Do open-access or openly readable studies show that placing a subject on rule-of-thirds points makes a picture preferred or attended to more?

In the sources actually read, evidence for a general preference for thirds placement is weak and mixed. A Palmer-lab conference abstract reports that single-object placement at the frame center, not at third-line intersections, was most preferred; a small eye-tracking study (30 participants) reports experts favoured thirds images while novices did not, but its reported statistic is internally inconsistent. A 2024 open-access eye-tracking study shows that leading lines, another composition cue, did change attention. Treat thirds as a heuristic with limited support, not a measured requirement. The Amirshahi et al. 2014 paper could not be read. Curator qualifications: 6 of 6 excerpts present. Torabi: '30 participants, 15 with photography education (Experts) and 15 without (Novices)'; 'An independent sample t-test was conducted to compare whether experts selected the rule of thirds or non-rule of thirds ... t(28)=3.33, p =0.02'; novices 'M=52,2249' (malformed number) ... 't(28)=0.46, p=0.651'. Curator arithmetic: t(28)=3.33 gives two-tailed p = 0.00245 (one-tailed 0.00122), so p = 0.02 is inconsistent with the statistic. Curator inference: comparing thirds vs non-thirds within the same 15 experts with df = 28 treats repeated measures as two independent samples; the test model does not match the design as described. Palmer-lab abstract: centre most preferred 'for every object in every facing direction'; 'although the ROT is empirically incorrect, it may be heuristically useful'. Evidence status: statements located in the cited sources by a correlated desk review (Claude); not executed; not independently reviewed.

{
  "citations": [
    {
      "url": "https://palmerlab.berkeley.edu/spatial.html",
      "note": "palmerlab.berkeley.edu/spatial.html, Palmer, Hara, Griscom VSS 2014 abstract (equals pin); Torabi & Teeravarunyou PDF (equals pin), abstract, method, Results 3.1; Europe PMC PMC11763225 fullTextXML (equals pin), abstract; two-tailed p for t(28)=3.33 by numerical integration of the t density (Python, curator)"
    },
    {
      "url": "https://soad.kmutt.ac.th/wp-content/uploads/2021/06/2021INTC3.pdf",
      "note": "palmerlab.berkeley.edu/spatial.html, Palmer, Hara, Griscom VSS 2014 abstract (equals pin); Torabi & Teeravarunyou PDF (equals pin), abstract, method, Results 3.1; Europe PMC PMC11763225 fullTextXML (equals pin), abstract; two-tailed p for t(28)=3.33 by numerical integration of the t density (Python, curator)"
    }
  ]
}