1.Introduction

Pragmatic markers come in a multitude of forms and are an integral part of spoken discourse as speakers use them to structure conversations, guide utterance interpretation, integrate other participants into a conversation, encode politeness, and express their own stance (Ajimer 2013, 4–8; Beeching 2016Beeching, Kate 2016 Pragmatic Markers in British English. Cambridge: Cambridge University Press. , 4; Brinton 2017Brinton, Laurel 2017 The Evolution of Pragmatic Markers in English. Cambridge: Cambridge University Press. , 2–8). In this paper, the term pragmatic marker is used as an umbrella term to describe markers that both structure discourse and express speaker stance. Despite their functional significance, they are commonly stigmatized (Schiffrin 1987Schiffrin, Deborah 1987 Discourse Markers. Cambridge: Cambridge University Press. , 45; Watts 1989Watts, Richard 1989 “Taking the Pitcher to the ‘Well’: Native Speakers’ Perception of Their Use of Discourse Markers in Conversation.” Journal of Pragmatics 13: 203–237. ). In addition, they are very difficult to classify in conventional categories of structural linguistics, as they are syntactically optional, only loosely connected to the syntax of an utterance, often short and phonologically reduced; they do not fit into any traditional word class, often do not change the truth-conditional meaning of an utterance, and instead fulfill a wide range of pragmatic functions. Pragmatic markers have been largely ignored in theoretical linguistics and many strands of applied linguistics have only paid more attention to them recently.

Increasing linguistic interest in pragmatic markers has led to fundamental work defining pragmatic markers (Schiffrin 1987Schiffrin, Deborah 1987 Discourse Markers. Cambridge: Cambridge University Press. ) – including discussions about terminology (see Pichler 2013Pichler, Heike 2013 The Structure of Discourse-Pragmatic Variation. Amsterdam: Benjamins. , 4–6; Beeching 2016Beeching, Kate 2016 Pragmatic Markers in British English. Cambridge: Cambridge University Press. , 3) – and a multitude of studies on the use and function of individual pragmatic markers. The advent of variational pragmatics (Schneider and Barron 2008Schneider, Klaus, and Anne Barron (eds) 2008 Variational Pragmatics. Amsterdam: Benjamins. ) has further pushed studies analyzing regional (e.g. Columbus 2010Columbus, Georgie 2010 “A Comparative Analysis of Invariant Tags in Three Varieties of English.” English World-Wide 31 (3): 288–310. ; Aijmer 2013Aijmer, Karin 2013 Understanding Pragmatic Markers in English. Edinburgh: Edinburgh University Press. , 127–147) and sociolinguistic (e.g. Anderson 2001Anderson, Gisle 2001 Pragmatic Markers and Sociolinguistic Variation. Amsterdam: John Benjamins. ; Beeching 2016Beeching, Kate 2016 Pragmatic Markers in British English. Cambridge: Cambridge University Press. ) variation of pragmatic markers. Many studies on pragmatic markers focus on individual forms, such as like (Anderson 2001Anderson, Gisle 2001 Pragmatic Markers and Sociolinguistic Variation. Amsterdam: John Benjamins. , 209–299; Beeching 2016Beeching, Kate 2016 Pragmatic Markers in British English. Cambridge: Cambridge University Press. , 126–154), to illustrate their functions and (sociolinguistic) variation in detail. The inclusion of alternative forms that may fulfill similar functions is often not possible due to the multitude of similar pragmatic markers and the problems of defining the envelope of variation for pragmatic phenomena (Pichler 2013Pichler, Heike 2013 The Structure of Discourse-Pragmatic Variation. Amsterdam: Benjamins. , 6–9). If sets of pragmatic markers are analyzed, such as general extenders (e.g. Aijmer 2013Aijmer, Karin 2013 Understanding Pragmatic Markers in English. Edinburgh: Edinburgh University Press. , 127–147), these studies are mostly based on predefined lists of forms assumed to function in the same way, analyze variation in forms in larger data sets but cannot investigate fine-grained differences in functions due to their a priori study design. Since pragmatic markers are most frequent in informal face-to-face conversations, most research has focused on this text type (e.g. Anderson 2001Anderson, Gisle 2001 Pragmatic Markers and Sociolinguistic Variation. Amsterdam: John Benjamins. ; Beeching 2016Beeching, Kate 2016 Pragmatic Markers in British English. Cambridge: Cambridge University Press. ). Studies that were able to utilize a wider range of spoken data have shown that text type is a decisive factor for the use of various pragmatic markers (e.g. Aijmer 2013Aijmer, Karin 2013 Understanding Pragmatic Markers in English. Edinburgh: Edinburgh University Press. ). The overwhelming amount of research on pragmatic markers in English is on British English (Anderson 2001Anderson, Gisle 2001 Pragmatic Markers and Sociolinguistic Variation. Amsterdam: John Benjamins. ; Aijmer 2013Aijmer, Karin 2013 Understanding Pragmatic Markers in English. Edinburgh: Edinburgh University Press. ; Beeching 2016Beeching, Kate 2016 Pragmatic Markers in British English. Cambridge: Cambridge University Press. ) and other countries where English is spoken as a native language, whereas there is comparatively less research on New Englishes, i.e. Englishes learned as a second/subsequent language or dialect, such as in anglophone Africa, Asia, or the Caribbean. Finally, there is an immense body of research on the use of pragmatic markers, but much less research on how speakers perceive them; exceptions include Watts (1989)Watts, Richard 1989 “Taking the Pitcher to the ‘Well’: Native Speakers’ Perception of Their Use of Discourse Markers in Conversation.” Journal of Pragmatics 13: 203–237. and Beeching (2016)Beeching, Kate 2016 Pragmatic Markers in British English. Cambridge: Cambridge University Press. .

This paper addresses these research gaps by analyzing the use and perception of question tags in Trinidadian English. I conceptualize question tags as a set of pragmatic markers that includes variant question tags, such as do I, and invariant ones, such as right. Earlier research has presented a general account of variation in question tag use in Trinidadian English (Wilson et al. 2017Wilson, Guyanne, Michael Westphal, Johanna Hartmann, and Dagmar Deuber 2017 “The Use of Question Tags in Different Text Types of Trinidadian English.” World Englishes 36 (4): 726–743. ), while I focus on specific question tag forms, presenting their usage and perceptual profiles. These forms include variant question tags and seven invariant forms: OK, right, you know as well as eh, ent, nah, and not so. These profiles are based on a variationist analysis of these forms in the Trinidad and Tobago component of the International Corpus of English (ICE-T&T) that considers general frequencies, pragmatic function, and text type variation. The perception study on the same eight forms (i.e. variant question tags and the seven invariant forms) is based on a questionnaire that combines a multiple-choice test, where participants were presented with different dialogue scenarios and selected the most appropriate form, and an indirect language attitude test, in which participants rated the use of the eight question tag forms. The paper addresses the following research questions:

• What are the usage profiles of variant question tags as well as OK, right, you know, eh, ent, nah, and not so?

• Which of these question tags do Trinidadians prefer in different situations?

• Which attitudes do Trinidadians hold toward the eight question tags forms?

The paper is structured as follows: Section 2 introduces question tags as a set of pragmatic markers, discusses previous research in this area, and gives a brief introduction to English and Creole in Trinidad. Section 3 presents the data and methods of both approaches. In Section 4, I present the usage profiles of the eight question tags, and Section 5 illustrates the perception study. Section 6 discusses the implications for future research on pragmatic markers.

2.Theoretical background

3.Data, participants, and methods

The usage profiles of variant question tags and the seven invariant forms OK, right, you know, eh, ent, nah, and not so are based on an analysis of all question tags identified in the ICE-T&T data. I did not use the entire one-million-word corpus but focused on 115 texts (each 2,000 words) from four dialogue text types: private dialogues (face-to-face conversations and phone calls), classroom lessons, broadcast dialogues (broadcast discussions and interviews), and legal cross-examinations. Table 1 provides an overview of this subcorpus, which has a size of 230,000 words.

These four text types represent different discourses that affect the use of question tags. Private dialogues are most informal, and speakers do not have fixed roles. All other text types are public dialogues, more formal, and speakers have specific roles. In classroom lessons, teachers dominate the dialogues, explain topics or tasks, and ask questions to the students. In broadcast dialogues, there is a host who moderates the exchange and asks questions to their guests, who discuss these questions. In legal cross-examinations, attorneys control the dialogue by asking questions to witnesses.

These data do not depict the full range of variation in Trinidad English/Creole but are biased toward the formal end of this continuum, because the ICE project aims to represent Standard English in countries where English has an official status. Standard English is defined via the speakers sampled in the corpus (Greenbaum 1996Greenbaum, Sidney 1996 “Introducing ICE.” In Comparing English Worldwide: The International Corpus of English, ed. by Sidney Greenbaum, 3–12. Oxford: Clarendon., 6): They are 18 years or older and have completed secondary education in English.

The usage profiles of variant question tags and the seven invariant forms OK, right, you know, eh, ent, nah, and not so are based on a variationist analysis of these eight forms in relation to all question tags used by Trinidadian speakers in the 115 dialogues. To include all forms, I read through all texts, identified forms functioning as question tags, and analyzed their function qualitatively. Besides being pragmatic markers, the decisive criteria for including forms into the analysis was that they fulfilled one of the three defining pragmatic functions: informative, facilitative, or punctuational.

This three-way distinction is a reduced system of the six functional categories by Tottie and Hoffmann (2006)Tottie, Gunnel, and Sebastian Hoffmann 2006 “Tag Questions in British and American English.” Journal of English Linguistics 34 (4): 283–311. , which was also used by Wilson et al. (2017)Wilson, Guyanne, Michael Westphal, Johanna Hartmann, and Dagmar Deuber 2017 “The Use of Question Tags in Different Text Types of Trinidadian English.” World Englishes 36 (4): 726–743. as well as Lange and Leuckert (2021)Lange, Claudia, and Sven Leuckert 2021 “Tag Questions and Gender in Indian English.” In Gender in World Englishes, ed. by Tobias Bernaisch, 69–93. Cambridge: Cambridge University Press.: aggressive, attitudinal, confirmatory, facilitating, informational, and peremptory. In these studies, confirmatory, facilitative, and attitudinal emerged as the three main categories: For example, in Tottie and Hoffmann (2006)Tottie, Gunnel, and Sebastian Hoffmann 2006 “Tag Questions in British and American English.” Journal of English Linguistics 34 (4): 283–311. , 93% of question tags fall into these three categories. A clear distinction between aggressive, attitudinal, and peremptory tags proved impossible in the current analysis because all three functions are emphatic, speaker-centered, and express stance; aggressive and peremptory tags express very specific stances and could be classified as subcategories of attitudinal tags. Hence, these three functions were summarized to the punctuational category. Similarly, it was mostly not possible to distinguish between informative and confirmatory uses: Both functions are interactional, addressee-centered, and require a response but can only be distinguished based on the level of speakers’ pre-knowledge. These two categories were summarized to informative functions. This classification system works on a high level of abstraction and is not useful for detailed qualitative analyses of the functions of question tags but proved very productive for quantitative analyses (Westphal 2020Westphal, Michael 2020 “Question Tags in Philippine English.” Corpus Pragmatics 4 (1): 401–422. , 2022Westphal, Michael 2022 “The Multilingual Pragmatics of New Englishes: An Analysis of Question Tags in Nigerian English.” Frontiers in Communication 6: 1–16. ).

Along my operationalization of function, speakers use question tags in an informative way to ask for a confirmation or seek new information. In (7), a talk show host is unsure about a date of an event and asks a guest for confirmation.

Facilitative question tags are used to integrate interlocutors into the conversation, either signaling a transition relevance place or checking whether the interlocutors are following along, which they can show by backchanneling or asking for clarification. In (8), a talk show host signals that he wants to hand over the turn to his guest using you know.

Punctuational tags are used stylistically, mostly to add emphasis, and no response is required. In (9), a speaker uses nah to add emphasis to his suggestion to leave.

In cases when it was impossible to clearly differentiate between two functions the question tag was coded for both functions.

In Section 4, I first give a brief overview of the frequencies of the most frequent forms identified as question tags in the data and then discuss each of the eight question tags regarding their general frequencies, the constraints of function and text type on their use, and I also provide examples of typical patterns of their use. The constraints of function and text type are analyzed via binary regression modelling: form is the dependent variable and is reduced to a binary distinction (e.g. right vs. all other forms); function (informative vs. facilitative vs. punctuational) and text type (private dialogues vs. classroom lessons vs. legal cross-examinations) are inserted as fixed predictor variables; speaker is used as a random factor to avoid Type I errors due to idiosyncratic variation (Johnson 2009Johnson, Daniel E. 2009 “Getting off the GoldVarb Standard: Introducing Rbrul for Mixed-Effects Variable Rule Analysis.” Language and Linguistics Compass 3 (1): 359–383. ). This approach describes the usage profiles of the eight forms in relation to all question tags and is capable to show differences and similarities between them. Tokens that were coded for two functions were removed from the analysis.

The perception of question tags was analyzed via a questionnaire that comprises two parts. The first part investigates which question tags the participants prefer in different situations. It is modelled after a Discourse Completion Task but employs fixed answer options. There is a brief description of a dialogic scenario followed by an incomplete utterance to which the participants add a question tag from a given list. The options include the respective variant question tag, invariant isn’t it, OK, right, you know, eh, ent, nah, and not so, as well as the option ‘other’, which allows participants to enter free text. There are seven different scenarios which reflect different communicative settings and target a specific pragmatic function:

• Scenario 1 depicts a private dialogue and targets an informative question tag.

• Scenario 2 is a classroom lesson, and the teacher uses a question tag facilitatively to check whether students have any questions.

• In scenario 3, participants imagine they are an attorney eliciting information from a witness in a legal across-examination using an informative question tag.

• Scenario 4 is a private dialogue and requires the use of a punctuational question tag.

• In scenario 5, participants assume the role of a teacher adding a punctuational question tag to an imperative targeted at a student.

• Scenario 6 is a private dialogue, which targets the use of a facilitative question tag.

• In scenario 7, participants assume the role of a student at university, asking their lecturer for information using a question tag.

The second part of the survey is a written Matched Guise Test adapted from Beeching (2016Beeching, Kate 2016 Pragmatic Markers in British English. Cambridge: Cambridge University Press. , 38–41), which investigates the participants’ language attitudes toward the eight question tags, as well as invariant isn’t it, which is not discussed in this paper. For each question tag there is a specific scenario which reflects its typical pattern of use. Each scenario includes a description of the context and two almost identical utterances produced by two different speakers (Speaker A and B). Speaker A does not use a question tag, and Speaker B does, which is highlighted in bold. The participants are asked to rate Speaker B in contrast to Speaker A on six personality traits, using six-point semantically differential scales (impolite vs. polite; reserved vs. outgoing; aggressive vs. gentle; indirect vs. direct; unfriendly vs. friendly; uneducated vs. educated). For each scenario participants could also add additional thoughts on the question tag. See Figure 1 for an example.

The results for the multiple-choice task are first described descriptively, and for forms which were selected across many different scenarios binary regression analyses were conducted: form (reduced to a binary distinction; e.g. right vs. all other forms) is the dependent variable; scenario (seven different levels) is the fixed predictor variable. For the Matched Guise Test, the mean scores of the ratings on the six different items are presented for all eight question tags. A Principal Component Analysis (with Varimax Rotation) was conducted to identify clusters of items, which are taken to represent the larger attitudinal dimensions that govern the participants’ answers. I then calculated mean values from the items that clustered together and used them in a regression analysis. These rating scores were employed as dependent variables and question tag was used as fixed predictor variable.

The perception study is based on 166 questionnaires completed by students at the University of the West Indies, St. Augustine Trinidad. Fieldwork was carried out by Ryan Durgasingh in February 2020. The students are from diverse study programs but there is a large group (fifty-eight; 34.9%) who indicated that they study English Literature and Language with Education or Linguistics and thus have advanced linguistic knowledge. In terms of age, gender, and ethnicity, 140 (76.5%) participants are between 18 and 25 years old, thirty-four (20.5%) are between 26 and 45, and only five (3.0%) are older than 46; 140 (84.3) participants are female, twenty-four (14.5%) male, and two (1.2%) are of diverse gender; sixty-two (37.7%) students self-identified as Afro-, forty-one (24.7%) as Indo-, sixty-two (37.3%) as Mixed-Trinidadian/Tobagonian, and one (0.6%) student indicated a different ethnic identity. This sample is biased toward young and female participants, but this fits the general student demography at St. Augustine.22. https://​sta​.uwi​.edu​/copir​/sites​/default​/files​/copir​/documents​/Quick%20Facts%202019​-2020​-Updated​.pdf In addition, university students are a specific demographic of Trinidadian society in terms of age and level of education: They are young (but over 18) and well-educated, having completed secondary education. They are like the speakers sampled in ICE-T&T, which improves the comparison of language use and perception, and similar biases apply to both samples.

4.Usage profiles of question tags in Trinidadian English

With 1,316 occurrences (302.7 per one million words), question tags are a highly frequent feature in the dialogues of ICE-T&T. Table 2 provides an overview of the frequencies of question tag forms. Right is the most frequent form followed by you know, alright, which is also included in the description of usage profiles, and OK. Trinidadians also use yeah and yes as question tags but do so infrequently. Variant question tags are shown to play a marginal role in the data. Trinidad English/Creole forms are varyingly employed by the speakers in the ICE-T&T dialogues: eh is most frequent followed by nah and not so, while ent is marginal.

Figures 2 and 3 show the direction and effect size of the different levels of text type and function on variant question tags as well as alright, OK, right, you know, eh, ent, nah, and not so as (centered) factor weights. Factor weights for the individual levels are only reported if text type or function have a significant effect on the selection of the specific form (see Table 3; Figures 2 and 3). For ent no regression analysis was run due to the low token count. Factor weights range from 0 to 1: Values above 0.5 indicate a preference for the specific question tag; values below 0.5 a dispreference. To facilitate the interpretation of the usage profiles lines between factor weights are included in the figures.

text type only had a significant effect on the use of OK, you know, and eh (Figure 2). For the five other question tags no significant effect was found; however, it has to be kept in mind that inserting speaker as a random factor makes the analysis very conservative. This effect is more pronounced for infrequent forms. For example, not so is especially frequent in legal cross-examinations (i.e. twenty out of thirty not so tokens occur in this text type) and nah in private dialogues (i.e. forty out of fifty-one nah tokens are in this text type), but these patterns of variation do not reach the level of significance as there are few speakers who use these tags frequently and hence these text type effects interact with idiosyncratic preferences. The analysis is further complicated by text types with a relatively small number of texts, such as legal cross-examinations.

The analysis shows that OK, you know, and eh have very different distributions across the four text types, as the lines intersect frequently. Eh is preferred in informal private dialogues and classroom lessons but dispreferred in the more formal text types, broadcast dialogues and legal cross-examinations. You know is preferred in private dialogues and strongly preferred in broadcast discussions, while there is a strong dispreference in classroom lessons. For OK there is a categorical absence in broadcast dialogues. Hence, this text type was removed from the regression model and the factor weight is represented as zero in Figure 2. The model shows the strongest preference of OK for legal cross-examinations but only a slight one for classroom lessons. However, these results need to be treated with some caution as there are only eighty-five question tags overall in legal cross-examinations and OK is used six times by five different speakers but still is not particularly characteristic for this text type. Instead, OK (eighty-seven out of 539) is the second most frequent question tag after alright (180 out of 539) in classroom lessons, but due to the high overall frequency of question tags in this text type and as one teacher uses OK thirty-three times, the statistical preference in this text type is not very pronounced.

Figure 3 shows the (variationist) functional profiles of the eight different question tags, as function has a significant effect on the use of all of them. Intersecting lines indicate that the respective forms have distinct profiles, while lines that run parallel point toward similar functional profiles. High factor weight differences or steeply ascending/descending lines suggest that the specific forms have very focused functional profiles, while flatter lines indicate less focused profiles. Variant question tags as well as invariant not so and right have very similar functional profiles, which is least focused for right: There is a strong preference for informative uses, while punctuational ones are dispreferred. Alright and OK are used similarly with a preference for facilitative uses, but the functional profile of OK is less focused. Nah, eh, and you know show a strong preference for punctuational functions and, except for eh, a strong dispreference for informative ones. You know is less focused on punctuational uses and demonstrates a preference for facilitative functions.

With only twenty occurrences variant question tags are generally marginal and not characteristic for Trinidadian English. They have a very focused functional profile because speakers almost exclusively use them for informative functions across all text types, such as the teacher in (10) asking her students to interpret a passage in a poem or the attorney in (11) using were you in a leading question. Although text type does not have a significant effect on the use of variant question tags due to their overall low frequency, there is a tendency toward the use of variant question tags in the more formal classroom lessons with English as a subject and in legal cross-examinations.

Like variant question tags, not so has a focused profile for informative uses, a somewhat formal character, and is rather infrequent in the data. Across all text types, it is predominantly used to receive information from the interlocutors and an answer is always given, for example, in an interview-style conversation (12). Although there is no significant effect of text type, not so is particularly frequent in legal cross-examinations (twenty out of the thirty question tags in this text type are not so), where attorneys use not so in leading questions, as in (13). In addition, Wilson et al. (2017Wilson, Guyanne, Michael Westphal, Johanna Hartmann, and Dagmar Deuber 2017 “The Use of Question Tags in Different Text Types of Trinidadian English.” World Englishes 36 (4): 726–743. , 735) state that not so is “promoted by parents and educators as a more acceptable alternative to the tag ent”.

Right has a similar functional profile to variant question tags and not so. However, right is the most frequent question tag in the ICE-T&T dialogues, it is used across all text types, and it is functionally more flexible. Right may be used punctuationally but informative uses, as in (14), are most frequent. In contrast to not so and variant question tags, speakers use right frequently as a facilitative tag in longer narrative sections to quasi integrate interlocutors into monologic passages (15), or teachers regularly insert right into explanatory sections to check whether students are following along (16).

OK is relatively frequent in the corpus and is mainly used for facilitative functions. However, it has the least clearly defined functional profile and is used in a variety of different situations. The statistical preference for legal cross-examinations is somewhat misleading as OK is particularly characteristic for classroom lessons together with alright. Teachers and lecturers use both tags in longer explanatory or narrative sections to keep the students integrated into the classroom discourse and often combine OK and alright with a pause to leave time for backchanneling or questions, as in (17). In this way, OK is a structuring device for classroom discourse, and it is used in a very similar fashion by speakers in conversations in monologic passages, as in (18), where a speaker soliloquizes about violence in Trinidad.

You know is the second most frequent question tag in the Trinidadian dialogue data. Despite its high frequency, you know is nearly never used in an informative way. Facilitative and punctuational uses are almost evenly preferred: Speakers use you know frequently to establish common knowledge or checking for agreement, in this way integrating the interlocutors into the conversation when continuing with their turn (15) or signaling a transition relevance place with you know. The latter facilitative aspect is rather common in phone calls where there are no visual cues to end turns. In (19), the speaker wants to end her long monologic turn several times signaling a transition relevance place with you know combined with a pause.

Speakers also use you know frequently to add emphasis to their arguments in discussions – without leaving a pause. This punctuational use is especially frequent in broadcast dialogues among guests who make an argument, as in (20). In contrast to the punctuational tags eh and nah, you know is not associated with Creole and hence speakers employ you know in these public discussions, while opting for other tags in more private informal situations.

Eh is the most frequent Trinidad English/Creole tag in the data and is mainly used in informal private dialogues between friends. It is sometimes used by teachers in more informal classroom lessons in school – not in university lectures, which are also part of this text type. Eh may be used for all functions, but there is a clear preference for punctuational uses, as in (21), where a teacher uses eh for extra emphasis when warning a student to not come again without homework. In addition, eh is often used when imitating informal speech, as in (22).

Hence, eh signals anti-formality and carries extra-metaphorical meaning besides emphasis in contrast to you know, for example signaling antagonism (21). Hosts and guests do not use eh in formal and polite broadcast discussions, but attorneys use eh – although rarely – as an anti-formal power move to pressure witnesses, as in (23).

Nah has a very similar usage profile to eh, also signaling anti-formality. However, nah seems to be more strongly marked for Creole as speakers rarely use it – and when they do it sticks out. Nah is almost exclusively used in private dialogues between friends and is often combined with colloquial language or other Creole features, as in (24), which contains copula absence. In (25), a teacher imitates a student speaking to a salesclerk and uses nah in his stylization of informal haggling. Similar to eh, attorneys also use nah as a marked power move, as in (26).

Ent is generally avoided by speakers in ICE-T&T and seems the most marked for Creole and anti-formality. From the three instances there is no clear functional pattern. In (27), a teacher uses ent facilitatively to check whether the students understand her explanations, and in (28), ent is used informatively in a private conversation.

5.The perception of question tags

Both the multiple-choice test, which illustrates the participants’ preferences for specific forms in different situations, and the Matched Guise Test, which demonstrates the Trinidadians’ attitudes toward variant question tags and the seven invariant forms OK, right, you know, eh, ent, nah, and not so, show a diverse perceptual picture, which partly overlaps with the usage profiles but also contrasts them.

The multiple-choice test shows that the students preferred different question tags in different scenarios (Figure 4). Across all scenarios, right was selected most frequently (247 times out of 1158; 21.3%), followed by variant question tags (16.7%), ent (15.7%), and not so (10.4%). These four question tags were also selected across all scenarios, which allows further regression analyses. In contrast, OK, you know, eh, and nah, were chosen only in specific scenarios and illustrate rather categorical variation. The participants mainly selected OK in scenarios depicting classroom lessons (scenarios 2 and 5), which confirms OK as a teacher tag. Very similar to the usage profile, you know was exclusively chosen for facilitative and punctuational uses in conversations (scenarios 4 and 6) but not for informative uses in a similar context. Eh was only chosen in larger frequencies in the context of a private dialogue with a punctuational function (scenario 4), and nah almost exclusively for punctuational use in a classroom lesson (scenario 5). In this scenario, the question tag emphasizes an imperative targeted at a student and carries an additional aggressive tone, which the participants felt is best expressed by nah and OK. In general, there is a tendency toward the selection of the Trinidad English/Creole tags eh, ent, and nah in more informal private conversations and the classroom lessons where an additional metaphorical meaning is targeted.

Right, variant question tags, ent, and not so were selected across most scenarios and were thus perceived to be more flexible. The patterns of variation are shown via regression analyses (Tables 4 and 5) with question tag (e.g. right vs. all others) as dependent variable and scenario (seven levels) as predictor variable. scenario has a significant effect on the selection of all four question tags in different ways. There is a preference for right for all informative uses and for a facilitative use in a classroom setting (scenario 2). For variant question tags, the strongest preference is found for the legal cross-examination, followed by a punctuational use in a conversation; there is a marked dispreference for facilitative uses.

Ent is preferred in two informal conversations targeting an informative and a punctuational function and surprisingly also in the more formal conversation between a student and a lecturer targeting an informative function (Table 5). In contrast, there is a strong dispreference for the public dialogue situations in classroom lessons and the legal cross-examination. This finding supports the descriptive observation that ent – as well as eh and nah – are tendentially avoided in more formal contexts. This is not the case for not so, which is strongly preferred in the legal cross-examination and scenario 2 depicting a classroom lesson but is dispreferred in the more personal scenarios. This supports the more formal character of not so.

The results of the Matched Guise Test show that the participants rated variant question tags and the seven invariant forms OK, right, you know, eh, ent, nah, and not so very differently (Figure 5). All ratings – especially outgoing, direct, and friendly – are biased toward positive values (i.e. 3.5 and higher). This may mean that speakers using question tags – irrespective of form – were generally perceived as more outgoing, direct, and friendly. However, these ratings may also be affected by an acquiescence bias (Garrett 2010Garrett, Peter 2010 Attitudes to Language. Cambridge: Cambridge University Press. , 45), i.e. the tendency among participants to agree to an item, and a social desirability bias (Garrett 2010Garrett, Peter 2010 Attitudes to Language. Cambridge: Cambridge University Press. , 44–45), i.e. to give socially appropriate answers, which in this case means choosing positive traits for speakers using question tags. The problem of a social desirability bias is further supported by many comments which stated that level of education does not play a role for the use of the different question tags, as in (29).

Variant question tags were generally rated very positively and not so more negatively, but due to the complexity of variation the results are difficult to interpret. To reduce this complexity a Principal Component Analysis was conducted, which demonstrated that the six items cluster together to two larger dimensions. Polite, gentle, friendly, and educated form one cluster, which was termed decency, has an Eigenvalue of 2.67, and explains 44.4% of the variance. Direct and outgoing cluster together to a second dimension, which is less decisive for variation in the data; this dimension was labelled expressiveness, has an Eigenvalue of 1.12, and explains 19.7% of the variance. Table 6 gives an overview of the factor loadings of the six items on the two dimensions; factor loadings below 0.3 are not shown. Ratings for decency and expressiveness (i.e. mean values of the ratings of the items clustering together to these two dimensions) were used for two regression analyses, with decency/expressiveness as the dependent variable and question tag as the predictor variable.

The analysis shows that the eight question tags were rated significantly differently both in terms of decency and expressiveness (Table 7). The range of variation is more pronounced for decency (∆ = 1.23) than for expressiveness (∆ = 0.33) and the modelling is much better. Variant question tags were rated most positively in terms of decency, followed by right, OK, and you know. In contrast, all Trinidad English/Creole tags were downrated in terms of decency, and some individual comments also stated that eh, ent, and nah are “too informal” or “harsh”. However, regarding expressiveness, nah and eh were rated most favorably, followed by OK and variant question tags. Not so was again downrated and was hence generally perceived the most negatively. The appreciation of Trinidad English/Creole tags, most notably ent, was also evident in many comments where participants praised these sentences as “truly Trini”, as in (30). However, participants made a distinction between formal not so and more informal Trinidad English/Creole tags, as in comment (31) on the use of not so.

Although the survey study worked well in the sense of showing meaningful patterns of variation for the preferences of and attitudes toward different question tags, authenticity of the examples turned out to be a problem. Though the exemplary sentences were based on corpus data, and I used typical Trinidadian personal names, such as Janielle, Sherry-Ann, or Ryan, and place names, such as Marabella, many participants complained that the exemplary sentences do not represent real Trinidad English Creole, as in (32). As much linguistic research in Trinidad is on Creole, the students seemed to be primed toward evaluating Creole and not English. Comments also displayed a strong pride of Creole and the Creole pragmatic markers ent and nah among many participants.

Another problem is that the ratings of the question tags are influenced by the general scenario, although the structure of the question tries to overcome this problem: Participants were asked to rate Speaker B in contrast to Speaker A, who both speak in the same context. This problem seems to have influenced the negative ratings of not so, as the attorney in the example who uses not so comes across as rather aggressive, as highlighted in (33).

6.Similarities and differences between use and perception of question tags

The analysis of the use and perception of variant question tags and the seven invariant forms OK, right, you know, eh, ent, nah, and not so shows that these two dimensions of linguistic variation often overlap but there are also marked differences. These incongruencies are most pronounced for variant question tags and the Trinidad English/Creole eh, ent, and nah. Variant question tags play a marginal role on the level of language use in Trinidadian English, similar to previous research on New Englishes (Westphal 2020Westphal, Michael 2020 “Question Tags in Philippine English.” Corpus Pragmatics 4 (1): 401–422. , 2022Westphal, Michael 2022 “The Multilingual Pragmatics of New Englishes: An Analysis of Question Tags in Nigerian English.” Frontiers in Communication 6: 1–16. ). Their use is restricted to informative functions, mostly in formal contexts. This usage profile is confirmed by the perception study, but participants showed a heightened preference for variant question tags and valued them highly in terms of decency. This attitudinal preference contradicts the marginal role in use but is not as pronounced as for Nigerian participants (Westphal 2022Westphal, Michael 2022 “The Multilingual Pragmatics of New Englishes: An Analysis of Question Tags in Nigerian English.” Frontiers in Communication 6: 1–16. ).

Ent, eh, and nah emerge as characteristic of Trinidad English Creole because they are restricted to informal contexts, similar to overt morpho-syntactic (Deuber 2014Deuber, Dagmar 2014 English in the Caribbean. Cambridge: Cambridge University Press. ) Creole structures. Ent plays a marginal role in use, while eh and nah are more frequent but their use is restricted to informal situations and metaphorical uses in formal contexts. All three tags have ideological value in terms of expressiveness. Although the participants expressed their appreciation for these three Creole tags, they also exhibited a high degree of register awareness: i.e. when the use of ent, eh, and nah is appropriate and when it is not. Overt pride in these Creole forms seems to interact with some degree of a more covert stigmatization. On the one hand, these findings point toward an increasing nativization of Trinidadian English since Creole forms are partially integrated into acrolectal usage and, on the other hand, toward a growing acceptance of local forms alongside continuing “residual conservatism” (Schneider 2007Schneider, Edgar 2007 Postcolonial English. Cambridge: Cambridge University Press. , 56). This conclusion corroborates Wilson’s (2014)Wilson, Guyanne 2014 The Sociolinguistics of Singing: Dialect and Style in Classical Choral Singing in Trinidad. Münster: Monsenstein und Vannerdat. description of Trinidad as striving toward endonormative stabilization but still being influenced by outside norms.

The analysis of use and perception illustrate that not so is an informative question tag, mainly used in formal situations. It is not associated with Creole and perceived generally rather negatively but also as a formal alternative to the Creole tag ent. Hence, not so emerges as a local Trinidadian tag that is mainly characteristic of the standard end of the Trinidad English/Creole continuum. The results for variant question tags and the Trinidad English/Creole forms illustrate that pragmatic markers may be insightful for variation in Creole-using communities (e.g. Winford 1997Winford, Donald 1997 “Re-Examining Caribbean English Creole Continua.” World Englishes 16 (2): 233–279. ) but have been underresearched in these contexts.

The usage and perceptual profiles of right, OK and you know match very well. Right is the most frequent Trinidadian question tag, used for a wide range of functions, is generally preferred in the multiple-choice test, and received high attitudinal ratings. Right emerges as the all-purpose Trinidadian question tag, with a preference for informative uses dominating over the more formal alternatives not so and variant question tags. Both approaches confirm OK as a teacher tag, which is generally used to integrate others into the conversation. You know is shown to be a frequent and flexible question tag but not for informative uses. The preference of you know for informal conversations is evident on both levels of linguistic variation but is more pronounced in the perception study. These three forms can be termed global invariant question tags and these profiles are very similar to other varieties of English (Westphal 2020Westphal, Michael 2020 “Question Tags in Philippine English.” Corpus Pragmatics 4 (1): 401–422. , 2022Westphal, Michael 2022 “The Multilingual Pragmatics of New Englishes: An Analysis of Question Tags in Nigerian English.” Frontiers in Communication 6: 1–16. ), albeit general frequencies vary: For example, right is substantially more frequent in Trinidadian English than in other varieties (see also Columbus 2010Columbus, Georgie 2010 “A Comparative Analysis of Invariant Tags in Three Varieties of English.” English World-Wide 31 (3): 288–310. ).

Instead of focusing on individual forms in isolation, I have illustrated a variationist approach to describing the usage profiles of question tags, conceptualized as a set of pragmatic markers. This approach has shown in how far individual forms overlap and diverge in their text type distribution and form-function relationship, which is often a matter of frequencies (Aijmer 2013Aijmer, Karin 2013 Understanding Pragmatic Markers in English. Edinburgh: Edinburgh University Press. , 148). As individual forms have their own distinct usage profiles, it is desirable to not combine all invariant question tags together but analyze them individually. Functionally similar forms, such as not so and right, are not identical but exhibit fine-grained differences in terms of contextual use or functional flexibility. The results have shown that it is essential to not restrict analyses of question tags to variant ones – especially in New Englishes contexts – and even studies that focus on variant forms should also investigate them in relation to ‘competing’ invariant tags, such as right, as also suggested by Tottie and Hoffmann (2006Tottie, Gunnel, and Sebastian Hoffmann 2006 “Tag Questions in British and American English.” Journal of English Linguistics 34 (4): 283–311. , 307).

The participants’ perceptions seem to be governed by three larger ideological patterns. First, students’ preferences and attitudes are guided by a Standard Language Ideology that is biased toward English canonical/variant question tags and devalues Creole forms. In contrast, the students exhibit pride in Trinidad English/Creole tags. Third, they exhibit a strong register awareness regarding when a tag is appropriate and when it is not: For example, not so is only accepted in formal contexts and ent more in informal ones. Hence, this register awareness interacts with the Standard Language Ideology and their pride in Creole since the level of formality is a decisive factor. However, register awareness is not restricted to a mere formal-informal dichotomy but influences the perception of right, OK, and you know, which are perceptually more neutral, in terms of speakers’ communicative needs.

Establishing a direct link from the participants’ attitudes to language use is rather difficult but the perceptual data provides an addition explanatory framework to understand some of the patterns of variation in language use. A Standard Language Ideology interacts with speakers’ register awareness and restricts the use of marked Creole forms to informal contexts or anti-formal metaphorical uses. Stereotypical pragmatic markers, such as ent, seem to be most affected by speakers’ language attitudes but most question tags – especially the most frequent forms right, OK, and you know – seem to operate below speakers’ level of awareness. Despite the high ideological value of variant question tags, speakers hardly use them in actual dialogues, which seems to be more guided by speakers’ immediate communicative needs in particular contexts than their ideological dispositions. These ideological dispositions nevertheless have some effect, as shown for the tendency to use not so and variant question tags more in formal situations. This conclusion means that it is imperative for future research on pragmatic markers to consider text type – or speakers’ roles in particular contexts – more earnestly beyond a formal-informal dichotomy (see also Aijmer 2013Aijmer, Karin 2013 Understanding Pragmatic Markers in English. Edinburgh: Edinburgh University Press. , 13–16, 125; Gómez González 2018Gómez González, María de los Ángeles 2018 “Variable and Invariable Tag Questions in Spoken British English.” In The Construction of Discourse as Verbal Interaction, ed. by María de los Ángeles Gómez González, and J. Lachlan Mackenzie, 109–144. Amsterdam: Benjamins. , 138).

The mix of two methods in the survey study, i.e. the multiple-choice test and the Matched Guise Test, is capable of illustrating the fine-grained perceptual differences between the eight forms. Although the study design is quite complicated, both approaches work well and produce meaningful patterns of variation, which substantiate the usage profiles. Ideally, the survey study is combined with metalinguistic interviews, to discuss the indexicality of individual forms in more detail, for example the associations with Creole, metaphorical uses like expressing aggressiveness, but also social characteristics, such as masculinity or femininity. These metalinguistic interviews may also show aspects not anticipated in the study design and may shed light on certain shortcomings, such as issues of authenticity of samples. If researchers are interested in studying sociolinguistic variation of pragmatic markers, including the speakers’ perspective seems highly valuable. Pragmatic markers might also be best suited for such an approach as speakers are aware of their different functions and have complex associations with them. Studying the perception of pragmatic markers in addition to their use may be insightful for single forms, such as like (Anderson 2001Anderson, Gisle 2001 Pragmatic Markers and Sociolinguistic Variation. Amsterdam: John Benjamins. , 209–299; Beeching 2016Beeching, Kate 2016 Pragmatic Markers in British English. Cambridge: Cambridge University Press. , 126–154), to provide a very detailed account of such individual forms. However, this approach also seems very promising for analyses of sets of pragmatic markers – as was done in this study – to provide an additional explanatory framework for the linguistic dynamics of the relationship between different pragmatic markers.