Cannabis Coverage and the Distribution of Voices in the Portuguese Media Ecosystem
Débora Paulino
Abstract
Media representations help define public issues, but their significance lies not only in how a subject is framed but also in whose perspectives receive visibility and authority. Cannabis offers a useful case for examining these dynamics in Portugal, where medical, industrial and recreational uses occupy different legal and social positions. This case study analyses 48 articles published by Público, Expresso and Observador between January 2019 and April 2020. Articles primarily focused on criminality were excluded to examine coverage beyond crime-centred narratives. The content analysis considered journalistic genres, information sources, main and secondary themes, principal protagonists, tone and framing. Medicinal cannabis was the main theme in 64.6% of articles, followed by recreational use (29.2%). Coverage was predominantly neutral in tone (60.4%) and informative in framing (81.3%). Public institutions were the principal protagonists in 31.3% of articles, while private institutions and corporations each accounted for 20.8%. Consumers and health professionals were principal protagonists in only one article each. The findings reveal a distinction between the tone of coverage and the distribution of speaking roles: predominantly neutral and informative reporting did not translate into a broad representation of perspectives. Situated within the Portuguese media ecosystem, the study highlights the value of examining institutional visibility alongside thematic framing and discusses the implications for pluralism in reporting on a socially stigmatised issue.
Review reports
xPeerd (AI review)
AI Review (xPeerd — DAReview)
Summary
This manuscript examines 48 non-crime-centred cannabis-related articles published in Público, Expresso, and Observador between January 2019 and April 2020. Its central claim is that apparently neutral reporting should not be equated with pluralistic representation, because institutional actors dominated the principal speaking roles while consumers and health professionals were rarely central protagonists. The paper is strongest when it clearly distinguishes “tone” from “distribution of voices” and when it explicitly limits its claims to a bounded corpus rather than broader causal conclusions about effects, intentions, or the whole Portuguese media ecosystem. The study is also well contextualized in Portuguese legal and media-system developments, and the article shows commendable caution in several places about overclaiming.
Its main weaknesses concern reproducibility, evidentiary transparency, and originality positioning. The manuscript openly acknowledges that no intercoder reliability coefficient is reported and that one source table contains denominator inconsistencies, both of which materially weaken confidence in the coding-based results. It is also a revised adaptation of a 2021 Master’s dissertation, which raises an originality and self-reuse question that should be addressed more directly for journal submission. Overall, the paper has a coherent argument and a plausible contribution, but it would benefit from substantial methodological clarification, tighter handling of the ecosystem section, and more explicit articulation of what is genuinely new in this article relative to the dissertation.
Potential Major Revisions
- Reproducibility is significantly weakened by the absence of intercoder reliability for interpretive variables.
The manuscript states that “The available dissertation does not report an intercoder reliability coefficient” and adds that “This is an important limitation, especially for interpretive categories such as tone, framing and principal protagonist.” It further concedes that “The coding should be understood as a structured analysis by the original researcher, rather than a multi-coder study with independently verified agreement.” This is not a minor caveat; it directly affects the trustworthiness of the article’s main findings, because the core contribution depends on interpretive coding of tone, framing, and protagonist prominence. Page 5, Section 4 therefore contains the strongest reason for editorial skepticism.
Because the article’s central conclusion rests on distinctions such as “neutral in tone” versus “institutionally centred,” the absence of independently checked coding reliability leaves readers unable to assess how stable those classifications would be across coders. The argument may still be plausible, but in its current form it is methodologically vulnerable. At minimum, the author should either:
a) conduct and report a fresh reliability test on the article dataset,
b) reframe the piece more explicitly as a single-researcher exploratory coding study, or
c) reduce the force of claims that rely on these interpretive categories.
- One quantitative component is internally unstable, and that instability affects confidence in the dataset as a whole.
The manuscript acknowledges that “the source table contains percentages that do not map neatly onto a denominator of 48” and states that these figures “should be checked against the original coding sheet before any further quantitative reanalysis.” Elsewhere it says that “Because the underlying coding sheet was not available for independent verification, these source percentages are not reproduced as quantitative findings here.” This is appropriately cautious, but it also reveals a serious data-audit problem. If one table cannot be reconciled with the denominator, editors and reviewers may reasonably ask whether other tables were also rechecked from the underlying sheet.
The manuscript attempts to contain this issue by excluding the problematic source percentages from reproduced results, but the article still relies on the same broader coding exercise. That means the paper should report more explicitly what was verified and what was inherited from the dissertation. Without that transparency, readers are asked to trust a selectively repaired dataset. A stronger revision would explain which variables were revalidated article-by-article and whether all percentages in Table 1 were recalculated from the article-level data or copied from prior work.
- The manuscript’s originality relative to the prior dissertation remains exposed and should be handled more directly.
The article explicitly states that “The article is a substantially revised, Portugal-focused adaptation of a broader comparative Master’s dissertation” and later notes: “Before submission, the author should confirm the public repository status of the dissertation and ensure that the article is sufficiently revised to meet the journal's originality requirements.” This is a prudent disclosure, but it also signals a potential editorial problem: the manuscript itself recognizes that originality may be questioned.
In its present form, the paper says it narrows the focus to Portugal and reorganizes the argument around tone, framing, and protagonists, but it does not sharply specify what analytical advance this article makes beyond extraction from the earlier dissertation. The statement “The article does not present the original corpus as newly collected data” is good ethical practice, yet it does not by itself establish novelty.
A journal-ready revision should explicitly differentiate:
- what parts of the coding derive from the prior dissertation,
- what analyses are newly performed for this article,
- what interpretive argument is new, and
- why a Portugal-only reframing materially advances the literature rather than merely repackaging earlier graduate work.
- The ecosystem section is useful but somewhat overextended relative to what the study actually tests.
Section 3 repeatedly states that current figures are not used as direct explanations of the 2019–2020 corpus. For example, the manuscript notes that contemporary indicators “should not be projected backwards onto the 2019–2020 sample” and that the study “did not measure newsroom resources, journalists' working conditions or editorial decision-making directly.” It also says ownership “does not by itself establish editorial bias.” These caveats are appropriate, but they reveal a structural issue: the ecosystem section may be broader than the evidence base can sustain.
Because the paper does not empirically test ownership, audience behavior, platform circulation, or newsroom conditions, the ecosystem profile risks functioning as context that implies explanatory relevance without demonstrating it. The discussion itself warns against using these conditions “as a shortcut to explain the content results.” That warning is analytically sound, but it also suggests that the article should either shorten Section 3 or integrate it more tightly into a clearly non-causal contextual framing.
- The exclusion of crime-centred articles is conceptually defensible but narrows generalizability more than the article’s framing sometimes suggests.
The manuscript states that excluding crime-centred items was “a deliberate analytical choice” and that the purpose was “to examine how the subject was represented when criminality was not the principal organising theme.” That is a coherent design decision. However, because cannabis coverage is often strongly shaped by criminal or policing narratives, excluding such items may remove a major domain through which media pluralism and voice distribution are actually manifested.
The manuscript does acknowledge this limitation, noting that it “cannot describe the overall balance between criminal, medical, recreational and industrial coverage.” Still, the author should further clarify that the findings concern a filtered sub-corpus, not cannabis coverage in Portugal more generally. Otherwise, readers may overgeneralize from a design that intentionally excludes one of the most socially salient framing dimensions.
- The claims are mostly careful, but the article still needs clearer operational grounding for “principal protagonist.”
The manuscript defines principal protagonists as the actor “receiving the greatest prominence in the article, based on references and statements in the headline, lead and main text.” This is a reasonable coding concept, but it remains vulnerable to subjective judgment unless the codebook criteria are summarized more concretely. For example, how were ties handled? Was prominence determined by frequency, narrative position, direct quotation, or interpretive salience?
Since the paper’s main conceptual innovation is distinguishing source diversity from protagonist prominence, that distinction requires especially transparent operationalization. Otherwise, the central variable appears theoretically appealing but empirically under-specified.
- Anti-fraud / coherence check
I do not see strong paper-mill or fabrication signals in the manuscript. The argument is coherent, the limitations are openly acknowledged, and the author repeatedly narrows claims rather than inflating them. That said, the combination of inherited coding, absent intercoder reliability, and one inconsistent quantitative table creates a traceability problem. This does not by itself indicate fraud, but it does create an evidentiary vulnerability that should be corrected before publication.
Potential Minor Revisions
- Title, summary framing, and contribution language
The title is clear and aligned with the findings. However, some contribution wording could be tightened. For example, on page 1 the manuscript says, “The article makes three contributions.” This is acceptable, but the phrasing would be stronger if each contribution were paired with a more explicit statement of what prior cannabis-media scholarship has not yet done in the Portuguese case.
- Typographic and stylistic issues
I do not detect major typographical or grammatical weakness in the extracted text. The prose is generally clean, formal, and controlled. Still, a few stylistic edits would improve concision and reduce repetition:
| Page | Section | Issue | Suggested revision |
|---|---|---|---|
| 1 | Introduction | Repetition of cautionary explanation about WHO recommendations and scheduling change | Condense the explanation while preserving legal accuracy. |
| 2 | Literature review | Repeated qualification that neutrality does not equal pluralism appears several times across the paper | Keep the point, but reduce restatement across Sections 2, 6, and 8. |
| 3–4 | Media ecosystem / Methods transition | Sectional movement feels slightly diffuse | Add a bridging sentence explaining exactly why ecosystem context precedes the coding design. |
| 7–9 | Discussion / Conclusion | Some claims are restated with minimal variation | Streamline for sharper emphasis and reduced redundancy. |
- Formatting issues
| Page | Section | Issue | Comment |
|---|---|---|---|
| 9 | Table 1 | Table is useful, but source-category inconsistency is relegated to note/text | Consider visually separating fully validated variables from non-reproduced source categories even more explicitly. |
| 10–11 | References | Reference list appears generally consistent | Final submission should check house style for capitalization, DOI presentation, and legal-source formatting. |
| Throughout | Headings and emphasis | Bold sublabels within sections are readable | Ensure consistency with target journal style, especially for subsection formatting. |
- Minor conceptual clarifications
| Page | Section | Issue | Suggested fix |
|---|---|---|---|
| 4–5 | Methods | “Visual prominence” is named among form variables but not discussed in findings | Either report it in results or remove it from the methods overview if unused. |
| 5–6 | Findings | “Corporations” appears to include “parties, associations or other organised bodies” | This category label may confuse readers because parties/associations are not typically “corporations” in a narrow sense. Rename or define more clearly. |
| 1 / 8 | Introduction / Conclusion | “Portuguese media ecosystem” is invoked at different levels of generality | Keep explicit distinction between studied outlets and wider ecosystem to avoid overreach. |
- AI content analysis for this post-2021 submission
Estimated percentage of AI-generated content: low to moderate likelihood, roughly 10–25%.
Rationale: the prose is highly uniform, polished, and caution-heavy, with repeated balancing formulas such as “does not establish,” “cannot conclude,” and “should not be treated as.” That pattern can occur in human academic revision, especially when a dissertation chapter is reworked for submission, but it can also resemble AI-assisted smoothing. I do not see obvious hallucinated references or incoherent transitions; the text is substantially more coherent than typical AI-generated low-quality academic prose. The more likely scenario, if AI was used, is limited editorial assistance at the level of polishing, restructuring, or hedging language rather than wholesale content generation.
Highlighted passages with possible AI-assisted stylistic smoothing include:
- “The study does not claim that the observed distribution proves deliberate exclusion or that it caused particular audience attitudes. Rather, it identifies a pattern in the published corpus and considers its implications for understanding journalistic representation.” page 1.
- “The design does not establish the intentions of journalists, causal effects of regulation, audience interpretations or the effect of the coverage on public opinion.” page 5.
- “They do not demonstrate that the articles were inaccurate, biased or intentionally exclusionary; rather, they identify a pattern that warrants closer examination...” page 6.
Assessed epistemic impact: limited, if any, on the core contribution. The main concerns here are not AI authorship but methodological traceability, originality relative to the dissertation, and coding reliability. Even if AI-assisted line editing occurred, the larger scholarly issue remains whether the empirical claims are reproducible and sufficiently novel for journal publication.
Recommendations
-
Revalidate the article-level dataset before submission. Recheck all counts and percentages from the original coding sheet and explicitly state which variables were independently verified for the article version. This is the single most important practical revision.
-
Add a reliability procedure. If possible, have a second coder independently code a subset or the full corpus and report an appropriate agreement statistic for tone, framing, and principal protagonist. If that is not feasible, explicitly reposition the study as exploratory single-coder research and temper the strength of inferential claims.
-
Make originality explicit. Add a short but direct paragraph specifying exactly how this article differs from the 2021 dissertation in scope, argument, analysis, and intended scholarly contribution. This will reduce editorial concern about redundant publication or insufficient transformation.
-
Tighten the ecosystem section. Keep only the contextual material that is necessary to interpret the corpus and move any broader current-market description to a shorter framing paragraph unless it is analytically integrated into the discussion.
-
Clarify the protagonist coding rule with operational detail and, ideally, a brief coding example. Since this is the manuscript’s key analytical distinction, readers need to understand exactly how one actor became the “principal protagonist” in borderline cases.
-
Reduce repetition across Introduction, Discussion, and Conclusion. The manuscript’s caution is a strength, but several formulations recur with minimal substantive gain. A leaner structure would sharpen the paper’s argumentative force.
-
Preserve the current restraint against overclaiming. One of the manuscript’s strengths is that it repeatedly avoids unjustified causal or normative conclusions. That discipline should remain in the revised version.
Overall recommendation: promising but not yet fully publication-ready. The manuscript has a coherent argument, a plausible niche contribution, and no strong signs of fabrication, but it requires substantial methodological strengthening and clearer originality positioning before it would be robust enough for external review or acceptance.
Reviewer3 (AI review)
AI Review (reviewer3.com)
Comments (7)
1. Single principal protagonist metric cannot measure overall voice pluralism — Critical
health professionals were principal protagonists in only one article each. The findings reveal a distinction between the tone of coverage and the distribution of speaking roles: predominantly neutral and informative reporting did not translate into a broad representation of perspectives. Situated within the Portuguese media
The authors argue that predominantly neutral and informative coverage did not translate into a broad or pluralistic representation of perspectives because institutional and organised actors occupied most principal protagonist roles while consumers and health professionals were rarely the main protagonists. However, coding only a single mutually exclusive "principal protagonist" per article measures which actor anchored the headline and lead, not the breadth or pluralism of perspectives represented across the article as a whole, and the companion "information source" variable neither categorised actor types (such as consumers or clinicians) nor survived data verification. Because an article triggered by a regulatory announcement or corporate event will naturally feature an institution in the lead even if it extensively quotes patients, consumers, or medical professionals in the body text, a low count of consumer or clinician principal protagonists does not establish that their perspectives were absent or lacked substantive representation. The authors should either recode the corpus to measure the presence and prominence of all quoted actor types across each article, or revise the headline conclusion to state strictly that institutional actors most frequently served as the primary narrative anchor rather than concluding that coverage lacked a broad representation of perspectives.
2. Overlapping and ambiguous institutional protagonist categories — Major
were the main protagonists in 15 articles (31.3%). Private institutions, organisations or companies were principal protagonists in 10 articles (20.8%), and corporations—such as parties, associations or other organised bodies—also accounted for 10 (20.8%). Public figures who were not classified as political figures
The authors conclude that institutional and organised actors accounted for the majority of principal speaking roles based on combining "Public institutions, organisations or companies" (31.3%), "Private institutions, organisations or companies" (20.8%), and "corporations—such as parties, associations or other organised bodies" (20.8%). In standard English usage, "corporations" refers to commercial companies, which overlaps directly with "Private institutions, organisations or companies," whereas defining corporations as "parties, associations or other organised bodies" appears to conflate civil society associations, advocacy groups, and political parties with corporate entities (and overlaps with "Political or governmental figures"). Because civil society associations (such as patient or consumer advocacy associations) play a very different representational role from commercial cannabis businesses or state agencies, grouping them under an undefined "corporation" label obscures who is actually speaking. The authors should provide explicit operational definitions and examples for each protagonist category, clarify the translation and meaning of "corporations," and distinguish commercial businesses from civil society or patient associations.
3. Cross-tabulation needed to link themes, genres, and protagonist roles — Major
exclusion, audience effects or a causal relationship between regulation and news framing. It does, however, identify an empirical pattern in a defined corpus and demonstrates the analytical value of examining thematic priorities, tone, framing and protagonist prominence together.
The authors argue for the analytical value of bringing together thematic framing, tone, and the distribution of principal protagonists to evaluate media pluralism on socially stigmatised issues. However, the empirical analysis only presents isolated univariate marginal totals for each variable and never examines their joint distribution within the 48 articles (until noting in Section 7 that checking whether protagonists differ across themes would require rechecking the dataset). Without cross-tabulating protagonist categories against main themes, genres, or tone, the study cannot show how these dimensions actually interact—for instance, whether the institutional concentration occurs specifically within neutral, medicinal news stories or across recreational and report genres as well. The authors should either perform simple descriptive cross-tabulations linking theme, genre, tone, and protagonist across the 48 articles, or temper the claim that the study empirically demonstrates the combined analytical relationship among these variables.
4. Unclear coding rules and mutual exclusivity for secondary themes — Minor
Secondary themes further demonstrate the prominence of policy and health. Legalisation and decriminalisation appeared as a secondary theme in 18 articles (37.5%), followed by research and health in 14 (29.2%), cultivation and production in 10 (20.8%), commerce, entrepreneurship and business in four (8.3%), and education in two (4.2%). Thus, the coverage frequently connected cannabis to changes in legal
The authors report five secondary theme categories across the corpus: legalisation and decriminalisation (18 articles), research and health (14), cultivation and production (10), commerce, entrepreneurship and business (4), and education (2). Although described in the plural as "secondary themes," these counts sum to exactly 48 (100%), indicating that every article was assigned exactly one secondary theme, and the variable is omitted from Table 1. Furthermore, the manuscript does not explain how coders separated the main theme of "medicinal cannabis" from the secondary theme of "research and health" or whether articles could lack a secondary theme. The authors should clarify whether secondary theme coding forced a single mutually exclusive choice per article, explain the boundary rules distinguishing main from secondary themes, and include the secondary theme distribution in Table 1.
5. Framing variable measures rhetorical mode rather than substantive frames — Major
Informative framing predominated, appearing in 39 articles (81.3%). Eight articles (16.7%) were coded as explanatory, while one (2.1%) was coded as ironic. The prominence of informative framing is consistent with
The authors ground their study in Entman's (1993) definition of framing—problem definitions, causal interpretations, moral evaluations, and treatment recommendations—yet operationalise the "framing" variable using the stylistic categories "informative" (81.3%), "explanatory" (16.7%), and "ironic" (2.1%), and code "tone" separately as "neutral" (60.4%) or "balanced" (35.4%). Coding framing as "informative" versus "explanatory" measures journalistic mode of presentation (largely duplicating the journalistic genre distinction between news and report) rather than substantive issue frames, and the paper provides no operational criteria explaining how a "neutral" article was distinguished from a "balanced" one. The authors should provide the explicit codebook criteria used to separate "neutral" from "balanced" tone and "informative" from "explanatory" framing, and re-label the latter variable as stylistic or rhetorical mode so it is not conflated with substantive framing in the theoretical discussion.
6. Missing search yield, exclusion criteria, and outlet-level breakdown — Major
Corpus and sampling. The corpus comprises 48 articles published by Público, Expresso and Observador between 1 January 2019 and 30 April 2020. The three outlets were selected to include established print
The authors state that the corpus of 48 articles was obtained by searching three outlets for cannabis-related terms and excluding items in which criminality was the principal angle, and note that form variables included the outlet, date, and visual prominence. However, the manuscript never reports how many total articles were initially retrieved, how many were excluded under the criminality criterion, the operational threshold used to decide when crime was the "principal angle" versus a secondary aspect of a recreational or legal-reform story, or how the 48 retained articles were distributed across Público, Expresso, and Observador. Without knowing the exclusion rate or whether a single outlet accounts for most of the 48 articles, readers cannot assess selection bias or whether the thematic and genre patterns reflect one newspaper's editorial format rather than all three outlets. The authors should report the initial search yield, the number of excluded crime-focused articles, the decision rule for exclusion, and the breakdown of the 48 articles by outlet.
7. Unavailable primary coding dataset prevents verification of empirical counts — Critical
several percentages in the original table do not map consistently onto a denominator of 48. Because the underlying coding sheet was not available for independent verification, these source percentages are not reproduced as quantitative findings here. This conservative treatment avoids presenting potentially
The study's empirical claims rely entirely on aggregate counts adapted from an earlier Master's dissertation, yet the text discloses that the underlying article-level coding sheet is unavailable for verification and already led to the removal of the information-source table due to denominator inconsistencies. Relying on an inaccessible secondary spreadsheet in which one of the core variables already exhibited unresolvable mathematical errors undermines confidence in the remaining frequency distributions and prevents basic cross-tabulations (such as checking whether protagonist distributions differ between medical and recreational stories or across outlets). To substantiate the empirical claims, the authors should retrieve or reconstruct the article-level dataset from the 48 published texts, verify all counts directly against the primary articles, and make the anonymised article list and coding sheet available.
XSci (AI review)
Summary
The manuscript reports a quantitative content analysis of 48 articles on cannabis published by Público, Expresso and Observador between 1 January 2019 and 30 April 2020, with items whose principal angle was criminality deliberately excluded. Each article was coded for genre, information source, themes, principal protagonist, tone and framing. Medicinal cannabis was the main theme in 31 articles (64.6%), 29 (60.4%) were coded neutral in tone and 39 (81.3%) informative in framing, while public institutions were the principal protagonists in 15 articles (31.3%), private institutions and corporations in 10 each (20.8%), and consumers and health professionals in one each (2.1%). From this the authors argue that a predominantly neutral tone should not be read as evidence of pluralistic representation, because the distribution of speaking roles remains institutionally centred.
Contribution and Significance
The empirical contribution is a first systematic description of how three Portuguese national outlets covered cannabis during the implementation of Law No. 33/2018 and Decree-Law No. 8/2019; no comparable Portuguese corpus appears in the published literature, and the medical-theme dominance in a non-crime-led corpus is a usable national datum. The analytical contribution the authors claim is the separation of information source from principal protagonist (Section 2), and the argument that tone and pluralism are distinct properties of a text.
That analytical claim is weaker than the manuscript presents it. The proposition that institutional and official actors occupy the authoritative speaking positions in news, and that neutral language does not imply a plural distribution of voices, is a founding result of the sourcing literature — Sigal, Reporters and Officials (1973); Hall, Critcher, Jefferson, Clarke and Roberts, Policing the Crisis (1978), on primary definers; Gans, Deciding What's News (1979). None of it is cited, and without that engagement the novelty claim is not sustainable. The nearest national precedent for the design is also absent: Langbecker and Catalan-Matamoros (2021), "Who guides vaccination in the Portuguese press? An analysis of information sources", International Journal of Environmental Research and Public Health, 18(4), 2189, which coded 670 sources across 300 Portuguese press items and found governmental organisations and professional associations dominant. What remains genuinely new is the Portuguese cannabis corpus itself — a modest but real increment.
Strengths
- The manuscript withholds the source percentages rather than publish figures it cannot verify (Section 5.2 and the note to Table 1), an unusually disciplined decision.
- It states plainly that no intercoder reliability coefficient exists and that the coding "should be understood as a structured analysis by the original researcher" (Section 4), and discloses the dissertation provenance of the corpus (Sections 1 and 7).
- Inferential restraint is consistent and well judged: Sections 5.3, 6 and 7 repeatedly decline causal, intentional and audience-effect conclusions that descriptive coding cannot license.
- The reported counts are internally consistent: every distribution in Table 1 sums to 48, and the derived figure that news and brief news together account for 56.3% matches 27 of 48.
Recommendation
Major Revision for Observatorio (OBS*). The corpus is worth publishing, but one of the two headline variables is withheld by the authors' own account, the other rests on single-coder judgement with no verified agreement, and the central pattern has no baseline against which it can be interpreted. These are remediable within the present approach — the 48 articles are public and can be re-retrieved, double-coded and re-reported — but the work required is substantial, and the novelty claim must be rewritten against the sourcing literature.
Major Concerns
-
The institutional concentration finding has no baseline, and the natural comparison was discarded (Sections 4, 5.2, 6, 7). The headline result is that institutional and organised actors hold 72.9% of principal protagonist roles while consumers and health professionals hold 4.2%. But institutional dominance of authoritative speaking positions is the ordinary condition of national news on almost every topic, and nothing here allows the cannabis figure to be read as high, low or typical. The crime-led articles excluded at the sampling stage were the obvious internal comparison, and would have answered the question Section 6 raises — whether protagonist distribution varies with thematic frame. Section 4 also never reports which archive was searched, how many items the search returned, or how many were excluded as crime-centred, so readers cannot judge what fraction of cannabis coverage the 48 articles represent. Without a contrast and without those counts, the finding supports the already-established general claim that news is institutionally centred, but not the manuscript's specific claim about cannabis or about stigmatised issues. Coding the excluded crime corpus as a comparison category, or contrasting a contemporaneous non-cannabis corpus from the same outlets, would resolve this.
-
The source variable, which the stated analytical contribution requires, is not reported (Section 5.2; Section 7; note to Table 1). Section 2 builds the paper on distinguishing "the source or provenance of information" from "the principal protagonist", and "source diversity" is an abstract keyword. Section 5.2 then states that "several percentages in the original table do not map consistently onto a denominator of 48" and that "the underlying coding sheet was not available for independent verification", so the source figures are withheld. The paper therefore measures only one side of its own distinction and argues the other conceptually; the candour is right, but the central analytical claim ends up asserted rather than demonstrated. Because the 48 articles are published and retrievable, the source variable can be re-coded from the texts; if it cannot, the distinction should be withdrawn and the paper reframed around protagonist prominence alone.
-
No reliability evidence for the variables carrying the claims (Section 4, final paragraph of "Unit of analysis and coding"; Section 7). Tone, framing and principal protagonist are the three most interpretive variables in the scheme, and all three were coded by a single researcher with no double-coding and no agreement statistic; the protagonist variable alone carries the paper's conclusion. The manuscript acknowledges this, but acknowledgement does not substitute for evidence, and at N = 48 the remedy is cheap: have a second coder independently code the corpus or a random 30% subsample and report Krippendorff's alpha, following Lombard, Snyder-Duch and Bracken (2002), Human Communication Research, 28(4), 587–604. The codebook should accompany the article as an appendix.
-
The "framing" variable does not measure framing (Section 4, "Discourse variables"; Sections 5.3 and 6). Section 2 defines framing through Entman (1993) as selection and salience promoting particular problem definitions, causal interpretations, moral evaluations or treatment recommendations. The variable actually coded takes the values informative, explanatory and ironic, which describe the register of the writing, not any of those four functions; Section 5.3 itself notes that the 81.3% informative result is "consistent with the genre distribution", indicating the variable largely tracks genre. Section 6 nonetheless claims the results are "consistent with a framing perspective in which selection and salience extend beyond evaluative language (Entman, 1993)", an inference the coded variable cannot support. Either re-specify the variable as frames in Entman's sense and re-code, or relabel it as rhetorical register and drop the theoretical claim built on it.
Minor Comments
- The three outlets are chosen for contrast and Section 3 devotes four sub-sections to their differing ownership and market positions, yet no result in Section 5 or Table 1 is broken down by outlet.
- Add the closest empirical precedents: Abalo (2019), "Rifts in the hegemony: Swedish news journalism on cannabis legalization", Journalism Studies, 20(11), 1617–1634; and Lynch (2021), "Themes and tones of cannabis news reports and legalization outcomes", Media, Culture & Society.
- No codebook or category definitions are supplied, leaving "neutral" (60.4%) and "balanced" (35.4%) tone undistinguished.
- The category "corporation—such as parties, associations or other organised bodies" (Section 5.2) overlaps with both institutional categories and places political parties apart from "political/governmental figure" (4.2%).
- The secondary-theme counts sum to exactly 48, implying one code per article, although Section 4 uses the plural; state whether coding was single- or multi-label, and add these rows to Table 1.
- That no article was coded clearly negative or clearly positive deserves comment, since the crime-led exclusion may have removed the negative pole by construction.
- One-decimal percentages for cells of n = 1 (2.1%) overstate precision at N = 48; lead with counts and, if percentages are kept, give confidence intervals.
- Section 3 is long relative to its analytic yield and rests on 2025–2026 indicators for a 2019–2020 corpus the manuscript says cannot be projected backwards; compress it and substitute period-appropriate figures.
- Section 7 still contains a note to the author — "Before submission, the author should confirm the public repository status of the dissertation and ensure that the article is sufficiently revised" — which belongs in the cover letter, with the check completed.
- Deposit the article-level coding sheet in an open repository; this would also address the verification gap described in Section 5.2.
- Correct the stray asterisks at "*Platforms and audiences." (Section 3) and "Analysis. Frequencies" (Section 4), the broken sentence "The corpus included items directly / addressing cannabis" (Section 4), and the misplaced italic markers in the reference list (for example "43(4)").
- Consider extending the period or adding a fourth outlet, since conclusions about representational pluralism drawn from 48 articles are fragile to small changes in sampling.