BGPT: Paper Review: Press Releases Shape Online Attention for Neuroscience Articles

Fuel Your Discoveries

Quick Explanation Copied

Core takeaway

Across retrospective (n=3,129) and a prospective pre/post experiment (n=25), the paper reports that issuing a press release (PR) is the dominant driver of spikes in news mentions and therefore a large increase in Altmetric Attention Score (AAS) for neuroscience articles, while isolated X posting shows limited or no additional effect after the PR is accounted for.

Critical note: the causal claim rests on a prospective within-center design, but the altmetric instruments (Altmetric’s source detection, language coverage, and weighting) and the limited prospective sample size create important uncertainty about generalizability and effect sizes beyond this setting.

Long Explanation

Paper review (science-communication impact metrics)

Paper: “Press Releases Shape Online Attention for Neuroscience Articles” .

1) Visual evidence map (what the paper measured)

Data sources & pipeline (as reported)

Retrospective: authors gathered neuroscience outputs via Scopus (affiliations; Dec 2021), pulled journal metrics from WoS JCR, and used Altmetric API for news/X/Facebook/Mendeley/AAS; PlumX citations were later discarded for mismatch in the workflow.
Second (in-house retrospective): 201 English original papers (2016–2021) with known PR vs NPR status (PR group n=28; NPR group n=173).
Prospective: 25 INc-UAB principal-authored articles (2022–2023) assigned to PR’ or NPR’ and monitored across time windows with manual Google checks for undetected news items.

2) Key quantitative results (replottable from the text)

3) What the results do and don’t establish

3.1 Retrospective correlations: “attention tracks outcomes” but causality is unresolved

The paper reports that citations correlate positively with multiple altmetric components (including X posts, Mendeley readers, AAS) and that journal impact factor (IF) correlates with citations and AAS.
These correlations do not separate “PR increases attention” from “newsworthy papers are more likely to receive PRs.” The paper itself notes this alternate explanation from prior PR/news literature and frames the need for prospective designs.

3.2 In-house retrospective: strong PR-vs-NPR separation persists after IF adjustment

The authors report PR articles showing dramatically higher means in news mentions, X mentions, Mendeley readers, and AAS relative to NPR articles (and also report statistical tests controlling for IF).

3.3 Prospective pre/post experiment: PR timing aligns with the “first-week spike” in news and AAS

The paper reports that after a ≥22-day wait, articles with PR begin receiving news mentions shortly after PR issuance, while NPR’ +22 articles receive none (after excluding Altmetric false positives).
They further report AAS increases within the first week are large in PR’ +22 articles (reported average +68.3) versus minimal in NPR’ +22 (+0.6), and that the interaction TIME×PR is statistically significant in ANCOVA/ANOVA style tests.

4) Skeptical critique: uncertainty, measurement limits, and plausible blind spots

4.1 Altmetric detection is incomplete and language-dependent (measurement bias)

The paper reports that Altmetric missed a large fraction of digital news mentions (authors’ manual Google searches found 105 overlooked mentions, said to be 63% of the total), and the distribution of “detected vs undetected” differs by whether stories cite the article and by language.
More broadly, prior work has examined precision/recall and differences among Altmetric news sources, supporting the possibility that “AAS” differences may partly reflect coverage/eligibility differences rather than attention magnitude.
The paper also notes Altmetric’s AAS is a composite weighting system and that the algorithm details (e.g., source influence and posting patterns) are not fully disclosed.
Altmetric has published explanations of how the AAS is calculated and how it incorporates different attention sources, which provides context for why news mentions can dominate score changes relative to social posts.

4.2 Selection and confounding: PR issuance isn’t randomized at the publication level (even in “prospective” work)

In study 3, the allocation into PR’ vs NPR’ is based on a “fork decision tree” incorporating general interest, author preference, and delay to campaign start; that can introduce confounding between “newsworthiness” and measured outcomes.
The authors argue for within-publication pre/post comparisons to mitigate some confounding, but unmeasured factors (e.g., concurrent media narratives, journal promotion outside the institution, or external societal attention) could still affect early-week spikes.

4.3 Social media mechanisms: “mentions” ≠ “engagement” ≠ “reading” ≠ “citing”

The paper reports limited impact of isolated X actions on AAS/news mentions in the prospective test, but it also notes Altmetric’s exclusion rules for posts that don’t explicitly cite the publication and highlights that engagement (likes/replies/clicks) wasn’t uniformly captured in the AAS weighting.
External literature has also raised the “empty tweeting” concern (tweets may be low-information or not translate into deeper diffusion), supporting caution when interpreting counts of social posts as proxy for knowledge uptake.

5) Practical implications (limited to what is directly supported)

If the goal is media/news visibility as captured by Altmetric’s news mention component, the paper’s reported prospective association is consistent with PR being a trigger for those mentions.
However, because Altmetric news detection is incomplete and language-dependent, “score increases” should be treated as instrument-dependent measures, not direct measures of “public attention” in an absolute sense.

6) Authorial transparency & conflict-of-interest (from provided text)

The abstract metadata provided indicates a salary/role relationship for one author (Eduscopi) and states that interactions with Altmetric/EurekAlert!/publishers were informational and did not influence outcomes.

7) Buttons for deeper BGPT exploration

Author review links (BGPT)

Feedback:

Updated: March 29, 2026