Convergence and divergence in the expression of partitivity in French, Dutch, and German

1 Introduction

In French, partitivity, i. e., the relation between a part and a whole/set, can be expressed in various ways: by proper partitive constructions as in (1), by partitive articles as in (2), and by the partitive pronoun en. ^[1]

In this paper, we focus on examples with the pronoun en ‘of.them’, as in (3), but also on examples like (4), which contains a partitive pronoun but which is quantitative in interpretation, in the sense that there is no whole/set implied, in contrast to (3).

As the above examples show, the term partitive pronoun for elements like en is a misnomer. However, in the absence of a better alternative, we will use that term for presentational reasons. By extension, partitivity may refer to contexts/expressions with a partitive pronoun, although the notion can also be used in a broader sense, covering, e. g., also examples such as (1)-(2). The use of the French pronoun en, which is a clitic, has been extensively studied, both from a descriptive and a theoretical point of view (Milner 1976; Pollock 1998; Ihsane 2013). Ihsane (2013) claims that en can replace different portions of the noun phrase. In (3) en pronominalizes a smaller portion of the DP than in (5), which does not contain a quantifier:

According to Ihsane (2013), en replaces the complement of the numeral in (3), whereas it pronominalizes the whole complement des noisettes ‘(some) nuts’ in (5). The author models the conditions of the use of partitive en in a DP structure in which various uses of en correspond to different portions of the DP, headed by different functional projections.

In Dutch, the standard use of the partitive pronoun er, which is also called quantitative pronoun, is its combination with a quantified indefinite noun phrase in object position, like the French (4) ^[2]:

De Schutter (1992) observes that in some Dutch dialects, especially in Belgium, but not in standard Dutch, er can replace the whole indefinite noun phrase, just like French en in (5) ^[3]:

In contrast to some German dialects, standard German does not possess a partitive pronoun like en/er. However, Strobel (2016) shows that standard German uses the pronominalized adjectival pronoun welch- (which can also be used as an interrogative pronoun: welche Frau ‘which woman’) to express an undefined quantity ^[4]:

Welch- is one of the four strategies of partitive-anaphoric reference in the German language (Glaser 1993). ^[5] In this paper, we only discuss this strategy, as it is the one found in contemporary Standard German.

To our knowledge, a relation between French en or Dutch er and the expression of partitivity in German has never been made explicit or investigated in the linguistic literature. One of the goals of this paper is to fill this gap. The perspective is therefore crosslinguistic, but, because of the richness and the variation in the use of the French pronoun en, we will take the limitations and possibilities of this pronoun as our starting point. What we want to know is how the ways in which en is used in French are expressed i) in languages that also possess a partitive pronoun like en/er, and ii) in languages that do not possess such a partitive pronoun. For the first type of languages we study Dutch, and for the second type of languages we take German as a representative, although this does not necessarily mean that other languages behave exactly in the same way. More generally, this paper will contribute to the understanding of the expression of partitivity in Dutch and German, about which relatively little is known, in contrast to French.

Our methodology consists in using and analyzing equivalent Grammaticality Judgment Tasks submitted to native speakers of French, Dutch and German. We will model our results for Dutch and German in DP structures comparable to the one proposed for French by Ihsane (2013). This will enable us to compare the expression of partitivity in the three languages in a proper way.

This paper is structured as follows. In Section 2, we unfold the subtleties of the use of the partitive pronoun in French, and we present Ihsane’s (2013) theoretical modelling that accounts for the various possibilities of its use. In Section 3, we present our methodology and in Section 4 the results of the Grammaticality Judgment Tasks that we submitted to native speakers of French, Dutch and German. Our results are discussed in Section 5 and modelled in DP structures comparable to the one proposed by Ihsane (2013) for French in Section 6. In Section 7, we summarize our findings and make some suggestions for future research.

2 The use and syntactic analysis of the partitive pronoun in French

3 Methodology

4 Presentation of the results

In this section, we present the analysis of the data per category, as classified in Section 3.1. ^[14]

Figure 1:

Definite article + adjective.

(i) Definite article + adjective

Figure 1 shows that in the three languages an elliptical noun phrase introduced by a definite article and containing an adjective, as in (13) or (33), is accepted by the participants. It is, however, significantly less accepted in Dutch than in French (p = 0.011) or in German (p = 0.001). We will come back to this point in the discussion. If the sentence contains a partitive pronoun, it is not accepted by the native speakers of French (0%) nor by the native speakers of Dutch (cf. (13)) vs. ((34)) below) in the test:

(ii) Non-referential plural NPs

Figure 2 shows that with non-referential noun phrases the German participants accept welch- (cf. (8)) where in French en is accepted (cf. (14)). In this context the Dutch participants do not accept er (cf. (35)). In Dutch (cf. (35)) the use of the definite pronoun is accepted significantly more than in French (p = 0.001) or in German (p = 0.001). For Dutch and German, we also investigated whether the participants accepted the use of a full NP (cf. (35c)), which they do, as the figure shows.

Figure 2:

Non-referential plural NPs.

(iii) Referential plural indefinite NPs

En, er and welch- are not accepted with referential NPs, as Figure 3 shows, but significantly more in French (p = 0.002), cf. (15), or in German (p < 0.001), cf. (36), than in Dutch (cf. (37)). In one of the three sentences of the Dutch and the German test the NP (not tested for French) was preceded by a definite article. The use of a bare NP is less accepted than the use of a definite article, both in Dutch (p = 0.001) and in German (p < 0.001).

Figure 3:

Referential plural NPs.

(iv) Mass nouns

As for mass nouns, we see in Figure 4 that whereas the use of en and welch is accepted in French (cf. (18)) and German (cf. (39)), the use of er is not in Dutch (cf. (38)). The use of a definite pronoun is more marginally accepted in German than in Dutch. This difference is significant (p < 0.001). In French, it is not accepted (0%). The acceptance of the use of NPs was only investigated for Dutch and German (cf. (30)). NPs are accepted for both languages.

Figure 4:

Mass nouns.

(v) Quantified elliptical object NPs

As illustrated in Figure 5, the use of the partitive pronoun is accepted both in Dutch (cf. (6)), and in French (cf. (11)). The omission of the partitive pronoun is, however, rejected in French (0% acceptance), as it is in Dutch (15%), although this difference is significant (p = 0.041). In German, welch- cannot be used in this context (not tested), but the omission of the noun is accepted only in 63% of the cases (cf. (40)), which is, however, significantly more than in French (p < 0.001) and in Dutch (p < 0.001).

Figure 5:

Quantified object NPs.

(vi) Indefinite article + adjective

The use of the partitive pronoun in combination with an indefinite elliptical NP + adjective is accepted by the French participants (cf. (12)), see Figure 6. By the Dutch participants, the omission of the partitive pronoun is preferred (cf. (41)). This also holds for the native speakers of German (cf. (42)).

(vii) Negated indefinite singular and plural NPs

In Figure 7, we see that with negated singular elliptical NPs the native speakers of French, cf. (16), and German, (cf. (44)), accept the use of en resp. welch-, but that the Dutch participants strongly prefer the definite pronoun (cf. (43)). The German native speakers marginally accept the use of the definite pronoun. The French native speakers do not accept it at all (0%). The acceptance of the use of an NP was investigated for Dutch and German, with a positive result. For German, the acceptance of the negator kein- was furthermore tested and yielded a positive outcome (cf. (32)).

Figure 6:

Indefinite article + adjective.

Figure 7:

Negated singular elliptical NPs.

A similar result is provided in Figure 8 for negated plural NPs, but the acceptance of welch- in this context is significantly lower than the acceptance of en (p < 0.001) and also than the acceptance of welch- with negated singular NPs (p < 0.001). ^[15] We will come back to this point in the discussion section.

(viii) Quantitative or partitive interpretation

Figure 8:

Negated plural elliptical NPs.

According to Figure 9, the quantitative interpretation is accepted by the participants for both en (cf. (9b)) and er (cf. (45b)) below), but also, although to a minor degree, in German, which does not use welch- in this context, the distinction is made (47). This also holds for the partitive interpretation (cf. (9a), (45a)), even though in both Dutch (cf. (31) and (45a)) and German (cf. (46)) the partitive interpretation can also be expressed by a pronominal adverb: ervan/davon. German does not significantly differ from French or Dutch in these three contexts.

Figure 9:

Quantitative or partitive interpretation.

5 Discussion

Besides a crosslinguistic comparison of the possibilities to express partitivity in the three languages under study, our test has allowed us to compare alternative constructions within the languages. Furthermore, we have not provided the intuitions of one native speaker per language, as is often done in the linguistic literature (often these are the author’s intuitions) nor have we simply copied judgments provided in grammars or the linguistic literature. For French we have checked judgments reported in the literature by means of a Grammaticality Judgment Task. For Dutch and German, we have provided very few judgments from the literature beforehand, the reason being that the 8 categories distinguished in Section 3 are barely discussed in Dutch and German grammars and literature. A Grammaticality Judgment Task is an appropriate tool to explore native speakers’ intuitions (Blume and Lust 2016) about the possibilities within these categories, which we distinguished on the basis of what has been described for French. Familiarity with this kind of tests or some metalinguistic knowledge may have influenced the results. The Dutch informants were undergraduate and graduate students of French, whereas most of the German participants did not have any notion of linguistics. The French informants were mainly, but not exclusively, linguists. Hereafter we discuss our results. For the reader’s convenience, we first represent the results presented in the previous section in Figures 10, 11 and 12 per language.

Figure 10:

French data.

Category (i): partitive pronoun 0%; (iv) definite pronoun 0%; (v) empty noun 0%; (vi) empty noun 0%; (viia) definite pronoun 0%; (viib) definite pronoun 0%.

Figure 11:

Dutch data.

Figure 12:

German data.

The context of definite article plus adjective (i) disallows the use of the partitive pronoun; an elided noun is preferred. ^[16] The lower acceptance of the omission of the noun in Dutch is due to the relatively low percentage of acceptance of one sentence (44%), which may be due to another reason than the omission of the noun.

With non-referential plural noun phrases (ii), standard Dutch does not accept er, in contrast to Dutch dialects (cf. (7)), but prefers a definite pronoun, significantly more than German (37,5%) and French (25%).

With referential noun phrases (iii) the use of a definite pronoun or a DP is preferred in the languages under consideration. The use of an NP is less accepted with referential noun phrases than with non-referential noun phrases. In Dutch the use of er is rejected, but the use of en in French and welch- in German is marginally accepted by the native speakers. The reason for this might be that some contexts did not completely exclude the use of a non-referential interpretation. Although the participants were categorical in rejecting (29) and (36) – 0% acceptance for French and 2,5% acceptance for German – the other two contexts may also have favored a non-referential interpretation. In one of the contexts en was accepted in 37,5% of the cases and welch- in 64%; in the other context en was accepted in 50% of the cases and welch- in 37,5%. ^[17] Notice however that 13 of the 40 German participants made the referential versus non-referential distinction in all contexts by accepting welch- in the 3 non-referential sentences and a definite pronoun in the 3 referential sentences. The significantly lower acceptance rate of er in Dutch with respect to en and welch- in referential contexts is due to the fact that er is not accepted in non-referential contexts either. The interpretation of a referential context as a non-referential context will not make the use of er more acceptable.

The results for mass nouns (iv) are the same as for non-referential plural noun phrases, but the use of a definite pronoun in Dutch is more accepted with mass nouns (81%) than with non-referential plural NPs (62%). For German the acceptance rate of a definite pronoun for mass nouns (36%) is comparable to the one for non-referential plural nouns (37,5%) and lower than for Dutch (probably because German offers the alternative welch-), but interestingly, we observed a difference among the singular definite pronouns: for es ‘it’ we observed 2,5% of acceptance, for ihn ‘him’ 40% and for sie ‘her’ 65%.

With quantified noun phrases (v) in French and Dutch the partitive pronoun is required according to prescriptive grammar. This is confirmed by our data. In German the omission of the noun is accepted in only 63% of the cases. If we take into consideration the omission of the noun in other categories under study, the percentage increases. In category (viii) the omission of the noun is accepted for German in 74% of the cases in the quantitative interpretation and in 82% in the partitive interpretation. In category (vii) the acceptance of the omission of the noun with the negator kein- is 86%. For the three categories the acceptance of the omission of the noun is on average 77%. The omission of the partitive pronoun is accepted in 15% of the cases in Dutch, which is a significant difference with French (0%). This may be due to the choice of the quantifier in two of the sentences, because in the third sentence (with the negator geen ‘no’) there was 0% of acceptance. ^[18]

Whereas in category (vi) for French the use of en with an indefinite article + adjective is accepted, its omission is not completely rejected either. This might be due to the fact that in spoken French, the omission of en is accepted in this context by some speakers of French. Dutch and German accept the simple omission of the noun without a supplier. For Dutch a rate of 29% of acceptance for the use of er with an indefinite article + adjective is somewhat surprising. In some Dutch dialects, the use of er with an indefinite determiner + adjective is, however, accepted (Sleeman 1998; Kranendonk 2010). Although our participants were probably not speakers of these dialects, the combination of er and indefinite article + adjective may not have shocked some of the participants for this reason. Out of the 23 Dutch participants, 6 accepted 2 or 3 sentences with er. That 7 participants accepted er in one of the sentences may also be due to other reasons. The small word er may not have been noticed when reading the sentence or it may have been interpreted as a locative er, although we added an explicit locative modifier to the test sentences in order to prevent a locative interpretation of er (cf. fn. 2).

With negated nouns (vii), the three languages behave as in the case of non-referential nouns and mass nouns. In German, however, welch- is accepted significantly less with plural NPs (71%) than with singular NPs (92%). ^[19] With the negated plural NPs, the use of a definite pronoun is accepted in 45% of the cases. This percentage comes close to the percentage of acceptance of the use of the definite pronoun with non-referential plural NPs (37,5%). In both cases the definite pronoun is the plural form sie. In the singular, there is a difference between the acceptance rate of a definite pronoun with negated indefinite singular NPs (12%) and mass nouns (36%). In the test sentences with a singular negated NP, the pronoun was twice ihn ‘him’ (22,5% and 11% of acceptance respectively), and once the neutral pronoun es ‘it’ (0% of acceptance). The rate of acceptance of sentences with ihn is lower with a negated indefinite singular NP than with a mass noun. With mass nouns, there was one sentence containing ihn (40% of acceptance). With mass nouns, the test sentence containing es was accepted in 2,5% of the cases and the one containing sie, i. e., a feminine singular pronoun, in 65% of the cases. The differences in the acceptance rates of the different forms of the definite pronouns are intriguing. In future work, we will investigate in more detail the role of gender and number in our results.

For Dutch, the definite pronoun was accepted to an equal rate with singular (mass, negated mass noun) and plural (non-referential, negated) elliptical NPs: 77% and 76%. In the singular we always used the neutral pronoun het ‘it’, because, according to Audring (2009), in present-day Dutch common, i. e., non-neuter, mass nouns are replaced by het and not by the masculine pronoun hem ‘him’ or the feminine pronoun haar ‘her’. In our test, het replaced twice a neuter noun and four times a common noun. The percentages of acceptance were comparable: 76% and 78% respectively, which corroborates Audring’s claim.

In the three languages a distinction between a quantitative and a partitive interpretation is made (viii). The relatively low percentage for the acceptance of the quantitative interpretation in German may be due to the results of one sentence, which was only accepted by 58% of the participants. ^[20]

In the next section, we make use of a broader classification to compare our results for the three languages.

6 Modelling the results

In Section 3, we grouped our subcategories (i–viii) into four broad categories:

A. The relation between partitivity and indefiniteness (i).

B. The relation between partitivity and non-referentiality (ii–iv).

C. The relation between partitivity and quantification (v–vi).

D. The distinction between a quantitative and a partitive reading (viii).

Our results show that in all three languages a distinction is made between partitivity, in the broad sense as used in this paper, and indefiniteness (A): en, er and welch- are only used in indefinite contexts.

In addition, the results show that there is a relation between partitivity and non-referentiality (B), at least for French and German: with non-referential plural nouns and with mass nouns en resp. welch- are preferred. Standard Dutch does not make the distinction: the use of the definite pronoun is preferred with both referential and non-referential noun phrases like kinderen ‘children’ (48), which is ambiguous between the two readings.

In French, the use of the definite pronoun is not or marginally accepted with non-referential noun phrases and with mass nouns. This is confirmed by the results of the negated noun phrases, which are non-referential. In terms of Ihsane’s (2013) DP structure presented in Section 2.2, this means that the distinction of a non-referential portion of the DP can be overtly expressed in French (en) and in German (welch-):

French and German distinguish a non-referential interpretation from a referential interpretation by using a definite pronoun in the latter case:

As for Dutch, it is similar to French and German regarding the use of a definite pronoun for the referential interpretation, suggesting that the referential portion of the noun phrase can be expressed by a definite pronoun as in (50). ^[21] The question that arises is whether Dutch also has a structure like (49) for non-referential constituents: indeed, since in such examples it is ze that is used, it is not obvious that Dutch has a specific structure for non-referential constituents. However, as kinderen ‘children’ in (48) can be both non-referential and referential, it shows that Dutch does make a semantic distinction between non-referential and referential noun phrases. Therefore, we propose that the structure of the former is as in (49), even if the distinction referential/non-referential is not made by means of different pronouns. These cases would be examples of syncretism where an element, like ze, is used in different contexts and represents different structures. ^[22] The fact that er is not accepted in non-referential constructions may explain the broad acceptance of definite pronouns in Dutch. ^[23]

Both French and Dutch use a partitive pronoun with a quantifier (C). The corresponding sentences in German have an elided noun. Although the percentage of acceptance is relatively low with quantifiers, the rate of acceptance of an elided noun with the quantifier kein- and in a quantitative or partitive interpretation shows that elision of the noun is a way to express partitivity in German, in the absence of a partitive pronoun. The structure would be as in (51), where FP_de is restricted to French (in Dutch er replaces the projection selected by the quantifier):

With negated constituents, the results differ from (51): welch- in German patterns with en in French and is accepted, whereas er in Dutch is rejected, as for context (ii). This is not surprising since these constructions are also non-referential (cf. (49)). These results support the fact that en in French depends on non-referentiality rather than on the quantificational character of the construction or on the underlying presence of de ‘of’ and suggests that it is also the case of welch- but not of er. What seems crucial for the use of er is the presence of a quantifier.

With classifying adjectives, er is often not accepted, suggesting that, in Dutch, classifying adjectives are not as quantifying as quantifiers; otherwise er should be fully accepted since a quantifier seems to be needed for er to be possible. In addition, Dutch allows nominal ellipsis with adjectives, like German. In both Dutch and German, nominal ellipsis depends on the inflectional properties of the adjective (e. g., Muysken and van Riemsdijk 1986; Kester 1996), whereas in French nominal ellipsis with or without en depends on the partitive interpretation of the adjective (Sleeman 1996; Bouchard 2002). In Dutch and German, the FP_de is not projected.

In structure (51), both French en and Dutch er can have a quantitative interpretation and a partitive interpretation (D). Even if in German there is no pronoun, both interpretations are accepted as well.

The above discussion shows that welch- in Standard German is found in the counterparts of the non-referential complements with the partitive article in French. The constructions in which welch- is found are all complete non-referential nominal phrases. In other words, welch- cannot replace subparts of the nominal phrase. Er, in contrast, only replaces a portion of the nominal structure, and this only with a quantified element. As for en, it is accepted in both contexts, and also with classifying adjectives.

7 Conclusion

In this paper, we have studied the notion of partitivity in a specific way: we have focused on the so-called partitive pronoun en in French and studied eight contexts in which en is accepted or disallowed. Our objective was to find out how other languages, with a similar partitive pronoun like Dutch, or without such a pronoun like German, behave in analogous contexts and to formalize the results in the model developed in Ihsane (2013) in which the partitive pronoun can replace different portions of the nominal structure.

This new crosslinguistic study is based on the results of a Grammaticality Judgment Test taken by native speakers of French, Dutch and German. Our overall finding is that the subtleties of the French partitive pronoun en can also be expressed in Dutch and German, but in partially different ways. Our investigations show that German, which does not have a partitive pronoun such as en or er, often uses no lexical item in the contexts studied, i. e., ⊘, where these overt pronouns are found, but that welch- partially has the same function as en, which could mean that standard German does have a partitive pronoun after all. Our results also show that (non)-referentiality correlates with partitivity: non-referentiality can be expressed by a pronoun, both in German and in French, with welch- and en respectively, although the former is much more restricted in the expression of partitivity than the latter. Dutch on the other hand does not explicitly make a distinction between non-referential and referential noun phrases, at least in its standard. Er cannot be used with non-referential noun phrases. However, our results suggest that definite pronouns are relatively more accepted with referential noun phrases and the repetition of the NP more with non-referential noun phrases. Although German has welch- to express non-referentiality versus referentiality, we found the same tendency for definite pronouns and NPs in German as in Dutch, although to a lower degree. We also observed that Dutch er and the simple omission of the noun in German both allow a quantitative and a partitive interpretation, just like French en. Another finding, on the other hand, is that classifying adjectives allow a partitive pronoun only in French, in contrast to Dutch. In Dutch and German, an elided noun is acceptable, but not in standard French.

In terms of structure, building on Ihsane (2013), we have proposed that different layers of the nominal structure correspond to different elements in the languages under study. This can be summarized as follows:

An interesting finding of this paper that deserves more research is the different acceptance rate of the various definite pronouns in non-referential contexts in German. For Dutch it might be furthermore investigated if, in regional varieties, the use of er in non-referential contexts is accepted.