1. ONE GENDER SYSTEM OR TWO?

The main idea of Enger (Reference Enger2004), on which he bases his later criticism of Josefsson (Reference Josefsson2009), is that there is only one gender system in Mainland Scandinavian (henceforth MSc), and that this system is semantic in nature. The main claim of Josefsson (Reference Josefsson2009), see also Josefsson (Reference Josefsson2006, Reference Josefsson2010, Reference Josefsson2012b, Reference Josefsson2013, Reference Josefsson2014), originally based on a proposal in Teleman (Reference Teleman and Teleman1987), is that there are two gender systems or gender dimensions in Swedish (and probably also in the other MSc languages). The systems or dimensions can be termed formal (or syntactic) gender and semantic gender. These two systems are independent, but they interact closely. Importantly, Josefsson (Reference Josefsson2009) does not argue, which is implied in Enger (Reference Enger2013:282), that formal gender is a system for nouns, whereas semantic gender operates within the pronominal domain. Things are more complex, as I will touch upon here. (For reasons of space the reader is referred to the references above for a more comprehensive discussion.) Following Josefsson's system, nouns have formal gender (which I take to be uncontroversial), but when we consider pronouns, there are pronouns that express formal gender and pronouns that express semantic gender. In what follows, I sketch the basic properties of the two systems.

Swedish has two formal genders, common gender and neuter; the formal gender is inherent to a nominal root or to a derivational suffix.Footnote ¹ For example, tiger ‘tiger’ and stol ‘chair’ are common gender nouns, and lejon ‘lion’ and bord ‘table’ are neuter nouns. A given nouns's formal gender is marked on the definite determiner, as shown in (1).Footnote ²

1. (1)

   

We may refer to the noun phrases tigern and lejonet by a personal pronoun that picks up the formal gender of the noun. In addition, a predicative adjective agrees with the formal gender of the subject:

1. (2)

   

Drawing on Bosch (Reference Bosch1983, Reference Bosch1986, Reference Bosch1988), I will call pronouns that refer back to linguistic entities (typically noun phrases or pronouns) Syn-pronouns (an abbreviation for ‘syntactic pronouns’). Consequently, den and det in (2) are Syn-pronouns, since they refer back to the noun phrases tiger and lejonet, respectively. Bosch mentions explicitly that Syn-pronouns, in his terminology called S-pronouns, can be thought of as akin to agreement; see Bosch (Reference Bosch1983:215). (Pronouns that refer to non-linguistic entities, for example deictic pronouns, are termed Ref-pronouns in Josefsson (Reference Josefsson2013, Reference Josefsson2014). Bosch refers to such pronouns as R-pronouns.Footnote ³)

Semantic gender is the gender that reflects properties of a referent, as viewed by a beholder. Let us consider a noun such as hund ‘dog’, which is a common (c) gender noun, as witnessed by the form of the definite article, hund-en (dog-c.def) ‘the dog’. We may talk about this animal in different ways:

1. (3)

   

If it is possible to use all four personal pronouns hon, han, den, and det in (3) to refer back to ‘dog’, it is clearly not appropriate to refer to these pronouns as instances of agreement with the noun hund. The very term agreement implies morphosyntactic ‘sameness’ between a feature of a controller and a target. If there is no restriction as to the feature content of the controller, the term agreement is devoid of content. A better way of describing the relation between the pronouns and their antecedents in (3) is as follows: If speakers know that a dog is female or male (and also think of dogs in terms of being basically animate/humanlike) they tend to use the pronouns hon ‘she’ or han ‘he’. With a ‘ground’ reading, as in (3d), the neuter pronoun det can be used.Footnote ⁴ The pronouns do not express agreement with the noun hund in any of these cases; the choice of pronoun expresses the view that the speaker takes on the referent that is talked about – these pronouns are therefore Ref-pronouns. However, the use of den in (3c) is slightly more complicated. In this context it could either be a Syn-pronoun parallel to (2a), or it could be an Ref-pronoun, where den refers to a bounded discourse entity, in essence similar to (4a) below.Footnote ^5, Footnote ⁶

It has been a corner-stone of Corbett's work that pronouns are expressions of agreement relations (Corbett Reference Corbett2006:21–22), a view that Enger (Reference Enger2013:280) claims to support. However, under the analysis presented in Josefsson (Reference Josefsson2012b, Reference Josefsson2013, Reference Josefsson2014), examples such as those in (3) do not lend support to this way of thinking. The pronouns den and det in (2a) and (2b) are indeed akin to agreement, but not hon, han and det in (3a), (3b), and (3d). The choice between den (3.sg.c) and det (3.n) in (2a–b) is due to the formal gender of a noun in the preceding clause, but this is not the case for hon ‘he’ and han ‘he’. (For a detailed argumentation showing that the pronouns han ‘he’ and hon ‘she’ lack a formal gender in Swedish, and also that nouns do not carry morphosyntactic features, such as feminine and masculine, see Josefsson Reference Josefsson2010:2100f.)

Deictic pronouns provide an even stronger argument that we have to differentiate between pronouns that refer to linguistic entities (Syn-pronouns), typically noun phrases and pronouns, and pronouns that refer directly to discourse referents (Ref-pronouns). Recall that den and det can be used as Syn-pronouns (see (2a) and (2b)). However, den and det can be used in a purely deictic way too, common gender den making reference to a bounded entity of some sort, neuter det being a default alternative. Both (4a) and (4b) below can be used as out-of-the-blue utterances, without the speaker knowing the appropriate denomination of the referent that is praised:

1. (4)

   

In view of examples such as those in (1)–(4), it is not at all clear why Enger (Reference Enger2013:283) argues against the splitting of gender into ‘one semantic and one non-semantic part’. It is obviously true that a pronoun in some cases conveys semantic information, but in other cases it does not do so. At the very least, the pronouns hon ‘she’, han ‘he’, and det ‘it’ in (3) and den and det in (4) tell us something about the semantics of the referent, whereas the distinction between the common gender den in (2a) and the neuter det in (2b) has no semantic significance. As far as I am aware, it is impossible to come up with any reasonable semantic distinction that would motivate the use of den as an anaphoric pronoun for stolen ‘the chair’ whereas det is used for bordet ‘the table’. The tendency that inanimate and substance entities are often neuter does not help us at all to explain this difference.

In addition to the preceding discussion, it is necessary to point out that Josefsson (Reference Josefsson2009) does not claim that formal gender, viewed as a system, is completely ‘asemantic‘, as implied by Enger (Reference Enger2013:283). What Josefsson (Reference Josefsson2009:40) claims is that formal gender is arbitrary in the sense that the formal gender of a noun is not predictable from its meaning; ‘there is simply no element of meaning shared by all neuter nouns’. The fact that tiger ‘tiger’ and stol ‘chair’ are common gender nouns, whereas lejon ‘lion’ and bord ‘table’ are neuter are not isolated exceptions. There are tendencies, of course, that countables and/or animates are often common gender, whereas non-countables and/or inanimates are often neuter; this was pointed out already in Josefsson (Reference Josefsson1997, Reference Josefsson1998) and mentioned also in Josefsson (Reference Josefsson2009:66). To the best of my knowledge, no one has denied that such a typicality relation holds. How to characterize the relation between the tendencies above and the semantic gender system, as expressed in personal and deictic pronouns, is an interesting problem that remains to be solved. However, simply stating that neuter as a formal gender of nouns and neuter on pronominal det (3.n) ‘it’ is the same does not seem to me to shed any light on the problem.

Enger (Reference Enger2013:282) ascribes to Josefsson (Reference Josefsson2009) the view that agreement on predicative adjectives could not carry semantic information. This is not correct; on the contrary, Josefsson (Reference Josefsson2009:38) makes the following claim:

Josefsson's system does not require the extra device that Enger (Reference Enger2013:276ff.) advocates: purely referential, but not morpho-syntactic agreement. This question will be discussed in more detail below.

Part of the discussion below will revolve around the construction sometimes referred to as ‘pancake sentences’. In short, pancake sentences are sentences in which there appears to be disagreement between the subject and a predicative adjective, where such an adjective occurs in the neuter form, completely independently of the gender and number on the noun phrase in the subject position. Three Swedish examples are given here in (5).

1. (5)

   

The main point of the solutions presented in Josefsson (Reference Josefsson2009, Reference Josefsson2010, Reference Josefsson2013, Reference Josefsson2014) is that neuter agreement on a predicative adjective in such pancake sentences is triggered by a null version of the pronoun det (3.n) assumed to head the subject XP, which, in turn, can be an NP, vP or a Small Clause). Crucially, this null det lacks a number feature, as do the corresponding overt instances of det in MSc. (One of the main points in Josefsson's above-mentioned analyses is that the feature content corresponding to non-countability is the radical absence of number.) The absence of a number feature corresponds morphologically to predicative agreement in the neuter, to the semantic interpretation of the subject as a non-countable entity, and to the blocking of canonical agreement. Naturally, it is impossible to prove that a null element is present in a structure, but judging from other construction types it is argued that this null det does exist in MSc. A corresponding overt det is possible in some varieties of MSc, for instance, in Jutlandic, and also spoken Danish, but not in Swedish; further discussion of pancake sentences follows in the next sections.

Enger (Reference Enger2013:282) claims that the solution proposed in Josefsson (Reference Josefsson2009) implies that neuter on the predicative adjective in the three examples below have different motivations, which would weaken Josefsson's position. (The sentences in (6) are the Swedish equivalents of Enger's sentences. Further, note that (6a) is a pancake sentence.)

1. (6)

   

As a matter of fact, examples, such as (6b), the doubling of a clause-initial element with a det that appears not to agree with the preceding DP, are not discussed at all in Josefsson (Reference Josefsson2009), so the question of their analysis does not arise in that paper. However, the construction is discussed at length in Josefsson (Reference Josefsson2010, Reference Josefsson2012a), where it is argued that the source for neuter on sun-t (healthy-n) ‘healthy’ in (6a) and det (3.n) ‘it’ in (6b) is the same: The proposed null version of neuter det, assumed to head the subject noun phrase in (6a), [Ø_detvodka], has the same feature content and interpretation as the overt det in (6b). In other words, neuter agreement on the adjective in (6a) has exactly the same motivation as the use of det in (6b). Importantly though, the syntactic structures in (6a) and (6b) are not identical. Following Eide (Reference Eide2011), I assume that the left dislocated element vodka in (6b) is located in a separate clausal domain, whereas the suggested null version of det in (6a) heads the noun phrase in the subject XP. (See Eide Reference Eide2011 for arguments regarding det in (6b).) Consequently, the subject det in (6b) triggers agreement on a predicative adjective in a canonical way in sentences such as Vodka, det är stark-t (vodka(c), 3.n is strong-n) ‘Vodka is strong’ – the left dislocated element vodka does not trigger agreement at all. The construction in (6c) is not discussed in Josefsson (Reference Josefsson2009) either. However, this does not lend support to Enger's (Reference Enger2013) criticism. The adjective in (6c) agrees with the overt infinitival phrase which (like clauses in general) is assumed to lack a number feature); the NP vodka is extracted from the infinitival phrase and fronted to a sentence-initial position – in other words, this is an instance of tough-raising.Footnote ⁷ It should be pointed out that an expletive det (3.n) ‘it’ can be optionally inserted in the subject position in (6c), yielding Vodka är det gott att dricka (vodka is expl.n good.n to drink) ‘It is good to drink vodka’. This paper is not the proper place for discussing whether the subject position in (6c) is radically empty or contains a null expletive neuter det; the reader is referred to Falk (Reference Falk1987, Reference Falk1993:270) and Engdahl (Reference Engdahl2010) for more discussion about this type of construction. It is fully possible that an expletive det is intimately related to other uses of det, but the question is complex. The only point I want to make is that a sweeping generalization, with a claim that there is but one gender system, without a detailed discussion concerning the different construction types does not automatically make an analysis stronger.

To conclude so far, we can say that personal pronouns cannot generally be viewed as exponents of agreement relations. Some pronouns, used in some contexts, refer back to noun phrases, and we could very well think of them as akin to agreement. The gender of the pronoun does not carry any semantic information in these cases. In other instances, a pronoun refers to a discourse entity – such pronouns are not agreement-like. In the latter case, common gender den is used to refer to countable entities and neuter det to non-countables, regardless of any noun that is conventionally used for the referents in question. In my view this shows that there are indeed two gender dimensions.

2. AGREEMENT INSIDE AND OUTSIDE THE NP

Enger (Reference Enger2013:287) implies that Josefsson (Reference Josefsson2009) argues that agreement inside the NP/DP is ‘asemantic’; this is either a misconception or a misreading. On the contrary, Josefsson (Reference Josefsson2006, Reference Josefsson2009, Reference Josefsson2010, Reference Josefsson2013, Reference Josefsson2014) argues that there is a ‘high’ position within the functional domain of the noun – thus inside the NP/DP – that may host pronouns expressing semantic gender. The formal gender of the noun, on the other hand, is inherent to the root; formal gender is arbitrary in the sense that it is not predictable from the meaning of the noun. The formal gender of the noun is used in the spell-out of definiteness, which is why we get den vita mjölk-en (def.c white.def milk-c.def) ‘the white milk’, and thus common gender occurs even on a noun that so clearly denotes a substance. The pre-nominal position in the functional domain of the noun, hosting what I have identified as a semantic gender marker (Josefsson Reference Josefsson2006, Reference Josefsson2009:39) or classifier (Josefsson Reference Josefsson2013:34ff.; Reference Josefsson2014), explains why we find Mjölk är vit-t (milk(c) be.prs white.n) ‘Milk is white’ – where ‘milk’ has a clear non-countable meaning – and not ?*Mjölk är vit (milk(c) be.prs white.c). (As shown above, I assume that the subject in Mjölk är vitt is headed by a null version of neuter det.)

First of all, evidence that a position for ‘gender markers’ or ‘classifiers’ does exist comes from the possibility of adding han ‘he’ and hon ‘she’ before a definite DP; han and hon are assumed to be expressions of semantic gender. (See Josefsson Reference Josefsson and van Riemsdijk1999 for a detailed discussion on this construction type.)

1. (7)

   

There is nothing in the noun vaktmästare ‘janitor’ or professor ‘professor’ that would motivate the use of han ‘he’ instead of hon ‘she’ or vice versa; these pronouns ‘reflect properties of the referent, as viewed by a beholder’, as formulated above.Footnote ⁸ Enger (Reference Enger2013:296) points out that the gender marker slot in question could not exist in Norwegian, since Norwegian does not have this particular construction (hon/han + pre-nominal determiner). However, in Norwegian, ‘gender markers’ are possible too, but without the prenominal determiner: ho nye professoren (3.sg.fem new.def professor.def) ‘she/the new professor’, and also ho professoren (without an adjective) – like in Swedish.Footnote ^9,Footnote ¹⁰ This shows that the slot in question is there in Norwegian too.

Secondly, one of the most important points in Josefsson (Reference Josefsson2013:76ff.; Reference Josefsson2014) is that there is another indication of a prenominal position in the functional sequence of the noun where semantic gender is expressed. However, this may not be so clear in Swedish as in other varieties of Mainland Scandinavian. What Josefsson refers to here is the use of the prenominal element det, which is obligatory in West Jutlandic, with a substance reading, as in det mælk (n milk) ‘milk’, and also found in East Jutlandic in the same type of construction. This construction type seems to be creeping into spoken standard Danish too, regardless of the formal gender of the head noun, according to Arboe (Reference Arboe, Hovmark, Sletten and Gudiksen2009), who reports expressions such as det regn (n rain(c)) ‘rain’ and det musik (n music(c)) ‘music’, as heard on Danish radio. This use of neuter det is definitely semantic, and the det in question also clearly a part of the noun phrase. The pronouns hon ‘she’ and han ‘he’ in (7) presumably occupy the same position as det in det mælk (n milk(c)) ‘milk’, det regn (n rain(c)), and det musik (n music(c)) ‘music’.Footnote ¹¹ What I have claimed is that Swedish, and presumably also Norwegian, has a null version of det in basically the same position as han ‘he’ and hon ‘she’ in (7), as well as det in det mælk, with the same meaning as an overt det, and also triggering agreement on the predicative, just like an overt det. Unfortunately, this null element cannot be seen, thus cannot be observed empirically, which is what Enger (Reference Enger2013:296) asks for. In addition, Swedish (and maybe also Norwegian) has overt non-pronominal classifiers or a classifier-like element. In these cases, the classifiers, not the head nouns, trigger agreement on predicative adjectives, as the examples below in (8) show:

1. (8)

   

Turning to pancake sentences, again, for obvious reasons, it is impossible to provide empirical proof of the existence of a null det heading the subject, for instance in pancake sentences, such as (5) above, for convenience repeated here:

1. (5)

   

However, the idea that there is a null element, a null version of det in (5), would provide an explanation for the semantic interpretation, for the observed agreement patterns, and for the blocking of canonical agreement on a predicative adjective. Note, too, that there seems to be complimentary distribution when it comes to the expression of neuter gender on pancake sentences. According to Ringgaard (Reference Ringgaard1971:31), there is no agreement on adjectives in West Jutlandic. However, as pointed out, the dialect allows for a neuter det (or another neuter element, such as noget ‘some’) to precede the head noun. In Swedish there is no overt det, but instead there is agreement on the adjective. The different sites for the overt expression of gender are illustrated in (9a) and (9b):Footnote ¹²

1. (9)

   

Intuitively, it makes sense that gender can be marked either on the subject or on the predicative, or in both places; different varieties of MSc makes different choices for different configurations. (This would work much like definiteness, where Swedish has double marking of definiteness on nouns, whereas Danish has single marking.)

Swedish does not have the det + noun construction, but, interestingly enough, the possibility of using neuter något (some.n) + noun, also with common gender nouns, such as those in (10).Footnote ¹³ The examples below are from the Internet:

1. (10)

   

The examples in (10) are very similar to parallel Danish cases, where the neuter pronoun noget (some.n) ‘some’ + noun can be used regardless of the formal gender of the noun, see Arboe (Reference Arboe, Hovmark, Sletten and Gudiksen2009).

I share the view in Enger (Reference Enger2013:294) that null elements should not be postulated easily. One principle restricting null elements is that it should be possible to replace the null element with an overt one. It might well be that it is not possible to use an overt det in Swedish or Norwegian, however, other lexical elements or pronouns may show up in the position in question, as shown by the examples in (7), (8) and (10) above. Furthermore, since Rizzi (Reference Rizzi1986) it is generally accepted that agreement may identify a null element, pro. I am convinced that not only generative grammarians agree with Rizzi's conclusion, even though the insight could be formulated differently in other theoretical frameworks; insofar as one accepts that languages such as Italian have null subjects, it should not be too difficult to accept that this subject is identified by agreement on the verb.Footnote ¹⁴ And if one accepts that verbal agreement may identify null subjects, it should not be so strange to assume that adjectival agreement too could identify a non-overt subject.

The Jutlandic gender system and the det + noun construction has developed in a historical context where inflection eroded (Skautrup Reference Skautrup1968, vol. IV:127ff.). In view of this, it is not unexpected that the det + noun construction, where det replaces neuter predicate agreement, is spreading in modern Danish, a language that is in the process of losing a great deal of its remaining inflection.

To conclude: contrary to what is claimed in Enger (Reference Enger2013), Josefsson shows that there is at least one designated position within the NP/DP that expresses semantics.

3. THE SUBJECT OF PANCAKE SENTENCES

The discussion about pancake sentences revolves around the question of the nature of the subject of the sentences, as well as how agreement works. Enger (Reference Enger2013:278f.) claims that the properties of pancake sentences can be explained by Corbett's Agreement Hierarchy (Corbett Reference Corbett2006:207, as presented in Enger Reference Enger2013:279):

1. (11)

   

According to Enger (Reference Enger2013:289–290), the figure in (11), taken together with the conclusion that the subject of pancake sentences are nouns with a low degree of individuation, explains the apparent ‘disagreement’ in pancake sentences. In his 2013 paper, the author refrains from commenting on the fact that not all subjects of pancake sentences can be analyzed as simple nouns/noun phrases. As pointed out in Josefsson (Reference Josefsson2006, Reference Josefsson2009, 2013, 2014), the possibility of adding adverbial modifiers to these ‘noun phrase subjects’ is what motivates Teleman, Hellberg & Andersson (Reference Teleman, Hellberg and Andersson1999, vol. 3:702f.) to treat pancake sentences as biphrasal clause equivalents (‘tvåledad nominal satsförkortning’).Footnote ¹⁵ In order to show that the subject of pancake sentences is or could be larger than a simple noun phrase, let us first consider the fact that Mainland Scandinavian is a V2 language. The reason why the sentences in (12), which are not pancake sentences, are ungrammatical is presumably that they violate the V2 criterion:Footnote ¹⁶

1. (12)

   

The corresponding pancake sentences are impeccable:

1. (13)

   

The possibility of having what indisputably are adverbial modifiers in the subject of pancake sentences, as shown in (13), indicates that the subject, at least in these two examples, has some kind of clausal properties and, consequently, is larger than would seem at first glance. The most straightforward solution would be to assume that the two preverbal constituents in (13) are embedded in a larger structure, in certain ways akin to VP-topicalization. As the translation shows, the subjects have a propositional interpretation: ‘to have two lovers every night’ for (13a) and ‘to have pancakes in the morning’ for (13b). It is not a coincidence that the translation contains the light verb have; this is frequently the case for pancake sentences. In fact, this ‘ha’ meaning is presumably what lies behind the somewhat cryptic formulation in Enger (Reference Enger2013:280) that Josefsson (Reference Josefsson2009) assumes that ‘many of the problems may be solved by using Butt's (Reference Butt2003, Reference Butt, Amberber, Baker and Harvey2010) concept of “light verbs”’.Footnote ¹⁷

More evidence that the subject of pancake sentences can be larger than would seem at first glance is that such a subject may contain reflexives, as in (14).

1. (14)

   

Assuming that reflexives have to be bound by some kind of subject (for example pro, PRO, a trace, or an operator), we may conclude that the subject of pancake sentence can be clausal in some sense, at least in some of the cases, and that it may contain a notion of a subject. Note too that it is possible to have what looks like a small clause as subjects, as in (15), repeated from (5c) above:

1. (15)

   

The pronoun henne ‘her’ in (15) is in the accusative case. As opposed to Danish and English, it is never possible to have accusative case pronouns as subjects in Swedish. This too indicates that the subject is larger than we see, and that it may contain some kind of non-overt case assigner. A subject, such as the one in (15), is particularly problematic to the approach in Enger (Reference Enger2013), which states that the subjects of pancake sentences are nouns with a low degree of individuation (page 292). Henne is a pronoun, and thus, according to Josefsson (Reference Josefsson2006:1363), devoid of formal gender features; it is not common gender, but definitely not neuter either. Furthermore, henne is both specific and definite – it is hard to see how it could be interpreted as having low degree of individuation, or, as Enger (Reference Enger2013:290) puts it: ‘Nouns that have a very general meaning trigger “pancake agreement”’.Footnote ¹⁸

The data that indicate that the subject of pancake sentences may have clausal properties is discussed at length in Josefsson (Reference Josefsson2006, Reference Josefsson2009, Reference Josefsson2012b, Reference Josefsson2013, Reference Josefsson2014), but Enger (Reference Enger2013) chooses not to include such data in his analysis. It clearly shows, however, that one of Enger's (Reference Enger2013) conclusions – that the subjects of pancake sentences are nouns denoting an entity with low degree of individuation – cannot explain the full set of data.

Finally, a note on Corbett's Agreement Hierarchy. Enger (Reference Enger2013) claims that this hierarchy explains the odd agreement on pancake sentences. What he calls referential (or semantic) agreement (page 290) will ‘rise monotonically towards the right [of the scale in (11)]’ (see Corbett Reference Corbett2006:207). There is no reason to doubt Corbett's observation from a typological ‘helicopter’ perspective, or from a diachronic perspective. However, it is hard to understand how this would explain the agreement pattern on pancake sentences. Josefsson (Reference Josefsson2009) suggests that predicative agreement is a relation that holds between a subject and an adjective. Agreement is morpho-syntactic, which means that agreement on the adjective reflects the features of its subject. In such a framework, there is only one way for agreement to end up on the predicative adjective. The ‘cost’ of this analysis is the assumption of a null det, which Josefsson (Reference Josefsson2013:60ff.; Reference Josefsson2014) suggests is a classifier, in all relevant aspects similar to other classifiers in (7), (8), and (10) above. Enger (Reference Enger2013) rejects the idea of the null det, but the cost of his analysis is that he has to assume two ways in which agreement may appear on an adjective, first of all by morpho-syntactic agreement in the canonical way (agreement sharing between a subject and a predicative adjective), and secondly by semantics being transformed into morpho-syntactic features, which determines overt agreement. Which of the analyses that best captures the data is a question that can be discussed, but Occam's razor would seem to argue against the solution proposed in Enger (Reference Enger2013).Footnote ¹⁹

4. CONCLUDING REMARKS

The main idea of Enger (Reference Enger2013:292) is that the subject of pancakes sentences is a noun with a low degree of individuation, and that this is why agreement in neuter is triggered. However, such a solution disregards evidence showing that pancakes sentences can in some sense be clausal, even have small clause subjects consisting of a personal pronoun in the accusative case, such as henne ‘her’ and a PP in (15). The semantic notion of low degree of individuation in Enger (Reference Enger2013) is captured in Josefsson (Reference Josefsson2009, Reference Josefsson2013, Reference Josefsson2014) by the assumption that the assumed null det, heading the subject of pancake sentences, lacks a number feature, just like the corresponding overt version of det. The same holds for clauses. The semantic interpretation of the absence of number is non-countability.

Enger (Reference Enger2013:294f.) dismisses a ‘light verb analysis’. Importantly though, the core meaning of the ‘light verb analysis’ is that the noun/noun phrase is generally interpreted as an object of the light verb have, used in a general and broad sense. A restricted set of other interpretations are available – corresponding roughly to the set of light verbs that have been suggested independently. A light verb analysis would account for the accusative case on ‘small clause’ subjects of pancake sentences, which would otherwise be hard to account for under Enger's (Reference Enger2013) analysis.

Enger (Reference Enger2013:283) assumes that Corbett's Agreement Hierarchy (Corbett Reference Corbett2006:207) can explain the properties of pancake sentences. In my view, the Agreement Hierarchy has nothing to say about pancake sentences per se. It seems to me that the point of disagreement first of all boils down to the question of the justification of postulating null elements, in this case a null version of det heading the subject of pancake sentences. The virtue of allowing this is that it restricts predicate agreement to a morpho-syntactic relation between a nominal element in the subject position and a predicative adjective. The cost is the assumption of a null element. The virtue of the approach in Enger (Reference Enger2013) is of course that we can do without a null det. However, the cost is that we need to assume two ways in which agreement features may end up on a predicate adjective: by morpho-syntactic agreement in the canonical way, namely subject–adjective agreement, which we need to assume in any case, and by so-called referential agreement, that is a referent in the world of discourse (the real world or the linguistic discourse) is interpreted in terms of morphosyntactic features that end up on the predicative adjective.

To clarify my point, let us consider the following situation: A person spots something undefined in front of her, which causes her to utter Det var ful-t! (3.n be.prs ugly-n) ‘It's ugly!’. The choice of the neuter pronoun det ‘it’ cannot be motivated by any agreement process – there is nothing with which det can agree morpho-syntactically. Det in this use is a pronoun with deictic reference; it is the default pronoun that we use when we talk about unspecified referents in the world, in particular when we do not want to assign them cognitive boundaries. The choice of the adjectival form ful-t ‘ugly’, on the other hand, is due to an agreement process; let us call it subject–adjective agreement. And in the case of pancake sentences it is obviously not the morpho-syntactic features of the overt noun phrase that give rise to t-agreement, but something else. All accounts of pancake sentences require some mechanism that blocks canonical agreement. Even in the theory proposed in Enger (Reference Enger2013), nouns such as pannkakor ‘pancakes’ should have gender and number. In many cases ‘ordinary agreement’ with the overt noun phrase in the subject position is simply not possible, so what is the blocker? For instance, why is *?Pannkakor är god-a (pancake(c).pl be.prs good-pl) out, or at least bad, whereas Hundvalpar är söt-a (puppy(c)-pl be.prs pretty-pl) ‘Puppies are pretty’ is impeccable. The same applies to *Tiger-n är brun-t (tiger(c)-c.def be.prs brown-n) with the intended generic meaning ‘The tiger is brown’ (compare Tiger-n är brun-Ø (tiger(c)-c.def be.prs brown-c) ‘The tiger is brown’). And why is *Tigr-ar är brun-t (tiger(c)-c.pl be.prs brown-n) not possible either? (This sentence should be compared to the well-formed example Tigr-ar är brun-a (tiger(c)-c.pl be.prs brown-pl) ‘Tigers are brown’.) It is difficult to understand in what sense hundvalpar ‘puppies’, tigern ‘the tiger’, or tigrar ‘tigers, used in a generic sense, should be more individuated, and therefore resisting t-agreement, than pannkakor ‘pancakes’. There has to be something more than just degree of individuation and Corbett's Agreement hierarchy at play here.

In light of the argumentation given above, I suggest that we abandon Corbett's (Reference Butt2006) idea of analyzing pronouns in general as instances of agreement. Some pronoun occurrences and pronoun–antecedent relations could indeed be thought of in such terms, in particular den and det in (2a) and (2b) above, but not the pronouns han, hon and det in (3), and definitely not den and det used as demonstrative pronouns, as in (4a) and (4b). In fact, my position could be stated even more strongly: The idea that pronouns in general are always instances of agreement has turned out to be a straightjacket when it comes to our understanding of this category, and only if we remove it, will we be able to achieve a more profound understanding of gender and pronouns – and pancake sentences.