{
"papers": [
{
"n": 27,
"doi": "10.1016/j.neuron.2017.12.037",
"value": "0.97",
"method": "retrograde tracer + cell counting",
"metric": "binary inter-areal connection density",
"n_analyzed": 27,
"ci_or_error": null,
"text_access": "abstract_only",
"n_definition": "retrograde tracer injections",
"scope_region": "mouse cortex (47-area parcellation)",
"study_system": "Mouse cortex (19 of 47 parcellated areas)",
"taxonomic_level": "area-to-area",
"scope_population": "inter-areal connections (E projections)",
"value_source_sentence": "The cortical network has a density of 97%, considerably higher than the 66% density reported in macaques.",
"experimental_conditions": "retrograde tract tracing, flat-mount histology"
},
{
"n": 1000,
"doi": "10.1038/s41586-019-1716-z",
"value": "shallow hierarchy across 43 isocortical areas + thalamus",
"method": "Cre-driver AAV anterograde + axon-pattern hierarchy fit",
"metric": "qualitative hierarchy depth",
"n_analyzed": 1000,
"ci_or_error": null,
"text_access": "abstract_only",
"n_definition": "Cre-driver anterograde tracer experiments",
"scope_region": "mouse isocortex + thalamus",
"study_system": "Mouse, Allen Mouse Brain Connectivity Atlas",
"taxonomic_level": "area + cell class",
"scope_population": "cell-class-specific projections",
"value_source_sentence": "Our results show that cell-class-specific connections are organized in a shallow hierarchy within the mouse corticothalamic network.",
"experimental_conditions": "AAV anterograde tracing"
},
{
"n": null,
"doi": "10.1162/netn_a_00345",
"value": "85–90%",
"method": "machine-learning prediction on weighted graph",
"metric": "weighted-link prediction accuracy (mouse)",
"n_analyzed": null,
"ci_or_error": null,
"text_access": "abstract_only",
"n_definition": "inter-areal link entries in the weighted matrix",
"scope_region": "mouse + macaque cortex (inter-areal)",
"study_system": "Mouse + macaque inter-areal cortical matrix",
"taxonomic_level": "area-to-area",
"scope_population": "weighted links",
"value_source_sentence": "Weighted medium and strong links are predictable with an 85%-90% accuracy (mouse) and 70%-80% (macaque), whereas weak links are not predictable in either species.",
"experimental_conditions": "ML imputation from retrograde tract-tracing data + projection-length"
}
],
"audit_issues": [
{
"dimension": "metric_definition",
"description": "Row 1 reports binary inter-areal density = 0.97 (Gămănuţ); row 2 reports a qualitative 'shallow hierarchy depth' (Harris); row 3 reports weighted-link prediction accuracy = 85–90% (Beul). These are three different scalars on different graphs.",
"entries_affected": [
"10.1016/j.neuron.2017.12.037",
"10.1038/s41586-019-1716-z",
"10.1162/netn_a_00345"
]
},
{
"dimension": "study_system",
"description": "Graphs differ: 47-area binary mouse cortex parcellation, 43-area Cre-driver AAV cortex+thalamus, ML-imputed mouse+macaque weighted matrix. Density values are computed on non-identical underlying matrices.",
"entries_affected": [
"10.1016/j.neuron.2017.12.037",
"10.1038/s41586-019-1716-z",
"10.1162/netn_a_00345"
]
}
],
"audit_verdict": "SPLIT",
"comparison_id": "mouse-cc-graph-density",
"comparison_name": "Mouse vs macaque inter-areal cortico-cortical graph density",
"comparison_type": "convergent evidence",
"what_it_reveals": "Three lines of analysis converge on the same global picture of the mouse cortico-cortical weighted graph: dense, log-normal-weighted, and rule-based but shallowly hierarchical.",
"homogeneity_check": {
"caveats": [
"Gămănuţ et al. report binary density on a 47-area mouse parcellation; Harris et al. operate on a Cre-driver tracer dataset across 43 isocortical areas; Beul et al. work on inter-areal weighted matrices in both mouse and macaque.",
"Numerical comparison should not equate '97% density' (binary) with the '85–90% weighted-link prediction accuracy' — these are different statistics on overlapping but non-identical matrices."
],
"n_definition_uniform": "false",
"scope_region_uniform": "false",
"taxonomic_level_uniform": "false",
"scope_population_uniform": "false"
},
"suggested_plot_type": "grouped bar",
"mandatory_caption_caveats": [
"Row 1's 0.97 is a binary graph density; row 3's 85–90% is a weighted-link prediction accuracy. The two scalars are not the same quantity and should not be plotted on a shared y-axis.",
"Row 2 reports a qualitative claim (shallow hierarchy depth) with no scalar to plot.",
"Phase 7 writer: SPLIT verdict — implement as: Convert to a table summarising each study's graph definition and scalar; or split into a binary-density panel and a separate prediction-accuracy panel."
]
}