1. Introduction
Most research conducted so far on linguistic variation in Albanian falls within the Albanology tradition and belongs to the broader field of dialectology. One of the goals of this line of research is to uncover the geography of dialects by defining areas where bundles of linguistic features are found, with particular (if not exclusive) attention paid to features produced by nonmobile, older, rural, male (NORM) speakers (Chambers & Trudgill, Reference Chambers and Trudgill1998; Wieling & Nerbonne, Reference Wieling and Nerbonne2015). This predominance of the Albanology tradition entails that some aspects of linguistic variation in Albania are not well-understood. For example, it is unclear how dialects are realized nowadays in urban areas, knowing that Albania experienced mass population movements, rapid urbanization, and rural exodus in its recent history, or how Albanians conceive of linguistic variation across their country. In this article, we present a perceptual dialectology study investigating lay listeners’ perceived correctness and pleasantness of Albanian spoken in the country’s 12 counties, as well as the connection between these perceptions and the counties’ majority dialects and relative urbanization.
1.1. The Albanian context
Albanian is an isolate language of the Indo-European family, spoken by nearly 7 million people around the world (Rusakov, Reference Rusakov and Kapović2017). It has traditionally been described as having two major dialects in Albania: Gheg, spoken in the central and northern parts of the country, including in the capital city Tirana, and Tosk, spoken in the south, with the Shkumbin river as approximate dialect boundary (Elsie & Gross, Reference Elsie and Gross1988; Gjinari, Reference Gjinari1988; Gjinari et al., Reference Gjinari, Beci, Shkurtaj, Gosturani and Dodi2007; Coretta et al., Reference Coretta, Riverin-Coutlée, Kapia and Nichols2023; see Figure 1). Albanian also has a standard variety based on Tosk, officially declared in 1972 during a Congress of Orthography orchestrated by the then-communist state (Byron, Reference Byron1976; Kostallari, Reference Kostallari1984; Moosmüller & Granser, Reference Moosmüller and Granser2006). Variation in the language has predominantly been described in terms of traditional dialectology, i.e. based on geographical areas and associated features, by focusing on documenting the most conservative forms, unaffected by ongoing change, and by collecting data primarily from older male informants living in rural areas (NORMs). The prime example of this methodology is found in Atlasi dialektologjik i gjuhës shqipe ‘The dialectological atlas of Albanian language’ (Gjinari et al., Reference Gjinari, Beci, Shkurtaj, Gosturani and Dodi2007), considered to this day to be the most fundamental reference on Albanian dialects (Çeliku, Reference Çeliku2010; Lamaj, Reference Lamaj2011). Table 1 shows relevant sociodemographic characteristics of the 148 informants for the atlas’s geographical points within Albania.Footnote 1
Figure 1. Map of Albania showing the Shkumbin river as division line between Gheg-speaking areas (north of Shkumbin) and Tosk-speaking areas (south of it). Labeled municipalities are those discussed in Section 1.1. Coordinates were retrieved from OpenStreetMap (Padgham et al., Reference Padgham, Lovelace, Salmon and Rudis2017).
Table 1. Characteristics of Albanian subject pool in Atlasi dialektologjik i gjuhës shqipe (Gjinari et al., Reference Gjinari, Beci, Shkurtaj, Gosturani and Dodi2007).
However, due to this focus on geographical variation among NORM speakers, there has been little exploration of other factors which may structure variation in Albanian. One factor that, in other languages, has been shown to be relevant from the point of view of both linguists and lay speakers is the divide between urban and rural speech (Britain, Reference Britain, Busse and Warnke2022). A small number of studies of Albanian indeed point in this direction. Urban speech of Lezha (Shkurtaj, Reference Shkurtaj1984) and Gjirokastër (Totoni, Reference Totoni1966), as well as rural speech of Koplik (Shkurtaj, Reference Shkurtaj1969), were found to be different from surrounding varieties, either due to contact with other varieties (cases of Gjirokastër and Lezha) or lack thereof (case of Koplik) (see Figure 1). While these studies captured variation at a time when Albanians were deprived of their freedom of movement due to national policies of the communist state (Misja & Vejsiu, Reference Misja and Vejsiu1990; Gedeshi & Jorgoni, Reference Gedeshi and Jorgoni2012), little work so far has investigated variation and the urban/rural divide following the fall of communism in the early 1990s. At that time, citizens regained their freedom of movement and flocked into urban centers in search of better economic opportunities in what has been described as rural exodus (Carletto et al., Reference Carletto, Davis, Stampini and Zezza2006; Lerch, Reference Lerch2016), resulting in rapid and drastic sociopolitical changes which reshaped the social structure across Albania. Addressing this contemporary issue, Riverin-Coutlée et al. (Reference Riverin-Coutlée, Kapia, Cunha and Harrington2022) recently observed that speech features found in the Gheg dialect spoken in Tirana differed from those of a rural Gheg-speaking community located just 15 kilometers away from Tirana. They attributed these differences to high contact among different varieties in urban Tirana, which tripled in size over the course of a few decades, but low contact in the rural area.
Turning to sociolinguistic attitudes expressed by Albanians, recent studies showed that the Tosk dialect, or what are perceived to be Tosk speech features, were attributed a higher status than the Gheg ones (Dickerson, Reference Dickerson2021; Riverin-Coutlée, Kapia & Gubian, Reference Riverin-Coutlée, Kapia and Gubian2024). For example, Gheg features or dialects were rated as less educated, less proper, or more rural by participants with both Gheg and Tosk backgrounds. Morgan (Reference Morgan2015) conducted sociolinguistic interviews with 19 Tosk-speaking participants who had moved to Gheg-speaking Tirana and reported that, consistent with the above, Gheg was described by the informants with terms such as isolated, rural, undeveloped, backwards, harsh, and uncultured/thick, while Tosk was qualified as standard, more developed, calm, and soft. Interestingly however, several of her informants distinguished the center of Albania, comprising the large cities of Tirana, Durrës, and Elbasan, from the Gheg-speaking north and the Tosk-speaking south (Morgan, Reference Morgan2015, Reference Morgan2022). One informant, who mentioned hearing Gheg speech features in Tirana, even hypothesized that these had been borrowed from more remote Gheg areas to give more strength (forcë) to urban speech, otherwise considered standard (Morgan, Reference Morgan2015:49). Given that Tirana, Durrës, and Elbasan are located in the Gheg-speaking area (see Figure 1), this is inaccurate, and the Gheg features that Morgan’s (Reference Morgan2015) participant mentioned are not recent borrowings from remote areas. The discrepancy between the participants’ low esteem of Gheg and their positive perception of urban speech seems to be resolved here through the borrowing rationale. Morgan (Reference Morgan2015) also observed among other informants a tendency to partition both Tosk-speaking and Gheg-speaking areas along the urban/rural divide. For instance, Shkodra, located in (low-status) Gheg-speaking territory (see Figure 1), was described as a developed city (qytet i zhvilluar; Morgan, Reference Morgan2015:42–43); whereas inhabitants of Tepelena, a small town in a (high-status) Tosk-speaking area (see Figure 1), were labeled as malokë, which the author translated as ‘hillbillies’ (Morgan, Reference Morgan2015:46–47).
In sum, in addition to the widely accepted Gheg/Tosk division, a few studies have suggested that an urban/rural divide could also be relevant within the Albanian linguistic landscape. Of particular interest to the current study is the still-little-explored, complex, and sometimes conflicting way in which the classic Gheg/Tosk division and the urban/rural one are brought together in Albanian speakers’ sociolinguistic representations, especially given the changing face of these different speech communities in the last 30 years or so.
1.2. Perceptual dialectology
Differently from traditional dialectology, which preoccupied itself with the notion of place as a geographical area with specific phonological and lexical features (Cramer & Montgomery, Reference Cramer and Montgomery2016), perceptual dialectology focuses on understanding how non-linguists think about and perceive language variation (Cramer, Reference Cramer2016). Within this area of research, lay speakers’ perceptions are considered important to understanding variation, where it comes from, how it unfolds in space, and how these differences may reflect and lead to language and social choices (Preston, Reference Preston1989, Reference Preston1999, Reference Preston, Chambers and Schilling2013; Britain, Reference Britain, Al-Wer and de Jong2009, Cramer, Reference Cramer2016).
The conduct of this research has been realized using different methods, such as map drawing, ranking of degrees of differences, ranking in terms of pleasantness and correctness, qualitative analyses, etc. We focus here on studies that asked speakers to rate varieties, including their own, in terms of how correct or pleasant they perceive them to be, as these are the methods we also used. This type of work has generally suggested that correctness is a measure for which participants of different backgrounds agree with each other (Cramer & Montgomery, Reference Cramer and Montgomery2016). For instance, Americans, regardless of whether they are from the North or the South, typically agree on labelling the Northern American English varieties as more correct and the Southern varieties as less correct (e.g. Niedzielski & Preston, Reference Niedzielski and Preston2000; Niedzielski, Reference Niedzielski, Long and Preston2002); similarly, Germans from both (former) East and West Germany label the Western German varieties to be more correct than the Eastern ones (Dailey-O’Cain, Reference Dailey-O’Cain and Preston1999). Ratings on the pleasantness scale, however, are less consistent and predictable. Several studies have found that listeners who rate their own variety as correct also rate it as pleasant, for example the West Germans mentioned above (Dailey-O’Cain, Reference Dailey-O’Cain and Preston1999; see also e.g. Michiganders in Preston, Reference Preston, Chambers and Schilling2013, and Parisians in Kuiper, Reference Kuiper and Preston1999). On the other hand, listeners who rate their own variety as less correct have been found to rate their own variety as either less pleasant (e.g. East Germans in Dailey-O’Cain, Reference Dailey-O’Cain and Preston1999), more pleasant (e.g. Alabamians in Preston, Reference Preston, Chambers and Schilling2013), or equally pleasant (e.g. Memphians in Fridland & Bartlett, Reference Fridland and Bartlett2006). Thus, speakers of different varieties may or may not agree with each other on which variety is more pleasant.
Furthermore, research in perceptual dialectology has reported that urban speech is generally perceived more positively than rural speech. Examining perceptions in the Korean context, for example, Jeon & Cukor-Avila (Reference Jeon, Cukor-Avila, Cramer and Montgomery2016) show that participants mostly associate the urban Seoul variety with the “standard” and “pleasant” labels, while rural areas’ speech is labeled as less pleasant and less standard. Similarly, Cramer (Reference Cramer2016) illustrates that Louisvillians, living in the sole urban area within the largely rural state of Kentucky, label Kentuckian urban speech as northern, and Kentuckian rural speech as southern. Along similar lines, Evans (Reference Evans, Cramer and Montgomery2016) demonstrates that the urban/rural dichotomy is also salient for Washington state informants, with the speech of the rural communities of Washington associated with perceptions of “country,” “southern,” and “gangsta,” but the urban speech with high status.
The results from these perceptual dialectology studies suggesting that the urban/rural divide is relevant to many participants in various linguistic settings prompt us to address the question in the Albanian context. It appears all the more relevant given the recent drastic changes in the urban/rural landscape of the country as well as some of Morgan’s Reference Morgan(2015) observations (see Section 1.1). However, while the consistent and clear-cut opinions that people seem to hold about urban/rural speech suggest a more general categorization mechanism between urban/rural that is both productive and tangible, framing the urban/rural divide with objective terms and measures is notoriously difficult (see e.g. comments from geographers collated by Britain, Reference Britain, Busse and Warnke2022:67–69). In this study, we resort to population density as a proxy for relative urbanization. Using population density has many advantages: it can be computed from easy-to-access data; it is continuously distributed, avoiding the need for discrete categorization; and it has been argued elsewhere to be a reasonable proxy for urbanization of a given community (e.g. Hooghe & Botterman, Reference Hooghe and Botterman2012; Hans & Koster, Reference Hans and Koster2018; Kuussaari et al., Reference Kuussaari, Toivonen, Heliölä, Pöyry, Mellado, Ekroos, Hyyryläinen, Vähä-Piikkiö and Tiainen2021). The small size and comparable area of Albania’s 12 counties (see Section 2.3) also entail that a large city within a given county noticeably affects that county’s density, and therefore that higher density means more urban. While some caveats of this metric will be discussed in Section 4.3, we use it to explore the connection between the urban/rural divide and ratings provided by Albanian participants on correctness and pleasantness scales.
1.3. Research questions and hypotheses
Our broad motivation for this study is to deepen our understanding of variation in Albanian beyond the parameters of traditional dialectology. More specifically, we want to know how the classic Gheg/Tosk division and the urban/rural divide interact in terms of how correct and pleasant Albanians perceive the language spoken across the country to be. We also investigate whether there are differences in these ratings depending on whether the participants originate from Gheg-speaking or Tosk-speaking areas.
Based on the literature outlined so far, we formulate the three following hypotheses for the correctness scale.
-
H1. Tosk counties will be rated higher than Gheg counties.
-
H2. Within a given dialect area, more urban counties will be rated higher.
-
H3. Participants with Tosk and Gheg backgrounds will produce similar ratings.
The outcome of pleasantness ratings is more difficult to predict based on inferences from prior work. We restrict our hypotheses to the two following, leaving open the question of whether Gheg-speaking participants will agree with Tosk-speaking ones regarding where language is more pleasant.
-
H4. Participants with a Tosk background will rate Tosk counties higher than Gheg counties.
-
H5. Within a given dialect area, more urban counties will be rated higher.
2. Methods
2.1. Procedure
For this pioneering perceptual dialectology study conducted in Albania, we took inspiration from the now well-established and widely used methodology put forth by Preston (Reference Preston1989), which asked participants to rate perceived speech on two scales, correctness and pleasantness, relating to the attitudinal dimensions of status and solidarity respectively (Fiske et al., Reference Fiske, Cuddy, Glick and Xu2002). Participants were sent by email or other messaging applications the map reproduced in Figure 2, where the 12 administrative counties of Albania are clearly delineated and labeled. They were first asked to rate the correctness (translated as gjuhë e saktë in Albanian) of the speech produced in each of the 12 counties on a scale from 1 to 10, where 1 was least correct and 10 was most correct. Second, using the same map, they rated the pleasantness (gjuhë e këndshme) of each county’s speech on a scale from 1 to 10, where 1 was least pleasant and 10 was most pleasant. They sent their ratings to the research team either as plain text, as annotated maps, or verbally over the phone.Footnote 2
Figure 2. Map of Albania with its 12 administrative counties distributed to participants for both the correctness scale and the pleasantness scale. Retrieved from the government website http://sipermarrjaime.gov.al/qarqet/ in 2022.
2.2. Participants
Participants were recruited among the first author’s and her students’ networks, then by snowball sampling, as part of an assignment for a master’s research seminar at the Ludwig-Maximilian University of Munich, Germany, in 2022 (students conducted an assignment based on this data). One hundred and twenty-three (123) participants completed the task. When sending their ratings to the researchers, they were also asked to provide some basic sociodemographic information. All participants declared that they spoke Albanian natively and lived in Albania at the time of the experiment; 75 were women, 48 were men; and they were aged 18 to 78 years old (mean 31, median 24). Their dialect background was determined on the spot as either Gheg or Tosk, which resulted in 77 participants categorized as Gheg and 46 as Tosk. Dialect background is the only sociodemographic information considered for further analysis.
2.3. Geographical considerations
Albania comprises 12 administrative counties (qarqe), and each of them can be categorized as comprising a majority of either Tosk or Gheg speakers, as shown in Table 2 and Figure 3. Population density for each county was calculated by dividing the number of inhabitants in January 2022 (INSTAT, 2023) by the county’s area. Note that relative to many other places around the world, Albanian counties’ areas are both small and comparable with each other, which means that any large urban center noticeably increases a county’s population density (see Section 1.2 for advantages of this proxy).
Table 2. Majority dialect, population size, area and density of the 12 administrative counties of Albania.
Figure 3. Map of the 12 administrative counties of Albania, density-colored. The majority dialect spoken in each county is indicated by G for Gheg and T for Tosk.
2.4. Statistical analyses
Ratings on the correctness and pleasantness scales were analyzed separately with linear mixed-effect regressionsFootnote 3 using the lmerTest package (Kuznetsova, Brockhoff & Christensen, Reference Kuznetsova, Brockhoff and Christensen2017) in R (R Core Team, 2024). In each model, the response variable was ratings from 1 to 10 on the relevant scale, treated as continuous. Three independent variables were considered: participants’ dialect (2 levels: Gheg, Tosk), county dialect (2 levels: Gheg, Tosk) and county density (continuous, log-transformed). Random intercepts per participant were also included, as well as random slopes for county dialect. Random slopes for county density were removed from the models due to persistent failure to converge. For each scale, the model of best fit was chosen by constructing models with increasing complexity and comparing their Aikake and Bayesian information criteria (AIC and BIC) as provided by the anova() function. For both scales, the best fitting model is that shown in (1), as expressed in R syntax. Post hoc comparisons were computed with emmeans (Lenth, Reference Lenth2023).
3. Results
The results are presented for the correctness scale first (Section 3.1) and the pleasantness scale second (Section 3.2). In each section, our analyses are guided by the outcome of the model selection procedure described in Section 2.4, which resulted in selecting models integrating an interaction between county dialect, county density, and participants’ dialect.
3.1. Correctness
Figure 4 illustrates mean ratings from Gheg and Tosk participants on the correctness scale for the 12 Albanian counties. In this figure (as well as Figure 6), the white color marks the middle point of the scale; blue indicates ratings above the middle point, towards the most correct label; red indicates ratings below the middle point, towards the least correct label. The darker the color, the closer to the label.
Figure 4. Mean ratings on the correctness scale from Gheg (left) and Tosk (right) participants for the 12 counties. The majority dialect spoken in each county is indicated by G for Gheg and T for Tosk.
Figure 5. Regression line extracted from the statistical model applied to ratings on the correctness scale, superimposed on real ratings provided by participants, split by participant dialect and county dialect, with county names indicated above their corresponding data.
Figure 6. Mean ratings on the correctness scale from Gheg (left) and Tosk (right) participants for the 12 counties. The majority dialect spoken in each county is indicated by G for Gheg and T for Tosk.
Figure 4 shows that on average, the least correct ratings were attributed to the Gheg counties of Kukës, Lezhë, and Dibër. Among the Tosk counties, Gjirokastër received the least correct ratings, especially from Tosk participants. The other Tosk counties, and the Gheg counties of Durrës, Tirana, and Elbasan, were rated fairly positively, with Tirana receiving the highest ratings overall. Some differences emerge between ratings provided by Tosk and Gheg participants. In particular, ratings of Tosk counties are less uniform when given by Tosk participants, while average ratings of northern Gheg counties are slightly less negative when given by Gheg participants.
The output of the statistical analysis is reproduced in Figure 5, Table 3, and Table 4. Figure 5 displays regression lines extracted from the model, superimposed on actual ratings from the participants. For Gheg counties, the trend is quite clear: higher density comes together with higher ratings on the correctness scale, regardless of the participant’s dialect background. Table 3 indeed shows that for a density increase of 1 unit on the (base-10) logarithmic scale, that is, every time the density of Gheg counties is multiplied by 10, correctness ratings increase by approximately 3 points (precisely, 3.001 when participants are Gheg, 2.954 when participants are Tosk). This increase is significantly different from 0 (rows 1 and 2 of Table 3), and does not differ between Gheg and Tosk participants (row 1 of Table 4).
Table 3. Output of the emtrends() function showing the estimated slope of density (continuous predictor) for each combination of participant dialect and county dialect, and whether that slope differs from zero, for the model applied to correctness ratings.
Table 4. Output of the emtrends() function comparing whether the slopes in Table 3 differ between each combination of participant dialect and county dialect for the model applied to correctness ratings.
For Tosk counties, Figure 5 suggests that ratings provided by Tosk participants also increase along with density. For a tenfold increase of density in Tosk counties, correctness ratings from Tosk participants increase by 1.752 points, a slope that is significantly different from 0 (fourth row of Table 3) and comparable to that of Gheg counties (fifth row of Table 4). Ratings from Gheg participants, however, are usually relatively high and do not seem to be influenced by county density, with a non-significant increase of 0.235 on the correctness scale when the density of Tosk counties is multiplied by 10 (third row of Table 3), which also differs from their ratings of Gheg counties (second row of Table 4). The difference between Tosk and Gheg participants, in this case, is to be taken as a trend, as it does not reach the threshold of significance after correcting for multiple pairwise comparisons (last row of Table 4).
3.2. Pleasantness
Figure 6 displays mean ratings on the pleasantness scale for the 12 Albanian counties, with means for Gheg and Tosk participants presented separately. Overall, the lowest ratings were attributed to the Gheg counties of Kukës, Lezhë, and Dibër, and the Tosk county of Fier. Two counties which stand out for the high ratings obtained on the pleasantness scale are both Gheg: Tirana and Shkodër. Among Tosk participants, the Tosk county of Gjirokastër was also rated high on the pleasantness scale.
The output of the statistical analysis is reproduced in Figure 7, Table 5, and Table 6. It appears from Figure 7 that for Gheg counties, low-density counties tend to be rated as less pleasant than high-density ones, and more markedly so by Gheg than Tosk participants. Table 5 indicates that when the density of Gheg counties undergoes a tenfold increase, pleasantness ratings increase by an estimated 2.331 points when given by Gheg participants, and by 0.971 point when given by Tosk participants. These two increases are significantly different from 0, and significantly different from each other (Table 6, row 1).
Figure 7. Regression line extracted from the statistical model applied to ratings on the pleasantness scale, superimposed on real ratings provided by participants, split by participant dialect and county dialect, with county names indicated above their corresponding data.
Table 5. Output of the emtrends() function showing the estimated slope of density (continuous predictor) for each combination of participant dialect and county dialect, and whether that slope differs from zero, for the model applied to pleasantness ratings.
Table 6. Output of the emtrends() function comparing whether the slopes in Table 5 differ between each combination of participant dialect and county dialect for the model applied to pleasantness ratings.
For Tosk counties, the opposite is found: Figure 7 shows that low-density counties are rated as more pleasant than high-density ones, and that the trend is more marked for Tosk than Gheg participants. When the density of Tosk counties is multiplied by 10, Table 5 reports an estimated decrease of 3.064 points in pleasantness ratings given by Tosk participants, and a decrease of 1.080 points in ratings given by Gheg participants, both of which are significantly different from 0 (Table 5) and different from each other (Table 6, row 6).
4. Discussion
The motivation for the present perceptual dialectology study was to deepen our understanding of variation in Albanian beyond the parameters of traditional dialectology. Our aim was twofold: to investigate in which way the classic Gheg/Tosk division and the urban/rural divide interact in how correct and pleasant Albanians perceive the varieties spoken across the country to be; and to examine whether there are differences in these ratings depending on whether the participants have a Tosk or a Gheg background. To verify five hypotheses formulated based on previous literature, we asked 123 participants to rate the correctness and the pleasantness of the speech produced in each of the 12 counties of Albania on scales from 1 to 10, where 1 was least correct or pleasant and 10 was most correct or pleasant. These ratings were examined in relation to three main predictors: the participants’ dialect, the counties’ majority dialect, and the counties’ density, used as a proxy for relative urbanization.
4.1. Correctness
The first part of our analysis focused on the correctness scale, for which we formulated three hypotheses. H1 was concerned with the counties’ majority dialect, and predicted that Tosk counties would be rated more correct than Gheg counties. H2 addressed relative urbanization and predicted that within a given dialect area, urban counties would receive higher ratings of correctness. H3 was focused on the participants’ dialect, predicting that similar ratings of correctness would be produced by both Tosk and Gheg participants. Since these three factors—participants’ dialect, counties’ majority dialect, counties’ relative urbanization—were found to interact with each other in the statistical model applied to responses on the correctness scale, our three hypotheses are only partly supported.
Regarding H1, which was concerned with counties’ dialects, the results showed, as predicted, that the lowest correctness ratings were attributed to the three Gheg counties Kukës, Lezhë, and Dibër (see Figure 4). However, the highest ratings overall were also given to two Gheg counties, Durrës and Tirana. Thus, Tosk counties were rated higher than the lowest-ranking Gheg counties, but lower than the highest-raking Gheg counties. Correctness ratings for Gheg counties appeared to be strongly modulated by relative urbanization, with more urbanized counties triggering much higher ratings of correctness than less urbanized counties (Figure 5). The observation that some Gheg counties fared lowest is consistent with prior sociolinguistic work concluding that Gheg has a lower status than Tosk in Albanian society (Morgan, Reference Morgan2015, Reference Morgan2022; Dickerson, Reference Dickerson2021; Riverin-Coutlée, Kapia & Gubian, Reference Riverin-Coutlée, Kapia and Gubian2024). On the other hand, the finding that Tirana and Durrës ranked so high resonates with Morgan’s (Reference Morgan2015, Reference Morgan2022) proposal of a tripartite division, where Middle Albania (Shqipëria e Mesme) is perceived by Albanians as a different entity than the North and the South. While this may not be compatible with the traditional Gheg/Tosk division within Albanology, which categorizes Tirana and Durrës as Gheg-speaking locations (e.g. Gjinari et al., Reference Gjinari, Beci, Shkurtaj, Gosturani and Dodi2007), it supports an idea expressed elsewhere that “transitional dialects” are spoken in these areas, i.e. varieties with some overlapping Gheg and/or Tosk features (Çeliku, Reference Çeliku1965; Beci, Reference Beci2002). Additionally, the participants’ intuition about these areas intersects with the commonly observed emergence of urban varieties in fast-growing cities conducive to contact (see e.g. Kerswill & Wiese, Reference Kerswill and Wiese2022). The presence of a distinct urban variety in Tirana and Durrës, and more generally the linguistic grounding of the tripartite division, have yet to be confirmed. Nevertheless, our results suggest that Albanians’ representations align, at least partly, with the well-documented Gheg/Tosk division, although not independently from an urban/rural divide, which brings us to H2.
The results partly confirmed H2 in showing that as the population density of Gheg counties increased, so did Gheg and Tosk participants’ correctness ratings (Figure 5). The same was observed for Tosk counties rated by Tosk participants. These results are in line with those from work conducted in other linguistic settings, which reported a more positive perception of urban than rural speech (Cramer, Reference Cramer2016; Evans, Reference Evans, Cramer and Montgomery2016; Jeon & Cukor-Avila, Reference Jeon, Cukor-Avila, Cramer and Montgomery2016). However, the trend was less clear for Gheg participants rating Tosk counties: they gave similar ratings to all Tosk counties, regardless of their degree of urbanization. We may advance a few tentative explanations for this result. First, we note that ratings of Tosk counties by Gheg participants were all relatively high, suggesting that even less urbanized areas could have benefited from a generally positive attitude towards Tosk (Morgan, Reference Morgan2015, Reference Morgan2022; Dickerson, Reference Dickerson2021; Riverin-Coutlée, Kapia & Gubian, Reference Riverin-Coutlée, Kapia and Gubian2024). Second, previous research in perceptual dialectology has shown that participants had more detailed representations of differences between speech areas closer to where they lived (e.g. Cramer, Reference Cramer2016). This could mean that Gheg participants had a less detailed representation of differences between Tosk counties than Tosk participants did, and gave similar scores to the whole area, while Tosk participants were more aware of differences in degrees of urbanization across Tosk counties. In any case, as already discussed in the above paragraph concerning H1, our results suggest distinct attitudes towards heavily urbanized Tirana and Durrës compared to other Albanian counties. Interestingly, this perceived distinctness of Tirana and Durrës resonates beyond language-related attitudes to other spheres of Albanian society, for example, in the political and economic concept of the “Durana corridor” introduced by the Albanian government as a model for urban planning and development (Fabi & Piovene, Reference Fabi and Piovene2016).
H3 predicted similar ratings by Tosk and Gheg participants based on previous findings in perceptual dialectology studies where speakers of different backgrounds generally agreed on which variety was most correct (Dailey-O’Cain, Reference Dailey-O’Cain and Preston1999; Niedzielski, Reference Niedzielski, Long and Preston2002; Cramer & Montgomery, Reference Cramer and Montgomery2016). H3 was partly confirmed in that speakers agreed on correctness for Gheg counties. For Tosk counties, however, while Tosk participants’ ratings varied with density, Gheg participants gave uniformly high ratings to all counties (Figure 5). While possible explanations to this have been proposed in the discussion of H2, we have to keep in mind that the difference between Tosk and Gheg participants’ ratings is weak at best (Table 4), and we do not conclude that it makes our study much different from those having found general agreement on correctness between participants of different backgrounds.
4.2. Pleasantness
The second part of our analysis was concentrated on the pleasantness scale, for which two hypotheses were formulated. H4 predicted that Tosk participants would rate Tosk counties as more pleasant than Gheg counties, but no parallel hypothesis was formulated for Gheg participants. H5 was concerned with relative urbanization and predicted that more urban counties within a given dialect area would receive higher pleasantness scores. Once again, the three tested factors—participants’ dialect, counties’ dialect, and counties’ relative urbanization—interacted, providing nuanced answers to our questions.
H4 was formulated for Tosk participants assuming that Tosk would be rated as more correct; the absence of a parallel hypothesis for Gheg participants was based on the assumption that Gheg would be rated as less correct. This was because previous studies in perceptual dialectology have repeatedly reported cases of participants rating their own varieties as both correct and pleasant, but participants rating their own varieties as less correct were somewhat unpredictable in their pleasantness ratings (e.g. Dailey-O’Cain, Reference Dailey-O’Cain and Preston1999; Kuiper, Reference Kuiper and Preston1999; Fridland & Bartlett, Reference Fridland and Bartlett2006; Preston, Reference Preston, Chambers and Schilling2013). However, ratings on the correctness scale were not as straightforward as expected, with Gheg counties scoring both lowest and highest depending on relative urbanization (see Section 4.1). Likewise, pleasantness ratings show more complex trends than predicted, i.e. Tosk participants gave the highest pleasantness scores to the counties of Shkodër (Gheg) and Gjirokastër (Tosk), while Gheg participants attributed the highest mean pleasantness ratings to Tirana and Shkodër (both Gheg). At the other end of the pleasantness scale, Tosk and Gheg participants agreed on ranking lowest the Tosk county of Fier and the Gheg counties of Kukës, Lezhë, and Dibër. Below, we discuss these particular cases in turn, starting with the counties that were rated least pleasant and least correct.
Gheg counties of Kukës, Lezhë, and Dibër were rated lowest on both correctness and pleasantness, by both groups of participants. This is a similar finding to that of Dailey-O’Cain (Reference Dailey-O’Cain and Preston1999), where Eastern German varieties were rated as less correct and less pleasant than Western German varieties by both Eastern and Western Germans. This might also correspond to the stereotypes that Morgan’s (Reference Morgan2015) participants had in mind when they described Gheg with terms such as uncultured/thick, harsh, or backwards, none of which carry any pleasant connotation.
The case of the Gheg county of Shkodër is radically different: despite being geographically close to Kukës, Lezhë, and Dibër, it was rated very positively on the pleasantness scale by Gheg and Tosk participants alike. One reason that might contribute to the high pleasantness score of the Gheg county of Shkodër across both Tosk and Gheg participants could be linked to the fact that this area has distinguished itself over time as a center of intellectual and cultural influence. For example, it has been argued that Shkodër is one of the earliest cradles of Albanian renaissance and independence (e.g. Demiraj, Reference Demiraj2012).
Tirana, the capital city and another past and present influential cultural center in Albania, was also rated as having pleasant speech, and especially so by Gheg participants. Knowing that Tirana ranked highest on the correctness scale too, our results indicate that Gheg participants rated their own variety (or at least the Tirana subvariety) as both the most pleasant and the most correct, a pattern which turns out to be very much compatible with that observed in other prior perceptual dialectology studies in different language contexts (Dailey-O’Cain, Reference Dailey-O’Cain and Preston1999; Kuiper, Reference Kuiper and Preston1999; Preston, Reference Preston, Chambers and Schilling2013).
The case of Tosk speakers’ high pleasantness ratings for the county of Gjirokastër was unexpected. It was rated low on the correctness scale, which we attributed to the largely rural character of the county in Section 4.1, with Gjirokastër having the lowest population density among the 12 counties. Its concomitant high pleasantness score seems to suggest that Gjirokastër speech benefits from covert prestige, that is, a positive attitude held by members of a speech community towards their own nonstandard variety (Trudgill, Reference Trudgill1972; Labov, Reference Labov2006; Hernández-Campoy, Reference Hernández-Campoy2008). Covertly prestigious varieties are valued in situations of social proximity; they typically score low on status-related dimensions like correctness (contrary to overtly prestigious varieties; Labov, Reference Labov2006), but high on solidarity dimensions like pleasantness (Fiske et al., Reference Fiske, Cuddy, Glick and Xu2002). If this were the case, it would be all the more surprising given that Gjirokastër speech is part of the variety considered by many scholars to be the one on which standard Albanian was based in 1972 (Byron, Reference Byron1976; Kostallari, Reference Kostallari1984). Hypothetically, Gjirokastër’s high pleasantness score could be an indirect consequence of various government-based initiatives aimed at boosting the attractiveness of Southern Albania as a leisure and touristic destination, for instance through the “100 Villages” financial program which helps revitalize villages across Albania, and from which Gjirokastër benefited more than any other county (Government of Albania, Ministria e Bujqësisë dhe e Zhvillimit Rural, 2024). The historical center of Gjirokastër city is also one of the first (and few) UNESCO World Heritage Sites in Albania.
A reversed situation is observed with the Tosk county of Fier, which has been found across both Gheg and Tosk participants to be rated high on correctness but low on pleasantness. One possible explanation for the high correctness scores could be its urbanization (see Section 4.1), i.e. Fier is the county with the third highest density, after Tirana and Durrës. Its low pleasantness score is currently difficult to explain, although a similar scenario has been observed in other linguistic settings. For example, Chalier (Reference Chalier2021:250) reported that French speakers from both Switzerland and Quebec rated Parisian French high on correctness but low on pleasantness.
With regard to H5, which predicted higher pleasantness scores for more urbanized counties in general, our results showed that this is true only for Gheg counties: low-density counties were rated as less pleasant than high-density ones, with a more pronounced trend from Gheg than Tosk participants. Similar trends were found in other perceptual dialectology studies addressing the issue of the urban/rural divide (e.g. Jeon & Cukor-Avila, Reference Jeon, Cukor-Avila, Cramer and Montgomery2016). For Tosk counties, however, our results showed the exact opposite: low-density counties were rated as more pleasant than high-density ones, and the trend was more marked for Tosk than Gheg participants. We have suggested, in our discussion of the Gjirokastër case, the possibility that government-led programs to develop the tourism sector could have enhanced the attractiveness of rural areas of Southern Albania (whereas Northern Albanian counties such as those of Lezhë, Kukës, and Dibër benefited much less from these programs; see e.g. Government of Albania, Ministria e Bujqësisë dhe e Zhvillimit Rural, 2024). This idea, however, would need to be validated in future work using other types of methodological tools, e.g. interviews.
4.3. Limitations
Our study has a few limitations worth mentioning. First, we asked participants to rate the degree of correctness and pleasantness of the language spoken in the 12 administrative counties of Albania. As such, participants did not have the freedom to make more distinctions (e.g. within a given county) or to suggest different boundaries. Another type of task, for instance the map-drawing one (Preston, Reference Preston1989; Niedzielski & Preston, Reference Niedzielski and Preston2000; Garrett, Reference Garrett2010), could have revealed more detailed representations in some cases (e.g. closer to where participants come from; Cramer, Reference Cramer2016), less detailed representations elsewhere (e.g. as hypothesized for Gheg participants rating Tosk in Section 4.1), and, of particular relevance for this study, potentially more nuanced distinctions between urban and rural speech than those allowed by the format of our task. However, the processing and analysis of mental maps pose well-known difficulties (Montgomery & Stoeckle, Reference Montgomery and Stoeckle2013; Cramer, Reference Cramer2021); we judged the 12-county option to be a reasonable compromise for this exploration of Albanians’ representations, which should be seen as a starting point to be refined in follow-up studies.
Second, population density was used as a proxy for relative urbanization, a concept known to be elusive and difficult to quantify (Britain, Reference Britain, Busse and Warnke2022). We acknowledge that the actual density numbers used in our statistical analyses may not mean much to most participants, who were likely not guided by these numbers when rating speech correctness and pleasantness. Still, density is arguably an important and intuitive component of relative urbanization, and the real question remains to what extent Albanians conceptualize speech (and other social categories) in terms of urban versus rural beyond our own post hoc categorization, which future work could establish more directly. We also judged that a density measure at the county level was appropriate in the Albanian context, given the small and comparable size of the 12 counties (see Sections 1.2 and 2.3), but it may be less appropriate in other regions of the world and should be used cautiously. In the advent of a follow-up study using mental maps, more local measures of density could also be considered.
Third, we did not consider whether participants came from urban or rural areas, but it is plausible that the urban/rural background (or other social characteristics) of the participants may have affected some (but not all) of their ratings, in the same way that having a Tosk/Gheg background was found to affect some (but not all) of their ratings.
5. Conclusion
This study is one of the first to examine how non-linguists perceive language varieties spoken around Albania. We found that Gheg and Tosk participants viewed areas traditionally classified as Gheg and Tosk in a way that seemed modulated by relative urbanization. Tirana and Durrës speech stood out in being rated as particularly correct and pleasant. If one of the purposes of perceptual dialectology is to challenge traditional views on language variation (Preston, Reference Preston1999), this latter finding should be taken as an incentive to carry out further linguistic inquiry on the matter. The participants’ strong and consistent intuitions may indeed be tied to the emergence of an urban contact variety in these densely populated zones which have undergone tremendous sociodemographic changes in the past 30 years, the linguistic features of which have yet to be discovered.
Acknowledgments
The first author gratefully acknowledges the support provided by the DFG grant # 499336853. The second author was supported by the DFG grant #520195671 during the writing phase. Many thanks go to our students for their help with data collection and to the editorial team and reviewers for their feedback. The authors are responsible for all the opinions expressed in this article.
Competing interests
The authors declare none.
Open access
