Abstract
The study examines lexico-grammatical variation among English texts within registers, empirically testing the claim that linguistic variation among texts is attributable to the situation of use. To this end, the study relies on prototype theory and identifies linguistically central texts, that is, representative of the register central trend, and peripheral texts that deviate linguistically from the central tendency. The study then identifies the basis for this linguistic variation. The findings show that register texts vary in their prototypicality, and linguistic variation among texts is explained in functional terms through communicative differences among texts. The identified patterns of functional correspondence between linguistic and communicative variation are demonstrated consistently in four registers.
Keywords
1. Introduction
1.1. Explaining Linguistic Variation Among Texts: The Role of Register
Texts—instances of spoken or written naturally occurring communication within a particular situation of use (Brown & Yule 1983:9; Fairclough 1992:4; Stubbs 1996:1-2; Biber & Conrad 2019:5)—are fundamental units of language. Indeed, language is not produced as an arbitrary stream of words and sentences; it constructs a meaningful whole—a cohesive unit of discourse that functions to achieve some social and communicative purpose. Halliday and Hasan (1976:23), for example, define a text as “coherent with regard to the context of situation [. . .]; coherent with regard to itself, and therefore cohesive.” For De Beaugrande (1980:18), a text is a “communicative occurrence” characterized by cohesion, coherence, intentionality, and situationality. For Egbert and Schnur (2018:162-163), a text is a recognizably self-contained (i.e., internally coherent) and functional unit of language. It is not accidental that a text is a fundamental unit of analysis in text-linguistics, and for many scholars, at the heart of the study of linguistic variation has been studying variation among texts.
Remarkable strides toward this goal have been made in register-functional text-linguistics (Biber 1988). In text-linguistic register research, linguistic variation among texts has traditionally been attributed to differences among registers—culturally recognized varieties associated with the situation of use (Biber, Egbert & Keller 2020:583). That is, registers (e.g., research papers, news reports, conversations) differ in their situation of use, typically characterized by configurations of situational factors, such as participant relationships and relative status, degree of interactivity, amount and nature of shared knowledge, setting, mode, production and comprehension circumstances, general and specific communicative purposes, and topic. These situational characteristics in turn explain the language of a register: The prevalence of particular linguistic resources in a given register is not random; rather, linguistic features are common (or uncommon) in a register in response to its situation and perform functions necessitated by that situation (Biber & Conrad 2019). A typical procedure in text-linguistic studies of register differences has consisted of three steps: (a) identifying the situation of the target registers; (b) recording the linguistic rates of occurrence for each text and computing measures of central tendency of these linguistic counts for the registers; (c) interpreting the linguistic results in functional terms with respect to the situation of the registers under analysis.
The claim underlying this focus on register categories as a predictor of linguistic variation among texts has been that differences between registers are greater than differences among individual texts within them. In fact, perhaps the strongest statement about the role of register in explaining variation among texts is that general accounts of language without attention to register are meaningless (Biber 2012; Seoane & Biber 2021:3), as the differences between situational varieties are so extreme that no such general account of language use can be true. Over the past decades, this research paradigm has produced a substantial body of work in support of such claims (see Barbieri & Wizner 2019 for an overview). Yet, recent findings have caused an important shift in this perspective toward an increased focus on variation among texts left unexplained by their register membership and attempts to account for this variation.
1.2. A Conceptual Shift: New Levels of Analysis
In traditional register research, texts of a single register have been considered unified by their situational characteristics. Situation has therefore been defined at the level of the register: All magazine articles have been generally assumed to share communicative purposes and audiences; all academic lectures tend to share settings, participant relationships, and modes by virtue of the fact that they belong to that register. As the situation of a register determines the language of that register (the key tenet of register research), it should follow that all magazine articles should also share linguistic characteristics, and the language of all academic lectures should generally be the same. However, it is far from uncommon for register studies to observe linguistic variation among texts of a single register. In fact, even highly conventionalized registers, such as academic writing, exhibit substantial linguistic variation among texts, documented even by early register studies (e.g., Biber 1988:170-198). Until recently, however, the question of the basis for this linguistic variation among register texts was not raised in register research, whose focus had been predominantly on situational and linguistic differences between register categories.
The question of unexplained linguistic variance within registers was brought to light when a number of studies conclusively pointed not only to linguistic but extensive situational (or communicative) variation among register texts. In a study of disciplinary variation in academic journals, Gray (2015) identified numerous important communicative contrasts among publications that extended beyond discipline, such as empirical and non-empirical; quantitative experimental, qualitative, and mixed methods studies; theoretical and evaluative articles, author interpretation and general theoretical articles, reviews, commentary/forums, and syntheses. Journal articles thus showed unusual communicative diversity.
A prime example of such diversity is a large-scale investigation of the searchable web (Biber & Egbert 2018), in which many of the texts coded for specific situational factors (mode, interactivity, and purpose) were characterized by recurring combinations of communicative purposes rather than a clear purpose for each text. As a result of this communicative variability, these texts were classified into “hybrids” of how to/opinion, opinion/informational description, narrative/opinion, narrative/informational description, and others.
This apparent situational complexity led Biber, Egbert, and Keller (2020) to approach the situational characteristics of online texts continuously and code texts for the extent to which they can be defined through particular situational parameters. That study showed that while some registers may be more situationally uniform (song lyrics, encyclopedia articles), the texts of others, such as opinion blogs or interactive discussion forums, are highly situationally diverse. The study then identified situational text types—groups of texts unified in situational characteristics regardless of their register—and showed that some especially varied registers (e.g., opinion blogs) can include multiple situational text types.
Communicative complexity is not unique to online language. In a study of conversation, Biber, Egbert, Keller, and Wizner (2021:22) segmented conversations of the British National Corpus into “sequences of coherent discourse” on the basis of purpose shifts that they systematically observed within each recording. The study then identified conversational text types united by the same purpose. Again, these findings foreground extreme situational versatility within the scope of a register and even within a single text.
Taken together, these studies have shown that registers are, in fact, not situationally uniform, and texts within the same register vary, at times widely, in their communicative characteristics. Integrating this new evidence into the state of the art, Biber and Egbert (2023) posit that since
(a) functional correspondence between situation and language has been established at the level of register categories (i.e., situational differences between registers correspond to linguistic differences between them);
(b) texts within registers can show extensive linguistic variation; and
(c) texts within registers can also exhibit communicative variation,
then linguistic variation among texts within registers is also not random and should be expected to have the same functional basis as linguistic variation between registers. Namely, linguistic variation among texts of a register should correspond to the communicative differences among its texts.
Reconceptualization of registers as situationally (as well as linguistically) varied and the idea of functional correspondence at the level of the text within registers marked a new direction in the text-linguistic enterprise of explaining linguistic variation among texts, making it clear that while register categories may be important predictors of textual variation, register differences do not capture its full range.
1.3. Functional Correspondence Within Registers
Building on this work, Egbert and Gracheva (2023) examined granular communicative distinctions among texts, such as discourse types in fiction (dialogue or narration), identified by Egbert and Mahlberg (2020); communicative purpose distinctions in university textbooks; and speech types in political speeches. Their study consistently demonstrated quantitative gains in explained linguistic variance as these granular communicative predictors were included in the model. Egbert and Gracheva’s results thus suggested that there should be a link between the observed linguistic gains and the granular situational factors that they accounted for, and attention to such specific communicative predictors within registers can reveal new functionally interpretable linguistic patterns.
Two large-scale investigations—of legal statutes (Wood 2023) and undergraduate student essays (Goulart 2024)—concerned themselves with linguistic descriptions of texts within their respective registers on the basis of communicative differences among texts. Wood (2023) observed a wide range of communicative variation among statutes and, by grouping texts into communicative text types on the basis of common communicative purposes, identified procedural guidelines, operational definitions, criminal offenses, permissions, prohibitions, duties, and impersonal rules. She then examined the lexico-grammatical features characteristic for each type. Similarly, Goulart (2024) observed variation in purpose in undergraduate student essays, coded essays for communicative purpose, and also undertook the text type approach, grouping essays coherent in purpose and examining the linguistic characteristics of the emergent communicative text types. These studies thus took important steps toward acknowledging the situational variability among the texts of these registers and providing linguistic accounts of the identified communicative text types rather than defining the situation of the registers in their entirety and producing an average for the linguistic characteristics of their texts.
1.4. Motivation for the Present Study
Despite the several first forays into the study of intra-register variation and the evidence pointing to the existence of functional links between the communicative variation among texts and their linguistic characteristics, these links remain a nascent area of research in text-linguistics, with studies tending to focus on a single register of interest, typically a particular specialized domain (undergraduate academic writing; legal statutes). Further, communicative purpose is so far overwhelmingly the most researched variable situational factor (see Sections 1.2-1.3), which raises the questions whether other such factors exist, whether they systematically predict language, and whether and how these communicative predictors differ from one register to another.
These questions remain in part because no study has yet begun with the observed patterns of linguistic variation among texts and proceeded to identify the communicative basis for that linguistic variation—thus asking the question what explains the language differences that exist among texts rather than coding the texts for a pre-determined communicative characteristic and then turning to the corresponding linguistic patterns—the approach adopted by previous work (Wood 2023; Goulart 2024). While the insights from those studies are critical, the methodological choice to code texts for an a priori selected communicative factor inevitably precludes the possibility that other factors may play a part in explaining linguistic variation.
To address these gaps, this study examines linguistic variation among texts of four distinct registers. Specifically, the study groups texts of each register by how widely they vary linguistically, focusing on texts that exhibit extreme linguistic differences from the registers’ central trends—peripheral texts, described in detail in the next sections. The study then performs systematic text analysis of these linguistic groupings to identify the possible communicative factors that may account for the linguistic variation among texts (see the Method section). The overarching goal of this bottom-up approach is to test the notion of functional correspondence between linguistic and communicative characteristics of texts in several discrete registers, addressing the questions: What accounts for the linguistic variation among texts within the registers of the study? Does this linguistic variation have a functional basis?
2. Method
2.1. Corpus
The study uses a subset of the Corpus of 100 Idiolects (Heini, Kredens & Pezik 2021), which includes seven undergraduate student registers. Three of them—essays, emails, and text messages—are naturally occurring, and four are elicited: evaluations (written responses on students’ likes and dislikes of university facilities), image descriptions (speech-to-text descriptions of three visuals), business memos (written responses to a scenario in which the students acting as personal assistants develop an itinerary for their CEO’s business trip), and oral interviews (see details about the corpus in Appendix 1). The benefit of including elicited and naturally occurring registers consisted in their different levels of expected situational constraints: As elicited registers are heavily constrained by their task, it is natural to expect less communicative variation in such registers and more in the naturally occurring ones. Their juxtaposition was thus informative to the goal of testing functional correspondence in their texts. Importantly, the study does not generalize the findings observed in elicited registers to the naturally occurring ones, but rather finds patterns characteristic for each.
2.2. Functional Linguistic Variation: Multidimensional Analysis
The study employs multidimensional analysis—a macroscopic analytical approach (Biber 1985), which identifies dimensions of functional linguistic variation among texts. At the core of these dimensions are patterns of linguistic feature co-occurrence in texts. Proposing the method, Biber argued that although microscopic analyses are fundamental for understanding the functions of individual features, which contribute to the complex patterns of the dimensions, such individual functions are too narrow to reflect the “overall parameters of linguistic variation within a set of texts” (Biber 1985:339). Biber claimed that the goal of capturing linguistic variation in a discourse domain is best served by a macro approach, as there is not an infinite number of specific communicative functions of individual features but a limited number of general, underlying communicative functions shared by frequent and pervasive features co-occurring in discourse and jointly contributing to its construction. Crucially, feature co-occurrence is never random but always functional and, therefore, interpretable (Biber 1988). Such underlying patterns of variation are “conceptually clearer than the many linguistic measures considered individually” (Biber 1985:340).
At the core of multidimensional analysis is exploratory factor analysis—a multivariate statistical technique of dimension reduction, reducing a large pool of linguistic features to a small number of latent variables (dimensions) based on feature co-occurrence. Dimension poles are comprised of sets of features in complementary distribution—that is, when features of the positive pole are common in texts, the features of the negative pole are relatively infrequent, and vice versa. Supplemental File 1 online contains a complete list of features initially included in factor analysis, and Supplemental File 2 online presents the four identified dimensions of linguistic variation and their linguistic structures—co-occurring features that comprise each pole: Oral Elaboration versus Information Density, Abstract versus Concrete Discourse, Others-Oriented Descriptive versus Self-Oriented Interactive Discourse, and Evidence-Based Stance (see Gracheva [2023:59-144] for a detailed account of the factor analysis and a description of each dimension). For space reasons, this paper illustrates all the intended analyses on the first dimension—Oral Elaboration versus Information Density—a cline from oral and elaborated texts to informational ones. This dimension, often referred to as the “oral/literate divide,” comprised of predominantly clausal features representing the fragmented, elaborated, and involved manner of oral discourse on the one hand and the predominantly nominal features of informational literate discourse on the other hand, has consistently emerged from multidimensional studies of numerous discourse domains across languages and has been named one of the “universal” dimensions of variation (Biber 2004). This dimension was therefore considered particularly interesting for examining intra-register variation. Figure 1 plots the registers in the corpus on the dimension and illustrates the register differences. Register explains an impressive 76 percent (R2 = .76) of linguistic variance among texts on the dimension (Supplemental File 2 online contains descriptive statistics for each register). Yet, Figure 1 (and the SDs in in Supplemental File 2) also show that even in this scenario of an extremely high importance of register, each category still reveals internal variation. The next section details the method for analyzing linguistic variation across each register’s texts.
Registers on Dimension 1: Online Oral Elaboration Versus Information Density
2.3. Intra-Register Linguistic Variation on Dimensions: Central and Peripheral Texts
To target variation within a register, this study develops a method that allows a contrastive analysis of texts that represent a register’s central tendency (hereafter, central texts) and texts that diverge from this tendency (hereafter, peripheral texts). This approach draws on prototype theory (Rosch 1976), according to which a category is structured around its central or most prototypical exemplars, while the other members are characterized by “graded membership [. . .] based on similarity to the prototype” (see Keller 2025 for a discussion of categorization with relation to register). Prototype theory is invoked here to examine register-internal variation systematically, viewing texts as specific instantiations of a register category and determining the degree of their prototypicality—their place in a category relative to the prototype, operationalized in this study as the register mean (Mr). Texts peripheral to a category are then indicators of significant intra-register variability. These texts and the basis for their peripheral status in their categories are the primary focus of this study. Importantly, because the mean captures statistical centrality but not necessarily cognitive salience, it is not implied here that central texts represent socially recognized prototypes. However, registers have traditionally been described in terms of their average characteristics and mean differences, and that motivated the choice of the mean as the reference point in these analyses.
Biber, Egbert, and Keller (2020) identified central and peripheral texts in a study of situational text types: texts that are central exemplars of their text types and texts not quite representative of their situational characteristics. This study applies these concepts to linguistic variation within a register and examines the position of each text in a register with relation to the mean of that register. Central texts are considered representative of the register central tendency. Peripheral texts, on the other hand, demonstrate deviation from the register trend. Importantly, in line with the idea of graded category membership, the central or peripheral position of a text in a register is not a dichotomy but a spectrum, where texts can be more or less removed from Mr. To place texts into central or peripheral groups, the following steps were implemented in each register (illustrated in Figure 2 in the register of emails).
Illustration of Central and Peripheral Emails Above and Below Mr
2.3.1. Measures of Distance
The study computes a measure of distance between each individual text score on the dimension and Mr (Distance = Individual text score − Mr). The assignment of texts into a particular prototypicality band was operationalized through the distance measures shown in Table 1:
Text Prototypicality: Operationalization
Supplemental File 3 online contains individual text scores, Mr, and the distances between all the individual texts and Mr in each register.
2.3.2. Direction of Deviation
In the case of Moderately Removed and the two Peripheral groups, the study identifies the direction of deviation relative to the central tendency—texts above or below Mr. Direction of deviation characterizes each text in terms of its tendency toward one or the other functional pole of the dimension. That is, if a text is noticeably higher than Mr (e.g., in a Peripheral group above Mr), this text shows a preference for features of oral elaboration. Conversely, if a text is in a Peripheral or Moderately Removed group below Mr, this text shows an inclination toward information density. Importantly, the position of a text on the dimension should be interpreted in relative terms. That is, a text’s position above Mr does not necessarily correspond to the positive range of the dimension (e.g., Mr itself may be quite low in the negative range, and a text above Mr may thus remain in the negative range). Rather, it shows this text’s tendency toward the positive pole relative to Mr. Similarly, a text’s position below Mr does not necessarily correspond to the negative pole (Mr may be positive and a text below Mr may also be in the positive range) but shows a tendency to informational discourse.
Figure 2 illustrates this approach in emails (where M = 0). The central tendency of the register says that overall emails combine the characteristics of oral elaboration and information density. However, from the box plot area of Figure 2, where several individual text scores are plotted, we see that some of these texts are more informational (below Mr) and others are more elaborated (above Mr), and the extent to which these text scores are removed from Mr varies. On the left are the distance groups into which these texts fall based on how far they are removed from the central tendency, and the arrow shows the direction of deviation.
Table 2 contains the number of texts in each register that fall in each of the groups above and below Mr. 1
Text Prototypicality Across Registers: Number of Texts per Distance Group and Direction of Deviation Within Each Register
Note: Bold indicates largest deviations in Moderately Removed and Peripheral groups.
It becomes apparent from Table 2 that (a) register texts vary in their prototypicality, and (b) registers allow quite different degrees of linguistic variability: Business memos, for instance, despite the highly specific task (providing recommendations on specified items), allow the most linguistic freedom, with the largest number of peripheral texts. While business memos are a generally informational register (M = −1.01; SD = 0.79), deviations from Mr are equally likely toward both increased information density (below Mr) and increased oral elaboration (above Mr).
Following business memos are image descriptions, emails, and essays—all registers with several peripheral texts. Image descriptions, generally an oral and elaborated register (M = 0.44; SD = 0.51), also allow deviation in both directions; in essays (M = −0.24; SD = 0.27) and emails (M = 0; SD = 0.38), the texts that show extreme deviation (peripheral groups) mostly show a tendency to informational discourse, although texts that are moderately removed from Mr do gravitate toward oral elaboration.
In contrast to the other registers, most of the texts in evaluations (M = 0.15; SD = 0.24) and interviews (M = 1.62; SD = 0.21) generally conform to the central tendency (Central and Minimally Removed groups); for this reason, these registers are excluded from the following analyses.
2.3.3. Qualitative Text Analysis
Having grouped text of each register on the basis of linguistic deviation and its direction, the study conducts qualitative analysis of texts to identify the basis for the observed patterns of deviation. As shown in Figure 2 by curly brackets, three groups of texts are of interest: central exemplars of a register, peripheral texts below Mr (i.e., tendency toward increased information density), and peripheral texts above Mr (i.e., tendency toward increased oral elaboration). Moderately Removed texts were examined in cases with few extreme deviations in some registers (e.g., essays) or to confirm the trend found in the Peripheral groups.
This final analytical step is directly informed by the notion of functional correspondence: If functional correspondence between communicative situation and language exists at the level of the text, the linguistic variation among texts should be attributable to communicative differences among these texts. To test this, the study undertakes a situational analysis of central and peripheral texts above and below Mr.
2.4. Applying the Framework for Situational Analysis
The study draws on the framework for situational analysis (Biber 1994:40-41; Biber & Conrad 2019:40) and analyzes the texts in the identified linguistic groups in terms of the following situational factors: the addressee (singular, plural, unenumerated), relationship between the participants (personal, professional, academic), their relative status (lower => higher or equal), degree of interactiveness (low, medium, high), degree of shared knowledge between participants (low, medium, high) and its nature (specialist or general), domain (school, work, personal), setting (public or private), whether the time and space are shared by participants, mode (written or spoken), specific medium (permanent, such as printed or recorded and transcribed, or transient, such as face-to-face real-time communication), production and comprehension circumstances (immediate online communication or offline with opportunities for editing and revision), general and specific purpose and topic (values of these are identified for each text). Table 3 summarizes the application of this framework to the registers of the corpus.
Application of the Framework for Situational Register Analysis (Biber 1994; Biber & Conrad 2019) to the Corpus of the Study
Note: Values of the situational factors were identified for the texts of the registers in this study (naturally occurring business memos or other types of interviews, for example, will differ in their situational characteristics from the ones in this corpus).
An important pattern that emerges from situational text analysis is the fact that the values of some situational parameters are the same for all texts of a register, while the values of other parameters vary across texts (marked with “–” in Table 3). For example, all essays are written texts. Mode is therefore held constant for all essays. Similarly, all emails in this corpus are interactive, so interactiveness is held constant across emails. In other cases, however, it is not possible to say that all texts of a register are the same with regard to a particular parameter. For example, essays and emails set different communicative goals and address a range of topics. The parameters of purpose and topic are variable rather than constant for these registers.
If we assume that linguistic variation among texts corresponds to situational variation, the variable slots in Table 3—the situational parameters that vary across texts of a register—provide a natural starting point in the search for reasons for linguistic variation, as variation in those situational parameters may correspond to linguistic variation across the texts of these registers. For example, linguistic variation within essays and emails may be due to purpose and/or topic variation among their texts. For this reason, the study begins its analysis with these variable situational parameters and systematically assigns topic and purpose labels to each essay and each email in the linguistic groups of interest (i.e., Central, Moderately Removed, Peripheral 1, Peripheral 2).
The pattern is less straightforward in image descriptions and business memos. According to Table 3, these registers are highly situationally well-defined, with no variable situational parameters. This uniformity may be intuitive in view of the narrow nature of the tasks—to describe the visuals and to give recommendations in response to a scenario. The outcome that is surprising, though, is the wide linguistic variation in these registers despite this situational uniformity, with peripheral texts above and below Mr (Table 2). A possible explanation may come from the nature of the framework. The framework for situational analysis applied here has been especially useful in explaining high-level between-register differences. However, it does not present an exhaustive set of situational parameters that determine language use, and it is likely that additional factors, not accounted for by the current framework, may be at play when textual variation within registers is examined. The aim of text analysis of central and peripheral business memos and image descriptions above and below Mr was therefore to identify such additional situational factors if they exist.
3. Results and Discussion
3.1. Variation Within Essays: Effect of Discipline and Personal Versus Technical Topics
As essays were examined for topic and purpose, a discipline label was additionally assigned, and it was discipline and the specific topics within disciplines that proved the most informative. Appendix 2 presents essays in each distance group according to discipline (psychology, sociology, law, business, and marketing) and the topics listed under each discipline.
Essays in the central tendency group do not reveal any notable situational trends. These linguistically central texts come from a variety of topics and a range of disciplines, and the same disciplines and topics seem to occur on either side of Mr. Example (1) is a central essay in sociology on trends in liberalism (features of elaboration (1) Central essay, Sociology, Dimension 1 score: −0.28 [Marx]
The passage makes use of several nominal sequences (underlined), which denote concepts in politics and social studies. On the other hand, features of elaboration, mainly the present tense and demonstrative pronouns (bolded), state the beliefs of prominent thinkers and refer to these beliefs in the following discussion. This essay (score of −0.28) reflects the general trend of the register (M = −0.24; SD = 0.27), showing a balanced need for information presentation and problematizing a topic, discussing theories, and analyzing their interrelationship, the latter achieved through oral elaboration.
Deviation from the central tendency in both directions reveals a stark contrast between the disciplines of sociology and psychology. Appendix 2 shows that ten of the twelve essays above Mr are sociology essays, requiring considerably more features of oral elaboration, while all the texts below Mr (n = 11) are topics in psychology, which prioritize informational features and are markedly less oral and elaborated.
Similar contrasts have been observed by research on disciplinary variation. Discipline has been studied extensively as a predictor of linguistic variation within university textbooks (Egbert 2015), journal articles (Gray 2015), and university student writing (Hardy & Römer 2013; Gardner, Nesi & Biber 2019). These studies used dimensions of variation either identified by Biber (1988) or by their own analyses as linguistic variables. In Gardner, Nesi, and Biber’s (2019) study, for instance, Dimension 1, Compressed Procedural Information versus Stance toward the Work of Others, resembles Dimension 1 of this study and the typically identified oral/literate divide. The Compressed Procedural Information end is comprised of nominal features of information density, pre-modifying nouns (also conveying informational focus in this study) being the feature most strongly associated with the dimension; Stance toward the Work of Others contains verbs and adverbial features, which also contribute to oral elaboration on Dimension 1 of this study. Gardner, Nesi, and Biber (2019) find that Life Sciences, which include psychology, occur higher on the informational end, while Social Sciences, which include sociology, gravitate toward the other pole. The relative position of the disciplines on the dimension thus reflects the same trend observed in this study—psychology incorporating more informational features and sociology calling for increased oral elaboration. However, Gardner, Nesi, and Biber do not examine the texts of these particular disciplines to account for these differences.
The analysis of the topics within each discipline in this study showed that the distinction between essays far above and below Mr appears to represent more and less personal topics, or rather, cohesive groups of topics or “topic types.” Topics related to daily experience, human qualities, skills, and behavior are predominantly found in sociology, constitute the background knowledge of a general audience, and lend themselves to a discussion: gender and race dynamics, leadership qualities, stress and anxiety, ethics and morality, effects of social media on self-perception, self-esteem, and societal norms. This personal nature of the topics lends itself to much higher rates of occurrence of features of elaboration than in central texts. Example (2), an essay on successful leadership, illustrates this trend (features of elaboration bolded).
(2) Sociology essay above Mr, Dimension 1 score: 0.28
Demonstrative pronouns are used for discourse organization as the author refers to aspects of personality development and relates them to leadership. Present tense is used to make general statements, and adverbs qualify propositions, organize ideas, and convey critical detail such as how leadership skills are formed and what comprises them. While the text does use informational features to refer to complex concepts (
In stark contrast with this, psychology topics are heavily focused on the scientific enterprise. Implementation of the visual attention task, anticipatory attention and attentional selection and suppression, Alpha Inhibition Hypothesis, or empirical testing of the link between neuroticism and eating habits are topics that involve descriptions of cognitive phenomena, such as brain regions and brain activity, steps of an experiment, and interpretations of findings. It is this scientific, technical focus that is associated with increased information density. Example (3), reporting a research procedure, presents the hypothesis tested (
(3) Psychology essay below Mr, Dimension 1 score: −0.87 The
This effect of topic in essays may perhaps also be considered in light of the degree of empirical focus in a discipline. While this question is not addressed here directly, the data of this study appear to suggest that the topics commonly addressed by particular disciplines lend themselves to different degrees of empirical focus, and that in turn leads to a different linguistic composition of empirical versus nonempirical texts.
The effect of topic extends beyond psychology and sociology, as essays from other disciplines confirm this trend of personal judgments, interpretations, and views calling for oral elaboration and scientific enterprise being associated with increased information density (see Gracheva [2023] for analyses of texts from Marketing and Business and Law on the topics listed in Appendix 2 and for a larger variety of topics in each discipline illustrating these conclusions). Further support for this interpretation comes from Egbert and Gracheva’s (2023) study, which examines geology and psychology texts in introductory university textbooks and observes a rather strong effect of discipline (R2 = .28) on Biber’s (1988) Dimension 1 (Involved versus Informational Production), whose structure mirrors the dimension of this study. Although both disciplines are in the informational range, psychology in that study (M = −8.22; SD = 6.94) is closer to oral involved production, while geology (M = −13.99; SD = 5.82) is much more informational. This pattern may at first appear contradictory to the results presented here; however, the relative position of the disciplines on the dimension confirms the conclusions regarding the effect of topics: While the psychology essays in this corpus are technical and focus solely on research procedures or specialist terminology, introductory psychology textbooks deal with non-technical topics (personality, perception, learning, motivation, and stress) and explain each through references to familiar experiences, relatable examples, and personal addresses that invite the reader to experiment. Expectedly, the personal nature of these texts results in elaboration. Such references are impossible in geology texts on the topics of plate tectonics, rivers, glaciers, metamorphic rocks, and radiometric dating. As a result, these impersonal texts are much more informationally dense.
This suggests that the observed trend should not be bound to a particular discipline, but rather could be found within other disciplines (as in psychology and geology in Egbert & Gracheva 2023). Importantly, it also becomes clear that a certain discipline by itself (e.g., psychology) should not be by default associated with oral elaboration or information density, as this functional preference varies based on specific situational characteristics (such as topics) of its texts in a given corpus.
It is illustrated further in the paper that the pattern of topic determining the balance between oral elaborated and informational features extends beyond academic discourse and therefore beyond disciplinary distinctions.
3.2. Variation Within Emails: Personal Topics Versus Official Matters and Class Content
In emails, again topic revealed the most informative and interpretable patterns. Appendix 3 lists the email topics in each analyzed distance group and the identified topic types: personal topics, official matters, and class content-related topics. There is a consistent pattern of personal topics requiring increased oral elaboration (nineteen out of the twenty-eight emails above Mr) and official matters and topics related to class content clearly prioritizing information density (twenty-one—fourteen and seven, respectively—out of the twenty-four emails below Mr) as follows from the number of emails in each group. Central emails exemplify the intermediate position the register occupies on the dimension (M = 0; SD = 0.38) and a balance of positive and negative features.
Example (4)—an email to a landlord—is a central text. The most salient features of elaboration include the present tense, used to report maintenance issues, conjuncts serving for discourse organization, adverbs of manner describing the malfunction, and causative adverbial clauses providing reasons for them, among others (demonstrative pronouns, the pronoun ‘it,’ and contractions). The author’s involvement is apparent through emphatic stance adverbs (all positive features bolded). However, informational features also serve important functions, showing partative relations in object references (
(4) Central email, Dimension score: 0.07
Example (5), a peripheral text above Mr, illustrates a topic personal for the author, who gives advice to a fellow student, reflecting on their experience with placement, training, and coursework. The email is replete with adverbs of stance, especially emphatics. Hedges are used to make suggestions or mitigate subjective judgments. Clausal coordination and causative adverbial clauses add detail and justify opinions. Features of inexplicit discourse—the pronoun ‘it’ and contractions—suggest that the email approximates oral communication, where background knowledge shared by the interactants allows lack of explicitness (positive features bolded).
(5) Peripheral 1, email above Mr, personal, Dimension score: 0.70 Hi,
In contrast, emails on official matters are concerned with university policies, attendance, enrollment, exam deferral, or job placement. In these emails, informational features comprise the very reason for writing and denote the objects or entities that constitute the official matter (the (6) Peripheral 1, email below Mr, official, Dimension score: −0.77 Dear Sir/Madam, I lost my
Neither are personal matters the subject of class content-related emails, which integrate topics from students’ classes: fashion and gender, origins of Liberalism, or marketing strategies in addressing consumer needs. Similar to the emails on official matters, content-related emails gravitate toward information density. However, the nominal sequences in these texts perform a different function—they refer to concepts borrowed from specialist fields and brought into email communication. A unique feature of class content-related emails is thus that the nominal features in them are the result of a transfer of the topic from a different register, normally uncharacteristic for emails. The email in Example (7), for instance, requests assistance with an essay outline on gender and social class in fashion. The informational features (underlined) directly reflect the topic of the essay; while these informational features may be expected in the register whose topic they reflect (essays, dissertations, etc.), they are uncommon for emails, which drives them to the periphery of the register. In this sense, class content-related emails are linguistic and situational hybrids, on the one hand representing email communication (typically with the professor, whose class the content is from), but on the other hand substituting the value of a situational parameter (bringing in the topic of an essay, dissertation, etc.) and integrating the language of that register.
(7) Peripheral 1, email below Mr, class content, Dimension score: −0.79 Hi,
The pattern observed here may seem reminiscent of the trend in essays: While the topics of emails are naturally different, it appears that the grouping of these topics has the same general basis: the distinction between more or less personal topics. Crucially, as in the case of essays, the key take-away from this analysis is the interpretability of the linguistic variation among texts through variation in their specific communicative characteristics.
3.3. Variation Within Image Descriptions: Content of the Visual Versus Subjective Commentary
In image descriptions variation is surprising. Situational analysis shows that the values of each situational parameter are defined for all texts of the register (Table 3). This specific task is additionally constrained by the content of the visuals, which constitutes the topic of all texts. Yet, despite these constraints, linguistically peripheral descriptions are not at all uncommon (Table 2).
Text analysis of the peripheral texts on either side of Mr revealed variation in a highly specific communicative parameter, unaccounted for by the framework—specifically, the extent to which the descriptions rely on the content of the visual versus the content being enhanced by subjective commentary. A central image description in Example (8) shows that both a concrete description tied to the content of the visual and subjective commentary fulfill its overall goal. Oral elaboration reflects authors’ judgments expressed through the present tense (here of numerous stance verbs) and emphatic adverbs (features of elaboration bolded). Nominal features, overall relatively infrequent (the register as a whole gravitates toward elaboration: M = 0.44; SD = 0.51), convey concrete detail (underlined).
(8) Central image description, Dimension 1 score: 0.50
Peripheral descriptions below Mr state only what is depicted—the ambience, interior design, people’s appearance, and objects immediately present in the visual—but do not offer interpretation or add to the content in any other way. Their strict focus on the visual results in noticeably lower rates of occurrence of features of elaboration. At the same time, the high level of visual detail leads to an increased informational focus, as illustrated in Example (9) (informational features underlined). Some elaboration is present; however, a closer look at the few oral features in the text shows that they still refer to what is immediately obvious from the visual. These adverbial features may even seem incongruous with the content of the proposition, as there is no apparent reason for hedging or emphasis with regard to a concrete physical object ( (9) Peripheral 2, image description below Mr, Dimension score: −0.24 I can see two females,
In contrast, texts above Mr enhance the content through subjective interpretations of the relationship between the people in the visual, the tasks they are engaged in, or reasons for their behavior. Such commentary naturally calls for increased oral elaboration. The detail observed in the peripheral descriptions below Mr, on the other hand, is not the focus of these texts, more concerned with analysis than the accuracy of the description. Example (10) is minimally focused on the image and only references the visual to anchor the narrative constructed around the characters (e.g., the laptop in the discussion of the characters’ reactions): their engagement in the conversation, facial expressions, and the nature of their work. Prominent features of elaboration include the present tense, causative adverbial clauses, clausal coordination, the pronoun ‘it,’ contractions, and adverbs. The functions of adverbs, specifically hedging (frequent use of (10) Peripheral 2, image description above Mr, Dimension score: 1.48 The guy
In essence, the additional parameter that accounts for variation in image descriptions—the focus on the content of the visual as opposed to the additional content incorporated through subjective interpretation—may be construed as variation in topic on a highly specific level. Regardless of the label for this communicative distinction, the analysis again confirms that linguistic variation across texts is explained through a shift in a specific situational characteristic, the communicative choice to focus on the visual or interpret it, integrating a personal perspective.
3.4. Variation Within Business Memos: Author Involvement and Addressee Focus
While business memos generally may vary in topic, addressee, and setting, the business memos in this corpus are highly constrained, which is reflected in the values of all situational parameters being specified for all texts (Table 3). Thus, as in the case of image descriptions, the peripheral business memos were examined in an attempt to identify the possible additional factor(s) corresponding to the surprisingly large linguistic differences across texts.
Central business memos provide information about the accommodation, cafés, museums, souvenir shops, and entertainment selected for the trip. These references inevitably require the use of nominal sequences, which are present in all texts and determine the overall informational focus of the register (M = −1.01; SD = 0.79). However, texts may also provide some detail about the selected options, which results in some features of oral elaboration. Example (11) illustrates this general register trend:
(11) Central business memo, Dimension score: −1.04 This hotel
Despite this general informational focus, the texts far below Mr present a rather striking contrast. Such texts adopt a minimalist approach to the task, which results in lists of selections, comprised of numerous nominal sequences denoting locations and activities, and no or minimal commentary. In fact, in many such memos, nominal sequences are the only linguistic feature constructing the texts. Example (12) illustrates this approach, with sparse comments from the author. While informational features were the driving force of the central text, by giving an even higher priority to information presentation, this text reveals an almost complete lack of author involvement.
(12) Peripheral 2, business memo below Mr, Dimension score: −2.00 2
In contrast, peripheral texts above Mr appear to reflect greater author involvement and the communicative choice to justify and relate the proposed options to the addressee’s needs and preferences. Example (13) illustrates this trend and the corresponding frequent use of features of elaboration.
(13) Peripheral 2, business memo above Mr, Dimension score: −0.01
Thus, variation in business memos too is explained by variation in a specific communicative factor, and, as in the case of image descriptions, different communicative strategies or approaches to a highly constrained task result in linguistic differences on the dimension.
4. Conclusions, Limitations, and Future Directions
The study carried out a close analysis of linguistic variation among texts within registers, with the goal of systematically identifying the basis for this, often extensive, variation. More specifically, the study examined linguistically peripheral texts relative to the register central trend and tested the existence of functional links between linguistic and communicative characteristics of these texts in each register. The contrastive analysis of central and peripheral texts showed that, in every case, the linguistically peripheral status of texts was linked to a change in the value of specific communicative parameters. It was thus shown that (1) register texts vary in their prototypicality; (2) importantly, in every register, intra-register linguistic variation is functionally motivated. Unlike previous work on communicative variation among texts, this study did not select potentially informative communicative characteristics a priori, but adopted the opposite approach to allow for the possibility of other situational factors explaining linguistic differences and, in some registers, identified variation in highly specific communicative characteristics, not accounted for by the existing situational framework. The observed functional patterns have been traced in several registers, have been shown to vary depending on the register, and even within situationally narrow, seemingly well-defined registers (in this dataset, business memos and image descriptions), linguistic variation corresponds to communicative differences among texts. These consistent links comprise the main outcome of this study—evidence of a systematic functional relationship between communicative considerations and the language that fits these communicative needs as language users produce texts.
Such new functional patterns, particularly the focus on peripheral texts, raise intriguing questions regarding the nature and extent of situational constraints imposed by registers and the existence of a certain reciprocal relationship between these situational constraints on the one hand and communicative freedom possible within registers on the other. That is, on the one hand, the peripheral texts in the corpus belong to their respective registers (at least in the present corpus design), as they (a) were produced in response to the task set by corpus compilers in a given register, and (b) thus represent these language users’ cultural understanding of that register. On the other hand, the linguistic and communicative deviations of these texts from many other texts in their registers show that the situation of use constrains language users only to an extent. The existence of peripheral texts suggests that registers are dynamic and flexible categories, allowing new/atypical members and redefining themselves.
The functional links between communicative situation and language at the level of the text reflect the recent conceptual shift in register research (Biber & Egbert 2023)—extending ways of explaining linguistic variation among texts from the focus on between-register differences to identifying new bases for linguistic variation. Importantly, this focus on register-internal variation does not contradict traditional register studies but builds on earlier developments. For example, the effect of sub-register has been long established in text-linguistics (e.g., Biber & Gray 2013), and sub-register differences correspond to shifts in situational characteristics. However, sub-register accounts for register-internal variation only in part, and, as in the case of this study, there are registers that do not have clear recognized sub-register categories (Biber 1988:171). This study, alongside other work that contributed to this emerging state of the art (Egbert & Gracheva 2023; Wood 2023; Goulart 2024; Hanks 2025), shows that functional correspondence is in fact much more pervasive and can be traced at the level of individual texts.
The focus on textual variation thus complements traditional register research and offers researchers a choice of higher-level, more granular, or combined analyses. In a recent paper, Egbert, Biber, Keller, and Gracheva (2024) bring together register membership and situational characteristics of individual texts (in that study, text scores on situational dimensions of variation) as predictors of linguistic variation in several case studies on online texts and compare the amount of linguistic variance explained by these predictors. The results unquestionably show that register membership matters. However, the study then also shows that individual communicative characteristics of texts explain unique linguistic variance in addition to the variance explained by register alone. The findings of this study from several registers are in line with that outcome. As noted in Section 2.2, a large amount of linguistic variation among texts is indeed explained by register differences (76 percent). Yet, it is consistently shown that this is not the full extent of variation among texts. This focus on register-internal variation and on peripheral texts does not undermine the importance of register, but the findings of this study, alongside Egbert, Biber, Keller, and Gracheva (2024), call for a shift toward accounting for more fine-grained patterns of linguistic variation among texts and doing that through their specific communicative characteristics.
Explanations for functional correspondence and intra-register variation have been offered not only by sociolinguistic/text-linguistic (Biber & Egbert 2023) but also by the psycholinguistic account of register variation (Keller 2021, 2022, 2025). Language users form associative links between situational cues and corresponding co-occurring linguistic structure; these associative links are acquired through repeated exposure to similar situations and their language, that is, through statistical/associative learning in the process of language use. This results in a mental model of a register. Thus, language users have “mental representations of a situation” and “the relevant knowledge about what is and is not appropriate for the situation” and what is possible within its other constraints (e.g., online/offline production), based on which they select the linguistic resources for this situation (Keller 2021:52). However, atypical situational cues may occur alongside situational characteristics that are part of such a mental model of a register. When that happens, these atypical situational cues activate linguistic features associated with them, thus resulting in a linguistic deviation from the language commonly associated with the register. At the same time, the linguistic features that are commonly associated with the register may no longer be as relevant, which leads to their suppression in favor of the new features that are now relevant and functional for the new situation. Furthermore, situational cues from different registers may simultaneously activate linguistic features associated with several registers, which would explain the existence of hybrid texts (such as those identified by Biber & Egbert 2018, or by this study). It is this activation of linguistic features corresponding to new situational cues that can explain intra-register variation (Keller 2022).
While the main focus of most recent work has been the study of variation within registers, a related implication of functional correspondence concerns variation beyond register categories. The idea of texts not belonging to any culturally recognized register may appear odd in light of the traditional view that members of a culture converge on some shared idea of register categories (Biber & Conrad 2019) and perceive language production and processing through the lens of such categories (e.g., “novels, articles, editorials . . . when people say they can read or write a language, it is this kind of text that they usually have in mind” [Hoey 2001:13]). Yet, recent studies have shown that texts that “do not belong” to any recognized category are not uncommon (e.g., in Biber & Egbert’s 2018 study, coders were not able to categorize a vast number of online texts into any register). In a case study devoted to these texts without a register, Egbert, Biber, Keller, and Gracheva (2024) examined them for the existence of functional links between their situation and language and showed that variation in situational dimension scores corresponded to variation in linguistic dimension scores, providing evidence that the links between communicative and linguistic characteristics of texts can in fact be traced outside registers. An intriguing question for future research is whether (or to what extent) language users draw on multiple known registers when producing such communicatively complex texts or whether they rely solely on their interpretation of specific situations but not on register category knowledge.
The focus on textual variation has triggered important methodological work in text-linguistics on analytical frameworks and methods for systematically accounting for patterns of functional correspondence. Studies like Biber and Egbert (2018) and Egbert et al. (2021) laid the foundation for developing highly detailed situational taxonomies and continuous coding of texts for the extent to which a particular communicative factor characterizes them. A newer study builds on that work and proposes a taxonomy and a coding method for text-level situational analysis across registers, with the goal of comparing registers not for their central tendency as in the traditional approach, but for the extent of communicative and corresponding linguistic variation within their scope (Gracheva & Egbert 2026). As a culmination of this research program, Egbert (under review) introduces a unified text-linguistic framework for functional correspondence analysis, which achieves a comprehensive account of linguistic co-variation patterns, communicative co-variation, and linguistic and communicative covariance in texts through the use of canonical correlation analysis. The increasingly granular focus on text-level variation in all these studies promises to uncover new functionally interpretable patterns, result in highly detailed linguistic descriptions, and bring us closer to the text-linguistic goal of explaining linguistic variation among texts.
Important future directions also emerge from the limitations of this study. For example, its findings naturally create the need for further investigations of the communicative factors that influence language use. One such factor highlighted here and not identified before is the topic type—specifically, the distinction between personal and impersonal topics. Considering that topic has been previously associated with lexical, rather than grammatical, variation (Biber & Conrad 2019:48), the effect of topic type is worth exploring further. A related question is that of the interrelationship between situational characteristics. Communicative purpose, for example, has been previously shown to be a highly variable factor and an effective predictor of linguistic variation across texts. The fact that patterns of purpose variation did not emerge from this analysis should not mean that variation in purpose (surely present in the texts as shown by the situational analysis) does not correspond to linguistic variation. What is more likely and intuitive is that purpose and topic are related situational parameters. It is then worth examining this interrelationship of topic types and purposes in order to distinguish linguistic variation due to each parameter in the future (Keller & Gracheva In preparation).
For space reasons, the analyses were demonstrated only with regard to one linguistic pattern—Oral Elaboration versus Information Density. While in a larger work Gracheva (2023) corroborates the trend on another linguistic dimension—Evidence-Based Stance—a question for future research is that of the extent to which different communicative factors contribute to variation in linguistic dimensions (as in this study) versus variation in individual linguistic features (Gracheva & Egbert Forthcoming).
Future studies could adopt a different view on the central tendency and, relatedly, a more precise operationalization of a prototype. A limitation of the current method lies in the choice to rely on the register mean as the benchmark against which distance from the central tendency was measured. As the mean is not a concrete point of reference but an amalgam of all text scores, including peripheral texts, studies could choose other measures of central tendency, perhaps more representative of the register “norm.” Alternatively, the central tendency could be computed excluding the peripheral exemplars.
Finally, the study worked with naturally-occurring as well as elicited registers. The elicited registers of the corpus showed an impressive situational and linguistic range, and the findings are even more likely to be observed in naturally occurring registers, which can be anticipated to be even more varied. Still, as in any study, the findings should be tested further on other registers.
The future directions overviewed here are only a few of the new avenues of research opened by functional correspondence. This array of unanswered questions suggests that functional correspondence does not only contribute to linguistic theory. The possibility of unveiling new functional patterns of linguistic variation has pushed researchers to methodological innovations, such as continuous coding for situational characteristics and development of new frameworks, questioned and challenged the status quo in an established research field, and perhaps most importantly, enabled a new, targeted, and rigorous study of texts, highlighting their primacy in human discourse.
Supplemental Material
sj-docx-1-eng-10.1177_00754242261432215 – Supplemental material for Prototypicality of Register Texts: The Language and Communicative Situation of Central and Peripheral Exemplars
Supplemental material, sj-docx-1-eng-10.1177_00754242261432215 for Prototypicality of Register Texts: The Language and Communicative Situation of Central and Peripheral Exemplars by Marianna Gracheva in Journal of English Linguistics
Footnotes
Appendix
Email Topics Within Each Distance and Direction Group. Dimension 1: Oral Elaboration Versus Information Density
| Distance group | Above M | Below M |
|---|---|---|
| Central (<0.1) | 1. Personal topics—9 emails: • Maintenance/living conditions • Questions re. Master’s program • Personal circumstances causing a delay in assignment submission • Student’s CV & ways to incorporate their marketing experience • Remote versus in-person work format during Covid-19 (personal & work-related circumstances) • Complaint about group work • Issues with assignment submission • Questions re. choice of format & student’s placement • Struggles with finding evidence/references (student’s thought process) |
1. Official matters—5 emails: • Company placement & job interview outcome • Course repetition deferral • Course enrollment (dropping out); incorrect submission • Job offer • Assessment submission deadline |
| 2. (Class) content-related topics + student’s thought process—1 email: |
2. Class content-related topics—5 emails: |
|
| 3. Personal topics—2 emails: |
||
| Moderately removed (0.3-0.6) | 1. Personal topics—17 emails: |
1. Official matters—10 emails: |
| 2. Official matters—3 emails |
2. (Class) content-related topics—4 emails: |
|
| 3. (Class) content-related topics—6 emails |
3. Personal topics—3 emails: |
|
| Peripheral 1 (0.6–0.9) | 1. Personal topics—2 emails: |
1. Official matters—4 emails: |
| 2. Content-related topics—2 emails: |
||
| Peripheral 2 (>0.9) | 1. Content-related topics—1 email: |
Declaration of Conflicting Interests
The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author received no financial support for the research, authorship, and/or publication of this article.
Supplemental Material
Supplemental material for this article is available online.
Notes
Author Biography
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
