6 Macedonian Heritage Speakers’ Strategies of Interpreting Specificity and Genericity

Univerza v Novi Gorici, Center for Cognitive Science of Language, Nova Gorica, Slovenia
metodi.efremov@student.ung.si

This study proposes a design to test how heritage speakers of Macedonian interpret generic plural nouns in their heritage language due to the influence of the majority language in this domain. Macedonian has the peculiar property of allowing both bare and definite plural NPs to refer to kinds (Tomic, 2012), contrary to the predictions made by Chierchia’s Nominal Mapping Parameter (NMP) (1998). For example, English only allows the bare plural option, whereas French only the definite one, as predicted by the NMP. We put forward a grammaticality judgment experiment to test this in Canadian French-Macedonian and Australian English-Macedonian heritage speakers, using monolingual speakers of Macedonian as controls. We have a two-tailed prediction concerning the possible effect the majority language has on the heritage one. If abstract economy conditions in line with Chierchia (1998) determine cross-linguistic influence, as Serratrice et al. (2009) suggest, we predict French-Macedonian speakers to retain generic bare plural NPs in Macedonian despite the effect of French, while English-Macedonian speakers will judge definite plural ones as unacceptable. If semantic overlap decides the nature of cross-linguistic influence (Kupisch, 2012), we predict French-Macedonian speakers to interpret definite plural NPs as generic, judging them as acceptable but not bare plural ones due to the effect of French. We expect to find the opposite pattern in English-Macedonian speakers; speakers will interpret bare plural NPs as generic, judging them as acceptable but not definite plural ones due to the effect of English.

1. Introduction

In this study, we propose an experiment to investigate how Australian English-Macedonian and Canadian French-Macedonian heritage speakers comprehend and judge in their heritage and dominant language bare plural subject NPs and definite plural subject NPs, with regards to specificity and genericity.
Regarding genericity, Macedonian is an interesting case as it allows a generic reference with definite and bare plural subject NPs (Minova-Gjurkova, 1994; Tomic, 2012).

1. Kucinja      sakaat                            da           si                  igraat. [generic V, specific *]
   Dogs.Pl.    like.3Pl.Pres.Imper Subj.CL Dat.Refl.CL play3Pl.Pres.Imper
2. Kucinjata         sakaat                           da           si                   igraat. [generic V, specific V]
   Dogs+the.Pl. like.3Pl.Pres.Imper Subj.CL Dat.Refl.CL play3Pl.Pres.Imper

Both examples allow the generic reading about dogs as a species, with sentence two also allowing a specific interpretation about a salient set of dogs in the discourse.
This behavior is atypical as it is contrary to the predictions made by Chierchia’s Nominal Mapping Parameter (NMP) (1998), by which, typologically, languages with articles differ in whether or not plural nouns can make a generic reference with the definite article.
On the other hand, English and French fall into the typological classifications defined in the NMP.

3. Dogs like to play. [generic V, specific *]
4. Les chiens               aiment           jouer. [generic V, specific V]
   The.Pl dogs.M.Pl like.3Pl.Pres  to play

French allows definite plural subject NPs to have a generic and a specific reference, with bare arguments being ungrammatical without exception, while English only allows bare plural subject NPs to have a generic reference.
English and French are, therefore, in a subset-superset relationship with Macedonian, not in terms of available syntactic structure, but possible semantic interpretation regarding generic reference and the question is how the majority language will affect the minority language in this aspect.
Due to this, Macedonian and its heritage speakers present a novel field of research as previous works on specificity and genericity in heritage populations (Serratrice et al., 2009; Montrul and Ionin, 2010, 2012; Kupisch and Pierantozzi, 2010; Kupisch, 2012) have only focused on languages where only one option for generic reference is available. Furthermore, it allows us to test two hypotheses about cross-linguistic influence: semantic overlap and abstract economy conditions in a new way.
This study aims to provide a theoretical account of genericity, overview existing experimental literature on this topic, give a brief survey of how Macedonian expresses genericity morphosyntactically and how it compares to English and French, to put forward an experiment and make predictions about what we expect based on previous findings, and to present the limitations of our proposal.

2. Semantic Framework

For our proposal, we adopt Chierchia’s NMP (1998) to account for the pattern found in English and French but not for Macedonian, and we will explain why in section four. By the NMP, the typological cross-linguistic differences found amongst languages in the use or absence of articles with singular, plural count and mass nouns is a result of the language-specific setting of the two semantic features [+/-argument] and [+/-predicate]. The motivation for these settings is that nouns can play a double role as they can be of type e (e is the type of individuals/entities) and an argument when they refer to a kind (as “the professor” in “the professor was late to class”) or of type <e,t> (<e,t> is the type of functions from individuals to truth values) and a predicate (as “a professor” in “Lazar is a professor”), and how these two options are actualized differently across languages. Additionally, depending on the specification of these two semantic features, languages differ in how they make generic reference.
Depending on their specification, we have four possible typological combinations: [-arg, -pred], [+arg, -pred], [-arg, +pred], and [+arg, +pred]. Of these four, only [-arg, -pred] is not attested as it is an impossible combination – nouns would then refer to nothing. Of the other three, we will not explain type 1 languages ([+arg, -pred]), like Chinese or Japanese, as they are irrelevant to our experiment.
The other two types that remain, type 2 ([-arg, +pred]) and type 3 ([+arg, +pred]), correspond to French and English respectively. As in type 2 languages ([-arg, +pred]), all nouns are predicates and cannot be an argument without a determiner, they necessitate an overt determiner to be an argument and to refer to a kind, thereby having generic reference. We exemplify this with the following pair of examples.

5.          Les       médecins                aiment
   The.Pl doctors.M.Pl  like.3Pl.Pres
   les           voitures                     chères. [generic V, specific V]
   the.Pl     cars.F.Pl.                  expensive.F.Pl
6.         *Médecins          aiment            les       voitures             chères.
   doctors.M.Pl  like.3Pl.Pres  the.Pl   expensive.F.Pl  cars.F.Pl

In these sentences, the definite article is a lexicalized iota i operator, which can be intensionalized and correspond to the type-shifting operator ∩. The iota i operator is interpreted as such that it picks out the greatest element from the predicate’s extension. If the iota operator i applies to a set of pluralities (for example, birds), it will refer to the largest possible one. In this case, with birds, to the one that consists of all birds.
On the other hand, in type 3 languages ([+arg, +pred]), nouns can freely function as both arguments and predicates, shifting back and forth with the available type-shifting operators. Nouns of type e [(+arg]) will be mass and be able to occur as bare arguments in any position, while nouns of type <e,t> ([+pred]) will be predicates and require a determiner to be grammatical, which we see with singular nouns. We illustrate this contrast with the following examples.

7. Water is dripping on the floor.
8. I see water in the distance.
9. The singer was ready for the concert.
10. *Singer was ready for the concert

Yet bare plural count nouns are allowed in any syntactic position, and we see this in generic sentences with individual-level predicates. This behavior is possible due to applying the type shifter ∩ to the plural noun, which yields a kind. If applied to singular nouns, it will be undefined. We show this with examples 11 to 16.

11. Cats like to eat mice. [Generic V, specific *]
12. *Cat likes to eat mice.
13. I find cats to be lovely. [Generic V, specific *]
14. *I find cat to be lovely.
15. Santa Claus gifts presents to kids. [Generic V, specific *]
16. *Santa Claus gifts presents to kid.

However, one problem arises from what we have adopted so far. English also has a definite article, so it should be possible for “the” to function as an iota i operator and to intensionalize the type-shifter ∩. Chierchia remedies this by resorting to principles fundamental to the architecture of grammar, such as the elsewhere condition, the blocking principle, and avoiding structure. By the elsewhere condition, “language-particular choices win over universal tendencies” (Chierchia, 1998, p. 360), and “don’t do covertly what you can do overtly” (Chierchia, 1998, p. 360).
From these two principles, we arrive at the blocking principle, which states that if a language has a particular morpheme which is the intensionalized equivalent of one of the type-shifting operators, then that morpheme should be preferred to the covert type-shifting operator. Romance languages follow this. English follows the avoid structure principle. It has a covert type-shifting operator that transforms bare nouns into kinds in generic sentences, which can apply as early as the NP level – thus is preferable in terms of economy conditions as we do not have to project an overt determiner and additional syntactic structure to get the interpretation we need.

3. Cross-Linguistic Influence and Previous Research on Genericity and Specificity in Heritage Speakers

A dominant topic in the study of heritage speakers nowadays is cross-linguistic influence and the effect that the majority language has on the heritage one, as research has shown that the language skills of heritage speakers significantly differ in some areas from monolinguals resulting from the imbalance between their minority language and majority language, as well as the amount of input received in both (Benmamoun et al., 2013; Polinsky, 2006).
One such domain that has received significant attention in the literature is the interpretation of specificity and genericity. These two categories can have a great degree of cross-linguistic variability, and languages can differ in their distribution of how they express these categories – hence giving a rich ground to research how one language can affect another.
This research has shown that while cross-linguistic influence is possible, it is not random or unrestrained, and several hypotheses have been proposed to account for the patterns found.
One such hypothesis is the structural overlap one, which states that cross-linguistic influence is possible only if the construction affected is at the interface between two modules of grammar (syntax and semantics or pragmatics), and the two languages must structurally overlap (subset-superset relationship) (Hulk and Müller, 2000; Müller and Hulk, 2001). Structural overlap is: if language A has construction X1-X2 and construction X2-X1, but they are used in different contexts, while language B only has X2-X1 for both contexts, speakers of both languages will overextend the use of X2-X1 in language A, even in inappropriate contexts for its use.
Another similar hypothesis is the semantic overlap one proposed by Kupisch (2012). The foundation of this hypothesis is that the interpretation of the structures and their appropriate semantics within a context determines cross-linguistic influence, not the surface structure of the structural representations. For example, in French, the semantic interpretation still has to be applied to the nominal (which can only be a DP as we saw in section 2), whether generic or specific. Semantic overlap is: if language A has construction X1-X2 which has both meanings Y and Z, while language B only has meaning Y assigned to X1-X2, speakers of both languages will overextend meaning Y of X1-X2 in language A, even in inappropriate contexts.
One alternative hypothesis, which does not concentrate on overlap, in the literature researching how heritage speakers interpret specificity and genericity is that abstract economy conditions, such as the avoid structure principle in section 2, decide how cross-linguistic influence occurs in the domain of genericity and specificity. Serratrice et al. (2009) were the first to propose this alternative, and their results suggest that this hypothesis might be correct.
Serratrice et al.’s participants were younger and older simultaneous bilingual English-Italian children (one group living in England, another living in Italy), Spanish-Italian children (living in Spain) and, as control groups, monolingual Italian and English children. As additional control groups, they had adult English and Italian monolinguals.
They predicted that if structural overlap drives the directionality of cross-linguistic influence, then Italian will affect English as it allows definite plural NPs, and English allows both definite and bare plurals NPs. In that case, the simultaneous bilingual English-Italian children will judge definite plural NPs with a generic reading in English as more acceptable than the English monolinguals. However, if abstract economy conditions determine the direction of influence, they predicted the opposite. English will affect Italian as English has the more economical alternative as it can have kind-referring bare plural NPs with covert type-shifting operators instead of projecting additional syntactic structure like Italian. In this scenario, there will be a larger acceptance rate of bare plural NPs in generic contexts by English-Italian bilinguals opposed to Italian monolinguals.
Serratrice et al. used a scalar grammaticality judgment task where participants judged six pairs of test sentences and three pairs of filler sentences by their acceptability. The pairs were one specific and one generic sentence, and all had a subject plural noun. Participants saw a prototypical picture or object for the animals introduced in the sentences. Before each sentence, they used an introductory phrase: for the specific ones, it was “Here”/”Qui”, whereas for the generic ones, it was “In general”/”In genere”. In Italian, the cue for grammaticality was the absence or presence of a definite article. In English, the cue for grammaticality was the adjunct as it determines the semantic context.
For English, the results showed that monolingual and bilingual children accepted bare and definite plurals in both generic and specific constructions respectively. Although, they also judged the bare plurals in specific contexts and the definite plurals in generic contexts as acceptable. Serratrice et al. reason that this might be the case due to the strength of the contextual cue, as the adjunct did not unambiguously determine the semantic context of the sentence. Further proof for this reasoning is that the adult group did not successfully recognize the semantic context by the adjunct 100% of the time.
For Italian, the results showed that the monolingual adult and children Italians performed at ceiling in both conditions. The English-Italian bilingual children accepted bare plural subject NPs in sentences with a generic reading, in contrast to the monolingual Italian children. They also found that exposure to the language plays a role, as the group living in the UK performed worse than the one living in Italy.
Based on these results, the authors conclude that abstract economy conditions play a role in cross-linguistic influence, not structural overlap, as Italian did not affect English, but English did affect Italian.
Other experiments on heritage speakers have come to similar conclusions. In cases where English is the subset and Spanish is the superset, Montrul and Ionin (2010, 2012) tested Spanish heritage speakers whose dominant language was English. They included Spanish L2 learners as a comparison group and Spanish native speakers as a control group. All their participants were adults. They tested their participants using a grammaticality judgment task to check their knowledge of articles and a truth-value judgment task testing how participants interpret said articles. Their results showed a transfer effect from English to Spanish in the L2 learners and the heritage speakers. Unlike the native Spanish speakers, the other two groups judged bare plural NPs as acceptable in a generic context, and they preferred the specific interpretation of definite plurals over the generic one.
So far, we have only overviewed works that have tested the structural overlap and the abstract economy conditions hypotheses but not the semantic overlap hypothesis. To the best of our knowledge, only one work has tested it, and that is Kupisch (2012) which tested the use and interpretation of specific and generic subject NPs in the Italian of both simultaneous bilingual German-Italian adults and second language learners. The German-Italian simultaneous bilinguals were divided into two groups. One group had Italian as their dominant language, while the other had German, but they all grew up in one parent – one language homes and regularly used both languages until school age.
Kupisch tested four distinct hypotheses about cross-linguistic influence: abstract economy conditions, structural overlap, semantic overlap, and dominance. The abstract economy conditions hypothesis predicts that German will influence Italian and speakers will overuse and overaccept bare NPs in Italian, especially in generic contexts. The structural overlap hypothesis predicts no influence to occur whatsoever. The semantic overlap hypothesis predicts that German will influence Italian and speakers will overinterpret definite plural and mass DPs in Italian as specific, as this interpretation is possible in Italian and German. The dominance hypothesis predicts that Italian will only be affected when German is the dominant language.
Kupisch used an acceptability judgment task and a truth value judgment task. In the first one, participants had to judge forty-two items with a specific or a generic subject context, with article usage being the cue for grammaticality. Of these items, thirty-four had a generic interpretation, while eight had a specific one. The generic interpretation items were split in two, with seventeen containing grammatical definite subject NPs, while the other half contained ungrammatical bare subject NPs or inappropriate indefinite-marked subject NPs. The specific interpretation items tested whether or not speakers accept a definite NP if the preceding context was biased towards a specific interpretation. In the truth value judgment task, they presented participants with twelve coloured pictures where every picture showed three objects or characters of one kind, with two anomalies in each one. Participants then had to judge thirty-six sentences, all of which were grammatical if they were true or false. Half of these sentences were true in connection with the picture but not the facts, and vice versa. Their goal with this was to test how the participants will interpret the definite subject NP in relation to the picture. In the case of canonical properties, if a participant judged a sentence as true, then they interpreted the definite subject generically. If they interpreted it as false, they interpreted it specifically. Conversely, in the case of atypical properties, as shown in the pictures, if the participants judged the sentence as true, they interpreted it generically. If they judged it as false, they interpreted it specifically. As a control, they included sentences with demonstratives and singular determiners.
Regarding genericity, their results in the acceptability judgment task showed that German-Italian bilinguals with Italian as the dominant language performed at ceiling for all conditions, as opposed to those with German as the dominant language which had problems with the generic bare subject NPs. The truth value judgment task also showed a difference between the two groups, even though it was not statistically significant. The German-Italian bilinguals with Italian as the stronger language had a stronger tendency to judge definite DPs as generic, unlike those with German as their dominant language. These results suggest that the abstract economy conditions hypothesis determines cross-linguistic influence, as well as dominance.
As to specificity, their results showed that definite subject NPs were more readily interpreted as specific in Italian by the bilingual group whose dominant language was German but not in the group whose dominant language was Italian, suggesting that the semantic overlap hypothesis determines cross-linguistic influence.
It is unclear why what determines cross-linguistic influence is different between the two categories, but we only mention this in passing as it is outside of the scope of our work.

4. Macedonian and its Strategies of Expressing Genericity

As mentioned in the introduction, Macedonian does not fit into the typological criteria of the NMP, as it allows both bare and definite plural NPs to have a generic reference. We will not attempt to develop or argue for an analysis of Macedonian in the NMP or pursue an alternative framework such as Dayal’s (2004), which builds on Chierchia’s. Neither will we attempt to see how Macedonian behaves according to Carlson’s (1977) diagnostics. We leave that for future work.
To be explicit, Macedonian is not the only language to behave in this regard. Brazilian Portuguese also does this. For a theoretical overview of Brazilian Portuguese and how it expresses genericity, readers can refer to Schmitt and Munn (1999, 2002) and Munn and Schmitt (2001, 2005). For an experimental investigation into how Brazilian Portuguese speakers interpret such constructions, readers can refer to Ionin et al. (2011). German is another language that allows the article option for generic reference, despite belonging to the Germanic language family. Although, this possibility is not universally acceptable across the whole language but more so in some dialects and is a result of other factors (Kupisch and Barton, 2013). We will not look into how Macedonian differs from them and why but only wish to mention that this pattern exists across languages, although it is rare. Our focus is to tackle the conditions under which bare and definite plural subject nouns refer to kinds if they are the external argument in transitive sentences, intending to eliminate all possible confounds for our experiment.
We will also explain how generic reference occurs with kind-referring subjects in copular constructions, intending to show that we do not fully understand their capacity for generic reference and why we will not use them in our experiment despite being commonly used in most of the literature. Furthermore, we will specify if singular bare, indefinite or definite nouns can refer to kinds in any of the constructions discussed in Macedonian but not French or English.
As mentioned in the introduction, we propose an experiment to test the semantic overlap hypothesis but not the structural overlap one. We will see that Macedonian patterns English syntactically, as the latter allows both bare and definite plural nouns like the former in these constructions. The difference between them is the semantic interpretation they can assign to the structure at hand. English bare plurals can have a generic reference, and definite plurals only have the specific one. Macedonian bare and definite plurals can have a generic reference, but only one option is available in English or French. Therefore, when we determine whether English and French are in a subset-superset relationship with Macedonian, we do so concerning the generic reading and with which option they actualize it, not syntactic structure. In determining the subset-superset relationship, we will only consider bare and definite plural nouns, not singular ones.
We will refer to the conclusions of this section again in the materials section. We acknowledge that our forthcoming conclusions are tentative, as a precise overview of more data is necessary. For the judgments given here, we consulted with speakers informally.

4.1 Transitive Constructions

In transitive constructions with animate external arguments, Macedonian allows generic interpretations with bare singular and plural, indefinite singular (with the indefinite article “eden”, which also functions as a numeral), and definite singular and plural nouns, but only if they are inflected with a definite article with the root t (Tomic, 2012). To explain what we mean by a definite article with the root t, Macedonian has a tripartite definite article system with three distinct roots. Articles with the root t are for anaphoric reference, articles with the root v refer to entities proximate to the speaker, and those with the root n refer to entities distant from the speaker and the addressee (Tomic, 2012). In line with that, we would like to clarify something. Every time we have referred to a noun or an NP as definite or with a definite article throughout this paper, we only meant those with the root t, as the others do not bring out a generic reading. We will continue doing so throughout the rest of our proposal.
We refer to her (2012, pp. 162-164) examples:

17. Lekar          ne      postapuva                            taka. [Generic V, Specific *]
    Physician not     behave.3Sg.Pres.Imper   so
18. Lekarot                           ne    postapuva                              taka. [Generic V, Specific V]
    Physican+the.M.Sg  not    behave.3Sg.Pres.Imper  so
19. Lekari                      ne  postapuvaat                         taka. [Generic V, Specific *]
    Physicians.Pl       not   behave.3Pl.Pres.Imper  so
20. Lekarite                           ne    postapuvaat                  taka. [Generic V, Specific ü]
    Physicians+the.Pl       not    behave.3Pl.Pres.Imper  so
21. Eden        lekar          ne     postapuva                               taka. [Generic V, Specific ü]
    A.M.Sg  physician  not    behave.3Sg.Pres.Imper  so

Of course, the examples with the indefinite singular, definite singular and plural NPs also have a specific reading.
With inanimate nouns, we disagree with Tomic (2012), who states that inanimate ones with the indefinite article cannot be kind-referring. However, the issue is with her data, as she uses proverbs (Tomic, 2012, p. 162).

22. Grom                     ne udira                                vo   kopriva. [Generic V, Specific V]
    Thunder.M.Sg  not hit.3Sg.Pres.Imper   in    nettle.F.Sg
23. Gromot                          ne udira                                vo   kopriva. [Generic V, Specific V]
    Thunder+the.M.Sg  not  hit.3Sg.Pres.Imper  in   nettle.F.Sg
24. Gromovi        ne udiraat                             vo   kopriva. [Generic V, Specific V]
    Thunder.Pl  not   hit.3Pl.Pres.Imper   in   nettle.F.Sg
25. Gromovite            ne udiraat                              vo    kopriva. [Generic V, Specific V]
    Thunder+the.Pl  not   hit.3Pl.Pres.Imper  in     nettle.F.Sg
26. *Eden         grom                      ne    udira                             vo   kopriva.
    A.M.Sg       thunder.M.Sg not   hit.3Sg.Pres.Imper  in    nettle.F.Sg

But if we take examples that are not proverbs, we find that inanimate nouns behave the same way as animate ones. Speakers we have consulted informally agreed with our judgments.

27. Nepodmackana            vrata                 ispusta
    Unlubricated.F.Sg      door.F.Sg        produce.3Sg.Pres.Imper
    zvuk. [Generic V, Specific V]
    sound.M.Sg
28. Nepodmackanata                 vrata                ispusta
    Unlubricated+the.F.Sg      door.F.Sg       produce.3Sg.Pres.Imper
    zvuk. [Generic V, Specific V]
    sound.M.Sg
29. Nepodmackani    vrati         ispustaat
    Unlubricated.Pl  door.Pl.  produce.3Pl.Pres.Imper
    zvuk [Generic V, Specific V]
    sound.M.Sg
30. Nepodmackanite         vrati        ispustaat
    Unlubricated+the.Pl  door.Pl  produce.3Pl.Pres.Imper
    zvuk. [Generic V, Specific V]
    sound.M.Sg
31. Edna         nepodmackana           vrata                ispusta
    A.F.Sg      unlubricated.F.Sg      door.F.Sg.      produce.3Sg.Pres.Imper
    zvuk. [Generic V, Specific V]
    sound.M.Sg

In these constructions, English only allows bare plurals and French only definite plurals to have a generic reference.

32. Gym-goers eat a lot of meat. [Generic V, Specific *]
33. The gym-goers eat a lot of meat. [Generic *, Specific V]
34. *Sportifs.          mangent       beaucoup de  viande.
    Athletes.M.Pl   eat.3Pl.Pres   a lot          of   meat.F.Sg
35. Les             sportifs          mangent          beaucoup de viande. [Generic V, Specific V]
    TheM.Pl  athletesM.Pl  eat.3Pl.Pres   a lot            of  meat.F.Sg

Subsequently, English and French are in a subset-superset relationship with Macedonian in these conditions.
Focusing on possible confounds for our experiment in distinguishing generic from non-generic contexts, while the use or absence of a definite article is a reliable cue cross-linguistically, there are other cues on which speakers can rely. Gelman and Raman (2003) have looked into which ones English-speaking children use. They identified morphosyntactic (tense, aspect and definiteness), pragmatic, and world-knowledge cues. Because the definite article is not as reliable in Macedonian, as speakers can make and infer generic reference with or without it, it is vital to ensure that the other cues do not act as confounds in our experiment. With that in mind, we will only use the 3rd person plural present imperfective forms of the verbs in our materials, as those forms never inhibit a generic reading. Due to brevity, we will not go into this in greater detail. We do not focus on pragmatic and world-knowledge cues as they are irrelevant to our experiment.

4.2 Genericity in Copular Constructions

Copular constructions are significantly more problematic as speakers’ intuition differ from standard accounts. According to Tomic (2012), subjects in these structures can only be kind-referring if they are definite singular or plural, and any other type of subject NP is not kind-referring. We illustrate this with the following pair of examples.

35. Mravkata                e                        insekt. [Generic V, Specific V]
    Ant+the.N.Sg       be.3Sg.Pres  insect.M.Sg
36. Mravkite              se                  insekti. [Generic V, Specific V]
    Ant+the.F.Pl      be.3Pl.Pres  insect.M.Pl

About copular constructions with either predicative adjectives or nominals, speakers we have consulted informally judged definite singular and plural NPs to be kind-referring. Contra Tomic (2012), the same speakers found bare singular and plural NPs to have a generic reference. Copular sentences with predicative adjectives and definite singular subject NPs received mixed judgments. Some speakers found them to be strictly specific, while others also allowed them to have a generic reference (hence two sets of brackets with example 40). All speakers interpreted copular sentences with predicative nominals and a definite singular subject NP as having a generic reference alongside the specific reading. They never interpreted copular constructions with an indefinite singular subject NP as generic, no matter the predicate type. We first give the examples with predicative adjectives and then those with nominals.

39. Pajak                e               ubav. [Generic V, Specific *]
    Spider.M.Sg  be.3Sg  pretty.M.S
40. Pajakot                       e              ubav. [Generic V, Specific V], [Generic *, Specific V]
    Spider+the.M.Sg   be.3Sg  pretty.M.Sg
41. Pajaci          se           ubavi. [Generic V, Specific *]
    Spider.Pl  be.3Pl  pretty.Pl
42. Pajacite              se            ubavi. [Generic V, Specific V]
    Spider+the.Pl  be.3Pl  pretty.Pl
43. Eden         pajak                e           ubav. [Generic *, Specific V]
    A.M.Sg  Spider.M.Sg  be.3Sg  pretty.M.Sg
44. Kuce            e              cicac. [Generic V, Specific V]
    Dog.N.Sg  be.3Sg  mammal.M.Sg
45. Kuceto                   e            cicac. [Generic V, Specific V]
    Dog+the.N.Sg  be.3Sg  mammal.M.Sg
46. Kucinja     se           cicaci. [Generic V, Specific V]
    Dogs.Pl   are.3Pl  mammals.Pl
47. Kucinjata        se              cicaci. [Generic V, Specific V]
    Dogs+the.Pl  are.3Pl   mammals.Pl
48. Edno     kuce              e           cicac. [Generic *, Specific V]
    A.N.Sg  dog.N.Sg  be.3Sg  mammal.M.Sg

In copular constructions with complex predicative nominals (of the type adjective plus noun), speakers we consulted found both bare and definite singular and plural nouns to refer generically. With indefinite singular nouns, speakers’ judgments differ. Some found them to have only specific reference, while others judged them as having both generic and specific reference (we use two sets of brackets to indicate this with example 53) as opposed to the copular sentences with predicative adjectives and nominals which were judged as strictly specific.

49. Slon                         e           golemo    zivotno. [Generic V, Specific V]
    Elephant.M.Sg  be3Sg  big.N.Sg  animal.N.Sg
50. Slonot                              e             golemo    zivotno. [Generic V, Specific V]
    Elephant+the.M.Sg  be.3Sg  big.N.Sg  animal.N.Sg
51. Slonovi               se          golemi  zivotni. [Generic V, Specific V]
    Elephants.Pl  be.3Pl   big.Pl   animals.Pl
52. Slonovite                   se         golemi  zivotni. [Generic V, Specific V]
    Elephants+the.Pl  be.3Pl   big.Pl   animals.Pl
53. Eden         slon                   e
    A.M.Sg  elephant.M.Sg   be.3Pl
    golemo      zivotno. [Generic V, Specific V], [Generic *, Specific V]
    big.N.Sg  animal.N.Sg

In these structures, with plurals, English only allows bare NPs to be kind-referring, whereas French only definite NPs.

54. Dogs are mammals. [Generic V, Specific *]
55. Les                chiens         sont                des                mammifères. [Generic V, Specific V]
    The.M.Pl   dogs.M.Pl be.3Pl.Pres of+the.M.Pl  mammals.M.Pl

Due to the divided judgments we received from speakers about these sentences, we refrain from exactly concluding how the subset-superset relationship is as more data and precise discernment are necessary.

5. Experiment

5.1 Participants

By the 2021 Australian census (“Macedonian Australians”, 2023) there are 111,352 Australians with Macedonian ancestry, 41,786 of whom are Macedonian-born. By the 2021 Canadian census (“Macedonian Canadians”, 2023), there are 39,435 Canadian with full or partial Macedonian ancestry.
We will screen our participants by doing a standardized questionnaire and proficiency tests.
We adopt Dunn and Fox Tree’s (2009) bilingual dominance scale questionnaire for the first screening test. To test speakers’ proficiency in their dominant language, we propose using cloze tests, a lexical decision task, and a task to test mean length of utterance (MLU).
To test speakers’ proficiency in their heritage language is much more difficult as these tests do not exist for Macedonian, to the best of our knowledge. As such, we would have to make our tests to assess their knowledge, which is extremely difficult. Moreover, these tests must be standardized, and the results divided into quotient groups to offer a precise and unbiased assessment.
Another option is to use the Centre for Testing and Certifying Macedonian as a Foreign/Second Language tests. These tests check the following four language skills: listening comprehension, reading comprehension, writing, and speaking. However, we cannot access these tests as they are not available to the public, so we cannot review them and see how they test these skills or if the results will be accurate and statistically usable.
We will only include participants who are intermediate or above speakers based on the proficiency tests. Ideally, we would have 20-30 participants per group for statistical power, but we need to gather and screen them to see if this is possible. We do not know how many generations of Macedonians live in these two countries or their age group. If possible, we would include multiple cross-generational groups to see the trajectory of divergent interpretations across time. As control groups, preferably, we would have both homeland and baseline speakers. We will include only monolinguals in the homeland control group, and it would be an easy task to find them. The second group would be more problematic, but it is manageable to find enough speakers as Macedonians still routinely move to Australia and Canada.

5.2 Materials

We adopt Serratrice et al.’s (2009) design for our experiment. About the number of items, we propose to have ten pairs of target sentences and five pairs of filler sentences, thus having twenty target pairs per language, of which twenty will have a specific context and twenty a generic one. In the Macedonian task, all target items will be grammatical. The use or absence of the definite article decides the grammaticality of the sentences in the French task, and in doing so, we will have ten grammatical and ten ungrammatical in both contexts. In the English task, the items for the generic context will be either grammatical or ungrammatical based on article usage, while all target items in the specific context will be grammatical.
Before each sentence, we will present participants with a picture or object prototypical of the noun with a generic reference in the item. We will have for every picture a pair of sentences: one with a generic context and another with a specific context. Same as in Serratrice et al.’s experiment, we will manipulate the presence or the absence of the definite article for English. In the Macedonian and French material, we will use the adjunct as a contextual cue that determines the semantic context of the sentence. However, we will strengthen the contextual cue for specific and generic contexts. As we saw in Serratrice et al’s experiment, despite the adjunct for the generic context being a clear indicator of the semantic context, the participants themselves did not integrate it with the semantic meaning of the definite article. Due to this, we will include one more introductory phrase in all sentences to strengthen the semantic context of the sentence. For the generic condition, we will also include “Kako rod/As a species/En tant qu’espèce” alongside the introductory phrase “Generalno/In general/En général”. We will modify the second introductory phrase based on the properties of the kind-referring entity, whether it is an animal, human, insect, or a group of some sort. For the specific condition, alongside the introductory phrase “Tuka/Here/Ici”, we will include a second phrase based on the temporal or spatial properties of the activity expressed in the sentence, something like “Momentalno/In this moment/À ce moment”.
We differ from Serratrice et al.’s (2009) experiment in one more way. We will not use copular constructions as we do not understand their capacity for generic reference well enough in Macedonian and whether French and English are in a subset-superset relationship with Macedonian regarding them, as we explained in section 4. 2. We will strictly use transitive sentences with a kind-referring plural subject. To exemplify our materials:

Generic context:

70a. Generalno, kako grupa,    advokati   sakaat                            skapi             koli. (Grammatical)
Generally   as       group.F  lawyer.Pl like.3Pl.Pres.Imper expensive.Pl cars.Pl
70b. Generalno, kako grupa,    advokatite
Generally   as     group.F  lawyer+the.Pl
sakaat                             skapi              koli. (Grammatical)
like.3Pl.Pres.Imper expensive.Pl  cars.Pl
71a. Generally, as a group, lawyers like expensive cars. (Grammatical)
71b. *Generally, as a group, the lawyers like expensive cars. (Ungrammatical)
72a. *Généralement, en tant que groupe,       avocats
Generally            as          a          group.M.   lawyer.M.Pl
aiment                                 les           voitures              chères. (Ungrammatical)
like.3Pl.Pres.Imper        the.M.Pl  expensive.F.Pl  cars.F.Pl
72b. Généralement, en tant que groupe,
Generally         as        a     group,M.Pl
less             avocats         aiment                         less            voitures            chères. (Grammatical)
the.M.Pl lawyer.M.Pl like.3Pl.Pres.Imper the.F.Pl. expensive.F.Pl cars.F.Pl

Specific context:

73a. Tuka,  vo ovaa  fabrika,         inzeneri
Here,  in   this   factory.F.Sg engineers.Pl
izmisluvaat                       novi      dizajni. (Grammatical)
invent.3Pl.Pres.Imper new.Pl designs.Pl.
74a. Tuka, vo   ovaa fabrika,         inzenerite
Here,  in   this   factory.F.Sg engineers+the.Pl
izmisluvaat                        novi     dizajni. (Grammatical)
invent.3Pl.Pres.Imper new.Pl designs.Pl.
75a. Here, in this factory, engineers invent new designs. (Grammatical)
75b. Here, in this factory, the engineers invent new designs. (Grammatical)
76a. *Ici,      dans cette  usine,            ingénieurs
Here,  in      this   factory.F.Sg  engineers.M.Pl
inventent                             de   nouveaux  designs. (Ungrammatical)
invent.3Pl.Pres.Imper  of     new.M.Pl  designs.M.Pl.
76b. Ici,     dans  cette  usine,                 les               ingénieurs
Here,  in      this    factory.F.Sg  the.M.Pl      engineers.M.Pl

inventent                     de     nouveaux      designs. (Grammatical)
invent.3Pl.Pres.Imper  of    new.M.Pl       designs.M.Pl

The participants task will be to judge whether or not the sentences are acceptable in their respective languages on a Likert scale of 1-7.

5.3 Procedure

We will divide participants into two groups to minimize the possibility of a carryover effect. One group will do the majority language version of the experiment, while the other only the heritage language one. Before doing the experiment, every participant will do Dunn and Fox Tree’s (2009) bilingual dominance scale questionnaire and the proficiency test for their respective languages. To minimize the yes-bias (Polinsky, 2018), we recommend that a corresponding heritage speaker should test the participants and only speak to them in the language of the corresponding test or experiment.
Before the experiment, the participants will receive instructions on what they have to do. They will do several practice tests to get accustomed and to see if they understood the instructions.

5.4 Predictions

We have a two-tailed prediction regarding how the majority language (English or French) will affect the heritage one (Macedonian).
If abstract economy conditions determine cross-linguistic influence, we predict French-Macedonian speakers to retain generic bare plural NPs in Macedonian regardless of the effect of French. Despite Macedonian being the minority language, which means less input for speakers, it still can have bare plural NPs with a generic reference due to the available type-shifter, which is more economical than projecting an overt determiner. Regarding English-Macedonian speakers, we predict that they will judge definite plural generic NPs as unacceptable due to both abstract economy conditions and the input of English, which only has the more economical bare plural generic. If semantic overlap decides the nature of cross-linguistic influence, we predict French-Macedonian speakers to interpret definite plural NPs as generic, thereby judging them as acceptable but not bare plural ones due to the effect of French. We also predict finding the opposite pattern in English-Macedonian speakers; speakers will interpret bare plural NPs as generic, judging them as acceptable but not definite plural ones due to the effect of English.
We expect proficiency to play a role, with advanced speakers performing similarly to the homeland and baseline speakers. We also presume frequency of use, frequency of input, and knowledge of other languages to have an effect. We predict a cross-generational effect, with the majority language more strongly affecting the judgments of each group following the first-generation one. Lastly, we predict the extent of formal teaching in Macedonian to play a role.

5.5 Limitations

The greatest limitation of our proposal is that we do not have a formal analysis of genericity in Macedonian in the scope of Chierchia’s NMP (1998) or a precise descriptive survey of what plays a role in various constructions whether nouns are kind-referring. Another issue is how speakers more easily interpret bare singular and plural NPs as generic, while the specific reading is much more dominant with the indefinite singular and definite plurals. We also do not understand how available bare arguments are in specific contexts. As noted in the materials section, 73a is also acceptable for the speakers we consulted. If other homeland speakers find them acceptable, it would be interesting to see how the heritage speakers behave because of the effect of the majority language, specifically French where this is ungrammatical. Due to this, there might be possible confounds in our experiment’s design that we cannot foresee.
We also did not focus on generic reference with “have” constructions and kind-referring objects due to conciseness, but as we have mentioned throughout section 4, speakers’ intuitions diverge from standard accounts, and it is most likely that this holds for these cases as well.
Another limitation is the non-existence of assessment tests for Macedonian. As far as we know, these tests have never been used for Macedonian whatsoever, even in monolingual populations.
Moreover, from what we know, there has been no linguistic analysis of the language of Macedonian heritage speakers, neither in Canada nor Australia. After all, the minority language of heritage speakers goes through either divergent acquisition or language attrition (Polinsky, 2008). It could be the case that we would include a particular form or morpheme that has gone through significant change in the Macedonian of Australian or Canadian Macedonians. If that is so, the reason for them judging a sentence as unacceptable is not due to the interpretation they have but another unforeseen factor.

6. Conclusion

As we showed in this proposal, despite its limitations and difficulties, Macedonian and its heritage speakers provide novel ground as it allows us to test the hypotheses for cross-linguistic influence in a new way. It is also essential to have a formal analysis of genericity in Macedonian and a better understanding of the heritage Macedonian of both the Australian English-Macedonian and Canadian French-Macedonian populations. If this is accomplished, it could help us answer whether abstract economy conditions, semantic overlap, or another third factor decide cross-linguistic influence.

References

Benmamoun, E., Montrul, S. & Polinsky, M. (2013). Heritage languages and their speakers: Opportunities and challenges for linguistics. Theoretical Linguistics, 39(3-4), 129-181.

Carlson, G. N. (1977). A unified analysis of the English bare plural. Linguistics and Philosophy, 1(3), 413–457.

Chierchia, G. (1998). Reference to Kinds Across Languages. Natural Language Semantics 6, 339-405.

Dayal, V. (2004). Number Marking and (in)Definiteness in Kind Terms. Linguistics and Philosophy, 27(4), 393–450.

Dunn, A. & Fox-Tree, J. 2009). A quick, gradient Bilingual Dominance Scale. Bilingualism: Language and Cognition, 12(3), 273-289.

Hulk, A., & Müller, N. (2000). Bilingual first language acquisition at the interface
between syntax and pragmatics. Bilingualism: Language and Cognition, 3(3), 227 – 244.

Ionin, Tania, Silvina Montrul, and Helade Santos. “An experimental investigation of the expression of genericity in English, Spanish and Brazilian Portuguese.” Lingua, vol. 121, 2011, 963-985.

Kupisch, T., & Pierantozzi, C. (2010). Interpreting Definite Plural Subjects: A Comparison of German and Italian Monolingual and Bilingual Children. Boston University Conference of Language Development, 2009, 245-254.

Kupisch, T. (2012). Generic subjects in the Italian of early German-Italian bilinguals and German learners of Italian as a second language. Bilingualism: Language and Cognition. 736-756.

Kupisch, T., & Barton, D. (2013). Generic Reference in Adult German Bilinguals: How Bilinguals Deal with Variation: Studia Linguistica 67(1), 1-27.

Minova-Gjurkova, Lj. (1994). Sintaksa na Makedonskiot Standarden Jazik. Rading.

Montrul, S., & Ionin, T. (2010). Transfer effects in the interpretation of definite articles by Spanish heritage speakers. Bilingualism: Language and Cognition, 13(4), 449-473.

Montrul, S., & Ionin, T. (2012). Dominant Language Transfer in Spanish Heritage Speakers and Second Language Learners in the Interpretation of Definite Articles. The Modern Language Journal, 96(1), 70–94.

Müller, N., & Hulk, A. (2001). Crosslinguistic influence in bilingual language acquisition: Italian and French as recipient languages. Bilingualism: Language and Cognition, 4(1), 1-21.

Munn, A. L., & Schmitt, C. (2002). Bare nouns and the morphosyntax of number. Amsterdam Studies in the Theory and History of Linguistic Science, 225–239.

Munn, A. L., & Schmitt, C. (2005). Number and indefinites. Lingua, 115(6), 821–855.

Polinsky, M. (2006). Incomplete Acquisition: American Russian. Journal of Slavic Linguistics, 14(2), 191–262.

Polinsky, M. (2018). Heritage Languages and Their Speakers. Cambridge University Press.

Serratrice, L., Sorace, A., Filiaci, F., & Baldo, M. (2009). Bilingual children’s sensitivity to specificity and genericity: Evidence from metalinguistic awareness. Bilingualism, 12(2), 239 – 257.

Schmitt, C., & Munn, A. L. (1999). Against the Nominal Mapping Parameter: Bare nouns in Brazilian Portuguese. North East Linguistics Society, 29(1), 339 – 355

Schmitt, C., & Munn, A. L. (2002). The syntax and semantics of bare arguments in Brazilian Portuguese. Linguistic Variation Yearbook, 2, 185–216.

Tomić, O. M. (2012). A grammar of Macedonian. Slavica.

Canadian Australians. (2023, February 19). In Wikipedia. https://en.wikipedia.org/wiki/Canadian_Australians

Macedonian Australians. (2023, March 19). In Wikipedia. https://en.wikipedia.org/wiki/Macedonian_Australians