Broad, strong, and soft: Using geospatial analysis to understand folk-linguistic terminology

Abstract

This study uses a modified online version of the “draw-a-map” task and Garrett, Williams, and Evans’ (2005b) “keywords” methodology to explore the geospatial distribution of different accent and dialect labels and descriptors in Greater Manchester, UK. Specifically, we consider the distribution of the three most frequent labels related to “accentedness”: Broad, Strong, and Soft, as provided by 349 Greater Manchester residents. This analysis finds that these descriptors were clustered in separate areas of Greater Manchester, suggesting that they were being used to describe perceptually distinct varieties of English. In order to uncover the nuances in these folk-linguistic terms, we consider how they correlate with other concepts emerging from the dataset, finding that they are being used to differentiate between varieties with contrasting social associations. By combining innovative approaches, this study demonstrates how the subtleties of folk-linguistic modes of awareness can be uncovered through in-depth analysis of the terminology employed to describe linguistic variation on a very local scale. In so doing, it paves the way for further development of draw-a-map techniques that will enable similarly nuanced analysis in different regions, thus pushing the sub-discipline forward.

1. Introduction

In both Niedzielski and Preston (2003) and Preston (2002:18), the authors present their view of speech perception as a continuum from conscious (aware, deliberate) reactions and comments on language use to unconscious (automatic, unaware) reactions. The study of folk-linguistics, as they define it, is concerned with uncovering and disambiguating those conscious reactions to language. These approaches can range from qualitative analysis of free-flowing conversations about language use to highly controlled experiments eliciting reactions to stimuli. Within folk-linguistics, the discipline of perceptual dialectology is concerned with “what non-linguists say about language variation, including where they think it comes from, where they think it exists, and why they think it happens” (Montgomery & Cramer, 2016:9). In other words, perceptual dialectology explores how non-specialists understand and talk about language variation and change. This paper sets out to explore the perceptual dialectology of Greater Manchester, both focusing on where, geographically, locals view different varieties to be spoken in the region and attempting to disambiguate the nuances in the terminology used to describe them.

One of the difficulties with this work is that non-specialists’ descriptions of language, or “folk metalanguage” (Preston, 2004), often do not align with the terminology used by linguists. For example, Preston (1996, 2004) discusses the folk use of the term “nasal,” apparently “incorrect,” in that the speech being described often does not align with the formal linguistic definition of nasalization. However, Preston notes the value in delinking the term from its formal definition and acknowledging the speech style and associated social types evoked by its folk-linguistic usage. Given the variation and complexity in the ways non-specialists describe language use, research in perceptual dialectology often seeks to code or group these descriptions in order to disambiguate more wide-reaching patterns. For example, researchers such as Long (1999), McKenzie (2006), and Montgomery (2007) have attempted to categorize folk-linguistic terminology as “positive,” “negative,” or (for some studies) “neutral.” While an effective method of quantitatively processing large amounts of varied and complex data, this research often acknowledges the complicating factor of the occasional ambiguity of these terms. The importance of consideration and understanding of folk-linguistic terminology terms is also brought to the fore by Cramer (2018) in her discussion of the benefits of combining the “emic” and the “etic” in perceptual dialectology. She argues that locally relevant labels accessed via methods, such as the draw-a-map task or more qualitative approaches, should be both carefully considered in interpreting the data and ideally incorporated into any subsequent tasks, such as voice placement and rating surveys. This careful examination of terminology then provides a much more nuanced understanding of attitudes, ideologies, and language use in a given region.

Overall, then, researchers in folk-linguistics are working toward a framework in which folk-linguistic terminology is not dismissed as an insufficient attempt to evoke specialist knowledge. Instead, understanding of this terminology is valued as both an entry point for exploring the social types linked to language use and a methodological tool for either grouping descriptions together to uncover wider trends or to be incorporated into the design of locally relevant surveys. As such, it is particularly important to not simply take this terminology at face value, but to attempt to disambiguate finer shades of meaning and gain a nuanced understanding of how folk descriptions of language are operating at a local level.

This disambiguation is the intention of the present paper, which explores the perceptual dialectology of Greater Manchester, a metropolitan county in the North West of England. This study uses an online version of the “draw-a-map” task (see section 2) to examine how the labels and concepts associated with different local English varieties pattern geographically in the region. Specifically, we explore how the terms Broad, Strong, and Soft (the three most frequent terms to describe a variety’s perceived difference from a non-local standard variety within the dataset) correlate with other salient concepts that emerged from the analysis, such as Status and Social Attractiveness. In doing so, we aim to disambiguate the locally relevant social meanings of these terms, demonstrating how they are not only being used to distinguish between different local varieties, but simultaneously act as a subtle evaluation of the variety and its associated social types. This paper begins with an overview of the draw-a-map methodology and the research location before describing the methodological design, introducing the dataset, and detailing the process of coding and analyzing this data. Section 5 (“Results”) then explores the geospatial distribution of the labels Broad, Strong, and Soft in Greater Manchester alongside other related concepts. Finally, the paper concludes by examining how these patterns align with previous research in language attitudes and perceptual dialectology and what this tells us about the social meaning of these examples of folk-linguistic terminology.

2. Draw-a-map tasks

Developed by Preston (1989), one of the key methodologies employed in perceptual dialectology is the draw-a-map task. Participants are asked to draw boundaries on a “base map” of a region representing where they consider different accents or dialects to be used, along with some combination of dialect labels, descriptions, or evaluations. This task seeks to elicit information about respondents’ mental maps by employing similar methods to those used in cultural geography (Gould & White, 1986). The method can be used to elicit a great deal of information—including the names, placements, and extents of dialect areas, as well as their linguistic features and social meanings. However, previous approaches to analyzing perceptual data have tended to be quantitative and to focus primarily on broad, generalized composite maps (Montgomery & Cramer, 2016:12).

The draw-a-map method, albeit relatively simple and straightforward, generates debate regarding its design, particularly regarding the level of detail that should be provided to respondents on base maps. Wales (2006:58–59) argues for the use of maps containing additional detail, so as to address recurring issues relating to respondents’ lack of geographical knowledge. The inclusion of some geographical cues is generally considered to be necessary, although the level of necessary detail provided naturally decreases when investigating larger geographical areas (Montgomery & Cramer, 2016). Previous USA- and UK-wide studies (e.g., Preston, 1989; Inoue, 1996) provided state and county lines on base maps, though found that this oriented respondents towards the lines. Romanello (2002) and Stoeckle (2012) both marked cities, roads, and administrative boundaries in their studies of smaller European regions. Although “the level of detail included on maps does not have a significant impact on the number of areas drawn” (Montgomery & Cramer, 2016:13), further research is necessary to determine whether the level of detail included on maps influences the placement and extent of the dialect boundaries drawn.

Perceptual dialectologists have recently moved away from examining clusters of lines (i.e., Preston, 1989) and from tracing lines using digitizing pads (i.e., Preston & Howe, 1987) towards the use of Geographic Information System (GIS) software (Montgomery & Stoeckle, 2013). This software offers a systematic approach to analyzing subjective spatial and attitudinal data and allows the latter to be explored in greater depth than in previous studies. Montgomery and Stoeckle (2013:52) posit that GIS software is “a stable and useable method” for aggregating, processing, and displaying perceptual dialectological data, emphasizing its benefits as a widely accessible software that produces high resolution maps and composite maps showing multiple perceptual areas. Various types of heatmaps can be created using GIS software, all of which typically show levels of agreement amongst respondents regarding dialect boundaries, with darker areas representing higher levels of agreement or perceptual “cores” (Montgomery & Stoeckle, 2013:74).

Recent developments in the use of GIS software within the field include coding practices to attach demographic information to the boundaries drawn, allowing researchers to query perceptual data and to show how dialect perceptions are stratified by factors such as age and gender (Cukor-Avila, 2018). They also include content analysis and the qualitative coding of responses, enabling researchers to analyze the social meaning of the spatial boundaries highlighted by respondents. Evans (2013), for example, created maps using thematic categories (e.g., country) and isolated regions (e.g., Seattle) in order to explore their associated labels in her work on Washington, USA, thus demonstrating an ability for researchers to prioritize, and work from, either geographical or social information.

To complement these significant enhancements in data analysis techniques, there exist opportunities to develop more sophisticated data collection techniques within the field. One area that shows great potential is the use of web-based approaches that enable the collection of larger volumes of data from greater numbers of more diverse participants across a wider geographical area. Online, digital data collection has the added benefit of removing what has been, to date, an essential step in the data analysis process: the digitization of hand-drawn maps prior to inputting them to GIS, which is often a time-consuming task and can introduce human error (in scanning position, for example).

Within the UK, there have been a number of studies applying the perceptual dialectological method to exploring folk perceptions of variation, beginning with Inoue (1996). This first study asked English students to draw their perceived dialect divisions on a map of the UK. These lines were then grouped to create a map of “subjective dialect divisions” in the region, broadly dividing by country, and then dividing England into Northern, Southern, Eastern, Western, and Midland. Subsequent work from Montgomery (2007, 2012, 2016) has used a similar method, asking participants to create hand-drawn mental maps of England, with a focus on Northerners’ perceptions of Northern English dialects. Instead of looking for the most frequent boundary groupings, Montgomery used an innovative method of digitally tracing the lines to create heatmaps of the different regions. In addition, Montgomery’s work looked beyond simply creating a composite map to explore the evaluation of the identified varieties and the cultural factors underlying these perceptions. Similarly, Coupland, Williams, and Garrett (1994) and Williams, Garrett, and Coupland (1996) explored the perceptual dialectology of English in Wales, uncovering not just the most prominent areas, but the various associated social types.

In a much more localized area, Pearce (2009) used an alternative perceptual dialectology approach (the “little arrow” method [Weijnen, 1946]) to explore perceptions of a relatively small region of the UK, asking participants in the North East of England to rate different local place names according to similarity or difference from their own hometown, deriving broad perceived dialect areas in the region. Pearce then compared these perceptual regions to existing research on speech production in the regions, providing evidence for clear linguistic factors underpinning these folk dialect divisions. In a similarly small region, Braber (2016) has explored how Nottingham speech is perceived in the context of the wider country by local sixth-formers (aged 16–18). In addition, Braber employed a “mind map” methodology, asking the students to annotate a map of the East Midlands region with their perceptions of language use in the region, followed by a group discussion of the task. Combined, these methodologies uncovered not only the lack of cultural prominence of the East Midlands dialect region, but the generally negative perception of the region and the students’ tedencies toward “denial” of the accent.

Taken together, these studies have advanced our understanding of the evaluation of regional varieties in the UK, providing insight into the linguistic, cultural, and social factors influencing perceptions. Notably, research such as Braber (2016) has demonstrated the effectiveness of employing these methods in a relatively small geographical region, particularly in combination with an in-depth examination of people’s more qualitative responses. In line with this approach, the present study uses the draw-a-map technique to examine language attitudes in a relatively small geographical area, Greater Manchester. As detailed in the following section, this region is particularly ripe for sociolinguistic investigation, encompassing a wide range of communities and social groups, and having undergone a great deal of social change in recent years.

3. Research location

Greater Manchester (Map 1) is a metropolitan county and city region in the North West of England with a population of 2.8 million across 493 square miles (Office for National Statistics, 2022). The population is 80% White British, with the largest ethnic minority groups being Asian (or Asian British) at 10% and Black (or Black British) at 3% (Office for National Statistics, 2013). Far from being a historically unified whole, Greater Manchester is made up of areas which, prior to 1974, were part of the historic counties of Lancashire, Yorkshire, Cheshire, or Derbyshire. Greater Manchester is divided into ten boroughs: Bolton, Bury, Manchester, Oldham, Rochdale, Salford, Stockport, Tameside, Trafford, and Wigan, which, due to their wide-ranging histories and traditions, have their own distinct social and cultural makeups. In recent years, the city-region has become increasingly established and prominent, electing its first mayor in 2017 and playing a key role in the UK Government’s Northern Powerhouse agenda. ¹

Map 1. Map of the ten boroughs of Greater Manchester and their location within the UK. This map was created using data derived from the following sources. Contains National Statistics data © Crown copyright and database right 2022. Contains OS data © Crown copyright [and database right] 2022.

The city of Manchester itself is well-known nationally and internationally, in part due to its historic role as the center of the cotton industry during the Industrial Revolution, but more recently due to its cultural and sporting influence in the areas of music, literature, and football. Other areas of the region have national prominence to greater or lesser degrees, but none have the reputation of Manchester itself.

There have been a variety of detailed speech production studies into various areas within Greater Manchester (e.g., Shorrocks, 1998; Moore, 2011; Baranowski & Turton, 2015; Baranowski, 2017; Turton & Baranowski, 2021; Dann, Ryan, & Drummond, 2022; Ryan, Dann, & Drummond, 2022); however, there has been a lack of research into the region of Greater Manchester as a whole. In particular, perceptions of dialect variation within and across Greater Manchester have received no attention in the literature to date, aside from as part of wide-scale surveys of much larger regions. However, in a discussion of one such large-scale survey, Montgomery (2016) notes the recent emergence of a Manc (Manchester) dialect area in the mental maps of UK respondents, which he attributes to the more recent cultural prominence mentioned above. The value in focusing on Greater Manchester as a whole, particularly with regard to perception research, lies in its ambiguous social status – is it a clearly defined region, or is it an administrative collection of geographic areas that are more strongly associated with historic counties?

4. Methodology

4.1. The online draw-a-map tool

Perceptual dialectology research requires the use of mixed methods, since it quantitatively aggregates boundaries drawn by hand—or digitally, in this case—and simultaneously captures the social meanings attached to spatial data. In essence, it makes generalizations regarding the perceived distribution and extent of dialect areas (quantitative) and explores individualized attitudes and the subjective nature of dialect perceptions (qualitative). The online software developed for this project (detailed below) was designed in such a way as to enable a mixed methods approach to data collection and analysis.

The study used a specially-designed online tool to collect perceptual dialectological data unsupervised between September 2019 and February 2021. Greater Manchester residents were recruited via social media and local newspaper and radio advertisements then directed to the online tool, where they were asked to provide some basic demographic information before progressing to the task. As shown in Figure 1, participants were presented with instructions asking them to draw boundaries, and then label and describe the accents/dialects within them. Next, as shown in Figure 2, they were shown a map of Greater Manchester with the ten boroughs labeled and invited to use their mouse or touchscreen to draw boundaries. After drawing each boundary, a comment box appeared, in which they could add any qualitative or evaluative comments. Participants also had the option of editing their boundaries and comments at any point before submitting.

Figure 1. Instruction presented to participants at the start of the task

Figure 2. Example of a map in the process of being completed (top), and the comment box presented to participants after drawing each boundary (bottom)

The digital, web-based data collection tool presented here has numerous benefits when compared to the traditional method of collecting hand-drawn maps in person. An early attempt to collect perceptual dialectological data online was made by Jones (2015), who compared data collected using Qualtrics survey software and ImageBot editing software with data collected via a hand-drawn map task. The digital component of his data collection methods contained three steps: an instructional video, the draw-a-map task, and a survey. The draw-a-map task involved the use of draw-and-text, if applicable features, although ImageBot also enabled participants to use auto-shape and color-coding features. Whilst the options made available by such digital technologies allow people to respond to the draw-a-map task in new and different ways, they also highlight the need for researchers to ensure that they are gathering comparable and usable data. Jones’s (2015) research instrument, albeit a notable advance in studies of language perception, led to respondent fatigue and low participation because respondents were required to download, save, and upload their maps, as well as complete additional tasks. Hence, Jones (2015:42) calls for “further refinement in terms of creating a tool that is user-friendly and not overly cumbersome for the respondents.”

That said, there are clear benefits to using an online approach. Firstly, online data collection has the potential to reach a wider audience across a larger geographical space. An online research instrument is more accessible to the sample population selected and, thus, has the potential to generate vast amounts of data (Lefever, Dal, & Matthíasdóttir 2007). Secondly, researchers are not limited to recruiting respondents through their social networks, which often lack diversity: “the use of crowdsourcing … is a promising approach to collecting more representative samples, as compared with the commonly used undergraduate participant pool” (Behrend et al., 2011: 802). Thirdly, web-based technology is currently used across numerous devices—e.g., PCs, laptops, tablets, and smartphones. As such, an online tool that is usable across various platforms is essentially available to anyone, anywhere, and at any time. This renders researchers uninhibited by issues relating to time and resources; the data collection is no longer limited to time spent in the field. Once an online tool is launched, it can collect data efficiently for an allocated period with minimal effort and planning on the part of the researcher. Of course, one of the disadvantages of online data collection is that respondents are self-selecting and may have a shared set of social and/or attitudinal characteristics; as such, their responses must be interpreted with caution (Coupland & Bishop, 2007).

4.2. Participants

Four-hundred-and-four respondents took part in the draw-a-map task after responding to calls for participants on social networking sites and in local media. The only exclusion criteria to take part in the study were age (participants had to be over five years old) and location (they must currently live in Greater Manchester). Of these 404 respondents, 43 were excluded because they did not currently live in Greater Manchester, 7 were excluded because they did not include any demographic data, and 5 were excluded because their maps were incomplete. This left a sample of 349 respondents. A full demographic breakdown of this sample can be seen in Table 1.

Table 1. Demographic details of the participants

Overall, this sample is relatively representative of Greater Manchester in terms of gender and borough, although Bury and Trafford are a little overrepresented and Salford a little underrepresented. Although the Manchester borough has more than double the participants of some of the other boroughs, this is in line with its higher population. In terms of the age distribution, there is a skew toward 26–45-year-olds. In addition, there is a clear skew toward White British respondents, and most other ethnicities are underrepresented in comparison to the demography of Greater Manchester. This is perhaps a drawback of the online sampling method, which relies heavily on self-selecting respondents. While it was relatively easy to target specific geographic areas through social media (e.g., local community Facebook pages), targeting specific ethnic groups and age groups beyond that was more challenging.

As a result of the demography of our sample, our participants’ knowledge is perhaps more likely to skew toward “traditional” dialects, as opposed to dialects spoken in more multicultural communities. Indeed, there were only 12 comments mentioning varieties associated with a specific ethnic group. However, the good geographical distribution of participants goes some way to mitigate proximity effects (Montgomery, 2012), where participants are likely to identify more accent/dialect variation in their local area.

4.3. Data coding and analysis

4.3.1. Coding the comments

Once we came to the end of our data collection period, the individual map comments were collated and analyzed in order to create a thematic coding scheme. We developed the coding scheme using a combination of Garrett et al.’s (2005b) proposed “keywords” methodology and more traditional approaches to language attitudes (e.g., Zahn & Hopper, 1985). Garrett et al. (2005b) describe the application of the keyword approach in two studies: Garrett, Evans, and Williams (2006), which collected cognitive responses to the word “globalisation,” and Garrett, Williams, and Evans (2005a), a study of attitudes towards Australian, US, British, and New Zealand Englishes. In both studies, the researchers performed a keywords analysis of the qualitative responses to each of the various concepts, grouping responses according to emerging themes, such as “unity,” “togetherness,” and “toughness.” Garrett et al. (2005b:44) note that a key benefit of using a keywords approach is that it enables the identification and interrogation of concepts that are infrequent in the dataset as a whole, but which “reflect some illuminating salience and importance in concentrated pockets of cognitive representations.” In other words, this approach avoids overlooking concepts that are important to only a small pocket of respondents or in one specific context. This differs to the approach of much language attitudes research which has either used keywords as a preliminary approach to identifying the most salient concepts (see, e.g., Preston, 1999) or asked participants to either rate varieties or draw on a map according to a set of attributes pre-defined by the researcher (e.g., Niedzielski & Preston, 2003; Montgomery, 2007).

Evans (2013) applied a keywords approach directly to perceptual dialectological data in a study of perceptions of English in Washington state and modified the approach for use with draw-a-map data, analyzing the qualitative comments together to find 20 distinct categories (Evans, 2013:277). These included linguistic descriptions (“Pronunciation,” “Slang”), regional labels (“Canadian”), and more social labels (“Country,” “Gangster”). Comments which fit in more than one category were coded into each. Although this method risks overrepresenting certain respondents’ comments, it avoids the subjective and reductive process of having to choose just one category for each comment. In addition, Al-Rojaie (2020) has used this technique in a study of the perceptual dialectology of Qassim in Saudi Arabia. After identifying salient dialect areas, Al-Rojaie used the keywords approach to group the various descriptors into six key categories, creating heatmaps for each to correlate them with the different varieties. Approaching the data in this way allowed for examination of the ideologies underpinning people’s perceptions of language use in the region.

In the present study, we took a similar approach to Garrett et al. (2005b), Evans (2013), and Al-Rojaie (2020), and first performed a keywords analysis of all the 1167 comments in our dataset. From this, we noted that most respondents named the dialect region/variety, and then provided additional evaluative comments and linguistic descriptions. Given this hierarchical structure of many of the comments, we did not follow Evans (2013) in treating dialect region labels as categories in themselves alongside the evaluative and descriptive labels. Instead, we identified the top 19 dialect regions noted by participants and coded this in the Dialect Region category (this approach is in line with Montgomery, 2007 and 2012). The specific label given to the variety (e.g., “Manc,” “Lanc,” “Wiganese”) was also preserved in a Dialect Label category. The linguistic descriptions of the variety were coded as one of four categories: Lexis, Morphosyntax, Phonology, or Aesthetic Quality. This latter category encompasses descriptors such as “Nasal,” “Fast,” “Slow,” and “Whiney”, in line with Long’s (1999:213) “paralinguistic” group.

A further 12 categories referred to non-regional or attitudinal social descriptors. These were created from the keywords analysis, but with reference to the more traditional categories used in language attitudes research of status and social attractiveness (e.g., Coupland & Bishop, 2007; Sharma, Levon, & Ye, 2022). First, we found that many of the comments fit well with Zahn and Hopper’s (1985) superiority factor, which combines social status, competency, and intellectual status. For example, “working/lower class,” “common,” “rough,” “chavy,” ² “unintelligent,” and “not proper English” were frequent labels in our dataset. On the flip side of this, “middle/upper class,” “posh,” “well spoken,” and “educated” also occurred frequently. These were coded in a Status category as either Low Status or High Status, respectively. In line with Zahn and Hopper’s (1985:119) attractiveness factor, we found that many respondents commented on the “social and aesthetic appeal” of the variety. As such, the Social Attractiveness category coded comments as either Attractive (e.g., “friendly,” “warm,” “down to earth,” “honest,” “trustworthy”) or Unattractive (e.g., “harsh,” “shrill,” “whiney,” “untrustworthy”). In contrast to Zahn and Hopper (1985), a dynamism category did not emerge from our keywords analysis. The remainder of the categories emerged directly from the keywords analysis. Some concepts, such as Rural (e.g., “country,” “farmers”) and Historical (e.g., “old fashioned,” “traditional”), represent categories in their own right. The focus of the present paper, the Accentedness category, relates to the folk-linguistic concept of accentedness, which assumes that “somewhere there is ‘accent-free’ speech” (Preston, 2016:180). As such, this category includes descriptors which appear to relate to the perceived intelligibility of the variety or distance from Received Pronunciation (RP), and comprises the terms Broad, Strong, Soft, Thick, and Heavy.

This approach to coding the dataset follows Cramer’s (2018) call to combine the “emic” (from the perspective of the local community) and the “etic” (more generalizable and meaningful to the scientific community) in perceptual dialectology. We have built the coding scheme from the bottom up, first considering which categories emerge organically from the dataset, then drawing upon the existing literature from language attitudes and perceptual dialectology to narrow down, group, and label our categories, enabling comparability with previous research. This method enables consideration of how key factors in language attitudes research, such as Status and Social Attractiveness, operate in this specific research context and of nuanced and locally relevant concepts, such as the differentiation of the Broad and Strong varieties described in this paper.

Using the coding scheme developed from this keywords analysis, each of the comments were coded into one or more of the categories. For example, the comment “Wigan. Broad, unique words for things. Old fashioned” was coded into the following categories: Dialect Region: Wigan; Dialect Label: Wigan; Accentedness: Broad, Unique, and Historical. And the comment “Lancastrian or Yonner accent. Think it sounds quite common. Long exaggerated vowel sounds, like noooo” was coded as: Dialect Region: Lancashire; Dialect Label: Lancastrian, Yonner; Status: Low Status; Phonological Feature: GOAT vowel. Where relevant, one comment could be coded into more than one category. For example, “shrill” would be coded as Social Attractiveness: Unattractive and Aesthetic Quality: Shrill/Whiney.

4.3.2. Processing the maps

At the end of this process, 1167 comments from 348 individual maps were labeled with the map and boundary number, as well as the demographic details of the respondent, and coded according to the keywords analysis. The next step comprised aggregating the coded data and the mental maps in ArcGIS Pro. Using a modified version of the method proposed by Montgomery and Stoeckle (2013), we first imported the 349 map images into ArcGIS Pro and georeferenced them, bringing them into alignment with an existing map of the region. Each of the lines of the map were then traced to create a dataset of polygons, labeled to cross-reference with their corresponding coded qualitative comment, and then “clipped” to the boundaries of Greater Manchester. Next, the demographic and coded qualitative data was “joined” to this dataset, attributing each polygon its corresponding information. As a result, the full dataset of polygons could be “queried,” and groups of polygons relating to a specified attribute could be extracted. For example, as shown in Figure 3, all the polygons coded as having the “Manc” dialect label could be selected. Finally, heatmaps were created by joining selected groups of polygons to a postcode sector map of Greater Manchester. The map could then be colored according to the number of polygons intersecting, or the “Join Count,” within a specified postcode boundary. This process is shown in Figure 3.

Figure 3. Workflow for the creation of heatmaps using traced “mental map” boundaries, from a full set of polygons (left), to a filtered set representing only those labeled as “Manc” (center), to a heatmap colored according to the Join Count of polygons intersecting within a postcode boundary (right)

These maps, and all subsequent ones, were created using data from the following sources. Contains National Statistics data © Crown copyright and database right 2022. Contains OS data © Crown copyright [and database right] 2022. Contains Royal Mail data © Royal Mail copyright and database right 2015.

5. Results

5.1. Accentedness

As noted in the previous section, the focus of this paper is the geospatial distribution of terms in the Accentedness category, which comprised Broad (n=93), Strong (n=33), Soft (n=32), Thick (n=9), and Heavy (n=2). Of these, there were too few comments to map the terms Thick and Heavy, so we focus this analysis on the former three, asking what is the social meaning of the folk linguistic terms Broad, Strong, and Soft in Greater Manchester?

Beginning with the comments themselves, Table 2 shows a sample of the comments for each of these categories. Taken together, these comments alone provide some insight into how these terms are being used to differentiate the different varieties of English in Greater Manchester. First, the use of either Strong/Broad or Soft appears to often be differentiating between more or less “accented” varieties. Indeed, some comments overtly put these labels on either end of a spectrum, such as the description of the Trafford accent as “softer, less broad.” There is also regularly an association between Soft and more prestigious varieties. At the other end of the spectrum, Broad and Strong both appear to be referring to more “accented” varieties. Indeed, both these terms are sometimes used in the same comment. However, a picture is also beginning to emerge of a subtle difference in the social meaning of these terms. In general, the Strong comments include more pejorative descriptions than the Broad comments, with terms such as “rough,” “common,” and “aggressive.” In contrast, some of the Broad comments appear to indicate a more socially attractive variety, with comments such as, “warm,” “friendly,” and “honest.”

Table 2. Sample of comments including the terms Broad, Strong, and Soft from the draw-a-map task

In order to explore this impressionistic finding in more detail, Maps 2, 3, and 4 show the geospatial distribution of the three terms, Broad, Strong, and Soft, in the region. It shows that these terms do indeed appear to generally be describing different varieties within the region, with the Broad boundaries generally clustering in the northern boroughs, Strong clustering in the Salford/North Manchester area, and Soft clustering in Trafford, Stockport and South Manchester.

Maps 2, 3, and 4. Heatmaps showing where the labels Broad (top left), Strong (top right), and Soft (bottom) were most used to describe accent/dialect areas in Greater Manchester. Postcode areas are colored by “Join Count,” or the number of intersecting boundaries in that area, with yellow being the most

Using this indication of where the varieties described by these terms are located, the following section explores how these areas are described more generally in the draw-a-map task. This provides us with an indication of the general conceptualization of these varieties as a whole, allowing us to make inferences about the social meaning of the folk-linguistic terms used to describe them.

5.2. Status

In the qualitative comments shown in Table 2 in the previous section, there was an indication that the labels Broad, Strong, and Soft, in particular the latter two, may carry connotations of the social status of the variety being described. It is also important to consider this dimension due to the wealth of language attitudes literature demonstrating the ubiquity of social status (along with social attractiveness) in the evaluation of language varieties (for a review, see Dragojevic et al., 2021). In order to explore this connection in more detail, Maps 5 and 6 show heatmaps of the geospatial distribution of comments that we have grouped as High Status and Low Status. In our coding scheme, High Status included comments such as “posh,” “upper class,” “educated,” “rich,” and “well spoken,” while Low Status included comments such as “rough,” “chav,” “working class,” “scally,” ³ “common,” and “poor.”

Maps 5 and 6. Heatmaps showing where comments coded as High Status (left) and Low Status (right) were most used to describe accent/dialect areas in Greater Manchester

Regarding the status of the different varieties of English in Greater Manchester, there is a clear North/South divide, with the southern boroughs of Trafford and Stockport, along with South Manchester, being described as High Status. In contrast, the Low Status labels cluster primarily in Salford, followed by Rochdale and Oldham, and then Tameside. The geospatial distribution of these categories provides some insight into the social meaning of the three Accentedness terms. The Soft area seems to correlate almost exactly with the High Status area, while the Strong area correlates with the Low Status region. Broad doesn’t correlate perfectly with either of these areas, but these maps indicate that it is certainly not seen as High Status, and the Broad areas of Rochdale and Oldham are sometimes considered Low Status.

5.3. Social attractiveness

Another important theme that emerged from the comments including the terms Broad, Strong, and Soft, and as well as being a key dimension in language attitudes more widely, was Social Attractiveness. We coded respondent comments as either Socially Attractive, including labels such as “friendly,” “warm,” and “pleasant,” or Socially Unattractive, including labels such as “ugly,” “harsh,” “shrill,” or “aggressive.” Maps 7 and 8 show heatmaps of where these comments were grouped.

Maps 7 and 8. Heatmaps showing where comments coded as Socially Attractive (left) and Socially Unattractive (right) were most used to describe accent/dialect areas in Greater Manchester

As with the Social Status categories, there is a slight correlation between the distribution of the Accentedness labels and the areas viewed as either Socially Attractive or Socially Unattractive. Once again, Salford and the closely surrounding areas, which were labeled as Strong and generally considered Low Status, are described in a more pejorative way. Language in this region, and particularly the more urban postcodes, is therefore generally considered to be both the lowest status and least socially attractive. The High Status and Soft area in the south of Greater Manchester was also sometimes labeled as Socially Unattractive. In comparison, the northern boroughs, and particularly the Bolton/Bury area, were most often considered Socially Attractive. The area described as Broad roughly correlates to this Socially Attractive region and was only very rarely labeled as Socially Unattractive.

5.4. Rural and historical

Finally, there were two further categories that, like the Broad label, were focused on the northern boroughs. As shown in Maps 9 and 10, these were Rural, which includes any mentions of concepts linked to rurality, such as “farmer,” “country,” “hillbilly,” or “yokel,” and Historical, which includes terms such as “traditional,” “old timey,” “classic,” and “quaint”.

Maps 9 and 10. Heatmaps showing where comments coded as Rural (left) and Historical (right) were most used to describe accent/dialect areas in Greater Manchester

Areas that were coded as Rural are clustered in the north-eastern boroughs of Bury, Rochdale, and Oldham, while the Historical areas are clustered in the north of Greater Manchester, including a little of North Salford, but are largely focused on Wigan and Bolton. Previous research in language attitudes in Britain has highlighted the importance of these two concepts in shaping perceptions of a variety. For example, Dann (forthcoming) discusses the general trend in large-scale language attitudes surveys for rural varieties to be upgraded relative to their urban counterparts, perhaps due to their association with more traditional/historical speech, unaffected by social and linguistic change in more urban areas.

6. Discussion

The previous sections have a provided an overview of the geospatial distribution of a variety of concepts that emerged from the draw-a-map task comments. It showed that the folk-linguistic Accentedness labels, Broad, Strong, and Soft, were being used to differentiate between three varieties within the relatively small region of Greater Manchester. Comparison of the distribution of these labels with four other key concepts that emerged from the keywords analysis, Status, Social Attractiveness, Rural, and Historical, demonstrated how these accent/dialect areas were being conceptualized more generally. This provides insight into the social meaning of these folk-linguistic terms.

First, we found that the terms Broad/Strong and Soft appeared to be differentiating between more and less “accented” varieties, respectively, with the less accented varieties likely invoking something similar to General Northern English (Cardoso et al., 2019; Strycharczuk et al., 2020). In this context, the term Soft correlated strongly with the areas that were described as High Status, with both of these descriptors predominantly used to describe the south of Greater Manchester. However, this is not to say that people were always describing something close to Received Pronunciation. Instead, the comments would often refer to a “softer” version of Mancunian or Northern varieties, presumably with fewer highly salient local features. Indeed, phonetic analysis of speech of speakers from South Manchester in a separate strand of the project (Drummond et al., 2022) found speakers from the boroughs of Trafford and Stockport to employ widely used regional features, such as the lack of a trap–bath split, but they were less likely to use features that were more local, such as front realizations of /u/ in pre-l environments, e.g., in “school.” This is a variety spoken across Northern England, often by more middle-class Northerners. As described by Strycharczuk et al. (2020:2), this variety “can be expected to display typically northern features, like the northern bath and strut, but not more narrowly defined northern features.” Beyond the term Soft discussed here, other descriptions of this variety that emerged from the draw-a-map task align with this characterization, including: “posh Manc,” “Trafford has a very very distinctive accent, the words are pronounced more eloquently even in perhaps rougher areas,” “metropolitan/well spoken with varying broadness,” “can tell their northern and from around the Manchester area but difficult to pin point exactly where the accent is from.”

In comparison, the terms Broad and Strong both seemed to be referring to more “accented” and localized varieties. However, the varieties described by these terms occupied different perceptual dialectological regions in Greater Manchester and appear to have contrasting social meanings. In contrast to the distinction with Soft, the distinction made between these two groups seemed to be between different accent groups, rather than simply the degree of accentedness. The term Strong was generally used to describe the urban variety spoken in the center of Greater Manchester and, particularly, in the Salford area. Broad, on the other hand, was used most often to describe accents/dialects in the northern boroughs of Greater Manchester. Again, this grouping correlates with what is known about actual speech variation in Greater Manchester, with the northern area of Greater Manchester belonging to a different dialect area to that of Manchester itself (Hughes, Trudgill, & Watt, 2012; Baranowski & Turton, 2015; Drummond et al., 2022). For example, the speech data collected in the Manchester Voices project (Drummond et al., 2022) shows that speakers from the north of Greater Manchester variably make use of a number of features which speakers from the central area do not—e.g., a fronter and closer mouth vowel, nurse∼square merger, and monophthongal face and goat.

The variety labeled as Strong was indicated to be both the most Low Status and Socially Unattractive in the region, with notably pejorative descriptions, such as “rough,” “common,” “harsh,” and “grating.” By contrast, while the region labeled as Broad was not considered High Status, it was not as Low Status as the Strong region, and the Bolton/Bury area in particular was most likely to be considered Socially Attractive. In other words, the Broad area was more likely to be described with terms such as “friendly” or “pleasant.” In addition, language in the west of the Broad area was often linked to the past, being described as “traditional” or “quaint,” while language in the east of the Broad area was often described in terms of rurality. In this way the term Broad is different from the terms Strong and Soft. Whereas Strong and Soft seem to be used directly in opposition to each other in terms of their perceived social status and accentedness, Broad does not seem to operate on this same scale, and instead carries different additional social meanings. This could perhaps be due to the fact that the area occupied by the term Broad traditionally has a different dialect to the area occupied by Strong. However, due to both Broad and Strong being “accented” varieties, additional social meanings are needed to distinguish these dialect areas from each other.

Research from Cooper (2019) in South Yorkshire, a nearby metropolitan county to Greater Manchester, has indicated that folk-linguistic uses of the term “broad” may have links to more “traditional” speech. He discusses how “broad” was often used to characterize speech in Barnsley as a notably old fashioned and unintelligible variety of Yorkshire English, as well as indexing “certain characterological figures like ‘Yorkshire farmers’” (Cooper, 2019:78). As such, it seems that in Greater Manchester, Broad is occupying a similar folk-linguistic function, delineating the more traditional, pleasant, and “folksy” variety perceived to be spoken in the more rural northern boroughs from both the less Socially Attractive, urban variety and the more middle-class, less regionally distinctive variety spoken in the more affluent south of Greater Manchester.

The distinction between these terms that emerged in this study mirrors wider patterns in British language attitudes research, in which urban varieties tend to be downgraded relative to rural varieties (e.g., Giles, 1970; Coupland et al., 1994; Coupland & Bishop, 2007). As discussed in the previous section, this rural/urban divide may be borne out of associations between the rural and an imagined, idyllic past, and between the urban and industrialization and undesirable social change (Dann, forthcoming). Relatedly, it could be argued that the recent rise to cultural prominence of the “Manc” variety (Montgomery, 2016), which is also associated with the central, urban area of Greater Manchester, and its connection to divisive characters such as the Gallagher brothers may have contributed to this more pejorative evaluation. Therefore, at least in the context of Greater Manchester, it appears that the use of either Broad or Strong to describe an “accented” variety may be a signifier of this differentiation, and a subtle evaluation of the variety’s social status.

7. Conclusions

This paper has presented a selection of results from the online draw-a-map task conducted as part of the Manchester Voices project. Methodologically, it has demonstrated the benefits of using an innovative, digital, web-based tool to collect this perceptual dialectology data, as it was easily shared and accessed, allowing us to collect large volumes of data unsupervised; and its novelty and simplicity also generated high levels of engagement. However, we did find that the reliance on participants self-selecting to take part resulted in a bias toward White British 26–45-year-olds. Therefore, future research using this method aiming to access a range of demographics may benefit from more targeted recruitment tactics.

The analysis explored the geospatial distribution of the three most frequent labels that emerged from the draw-a-map task relating to Accentedness (Broad, Strong, and Soft) and found that they were each clustering in a different region of Greater Manchester. We compared where, geographically, these descriptions clustered to other key categories (Status, Social Attractiveness, Rural, and Historical) in order to better understand the social evaluation of the varieties they were describing, providing insight into the social meaning of these terms. We suggest that, within a large and linguistically varied city-region such as Greater Manchester, these labels and their associated social meanings offer a particularly important way of enabling locally relevant distinctions to be made across multiple dimensions.

Preston (1996) discusses how, although non-linguists often lack the terminology and detailed knowledge to accurately describe linguistic variation, they are by no means “unaware.” The present paper demonstrates this further and adds to the evidence that folk-linguistic terminology should not be dismissed as an insufficient attempt to evoke specialist knowledge. Our approach here has been to treat the beliefs and attitudes expressed by our respondents as important for processes of language variation and change. We therefore responded to Preston’s (2016:195) calls for the use of “increasingly subtle techniques” to tease out the complexities of folk-linguistic awareness.

The insight provided by this analysis into the subtle nuances of folk-linguistic terminology in Greater Manchester was made possible by our adaption of the keywords methodology, first proposed by Garrett et al. (2005b). This approach enabled us to look beyond simply mapping the highest frequency accent/dialect labels, to explore how the region was divided by correlations of a variety of perceptual categories, including those that occurred relatively infrequently. We have also demonstrated the effectiveness of asking respondents to name, evaluate, and describe the variety as part of the map task, rather than simply asking them to add labels to the regions. This elicited rich and detailed data, providing us with a basis to explore the indexical relationships between different meaningful units for our respondents (i.e. between “dialect name” and “dialect evaluation”).

In this paper we have demonstrated one way that non-linguists characterize their awareness of linguistic variation, uncovering subtleties in the meaning of folk-linguistic terminology that could easily have been overlooked.