Lexical Multidimensional Analysis (LMDA), by Tony Berber Sardinha and Shannon Fitzsimmons-Doolan, presents an extension of Douglas Biber’s (Reference Biber1988) Multidimensional Analysis (MDA) framework, which was originally set to analyse register variation through co-occurring linguistic features. While traditional MDA (henceforth TMDA) focuses on functional variation (how grammatical features reflect communicative purposes), Berber Sardinha and Fitzsimmons-Doolan’s book discusses how LMDA uses lexical variables (words, lemmas and collocations) to uncover discursive and ideological structures within texts. One of the book’s main contributions lies in demonstrating how lexical choices can indicate socially shared ideas, values and stances on social issues such as climate change and migrant education. The Element consists of five chapters, which cover the key theoretical foundations of LMDA, its core methodological procedures and two specific case studies; it ends with a general conclusion with implications for future research.

In the introductory chapter (pp. 1–13), the authors discuss several studies which explain how LMDA was elaborated, pointing out the differences between TMDA and LMDA. While the former focuses on lexico-grammar, the latter examines the lexis to evaluate discourse. They point out that LMDA has been independently applied to diverse domains – such as education policy, national representation, the infodemic, sustainability and literary style – based on theoretical foundations which combine quantitative and qualitative corpus linguistic methods. Employing statistical procedures, the model identifies clusters of co-occurring lexical features using factor analysis, while interpretation relies on discourse analysis to label each dimension as a particular discourse or ideology. Each dimension thus represents a macro-level construct (e.g. neoliberalism, environmental denial, inclusivity discourse) encoded in lexical co-occurrence patterns. The chapter is clearly organised and situates LMDA between Critical Discourse Analysis (CDA) and Corpus-Assisted Discourse Studies (CADS). The authors argue that while CDA often lacks empirical breadth, LMDA integrates large-scale quantitative analysis that robustly supports its results. At the same time, while CADS highlights collocational patterns, LMDA interprets them as systematic ideological configurations. The chapter effectively leads the reader to grasp the principles of LMDA in a clear and theoretically coherent manner.

Chapter 2, titled ‘Synthesizing existing LMDA scholarship’ (pp. 13–26), reviews the trajectory of LMDA-related studies and highlights methodological diversity and theoretical innovation. Early foundations come from Biber (Reference Biber1993), who employed factor analysis of collocations to identify word senses, and Crossley & Louwerse (Reference Crossley and Louwerse2007), who explored bigram frequencies across registers. These studies established the viability of statistical modelling of lexical co-occurrence. Subsequent research by Fitzsimmons-Doolan and Berber Sardinha expanded LMDA into applied domains. Fitzsimmons-Doolan (Reference Fitzsimmons-Doolan2014, Reference Fitzsimmons-Doolan2019) analysed language ideologies in US education policy, identifying latent constructs such as the idea that language acquisition is monolingual and systematic. Berber Sardinha (Reference Berber Sardinha, Friginal and Hardy2021) employed LMDA to trace national representations in Google Books and academic discourses in Applied Linguistics journals, revealing historical shifts (e.g. applied linguistics as empirical science vs educational practice); Kauffmann & Berber Sardinha (Reference Kauffmann, Berber Sardinha, Friginal and Hardy2021) applied LMDA to literary style, identifying thematic dimensions (e.g. romance, morality, introspection) in Machado de Assis’s novels; and Clarke et al. (Reference Clarke, McEnery and Brookes2021) introduced keyword co-occurrence analysis via Multiple Correspondence Analysis to track representational discourses (e.g. Islam in British media, climate change denial in pseudoscience). Against this backdrop, Berber Sardinha and Fitzsimmons-Doolan discuss methodological issues regarding the results obtained in earlier work, and outline approaches to manage them, particularly regarding zero counts in the data and the interpretation of poles with strongly positive and strongly negative lexical loadings. They report the following common research outcomes: (1) identifying latent discursive constructs (ideologies, representations, stylistic dimensions); (2) exploring their distribution over time, register or source; and (3) deriving secondary measures (e.g. clustering historical eras or discourse shifts). Together, these studies indicate that LMDA’s findings depend on design choices – lexical variables (collocations, lemmas or keywords), measurement type (frequency, presence/absence) and corpus composition – allowing flexibility across disciplines such as linguistics, literary studies and social science. This chapter is particularly valuable for bringing these diverse applications together and articulating a coherent methodological framework for LMDA research.

In contrast to the theoretical orientation of the preceding sections, chapter 3 (‘How to conduct a lexical multidimensional analysis’, pp. 26–33) provides a step-by-step guide to LMDA, outlining a methodological procedure that parallels traditional MDA while relying exclusively on lexical data. The procedure is summarised as follows: (1) Corpus construction: selecting and preparing a corpus suited to the research question (balanced across registers, time periods or ideological stances); (2) Feature identification: choosing lexical variables, namely words, lemmas, n-grams or collocates (sometimes filtered by part of speech or grammatical role); (3) Counting occurrences: computing normalised frequency of each variable per text; (4) Statistical modelling: applying factor analysis to detect sets of co-occurring lexical features; and (5) Interpretation: examining texts with high or low factor scores qualitatively to label each factor as a dimension representing a discourse or ideology. Although the chapter succeeds in mapping out the methodological sequence, its complexity may require readers to revisit the outlined steps and rely on the following chapters for a more practically grounded appreciation of how the procedures function.

Some key assumptions relate to variation, since language use varies systematically with context; co-occurrence, since lexical features that co-occur reflect shared discursive or ideological functions; multidimensionality, in that several dimensions operate simultaneously within each text; parsimony, considering that a small number of dimensions explain most lexical variation; and comparability, since dimensions allow comparison across texts, genres or periods. As previously noted, while TMDA interprets dimensions functionally (e.g. ‘informational vs involved production’), LMDA interprets them discursively (e.g. ‘climate skepticism vs environmental concern’). In this approach, qualitative interpretation is essential for meaningful labelling. Researchers should not rely only on lexis but should also consider words in context, returning to the texts to check whether phenomena (such as polysemy) obscure the author’s intended meaning, or introduce ambiguity or sarcasm.

Chapter 4 is one of the most engaging sections in the book, demonstrating LMDA’s capacity to detect ideological discourses in online communication. Case Study 1 (pp. 33–52), which is concerned with the discourses around climate change on a conservative social media platform, applies LMDA to texts from GETTR (a conservative social media network similar to Twitter) to uncover the discourses on climate change circulating among users sceptical of environmental activism. The corpus consists of thousands of posts collected over a defined period. Lexical variables (e.g. climate, hoax, agenda, freedom, truth, science, control) were lemmatised and normalised by frequency. Berber Sardinha explains how variables are loaded in different factors and how to determine the factors that carry most variables to be interpreted. A factor analysis of lexical co-occurrence revealed five distinct dimensions of discourse, labelled as follows: Dimension 1 ‘Discrediting Climate Change vs Political Motivation’, Dimension 2 ‘Activism Alarmism vs Progressive Measures Dismissal’, Dimension 3 ‘Climate Collusion vs Anti-Globalism’, Dimension 4 ‘Anti-Chinese Campaigns’ and Dimension 5 ‘Regulatory Agency Distrust vs Hashtag Rejection of Anthropogenic Global Warming (AGW)’. The author draws attention to the fact that some dimensions may overlap carrying similar variables in different poles, and this explains the importance of reading texts carefully. A clear example involves Dimensions 1 and 5, which seem to discredit climate change as a whole; however, the former dimension represented the anger and popular sentiment while the latter repelled climate activism based on business credentials. The analysis illustrates how lexical patterns (e.g. collocations of ‘climate + hoax/fraud/alarmism’) encode ideological stances. Posts combining these features reveal a discursive ecosystem that naturalises denial and delegitimises environmental science. By quantifying these patterns and interpreting them qualitatively, the case study shows that LMDA can uncover ideological alignments in digital communication spaces, complementing qualitative CDA while maintaining empirical rigour. One of the strengths of this chapter is the authors’ ability to demystify factor analysis for readers who may not be statistically trained.

Chapter 5 discusses the second empirical LMDA case study with a focus on policy and media texts, and tracking ideological discourses about US migrant education across registers and time (pp. 52–74). The corpus includes policy documents, government web texts, and national and regional newspaper articles (2003–17). Following her own earlier studies, Fitzsimmons-Doolan extracted 114 collocates of the abbreviation MIGR, for migrant (e.g. worker, family, law, reform, student, service, policy). Factor analyses identified eleven ideological discourses, four of which were analysed in depth: (1) government programmes serve children in need, (2) immigrant lives are narratives (usually criminal or uplifting), (3) top-down immigration laws are punitive or permissive, and (4) immigrants and governments have a mutual relationship founded on acts of service and work. Statistical analyses (ANOVAs) revealed significant variation by register and time period. Institutional web texts expressed more ‘managerial and accountability’ discourses, while media texts displayed ‘deficit and conflict’ frames. Over time, equity-oriented discourses gained modest prominence, but deficit framing remained dominant. This case illustrates LMDA’s potential to map ideological evolution in public policy and media communication, revealing how lexical co-occurrence tracks shifts in socio-political attitudes toward migration. Although the chapter presents its main idea with clarity, the interpretive steps involved in handling the statistical data are intricate, and a fuller account of these procedures could further support readers’ understanding.

The final concluding remarks (pp. 74–5) outline several promising directions involving the theoretical and methodological contributions of LMDA. As a framework, LMDA merges the quantitative precision of MDAs with the interpretive depth of discourse analysis, thus providing a replicable method for uncovering discourses and ideologies encoded in lexical patterns. It extends corpus linguistics by focusing not only on frequency but also on associational structure, capturing how words cluster around shared meanings in social contexts. The authors emphasise that LMDA bridges corpus linguistics and CDA, avoiding the subjectivity of small-sample qualitative work. It can also identify and compare ideological discourses across registers, time or media, offering flexibility in variable selection and allowing application across domains (education, literature, digital media, sustainability). Additionally, it encourages mixed-methods interpretation, combining statistics with textual reading.

Future directions include expanding LMDA to multimodal corpora (e.g. integrating visual data from social media), applying it to large-scale datasets through computational processing and refining procedures for interpretive reliability. Ultimately, LMDA positions lexis as a marker of ideology, showing that word co-occurrence patterns embody socially shared belief systems. The method reveals that language, far from neutral, functions as a site of ideological struggle – a premise consistent with discourse theory but now empirically demonstrable through corpus-based MDA.

The final chapter effectively synthesises the book’s central theoretical and methodological contributions, positioning Lexical Multidimensional Analysis as a robust framework for investigating how lexical systems encode social meaning. By extending Biber’s multidimensional approach from communicative to ideological functions, the authors further align corpus linguistics with critical and cognitive traditions in discourse studies. This Element will therefore be of particular interest to researchers in corpus-based discourse studies, applied linguistics and digital media analysis. Overall, by integrating statistical rigour with qualitative interpretation, Berber Sardinha and Fitzsimmons-Doolan offer a powerful and flexible tool for examining how language constructs, sustains and challenges ideologies across diverse domains.