Direct pseudo-partitives in US English

1 Introduction

Whether or not English uses direct pseudo-partitives – for example, cup water – as opposed to indirect pseudo-partitive constructions – for example, cup of water – is a matter of debate: some say direct pseudo-partitives are nonexistent (e.g., van Riemsdijk 1998: 18), others say their use is marginal (e.g., Koptjevskaja-Tamm 2001: 550), and others simply state that English has direct pseudo-partitives (e.g., Jackendoff 1977: 138), as if their use is common and uncontroversial. In addition, syntactic analyses of pseudo-partitives differ in how to treat of (e.g., Hankamer and Mikkelsen 2008; Jackendoff 1977; Rothstein 2009; Selkirk 1977). We present corpus evidence of direct pseudo-partitive constructions in American English, showing that while their frequency is low in English, they are not necessarily as marginal as previously assumed. Moreover, while indirect pseudo-partitives in English are often ambiguous between container+contents readings, where both the container (e.g., a cup) and its contents (e.g., water) are referred to, and measure/portion readings, where just the amount of, for example, water is referred to (though this is a simplification; see, e.g., Sutton and Filip 2021, among others), we show that direct pseudo-partitives in English are more frequently associated with measure/portion readings than container+contents readings. We show how this can straightforwardly be accounted for by theories that assume distinct syntactic structures for container+contents and measure/portion readings (e.g., Selkirk 1977: 310–313).

2 Background

Most Germanic languages juxtapose nouns for pseudo-partitives – for example, Gläser Wein (lit. ‘glasses wine’, German) – and English is counted among languages that rarely, if ever, do this (Koptjevskaja-Tamm 2001; van Riemsdijk 1998; Wood 2020, among others). However, it has been noted (e.g., Jackendoff 1977: 138) that direct pseudo-partitives do occur in English, as in (1), particularly in cookbooks (Landman 2004: 237).

English direct pseudo-partitives are often used to justify assuming of serves a functional purpose (e.g., case marking) rather than carrying lexical meaning such as a relation between the two referents surrounding it (e.g., Jackendoff 1977: 138). This assumption is made in many analyses of English pseudo-partitives (e.g., Alexiadou et al. 2007; Borer 2005; Grestenberger 2015; Rothstein 2009; Stickney 2009), which are reviewed below and compared to Danish and German.

Selkirk (1977) argues that pseudo-partitives can take two forms, one of which contains a preposition phrase headed by of. She argues that extraposition is possible for the preposition phrase of a pseudo-partitive when it is the complement of the unit-specifying head noun, as in (2); extraposition is not possible when the head noun specifies that which is measured by a measure noun, as in (3).

(2)

a.

A review of (certain) answers to your argument was given. (Selkirk 1977: 309)

b.

noun complement (Selkirk 1977: 313)

(3)

a.

A number of answers to your argument was given. (Selkirk 1977: 309)

∗ ˗˗˗˗˗˗˗˗˗˗˗˗˗˗˗˗˗˗˗→

b.

pseudo-partitive (Selkirk 1977: 313)

Selkirk (1977) also argues these two structures underlie the difference between container+contents readings and measure/portion readings. In (4a), the container+contents reading is specified in that it is the cup that is smashed on the floor, so cup must be the head noun and therefore have the structure in (2). In (4b), the measure/portion reading is specified in that sugar is what is strewn on the floor, so sugar must be the head noun and have the structure in (3).

Selkirk (1977) therefore proposes that, while some English pseudo-partitives are genuine, others have noun complement structure. Further evidence for distinct structures includes subject-verb agreement from the Great Britain section of the International Corpus of English (Greenbaum 1996): in (5a), the verb, were, agrees with the second noun (N₂), samples, rather than the first noun (N₁), number, while in (5b), the verb, was, agrees with N₁, number.

Keizer (2007) provides further evidence for distinct structures via alternations in subject-verb agreement between N₁s and N₂s in pseudo-partitives in British English (see also Fernández-Pena 2020, among others). We provide corresponding data from US English nouns that can refer to containers in (6), from the Corpus of Contemporary American English (Davies 2008–): in (6a), the verb appear agrees with N₁, the container cups, while in (6b), the verb prevents agrees with N₂, the measured entity coffee.

In contrast, for both container+contents and measure/portion readings, Borer (2005) assumes the basic structure in (7), where #P is a nominalizing quantity phrase and “the merger location of all other determiners, including strong determiners, weak determiners, cardinals, and … the definite article” (Borer 2005: 59). Thus, two bottles of milk has the structure in (8), where CL, the classifier phrase, generates countable structures, and mass structures exist in its absence.

Because container+contents readings can take adjectival modification and certain modifiers before N₂, as in (9), while measure readings cannot, as seen in (10), Borer (2005) assumes N₁ is fully lexical (an L-head) in container readings but a “quasi-functional L-head” in measure readings (Borer 2005: 258).

This sort of analysis is found also in Alexiadou et al. (2007) and Stickney (2009). While Stickney (2009) focuses on adjectival modification of N₁ and N₂ as the reason for two structures, rather than container and measure readings as such, she does assume that of heads a functional node, FP, for both structures. However, in the case of semi-lexical N₁s (measure readings in Borer 2005), the FP head can be null, as in (11).

Since Borer (2005), most analyses seem to argue for minimal differences between readings of English pseudo-partitives. Following Landman (2004: 12–13), Rothstein (2009: 112–113) characterizes the differences between container+contents readings and measure/portion readings in terms of counting versus measuring, with similar syntax to Selkirk (1977). However, Landman (2004) presents no evidence for the distinct structures in English, and Rothstein (2009) argues available evidence is weak: she reports that, for at least some speakers of English, a sentence in which one attempts to have a modifier scope over a container+contents construction, such as (12a), is infelicitous, while it is felicitous to have a modifier scope over a measure/portion reading, as in (12b) and (12c).

Rothstein (2009) also shows it is possible to modify a container, as it is a nominal head, but not a measure unit.

Given judgments are not strong for the aforementioned sentences, Rothstein (2009) assumes this evidence is insufficient to justify distinct structures.

In contrast to English, recent analyses of Danish and German motivate distinct structures. Hankamer and Mikkelsen (2008) analyze direct pseudo-partitives – for example, en gruppe turister (lit. ‘a group tourists’) – and indirect ones – for example, en gruppe af turister (lit. ‘a group of tourists’) – in Danish with distinct structures; direct without a fully lexicalized first noun, and indirect with a PP complement to a DP with a fully lexical first noun. This is done to explain why the N₁ of indirect pseudo-partitives can take the definite affix, while that of direct pseudo-partitives cannot. Grestenberger (2015) has shown that, in Viennese German, container+contents readings and measure/portion readings are distinguished in that the former marks the first noun plural – for example, zwei Gläser Wein (lit. ‘two glasses wine’) – while the latter do not – for example, zwei Glas Wein (lit. ‘two glass wine’). To account for the two different forms, Grestenberger (2015) assumes distinct structures underlie the two readings in German. For measure/portion readings, N₁ is the head of a functional category #P (NumP) that takes an NP complement which contains N₂, as seen in (14). It is because N₁ is the head of #P rather than NP that it cannot take number agreement – that is, it is a measure word rather than a noun in this context.

(14)

(Grestenberger 2015: 102)

For container readings, N₁ is the head of an NP with a φP (functional phrase) complement, which has N₂ in an NP daughter, as in (15). Because N₁ heads an NP, it does indeed take number agreement.

(15)

(Grestenberger 2015: 103)

For English, Grestenberger (2015) does not find Stickney’s arguments for two structures convincing. Even assuming that recursion is the best syntactic argument for assuming different categories and that the recursive measure/portion readings in (16) are awkward (compare (17)), Grestenberger (2015) proposes that this awkwardness is because N₁s in measure/portion readings are inherently relational, and there may be general restrictions on recursion with such nouns.

Grestenberger (2015) therefore assumes a single structure for English, detailed in (18), and two different nominal semantics for nouns like glass, one referring to a container, the other to a measured amount.

(18)

(Grestenberger 2015: 125)

An anonymous reviewer points out that additional evidence of distinct structures in related languages might be found in Romance languages. For example, in French, “the partitive pronoun en seems only compatible with the amount interpretation (Il en a bu un verre ‘He drank a glass (of it)’), but not with the container reading (*Il en a posé un verre sur la table ‘He has put a glass (of it) on the table’)”.

We highlight three takeaways from this discussion: (i) the status of direct pseudo-partitives like two cups water in English is unclear in structural terms and unknown in quantitative terms; (ii) container+contents and measure/portion readings are assumed to be minimally different – that is, they differ in N₁ being fully lexical/referential or not; and (iii) the preposition of in pseudo-partitives is semantically empty, though possibly functional in some way such as a marker of case. At the same time, there is nascent evidence that (i) direct pseudo-partitives constructions occur in limited – that is, cooking – contexts (Landman 2004; Stickney 2009) and (ii) certain English pseudo-partitives – for example, two parts water – are always direct (Jackendoff 1977). Moreover, while container+contents and measure/portion readings are often assumed not to have structures that differ very much (e.g., Borer 2005), some have justified a distinction between two structures (e.g., Selkirk 1977).

3 Data

We tested claims about the extent to which direct pseudo-partitives occur in English with a corpus study. While direct pseudo-partitives were rare in Middle English (Wood 2020), we show they occur with a frequency suggestive of productivity in present-day US English. We also show that there is no positive evidence of container+classifier readings among direct pseudo-partitives.

The Corpus of Contemporary American English (COCA; Davies 2008–) was chosen because it is balanced by genre, making it less likely that forms are over- or underrepresented if they belong to a particular genre. The genres are TV/Movies, Spoken, Fiction, Magazine, Newspaper, Academic, Web-Blog, and Web-General, and there are an average of 125,361,219 words per genre (min. 119,505,305, max. 129,899,427). The transcription of the spoken section is considered to be accurate (no recording errors) by Davies (2008–).

First, a general search for “NOUN of NOUN” forms was done to see which pseudo-partitives are most frequent. Focusing specifically on pseudo-partitives in which the first noun can refer to a container and the second a substance, 14 were found in the 1,000 most frequent “NOUN of NOUN” results – for example, cup of coffee occurs 5,348 times. These are the first 14 listed in Table 1 (from CUP COFFEE to GLASS CHAMPAGNE). Second, “NOUN NOUN” constructions were searched, and only four container–substance forms were found in the top 1,000 “NOUN NOUN” constructions: cup water, cup sugar, teaspoon salt, and tsp salt. Note that, of the 14 indirect pseudo-partitives in the top 1,000 “NOUN of NOUN” constructions, only one (cup of water) had a direct form (cup water) in the top 1,000 “NOUN NOUN” forms. Independent searches were conducted for direct versions of the other 13 indirect pseudo-partitives not present in the top 1,000 “NOUN NOUN” forms – for example, cup coffee – to find their frequency and similarly for indirect pseudo-partitives corresponding to the other three frequent direct pseudo-partitives. While all of the “NOUN of NOUN” forms that were found and searched for were assumed to be indirect pseudo-partitives, the “NOUN NOUN” forms were manually inspected to be sure they were genuine direct pseudo-partitives rather than another form such as a proper noun – for example, Casual Cup coffee – yielding the actual number of direct pseudo-partitives for a given NOUN NOUN form, which is reported in Table 1. For the direct pseudo-partitives that occurred more than 100 times – that is, cup water, cup sugar, teaspoon salt, tsp salt – 100 random instances of each were analyzed to see what percentage were genuine direct pseudo-partitives, and this was used to approximate the total number of direct pseudo-partitives of each form – for example, cup water occurred 1,295 times; of 100 random instances, 98 were genuine direct pseudo-partitives, so it is assumed that 1,269 cup water occurrences are direct pseudo-partitives. Taking all of the pseudo-partitives together from these results, 24,611 (33%) are direct pseudo-partitives.

To understand the distribution of container+contents and measure/portion uses of pseudo-partitives in English, 10 examples were randomly selected, when possible, from each of the pseudo-partitive forms listed in Table 1. Each example was annotated as one of (a) definitely measure/portion, as in (19a); (b) definitely container+contents, as in (19b); (c) potentially ambiguous between measure/portion and container+contents, like (19c); or (d) other, such as when cup of tea is used as an expression of preference (Opera is not my cup of tea).

While categorizing pseudo-partitives as container+contents readings and measure/portion readings generalizes over the possible readings of these constructions that semanticists have identified (see, e.g., Sutton and Filip 2021, among others), this generalization follows syntactic analyses and therefore will be the focus here as well. These readings can sometimes be distinguished with NP external predication, as in (19a) where stir is only appropriate with a measure/portion reading and into the cake batter reinforces that no actual coffee cup need be present. In (19b), grip is only appropriate with a container+contents reading, given the coffee itself cannot be gripped. Of course, it is not often clear if either reading is being explicitly communicated, as in (19c) where it could easily be either: the sentence could be true in a context where Carolyn is holding a coffee pot which contains the amount of one cup of coffee, or true in a context where she is holding a cup filled with coffee in addition to one presupposed.

As shown in Table 2, for the indirect pseudo-partitives, measure/portion readings make up 41.18%, container+contents readings account for 10.59%, ambiguous readings account for 40.59%, and other readings account for 7.65%. For the direct pseudo-partitives, measure/portion readings make up 17.14%, container+contents readings account for 0.00%, and ambiguous readings account for 15.24%, the other 39.05% being constructions that are not direct pseudo-partitives. Not included in the above percentages is the 28.57% of the total examples where the forms are in lists of ingredients – for example, 1 cup water, 1/4 teaspoon sugar – so the presence of the container is arguably only very weakly implied, if at all.

Lastly, we also noted whether the tokens occurred in a cooking context or not, as with (20a) vs. (20b), where a “cooking context” was taken to be a description of a complex cooking event with multiple steps as opposed to mentioning preparing, say, a cup (of) tea while conversing about something else, as occurs in (20b). Non-cooking direct pseudo-partitives occurred relatively evenly distributed across the genres of the corpus: Academic (n = 1), Fiction (n = 2), Movies (n = 4), Newspaper (n = 1), Spoken (n = 2), TV (n = 2), Web-Blog (n = 2), and Web-General (n = 2). While the spoken data is approximately one-seventh of the total corpus, most of the written examples resemble spoken ones. Six come from TV and movie scripts and the instance from an academic publication (see (20c)) is a recorded quote. One instance from the web, (20d), is from dialogue in a short story. The fiction examples like (20e) are also quoted or reported speech. The remaining examples are from two blogs (e.g., (20f)), a genre characterized as more like spoken language than traditional written texts (see Squires 2010, among others), and from one newspaper (see (20g), which resembles a list of details, albeit in prose rather than bullets). All of this might suggest direct pseudo-partitives outside of cooking contexts are largely found in speech-like domains rather than formal writing.

We end up finding essentially opposite patterns between indirect and direct pseudo-partitives in terms of use in cooking versus non-cooking contexts. For indirect pseudo-partitives, 20.73% occur in cooking contexts while for direct pseudo-partitives, it is 74.60% that occur in cooking contexts (see Table 3). Note that this counts occurrences in lists of ingredients as cooking contexts. If we were to exclude these then it would be 51.52% of direct pseudo-partitives that occur in cooking contexts, which is nearly equal to the 48.48% of direct pseudo-partitives that do not occur in cooking contexts. While this is still notably different from the distribution of indirect pseudo-partitives across cooking and non-cooking contexts, it might be taken as somewhat stronger evidence that direct pseudo-partitives are not relegated entirely to cooking contexts.

Putting these results into the context of previous claims about direct pseudo-partitives in English, there is strong support for the regular use of direct pseudo-partitives in English as they account for 33% of pseudo-partitives in the present dataset. At the same time, it is not the case that direct pseudo-partitives are restricted solely to cooking contexts, though the majority of them are, whether we include ingredient lists for recipes or not. Lastly, and similar to Viennese German, we seem to find a semantic difference between pseudo-partitives, namely that there is only positive evidence of measure/portion readings of direct pseudo-partitives and ambiguous readings, but not of container+contents readings.

4 Analysis

This data could suggest that US English has a direct pseudo-partitive construction that is primarily, though not exclusively, used in the context of cooking. It could also weakly suggest that direct pseudo-partitives are more strongly, if not exclusively, associated with measure/portion readings, given there was insufficient context to ensure that the container+contents readings were used in the ambiguous cases. This begs the question, why is it the case that we found no direct pseudo-partitives with container+contents readings? Or, why does it seem that of must be present for container+contents readings?

Assuming of carries no semantic content, these results tell us nothing new about the distribution of of, or the structure of pseudo-partitives. However, assuming of is lexical for container+contents readings can explain their absence among direct pseudo-partitives – that is, container+contents readings cannot occur in a direct pseudo-partitive because they require of to specify the relationship between N₁ and N₂. The idea that container+contents readings require of to specify the relation between the two nouns is in line with Selkirk (1977), and with the parallel analysis of Hebrew in Rothstein (2009). Furthermore, while Hankamer and Mikkelsen (2008) do not discuss different readings, they too assume that the preposition heads a PP in Danish indirect pseudo-partitives, which are distinguished in that way from direct pseudo-partitives, meaning that assuming the same in English would make the language more like Danish than commonly assumed. This is also in line with the analysis of Viennese German in Grestenberger (2015), namely that container+contents readings have more complex syntax than measure/portion readings do, again making English more alike to German than was previously assumed. Taken together, by assuming that of is lexical in US English container+contents readings, then we not only explain why we see no such readings in direct pseudo-partitives, but we also see similarities in syntax across US English, Danish, and Viennese German.

For pseudo-partitives in US English, we build on Grestenberger (2015), albeit with a distinction between the structures for container+contents readings and measure/portion readings. We will also build on the formal semantic analysis of Sutton and Filip (2021), in which container+contents readings are formalized in terms of a phonologically null relation that takes two nouns and specifies (i) that their referents are a mereological sum that contains nothing else, (ii) that each container has its own contents, and (iii) that all of the contents is within some container (note that some of the finer details are omitted). We assume that of contains this relation, and that it is the head of a prepositional phrase that is sister to the #P that dominates N₁; see (21).

(21)

Container + contents

For the structure of measure/portion readings, presented in (22), N₁ specifies a measurement as head of #′ rather than NP in order to derive the distributional modifiers discussed at length by Alexiadou et al. (2007), Borer (2005), Stickney (2009), and others. In measure/portion readings, we assume of is an optional functional particle, and following the aforementioned research, only N₂ is a fully lexicalized noun.

(22)

measure/portion

The two different structures proposed here capture the complementary distribution of direct pseudo-partitives and readings that are unambiguously container+contents readings based on semantic selection of surrounding predicates seen in the corpus study.

4.1 Addressing previous claims

A great deal of evidence has been presented for and against distinct structures for container+contents and measure/portion readings. The main argument against significantly different structures for container+contents and measure/portion readings is a lack of convincing evidence that the two are, in fact, structurally different. Rothstein (2009) argues judgments are weak, and Grestenberger (2015) argues recursion is restricted for relational nouns. Here we address these issues in turn, after recalling the positive evidence from the background section. We will also include introspective data that further supports assuming two different structures.

First, recall that Jackendoff (1977) shows that certain measure phrases use no preposition, and, in line with this, we show that of is indeed optional for similar measure/portion readings but not for container+contents readings.

(23)

a.

The drink is two parts tequila, one part lime juice, and one part triple sec.

b.

The drink is two shots (of) tequila, one shot (of) lime juice, and one shot (of) triple sec.

c.

The crate contained two bottles #(of) tequila, one bottle #(of) lime juice, and one bottle #(of) triple sec.

Furthermore, Selkirk (1977) shows that extraposition justifies different structures for noun complements and pseudo-partitives, and that these structures distinguish between container+contents and measure portion readings. We recreate the extraposition here explicitly with the sort of sentences at the center of the present discussion to illustrate the continued relevance of this test that is curiously unmentioned in subsequent work.

(24)

A cup of (single-origin) coffee was broken in the sink.

(25)

A cup of (single-origin) coffee was poured (in the sink).

# ˗˗˗˗˗˗˗˗˗˗˗˗˗→

In line with Selkirk (1977), we argue that extraposition is marked for measure/portion readings compared to container+contents readings.

Recall also that Rothstein (2009) presents sentences with marginal judgments from a portion of speakers, showing potential differences between the two readings. While judgments might be marginal, considering it alongside the corpus data might be more convincing.

In Grestenberger (2015), the difference between container+contents readings and measure/portion readings in English is because N₁s in the latter are number-seeking relational nouns rather than fully lexical nouns. This assumption and the proposal that there are restrictions on recursion for relational nouns are offered to explain the difference in acceptability of recursive container+contents readings and measure/portion readings in English. However, we note that it is not clear that such a restriction on relational nouns does indeed exist given the possible recursion with possessive nouns – for example, their sibling’s spouse’s employer’s friend – and familial nouns – for example, sister wife, baby daddy – to name a few. In other words, a structural difference accounts for both recursion, and the novel data that has been presented here.

Finally, an anonymous reviewer suggested that the distinction between the measure/portion and container+contents interpretation could also be seen in the use of cuppa, which is more likely to be used in the former rather than the later kind of interpretation, but still allow a metonymic relation between the two. A search for “cuppa NOUN” in COCA yielded 26 results, nine of which (34.62%) were measure/portion readings, and two (7.69%) were other readings, while the remainder were ambiguous (n = 15; 57.69%). Cuppa occurs a total of 138 times in COCA (including the direct pseudo-partitives just mentioned), and consists of six (4.32%) container+contents readings such as (26a), 29 (20.86%) measure/portion readings like (26b), 87 (62.59%) ambiguous readings such as (26d), and 17 (12.2%) other readings, like (26c).

(26)

a.

sitting here in my nightie with a huge cuppa

b.

walk to the corner for a well-brewed cuppa

c.

It seems our Miss Jadov owes Mr. Givens a cuppa

d.

get chocolate bombs from Cuppa Joe’s

Direct partitives with cuppa pattern like the other direct partitives targeted in this study, and cuppa meaning ‘cup of tea/coffee’ patterns quite like indirect pseudo-partitives. It might be assumed that -a is a reduction of of and that its presence in constructions that look like direct pseudo-partitives undermines the two-structure analysis. However, we might also assume that cuppa is fully lexicalized in meaning ‘a cup of tea’ or ‘a cup of coffee’ rather than a cup of anything else, whereas a pint could refer to ‘a pint of beer’ if uttered in a bar or ‘a pint of raspberries’ if uttered in a supermarket. In other words, in addition to being a measure term, cuppa seems to be a count noun in (26), rather than a measure or container noun with a null N₂.

In summary, we have presented arguments for and against container+contents readings as noun complements with additional introspective data. While introspective judgments might be called marginal, when considered together with the corpus study that shows distinct distributions of measure/portion and container+contents readings among direct pseudo-partitives, this evidence might be taken as convincing enough to warrant the two structures proposed in (21) and (22).

5 Conclusions

We have presented novel data on pseudo-partitives in English, specifically targeting those that can be ambiguous between container+contents readings and measure/portion readings such as cup of water and glass of wine. In doing so, we have addressed two contentious issues: (i) the status of direct pseudo-partitives in English and (ii) the structure of pseudo-partitives in English. First, some have said that English does not have direct pseudo-partitives while others have claimed it does, at least in a limited capacity, and we have provided evidence that suggests that the latter is true, and that perhaps the class is even expanding beyond being exclusive to cooking contexts. Second, while some have said that container+contents and measure/portion readings differ in structure and others only in realization of N₁ as a measure term rather than referring to concrete objects, we have provided evidence that could be taken to support the two-structure analysis.

Moreover, Selkirk (1977) and others have distinguished container+contents readings from “genuine” pseudo-partitives because of the complement structure in which N₁ refers to an object as opposed to a measure. In her typological and diachronic survey of pseudo-partitives, Koptjevskaja-Tamm (2001) characterizes pseudo-partitives as those in which the container/measure term “creates a unit of measure” (p. 529) and “quantifies over the kind of entity (‘tea’, ‘books’) indicated by the other nominal” (p. 527). A strong interpretation might require what is formalized as a nominal measure term as in Borer (2005) or Sutton and Filip (2021). Under this interpretation, then here, as in Selkirk (1977), container+contents readings are not pseudo-partitives. A weak interpretation might allow for reference to possible quantities, and assuming ad hoc measure interpretations are derived from container+contents readings in parallel to the use of the suffix -ful as in cupful of water (cf. Sutton and Filip 2021, among others), one might assume container+contents readings are pseudo-partitives given this possible derivation.

Semi-lexical material has long puzzled linguists establishing dichotomies both in syntax and semantics (cf. Corver and Van Riemsdijk 2001). The current work begins to show that interface conditions (see, e.g., Chomsky 1995, among many others) may have an impact on the surface realization of such semi-lexical material. In actual use, it comes as no surprise that such conditions actualize as gradual processes (in the sense of Kroch 1989) rather than all-or-nothing conditions. But crucially, if our analysis is on the right track, the syntax-semantics interface may also impact on usage patterns and not just vice versa. With more resources, it would be insightful to ascertain in future work to what extent such possibly incipient patterns bear resemblance in terms of quantities and qualities to other paths of change towards direct pseudo-partitives that have been taken within Germanic and beyond and whether they went into completion and full regularization (as in German) or whether they have become obsolete and possibly interrupted (as in Middle English). In particular when a change in grammar is at a potentially incipient change and its frequency is low, additional empirical verification in future research can also come from experimental data (in addition to the mentioned increased corpus work with more resources). But rather than only using classical graded grammaticality judgments of different kinds, one might use recently developed experimental paradigms that have the potential to detect significant grammatical distinctions in critical areas that appear to be below the radar of the usual norms (cf. Gergel et al. 2021).