CARS Model¶
The CARS model reads a research introduction as establishing a territory, carving a niche in prior work, and positioning the present study to occupy it.
Core Idea¶
John M. Swales's Create a Research Space (CARS) model describes the rhetorical work of a research-article introduction through three moves: establish a territory by showing a field and its prior work; establish a niche by locating a gap, question, counterclaim or line to continue; and occupy the niche by positioning the present study's purpose, work or findings as a response. The sequence is a functional account of why a new article has a place in a research conversation, not a requirement that every introduction consist of exactly three paragraphs.[1][2][3]
Swales's 1990 book is the origin identified by the bibliographic record, but its full CARS chapter was not available for direct inspection in this pass. A university teaching adaptation gives the moves and their instructional use; Laurence Anthony's original corpus study applies and tests them in software-engineering introductions. The latter is important because it shows repeated moves and locally important features the simplest template misses.[1][2][3]
Structural Signature¶
Sig role-phrases:
- Research territory: the significant field, problem or prior-work context in which the paper asks to be read.
- Prior-work relationship: the cited or shared work relative to which a new contribution can be positioned.
- Niche claim: a gap, question, disagreement or continuation that opens research space inside that territory.
- Occupation by present study: a purpose, method, result or article trajectory that answers or extends the identified niche.
- Move analyst or writer: a reader who codes the actual functions, or an author who uses the scheme as a drafting heuristic, without mistaking step labels for a universal fixed order.[2][3]
The niche is the hinge. A territory followed by a list of methods can leave unclear why the present work is necessary. A gap without a present-study response is not yet occupied. In actual texts, territory and niche may alternate more than once before and after an announcement of the study; Anthony's coded introductions make this departure from a one-pass outline visible.[3]
What It Is Not¶
CARS is not IMRaD's Introduction–Methods–Results–Discussion structure of an entire paper. It explains rhetorical functions within an introduction. Nor is it a mechanical keyword detector: “however” may signal a gap, but a sentence's function depends on how it relates to previous and present work. The optional step labels are also not a guarantee that every research article announces findings or previews every section in the same way.[2][3]
It is not a quality certificate. An introduction could display the three move functions while misrepresenting the cited literature or claiming a non-existent gap. Conversely, a discipline may use a legitimate introduction that cycles or front-loads a finding. Anthony's software-engineering study found all three broad moves among its twelve sampled introductions yet judged several detailed steps inadequate for that corpus. That is a bounded finding, not a verdict on every discipline.[3]
Scope of Application¶
The model originated in applied linguistic study of academic genres and has a practical teaching use: a reader can ask where an introduction sets the field, what problem is left open, and how the study responds. A writer can use those questions to diagnose a draft that reviews literature but never explains its own contribution. The University of Rochester presents it as a framework for analyzing scholarly introductions and for instruction, explicitly as an adaptation of Swales rather than a new original theory.[1][2]
Anthony tested how well CARS described twelve article introductions from IEEE Transactions on Software Engineering. His move table shows all three broad functions, but often in recursive sequences rather than one linear 1→2→3. He also found that the model's step categories did not clearly accommodate definitions, extended examples and evaluation of the present work in those introductions. His proposed modified CARS retains broad moves while adding or sharpening a locally important evaluation step.[3]
Clarity¶
Consider a constructed introduction: “This journal has established that software inspections find some defects” marks territory and prior work; “but studies have not compared inspection yield after remote handoffs” marks a niche; “we compare two handoff protocols in three teams” occupies it. The three sentences are an author illustration of functions, not a source-reported article or a claim that the hypothetical comparison was conducted. A mere list of prior inspection papers would not by itself mark the niche; an announced experiment without the comparison gap would be harder to place.[2]
Now compare actual coding rather than an invented paragraph. Anthony's Table 1 gives the anonymized introduction 1.2 a six-part move path 1–2–1–3–1–3 across an 803-word text. A reader who expects exactly one territory, then one niche, then one occupation would miss the later return to field context and the second occupation. The table records functional coding, not the raw article sentences; the anonymous source article was not directly checked.[3]
Manages Complexity¶
The three moves compress many local writing choices into one research-space relation. Territory gives a reader enough common ground to judge the question; niche makes the new work's necessity legible; occupation turns that necessity into a promised contribution. As a diagnostic, it distinguishes a long literature review from a justified research introduction without requiring a single mandatory sentence at each step.[2][3]
The compression also risks erasing disciplinary detail. Anthony reports that, in his twelve software-engineering introductions, all three broad moves occurred, but definitions and examples did important work and an explicit evaluation of the present research was prominent. His Table 1 shows repeated territory/niche cycles, and his modified model responds to step-level mismatches. A rigid classroom outline would save teaching time but could teach an inaccurate account of those texts.[3]
Abstract Reasoning¶
To apply CARS descriptively, segment an introduction by rhetorical purpose, not just paragraph boundary. Ask whether a span makes the field relevant (M1), creates a question or insufficiency relative to it (M2), or states how the present paper responds (M3). Allow returns such as M1 after M2 when an author adds more background, and permit a second M3 when purpose and later contribution are stated separately. Anthony's coded sequences demonstrate why counts and order must be reported, not assumed.[3]
To apply it as a writing aid, check the link between moves. Is the niche honestly supported by prior work rather than a formulaic “little is known”? Does the announced study actually address that niche? Does the territory supply the reader with enough context for the gap to matter? Anthony's critique warns that a complete analysis may require added discipline-specific functions rather than squeezing all material into the original step inventory.[2][3]
Knowledge Transfer¶
The model can orient readers and novice writers in other research disciplines, but its detailed step frequency and typical order should be tested in the target genre. Anthony's result for software engineering is an application with a sample of twelve, not evidence that all research introductions cycle in the same way. The university teaching adaptation is a teaching tool, not a substitute for reading actual disciplinary articles.[2][3]
The wider metaphor of “creating space” may apply to grant proposals or dissertations, but those genres have different constraints and audiences. Using CARS there is an imported heuristic until original genre analysis shows comparable territory/niche/occupation functions. Even within research articles, a move's function may recur several times, so transfer should preserve the role relation rather than a three-box page layout.[3]
Examples¶
Anthony's coded software-engineering introduction 1.1¶
Anthony's primary corpus table labels one anonymized IEEE Transactions on Software Engineering introduction 1.1. Its 1,479-word path contains multiple M1/M2 alternations before and around M3 material, far more complex than one pass through the three labels. His analysis treats it as one of twelve sampled article introductions exhibiting all broad moves. The available proceedings report the code sequence and study interpretation; they do not provide enough raw article text here for an independent sentence-by-sentence recoding.[3]
Mapped back: research territory is repeatedly coded M1; prior-work relationship is made and remade as the text returns to M1; niche claim is coded in several M2 segments rather than one isolated gap; occupation by present study is M3 material after earlier moves; the move analyst is Anthony, whose table records the actual sequence. If those M2 spans did not relate to prior work, this would not be a supported CARS application merely because the numeral 2 appeared.
Anthony's coded introduction 1.2¶
The same original study reports another anonymized introduction, 1.2, at 803 words with sequence 1–2–1–3–1–3. The shorter record still revisits territory after an occupation move. This is a separate coded text, not another invented example and not a claim that its exact wording has been inspected in this pass. It demonstrates that even a text retaining all three functions may interleave them instead of marching through one fixed order.[3]
Mapped back: the initial and returning M1 segments are territory/prior-work positioning; M2 marks the niche; the two M3 segments are occupation; Anthony is the analyst applying function codes to the article. A teaching outline that insisted on only one M1 at the start would misdescribe the table's reported path.
Structural Tensions¶
Portable heuristic versus disciplinary fidelity. A simple three-move sequence makes the research-space logic memorable and can help a novice discover a missing justification. The cost of enforcing it as a complete, fixed template is that valid introductions may cycle, define technical terms or evaluate present research in ways its steps do not capture. Detailed corpus-based coding gains fidelity to a disciplinary genre but loses the simplicity of a universal writing checklist. Anthony's software-engineering results establish this as a real use-and-analysis tradeoff in that sample, not a proof that CARS is useless. Diagnostic: are the moves guiding a reader's questions, or being treated as a compulsory one-pass order that overrides observed text?[2][3]
Structural–Framed Character¶
CARS is mixed but leaning framed on the structural–framed spectrum. The territory–niche–occupation relation is abstract enough to travel across topics; its function does not depend on a particular software-engineering fact. Yet genre recognition is a social reading practice: disciplinary writers and reviewers decide what counts as sufficient prior work, a credible gap and a meaningful contribution. It bears evaluative weight because the model privileges a rhetoric of novelty and fit, not a neutral count of text spans.[1][3]
Its institutional origin is academic research writing and teaching, not a mathematical law. The vocabulary travels to other genres only with care. A grant proposal that claims a gap may be a useful import of the model, but it is not automatically an observed research-article introduction. Anthony's corpus provides recognition of the three moves within one discipline and a reason to qualify the detailed step inventory. Its character: a socially interpreted genre model with a portable functional skeleton and discipline-sensitive realization.[2][3]
Structural Core vs. Domain Accent¶
The skeleton is locate a field and prior work → mark a research space within it → position present work as answering or extending it. The domain-bound mechanism is the scholarly introduction, with citations, gaps, audiences, discourse conventions and publication stakes. Step variants—counterclaim versus question, findings announcement versus article roadmap—are accents rather than a requirement that every text use each step. Anthony's repeated coding shows that even order is more flexible than a naive three-paragraph template.[3]
The named CARS entry fails the prime bar because removing research-article introductions and scholarly contribution claims removes its identity. The method presupposes the research-introduction Genre as a socially recognized artifact, but is not itself one genre instance or a universal text template. A broad future prime about claiming a niche might be imaginable, but would need its own evidence and boundary tests.
Instantiates / Related Primes¶
This entry presupposes Genre.
Strict compositional prerequisite: Genre (presupposes, strict). CARS analyzes recurrent rhetorical moves in the research-introduction genre; many genres do not exhibit this specific territory–niche–occupation model. Rhetorical Situation and Rhetorical stance remain contextual neighbors, not additional edges. This link does not impose a rigid three-paragraph template. The original Swales chapter remains bibliographically verified but not fully inspected; later adaptations and Anthony's study have separate evidentiary standing.
Relationships to Other Abstractions¶
Current abstraction CARS Model Domain-specific
Parents (1) — more general patterns this builds on
-
CARS Model presupposes Genre Domain-specific
CARS presupposes the research-introduction genre.Its territory-niche-occupation analysis needs a recognized research-introduction context; genre identity exists without CARS.
Hierarchy path (1) — routes to 1 parentless root
- CARS Model → Genre → Schema → Abstraction
Neighborhood in Abstraction Space¶
CARS Model sits in a sparse region of the domain-specific corpus (84th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.
Family — Qualitative Research & Fieldwork Methods (29 abstractions)
Nearest neighbors
- Reflexive Journal — 0.83
- Ethics dumping — 0.81
- Information Seeking — 0.81
- Research Program — 0.81
- Planck's Principle — 0.81
Computed from structural-signature embeddings · 2026-10-08
Not to Be Confused With¶
CARS analyzes moves in an introduction, not the full article's IMRaD sections. A move is a rhetorical function, not a paragraph number or an automatically detected phrase. The original 1990 book is the model's bibliographic origin; the Rochester page is a later teaching adaptation, and Anthony's proceedings paper is a separate empirical test. The anonymous introduction codes are reported analyses, not independently inspected full source articles.[1][2][3]
References¶
[1] John M. Swales, Genre Analysis: English in Academic and Research Settings (Cambridge University Press, 1990), original-book bibliographic record; full model chapter not inspected in this pass. registry ↩a ↩b ↩c ↩d ↩e
[2] University of Rochester Writing, Speaking, and Argument Program, “CARS Intro”, institutional teaching adaptation of Swales's moves; search-visible page, direct fetch timed out (access attempted 2026-10-02). registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i ↩j ↩k ↩l
[3] Laurence Anthony, “Preaching to Cannibals: A Look at Academic Writing in the Field of Engineering,” 1997 proceedings, pp.75–83, original corpus study in ERIC-hosted proceedings, especially Figure 2, Tables 1–2, discussion and conclusion (accessed 2026-10-02). registry ↩a ↩b ↩c ↩d ↩e ↩f ↩g ↩h ↩i ↩j ↩k ↩l ↩m ↩n ↩o ↩p ↩q ↩r ↩s ↩t ↩u