect meaning is established via the relations between the two T-projections.

5. From present perfect HAVE got(ten) to possessive HAVE GOT

42As outlined in section 1, differences between present perfect HAVE got(ten) and possessive HAVE GOT can be found with respect to the temporal reference of the expressions, the thematic role of the subject, the meaning contribution of GET, the status of got(ten) as a participle and the auxiliary properties of HAVE, as summarized in table 1.

43The main aim of the present section is to frame these differences as direct consequences of the conventionalization of the conversational implicature ‘stative possession’ as outlined in sections 2 and 3, following the rationale that the conventionalization of a formerly only implied meaning can indeed have consequences for the structural make-up of an expression.

44Most of the changes will be argued to be ultimately due to the difference in temporal reference between the original coded meaning ‘onset of possession’ and the conventionalized or semanticized meaning ‘stative possession’.

Table 1. Differences between present perfect HAVE got(ten) and possessive HAVE GOT

	HAVE got(ten)	HAVE GOT
temporal reference	pre-present	present
thematic role of subject	recipient	possessor
GET	‘onset of possession’	‘possession’
status of got(ten)	past participle	verb
status of HAVE	auxiliary	defective auxiliary

45According to Declerck (2006), the present perfect locates an event in the so-called pre-present time zone, a time zone which precedes the utterance time t₀. The present tense, on the other hand, locates an event not prior to but directly at t₀ (Declerck 2006: 148-149).

46The coded and the conversationally implied meanings of HAVE got(ten) locate the event in different “time zones”. The coded meaning ‘have received’ locates the event in the pre-present time zone. The conversationally implied meaning ‘stative possession’, on the other hand, locates the event in the present time zone. During the conventionalization process the location of the event eventually switches from the pre-present to the present time zone.

47Following Giorgi and Pianesi (1997), what changes during the conventionalization process is the ‘consequent state’ variable in T2/Asp, which introduced the meaning component of a set of eventualities situated to the right of the temporal boundary of the event denoted by vP.

48HAVE GOT ‘stative possession’ no longer contains the meaning component of a set of eventualities following the event denoted by the vP. The loss of the ‘consequent state’ variable has direct consequentes for V_aux, as the eventive variable in V_aux can no longer enter into a relationship with the eventive variable of ‘consequent state’ (cf. the discussion of the present perfect in section 4.2).

49These changes have far-reaching consequences for the structural make-up of the present perfect expression HAVE got(ten) as depicted in (17). The consequences will be discussed in turn, starting at ground zero of the structure, as it were.

50One of the few things which remain relatively intact is the expression of possessive meaning. While there is a change in the semantic role of the subject from recipient to possessor, it is a fairly straightforward one. Semantically, the role of recipient is closely related to the role of possessor. Recipients have been described as “prospective possessors” (Pinker 1989: 48) and as “projected possessors” (Goldsmith 1980: 429, see also Green 1974, Goldberg 1995).

51Structurally, both recipients and possessors are base-generated in Spec P_have, as illustrated in (14) for HAVE ‘possess’ and (16) for GET ‘receive’ above. The structural configuration of P_have, Spec P_have, and the complement of P_have remains intact during the conventionalization process and accounts for the possessive meaning of HAVE GOT.

52The change in the meaning contribution of GET is similarly straightforward. The inchoative component v_become is not compatible with the semanticized meaning ‘stative possession’ and is dropped (cf. Gronemeyer 1999: 26). The absence of the eventive component v_become from HAVE GOT can be illustrated with the help of the adverbial modification test (cf. Embick 2004: 357, 363). While HAVE got(ten) allows modification by a manner adverbial, possessive HAVE GOT behaves like HAVE_poss in that it does not allow this type of modification, as illustrated in (18)-(20).

53The absence of v_become immediately raises questions about the make-up of the lower incorporation site of the expression. As outlined above, the structural configuration of P_have and its specifier and complement is still intact and signals possession. The spellout remains intact as got. It will be argued here that during the semanticization process v_be is substituted for v_become and serves as a new incorporation site for the lower P_have, as illustrated in (21).

54The spellout got at PF does not change to have (as might be expected for the incorporation of P_have into v_be) but is kept due to the shared meaning component of possession in both HAVE_poss and GET ‘receive’ (cf. Quinn 2009: 228 for a similar argument within a slightly different framework).

55The change in the value of the aspectual variable in T2/Asp can be observed most clearly in American English, where the past participle of GET spells out as gotten, while got is reserved for the possessive expression. The absence of the participial ending -en on the surface level clearly signals a loss of the notion of ‘consequent state’ here. In British English, on the other hand, the loss of ‘consequent state’ is evidenced only in the change of temporal reference but does not have any immediate repercussions on the surface level as got is the spellout in both the present perfect and the possessive expresion (cf. section 1).

56Variation between got(ten) and got as participle forms in earlier stages of British English as reported in Crowell (1959: 285) may have aided the loss of the meaning component of ‘consequent state’ through the process of exaptation, as learners assigned different meanings to the two forms (McIntyre 2010: 22, footnote 23).

57There are good reasons for the assumption that it is not only the value ‘consequent state’ which is lost. As only complex tenses in English have both a T1 and a T2 projection (cf. section 4.2), the switch from present perfect to simple present during the semanticization process entails a loss of the whole T2/Asp projection. Structurally, this leaves HAVE GOT hanging, as it were, as a frozen hybrid of two different structures, as illustrated in (22) above.

58The lower part is identical to that for HAVE_poss as illustrated in (14) save for the special spellout got (as per the discussion above). The vP, however, cannot behave like a regular vP as it is prevented from combining directly with Agr/T1 and being assigned person, number and tense. V_aux is still present either as a full form or a clitic, with a preference for the latter form (McIntyre 2010: 22).

59Cases of complete elision of V_aux have been reported for American English, British English dialects and for African American Vernacular English but will not be discussed in detail here (Crowell 1959; Tagliamonte 2003; Howe 2005).

60V_aux still has some of its auxiliary verb properties in HAVE GOT, as it negates directly and raises to Agr/T1 to be assigned person, as illustrated in examples (23a)-(24b). Dialectal material from Shropshire, Nottinghamshire and Leicestershire in Great Britain also shows Agr/T1 assigning tense, with past tense instances of HAVE GOT as illustrated in examples (25)-(27) (cf. Schulz 2012a).

61On the other hand, HAVE GOT does not have the gerund available to it, as illustrated in (28), which does not allow for a possessive reading, and has been reported to be marginal in combinations with other auxiliaries, as illustrated in (29).

62It is argued here that this type of partial defectiveness can be attributed to the impact of the loss of Asp/T2. In section 4.2 it was argued that the configuration of P_have and its Asp/T2 complement assigns auxiliary verb status to HAVE. The loss of Asp/T2 destroys this configuration. P_have still incorporates into vP_be, spells out as have or has and can precede a NEG head. Combinations with other auxiliaries, however, and infinitival forms are no longer an option. At the present stage we can only speculate why some auxiliary properties “survive” and others do not.

63Agr/T1 still performs some of the functions it performed in the present perfect expression. It checks person and number, assigns nominative case and makes sure that the eventive variable of the auxiliary is coextensive with the speech event. Coextensiveness with the speech event effectively signals present tense, as Agr/T1 can no longer enter into a relationship with the consequent state variable of T2/Asp, which established reference to the pre-present time zone in the present perfect expression.

64On the other hand, the loss of the relationship between Agr/T1 and T2/Asp renders Agr/T1 defective in some sense, freezing it in the form it had in those contexts in which ‘stative possession’ was semanticized. It cannot usually, for example, assign any other tense than present, as past tense uses of HAVE GOT are reported to be exceedingly rare to non-existent (see section 1).

6. Conclusion and outlook

65The present paper has argued that possessive HAVE GOT developed out of present perfect HAVE got(ten) ‘onset of possession’ via the conventio­nalization or semanticization of the conversational implicature ‘stative possession’. The conventionalization and subsequent semanticization of ‘stative possession’ trigger a change in temporal reference from the pre-present to the present, a loss of the participial status of got(ten), a loss of the eventive component v_become in GET and a change of the thematic role of the subject from recipient to possessor. Structurally, HAVE GOT can be argued to be a hybrid of a present perfect and a present tense expression, incorporating structural features of both. The heart of the changes can be located in the loss of the Asp/T2 projection, which renders both V_aux and Agr/T1 defective and limits HAVE GOT to non-modalized, finite, present tense affirmative or negative statements.

66As an anonymous reviewer of this paper has pointed out, the pragmatic, semantic and structural changes outlined here are idiosyncratic to stative HAVE GOT, rather than a productive pattern found throughout the language. This is mainly due to the expression-specific interaction between the meanings carried by the present perfect and the semantics of GOT, where ‘onset of possession’ develops into ‘stative possession’ and triggers a number of structural changes. The processes outlined here do have some value beyond an exercise in the development of an idiomatic verb complex in English, though, as they provide further evidence of work in grammaticalization that postulates effects of changes in the pragmatics and semantics of an expression on its syntax (cf. Diewald 2002).