3.2.1 Gender

It is a fact that the gender of a noun cannot be overriden, even in cases where the noun refers to a human and intuitively has a real-life gender. For example, the masculine noun student ‘student’ in Norwegian cannot be used with feminine morphosyntax, regardless of the semantic gender of the student:

This is a piece of evidence suggesting that gender is a part of the lexical entry of the noun itself.¹⁰ It is inherent to the noun student ‘student’ that it has masculine grammatical gender. This is the basis of the approach that has been presented before in which gender is assigned on the N head itself (Roca 1989; Harris 1991; Carstens 2000; 2010; 2011; Ralli 2002; 2003; Alexiadou 2004; see also Wechsler & Zlatić 2003), or, according to the Distributed Morphology framework, on a noun-categorizing head n (Lecarme 2002; Ferrari 2005; Kihm 2005; Acquaviva 2008; 2009; Lowenstamm 2008; Kramer 2009; 2014; 2015; 2016; Percus 2011; King 2015; Deal 2016; Fathi & Lowenstamm 2016). From the DM perspective, placing gender in n can be justified because of the fact that gender collaborates with other processes that are associated with n, such as nominalization, inflection and declension class (Kramer 2016).

In my view, one issue with the n approach in its current form is that gender would need to always be in the nominal structure if we see a noun. This is problematic when we consider bare singular nouns in Norwegian, which do not show overt signs of gender agreement. I repeat example (18) below: there is no morphosyntactic marking showing that appelsin ‘orange’ is masculine, plomme ‘plum’ is feminine and eple ‘apple’ is neuter:

This would mean that there is no n in these nominal structures, but that goes against the principle of DM in that n needs to be present for the root to be integrated into the larger syntactic configuration. Without n, there is no way to tell that the word is a word, and which part of speech it belongs to. A more satisfying approach would be one in which gender is part of the categorizing of a word without it being the only way for the noun to be insertable into the syntax. Still, I am in favor of the main message behind the N/n analysis, namely that gender is part of a noun’s lexical entry. The main difference between my approach and others is that, following the noun phrase structure I have laid out, a feature can be part of a noun’s lexical entry without needing to be located on the noun head itself. Instead of lexical information being an inert conceptual item, lexical information is inserted through a contiguous head-complement sequence with its root at the bottom of that sequence (Halle 1997; Svenonius 2012; 2016).

In the previous subsection, I presented a head Noun Marker (NM) located immediately above the kind-denoting NP level. The purpose of this head is to provide identity to and categorize the entity denoted by the noun. Based on this description, it seems natural that it is the NM head that contains the gender-assigning feature ±fem, as well. NM is similar to n in that its role is to give the noun its identity and assign it a part of speech. The difference is that NP is a syntactic constituent on its own before NM is applied, meaning that it can take the role of argument and be selected by a verb. Still, if a full determiner phrase needs to be formed, NM is necessary: a determiner cannot select an NP directly, as it needs an individual with reference. This means that whenever we see a DP, the nominal structure contains an NM for D to select.

The purpose of NM is not only to carry a gender feature – instead, it provides the identity of an individual, meaning that it can contain other features that are crucial to the lexical item. This also makes NM different from an approach where gender comes out through a specialized Gender head, in that it categorizes nouns in more than one way, meaning that NM’s presence does not depend on the presence of gender. In the way I envision NM, nongendered languages also have NM in their nominal structure, since nouns can be categorized in different ways in different languages. NM is then a universal part of the formation of a nominal projection.

Despite all of this laying-out of the nature and use of NM, I wish to emphasize that its existence is not the main thesis of the article. I do not wish to make a bold statement about gender itself or where it is located. In several ways, NM is equivalent to n, and there may well be ways to explain the issues I come across with a DM framework, which I welcome suggestions for. The main reasons that I use this NM-based approach are that 1) lexical content can be located outside of the nominal head itself; 2) NM can be left out in certain situations, making it possible to shrink the nominal structure down to just an NP; and 3) the noun head itself does not carry a gender feature.

In the following, I ask: what else can be found on the NM head? The idea that I will put forward is that the count-mass distinction is another binary feature that can be located on NM, at least in some languages. To build up to this, we need to understand some earlier perspectives about where countability is located.

3.2.2 Countability

One major question in research on the count/mass distinction is whether or not the divisibility of a noun is encoded into the lexical entry of the noun itself. The alternative option would be that countability is introduced through a noun-external head, such as a number or classifier head.

I will work under the assumption that countability is binary and that a noun can either be count or mass, rather than there being a wider, more nuanced spectrum (cf. Allan 1980; Pelletier 2012). One reason for this choice is syntactic: if countability is binary, it can be translated into a syntactic feature ±atomic. One question, however, is whether each noun can be associated with one and only one ±atomic value. This matters for determining the feature’s location: according to my conclusion about gender so far, the gender value of a noun is fixed because it is part of the noun’s lexical entry, whereas a feature like ±pl can be freely valued as positive or negative (e.g., horses versus horse). So, whether a noun’s ±atomic value is rigid may be an indication of whether the feature is located in NM or not, helping us draw a parallel to the gender feature.

There are two particular approaches of interest relating to the syntactic location of countability. On one end, there is the approach that I call the inherent mass approach. Borer (2005) argues that nouns themselves are always mass-denoting and that the count meaning is caused by a functional head in the nominal projection that divides the mass, most often a classifier or plural marker (see also Zamparelli 2000). One opposing view is that all nouns are lexically specified for countability, mentioned earlier (Pelletier 1975; Krifka 1989; Cheng et al. 2008; Rothstein 2010; 2017). I label this the lexical specification approach. In the following, I will discuss the applicability of each approach when interacting with the Norwegian case study that I have presented.

Supporters of the inherent mass approach argue that all nouns are “born” with a mass denotation and that they can only denote countable units if a dividing head is added to the nominal syntactic structure. The ability to interact with the count system is then provided by syntax. In a language like English, the dividing head could for example be the plural suffix -s. Borer (2005) uses Mandarin Chinese for her argument, a language that supposedly requires a classifier for the noun to receive a count reading. Below are examples of this: in example (45a), the noun ren ‘person’ requires the classifier ge to get the count reading; in example (45b), the noun mi ‘rice’ is treated as a divisible unit because of the classifier l.

Nouns that receive a mass reading are not accompanied by a classifier:

Because the mass reading involves no classifier, and the count reading obligatorily involves one, the argument is that the mass interpretation is the one that is at the “bottom” of the structure, i.e., that nouns are inherently mass. The syntactic structure behind Borer’s (2005) argument is one in which a dividing head provides count meaning, if there is one. I translate this into the following underlying structure for a Norwegian phrase like to personar ‘two people’:

Please note that Num is equivalent to Div. Various labels have been used for the head that divides an object and introduces its count reading. In Mandarin Chinese, classifier heads may fulfill this purpose. In languages like English and Norwegian, Num has a similar role. The Num head to ‘two’, according to Borer’s (2005) view, transforms the inherently mass-denoting noun personar ‘people’ into a count-denoting one. Count readings are thus introduced by the syntax. The count interpretation is crucially not located on NM here: since NM is part of the noun’s lexical projection, it cannot contain a separate word that serves as a divider, like a numeral.

Let us consider what the inherent mass approach would mean for Norwegian data. Before starting to consider grinding and portioning, I will generalize that Borer (2005) would predict that nouns are mass-denoting whenever they are bare. For Norwegian, this is not true:

The meaning of hund ‘dog’ above, without an accompanying indefinite article, can certainly have a mass reading (likely referring to dog meat), but it is fully possible, and even likelier, that one is referring to a countable entity dog.

When considering grinding and portioning in Norwegian specifically, we encounter more issues. The inherent mass approach predicts that “grinding” and “portioning” are the result of the unexpected addition, or lack thereof, of a dividing head. In principle, according to Borer’s (2005) approach, all nouns should be receptive to a count or mass reading, since the only difference between them is the existence or nonexistence of a head that is already a natural part of the noun phrase. The fact that these “ground” or “portioned” readings sound less natural would come from world knowledge rather than the presence of additional functions that force a change of countability value. While masculine and feminine nouns cannot be used as grounds to dispute the inherent mass approach, since either reading is available, the approach cannot explain the fact that neuter nouns are locked into one of the two countability values. I ultimately dismiss the proposal that countability comes from noun-external projecting heads.

On the other end lies the lexical specification approach, whose supporters believe that nouns have one countability value that makes up a fixed trait of that noun (Krifka 1989; Cheng et al. 2008; Landman 2011; Rothstein 2017). This is not an unintuitive idea: if one thinks of the concept apple, it is reasonable to think of apple as a countable object. Cheng et al. (2008) use grinding and portioning to argue for the lexical specification approach, using data from Mandarin. They point out that Mandarin does have appearances of bare singular count nouns. Below are some examples:

The detail that Cheng et al. (2008) highlight is that, even though these count nouns are bare, they are restricted to a count reading and cannot have a “ground” interpretation. In the example below, gǒu ‘dog’ can only refer to an intact, countable unit (i.e., the animal itself). It cannot refer to some kind of dog mass (such as dog meat):

Mandarin would need to form a compound like gǒu-ròu ‘dog-flesh/meat’ for the desired mass reading to come through. According to the inherent mass approach, there should be no problem with interpreting gǒu ‘dog meat’, because there is no dividing head.¹¹

Cheng et al. (2008) argue that grinding and portioning are actually a form of coercion and that this coercion is realized in different ways in different languages. Languages that overtly show a mass-count distinction will use the morphosyntax associated with the unexpected countability value: if an English speaker says that there was dog all over a wall, the lack of a count-specific article and/or plural marker leads the listener to a “last resort” solution whereby dog ends up referring to dog meat. On the other end, languages that do not grammatically show such a distinction will rely on context and world knowledge. In the example involving a noun meaning ‘dog’ being on a wall, it may be the case that ‘dog’ is so clearly tied to a countable unit that the count reading is more probably to be relevant, and, without morphosyntactic cues, the count reading becomes the dominant one.

The lexical specification approach can be translated in syntactic terms in the following way. As I laid out in the previous section, my syntactic framework treats lexical information as a sequence of heads introduced in the syntax, and the noun head itself does not carry lexical features. I also presented a head named Noun Marker (NM) which categorizes a noun and supplies it with identifying feature(s), such as gender. The way I apply the lexical specification approach is then that NM can also contain an interpretable feature that informs the noun’s countability value, which I call ±atomic. When this feature is positively valued, the entity represented by the noun denotes something that can be split up. When the feature is negatively valued, the entity denotes a mass. Below is an illustration of the syntactic structure behind gǒu ‘dog’, before additional heads are added to the nominal phrase. Keep in mind gǒu ‘dog’ is always considered a countable noun:

The lexical specification approach is compelling when we consider the data presented by Cheng et al. (2008). It sounds reasonable if the only Norwegian data available to us is neuter nouns. Considering how neuter nouns like eple ‘apple’ or vatn ‘water’ are fixed to one countability value, the natural conclusion would seem to be that this value is inherently tied to the lexical entry of the noun itself. However, we cannot escape the fact that masculine and feminine nouns are not tied to one value.

An approach that fully depends on lexical specification, then, cannot give the full, universal story. The Norwegian data presented in this paper may muddy the water and lead to an argument that takes variation into account, both across languages and within the same language. Upon investigating the Norwegian data further, I will conclude that it is possible to introduce countability as a feature in the lexical entry, but it is not an obligatory part of every noun in every language. I now present my proposal.

4.2.1 Indefinite form

I use minimal comparisons between masculine, feminine and neuter nouns in support of the claim that neuter mass nouns cannot be portioned in Norwegian. This specific example was presented earlier:

Based on the structure I have advocated so far, my explanation for this distinction between masculine/feminine nouns and neuter nouns is that nouns like vatn ‘water’ are lexicalized with a –atomic value located in its NM. The reason for the unacceptability of eit vatn ‘a.neut water.neut’ would then be reminiscent of what we saw for grinding: neuter mass nouns are inherently –atomic, so they cannot be portioned.

Although the final conclusion will be that all neuter nouns are locked to one countability value, the story is not as simple, as the following empirical details will show. The first detail is that speakers find the “portioning” of neuter mass nouns marginally acceptable if the indefinite article takes the masculine form:

The permissibility of this construction heavily depends on context, in that it must already be firmly established what kind of container has been left unpronounced. Example (61) further illustrates the need for adequate context. One may imagine the sentences below if the speaker is at a fast food restaurant and asks for small sachets of salt and pepper:

A crosslinguistic perspective may help us understand why the masculine article is allowed here. German has also been found to have a mismatch between the indefinite article and the noun in portioned noun phrases, where the articles are always masculine regardless of the actual gender of the noun (Wiese & Maling 2005). This has led some to argue that mass nouns are portioned through a nonovert “classifying” head that forces a count reading.¹⁴ What is indefinite in ein salt ‘a.masc salt.neut’ is then the thing that contains the salt mass, not the salt itself. Portioned noun phrases like a coffee would be structurally identical to a noun phrase like a cup of coffee.

Note that the insertion of a nonovert nominal head is different from the idea that cup of is a dividing head in the Borerian (2005) sense: a cup of coffee is a biphrasal structure in which the head of the noun phrase is cup, and of coffee is a PP complement of this head. This is not a noun phrase in which cup of is a classifier in the nominal projection of a head coffee. While Borer (2005) argues cup of to be an example of a classifier in English, the syntactic structure of two cups of tea suggests that cups is instead a nominal head. Here is an illustration:

Cups is in a recursive structure in that it is the head of a functional sequence that is on top of the functional sequence of (of) tea.

The masculine form on the indefinite article is not the only reason to argue for a nonovert nominal head. Another piece of evidence is the fact that, when a plural numeral is used to count the portions, there is surprisingly no plural suffix on the noun. This is true regardless of whether the noun is gendered or neuter:¹⁵

The lack of plural marking suggests that the numeral to ‘two’ is not applied to the mass nouns themselves; instead, they count the nonovert containers of a substance. I label this container noun Unit. For the counting of a masculine noun like vin ‘wine’, we can imagine the following underlying structure:¹⁶

Unit is an N head that projects its own NM. The NM of Unit crucially does not have a gender feature. This is because Unit does not directly equate any overt noun, meaning that there is no noun that can be assigned a gender. NM ends up unvalued for gender.

With the concepts developed so far, I will tackle the initial question of the subsection: why can masculine and feminine nouns be portioned with a gender-agreeing indefinite article, while neuter nouns can only be portioned if a masculine article is used? I argue that the article agreement we see comes from different sources: for masculine and feminine nouns, the determiner has access to the gender value of the mass noun’s NM, but for neuter nouns, the determiner cannot find a gender value and thus picks masculine as a default form. This tracks with my proposal that gendered nouns have gender in their lexical entry and neuter nouns have countability in their lexical entry.

First, let us consider a portioned feminine noun phrase like ei mjølk ‘a.fem milk.fem’. The general process of agreement that I follow is one in which a probe may continue to search if it is not able to find the desired feature to agree with. In this case, the determiner agrees with the gender of the lower NM (i.e., the NM that categorizes mjølk ‘milk’) because the most local NM (i.e., the NM of Unit) does not contain a gender feature. The structure looks like this:

The feminine form on the indefinite article then comes from the +fem feature value on the NM of mjølk ‘milk’, not from the closer NM that categorizes the NP headed by Unit.

With this analysis of ei mjølk ‘a.fem milk.fem’ in mind, we have an explanation of the masculine form on the indefinite article in cases of portioned neuter mass nouns, as in ein vatn ‘a.masc water.neut’. The reason for the use of the masculine form comes from the fact that masculine is the default assignment gender in Norwegian. I wish to emphasize that masculine is not the “default” form in the same manner as I have so far argued the neuter form to be. A convincing line of research distinguishes between the concepts of default assignment gender and default agreement gender in natural languages (see especially Lohndal & Westergaard 2021 for a focus on Norwegian; consider Fraser & Corbett 1997; 2000 and Adamson & Šereikaitė 2019 for other languages). As I stated earlier, the neuter form appears in agreement when the nominal constituent does not have an assigning head that can carry a gender value, like in copular sentences where an infinitive clause is described by a predicative adjective. That is, neuter appears when there is no gender feature but agreement is required. On the other end, we find default assignment genders: this is the gender that is assigned to a noun that otherwise has not been assigned one, within a given speaker’s grammar. In this case, a noun may be headed by an NM, but its gender feature has not been valued. The gender feature is valued with the default assignment gender. There is evidence to suggest that masculine is the default assignment gender in Norwegian: for example, a majority of English loanwords are found to be masculine (Graedler 1998; Johansson & Graedler 2002), first language learners assign masculine gender to nouns that they have not yet acquired the gender of (Rodina & Westergaard 2013; 2015)¹⁷ and the feminine gender is consistently in decline in favor of the masculine (van Baal et al. 2024). The best explanation for the default use of masculine in these cases, and the default use of neuter in others, is that there are two types of default: one for assignment and one for agreement.

Returning to the “paradox” of vatn ‘water.neut’ combining with ein ‘a.masc’, we can gather that the default assignment form is triggered because there is no gender value to agree with in the entire noun phrase structure: neither the NM of Unit nor the NM of vatn ‘water’ contains a ±fem feature, as upper NM contains no class specification and the lower NM only contains ±atomic. The determiner, looking for a class feature to agree with but not finding one, ends up going for gender agreement, and because masculine is the default assignment gender, this is the form that appears on the determiner, resulting in ein ‘a(n).masc’. Below is an illustration of this agreement relationship:

The NM of Unit is not inherently valued –fem, but this value is inserted as an “escape hatch” solution to the lack of a gender feature for the determiner to agree with.

After having considered portioning in the indefinite form, the conclusion again becomes that neuter nouns do not have a gender feature in their NM but instead an ±atomic feature. The reason for confusion about this fact stems from the acceptable use of the masculine indefinite article, which I have shown is a misleading issue. Portioning is a slightly more complicated syntactic process than grinding in that it involves a recursive nominal structure in which a nonovert noun has been inserted that denotes a carrier or container of the entity denoted by the visible mass noun. The reason that neuter nouns “can be portioned", albeit with the masculine indefinite article, does not come from the fact that the –atomic value has been altered in the NM of neuter mass nouns. Vatn ‘water’ remains lexically specified as having a mass interpretation.

4.2.2 Definite form

Some may view the following piece of data as evidence that neuter mass nouns can in fact be portioned while showing neuter agreement:

Here, the definite suffix cannot take the masculine form, as opposed to what we saw for the indefinite article. This indicates that the “portioning” of definite neuter mass nouns is a different phenomenon. Namely, I contend that this is not a sign of portioning at all, at least not in the sense I have laid out so far where a nonovert noun is inserted in order to provide a unit for the substance to be divided into. One piece of data that supports this claim relates to the use of numerals and plural marking. Remember that in the indefinite form, the use of plural numerals did not cause the use of a plural suffix on the visible mass noun, suggesting that there is an invisible Unit head that is pluralized rather than the mass noun. This is different in the definite form: when a numeral is used in the definite form, the noun is pluralized. This happens regardless of whether the noun is gendered or neuter:

This leads us to the argument that the example of a “portioned” neuter mass noun in (67) is not an example of portioning at all. Instead, the intuition that the mass nouns have been portioned is a natural effect of the use of the definite form as an establisher of some “contextually identifiable medium” (Jackendoff 1991). The examples above can be compared to the definite form of water in the second sentence in example (69):

The same interpretation arises if the sentence is translated to Norwegian:

There is no way to make a mass noun definite without restricting the location or amount of the substance: the definite article makes its expected contribution of establishing familiarity and uniqueness. Vatnet ‘the water’ in the example above, however, does not refer to a contextually relevant container of water, meaning that portioning has not taken place. It cannot be the case that the definite article always portions a mass. For example, if this were the case, we would expect vatnet ‘the water’ to not be the subject of a verb like accumulate, since a container cannot accumulate. But this is not borne out:

The same is the case In Norwegian with a verb like samla seg ‘gather’:

It is therefore possible for a definite mass noun phrase to denote a mass and not a portion of a mass. I argue that the noun phrase vatnet ‘the water’ in example (67) is not a portioned mass noun, but simply the definite form of a mass noun. This means that vatn ‘water’ preserves the –atomic value in its categorizing head NM, and the definite suffix shows neuter agreement because there is no gender feature in NM. Here is an illustration:

The tree diagram shows that the structure of the “portioned” definite mass noun involves no portioning at all. The noun phrase is simply the definite form of a mass noun.

4.3.1 A note on sorting

In subsection 2.5, I showed one crucial way in which portioning and sorting are different in Norwegian, namely that neuter mass nouns have no issue with undergoing a “sorting operation” whatsoever, and that the sorting of a neuter noun does not restrict the agreement form that appears on indefinite articles, definite suffixes or adjectives. Below are examples of the neuter noun vatn ‘water’ when sorted versus portioned. As you can see, the sorted version comes with the neuter indefinite article eit ‘a(n)’ and the neuter adjective suffix -t, whereas this is impossible when vatn ‘water’ refers to a unit of water:

Another piece of data that shows a difference between the two processes is that, unlike portioned nouns, sorted nouns do in fact show plural marking. It is unacceptable to drop the plural suffix in combination with a plural number like to ‘two’. Below is a comparison (drawn from subsection 2.5):

This difference matters because of my argument in subsection 4.2.1 that portioned nouns cannot have plural marking because the plurality is in a nonovert Unit head, not the overt mass noun itself. The fact that there is plural marking on sorted nouns is a sign that there is no Unit in sorted noun phrases, at least in Norwegian.

Sorting behaves differently from portioning, as we can see from non-English data. I take the facts we have about sorting to mean that sorting is a form of polysemy rather than a symptom of some complicated operation (see Falkum 2010). This kind of polysemy is similar to how nouns like glas ‘glass’ have dual meanings that are related to each other, where one is count and the other is mass. In a way, the “mass nouns” that have this kind of “sortal” meaning are count nouns.¹⁹