Hogback Stones in northern Britain — a Wikidata exploration

This notebook queries Wikidata for hogback stones — a distinctive group of tenth- and early-eleventh-century carved stone grave-covers from northern England and southern Scotland. Four views build on each other: a marker map, a hex-binned density grid revealing the core areas, a site-level bar chart that reconstructs Williams’ (2015) observation that a handful of sites produce the bulk of the corpus, and a completeness chart comparing how much of Wikidata’s modelling machinery is actually populated for this small dataset.

The data is curated by the Wikidata WikiProject Hogback; see Wikipedia on Hogback (sculpture) for the wider archaeological background.

Note

On first load, your browser downloads the Python runtime (Pyodide, ~10 MB). Please allow a moment for it to initialise.

About this notebook

Why this dataset?

Hogback stones are a small, well-bounded corpus — around four dozen modelled items in Wikidata at time of writing — which makes them ideal for teaching analytical techniques that need less heavyweight presentation than a large dataset:

  • Small-n archaeology as a teaching case — Williams (2015) stresses that hogbacks crop up in “discrete clusters” along maritime networks, with three or four sites dominating the surviving corpus. A forty-item dataset makes that clustering visually immediate; the same pattern in a dataset of four thousand points would wash out in a density heatmap.
  • Label-driven site extraction — item labels follow semi-regular patterns (“Hogback von Lythe”, “Gosforth Hogback 1”, “Govan Stone 02 (Hogback)”), which lets us reconstruct site-level groupings from strings. This is not how you would do it in production (a proper P276 location join would be cleaner), but it shows honestly what happens when an ontology is under-used and you have to recover structure from labels.
  • Cross-referenced to Commons and OSM — like the holy wells, most hogbacks link to a Wikimedia Commons image and an OSM node, so the citizen-science completeness question is as pertinent here as there.

Data-context notes

A few features of the result worth flagging before we load it:

  • There are more items than photographed monuments: some items represent a group (e.g. St Helen's Church, church & hog-backed monuments is a single Wikidata item for a multi-monument site).
  • The 43 items are not evenly distributed — core sites (Brompton, Lythe, Sockburn-on-Tees, Govan) dominate, matching Williams’ core-area observation. This will be visible on both the hex grid and the site-level bar chart.
  • Coordinates are returned as Point(lon lat) — WKT with longitude first. A defensive regex parser is used, same as in the holy-wells notebooks.

Tooling notes

Same browser stack as the holy-wells notebooks: pyodide.http.pyfetch for SPARQL (Pyodide cannot use requests or SPARQLWrapper), Leaflet via inline HTML for the two maps, matplotlib for the bar charts.

Step 1 — Define the SPARQL query

The query is short because the modelling of hogbacks in Wikidata is straightforward: anything that is wdt:P31 wd:Q1570646 (instance of hogback) counts, and we pull three optional enrichment properties alongside the label.

Step 2 — Load the data

Because of the [AUTO_LANGUAGE],mul,en label fallback, labels come back in whatever language the browser prefers, with mul and en as backup. The site-extraction step below handles both the German “Hogback von X” and the English “X Hogback” / “X hogback NN” patterns with a small regex cascade.

Step 3 — Visualise

Step 3a — Marker map with popups

Each hogback is a clickable circle marker. The popup shows the item’s label, a thumbnail of its Wikimedia Commons image (if one exists), and hyperlinks to Wikidata and OpenStreetMap.

Step 3b — Hex-binned density grid

Hogback sites cluster sharply — a handful of core locations (Brompton, Lythe, Sockburn-on-Tees, Govan, Gosforth) account for most of the surviving corpus. With only ~40 points spread across northern Britain, a smaller hex size would produce mostly singletons; we use a larger cell (~0.4° ≈ 30 km at 54°N) so that multiple items at one site fall into the same hex and the clustering message comes through.

Step 3c — Items per site

Williams (2015) and earlier surveys point to three or four locations as the “core” of hogback production and use: Brompton, Lythe and Sockburn-on-Tees in northern England, with a further concentration at Govan in southern Scotland. If the dataset is well-modelled, a simple GROUP BY site over the item labels should reproduce that shape. The site names below are extracted from labels with the regex from Step 2 — so the particular counts depend on a string trick, but the shape of the distribution does not.

Step 3d — Completeness of cross-references

How much of the auxiliary Wikidata information is actually present on these items? As with the holy wells, this is the question a data steward asks before drawing quantitative conclusions from a citizen-science dataset.

Step 4 — Explore

The DataFrame df stays in scope. One starting point: list all items at a specific site.

Part of an Open Educational Resource series on knowledge graphs and linked open data, produced in the context of NFDI4Objects. Data: Wikidata WikiProject Hogback, CC0. Tiles: OpenStreetMap contributors, Esri. Images: Wikimedia Commons contributors. Archaeological background follows Williams, H. 2015, ‘Hogbacks: the Materiality of Solid Spaces’ in Early Medieval Stone Monuments: Materiality, Biography, Landscape, Boydell, pp. 241–68.