1. Introduction
Dedicated impersonal pronouns like German man are structurally deficient. Evidence for this comes from the observation that they lack phi-features. Consider for example the contrast in (1). In (1a), man receives a plural interpretation, as shown by the fact that it can be the antecedent of the reciprocal einander; however, the ungrammaticality of (1b) indicates that it is not morphosyntactically associated with a plural feature: it cannot bear overt plural marking and it does not trigger plural agreement on the verb.
(1)
We assume that 3rd person singular agreement on the verb in (1a) is simply a default value that arises in the absence of phi-features on the subject.
It is well known that, in addition to dedicated impersonal pronouns we also find impersonal uses of personal pronouns, as illustrated in (2). The 2nd person pronoun du is impersonal in (2a): it is used to make a general statement about what people do in Austria. Impersonal du in (2a) does not establish a link to the current speech-act situation; that is, it does not refer to the current addressee. It is for this reason that the continuation in (2b), where du receives a personal interpretation and does pick out the addressee, is felicitous.
(2)
Notice that in both (2a) and (2b), du triggers 2nd person singular agreement on the verb, which indicates that, unlike man, impersonal du has inherent phi-features, specifically 2nd person singular. As a consequence, default 3rd person agreement on the verb in (3a) is ungrammatical.
(3)
It is widely assumed that the impersonal interpretation of dedicated impersonal pronouns reduces to a structural deficiency (Cardinaletti and Starke Reference Cardinaletti, Starke, Cardinaletti, Starke, Riemsdijk, Bosong and Comrie1999). This structural deficiency also accounts for their lack of phi-features. The question we address in this paper is whether the impersonal interpretation of personal pronouns is also attributable to a structural deficiency. Given that even in their impersonal use, these pronouns have phi-features, it follows that their structural deficiency must be of a different kind from that of impersonal pronouns, such as man.
We propose that nominals that refer to specific individuals have speech-act structure, as in (4a), and that this is crucially missing in du-type impersonal pronouns, which are simply DPs, as in (4b). We hypothesize that DP pronouns have a full set of phi-features, but that they cannot directly establish a link to the speech-act participants or the speech-act situation more generally, precisely because of their structural deficiency. Man-type impersonals are even more structurally deficient, in that they lack not only speech-act structure, but also the functional categories responsible for phi features: DP, PhiP, and nP, as in (4c).Footnote 2
Essentially this hypothesis can be reduced to two claims: First, all impersonal pronouns are structurally deficient in that they lack speech-act structure.Footnote 3 Second, impersonal pronouns can vary in the degree of structural deficiency they manifest – some lack only speech-act structure, while others lack additional layers of functional structure.
The rest of the paper is organized as follows: Section 2 provides additional arguments in support of the analysis schematized in (4) above for man, impersonal du, and personal du. In section 3, we motivate speech-act structure on theoretical and empirical grounds. In section 4, we develop a typology of structure deficiency for impersonal pronouns. We conclude in section 5 with a summary of the paper.
2. Degrees of structural deficiency in impersonal pronouns
The goal of this section is to demonstrate that both man and impersonal du are structurally deficient. In section 2.1, we provide additional arguments in support of the claim that man is a bare NP; in section 2.2, we discuss a range of similarities and differences between personal and impersonal du, offering additional arguments for the claim that impersonal du is a DP, and that it is nevertheless structurally deficient because it lacks speech-act structure, which is present in personal du.
2.1 Structural deficiency #1: man-type impersonals
In section 1, we demonstrated that the impersonal pronoun man cannot bear a plural suffix or trigger plural agreement on the verb. We interpreted these facts as evidence that the phrase headed by man lacks number specification. The most straightforward explanation for the lack of number specification is that the functional category associated with number features; that is, PhiP,Footnote 4 is simply not present. In the remainder of this subsection, we summarize arguments first introduced in Ritter and Wiltschko (Reference Ritter, Wiltschko and Hracs2016) that man also lacks person and case specification, due to the absence of other nominal functional categories, notably DP and PhiP. We also assume that man is genderless. However, this claim has no empirical consequences as nouns are not marked for gender, and German predicates do not inflect for gender either. The only way to detect gender is on the basis of determiners and nominal modifiers (adjectives and numerals), which crucially cannot co-occur with man.
The claim that man-type impersonal pronouns lack inherent person features has been well established in the literature (see Egerland Reference Egerland2003 for Swedish, Ackema and Neeleman Reference Ackema and Neeleman2018 for Dutch, among others). One argument in support of this claim is based on examples like (5), where the referent of man includes the speaker (1st person), the addressee (2nd person), as well as other humans (3rd person). This is evident from the fact that the speaker in this example asserts that a particular prohibition applies to both interlocutors as well as everyone else.
(5)
We assume that person specification is associated with D; thus, the fact that man is personless indicates that it lacks DP.
Man is also caseless. This can be deduced from the fact that man can function neither as an accusative case-marked direct object, as in (6a), nor as the oblique case-marked object of a preposition, as in (6b). The only position where man is licit is the subject of a tensed clause, as illustrated in (1) and (5) above.
(6)
Following Bittner and Hale (Reference Bittner and Hale1996), McFadden and Sundaresan (Reference McFadden and Sundaresan2010), Preminger (Reference Preminger2011, Reference Preminger2014), and Kornfilt and Preminger (Reference Kornfilt, Preminger, Joseph and Predolac2015), we assume that, at least for German, nominative case is in fact the absence of case. The reason for the ungrammaticality of the examples in (6) is that man cannot appear in positions that require structural (accusative) or lexical (oblique) case. Hence man can only occur in caseless positions, and is spelled out with the default morphological form.
On independent grounds, we know that German default case is nominative: it is morphologically unmarked, and in the absence of a case-assigner, nominals are realized as nominative.Footnote 5
(7)
(8)
A final piece of evidence for the claim that man is a bare NP comes from coordination. As illustrated in (9), man cannot be coordinated with a full DP such as seine Gäste ‘his guests’.Footnote 7, Footnote 8
(9)
Assuming that coordination requires categorial identity of the two conjuncts, the ungrammaticality of (9) follows on the assumption that man is a bare NP while the possessed nominal, seine Gäste, is minimally a DP. Similarly, man cannot be coordinate with a bare plural, which we assume is a PhiP.
(10)
Thus, the evidence supports our claim that man is structurally deficient. More specifically, we conclude that man is a bare NP lacking all functional projections. As a consequence of this structural deficiency, man lacks all nominal features (person, number, and case) and cannot be coordinated with nominals that include functional categories.
2.2 Structural deficiency #2: impersonal du vs. personal du
We now turn our attention to the structure of du, on its impersonal use. The goal here is to establish that impersonal du is structurally deficient as compared to its personal use. However, the argument will have to proceed differently than in the case of man. This is because du-type impersonals do not behave like man-type impersonals. Notably, unlike man, impersonal du triggers subject-verb agreement (2nd person singular) as illustrated in (11). Note also that (11) is ambiguous between a personal and an impersonal reading.
(11)
Another difference between man and impersonal du concerns case. Unlike man, and like personal du, impersonal du is not restricted to the subject position of a tensed clause, which indicates that it can bear structural (or lexical) case. Moreover, the pronoun is morphologically marked for case (du is nominative, dich is accusative).
(12)
The fact that impersonal du has case and phi-features points to the conclusion that it is (at least) a DP, and hence not structurally deficient in the same way as man. This leads to the prediction that unlike man, impersonal du can be coordinated with a full DP. This prediction is borne out as illustrated by the following example:
(13)
The evidence adduced above establishes that impersonal du behaves like a DP, which in turn raises the question as to whether it is structurally deficient, and if it is, what the nature of its structural deficiency might be. A first indication that it is indeed structurally deficient comes from differences in the interpretation possibilities of strong and weak 2nd person pronouns in Dutch. As first discussed by Gruber (Reference Gruber2013), the weak version may be interpreted as either a personal or an impersonal pronoun, but the strong version is necessarily personal, as illustrated below.
(14)
Our hypothesis is that the contrast between (14a) and (14b) is due to the fact that strong pronouns are structurally more complex, and specifically, that they are dominated by a speech act structure. Weak pronouns, which lack speech act structure, may be considered structurally deficient, even if they are full DPs. Note, however, that structural complexity cannot always be deduced from form. As an anonymous reviewer points out, in both Russian (15) and Georgian (16), overt and null pronominals (pro) can be used impersonally.
(15)
(16)
We interpret this fact as evidence that the difference between overt and covert pronouns does not directly reflect structural complexity. Given that the pronouns in these languages can be used as impersonals, we predict that they may lack speech act structure.
A related fact, pointed out by a second anonymous reviewer, is that in Dutch when a 2nd person impersonal pronoun is used contrastively, it is spelled out as the strong form (jij), as shown in (17).
(17)
We interpret this to mean that there are two routes to the spellout of a “strong” pronoun: either via structural complexity or via prosodic weight. Significantly, jij must be stressed when used as an impersonal pronoun, though it need not be when it is used as a personal pronoun. We submit that despite the phonological similarity, the structural contrast remains, that is, the former is still a DP while the latter contains the speech-act layer.
In sum, we have argued that man is structurally deficient: it lacks all phi-features, a deficiency which we propose is due to the lack of functional super-structure. As a consequence, it is restricted to subject position. We have also argued that impersonal du is a full DP, based on the fact that it has all phi-features and is not restricted to subject position. The question then arises as to whether there is a common source for the impersonal interpretation found with both man and impersonal du. In what follows, we argue that there is. More specifically, we argue that the structure of those personal pronouns that refer to the current speech-act participants must include a nominal speech-act structure, and that personal pronouns interpreted impersonally lack this layer of structure. In other words, we hypothesize that the distinguishing property shared by all impersonal pronouns is their structural deficiency. Man-type impersonals lack all nominal functional categories, while du-type impersonals lack those functional categories that make up the speech act structure.
3. Motivating nominal speech act structure
In this section, we elaborate on how the presence of speech-act structure correlates with the personal interpretation of personal pronouns, and on why all types of impersonal pronouns lack this layer of structure. In section 3.1, we offer theoretical motivation for a speech-act structure in the structure of referential nominals. In section 3.2, we introduce our hypothesis regarding the contents of nominal speech-act structure. Specifically, we extend Wiltschko's (Reference Wiltschko, Coon, Massam and Travis2017) model for speech-act structure to nominals, and provide empirical evidence that this layer is absent in impersonal pronouns.
3.1 Theoretical motivation for nominal speech act structure
Ever since Chomsky (Reference Chomsky, Jacobs and Rosenbaum1970), some degree of parallelism between nominal and clausal syntactic structure has been assumed. Chomsky observed similarities in the argument structure of related nouns and verbs, and attributed this to category-neutral phrase structure rules (i.e., X-bar Theory). Subsequent work introduced the idea that these same phrase structure rules apply to clausal functional categories, notably COMP and INFL (generalized X-bar Theory, Travis Reference Travis1984, Chomsky Reference Chomsky1986). And indeed, just as there is evidence for a parallel between the lexical projections of VP and NP, there is also evidence that such parallelism exists in the functional architecture of the nominal and clausal spines. For example, in his development of the hypothesis that determiners host their own functional category (DP), Abney (Reference Abney1987) suggests that DP is the nominal counterpart of the clausal functional category IP (later re-conceptualized as TP; Pollock Reference Pollock1989). The structural and functional parallel between the nominal and the clausal spine has since been widely adopted, with a variety of empirical and conceptual arguments (e.g., Szabolcsi Reference Szabolcsi, Kiefer and Kiss1994, Lobeck Reference Lobeck1995, Chomsky Reference Chomsky2015).
For the purposes of this paper, we adopt the Universal Spine Hypothesis (henceforth USH, Wiltschko Reference Wiltschko2014) which asserts that Universal Grammar contributes a set of hierarchically organized, abstract categories defined by their interpretive function. For the USH, the parallel between nominals and clauses follows from the assumption that language-specific categories are derived from the interpretive spinal functions and language-specific units that give these abstract categories further substantive content. Crucially, the universal spinal functions are neutral with respect to whether they apply to lexical categories denoting eventualities (verbs) or individuals (nouns), and hence are category-neutral: they define both the extended verbal phrase (i.e., the clause) and the extended nominal phrase. This leads to the expectation that all functionally defined categories in the clausal domain will have an analogue in the nominal domain, and vice versa.
In the last couple of decades, a significant body of research has explored the syntactic representation of elements of the speech act, notably the speaker and addressee (e.g., Benincà Reference Benincà, Cinque and Salvi2001; Speas and Tenny Reference Speas, Tenny and Di Sciullo2003; Garzonio Reference Garzonio2004; Tenny Reference Tenny2006; Hill Reference Hill2007a, Reference Hill2007b; Haegeman and Hill Reference Haegeman, Hill, Folli, Sevdali and Truswell2013; Zu Reference Zu and Arapicio2013, Reference Zu2018; Haegeman Reference Haegeman2014), or the beliefs and intentions of the speaker and addressee (Wiltschko and Heim Reference Wiltschko, Heim, Kaltenböck, Keizer and Lohmann2016, Wiltschko Reference Wiltschko, Coon, Massam and Travis2017). Though the proposals differ in their details, they all converge on the idea that elements of the speech act are represented in the functional architecture at the top of the clausal spine. In other words, the speech act structure makes up the topmost layer of the sentence. Combining the assumption that the structural parallel between clauses and nominals is due to the USH with the assumption that the clausal structure contains a speech-act structure, we are led to postulate speech-act structure in nominals as well. This is schematized in (18).
(18) Parallel clausal and nominal structure?
Much of the work that focuses on clausal speech act structure takes as its point of departure Ross's (Reference Ross, Jacobs and Rosenbaum1970) performative hypothesis; that is, the hypothesis that the speaker and addressee, along with illocutionary force, are syntactically represented. Ross proposes that even declarative clauses are explicitly performative in that they encode what the speaker is doing with the utterance, namely “telling the addressee the proposition”. In his analysis, the Deep StructureFootnote 9 of every declarative sentence contains a matrix clause with the speaker as the agent (subject), the addressee as the goal (indirect object) and the utterance content as the theme (direct object). The representation of the speech act is subsequently deleted so that only the proposition itself is represented at Surface Structure, as illustrated in (19).
(19)
The performative hypothesis was criticized from the very outset. One consequential problem is that it is not clear how to stop an infinite regress with examples like (19), where each performative layer would be embedded in another performative layer (I tell you that I tell you that I tell you that …). Given counter-arguments such as this, and with the demise of the generative semantics framework that Ross (Reference Ross, Jacobs and Rosenbaum1970) assumed, the performative hypothesis was fully abandoned. However, many of the observations that Ross's proposal aimed to capture were then left without explanation.
With the rise of functional architecture, the door was opened to analyse Ross's speech-act structure as part of the functional architecture, thus overcoming the problem of infinite regress. Starting in the mid 1990s, several researchers re-analysed the performative hypothesis in terms of functional architecture, re-interpreting Ross's matrix clause structure as follows: speaker and addressee are assumed to be speech act roles, associated with one or more functional speech act heads. The functional heads that introduce these roles replace the verb of communication proposed by Ross (tell in (19)). Take, for example, Speas and Tenny's (Reference Speas, Tenny and Di Sciullo2003) version of this analysis: they assume that speech-act structure makes use of the same formal primitives as argument structure. To introduce an assertoric speech act of the sort “I tell you that”, they assume that the structure is akin to a double-object construction with an articulated speech act shell as in (20).
(20) [saP Spkr [sa [SAP Utt [SA Hearer]]]
Wiltschko and Heim (Reference Wiltschko, Heim, Kaltenböck, Keizer and Lohmann2016) refer to this recasting of Ross's original proposal as the neo-performative hypothesis. The problem with this particular implementation of the idea that there is a speech-act structure dominating the clause is that it adopts the version of speech act theory that was prevalent at the time Ross first introduced the idea, namely the version proposed by Austin (Reference Austin1962) and Searle (Reference Searle1969). However, since then, speech act theory has developed in various ways (see Clark and Brennan Reference Clark, Brennan, Resnick, Levine and Teasley1991 for a recent overview). What emerges from the more recent literature is that one cannot ignore the role of the addressee. In particular, even an assertion conveyed by a declarative with falling intonation cannot be regarded as successful transfer of information from the speaker to the addressee. The felicitous performance of an assertion is dependent both on the speaker giving the addressee the content of an utterance, (i.e., a piece of information), and on the addressee receiving and accepting, rejecting or further discussing this piece of information. It is only after the addressee accepts this information that the proposition can be added to the common ground (in the sense of Stalnaker Reference Stalnacker2002). This process of ensuring mutual understanding between two interlocutors defines dialogical interaction. It is what Clark and Brennan (Reference Clark, Brennan, Resnick, Levine and Teasley1991) refer to as grounding. Crucially, natural languages have forms dedicated to managing grounding in this sense, such as confirmationals which are specifically used to request acceptance of the host utterance into the common ground (I have a new dog, eh?), as well as response markers which are used to indicate acceptance (Yeah, I know!) or rejection (No! You're wrong!).
To accommodate the interactive dimension of meaning, Wiltschko and Heim (Reference Wiltschko, Heim, Kaltenböck, Keizer and Lohmann2016) and Wiltschko (Reference Wiltschko, Coon, Massam and Travis2017) develop a framework within which to explore the syntax of interactional language. They propose that the speech act structure consists of two layers, each associated with an abstract spinal function. The grounding layer (GroundP) is responsible for encoding the speaker's attitude towards what is being said. This layer can be articulated to further encode the speaker's assessment of the addressee's attitude. The responding layer (RespP) is responsible for encoding what the speaker wants the addressee to do with what is being said (i.e., respond or not respond). In English, the request for a response is typically realized with intonational tunes. Evidence for the assumption that speech act structure includes representations of the speaker's and addressee's attitude (i.e., that it realizes the relation of the proposition to the interlocutors’ Ground), rather than the speech-act roles, comes from the fact that speech act modifiers, such as confirmationals, modify the speech act participants’ attitudes rather than the speech act participants themselves. For example, within and across languages we find confirmationals that differ in terms of the timing of belief (i.e., whether the speaker came to believe the proposition before or during the time of the conversation) or in terms of the robustness of belief (i.e., how strongly the speaker believes the proposition). This type of modification is precisely what we would expect if the relevant functional category represents the speech-act participants’ grounds. The interactive structure dominating the propositional structure encodes what we do when we utter sentences.
3.2 The speech-act structure of nominals
The conceptual argument introduced in the previous section leads us to expect a nominal speech-act structure. In this section, we show that this expectation is borne out. Specifically, we provide empirical evidence that grounding applies not only to propositions but also to individuals (section 3.2.1). Furthermore, we argue that impersonal pronouns, precisely because they are not part of the common ground, lack a grounding structure (section 3.2.2).
3.2.1 Grounding of nominals
According to Bavelas et al. (Reference Bavelas, De Jong, Korman and Jordan2012: 1), grounding is the “fundamental, moment-by-moment conversational process by which speaker and addressee are constantly establishing mutual understanding.” The goal of a conversation is to establish common ground so that the interlocutors know what they each believe, and sometimes also why they believe what they do, how they came to these beliefs, how strongly they hold them, and their status in the common ground. Thus, the common ground contains propositions and the interlocutors’ attitudes towards these propositions (i.e., propositional attitudes). Crucially for our purposes, the common ground contains not only propositions but also discourse participants, the individuals we talk about (Karttunen Reference Karttunen and McCawley1976, Webber Reference Webber1979). And just as with propositions, interlocutors will establish who it is they are talking about, sometimes how they relate to these individuals, and their status in the common ground. These are the interlocutor's attitudes towards these referents and we suggest that these constitute the interlocutors’ referential attitudes.
Different types of DPs express different referential attitudes. For example, the difference between definite and indefinite DPs can be characterized as encoding whether or not a particular discourse referent is new or old relative to the common ground, a difference that clearly mirrors the above-mentioned encoding of timing of belief. Thus, we might expect that similar mechanisms for managing the grounding of referents can be found in the nominal domain. And this is in fact so, as the following conversation illustrates.
(21)
This conversation illustrates that reference is a collaborative act between interlocutors. What appears to be a disfluency in A's contribution is in fact an act of requesting confirmation that the individuals under discussion are in the common ground. That is, A wants to say something about Mr. and Mrs. Ford, but first must establish that B knows who A is talking about. In order to do so, A uses a pause, rising intonation and hesitation particles (uhm tch) to indicate that they are seeking confirmation that B knows who the Fords are, and in particular who Mrs. Ford is. A begins by naming the individuals under discussion, and then follows this with a description of the focal individual, and asks specifically if B knows who A is talking about – the cellist, Mrs. Holmes Ford. B responds in the affirmative, and then uses a pronoun to refer back to this individual, and to confirm that they know who she is. Having established the identity of the persons under discussion, A can proceed to tell B something about them. We propose that the three strategies used by A and B in this conversation, that is, naming, describing, and tracking constitute nominal speech acts, and that they are typically expressed by different kinds of nominals – proper names, common noun phrases, and pronouns, respectively.Footnote 10 We propose that like clausal speech acts, nominal speech acts are syntactically represented. In other words, we expect a dedicated layer of structure to encode what we do when we utter nominals. Recall, however, that the main focus of this paper is to demonstrate that impersonal pronouns are structurally defective in that they lack at least speech-act structure. We contend that the contrast between impersonal pronouns and other types of nominals provides us with indirect evidence for a speech-act structure in the nominal spine. In the next section, we provide empirical evidence that this is the essence of the structural deficiency in all impersonal pronouns.
3.2.2 Grounding of impersonal pronouns
One of the key functions of speech-act structure is to manage turn-taking, in that it marks the request for a response from the interlocutor. This implies that utterances that make up turns have to be dominated by speech-act structure, which in turn predicts that phrases without speech-act structure, for example impersonal pronouns, cannot be complete turns.
Evidence that this is indeed the case comes from answers to wh-questions. At least in English, the most natural answer to a wh-question asking about an individual person or object is just a nominal that names, describes, or refers to the individual, as in (22). Thus, a fragment answer consisting of a nominal only (A1) is preferable to a full sentence answer (A2).Footnote 11 Note that our interest here is not in the preference for A1 over A2, but in the well-formedness of A1.
(22)
In (22), names, definite descriptions, and pronouns are well-formed answers because they can serve as turns, and thus, by hypothesis, they have speech-act structure in their syntactic representation.
Interestingly, while most types of nominals are completely acceptable as sentence fragment answers to English wh-questions (22, A1), impersonal pronouns are not (22, A3). This is also true for German, where all types of noun phrases except impersonal pronouns can serve as answers to a wh-question, as shown in (23).
(23)
Note, however, that even in full answers, impersonal pronouns are not always possible.
(24)
It seems that impersonal pronouns are only possible in answers to questions that are clearly generic, and perhaps rhetorical. (We thank an anonymous reviewer for this observation.) Furthermore, an appropriate answer seems to require the negation of the proposition. Significantly, the same generalization holds for fragment answers: impersonal pronouns are not possible, but other types of nominals are.Footnote 12
(25)
Summarizing the discussion in this section, we have introduced the idea that there is a speech-act structure in the nominal spine, parallel to the speech-act structure in the clausal spine. Impersonal pronouns provide empirical motivation for this hypothesis to the extent that their defective distribution can be attributed to the absence of this layer of nominal structure. What this means is that impersonal pronouns cannot be used to perform nominal speech acts because they do not name, describe, or refer to specific individuals, and thus, they cannot function as sentence fragment answers.
4. A typology of structural deficiency in personal pronouns
The defining property of impersonal pronouns is that they do not relate to the speech act participants. We propose that this property derives from the lack of speech-act structure. Thus, we propose that impersonal pronouns are characterized by structural deficiency. However, there is – in principle – no reason to assume that their structural deficiency would be limited to the lack of speech-act structure. In this section, we demonstrate that all logical possibilities are attested. Assuming that a structurally complete nominal has the full structure in (26), then we expect that impersonal pronouns can be deficient in several ways.Footnote 13 Hence, we expect to find DP-impersonals that lack only GroundP; PhiP-impersonals that lack GroundP and DP; nP-impersonals that lack GroundP, DP, and PhiP; and finally, NP-impersonals that lack all functional categories – GroundP, DP, PhiP, and nP. These options are indicated by the arcs in (26).
(26) The typological space for impersonal pronouns
We have already seen that impersonal du and impersonal you are DP-impersonals, and that man-type impersonals are NP-impersonals. Below we provide evidence for PhiP-impersonals and nP impersonals, in turn.
In Ritter and Wiltschko (Reference Ritter, Wiltschko and Hracs2016), we discuss the distribution of Modern Hebrew impersonal pro as seen in (27).
(27)
The first thing to note is that Hebrew impersonal pro triggers plural agreement on the verb. Assuming that the locus of plural marking is PhiP, this would indicate that Hebrew impersonal pro is minimally a PhiP. Significantly, impersonal pro is available in all tenses (present, past, and future), as illustrated in (27). However, in Hebrew referential pro is only available in past and future tense, as shown in (28).
(28)
The tense restriction on referential pro is widely attributed to the fact that in Hebrew, past and future verbs are inflected for all phi features (person, number, and gender), while present tense verbs are inflected for number and gender, but crucially not for person. This is exemplified in (25) and in the verb paradigms in Tables 1–3.
Verbal agreement for past tense
Verbal agreement for future tense
Verbal agreement for present tense
We assume that Phi is the locus for number and that D is the locus for Person. The facts of Modern Hebrew thus provide striking evidence that impersonal pro is a PhiP; it is specified for number (and gender), but not for person, and thus instantiates the PhiP-impersonal in the typological space predicted by our analysis.
Evidence for the fourth type of impersonal pronouns (nP-impersonals) comes from a comparison of Standard German man and its Austrian German cognate ma. Both man and ma can have a universal interpretation, that is, they can both be used to refer to all possible persons, as illustrated in (29)–(30).
(29)
(30)
However, the two dialects differ in that Standard German allows for an existential interpretation of the impersonal, but Austrian German does not. In other words, Standard German man can be used to refer to an individual who is not identified in the discourse, but Austrian ma cannot. (In Austrian German, the wh-word that functions as an indefinite pronoun is used instead.) There are two cases to be considered. First, in (31a) and (32a), the impersonal/indefinite pronoun is used to refer to one specific individual who the speaker cannot or will not identify. Second, in (31b) and (32b), the impersonal/indefinite pronoun is used as a variable over individuals that co-varies with events. In these examples, for every instance of the speaker's visiting Vienna, there is an individual who steals their bike.
(31)
(32)
We assume that the existential interpretation requires a contextual variable to be associated with the nominal reference, whereas the universal interpretation is not contextually restricted. Following Wiltschko (Reference Wiltschko2013), we further assume that contextual variables are associated with nP. We propose that the contrast between the two varieties of German can be attributed to a categorial difference between man and ma: man is an nP-impersonal, and thus is associated with a contextual variable; ma is an NP-impersonal, and thus lacks a contextual variable. As a consequence, only man can be used existentially.
In sum, we have now seen that impersonal pronouns come in several different guises. What they all have in common is that they lack speech act structure (GroundP) as they cannot be linked to the speech act participants. However, all other possibilities are attested: impersonal du/you instantiates DP-impersonals, Hebrew pro instantiates PhiP-impersonals, Standard German man instantiates nP impersonals, and Austrian ma instantiates NP-impersonals. This typology is represented in (33).
(33) The typology of impersonal pronouns
Note that in our small sample of impersonal pronouns, PhiP-impersonals stand out as they are instantiated by pro rather than an overt pronoun. There is nothing in our proposal that would predict this to be the case, and thus it is an empirical question whether we might also find overt PhiP-impersonals and similarly, whether there are null versions of the other types of impersonal pronouns.
5. Conclusion
The goal of this paper was to introduce the idea that there is a speech act structure in the nominal spine, just as there is in the clausal spine. Its function is to encode what we do when we utter a nominal: we name, describe, or track individuals. In this way, the speech-act structure establishes a link between the discourse referent and the speech-act situation. The evidence discussed in this article comes from nominals that lack this speech-act structure, namely impersonal pronouns. Specifically, what all impersonal pronouns have in common is that they lack the ability to connect to the speech-act participants; they do not name, describe, or track discourse referents. However, impersonal pronouns are otherwise not a natural class. They vary in syntactic category and as a consequence, they differ in morphosyntactic features, and in distribution. Note that it has long been established that there is variation in the structure of personal pronouns (e.g., Ritter Reference Ritter1995, Cardinaletti and Starke Reference Cardinaletti, Starke, Cardinaletti, Starke, Riemsdijk, Bosong and Comrie1999, Déchaine and Wiltschko Reference Déchaine and Wiltschko2002) and hence, it comes as no surprise that impersonal pronouns also differ structurally.
As pointed out by an anonymous reviewer, cross-linguistically, it is the second person singular that often can be used impersonally, and, it appears, never the first person singular or the third person singular. At this stage, our analysis has nothing to say about this intriguing contrast and we will leave it for future research.
The hypothesis that nominals, like clauses, have speech-act structure fits squarely within current investigations of the syntax-pragmatics interface. It allows for a novel approach to the discourse properties of nominals, which goes beyond their role as thematic arguments and modifiers of predicates. As such, it opens up a new research agenda with a plethora of new research questions regarding the syntax, semantics, and pragmatics of different types of nominals.
