A register approach to negative concord versus negative polarity items in English
-
Stephanie Rotter
and
Mingya Liu
Abstract
Negative concord (NC) is used in many English varieties but usually considered ungrammatical in ‘standard’ contemporary English, where negative polarity items (NPIs) are used. In this paper, we take a novel experimental approach to the use of NC versus NPI constructions in relation to register, i. e., sets of speech repertoires linked to specific situational and functional parameters. We report on two rating experiments with American and British English participants using interlocutor relations as a formality manipulation. Results show that (i) across both samples, NC constructions were rated as less appropriate than NPI associates, and (ii) there was a register effect in the American English data in that NC was rated less appropriate in formal than informal contexts. Our study is the first to provide experimental evidence for the register-sensitivity of NC constructions in American English.
1 Introduction
Negation is a universal feature of human language and communication, and central to human cognition. Across languages, varying lexical and grammatical devices are used to express negation. In English, a negative meaning can be expressed by at least the following three constructions: sentence negation not either with a bare indefinite (BI) as in (1), with a negative polarity item (NPI) as any in (2), or with a negative concord item (NCI) as no in (3).
| Mary didn’t see apples. |
| Mary didn’t see any apples. |
| Mary didn’t see no apples. |
The focus of the paper is the phenomenon of negative concord (NC), which describes cases in which two (or multiple) negative expressions co-occur with the semantic meaning of solely one (Blanchette 2017; Giannakidou 2007; Horn 2009; Tubau 2016; Zeijlstra 2004). More specifically, we focus on NC constructions in which a verbal negator (n’t) is accompanied by the NCI (no) resulting in a single negation (SN) interpretation, as illustrated in (3). In contrast, the co-occurrence of two negations can also trigger a double negative (DN) interpretation in that each negative marker contributes a semantic negation, which results in an affirmative statement by canceling each other out (Horn 2010; Penka 2020). While multiple co-occurring negations trigger the SN interpretation in some non-standard varieties of English, i.e., they form NC constructions, standard English is claimed to only access DN interpretations (Penka 2020; Thornton et al. 2016).
Relatedly, NPIs such as any and ever are items that require licensing through some kind of negative context (Blanchette 2015; Krifka 1995; Penka and Zeijlstra 2010). NPIs, unlike NCIs, are not semantically negative themselves (see Breitbarth et al. 2020: 14–15; van der Auwera 2022 for discussion). Furthermore, NPI constructions are present in both standard and non-standard[1] varieties of English. Moreover, NPIs often function as the standard English translation of NC constructions (Blanchette 2015; Nevalainen 2006). This points towards the functional equivalence of the negation types: Semantically, NC and NPI (as well as BI) constructions in co-occurrence with a verbal negator express SN (Childs 2017). Thus, the presented forms constitute variants of a linguistic variable with the same functionality (Labov 1972c; Tagliamonte 2006). This makes them particularly suitable for studies on linguistic variation and register, which the paper aims to address. Here, we use the notion of register to refer to patterns of intra-individual variation influenced by situational-functional settings (Biber 2012; Biber and Conrad 2019; Lüdeling et al. 2022; Pescuma et al. 2023). For instance, the same person might use the word “dinner” when talking to their boss and “supper” when talking to their spouse. While referring to the same event, i.e., eating food in the evening, the words differ with respect to their social connotations and formality level (cf. Rotter and Liu 2023).
NC and NPIs have received much attention in various linguistic sub-fields such as syntax, semantics, pragmatics, and sociolinguistics (cf. Childs 2017; Giannakidou 2000; Horn 2009; Moore 2021; Smith 2001; Zeijlstra 2004 among others). For instance, sociolinguistic studies on the usage of NC in different populations of Detroit argue for the social stratification and stylistic uses of NC structures: While NC use stratifies along socioeconomic status in African American Vernacular English-speaking preadolescents and adults, adolescents did not show this pattern (Eckert 2019; Wolfram 1996). A study on high school students revealed the stylistic use of NC related to adolescent social identity in the school culture: Jocks (i.e., school-oriented adolescents constituting middle class culture) used NC significantly less than burnouts (i.e., school-alienated adolescents) (Eckert 1989: 68, 2019). Similar findings were reported for a high school in Bolton, England: Here, the social group called townies which is associated with rebellious and anti-school behavior – similarly to the burnouts – were found to use NC more frequently than the more pro-school oriented Geeks or Eden Villages (Moore 2021). A study on Buckie, a Scottish English dialect, revealed that the speakers use NC with outgroup interlocutors less frequently than with ingroup interlocutors, an effect independent of gender (Smith and Holmes-Elliott 2022). Thus, the variation of NC usage is manifold and does not only reflect stratification but also the dependency on situational parameters, for instance, ingroup versus outgroup dynamics, or social meanings (Beltrama 2020; Hall-Lew et al. 2021).
Yet, not much is known on the variation of negation types in different situational contexts and their pragmatic effects across different varieties of English. Furthermore, the study of negation and polarity variants can enrich the knowledge on the recruitment of alternations for register purposes and the choice of language users among variants.
In this paper, we take an experimental approach to the potential register effects of negation and polarity guided by two questions: First, are NC and NPI phenomena register-sensitive (i.e., sensitive to formal vs. informal contexts)? Second, is the interpretation of NCI (and NPI) constructions register-sensitive (i.e., sensitive to formal vs. informal contexts)? In order to approach these questions, we conducted two rating experiments in American and British English in which participants judged the usage and interpretation of NCI, NPI, and BI constructions. To access a formal versus informal register, we used labels of different social relations of the interlocutors, such as boss versus sister.
The paper is structured as follows: Section 2 presents the relevant theoretical and experimental background on the negation and polarity phenomena as well as on registers. Section 3 deals with the methodological approach we employed in the rating studies in American and British English. Section 4 represents a combined discussion of the results, and Section 5 concludes the paper.
2 Background
2.1 Multiple negations and their interpretation in varieties of English
Syntactic approaches to multiple negations differentiate between NC and DN languages. Languages of the former type typically employ multiple negative markers in a syntactic dependency resulting in the SN interpretation (Zeijlstra 2004), such as in Romance languages and Hungarian (Puskás 2012). DN languages, in contrast, do not show such concord relations and each negative element contributes a semantic negation (de Swart 2020). Typical examples include Dutch and German, but also English has been treated as a DN language (de Swart 2020; Giannakidou 2020; Zeijlstra 2004: 75). However, some researchers take this classification to be based on the prescriptive perception that NC constructions are ungrammatical in standard English (Blanchette and Lukyanenko 2019a; Childs 2017; Thornton et al. 2016). Multiple negations with a SN interpretation appear ubiquitously in non-standard varieties of English across the globe (Childs 2017; Childs et al. 2015; Horn 2009; Labov 1972b; Smith 2001; Szmrecsanyi and Kortmann 2004; Tubau 2016) – see (4) for examples from American English varieties and (5) from British English varieties.
| He didn’t see nothing . |
| ‘He didn’t see anything.’ |
| (African American vernacular English [Pullum 1999: 48]) |
| There wasn’t nothing for them to get into. |
| ‘There wasn’t anything for them to get into.’ |
| (Appalachian English [Blanchette and Lukyanenko 2019a: 2]). |
| She didna tak nae money fae us. |
| ‘She didn’t take any money from us.’ |
| (Buckie dialect [Smith 2001: 110]) |
| You didn’t have nobody to learn you in they days. |
| ‘You didn’t have anybody to teach you in these days.’ |
| (Somerset dialect [Tubau 2016: 145]). |
The appearance of SN and DN interpretation for multiple negative elements in different English varieties lead some linguists to theorize that two grammars for the English language exist: a DN grammar for standard and a NC grammar for non-standard English varieties (de Swart 2020; Puskás 2012). Such multiple grammar systems result in a distinction in that standard English speakers who only have access to the DN grammar would reliably interpret double negations as an affirmative.
However, the fundamental assumption that a language can possess either a DN or a NC grammar has been contested by experimental evidence. Even in typical NC languages, in which speakers should only access the SN interpretation, different prosodic and syntactic cues can give rise to the DN interpretation, such as in French (Déprez and Yeaton 2018), Spanish (Déprez et al. 2015; Espinal et al. 2015), Catalan (Déprez et al. 2015; Espinal and Prieto 2011), or Hungarian (Puskás 2012). For instance, in (6), where the arrows refer to rising (↑) or falling (↓) intonation in the preverbal ningú (nobody) in Catalan, the contradictory countour (Espinal and Prieto 2011), a fall-rise intonation in the middle towards the end of the word, triggers the DN reading, whereas the regular rising intonation in the middle of the word coveys the SN reading (Espinal et al. 2015). Similarly, a strong primary stress on the neg-word senkinek (nobody) in Hungarian, as indicated by the capitalized letters in (7), conveys the DN interpretation while regular emphatic stress, as in the non-capitalized version, triggers the SN interpretation.
| Ni↑n↓gú↑/Ni↑ngú↓ no ha menjat postres. |
| nobody not has eaten dessert |
| ‘Everybody/Nobody ate dessert.’ |
| (Catalan [Espinal et al. 2015: 146–147]) |
| Kolozs senkinek/SENKINEK nem mondott semmit. |
| Kolozs n-person NEG said n-thing |
| ‘Kolozs didn’t say anything/nothing to anybody/NOBODY.’ (NC/DN) |
| (Hungarian [Puskás 2012: 619]) |
Furthermore, similar observations have been attested for non-standard English speakers: Dialect speakers stress both negation elements, thus employ prosodic marking, in order to obtain the DN interpretation (Labov 1972b; Thornton et al. 2016).
Experiments with varying methodological approaches concluded that even in standard English as supposedly a DN language, both structures are accessed. In a study involving artificial language learning of English native children and adults, the authors found evidence that NC languages are learned easier in comparison to DN languages (Maldonado and Culbertson 2021). In an eye-tracking reading study, participants reliably processed NC constructions, which suggests that NC grammar is present in speakers of standard English (Blanchette and Lukyanenko 2019b). A series of acceptability judgment experiments with native speakers of standard American English were reported in Blanchette (2017). This study shows on the one hand that most participants self-reported NC as not being part of their repertoire. This was investigated in a forced choice task, see (8), where people indicated whether they would use object NCI constructions (a/c) or solely NPIs (b) (Blanchette 2017). The results indicate that less than 2 % self-reported using NCI constructions. On the other hand, participants judged NC more acceptable if the NCI was in object position compared to subject position; in the latter, the DN interpretation was favored, see 9b. The author concludes that the self-report results are due to the social stigma of NC so that many speakers would avoid admitting their usage or would not be aware of it altogether, and more crucially, that DN was not generally favored over NC, but rather that the acceptability of NC is sensitive to the syntactic structure of the context or constructions.
| Imagine a situation in which you have finished dinner, and you want to tell someone that dessert was not a part of your meal. Which of the following would you be more likely to say? | |
| a. | I didn’t have no dessert. |
| b. | I didn’t have any dessert. |
| c. | Either (a) or (b). |
| (Blanchette 2017: 13) |
| Nobody didn’t help patients on that day. |
| (negative quantifier in the subject position) |
| He didn’t take nobody on the trip. |
| (negative quantifier in the object position) |
| (Blanchette 2017: 10) |
Therefore, it seems to be the case that speakers of standard English can reliably access and interpret NC constructions as SN and even prefer the SN over DN interpretation given syntactic cues. That is, DN and SN interpretations seem to be present in the English grammar for multiple negative elements, whereas prescriptive and normative forces might shape the perception and usage of NC (Blanchette 2013; Horn 2010; Smith and Holmes-Elliott 2022).
2.2 Negative polarity items and negative concord items
NPIs do not employ overt negation when co-occurring with sentence negation (Blanchette and Lukyanenko 2019a; de Swart 2020) and therefore act similarly as NCIs in NC languages (Childs 2017; Giannakidou 2007). Yet, in the history of English, NPIs emerged later in comparison to NC constructions (Iyeiri 1999; Nevalainen 1998). Up until the period of Middle English, multiple negations with a SN interpretation had been common before professional and educated members of higher social ranks introduced the usage of NPIs (Nevalainen 1998, 2006). Prescriptive forces propagated multiple negations as “illogical” (cf. Horn 2010: 112; Labov 1972a: 225). Consequently, NC became obsolete in written but remained in spoken modalities.
In contrast to NC, NPI phenomena are not strictly associated with a certain type of English variety, thus NPIs[2] appear in standard as well as in non-standard English. NPI and NC can even co-occur within the same non-standard English sentence (Blanchette and Lukyanenko 2019a), see (10).
| I didn’t have no lice, and I didn’t have any itch. |
| ‘I didn’t have any lice, and I didn’t have any itch.’ |
| (Blanchette and Lukyanenko 2019a: 2) |
In order to test whether NPI and NC constructions receive similar interpretations, that is the SN reading, by American English speakers, Blanchette and Lukyanenko conducted acceptability and felicity rating studies. On a 7-point Likert scale (1 = completely unnatural/makes zero sense, 7 = completely natural/makes total sense), participants rated sentences with NPI or negative noun phrases (NP), such as anything versus nothing in conditionals or under sentence negation. Negative NPs in the conditional contexts are infelicitous, see (11a), and under sentence negation render a NC construction, which is felicitous if the SN reading is accessed, see (11b).
| a. | If my older sister leaves anything/nothing in her locker, then her backpack is gonna be a bit lighter during her walk home. |
| b. | My older sister didn’t leave anything/nothing in her locker, so her backpack is gonna be super heavy during her walk home. |
| (Blanchette and Lukyanenko 2019a: 7) |
The study showed, among others, two interesting findings: (i) under sentence negation, negative NPIs, that is NC constructions as in (11b), received lower acceptability ratings than NPI constructions; (ii) the felicity was rated higher in the sentence negation than in the conditional contexts for negative NPs, indicating that participants interpreted NC constructions as SN. Predominant DN readings would have rendered lower felicity ratings similar to the conditional context condition. The authors concluded that American English speakers show an asymmetry between NC acceptability and the readiness to interpret those constructions.
The current study has the following goals: First, we aim to replicate the findings about the usage of NC in American English (Blanchette 2017; Blanchette and Lukyanenko 2019a) and extend the investigation to British English. Secondly, we aim to look into not only the usage but also the interpretation of multiple negative elements in comparison to other negation types, e.g., verb negator with NPI or BI, in these two varieties of English. Thirdly, we take the first step of studying the usage and interpretation of these varying negative constructions in relation to their context of use, with a focus on register, which we introduce below.
2.3 Taking a register approach to NC versus NPIs
Registers are defined as sets of speech repertoires linked to specific situational and functional parameters, e.g., the identities and relations of speakers and interlocutors, circumstances and the setting of the conversation, as well as communicative purposes (Agha 2007: 147; Biber 2009; Lüdeling et al. 2022; Pescuma et al. 2023). The formation of registers is connected to social norms, in that a re-occurring linguistic behavior is acknowledged and normalized by a communicative population. Social norms bring along standardization which includes judgments based on appropriate and non-appropriate behavior in a specific culture (Agha 2007: 148–150).
Earlier literature has sometimes made a distinction between the registers of standard versus non-standard languages, or of formal versus informal languages. The register of standard languages, for example, is appropriate in specific types of interactions and situations, e.g., public or official communication (Agha 2007: 146) which require special attention to precision and form of the utterances (Heylighen and Dewaele 1999, 2002). Such formal contexts include sermons, political speeches, or court cases (Atkinson 1982; Biber et al. 2020; Irvine 1979). Contexts in which the register of non-standard languages is appropriate include informal settings, i.e., conversations with friends and family (Ferguson 1959; Heylighen and Dewaele 1999). Therefore, the degree of formality or informality in a given situation seems to be one of the dimensions distinguishing between the appropriateness of register choices.
The level of formality is a dynamic interaction between extra-linguistic (e.g., constellation of situational parameters) and intra-linguistic (e.g., linguistic markers of formal or informal speech) aspects (cf. Rotter and Liu 2023). Decisions concerning the degree of formality are taken “quickly and easily by relying on nothing more than our taken for granted competence as speakers of English” (Atkinson 1982: 91). However, given the dynamic interaction to formality, there might not be a clear cut when language users change register, i.e., from a standard to a non-standard language. Specifically in the English language, given its widespread usage across the globe, speech communities and cultures potentially have developed different intuitions about the appropriate register in a specific situation (Agha 1999, 2015; Berruto 2010).
Potential differences between American and British English speakers and their formality levels based on extra-linguistic parameters was investigated using labels of social relations between interlocutors (Rotter and Liu 2023). Participants read short stories (see [12]) with distant (i.e., judge) or close interlocutors (i.e., mother) and rated the level of formality on a 7-point Likert scale. The result showed similar rating patterns in American and British English speakers with a clear separation between formal and informal contexts. Distant social relations were reliably associated with more formal levels while the close social relation labels were related to informal levels, showing that this specific parameter suggests cross-cultural similar levels of formality.
| (S1) | Jennifer Patel is accused at court. |
| (S2) | The accusation is false. |
| (S3) | She says to her {judge/mother}: … |
| (Q1) | Is Jennifer Patel going to talk formally? |
As registers are highly culture-dependent, intra-linguistic aspects, linguistic markers appropriate for extra-linguistic parameters could vary across different English-speaking communities. More specifically, even though similar situational parameters suggest similar formality levels, American and British English speakers can still show differences in their usage of different registers.
2.4 The current study
In light of the rich theoretical background and the recent experimental work as outlined above, we saw that NC and NPI constructions show different characteristics in speakers’ usage. Both NC and NPI structures seem to be present in the standard English grammar while NPIs are preferred. This gives rise to a sensitivity with respect to situational parameters in that, putting the issue of dialects aside, NCI structures are more strongly associated with informal than formal situations, in comparison to NPI structures. We expect NCI but not NPI constructions to be sensitive to the contextual parameters relating to the formality level of the situation, i.e., the social relations of interlocutors.
Furthermore, experimental evidence suggests the availability of DN and SN interpretations for constructions with NCI in standard English grammars, but also the sensitivity to syntactic environments. Thus, we tentatively hypothesize that the register could also play a role in that formal situations could favor the DN over the SN interpretation.
Our experiments aimed to explore register effects in the usage and interpretation of NCI and NPI constructions in different English varieties. More specifically we asked the following two research questions:
Are NCI (and NPI) constructions register-sensitive (i.e., does their appropriateness vary in relation to formal vs. informal situations)?
Is the interpretation of NCI (and NPI) constructions register-sensitive (i.e., does their interpretation vary in relation to formal vs. informal situations)?
Within the context of the phenomena and register variation we have outlined, we formulated the hypotheses about the appropriateness (H1) and about the interpretation (H2) of NC and NPI constructions:
Appropriateness
NCIs are rated less appropriate than NPIs (NCI
NCIs are rated less appropriate in formal than in informal conditions (NCI formal < NCI informal ).
NPIs are rated similarly in formal and informal conditions (NPI formal ≈ NPI informal ).
Interpretation (i.e., lower ratings indicate SN interpretation, higher ratings point at ambiguities between SN and DN interpretations)
NCIs receive higher interpretation ratings than NPIs (NCI
NCIs receive higher interpretation ratings in formal conditions than in informal conditions (NCI formal < NCI informal ).
NPIs are interpreted similarly in formal and informal conditions (NPI formal ≈ NPI informal ).
In order to test the hypotheses, we conducted two rating studies in American and British English. We chose these two varieties of English as NC constructions are well attested to be part of their regional dialects. Rating studies showed to be a suitable methodological approach to the current questions. Registers are connected to normative judgments, which are done quickly based on the given linguistic and situational information. Additionally, the phenomena in question have been previously approached by acceptability ratings (cf. Blanchette and Lukyanenko 2019a, 2019b), which showed that the method is applicable to measure differences between NCI and NPI constructions. We adopted appropriateness judgments in order to avoid prescriptive judgments and draw attention to the contextual effects of the usage. Furthermore, we collected information on the demographic and language background, including their preferences for NCI or NPI constructions, similar to the forced choice task by Blanchette and Lukyanenko (2019a; see [8]). This task was included to explore whether we can: 1) replicate the findings by the original study in our American English data set, and 2) whether the self-reports in the American and British English data set are comparable.
3 Experiments
In order to test the aforementioned hypotheses, we conducted two rating studies with native speakers of American and British English. Since the methodological approach in both studies was comparable, a detailed description of the method for the American English study is given in Section 3.1. Section 3.2 presents the results. Section 3.3 addresses the differences between the British and the American experimental design. Section 3.4 presents the results of the British English study.
3.1 Methods: experiment 1
3.1.1 Design and material
The experiment was implemented using PCIbex (Zehr and Schwarz 2018) and hosted on the PCIbex farm platform (https://farm.pcibex.net/). The experimental setup was a 2 × 2 factorial design with the factors[3] FORMALITY (formal vs. informal) and NEGATION (NCI vs. NPI). Furthermore, we added BI as a control condition; it will also be analyzed for explorative reasons, but this paper will not go into further details on this. We employed a within-subject design and counterbalanced the items across participants. We used 36 critical and 66 filler items in the form of short stories, see (13). (S1) and (S2) set a context. (S3) contains the formality manipulation via social relations of interlocutors. (S4) contains the negation manipulation; we used simple past tense (see Blanchette and Lukyanenko 2019a) with object position for the negation manipulation, since this is the most frequent NC construction (Smith 2001). Furthermore, (S2) was created in such a way that the SN reading of (S4) is the more coherent one with the context, both with NCI and NPI. This manipulation was necessary as we are primarily interested in the usage of NCI and NPI in formal versus informal contexts when they are functionally equivalent. (Q1) asks about the appropriateness of (S4) in the given context, and (Q2) asks about the interpretation of (S4).
| (S1) | George Henderson works in a shop. |
| (S2) | The shop is deserted. |
| (S3) | He says to his { manager formal / mother informal }: |
| (S4) | “I didn’t see { no clients NCI / any clients NPI / clients BI } here.” |
| (Q1) | Is the sentence appropriate within the given context? |
| (Q2) | Did George Henderson see clients here? |
The validity of the appropriateness ratings and the social relation manipulations were assessed in multiple pre-tests in both American and British English. One pretest focused on the interpretation of (S2), and the other focused on the formality manipulation via interlocutor relation. They indicated that both manipulations were effective in that (S2) primed a SN reading in (S4) and interlocutor relation reliably predicted the level of formality in conversation in that distant relations were associated with more formal speeches than close relations (Rotter and Liu 2023).
3.1.2 Procedure
Participants were first informed about the nature and order of the four parts[4] in the study: (P1) participation information and consent, (P2) the experiment, (P3) short survey on their demographics, and (P4) scenarios and questions. They first gave their informed consent (P1) before accessing the study. (P2) the experiment started with four practice stories to familiarize participants with the setup. Participants indicated that they had read the sentence by pressing on the space key before the next one appeared in the middle of the screen. (S1) to (S3) were shown separately after each other. (S4) remained on the screen while (Q1) and (Q2) appeared separately under the sentence together with the scale. The answer to the question was given on a 7-point Likert scale in which all points were labeled, i.e., “Certainly/Probably/Possibly not” (1/2/3), “unclear” (4), “Possibly/Probably/Certainly yes” (5/6/7). Participants used the applicable keys to indicate their answer. After each story, a fixation cross appeared for 400 ms in order to enforce a proper distinction between the stories. After the experimental part, the participants took part in (P3) a short demographic survey including their language background, i.e., self-reports of being non-dialect or dialect speakers with the possibility to provide the specific dialect by text entry. This information solely served to describe the linguistic background of participants and we did not conduct inferential statistics based on these self-reports. Lastly, we conducted (P4) a forced choice task using the same item (see [8]) as in Blanchette (2017) to test for participants’ self-reports about their NC usage. This task was placed at the end of the study to not interfere with the appropriateness and interpretation judgments of (P2).
3.1.3 Participants
Participants were recruited on the crowd-sourcing platform (https://www.prolific.co/) with the following pre-screeing criteria: current location and country of birth is the US, first language is English with no language disorders. They received monetary compensation for their participation. The experiment took roughly 35 min. In total, we collected data from 120 American English native speakers (mean age = 36.2, SD = 12.1, range = 18–65; women: N = 60, men: N = 60). All participants provided their informed consent as approved by the Ethics Committee of the Deutsche Gesellschaft für Sprachwissenschaft (DGfS) in the context of SFB 1412 ‘Register’.
In total, 9.2 % of the participants reported to be dialect speakers with four different dialect backgrounds (i.e., both African American Vernacular and Southern English, [N = 1] as well as Midwestern American [N = 3], Southern [N = 6] American English dialect). The majority had completed college (46.6 %); 33.4 % had completed high school; 19.2 % had a graduate degree, and below one percent of the participants did not finish high school. More than half of the participants grew up in a suburban environment (56.6 %), while roughly the same amount grew up in a rural (20 %) or urban (23.4 %) environment.
In the forced choice task, the majority (N = 115) of the participants indicated that they prefer the NPI construction. Two participants indicated that they prefer the NCI construction, and three indicated no preference between NPI or NCI constructions – none of these participants reported to be dialect speakers.
3.1.4 Data analysis
The data was processed and analyzed with the software ‘R’ (R Core Team 2023; Version 4.3.1) in the RStudio environment (RStudio Team 2023). Our inclusion criteria were that participants are native English speakers and aged between 18 and 65 years, which all participants matched with. The critical measurements were ratings for (Q1/2). We also collected data of reading times for (S1) to (S4) and of reaction times for (Q1/2). We used the reading times for (S1) to (S4) and the reaction time for (Q1/2) to exclude entire trails if they were below 400 ms – meaning that participants did not attend to the sentence. This resulted in the removal of 921 (21.3 %) entire trails. If more than 25 % of a participant’s trail set had to be removed, we excluded the entire data of the participant (low attendance rate), which resulted in the data removal of 11 participants.
The remaining data was analyzed using the package ‘ordinal’ (Christensen 2019) in the cumulative link function model framework (see Howcroft and Rieser 2021; Liddell and Kruschke 2018 for discussion).
We sum coded the main effect FORMALITY with its levels formal (0.5) and informal (−0.5). The three levels of NEGATION were treatment coded such that NCI (1) and BI[5] (1) were compared with the reference level NPI (0). We added 2-way interactions between FORMALITY and NCI versus NPI as well as BI versus NPI. If the 2-way interactions turned out significant, we split the data along the NEGATION factor to further identify the direction of the interaction.
The link functions were identified by comparing the loglikelihood values of each of the five possible link functions (i.e., probit, logit, cauchit, loglog, and cloglog) with each other. The function with the highest value was chosen for the model (see Christensen 2019). This method has the advantage that the chosen link function follows the shape of the data and accounts for skewness to maximize the model’s fit. The used link functions for each model of the analyses are indicated in the result section. Random effect structures were obtained with the help of the most parsimonious model approach (cf. Bates et al. 2018); the used models are indicated in the result section. P-values were obtained with the help of loglikelihood ratio test comparisons of nested models (Bates et al. 2018). If models did not converge during the significance testing, we constructed simpler models taking the current random effect structures into consideration and compared their AIC (Akaike information criterion) scores. The scores indicate the fit of the statistical models on the same data set; the smaller the score the better is the quality of the model for the given data set.
All statistical values of means, estimates, and the like are rounded to the second decimals except for p-values smaller than 0.01.
3.2 Results
The ratings in (Q1) and (Q2) are depicted in Figure 1 with the descriptive statistics in Table 1. The overall means in comparison to by subject means are depicted in Figure 2.
Means and error bars of appropriateness (Q1, left) and interpretation (Q2, right) ratings on a 7-point Likert scale (1 = Certainly not, 7 = Certainly yes) for the six conditions of Exp. 1 (American English study). Colors indicate the levels of the NEGATION factor: Blue lines and bars are for negative concord item (NCI) conditions. Orange lines and bars are for negative polarity item (NPI) conditions. Yellow lines and bars are for bare indefinite (BI) conditions. The factor FORMALITY is represented on the right on the x-axis of each graph.
Descriptive statistics of Exp. 1 (American English study). Total amount of data points (N), mean, standard deviation (SD), and median for the appropriateness (Q1, left) and interpretation (Q2, right) ratings on a 7-point Likert scale (1 = Certainly not, 7 = Certainly yes) for the six conditions. NCI abbreviates negative concord items, NPI negative polarity items, and BI bare indefinite.
| Q1 | Q2 | ||||||
|---|---|---|---|---|---|---|---|
| Formality | Negation | N | Mean (SD) | Median | Mean (SD) | Median | |
| 1 | Formal | NCI | 556 | 4.86 (2.18) | 6 | 1.48 (1.12) | 1 |
| 2 | Formal | NPI | 564 | 6.08 (1.46) | 7 | 1.29 (0.78) | 1 |
| 3 | Formal | BI | 547 | 5.79 (1.62) | 6 | 1.31 (0.82) | 1 |
| 4 | Informal | NCI | 543 | 5.05 (2.16) | 6 | 1.46 (1.14) | 1 |
| 5 | Informal | NPI | 561 | 6.14 (1.32) | 7 | 1.22 (0.64) | 1 |
| 6 | Informal | BI | 548 | 5.99 (1.44) | 7 | 1.28 (0.84) | 1 |
By subject means of the appropriateness ratings (Q1) in comparison to overall means of Exp. 1 (American English study). The x-axis depicts the factor NEGATION with negative concord items (NCI) to the left, negative polarity items (NPI) in the middle, and bare indefinite (BI) to the right. The y-axis depicts the ratings on a 7-point Likert scale (1 = Certainly not, 7 = Certainly yes). Transparent dots represent by subject means, opaque dots show the overall mean with error bars. Color indicates the levels of the factor FORMALITY: Dots in yellow represent the formal condition and dots in blue the informal condition.
3.2.1 Q1 – appropriateness rating
The output of the model with the appropriateness ratings (Q1) as dependent variable is shown in Table 2. The logit-link function model with random subject intercepts and slopes for NCI versus NPI as well as random item intercepts with slopes for FORMALITY and NCI versus NPI fit the data best.
(Q1) – Appropriateness ratings of Exp. 1 (American English study): output of the main analysis using cumulative link mixed models. NCI abbreviates negative concord items, NPI negative polarity items, and BI bare indefinite conditions. The symbol ‘*’ indicates significant results. The last column represents the used formula with fixed and random effect structure for the analysis.
| Estimates | Fixed effects | Model comparison | |||
|---|---|---|---|---|---|
|
|
Std. error | z-Value | χ2 (1) | p-Value | |
| FORMALITY |
|
0.14 | 0.04 | 0.01 | 0.97 |
| NCI versus NPI | −1.84 | 0.25 | −7.27 | 42.10 |
|
| BI versus NPI | −0.44 | 0.09 | −4.89 | 24.02 |
|
| 2-way interactions | |||||
| FORMALITY:NCI versus NPI | −0.37 | 0.18 | −2.05 | 4.19 | 0.04* |
| FORMALITY:BI versus NPI | −0.31 | 0.18 | −1.70 | 2.89 | 0.09 |
|
|
|||||
| clmm(measurement ∼ 1 + FORMALITY + NCI_NPI + BI_NPI + FORMALITY:NC_NPI + FORMALITY:BI_NPI + (1 + NCI_NPI| subj) + (1 + FORMALITY + NC_NPI| item), link = logit) | |||||
The results showed that NCI was rated significantly lower than NPI (
The models’ output of the sub-analysis is shown in Table 3. For the NCI model, the logit link function with random subject and item intercepts fit the data best. The result revealed a significant main effect of FORMALITY in that the formal conditions received lower ratings than the informal conditions (
(Q1) – Appropriateness ratings: output of the sub analysis using cumulative link mixed models of Exp. 1 (American English study). NCI abbreviates negative concord items, NPI negative polarity items, and BI bare indefinite conditions. The symbol ‘*’ indicates significant results. The last column of each analysis represents the used formula with fixed and random effect structure for the analysis.
| Analysis | Estimates | Fixed effects | Model comparison | |||
|---|---|---|---|---|---|---|
|
|
Std. error | z-Value | χ2 (1) | p-Value | ||
| NCI-model | FORMALITY | −0.41 | 0.13 | −3.22 | 10.40 | 0.001* |
| clmm(measurement ∼ FORMALITY + (1| subj) + (1| item), link = logit) | ||||||
| NPI-model | FORMALITY | 0.06 | 0.12 | 0.51 | 0.26 | 0.61 |
| clmm(measurement ∼ 1 + formality + (1| subj) + (1 + FORMALITY| item), link = cloglog) | ||||||
3.2.2 Q2 – interpretation rating
The output of the model with the interpretation (Q2) ratings (i.e., lower ratings indicate SN interpretation, higher ratings point at ambiguities between SN and DN interpretations) as dependent variable is shown in Table 4. The cloglog-link function model with random subject intercepts fit the data best.
(Q2) – Interpretation ratings of Exp. 1 (American English study): output of the main analysis using cumulative link mixed models of the American data set. NCI abbreviates negative concord items, NPI negative polarity items, and BI bare indefinite conditions. The symbol ‘*’ indicates significant results. The last column represents the used formula with fixed and random effect structure for the analysis.
| Estimates | Fixed effects | Model comparison | |||
|---|---|---|---|---|---|
|
|
Std. error | z-Value | χ2 (1) | p-Value | |
| FORMALITY | 0.18 | 0.15 | 1.16 | 1.34 | 0.25 |
| NCI versus NPI | 0.59 | 0.10 | 5.95 | 36.06 |
|
| BI versus NPI | 0.16 | 0.10 | 1.48 | 2.20 | 0.14 |
| 2-way interactions | |||||
| FORMALITY:NCI versus NPI | 0.01 | 0.20 | 0.03 |
|
0.98 |
| FORMALITY:BI versus NPI | 0.03 | 0.21 | 0.16 | 0.03 | 0.88 |
|
|
|||||
| clmm(measurement ∼ 1 + FORMALITY + NCI_NPI + BI_NPI + FORMALITY:NCI_NPI + FORMALITY:BI_NPI + (1 | subj), link = cloglog) | |||||
The results showed that the NCI conditions received significantly higher ratings than the NPI conditions (
3.3 Methods: experiment 2
3.3.1 Participants
Participants were recruited on the crowd-sourcing platform with the following pre-screeing criteria: current location and country of birth is the UK, first language is English with no language disorders. They received monetary compensation for their participation. In total, we collected data from 120 British English native speakers (mean age = 37.1, SD = 11.4, range = 20–63; women: N = 60, men: N = 60). The experiment took roughly 35 min. We received informed consent from all participants.
In total, 17.4 % of the participants reported to be dialect speakers, eliciting 13 different dialects (i.e., Bristolian [N = 1], Brummie [N = 1], Cardiff [N = 1], Derbyshire [N = 1], Doric [N = 1], Estuary [N = 1], Geordie [N = 1], Leicester [N = 1], Midlands [N = 1], Norfolk [N = 1], Northern [N = 1], Scottish [N = 2], Sussex [N = 1], Welsh [N = 1], and Yorkshire [N = 3] English dialect). As for the education levels, we used the same terms as in the US study: The majority had a graduate degree (46.6 %); 29.2 % had completed college; 23.4 % had a high school degree, and below one percent of the participants did not finish high school. However, we realized after collecting the data that the terms ‘college’ and ‘graduate degree’ are used differently in the UK in that college refers to students aged between 16 and 18 who are taking their A-Levels, and the master’s degree and the doctoral degree are called postgraduate degrees. Thus, the indicated percentage does not reflect a realistic distribution of the participants’ education levels. In other words, 46.6 % of the participants might consist of those of who have a bachelor’s and those who have a postgraduate degree. More than half of the participants grew up in a suburban environment (52.4 %), while roughly a third (32.3 %) grew up in an urban and 15 % in a rural environment. We did not perform any inferential statistics based on the demographic information of our participants.
In the forced choice task, the majority (N = 116) indicated that they prefer the NPI construction. Two participants, who self-reported to be non-dialect speakers, indicated that they prefer the NC construction. Two participants, who were speakers of Scottish or Derbyshire dialects, indicated that they would use either construction.
3.4 Results
All participants matched our inclusion criteria. In total, 845 (19.6 %) entire trails were removed based on the reading times. Nine participants showed an attention rate below 75 %, their data was excluded from the analysis. The ratings in (Q1) and (Q2) are depicted in Figure 3 with the descriptive statistics in Table 5. The overall means in comparison to by subject means are depicted in Figure 4.
Means and error bars of appropriateness (Q1, left) and interpretation (Q2, right) ratings on a 7-point Likert scale (1 = Certainly not, 7 = Certainly yes) for the six conditions of Exp. 2 (British English study). Colors indicate the levels of the NEGATION factor: Blue lines and bars are for negative concord item (NCI) conditions. Orange lines and bars are for negative polarity item (NPI) conditions. Yellow lines and bars are for bare indefinite (BI) conditions. The factor FORMALITY is represented on the x-axis on each graph.
Descriptive statistics of Exp.2 (British English study). Means and standard deviations are rounded to the second decimal. Total amount of data points (N), mean, standard deviation (SD), and median for the appropriateness (Q1, left) and interpretation (Q2, right) ratings on a 7-point Likert scale (1 = Certainly not, 7 = Certainly yes) for the six conditions. NCI abbreviates negative concord items, NPI negative polarity items, and BI bare indefinite.
| Q1-ratings | Q2-ratings | ||||||
|---|---|---|---|---|---|---|---|
| Formality | Negation | N | Mean (SD) | Median | Mean (SD) | Median | |
| 1 | Formal | NCI | 560 | 4.88 (1.99) | 5 | 1.99 (1.45) | 1 |
| 2 | Formal | NPI | 566 | 5.73 (1.51) | 6 | 1.56 (1.14) | 1 |
| 3 | Formal | BI | 575 | 5.52 (1.64) | 6 | 1.56 (1.17) | 1 |
| 4 | Informal | NCI | 568 | 4.93 (1.91) | 5 | 2.04 (1.49) | 1 |
| 5 | Informal | NPI | 567 | 5.78 (1.47) | 6 | 1.63 (1.31) | 1 |
| 6 | Informal | BI | 563 | 5.60 (1.59) | 6 | 1.52 (1.06) | 1 |
By subject means of the appropriateness ratings (Q1) in comparison to overall means of Exp. 2 (British English study). The x-axis depicts the factor NEGATION with negative concord item (NCI) to the left, negative polarity items (NPI) in the middle, and bare indefinite (BI) to the right. The y-axis depicts the ratings on a 7-point Likert scale (1 = Certainly not, 7 = Certainly yes). Transparent dots represent by subject means, opaque dots show the overall mean with error bars. Color indicates the levels of the factor FORMALITY: Dots in yellow represent the formal condition and dots in blue the informal condition.
3.4.1 Q1 – appropriateness rating
The output of the model with the appropriateness ratings (Q1) as dependent variable is shown in Table 6. The logit-link function model with random subject intercepts and slopes for NCI versus NPI and BI versus NPI, as well as random item intercepts with slopes for FORMALITY and NCI versus NPI fit the data best.
(Q1) – Appropriateness ratings of Exp. 2 (British English study): output of the main analysis using cumulative link mixed models. NCI abbreviates negative concord items, NPI negative polarity items, and BI bare indefinite conditions. The symbol ‘*’ indicates significant results. The last column represents the used formula with fixed and random effect structure for the analysis.
| Estimates | Fixed effects | Model comparison | |||
|---|---|---|---|---|---|
|
|
Std. error | z-Value | χ2 (1) | p-Value | |
| FORMALITY | −0.10 | 0.14 | −0.69 | 0.48 | 0.49 |
| NCI versus NPI | −1.20 | 0.20 | −6.09 | 31.06 |
|
| BI versus NPI | −0.29 | 0.11 | −2.52 | 6.07 | 0.01* |
| 2-way interactions | |||||
| FORMALITY:NCI versus NPI | −0.01 | 0.17 | −0.02 |
|
0.98 |
| FORMALITY:BI versus NPI | −0.13 | 0.17 | −0.76 | 0.58 | 0.45 |
|
|
|||||
| clmm(measurement ∼ FORMALITY + NCI_NPI + BI_NPI + FORMALITY: NCI_NPI + FORMALITY:BI_NPI + (1 + NCI_NPI + BI_NPI| subj) + (1 + FORMALITY + NCI_NPI| item), link = logit) | |||||
The results showed that NCI was rated significantly lower than NPI (
3.4.2 Q2 – interpretation rating
The output of the model with the interpretation ratings (Q2) (i.e., lower ratings indicate SN interpretation, higher ratings point at ambiguities between SN and DN interpretations) as dependent variable is shown in Table 7. The logit-link function model with random subject and item intercepts fit the data best.
(Q2) – Interpretation ratings of Exp. 2 (British English study): output of the main analysis using cumulative link mixed models. NCI abbreviates negative concord items, NPI negative polarity items, and BI bare indefinite conditions. The symbol ‘*’ indicates significant results. The last column represents the used formula with fixed and random effect structure for the analysis.
| Estimates | Fixed effects | Model comparison | |||
|---|---|---|---|---|---|
|
|
Std. error | z-Value | χ2 (1) | p-Value | |
| FORMALITY | 0.05 | 0.14 | 0.38 | 0.14 | 0.71 |
| NCI versus NPI | 1.00 | 0.10 | 10.34 | 110.20 |
|
| BI versus NPI | −0.10 | 0.10 | −0.98 | 0.95 | 0.33 |
| 2-way interactions | |||||
| FORMALITY:NCI versus NPI | −0.14 | 0.19 | −0.72 | 0.52 | 0.47 |
| FORMALITY:BI versus NPI | 0.02 | 0.20 | 0.10 | 0.01 | 0.92 |
|
|
|||||
| clmm(measurement ∼ FORMALITY + NCI_NPI + BI_NPI + FORMALITY: NCI_NPI + FORMALITY:BI_NPI + (1| subj) + (1| item), link = logit) | |||||
The results showed that the NCI conditions received significantly higher ratings than the NPI conditions (
4 Discussion
The current study takes a first step to explore the effects of register on the usage and interpretation of negation and polarity constructions in American and British English. The results of both experiments are summarized in Table 8.
Overview of the statistical results of the American (AE) and British (BE) studies. The symbol
| Question | Analysis | Level comparison | AE | BE |
|---|---|---|---|---|
| Q1 | Main | NCI versus NPI | NCI
|
NCI
|
| BI versus NPI | BI
|
BI
|
||
| int F:NEG | Significant | – | ||
| Sub | NCI | NCI formal < NCI informal | – | |
| NPI | – | – | ||
| Q2 | Main | NCI versus NPI | NCI
|
NCI
|
In both experiments, we found evidence that NCI structures were rated less appropriate than NPI structures; thus, hypothesis (H1a) is confirmed. With respect to the register sensitivity, only the American data set revealed effects in NCI constructions in that the ratings were higher in informal than in formal conditions; thus hypothesis (H1b) is only partially confirmed. In both studies, no register effects of NPI constructions were found, this null result is in line with the hypothesis (H1c) that such structures are not specific to a register.
One could argue based on the lack of a register effect in NCI of the British study that such constructions are just not used in British English independently of the contexts. However, the overall mean of NCI is above the scale mean, therefore in general NCI constructions are considered as rather appropriate. This could suggest that the register sensitivity of NCI is less pronounced in British English.
With respect to the interpretation, we found significant differences between NPI and NCI constructions in both data sets. NCI received higher ratings than NPI, indicating that the DN reading was available; this is in line with the hypothesis (H2a).
The null result of the difference BI versus NPI suggests that these forms are similar with respect to their interpretation. However, no register difference in the NCI, NPI, or BI interpretation was found in either data set; therefore, while the null finding is in line with the hypothesis (H2c), (H2b) could not be confirmed. The absence of a register effect in NC interpretations can be explained by our items. As discussed in Section 3.1.1, (S2) controlled for the SN reading in the critical sentence, therefore, it is not surprising that the DN interpretation was not reliably competing. Furthermore, the critical sentences were constructed with the object position for the second negative marker; this was previously found to be a syntactic cue for the SN interpretation with multiple negative elements (cf. Blanchette and Lukyanenko 2019b; Thornton et al. 2016).
The results of the forced choice task showed that in both experiments, the majority of the participants reported to prefer NPI over NCI constructions. Crucially, only in the British data set did the result match with the linguistic background: Two participants who indicated being a dialect speaker also reported using both structures, NCI or NPI. The rest of the participants who reported using NCI or having no preference were non-dialect speakers. This result could hint towards the direction that NCI constructions are part of the grammar for some self-reported non-dialect speakers. In general though, our overall distribution in both studies, namely that the vast majority prefers NPI constructions, replicates the findings by Blanchette and Lukyanenko (2019a) for American English and extends it to British English. Due to the prescriptive forces and social stigma attached to NCI constructions, these findings might not represent the real distribution of NCI use (prescriptive forces have been at play for centuries, see Tieken-Boon van Ostade 1982). Participants might either not be aware of their own usage or they might consciously deny it to avoid social stigma. Crucially, the overall ratings of NCI were above the mean of the scale (e.g., NCI: meanexp1 = 4.92, meanexp2 = 4.89; NPI: meanexp1 = 6.10, meanexp 2 = 5.74), indicating that most participants accepted NCI constructions in either context. However, Figure 2 for the American English and Figure 4 for the British English study show a large variation with regard to the mean rating of the NCI condition across both formality levels. The vast majority of the subject means are located above in the higher appropriateness ranges and thus above the scale’s neutral point (4) for the NPI and BI conditions. The mean ratings of the NCI conditions appear to be more spread out with most subjects in the higher appropriateness ranges, but with some means in the inappropriatenes ranges. Thus, NCI constructions appear to be more subject to inter-individual variation than NPI or BI constructions, which can stem from prescriptive or social forces, e.g., some subjects might have rated NCI constructions as less appropriate due to the prescriptive ungrammaticality or due to a specific social meaning attached to it.
Overall, though, the clear difference of NCI versus NPI replicates the findings of previous acceptability studies (Blanchette and Lukyanenko 2019a) for American English and extends it into a British context. The difference between NPI and BI constructions in the British data set needs to be subject to future research: We speculate that it might result from the markedness of BI in negative contexts, e.g., I didn’t see clients here. might generate a contrastive focus reading (I didn’t see clients here … but salespersons) which involves more complex representations and higher processing costs.
The research questions guiding our experimental process could partially be answered. With respect to (RQ1), we note that NCI was confirmed to be register sensitive for only the American data set in that informal contexts (with close interlocutor relations) increase the appropriateness of such constructions. NPI constructions turned out to be non-sensitive to the different situational parameters in American or British English. With respect to (RQ2), we note that the interpretation of NCI was found to be independent of the situational parameters, suggesting that the standard language register does not promote the DN interpretation in our setup. However, NCI constructions rather foster the SN interpretation if the second negative marker appears in object position compared to subject position, thereby suppressing the DN interpretation. Since the items included NCIs in object position, while also controlling for the SN interpretation with the information provided in the second context sentence, the DN interpretation might have not been accessed. Nevertheless, we found evidence that NCI generally is rated higher than NPI. This suggests that while NPIs (and BI) are interpreted as an SN, the interpretation of NCI hints at an ambiguity. This needs to be further researched taking the position of the negative marker as well as the item structure into consideration. The interpretation of NPIs in both experiments turned out not to be dependent on the formality of the situation, either.
Before we finish the paper, we would like to briefly address the scope and limits of the study, as well as potential outlooks extending the register approach to the given and additional phenomena of negation and polarity. First, registers are multidimensional, meaning that they are the product of a set of functional and situational parameters and different linguistic features (Biber 2009). Since we controlled one single situational parameter, namely the interlocutor relation, we need to interpret the results with care. Future studies should take additional situational and functional parameters into account to gain a more naturalistic picture of language use.
Secondly, our items were controlled for the SN interpretation in the critical sentence with (S2) but also with a syntactic cue. Hence, further research with varying syntactic structures and without a contextual cue for the interpretation in combination with registers needs to be conducted for further validation of the hypotheses.
Thirdly, our study focused on the NCI usage in specific situational parameters, however, NC usage is also subject to stratification and reflects social meaning (Eckert 2019; Moore 2021). NC constructions were reliably attested in different regional American and British dialects (cf. Smith 2001) but not in their standard forms. Thus, including the individual differences in interaction with situational parameters and the resulting social meanings of NC use would shed further light on the variation in the American and British contexts.
Fourth, studies have shown that the ethnicity of language users plays a role in their perception of NC in different varieties of English (Eckert 1988, 1989; Moore 2021; Wolfram and Schilling-Estes 2016b). This data was not collected in the reported studies, therefore, we cannot draw any conclusions in this regard.
Lastly, while our study did not show register-sensitivity of the NPI any, we are far from being able to make a general claim about NPIs, since there are different kinds of NPIs with varying rhetorical functions, such as emphasis and attenuation (Israel 1996). Emphatic minimizer NPIs such as lift a finger/sleep a wink (e.g., He didn’t lift a finger to help me./He didn’t sleep a wink last night.) might differ from the emphatic NPI any. Such emphatic NPIs might differ from attenuating ones all that/much (e.g., He isn’t all that smart/He didn’t sleep much). As these expressions result in understatements or overstatements of varying degrees, the choice of one over the other might be related to register factors, at least to some extent, see e.g., Brown and Levinson (1987: 68–69) on over- and understatement as face-saving or threatening strategies. Despite the fact that we were not able to address these questions in detail, the current study takes a novel approach to the long studied topic of negation and polarity. Based on the strictly controlled contexts, our studies show a first impression of how and to what extent situational parameters could affect the usage and interpretation of negation and polarity constructions. These findings are relevant for better understanding not only the grammatical knowledge of negation and polarity but also register knowledge.
5 Conclusion
In this paper, we investigated the register sensitivity of NCI and NPI constructions in American and British English with a focus on their usage and interpretation in relation to formal and informal situational parameters. While NCI constructions were assumed to show sensitivity with respect to the formal and informal situational parameters, NPI constructions were assumed to be independent of such parameters. One of the clearest finding was that NCI and NPI differ with respect to their general appropriateness independently of the level of formality. Moreover, we found a register effect in the American study. Here, NCI constructions were judged as less appropriate for situations associated with higher formality levels. Our study is among the first to experimentally address the register sensitivity of negation and polarity phenomena in American and British contexts.
Funding source: Deutsche Forschungsgemeinschaft
Award Identifier / Grant number: 416591334
Acknowledgment
We thank the anonymous reviewers of Linguistics for their critical but constructive comments on previous versions of the paper. Many thanks also go to Juliane Schwab and Felix Golcher for their feedback on the experiments and data analysis, to Johan van der Auwera for helpful discussions, and to our student assistants Małgorzata Kielak and Charlotte Sorg for their support in stimuli creation and data acquisition.
-
Research funding: This work was supported by Deutsche Forschungsgemeinschaft (http://dx.doi.org/10.13039/501100001659, 416591334).
-
Data repository: The datasets for this study can be found in the online repository: https://osf.io/2pxe4/.
References
Agha, Asif. 1999. Register. Journal of Linguistic Anthropology 9(1–2). 216–219. https://doi.org/10.1525/jlin.1999.9.1-2.216.Search in Google Scholar
Agha, Asif. 2007. Language and social relations. Cambridge: Cambridge University Press.Search in Google Scholar
Agha, Asif. 2015. Enregisterment and communication in social history. Registers of Communications. 27–53. https://doi.org/10.2307/j.ctvggx2qk.6.Search in Google Scholar
Atkinson, John Maxwell. 1982. Understanding formality: The categorization and production of ‘formal’ interaction. The British Journal of Sociology 33(1). 86–117. https://doi.org/10.2307/589338.Search in Google Scholar
van der Auwera, Johan. 2022. Nominal and pronominal negative concord, through the lens of Belizean and Jamaican creole. Linguistics 60(2). 505–540. https://doi.org/10.1515/ling-2020-0137.Search in Google Scholar
Bates, Douglas, Reinhold Kliegl, Shravan Vasishth & Harald Baayen. 2018. Parsimonious mixed models. arXiv 1506. https://doi.org/10.48550/arXiv.1506.04967.Search in Google Scholar
Beltrama, Andrea. 2020. Social meaning in semantics and pragmatics. Language and Linguistics Compass 14(9). 1–20. https://doi.org/10.1111/lnc3.12398.Search in Google Scholar
Berruto, Gaetano. 2010. Identifing dimensions of linguistic variation in a language space. In Peter Auer & Jürgen Erich Schmidt (eds.), Volume 1 theories and methods: An international handbook of linguistic variation, 226–241. Berlin: De Gruyter Mouton.10.1515/9783110220278.226Search in Google Scholar
Biber, Douglas. 2009. Multi-dimensional approaches. In Anke Lüdeling & Merja Kytö (eds.), Corpus linguistic an international handbook, vol. 2, 822–855. Berlin & New York: De Gruyter Mouton.10.1515/9783110213881.2.822Search in Google Scholar
Biber, Douglas. 2012. Register as a predictor of linguistic variation. Corpus Linguistics and Linguistic Theory 8(1). 9–37. https://doi.org/10.1515/cllt-2012-0002.Search in Google Scholar
Biber, Douglas & Susan Conrad. 2019. Register studies in context. In Douglas Biber & Susan Conrad (eds.), Register, genre, and style, 2nd edn. (Cambridge Textbooks in Linguistics), 297–317. Cambridge: Cambridge University Press.10.1017/9781108686136Search in Google Scholar
Biber, Douglas, Jesse Egbert & Daniel Keller. 2020. Reconceptualizing register in a continuous situational space. Corpus Linguistics and Linguistic Theory 16(3). 581–616. https://doi.org/10.1515/cllt-2018-0086.Search in Google Scholar
Blanchette, Frances. 2013. Negative concord in English. Linguistic Variation 13(1). 1–47. https://doi.org/10.1075/lv.13.1.01bla.Search in Google Scholar
Blanchette, Frances. 2015. English negative concord, negative polarity, and double negation. New York, NY: CUNY Graduate Center PhD thesis.Search in Google Scholar
Blanchette, Frances. 2017. Micro-syntactic variation in American English negative concord. Glossa: A Journal of General Linguistics 2(1). 65. https://doi.org/10.5334/gjgl.188.Search in Google Scholar
Blanchette, Frances & Cynthia Lukyanenko. 2019a. Asymmetries in the acceptability and felicity of English negative dependencies: Where negative concord and negative polarity (do not) overlap. Frontiers in Psychology 10. https://doi.org/10.3389/fpsyg.2019.02486.Search in Google Scholar
Blanchette, Frances & Cynthia Lukyanenko. 2019b. Unacceptable grammars? An eye-tracking study of English negative concord. Language and Cognition 11(1). 1–40. https://doi.org/10.1017/langcog.2019.4.Search in Google Scholar
Breitbarth, Anne, David Willis & Christopher Lucas. 2020. The history of negation in the languages of Europe and the Mediterranean. Oxford University Press.10.1093/oso/9780199602544.001.0001Search in Google Scholar
Brown, Penelope & Stephen C. Levinson. 1987. Politeness: Some universals in language usage. Cambridge: Cambridge University Press.10.1017/CBO9780511813085Search in Google Scholar
Childs, Claire. 2017. Integrating syntactic theory and variationist analysis: The structure of negative indefinites in regional dialects of British English. Glossa: A Journal of General Linguistics 2(1). https://doi.org/10.5334/gjgl.287.Search in Google Scholar
Childs, Claire, Christopher Harvey, Karen Corrigan & Sali Tagliamonte. 2015. Comparative sociolinguistic insights in the evolution of negation. University of Pennsylvania Working Papers in Linguistics 21(2). 4.Search in Google Scholar
Christensen, Rune Haubo Bojesen. 2019. Ordinal—regression models for ordinal data. R package version 2022.11-16. Available at: https://CRAN.R-project.org/package=ordinal.Search in Google Scholar
Déprez, Viviane, Susagna Tubau, Anne Cheylus & Teresa Espinal. 2015. Double negation in a negative concord language: An experimental investigation. Lingua 163. 75–107. https://doi.org/10.1016/j.lingua.2015.05.012.Search in Google Scholar
Déprez, Viviane & Jeremy Yeaton. 2018. French negative concord and discord: An experimental investigation of contextual and prosodic disambiguation. In Lori Repetti & Francisco Ordóñez (eds.), Romance languages and linguistic theory 14: Selected papers from the 46th linguistic symposium on Romance languages (LSRL), 35–51. Amsterdam: John Benjamins.10.1075/rllt.14.03depSearch in Google Scholar
Eckert, Penelope. 1988. Adolescent social structure and the spread of linguistic change. Language in Society 17(2). 183–207. https://doi.org/10.1017/s0047404500012756.Search in Google Scholar
Eckert, Penelope. 1989. Jocks and Burnouts: Social categories and identity in the high school. New York: Teachers College Press.Search in Google Scholar
Eckert, Penelope. 2019. The limits of meaning: Social indexicality, variation, and the cline of interiority. Language 95(4). 751–776. https://doi.org/10.1353/lan.0.0239.Search in Google Scholar
Espinal, Teresa & Pilar Prieto. 2011. Intonational encoding of double negation in Catalan. Journal of Pragmatics 43(9). 2392–2410. https://doi.org/10.1016/j.pragma.2011.03.002.Search in Google Scholar
Espinal, Teresa, Susagna Tubau, Joan Borràs-Comes & Pilar Prieto. 2015. Double negation in Catalan and Spanish. Interaction between syntax and prosody. Language, Cognition, and Mind 145–176. https://doi.org/10.1007/978-3-319-17464-8$\_$7.10.1007/978-3-319-17464-8_7Search in Google Scholar
Ferguson, Charles A. 1959. Diglossia. WORD 15(2). 325–340. https://doi.org/10.1080/00437956.1959.11659702.Search in Google Scholar
Giannakidou, Anastasia. 2000. Negative … concord? Natural Language & Linguistic Theory 18. 457–523. https://doi.org/10.1023/A:1006477315705.10.1023/A:1006477315705Search in Google Scholar
Giannakidou, Anastasia. 2007. N-words and negative concord. The Blackwell Companion to Syntax 3. 327–391.10.1002/9780470996591.ch45Search in Google Scholar
Giannakidou, Anastasia. 2020. Negative concord and the nature of negative concord items. In Viviane Déprez & M. Teresa Espinal (eds.), The Oxford handbook of negation, 458–478. Oxford: Oxford University Press.10.1093/oxfordhb/9780198830528.013.25Search in Google Scholar
Hall-Lew, Lauren, Emma Moore & Robert J. Podesva. 2021. Social meaning and linguistic variation: Theoretical foundations. Social Meaning and Linguistic Variation. 1–24. https://doi.org/10.1017/9781108578684.001.Search in Google Scholar
Heylighen, Francis & Jean-Marc Dewaele. 1999. Formality of language: Definition, measurement and behavioral determinants. Internal Report, and Léo Apostel. Available at: http://pespmc1.vub.ac.be/Papers/Formality.pdf.Search in Google Scholar
Heylighen, Francis & Jean-Marc Dewaele. 2002. Variation in the contextuality of language: An empirical measure. Foundations of Science 7. 293–340. https://doi.org/10.1023/A:1019661126744.10.1023/A:1019661126744Search in Google Scholar
Horn, Laurence. 2009. Hypernegation, hyponegation, and parole violations. Annual Meeting of the Berkeley Linguistics Society 35(1). 403. https://doi.org/10.3765/bls.v35i1.3628.Search in Google Scholar
Horn, Laurence. 2010. Multiple negation in English and other languages. In Laurence Horn (ed.), The expression of negation, 111–148. Berlin, New York: De Gruyter Mouton.10.1515/9783110219302.111Search in Google Scholar
Howcroft, David M. & Verena Rieser. 2021. What happens if you treat ordinal ratings as interval data? Human evaluations in NLP are even more under-powered than you think. In Proceedings of the 2021 conference on empirical methods in natural language processing, 8932–8939 [Online]. Punta Cana, Dominican Republic: Association for Computational Linguistics.10.18653/v1/2021.emnlp-main.703Search in Google Scholar
Irvine, Judith T. 1979. Formality and informality in communicative events. American Anthropologist 81(4). 773–790. https://doi.org/10.1525/aa.1979.81.4.02a00020.Search in Google Scholar
Israel, Michael. 1996. Polarity sensitivity as lexical semantics. Linguistics and Philosophy 19. 619–666. https://doi.org/10.1007/BF00632710.Search in Google Scholar
Iyeiri, Yoko. 1999. Multiple negation in Middle English verse. In Ingrid Tieken-Boon van Ostade, Gunnel Tottie & Wim van der Wurff (eds.), Negation in the history of English, 121–146. Berlin & Bosten: De Gruyter Mouton.10.1515/9783110806052.121Search in Google Scholar
Krifka, Manfred. 1995. The semantics and pragmatics of polarity items. Linguistic Analysis 25. 209–257.Search in Google Scholar
Labov, William. 1972a. Language in the inner city: Studies in the black English vernacular. Philadelphia: University of Pennsylvania Press.Search in Google Scholar
Labov, William. 1972b. Negative attraction and negative concord in English grammar. Language 48(4). 773–818. https://doi.org/10.2307/411989.Search in Google Scholar
Labov, William. 1972c. Sociolinguistic patterns. Philadelphia: University of Pennsylvania Press.Search in Google Scholar
Liddell, Torrin M. & John K. Kruschke. 2018. Analyzing ordinal data with metric models: What could possibly go wrong? Journal of Experimental Social Psychology 79. 328–348. https://doi.org/10.1016/j.jesp.2018.08.009.Search in Google Scholar
Lippi-Green, Rosina. 1994. Accent, standard language ideology, and discriminatory pretext in the courts. Language in Society 23(2). 163–198. https://doi.org/10.1017/s0047404500017826.Search in Google Scholar
Lippi-Green, Rosina. 2011. The myth of non-accent. In English with an accent: Language, ideology and discrimination in the United States, 2nd edn., 44–54. London & New York: Taylor & Francis Group.10.4324/9780203348802-4Search in Google Scholar
Lüdeling, Anke, Artemis Alexiadou, Aria Adli, Karin Donhauser, Malte Dreyer, Markus Egg, Anna Helene Feulner, Natalia Gagarina, Wolfgang Hock, Stefanie Jannedy, Frank Kammerzell, Pia Knoeferle, Thomas Krause, Manfred Krifka, Silvia Kutscher, Beate Lütke, Thomas McFadden, Roland Meyer, Christine Mooshammer, Stefan Müller, Katja Maquate, Muriel Norde, Uli Sauerland, Stephanie Solt, Luka Szucsich, Elisabeth Verhoeven, Richard Waltereit, Anne Wolfsgruber & Lars Erik Zeige. 2022. Register: Language users’ knowledge of situational-functional variation–frame text of the first phase proposal for the CRC 1412. Register Aspects of Language in Situation (REALIS) 1(1). 1–57. https://doi.org/10.18452/24901.Search in Google Scholar
Maldonado, Mora & Jennifer Culbertson. 2021. Nobody doesn’t like negative concord. Journal of Psycholinguistic Research 50(6). 1401–1416. https://doi.org/10.1007/s10936-021-09816-w.Search in Google Scholar
Moore, Emma. 2021. The social meaning of syntax. In Lauren Hall-Lew, Emma Moore & Robert J. Podesva (eds.), Social meaning and linguistic variation: Theorizing the third wave, 54–79. Cambridge: Cambridge University Press.10.1017/9781108578684.003Search in Google Scholar
Nevalainen, Terttu. 1998. Social mobility and the decline of multiple negation in Early Modern English. In Jacek Fisiak & Marcin Krygier (eds.), Advances in English historical linguistics, 263–292. Berlin & New York: De Gruyter Mouton. https://doi.org/10.1515/9783110804072.263.Search in Google Scholar
Nevalainen, Terttu. 2006. Negative concord as an English “vernacular universal”. Journal of English Linguistics 34(3). 257–278. https://doi.org/10.1177/0075424206293144.Search in Google Scholar
Penka, Doris. 2020. Negative indefinites and negative concord. In Lisa Matthewson, Cécile Meier, Hotze Rullmann & Thomas Ede Zimmermann (eds.), The Blackwell companion to semantics, 1–23. Hoboken, NJ: Wiley-Blackwell.10.1002/9781118788516.sem095Search in Google Scholar
Penka, Doris & Hedde Zeijlstra. 2010. Negation and polarity: An introduction. Natural Language & Linguistic Theory 28(4). 771–786. https://doi.org/10.1007/s11049-010-9114-0.Search in Google Scholar
Pescuma, Valentina N., Diana Serova, Julia Lukassek, Antje Sauermann, Roland Schäfer, Aria Adli, Felix Bildhauer, Markus Egg, Kristina Hülk, Aine Ito, Stefanie Jannedy, Valia Kordoni, Milena Kühnast, Silvia Kutscher, Robert Lange, Nico Lehmann, Mingya Liu, Beate Lütke, Katja Maquate, Christine Mooshammer, Vahid Mortezapour, Stefan Müller, Muriel Norde, Elizabeth Pankratz, Angela G. Patarroyo, Ana-Maria Plesca, Camilo Rodríguez-Ronderos, Stephanie Rotter, Uli Sauerland, Gohar Schnelle, Britta Schulte, Gediminas Schüppenhauer, Bianca Maria Sell, Stephanie Solt, Megumi Terada, Dimitra Tsiapou, Elisabeth Verhoeven, Melanie Weirich, Heike Wiese, Kathy Zaruba, Lars Erik Zeige, Anke Lüdeling & Pia Knoeferle. 2023. Situating language register across the ages, languages, modalities, and cultural aspects: Evidence from complementary methods. Frontiers in Psychology 13. https://doi.org/10.3389/fpsyg.2022.964658.Search in Google Scholar
Pullum, Geoffrey. 1999. African American vernacular English is not standard English with 35 mistakes. In Rebecca S. Wheeler (ed.), The workings of language: From prescriptions to perspectives, 39–58. Westport, CT: Praeger.Search in Google Scholar
Puskás, Genoveva. 2012. Licensing double negation in NC and non-NC languages. Natural Language & Linguistic Theory 30(2). 611–649. https://doi.org/10.1007/s11049-011-9163-z.Search in Google Scholar
R Core Team. 2023. R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing. Available at: https://www.R-project.org/.Search in Google Scholar
Rotter, Stephanie & Mingya Liu. 2023. Interlocutor relation predicts the formality of the conversation. Register Aspects of Language in Situation 2(2). 1–27. https://doi.org/10.18452/26192.Search in Google Scholar
RStudio Team. 2023. RStudio: Integrated development environment for R. Boston, MA: RStudio, PBC. Available at: http://www.rstudio.com/.Search in Google Scholar
Smith, Jennifer. 2001. Negative concord in the old and new world: Evidence from Scotland. Language Variation and Change 13(2). 109–134. https://doi.org/10.1017/s0954394501132011.Search in Google Scholar
Smith, Jennifer & Sophie Holmes-Elliott. 2022. Mapping syntax and the sociolinguistic monitor. In Tanya Karoli Christensen & Torben Juel Jensen (eds.), Explanations in sociosyntactic variation (Studies in Language Variation and Change), 58–89. Cambridge: Cambridge University Press.10.1017/9781108674942.004Search in Google Scholar
de Swart, Henriëtte. 2020. Double negation readings. In Viviane Déprez & M. Teresa Espinal (eds.), The Oxford handbook of negation, 478–496. Oxford: Oxford University Press.10.1093/oxfordhb/9780198830528.013.26Search in Google Scholar
Tagliamonte, Sali A. 2006. The linguistic variable. Analysing Sociolinguistic Variation. 70–98. https://doi.org/10.1017/cbo9780511801624.007.Search in Google Scholar
Thornton, Rosalind, Anna Notley, Vincenzo Moscati & Stephen Crain. 2016. Two negations for the price of one. Glossa: A Journal of General Linguistics 1(1). https://doi.org/10.5334/gjgl.4.Search in Google Scholar
Tieken-Boon van Ostade, Ingrid. 1982. Double negation and eighteenth-century English grammars. Neophilologus 66(2). 278–285. https://doi.org/10.1007/bf02050617.Search in Google Scholar
Tubau, Susagna. 2016. Lexical variation and negative concord in traditional dialects of British English. The Journal of Comparative Germanic Linguistics 19. https://doi.org/10.1007/s10828-016-9079-4.Search in Google Scholar
Wolfram, Walter A. 1996. A sociolinguistic description of detroit negro speech. Urban language series. Washington D.C: Center for Applied Linguistics.Search in Google Scholar
Wolfram, Walt & Natalie Schilling-Estes. 2016a. Dialects, standards, and vernaculars. In American English: Dialects and variation, 3rd edn. Malden, MA: Wiley-Blackwell.Search in Google Scholar
Wolfram, Walt & Natalie Schilling-Estes. 2016b. Ethnicity and American English. In American English: Dialects and variation, 3rd edn., 183–211. Malden, MA: Wiley-Blackwell.Search in Google Scholar
Zehr, Jérémy & Florian Schwarz. 2018. PennController for internet based experiments (IBEX). Available at: https://doc.pcibex.net/.Search in Google Scholar
Zeijlstra, Hedde. 2004. Sentential negation and negative Concord. Utrecht: LOT. Amsterdam: University of Amsterdam Ph.D thesis.Search in Google Scholar
© 2024 the author(s), published by De Gruyter, Berlin/Boston
This work is licensed under the Creative Commons Attribution 4.0 International License.
Articles in the same Issue
- Frontmatter
- Research Articles
- A register approach to negative concord versus negative polarity items in English
- Sociolinguistic monitoring and L2 speakers of English
- The controller-first constraint beyond the Basic Variety: how do instructed learner varieties solve contexts of competition?
- The count-mass distinction in Hong Kong Sign Language: an intra- and cross-modal comparison
- Resultative secondary predicates in cooking recipes: an empirical study of Germanic and Romance languages
- A formal approach to reanalysis and the Early Semantic Stability Hypothesis: exploring the test case of the negative counterfactual marker ʾilmale in Hebrew and Aramaic
- The colonisation of the colour pink: variation and change in Māori’s colour lexicon
- Corrigendum
- Corrigendum to: Diachronic evolution of the subordinator kak in Russian
Articles in the same Issue
- Frontmatter
- Research Articles
- A register approach to negative concord versus negative polarity items in English
- Sociolinguistic monitoring and L2 speakers of English
- The controller-first constraint beyond the Basic Variety: how do instructed learner varieties solve contexts of competition?
- The count-mass distinction in Hong Kong Sign Language: an intra- and cross-modal comparison
- Resultative secondary predicates in cooking recipes: an empirical study of Germanic and Romance languages
- A formal approach to reanalysis and the Early Semantic Stability Hypothesis: exploring the test case of the negative counterfactual marker ʾilmale in Hebrew and Aramaic
- The colonisation of the colour pink: variation and change in Māori’s colour lexicon
- Corrigendum
- Corrigendum to: Diachronic evolution of the subordinator kak in Russian
