A Plural Indefinite Article in Heritage Greek: The Role of Register

Abstract

This paper investigates the use of kati “some” by Greek Heritage Speakers (HSs) in comparison to monolinguals. While all Greek determiners are marked for gender, case, and number, and agree with their nominal complement, kati is an exception, as it lacks agreement and combines only with plural nouns. Building on the existing literature, we show that its function is to remain vague about the number of individuals/entities denoted. Our hypothesis is that vague language (VL) is a feature of informal conversations and of the spoken language. To this end, we conducted a study in which Heritage Speakers of Greek and monolingual speakers of Greek participated in a production task held in two distinct settings and modalities. In addition, we performed corpus searches to see how both monolingual and Heritage Speakers use kati. The results show that monolingual speakers do indeed prefer kati in the informal register, while Heritage Speakers overgeneralize its use across registers. Our findings confirm the use of vague language in informal registers and oral modality and support claims in the literature on register levelling by Heritage Speakers. Focusing on monolinguals’ repertoire, a judgment task with different levels of formality was additionally performed. These results in principle align with our hypothesis and signal that neither frequency nor other informality contexts trigger a higher rate for kati.

1. Introduction

This paper is concerned with the properties and the distribution of the indefinite determiner kati “some” in Greek. While kati is mentioned in some work on Greek indefinite NPs, see e.g., Haspelmath (2001); Lazaridou-Chatzigoga (2009); Giannakidou (2012); Alexopoulou and Folli (2019); and Etxeberria and Giannakidou (2021), among others, systematic studies of this determiner are scarce, with Tsoulas (2021) being a notable exception. The behavior of kati is intriguing, as it combines uniquely with plural nouns, although diachronically, as Holton et al. (2019) report, other combinations were also possible. Building on Tsoulas and Etxeberria and Giannakidou, and on the basis of a production study, corpus investigations, as well as experimental data, we will show that kati is a marker of vague language and is favored in informal contexts by monolingual speakers. We will then turn to examine its use by Heritage Speakers, which points to an area of register levelling.

Our study breaks new empirical ground in many areas. First, we investigate monolingual productions and corpora, and conduct a judgment study focusing on monolingual’s repertoires, shedding light on the relation between kati and different levels of formality. Second, we compare monolingual speakers of Greek to Heritage Speakers of Greek (HSs). In the literature, HSs are defined as speakers who are bilingual with a minority language at home in addition to the majority language in the larger environment they live in (Montrul and Polinsky 2011). They have emerged as an important population to study, as they provide insights into the concept of nativeness as well as into language variation and change. Furthermore, in the acquisition of two or more languages, various metalinguistic factors such as the age of onset to bilingualism, the qualitative and quantitative input, the instruction of these languages in an educational setting and the generation and the socioeconomic status of the parents have been argued to play an important role (Unsworth et al. 2014; Flores et al. 2017; Meir and Armon-Lotem 2017; Daskalaki et al. 2019; Dosi and Papadopoulou 2019; Kupisch 2019, among others). Heritage languages are spoken usually within the core members of the family and are characterized by a casual, conversational speech style focused on everyday topics, leading to a gradual register narrowing (Dressler 1991, pp. 101–2; Chevalier 2004). Evidence concerning register levelling in heritage languages has recently been discussed in Wiese et al. (2022) for a variety of phenomena and heritage languages (see also Rothman 2007 and Polinsky 2018); one such phenomenon presented in Wiese et al. is the overgeneralization of the indeclinable complementizer pu for the introduction of relative clauses in Greek. Another is the use of light verb constructions for the expression of the perfective Aspect instead of lexical verbs (Alexiadou and Rizou, forthcoming). The comparative study of kati in different varieties of Greek will add to this literature and will be very informative about processes of register levelling as well as language interference, as we study HSs of Greek in different majority language contexts. In addition to these factors, tracking the psychological personality profile of participants has brought revolutionary and well interpretable findings in different scientific fields (neuropsychiatry: Atak 2013; medical field: Karfopoulou 2016; L2 learners: Chen et al. 2021). In the aforementioned studies, the variation of the results can be explained on the basis of The Big Five Personality Test (extended to Ten Item Personality Inventory) by tracing the aspects of the participants’ characters. Thus, we can correlate our findings to the personality traits of our participants, contributing to the study of social meaning (Beltrama 2020).

The paper is structured as follows: in Section 2, we discuss the properties of kati in Greek and conclude that it appears to be a marker of vague language. We present our research questions in Section 3, while in Section 4, we discuss our production study. In Section 5, we report the results of our corpora investigation, and in Section 6, we present our judgment study. Finally, in Section 7, we present the conclusion.

2. Kati as a Marker of Vague Language

Kati is an indefinite element, which has both a determiner and a nominal1 function. In this paper, we will only deal with the former use. Kati shows no agreement in case and gender, unlike all other Greek determiners, and is compatible with nouns in all genders and cases, the condition being that these appear in the plural (1)–(3).

(1a)	irthan	kati	pedia
	came.3PL	some	children.N.PL
	Some children came
(1b)	ida	kati	ginekes
	saw.1SG	some	women.F.PL
	I saw some women

(2)	kati	pedia/	pedion
	some	children.ACC/	children.GEN

(3)	*ida	kati	gineka/	kati	pedi
	saw	some	woman.SG/	some	child.SG

As Tsoulas (2021) states, even mass nouns must appear in the plural in the context of kati (4). We note that the abundance of reading associated with plural mass nouns in Greek (see Alexiadou 2011) is not present in this case.

(4)	agorasa	kati	krasia
	Bought.1SG	some	wines
	I bought some wines

Tsoulas remarks that the plurality requirement is puzzling, as there is nothing in the morphology of kati that would require plurality. Etxeberria and Giannakidou (2021) suggest that because kati is vague, plurality is associated with this feature.

Tsoulas convincingly shows that kati behaves like a plural indefinite determiner by applying Le Bruyn and Pozas-Loyo’s (2015) four properties associated with indefinite articles (singular or plural). According to Le Bruyn and Pozas-Loyo (op.cit.), the first property diagnoses ties with the number scale, e.g., whether a particular article can figure as an answer to a how many question. As Tsoulas points out, kati bears no connection to the number scale, since it is ungrammatical when used as an answer to a “how many” question (5a), behaving like the English a and not like the English one. The second property is related to the first one and concerns partitive implicatures. In (5b) kati, like English a, does not imply that there were students who did not come to see me:

(5a)	A:	Posi	fitites	irthan	na	se	dun;
	A:	How-many	students.PL	came.3PL	to	you	see?
	How many students came to see you?
	B:	*Kati	fitites	irthan	na	me	dhun.
	B:	Some	students.PL	came.3PL	to	me	see.
	Some students came to see me.

(5b)	Kati	fitites	irthan	na	me	dun.
	Some	students.PL	came.3PL	to	me	see.
	Some students came to see me.

The third property relates to compatibility with other operators, building on Farkas (2002). Indefinite articles are compatible with any operator, unless this operator requires a plural. Singular indefinite articles are fine in the scope of a generic operator, while kati cannot be used in generic contexts, as shown in (6); this is a property which, as Tsoulas mentions, needs to be explained, and one which we will examine towards the end of this section.

(6)	*kati	fokies	ine	eksipnes
	some	seals.PL	are	intelligent

The final property concerns the complementarity in distribution with covert type shifting devises (Chierchia 1998). As Tsoulas discusses at length, Greek allows the presence of bare plural arguments in contexts that escape this restriction, namely coordinated bare DPs, focus positions, and lexically governed positions, thus leading him to conclude that kati is indeed a plural indefinite determiner.

In order to substantiate the point that kati is a D° element, i.e., a head and not a phrasal element, Tsoulas discusses two further contexts: first, kati cannot co-occur with other determiners or demonstratives, as shown in (7):

(7)	*afta/*ta/kati	pedia
	these/the/some	children.PL

Second, unlike other indefinite phrasal elements and numerals in Greek, kati does not allow ellipsis of its nominal complement. As Tsoulas points out, this property of kati is also shared by the definite determiner, leading him to conclude that kati must be a plural indefinite D° element, see (8a–c):

(8a)	*kati	pedia	epezan	ke	kati	∅ tragudusan
	some	children	played	and	some	sang
(8b)	merika	pedia	epezan	ke	merika	∅ tragudusan
	some.PL	children	played	and	some.PL	sang
(8c)	*ta	pedia	epezan	ke	ta	∅ tragudusan
	the	children	played	and	the	sang

Looking at the semantic contribution of kati, Giannakidou (2012, p. 303) as well as Etxeberria and Giannakidou (2021) have argued that Greek kati is similar to Spanish unos, while Greek kapj—‘some’ + NP_plural—is similar to Spanish algunos, cf. Martí (2008). The point that Etxeberria and Giannakidou make is that kati introduces vagueness and not ignorance. The following provides evidence for this: as shown in (9), the NP is found as a complement to a cognitive verb like ksero “know”; it can hardly be suggested that kati pedia “some children” denotes complete ignorance (Tsoulas 2021). Rather, in (9), the speaker implies that they know some specific children, but they do not want to name them, remaining vague, see Etxeberria and Giannakidou (2021):

(9)	Ksero	kati	pedia	pu	zografizun	poli	orea
	know.1SG	kati	children	that	draw	very	nice
	I know some children who draw very nice

Recall that we pointed out that kati does not allow for generic readings. The reason for this is that kati has an existential meaning and seems similar to English a + few, as described in Solt (2009). Specifically, Solt claimed that a + few is defined by its lower bound, thus it has an existential meaning (it must be non-zero) which corresponds to at least a small number. Solt points out that sentences such as a few students came to the party can be continued with “in fact many did”, but not with “in fact no one did”. Similar observations hold for kati. For Solt, a few is purely cardinal; it specifies a small number of individuals. Solt further shows that a in a few does not combine with numerals and the quantitative adjective many (10a,b) and (10c,d). The reason for this, according to Solt, is because a makes a semantic contribution so that few and a few have distinct semantics. We note here a similar distribution for kati:

(10a)	*A many students came to the party.
(10b)	*A three students came to the party.

(10c)	*kati pola vivlia
	*some many books
(10d)	agorasa	*kati tria vivlia
	bought.1SG	*some three books
	I bought some three books

Although kati specifies a small number of individuals, its cardinality nevertheless remains vague. For Etxeberria and Giannakidou (2021), it is the prefix ka- that introduces vagueness in kati and the plurality requirement.2 Kati is vague as it lacks sharp boundaries, see Veloudis (2017). Specifically, building on Delveroudi (1989), Veloudis (2017, p. 143) argues that by using certain prefixed ka- elements, the speaker “gives the impression that she wants to refrain not simply from accurate quantity estimations but from her very conversational duty to refer with accuracy, especially regarding well-known referents.”; in support of its vague nature, kati behaves as a positive polarity item, as shown in Tsoulas (2021). As illustrated in (11), kati cannot appear in the context of negation, a characteristic which is also shared with, e.g., vague degree attenuators, see Solt (2021) for discussion. Importantly, kati as a determiner is an alternative and weaker expression of cardinality: speakers use kati to make a different contribution than the one they would make by using a cardinal or a definite article; the exact number of referents remains imprecise.3

(11)	*den	ida	kati	klidia
	NEG	saw.1SG	some	keys
	I didn’t see any keys

To conclude this section, the function of kati is to remain vague about the number of individuals/entities denoted. Thus, kati is a marker of vague language (henceforth VL). VL is used to avoid an excess of precision, or to achieve “imprecision” or “imprecise language use” (Crystal et al. 1975, pp. 112–4). As Brown and Yule (1983, pp. 8–9) pointed out, spoken language usually contains a lot of “general, non-specific” vocabulary. As McCarthy (2003, pp. 108–18) states, VL contributes to “naturalness and the informal, convergent tenor of everyday talk”. VL allows “a speaker to take refuge in strategic imprecision”, according to Leech (2000, p. 695). This leads to the hypothesis that speakers would choose the vague alternative in informal contexts.

3. Research Questions

We now turn to the exploration of our hypothesis that kati is a marker of VL and is hence associated with informal registers. To this end, we will examine its use in different groups of Greek speakers, through production and judgment tasks, in order to probe into the interaction of register variation with kati. We carried out two studies and complemented our investigations with corpora searches for the use of kati.

The aim of our production study is to explore how native speakers of Greek use kati and whether HSs deviate from monolingual speakers regarding the production of kati in different communication situations. Given the fact that HSs tend to overgeneralize features that are attributed to specific registers due to a lack of knowledge that could be transmitted via formal education and due to their exposure to less formal communication settings (Rothman 2007; Montrul and Polinsky 2019), we predict that HSs will overgeneralize the indefinite determiner kati in formal registers, unlike monolingual controls. More specifically, based on Dressler’s (1991) and Chevalier’s (2004) claims that heritage languages are usually spoken among family members, those speakers are not exposed to any other register variation but the informal one. As the latter variation characterizes the vernacular, namely the everyday oral speech, vagueness is one of the most salient elements of it, and the indefinite determiner kati is sui generis vague, as we have proposed in Section 2. Thus, we expect monolinguals to avoid its use in formal settings, as they are sensitive to register variation due to their exposure and their acquaintance with different communication settings. Comparatively, HSs are expected to exhibit a register levelling regarding the production of kati, due to their lack of sensitivity to register variation. On this basis, we aim to explore to which extent HSs’ productions indeed deviate from those of monolinguals. Furthermore, regarding the modality, we expect monolinguals to produce more kati instances in the spoken mode rather than in the written one. Concerning HSs, we expect modality to not play such an important role as register does. Based on the previous literature, we hypothesize that HSs will overgeneralize the everyday informal speech, regardless of the modality variation. Thus, we also explore whether spoken and written registers vary among monolingual and HSs with respect to the determiner kati.

Investigating further monolinguals’ repertoires, we explore whether different determiners are preferred in different communicative settings. As explained in Section 2, kati is a marker of VL which characterizes informal registers. Thus, our prediction is that in informal settings and especially in informal settings with one or more markers of informality—which is the diminutive construction—the indefinite determiner kati will be preferred. With respect to the two conditions within the formal register spectrum, we expect our participants to rate kati with a higher score in the condition with the more frequent lexical items compared to the other with the less frequent items.

4. Production Study in Heritage and Monolingual Speakers of Greek

The production study targets to elicit monolinguals’ and HSs’ repertoires, using a particular methodology that provides data for a variety of phenomena, unlike controlled experiments, which tap into the explicit (written task) and implicit (oral task) knowledge of both monolinguals and HSs (Montrul 2011). In the spirit of Labov (1972), under proper circumstances, subjects can easily adapt in communication settings. Assuming that in principle there are distinct realizations associated with different levels of formality, these data allow us to see to what extent monolingual speakers and HSs can adapt.

4.1. Method and Procedure of the Production Study

According to Wiese (2017), the “language situation” setting is a method that allows researchers to elicit naturalistic data. This set-up provides comparable oral and written data and in different levels of formality (data sets 2 × 2). A short video (00:42″) of a fictional event was shown to every participant. In the video, a non-severe car accident took place in a parking lot and the task was to retell what happened to different people, with the participant imagining that they witnessed the incident. As we wanted to test how formality and mode affected production, they had to produce both an oral and a written narration in two distinct communication settings.

For the elicitation, we simulated two distinct communication settings, in which every participant took part. The elicitation of the formal setting took place in an office with an elicitor in suit. The participants sat opposite the elicitor, keeping a proper distance from each other. The elicitor of the formal register used a standardized language with honorifics when asking the participants to testify to the police in detail how the accident happened. The oral narration of the formal register involved was to leave a voice mail on the answering machine of the police department, while the written one was to type a testimony on the police laptop as they witnessed the incident.

The elicitation of the informal part took place in a different office, where the two interlocutors were sitting close to each other. The elicitor for the informal setting was casually dressed, very talkative, and did not use any honorifics. A 20 min chitchat preceded the elicitation of the data in order to become acquainted with one another. While they were chatting, the elicitor offered treats to the participants in order to make them feel comfortable. After this conversation, the main part of the study took place. In this communication situation, the participants had to narrate in a voice message in WhatsApp what happened to a close friend. The written mode was to text the same close friend about the accident, again in WhatsApp. The informal elicitor always reminded their appointment one day before via text message in a friendly and familiar way. In order to balance the elicitation process, we created 16 elicitation orders, half of which started with the heritage language and half with the majority one (see Appendix A.1). The whole process was recorded.

4.2. Participants in the Production Study

Three groups of participants took part in the production task, which was divided into two age groups each. Two groups consisted of Greek HSs, one located in Germany and one in the US (Chicago and NY), and one group consisted of monolingual speakers of Greek. Within those groups, two different age groups were tested, namely adolescents and adults.

Table 1. Metalinguistic data across groups and age groups.

		HSs in Germany		HSs in the US		Monolinguals
		Adults	Adolescents	Adults	Adolescents	Adults	Adolescents
N		27	21	31	32	32	32
		(17 Females)	(7 Females)	(18 females)	(16 females)	(16 females)	(16 females)
Chronological4 Age		Mean 28; 5	Mean 16; 3	Mean 29; 9	Mean 16; 2	Mean 27; 6	Mean 15; 3
		(min 21–max 36)	(min 14–max 19)	(min 24–max 35)	(min 14–max 18)	(min 24–max 35)	(min 13–max 18)
		SD 4.108	SD 1.717	SD 3.224	SD 1.408	SD 3.003	SD 1.755
Age of onset		Mean 2; 3	Mean 1; 3	Mean 1; 7	Mean 1; 0	-	-
		(min 0–max 8)	(min 0–max 4)	(min 0–max 6)	(min 0–max 5)	-	-
		SD 2.404	SD 1.720	SD 2.715	SD 1.692	-	-
Current use in Gr		Mean 1.0	Mean 1.2	Mean 0.8	Mean 0.7	-	-
		SD 0.455	SD 0.476	SD 0.344	SD 0.347	-	-
Self-ratings in Gr		Mean 3.9	Mean 4.2	Mean 3.8	Mean 3.3	Mean 4.8	Mean 4.8
		(min 0–max 5)	(min 0–max 5)	(min 0–max 5)	(min 0–max 5)	(min 0–max 5)	(min 0–max 5)
		SD 0.853	SD 0.865	SD 0.973	SD 0.844	SD 0.353	SD 0.269
Literacy practices in Gr		Mean 1.0	Mean 1.2	Mean 0.8	Mean 0.7	Mean 1.6	Mean 1.4
		(min 0–max 1.6)	(min 0.3–max 1.8)	(min 0.2–max 1.5)	(min 0.1–max 1.3)	(min 0.3–max 2)	(min 0.3–max 2)
		SD 0.4556	SD 0.476	SD 0.344	SD 0.347	SD 0.494	SD 0.422
Parents’ Generation5	Both 1st	18 prt	6 prt	20 prt	10 prt	-	-
	One 1st	-	-	1 prt	2 prt	-	-
	One 1st one 2nd	2 prt	1 prt	8 prt	10 prt	-	-
	One 1st one foreign	7 prt	9 prt	-	2 prt	-	-
	Both 2nd	-	1 prt	2 prt	6 prt	-	-
	One 2nd	-	-	-	1 prt	-	-
	One 2nd One foreign	-	2 prt	-	1 prt	-	-
Visits to the country of heritage		Mean 1.5	Mean 1.5	Mean 1.0	Mean 1.0	-	-
		(min 1–max 2)	(min 1–max 2)	(min 1–max 2)	(min 0–max 2)	-	-
		SD 0.506	SD 0.511	SD 0.183	SD 0.407	-	-
Years of education in the HL		Mean 6; 0	Mean 8; 5	Mean 7; 7	Mean 10; 4	-	-
		(min 0–max 12)	(min 3–max 12)	(min 0–max 12)	(min 8–max 12)	-	-
		SD 4.301	SD 2.673	SD 4.266	SD 1.319	-	-
Hours of education in the HL		Mean 5352	Mean 6884	Mean 1664	Mean 2671	-	-
		(min 0–max 12480)	(min 312–max 12480)	(min 0–max 3120)	(min 1872–max 3120)	-	-
		SD 4979.337	SD 4262.286	SD 977.682	SD 355.319	-	-

presents the metalinguistic data that have been collected in the form of a questionnaire at the end of the elicitation task. Gender and mean chronological age at the time of testing have been reported for every participant. With the term literacy practices, we refer to the input they receive from media (TV, radio, and blogs in Greek), ranking in a three-point scale (0 to 2), measuring the frequency from never to often. The composite variable named current use is the outcome of the variables regarding the language input and language output. Language output measured how often participants speak Greek to different members of their family and friends, whereas language input measured how often family members and friends speak Greek to the participants. This was measured in two three-point scales from rarely to daily (1 to 3), computing one score per participant. The self-ratings have been calculated out of four questions on oral comprehension, reading, writing, and speaking the heritage language on a scale from very easy to very difficult (1 to 5) for each question. Some of the questions, e.g., the age of onset to bilingualism, have been completed only by HSs as they do not apply to monolingual-raised participants. The mean years of formal bilingual education were measured only for the two groups of HSs as monolinguals have attended Greek schools. The differences concerning education measured in both hours and years can be explained by the different curricula followed in American and European bilingual schools. Furthermore, the asymmetry observed between years and hours of formal education is due to the different educational setting in which every participant took part; an example being the bilingual school where students are taught Greek for more hours compared to the afternoon or weekend schools. In addition, we collected data on the generation of HSs’ parents. An average score based on the frequency of visits to the country of heritage was also calculated for both HSs groups, with the scale ranging from none to several visits per month (0–4). Finally, the elaborated version of The Big Five Personality Test, the Ten Item Personality Inventory, was performed in order to measure participants’ personality dimensions as shown in

Table 2. Ten Item Personality Inventory (TIPI) scores across groups and age groups.

		HSs in Germany		HSs in the US		Monolinguals
		Adults	Adolescents	Adults	Adolescents	Adults	Adolescents
Personality test	Critical	Mean 3.44	Mean 3.11	Mean 3.82	Mean 4.06	Mean 3.65	Mean 3.74
		(min 1–max 7)	(min 1–max 6)	(min 1–max 7)	(min 1–max 7)	(min 1–max 7)	(min 1–max 7)
		SD 1.502	SD 1.568	SD 1.806	SD 1.435	SD 1.704	SD 1.653
	Self-disciplined	Mean 5.64	Mean 5.39	Mean 6.29	Mean 5.88	Mean 6.16	Mean 5.94
		(min 2–max 7)	(min 2–max 7)	(min 4–max 7)	(min 3–max 7)	(min 5–max 7)	(min 3–max 7)
		SD 1.377	SD 1.539	SD 0.897	SD 1.100	SD 0.779	SD 0.929
	Extraverted	Mean 5.15	Mean 5.35	Mean 5.93	Mean 5.44	Mean 5.56	Mean 5.27
		(min 1–max 7)	(min 3–max 7)	(min 2–max 7)	(min 1–max 7)	(min 3–max 7)	(min 2–max 7)
		SD 1.377	SD 1.115	SD 1.215	SD 1.501	SD 1.343	SD 1.311
	Anxious	Mean 3.24	Mean 3.39	Mean 3.29	Mean 3.84	Mean 4.55	Mean 4.29
		(min 1–max 6)	(min 1–max 6)	(min 1–max 6)	(min 1–max 7)	(min 2–max 7)	(min 1–max 7)
		SD 1.508	SD 1.335	SD 1.652	SD 1.791	SD 1.546	SD 1.792
	Open to new experiences	Mean 6.28	Mean 6.06	Mean 6.00	Mean 5.69	Mean 5.87	Mean 5.52
		(min 5–max 7)	(min 3–max 7)	(min 3–max 7)	(min 2–max 7)	(min 2–max 7)	(min 2–max 7)
		SD 0.792	SD 1.259	SD 1.122	SD 1.281	SD 1.100	SD 1.313
	Reserved/quiet	Mean 3.68	Mean 3.06	Mean 3.43	Mean 3.72	Mean 4.29	Mean 3.97
		(min 1–max 7)	(min 1–max 7)	(min 1–max 7)	(min 1–max 7)	(min 1–max 7)	(min 1–max 7)
		SD 1.725	SD 1.862	SD 1.709	SD 1.836	SD 1.736	SD 1.602
	Careless	Mean 2.44	Mean 2.78	Mean 2.79	Mean 2.69	Mean 2.97	Mean 3.48
		(min 1–max 5)	(min 1–max 5)	(min 1–max 6)	(min 1–max 5)	(min 1–max 7)	(min 1–max 7)
		SD 1.356	SD 1.353	SD 1.729	SD 1.469	SD 1.816	SD 1.671
	Sympathetic/compassionate	Mean 6.20	Mean 5.72	Mean 5.75	Mean 5.69	Mean 5.74	Mean 4.29
		(min 4–max 7)	(min 3–max 7)	(min 1–max 7)	(min 3–max 7)	(min 1–max 7)	(min 1–max 7)
		SD 0.913	SD 1.320	SD 1.506	SD 1.330	SD 1.460	SD 1.716
	Calm	Mean 4.84	Mean 5.28	Mean 5.71	Mean 5.31	Mean 4.74	Mean 2.39
		(min 1–max 7)	(min 2–max 7)	(min 3–max 7)	(min 3–max 7)	(min 2–max 7)	(min 1–max 6)
		SD 1.748	SD 1.447	SD 1.301	SD 1.469	SD 1.548	SD 1.202
	Conventional	Mean 2.28	Mean 3.33	Mean 2.89	Mean 2.59	Mean 4.84	Mean 2.74
		(min 1–max 6)	(min 1–max 7)	(min 1–max 6)	(min 1–max 5)	(min 2–max 7)	(min 1–max 6)
		SD 1.429	SD 1.879	SD 1.641	SD 1.241	SD 1.485	SD 1.182

(Gosling et al. 2003; adaption in Greek Karfopoulou 2016). These data are illuminating in showing that kati contributes social meaning (Beltrama 2020).

It is important to mention that although the descriptive statistics presented in Table 1 appear to be quite similar across groups and age groups, we find some significant differences between the communities based on several non-parametric Mann–Whitney t-tests, which run pairwise comparisons on the individual samples. Comparing the two adolescent HSs groups, we find that there is a significant difference regarding the self-ratings in the heritage language (U = 133.500, Z = −3.139, p = 0.001), the years of education in the heritage language (U = 171.500, Z = −2.629, p < 0.05), and the literacy practices (U = 148.000, Z = −2.884, p < 0.05). Moreover, significant differences are reported in both age groups concerning the current use (adolescents: (U = 122.500, Z = −3.352, p = 0.001), adults: (U = 280.000, Z = −1.983, p < 0.05)), hours of instruction in the heritage language (adolescents: (U = 128.000, Z = −3.445, p < 0.001), adults: (U = 291.000, Z = −1.996, p < 0.05)), and the visits to the heritage country as well (adolescents: (U = 143.000, Ζ = −3.505, p = 0.001), adults: (U = 193.500, Z = −4.343, p < 0.001)). As we can observe firstly from the descriptive statistics and secondly from the tests, HSs in Germany appear to have higher ratings in all metalinguistic data. Finally, there is also one more significant difference with respect to the chronological age of adult HSs in the US and adult monolingual controls at the time of testing (U = 294.000, Z = −2.795, p < 0.05).

4.3. Results of the Production Study

This section presents the results of the distribution of kati in the production task across groups, age groups, communication settings, and modalities. The distribution of the indefinite determiner kati preceding a nominal head per setting and modality across groups can be seen in

Table 3. Distribution of kati and number of tokens per setting and modality across groups.

Setting	Modality	HSs in Germany		HSs in the US		Monolinguals
		Adults	Adolescents	Adults	Adolescents	Adults	Adolescents
Formal	Spoken	2/3413	1/1772	2/2821	0/2965	0/3471	0/3462
Formal	Written	1/2785	0/1350	0/2147	0/2362	0/3156	0/3426
Informal	Spoken	1/2745	1/1612	4/2243	1/2729	3/3202	4/3209
Informal	Written	2/2014	0/1065	0/1517	1/1772	1/2371	2/2315

.6 The instances found in our data, as the one in example (12), are few, because the production task did not explicitly target the elicitation of indefinite NPs involving kati.

(12)	Sti	sakula	ihe	kati	mila	Adult HSs in Germany
	In-the	bag	had.3SG	some	apples.PL
	In the bag there were some apples

Results show that, although all groups align in the sense that kati is preferred in the oral modality, monolinguals use it only in the informal context, while HSs, especially the German group, tend to overgeneralize across settings. Despite the small number of instances found, this pattern signals an asymmetry regarding register levelling in HSs who use an informal element such as kati across different settings, as was found in previous studies as well (see Rothman 2007).

In order to further explore our findings, we performed nonparametric Kendall’s correlations per group, as we had only a few quantitative findings concerning the kati constructions in SPSS v.25. In all three groups, there was a strong 2-tailed Kendall’s tau-b correlation coefficient with register (HSs in Germany: τ_b = 0.996, p < 0.0001, HSs in the US: τ_b = 0.993, p < 0.0001, and control group: τ_b = 0.992, p < 0.0001) and modality (HSs in Germany: τ_b = 1.000, HSs in the US: τ_b = 0.983, p < 0.0001, and control group: τ_b = 0.988, p < 0.0001). Furthermore, a strong 2-tailed Kendall’s tau-b correlation with some variables of participants’ personality was observed in the group of HSs in the US (careless: τ_b = 0.335, p < 0.005), while two weak 2-tailed Kendall’s tau-b correlations, one being positive and the other negative, were found in the control group (extraverted: τ_b = 0.245, p < 0.05 and quiet: τ_b = −0.278, p < 0.05). As mentioned in Section 2, kati is an indicator of informal settings and more precisely a marker of VL, which is a characteristic of colloquial speech, and this is confirmed from the correlations in our first study. Furthermore, the positive correlations of the score of two characteristics from the personality test in two out of the three groups, namely careless and extraverted, correlated significantly, although the second weakly, with the kati instances that are found. These two characteristics belong in the conscientiousness and the surgency/extraversion continuum, respectively, and both are on the same side of the axes (Goldberg 1992; Gosling et al. 2003). Being careless and extraverted simulates the informal register and the colloquial speech while the feature quiet or introverted simulates the formal one. This suggests that the use of kati can be viewed as a marker of social meaning (Beltrama 2020).

The instances of kati found in the narration task are produced by four adult and two adolescent HSs in the US, two adult and two adolescent HSs in Germany, and finally four adult and five adolescent monolingual speakers, as some participants used the indefinite determiner in more than one setting and/or modality. Looking further into the metalinguistic data of HSs who used this indefinite determiner, we could not find any significant correlation with the metalinguistic data (years of bilingual education received (τ_b = −0.3462, p = 0.380), generation of their parents (τ_b = −0.118, p = 0.343), age of onset to bilingualism (τ_b = −0.298, p = 0.488), current use (τ_b = −0.239, p = 0.546), and literacy practices in Greek (τ_b = −0.248, p = 0.546)).

Comparing the two HSs groups, we observe that instances of kati appear in both modalities (oral and written) and formality contexts in the German heritage group. Specifically, while US HSs do not use kati in the formal written mode, there is one instance of kati in the formal written mode produced by the German group, pointing to an overgeneralization of its use. We also checked their productions of the German ein paar ‘a few’, which seems the closest counterpart of kati. According to Sauerland (2022), this is more frequent in informal spoken registers. However, we could not identify a similar pattern of overgeneralization. Thus, language interference from German can be excluded. We interpret this as suggesting that German HSs speakers fully align with our predictions, showing a pattern of overgeneralization that is well established in heritage linguistic research. However, because this is a minimal difference between the two groups, we cannot construct a more theoretical explanation of this pattern. It could simply be related to the fact that the US corpus is relatively smaller (18,556 tokens produced by 63 speakers) than the German corpus (16,756 tokens produced by 48 speakers).

4.4. Interim Summary—Production Study

In conclusion, monolingual speakers of Greek and different groups of HSs took part in the production task, in which several constructions with the indefinite determiner kati are found. As kati is a marker of VL and is preferred in informal settings by monolingual speakers of Greek, we argue that HSs show a register levelling by overgeneralizing kati in the different communication situations. Regarding the modality, although the numerical difference is very small, we can observe a preference of kati in the oral modality over the written both in monolinguals and in HSs.

5. Kati in Greek Corpora

To further substantiate our hypothesis, and as we had few occurrences of kati in our corpus, we performed a search of the Corpus of Spoken Greek compiled in Aristotle University of Thessaloniki and the Greek Heritage Language Corpus (G.H.L.C). The former provides oral data from monolingual Greek speakers in different communication situations; the latter provides oral data from three groups of HSs (with one group having American English as its majority language and the other two groups Russian) who took part in (1) a production task and (2) an informal conversation with an elicitor.

The Corpus of Spoken Greek7 contains data from both formal communication settings (e.g., interviews with politicians broadcasted in television) and informal settings (e.g., everyday conversations among friends and relatives). The number of tokens available online is 671.543 and the instances of kati followed by a nominal head are 52, thus the frequency of them is 0.07 per thousand tokens.

Table 4. Distribution of kati tokens per communication situation in the online available material of the Corpus of Spoken Greek.

Communication Situation	Subcorpus/Occasion	Instances of kati Tokens
Formal	Interviews in television	3
Informal	Telephone calls among friends and relatives	14
	Everyday conversations among friends and relatives	35

shows the distribution of kati tokens per subcorpus. Despite the small number of kati instances found, we notice that Greek monolingual speakers tend to produce these constructions mostly in informal communication settings.

Furthermore, we explored the Greek Heritage Language Corpus (G.H.L.C), which comprised 130.000 tokens in total. Having performed a thorough search of kati proceeding a plural noun (13), we found 10 instances in the different groups of HSs, as shown in

Table 5. Distribution of kati tokens per HSs subcorpus in raw instances and frequency.

G.H.L.C Subcorpora	Number of Κati Tokens	Total Number of Tokens	Frequency per a Thousand Tokens
HSs in Chicago	2	85.000	0.02
HSs in Moscow	5	20.000	0.25
HSs in Saint Petersburg	3	25.000	0.12

:

(13)	Oti ine san	kati paramithia	HS in Chicago
	That are like	some fairy tales.PL
	That are like some fairy tales.

The findings from the Corpus of Spoken Greek support our claim that kati is primarily used in informal communication situations in Greek monolinguals’ repertoire. However, the few occurrences in the G.H.L.C. cannot help us make any conclusions with respect to the use of kati by these speakers.

6. Judgment Study in Monolingual Speakers of Greek

In order to further test our hypothesis, we conducted a judgment study with the aim to examine to what extent different communication settings trigger a preference for kati in monolingual speakers of Greek.

6.1. Design and Materials of the Judgment Study

We designed a judgment task in the sense of sentence completion with multiple answers to be scored. In the upper part of the screen appeared a context in blue, in which the sentence that proceeded below is uttered. This context informed participants that an oral utterance follows from different people. Moving on to the experimental items, there was a blank in every sentence and participants had to evaluate all five different possible options based on their preference regarding the context. Participants had to evaluate each option on a Likert scale from 1 (I don’t like it at all) to 7 (I like it very much), in order to increase variability (Rasinger 2013) (see Figure 1).

We created 2 lists of experimental items (see Appendix A.2), each one containing 32 items (total 64 items). Half of them (16 sentences) describe a formal communication situation, while the other half describe an informal communication situation; all participants were exposed to items of all different context conditions. In both settings, we also balanced the distribution of nouns concerning the gender they bear. Only common nouns were included in the items (proper, abstract, and mass nouns were excluded).

In order to adapt the sentences so that they reflect the formal communication situation, we manipulated the wording by creating two versions expressing the same meaning, one with less (formal 1) and one with more frequent words (formal 2). The less frequent lexical items are built from Ancient Greek (henceforth AG) roots (14a), (15a), while those that are more frequent belong to the demotic register (14b), (15b); for each we calculated the relevant frequency rate in the Hellenic National Corpus of Greek Language—HNC—which consists of written texts (total number = 97,000,000 tokens), see examples (13a,b).

(14a)	ipodimata	Frequency in HNC 0.0026 per thousand tokens
	shoes—AG root
(14b)	papoutsia	Frequency in HNC 0.0169 per thousand tokens
	shoes—Turkish root

(15a)	Taraksies	Frequency in HNC 0.0018 per thousand tokens
	Agitator—AG root
(15b)	Mpahalakides	No results
	Agitator—Modern Greek nominalization

A comparable example of the two formal settings that includes the two lexical items from example 14 can be seen in example 16, in which a. represents the item with less frequent words or words that have an Ancient Greek root and b. is the item with more frequent words. Even the two verbs anasiro “drag up” and entopizo “spot/find” are of different frequencies; the first having 0.0044‰ while the second 0.0416‰, respectively. The items have been carefully created not only in terms of frequency but also in terms of word order, detailed formulation, and passivization of the sentence.

(16)	Περίσταση επικοινωνίας: Τηλεοπτική αναγγελία σεισμολόγου Communication situation: Seismologist during the news on TV
	Formal 1	Formal 2
a. Σύμφωνα με τις τελευταίες πληροφορίες της Πυροσβεστικής έχουν ανασυρθεί μεταξύ των ερειπίων, ___ κατεστραμμένα υποδήματα.		b. Oι πυροσβέστες έχουν εντοπίσει στα ερείπια του κτιρίου ___ κατεστραμμένα παπούτσια.
a. According to the latest information from the Fire Department ___ damaged shoes have been dragged up.NACT among the rubble.		b. Firefighters have found.ACT ___ damaged shoes in the ruins of the building.

In the informal communication setting, the items were also manipulated. Of the items, 16 exhibited a nominal head (informal 1), in which participants had to indicate their preference regarding the determiners, and another 16 items were created with the same noun denoting a diminutive (informal 2). In our study, different types of diminutives are included, such as diminutives ending in -aki and in -itsa, -ula, (Alexiadou 2020). As Alexiadou (2020) mentions, the former type is the most frequent one and applies to all nouns triggering a gender shift—it always yields the neuter nouns—and a declension class (DC) shift. By contrast, the nouns combining with the latter type preserve their gender and their DC. As mentioned in Section 2, kati is a marker of VL and diminutives are indicators of informal setting as well (Makri-Tsilipakou 2003; Καβαρνού and Μπέλλα 2012), or have ironic/sarcastic connotations when used in literature (Katsaros 2018). Appendix A.3 exhibits the distribution of the nouns in both formal and informal settings.

The five possible responses (forms) which participants had to rate were three determiners, namely the definite, indefinites from the ka- series and the indefinite kati, a numeral, and the option of a determiner omission. All options were grammatical building agreement patterns between the nominal head and its complements, and all items have been tested in a pilot study prior to the main one.

In every list, 4 pilot items and 28 filler sentences of the same format/design were included, an example of which can be seen in 17. We used the Gorilla Experiment Builder (www.gorilla.sc, accessed on 10 August 2021) to create and host our experiment (Anwyl-Irvine et al. 2020), as this particular design fulfilled the presentation of the context, the sentence, and the five conditions that participants had to rate. Trials, fillers, and the respective five possible responses in every item were automatically pseudorandomized.

(17)	Context: Policeman to a citizen
Παρακαλώ μετακινήστε το όχημά σας ___ διότι εδώ απαγορεύεται η στάθμευση.
Please move your car ____ because parking is prohibited here.
αμέσως	γρήγορα	άμεσα	το συντομότερο δυνατό	επιτόπου
immediately	quickly	promptly	as soon as possible	right away

6.2. Participants in the Judgment Study

Monolingual speakers were asked to confirm some eligibility criteria, e.g.,: 1. if they had grown up monolingual with monolingual education, 2. if they currently live in Greece, 3. if they had not spent more than a year in a row abroad, and 4. if they are not diagnosed with any speech disorders; they were recruited via mailing lists and social media platforms (mailing lists/Facebook) all over Greece. This sample of participants is different from the one that participated in the production study. The randomizer function allocated 25 participants to fill in List A and 25 to List B. In

Table 6. Participants’ profile in judgment study.

	Monolingual Greek Speakers
N	50
	(33 Females)
Chronological Age	Mean 28; 9
	(min 19–max 35)
	SD 4.1837
Highest Education Completed	High school	7 prt
	Secondary technical school	1 prt
	Technical education	3 prt
	Bachelor’s degree	19 prt
	Master’s degree	20 prt
Personality Test	Critical	Mean 4.70
		(min 2–max 7)
		SD 1.542
	Self-disciplined	Mean 1.98
		(min 1–max 7)
		SD 1.059
	Extroverted	Mean 2.94
		(min 1–max 6)
		SD 1.331
	Anxious	Mean 3.38
		(min 1–max 7)
		SD 1.783
	Open to new experiences	Mean 2.76
		(min 1–max 5)
		SD 1.188
	Reserved/quiet	Mean 3.04
		(min 1–max 7)
		SD 1.355
	Careless	Mean 5.16
		(min 1–max 7)
		SD 1.376
	Sympathetic/compassionate	Mean 2.06
		(min 1–max 4)
		SD 0.843
	Calm	Mean 3.12
		(min 1–max 7)
		SD 1.438
	Conventional	Mean 5.30
		(min 1–max 7)
		SD 1.313

, the profile of our participants is presented.

6.3. Results of the Judgment Study

Table 7. Ratings of the responses per context.

Context	Response	Mean	SE	Lower Conf. L.	Upper Conf. L.
Formal 1	definite	3.69	0.167	3.36	4.02
	indefinite	3.73	0.167	3.40	4.06
	kati	3.40	0.167	3.07	3.72
	numeral	4.39	0.167	4.06	4.72
	omission	4.15	0.167	3.82	4.48
Formal 2	definite	3.64	0.167	3.31	3.97
	indefinite	3.56	0.167	3.23	3.89
	kati	3.25	0.167	2.92	3.58
	numeral	4.82	0.167	4.49	5.15
	omission	3.93	0.167	3.60	4.26
Informal 1	definite	4.81	0.167	4.48	5.14
	indefinite	4.36	0.167	4.03	4.68
	kati	5.05	0.167	4.72	5.38
	numeral	4.88	0.167	4.55	5.21
	omission	4.21	0.167	3.88	4.54
Informal 2	definite	4.48	0.167	4.15	4.81
	indefinite	4.41	0.167	4.08	4.74
	kati	4.67	0.167	4.34	5.00
	numeral	4.45	0.167	4.12	4.78
	omission	4.92	0.167	4.59	5.25

reports the output of the ratings of the responses per context while Figure 2 presents the means of each response form per context condition.

For the statistical analysis, we fit a linear mixed effects model using the lme4 package in RStudio (RStudio Team 2020). Form (factor with 5 levels: definite, indefinite, kati, numeral, and omission), Context (factor with 4 levels: formal 1, formal 2, informal 1, and informal 2), and their interaction were used as fixed effects, while subjects and items were used as random effects. First, a model comparison was conducted to define the best model in terms of random effects structure. Following Barr et al. (2013), the random effects structure was gradually simplified until convergence was reached, resulting in the simplest model, i.e., with varying intercepts by subject and by item. Then, the fixed effects structure of this model was checked, with an interaction version being compared to an additive version via a likelihood ratio test. The former was proven to provide a better fit to the data (χ²(12) = 63.4, p < 0.0001). The syntax of the best model used in the analysis is shown in (18):

(18)	Rating ~ Form * Context + (1 | subject) + (1 | item)

First, an analysis of deviance (Anova-Type II Wald chi-square test) was performed on the best model to detect the main effects. The results revealed a significant main effect of Form (p < 0.0001) and Context (p < 0.0001), as well as a significant interaction between Form and Context (p < 0.0001). Then, post hoc pairwise comparisons were run using the RStudio package emmeans, with adjusted alpha levels (Tukey’s method to control for multiple comparisons) in order to detect (1) significant differences between the different scores given for kati across contexts and (2) significant differences between kati and the other forms within each context.

Overall, results show that kati obtained significantly lower scores in the two formal contexts (formal 1 and formal 2) than in the two informal ones (informal 1 and informal 2) (β = −1.5, SE = 0.3, z = −4.5, p < 0.0001). Within the formal 1 context, kati yielded the lowest mean rating overall (3.4), but only its difference to numeral (β = −0.9, SE = 0.2, z = −6.3, p < 0.0001) and to omission (β = −0.8, SE = 0.2, z = −4.8, p < 0.0001) reached significance in our data. Similar is the picture in the formal 2 context, where kati indicated the lowest mean score again (3.2), differing significantly from both numeral (β = −1.6, SE = 0.2, z = −9.9, p < 0.0001) and omission (β = −0.7, SE = 0.2, z = −4.3, p = 0.0002). However, these patterns change in the two informal contexts. For informal 1, kati indicated the highest mean rating (5.05) and showed significant differences from the indefinite (β = 0.7, SE = 0.2, z = 4.4, p = 0.0001) and omission (β = 0.9, SE = 0.2, z = 5.3, p < 0.0001). For informal 2, kati was rated high (4.7), but was still numerically lower than omission (4.9); however, no significant differences were detected here.

Focusing exclusively on kati, some additional analyses were run to define the role of word frequency (formal 1 vs. formal 2), as well as that of the diminutive presence (informal 1 vs. informal 2). Regarding the former, results revealed that high word frequency does not appear to trigger better scores for kati than low word frequency in formality (β = −0.2, SE = 0.2, z = −0.9, p = 0.4). Upon the presence of a diminutive ending in an informal context, on the other hand, kati was rated significantly lower compared to those cases where no diminutive ending is present (β = −0.4, SE = 0.1, z = −2.8, p = 0.01).

6.4. Discussion of Judgment Task

Our prediction about kati is partially confirmed in formal and informal contexts. In both formal contexts (formal 1 and formal 2), kati receives the lowest score, differing significantly from numerals and omission. However, comparing the scores of kati in formal 1 and formal 2 contexts, i.e., low vs. high frequency, we observe that our hypothesis regarding kati receiving a higher score in the more frequent items (formal 2) is not confirmed. This suggests that frequency does not appear to be a trigger. Moving on to the informal contexts, we observe that in informal 1 context, kati does indeed receive the highest score, which differs from the score of other responses (indefinite and omission). However, in the informal 2 context, kati receives the second highest score numerically, but does not differ significantly from the other forms. Thus, our hypothesis that kati receives a higher rating when it is followed by another marker of informality—in our case diminutive—is not confirmed here. This could be attributed to the fact that speakers judge the presence of the diminutive as already sufficient to signal informality, and thus a further marker does not seem necessary.

7. Conclusions

In this paper, we first discussed the semantic contribution of kati and claimed that it is a marker of VL. To substantiate this, we investigated its distribution in monolingual and heritage varieties of Greek in production and judgment tasks. In the former tasks, Greek HSs, both in the US and in Germany, overgeneralize the informal marker in both registers, while monolingual speakers pattern differently by using it solely in the informal register. Thus, HSs exhibit register levelling, and they are unfamiliar with register differentiations, which can be transmitted via formal education, as stated in previous studies as well (Rothman 2007). We were also able to show the use of kati as a marker of social meaning. In the latter task, monolingual Greeks indicated their preference for kati in the informal contexts rather than in the formal ones. Although our hypotheses for formal (1 and 2) and informal (1 and 2) are partially confirmed, the results still point to a difference between the formal and the informal communicative situations.