← Wiki Index / Archaeology & Anthropology / Harris Matrix
🏺 Archaeology & Anthropology

Harris Matrix

The Harris matrix is a topological diagramming method invented by Edward C. Harris in 1973 that represents the relative chronological sequence of archaeological stratigraphic units and interfacial surfaces as a transitive-reduced directed acyclic graph.

Harris Matrix (also called the Winchester–Harris Matrix) is a formal topological method and two-dimensional schematic diagram used in archaeology to represent the four-dimensional physical and chronological relationships among all stratified deposits, architectural features, and non-material erosion or excavation surfaces ("cuts") at an archaeological site. Invented in February 1973 by the Bermudian archaeologist Edward C. Harris while analyzing the complex urban sequence of the Lower Brook Street excavations in Winchester, England (and codified in his 1975 World Archaeology paper and 1979 monograph Principles of Archaeological Stratigraphy), the method models an archaeological site as a partially ordered set (poset) rendered as a transitive-reduced directed acyclic graph (DAG) or Hasse diagram. In a Harris matrix, every individual stratigraphic unit (context) is enclosed in a numbered rectangular box and connected by vertical and horizontal lines so that younger contexts sit above older contexts according to their relative order of creation. Together with single-context recording, the Harris matrix transformed modern field excavation from vertical section-wall drawing into three-dimensional stratigraphic disassembly and now serves as the topological backbone for Bayesian chronological modeling of radiocarbon dating, dendrochronology, and luminescence dating sequences.

Stratigraphic Concept / ElementPhysical Nature on SiteGraph-Theoretic Representation in the Harris MatrixChronological Role
Positive Depositional UnitSoil layer, pit fill, hearth ash, floor makeup, masonry wall, post-packingRectangular node [Context #] with 3D volumeRecords an accumulation or construction event
Negative Feature Interface ("Cut")Pit cut, robber trench, posthole cut, foundation trench, ditch truncationRectangular node [Cut #] with zero volume (2D boundary surface)Marks an episode of digging/destruction that postdates all layers it cuts and predates all fills inside it
Superpositional EdgeContext A directly overlies or fills Context B (A is younger than B)Single vertical line connecting [A] above to [B] belowEnforces strict inequality t(A) < t(B) in years Before Present
Stratigraphic Correlation (===)Two separated deposits (C and D) once formed a single continuous layerDouble horizontal bar (===) linking [C] and [D] at the same levelEnforces contemporaneity t(C) = t(D) after truncation by a later pit
Multilinear BranchingContexts in separate rooms or pits with no physical contact between themParallel independent vertical branches between shared horizon boundariesPreserves partial order without forcing artificial relative dates

From Wheeler–Kenyon Baulks to Single-Context Stratigraphy

Prior to the 1970s, scientific stratigraphic excavation was dominated by the Wheeler–Kenyon box-grid system developed by Sir Mortimer Wheeler and Dame Kathleen Kenyon at Maiden Castle, Verulamium, and Jericho. Excavators dug a checkerboard of square trenches separated by standing unexcavated earthen walls (baulks, typically 0.5 to 1.0 m wide), relying on vertical profile drawings of the four baulk faces to reconstruct the site's stratigraphy.

When large-scale rescue excavations began in deeply stratified European cities such as Winchester and London in the late 1960s—where Roman, Anglo-Saxon, and medieval pits, wells, cellars, and robber trenches intersected across thousands of discrete contexts—the Wheeler–Kenyon grid broke down:

  • Most small features (such as postholes, stake-holes, shallow hearths, and short wall foundations) lay wholly between the standing baulks and never appeared in any vertical section drawing, while crucial intersections between pits were frequently hidden inside the unexcavated baulks themselves.
  • Switching to open-area excavation (pioneered by Philip Barker at Wroxeter and Martin Biddle at Winchester) removed the baulks, requiring a non-pictorial mathematical notation capable of tracking thousands of three-dimensional stratigraphic relationships across an entire city block.

Harris solved this problem by separating the topological sequence of events from the two-dimensional geometry of section drawings and introducing a standardized Stratigraphic Unit Sheet on which the excavator records each context's direct physical relationships (Above / Overlies, Below / Underlies, and Equal to / Correlated with).

The Four Laws of Archaeological Stratigraphy

While geological stratigraphy since Nicolas Steno (1669) had relied on three laws developed for vast, lithified sedimentary basins, human-made archaeological strata include standing masonry walls, dug-out pits, and artificial room enclosures that violate simple geological assumptions. Harris (1979) therefore adapted Steno's first three axioms to anthropogenic deposits and formulated a fourth axiom unique to archaeological graph theory:

  1. Law of Superposition: In a series of layers and interfacial features, as originally created, the upper units of stratification are younger and the lower units are older, for each must have been deposited on, or created by the removal of, a pre-existing mass of archaeological stratification.
  2. Law of Original Horizontality: Any archaeological layer deposited in an unconsolidated form will tend toward a horizontal disposition. Strata which are found with tilted surfaces were originally deposited that way, or lie in conformity with the contours of a pre-existing basin of deposition (such as the sloping sides of a ditch or pit).
  3. Law of Original Continuity: Any archaeological deposit, as originally laid down, or any interfacial feature, as originally created, will be bounded by a basin of deposition (such as a room wall or pit edge), or will thin down to a feather-edge. Therefore, if any edge of a deposit or interfacial feature is exposed in a vertical view with a truncated thick face, a part of its original extent must have been removed by later excavation or erosion, and its continuity must be sought—or its absence explained—in the stratigraphic sequence.
  4. Law of Stratigraphical Succession (Harris, 1979): A given unit of archaeological stratification takes its place in the stratigraphic sequence of a site from its position between the undermost (or earliest) of the units which lie above it and the uppermost (or latest) of all the units which lie below it and with which the unit has a physical contact, all other superpositional relationships being redundant.

Graph Theory: Transitive Reduction and Interfacial Units ("Cuts")

Transitive Reduction of Physical Contacts into a Hasse Diagram

The Law of Stratigraphical Succession is the direct archaeological statement of transitive reduction in mathematical graph theory. Suppose an ancient pit cut cuts down through three successive horizontal floors: an upper Roman mosaic [10], a middle clay floor [20], and a lower gravel surface [30], where [10] overlies [20] and [20] overlies [30]. In the trench, the pit fill [5] touches all three floors ([10], [20], and [30]).

If an archaeologist drew lines for every physical contact ([5] → [10], [5] → [20], and [5] → [30]), a site with 2,000 contexts would become an unreadable tangle of crossing lines. By the Law of Stratigraphical Succession, because [10] is already above [20] and [20] is above [30], the transitive property of time (t5 < t10 < t20 < t30) ensures that [5] is younger than [20] and [30] once we draw the single link from [5] to the latest unit below it ([10]). Pruning all redundant transitive edges (u → w whenever a path u → v → w exists) produces the unique Hasse diagram of the site's partial order. Furthermore, if a cycle ever appears in the graph ([A] above [B] above [C] above [A]), the directed acyclic graph condition is violated, alerting the field team to an excavation or recording error (such as an unrecognized animal burrow or undercutting feature).

Non-Material Interfacial Units ("Cuts" and Destruction Surfaces)

Before Harris, archaeologists treated a pit or ditch as a single unit ("Pit 1") and assigned all pottery found inside it to the creation of the pit. Harris demonstrated that every pit, posthole, grave, or robber trench actually comprises at least two distinct stratigraphic events:

  1. The Feature Interface ("Cut"): The negative act of digging out and removing earlier soil, which creates a three-dimensional boundary surface of zero physical volume ([Cut 15]). The digging of the cut destroys earlier strata (carrying old, residual artifacts upward into the spoil heap) and marks the actual construction or activity event.
  2. The Backfill Deposit(s): The positive soil layers ([Fill 14], [Fill 13]) that subsequently wash or are dumped into the open hole—sometimes days, decades, or centuries after the hole was dug.

Giving every negative cut its own context number and node in the Harris matrix resolved the classic paradox where the soil at the bottom of a deep medieval well lies topographically lower (deeper below modern ground level) than the Neolithic layers through which the well shaft was cut, yet sits stratigraphically higher (younger) than those Neolithic layers in the Harris matrix because it overlies the medieval well-cut interface.

Integration with Bayesian Radiocarbon Chronological Modeling (OxCal)

By itself, a Harris matrix expresses only relative time (ordinal ranking: earlier vs. later) without specifying whether the gap between two contexts lasted five minutes or five centuries. Conversely, calibrated radiocarbon dating probability distributions are non-normal and multimodal due to wiggles in the atmospheric IntCal20 calibration curve, often spanning 150 to 300 calendar years at 95.4% confidence (2σ).

Starting in the 1990s, statistician Caitlin E. Buck and physicist Christopher Bronk Ramsey (creator of OxCal) united the Harris matrix with Bayesian Markov Chain Monte Carlo (MCMC) chronological modeling via Bayes' theorem:

P(θ | X, M) ∝ P(X | θ) × P(θ | M)

  • Likelihoods P(X | θ): The individual calibrated radiocarbon, luminescence dating, archaeomagnetic dating, or coin-date probability distributions measured on samples from specific contexts.
  • Prior Beliefs P(θ | M): The topological partial order of the Harris matrix (M), translated directly into OxCal Sequence blocks (for contexts linked vertically in superposition, enforcing t1 > t2 > t3) and unordered Phase blocks bounded by Boundary parameters (for separate branches between two horizon-wide floors or destruction surfaces).

When MCMC sampling rejects all calendar-age combinations that violate the Harris matrix order—and identifies residual (reworked older charcoal) or intrusive (burrowed younger charcoal) outliers via Agreement Indices (Amodel > 60%)—the 95% Highest Posterior Density (HPD) intervals routinely shrink from ± 150 years down to ± 15–35 calendar years, enabling generational-scale histories of Neolithic monuments and urban settlements (Bayliss, 2009).

Frequently Asked Questions

Why can't physical depth below the ground surface ("arbitrary spit digging") replace a Harris matrix?

Digging an archaeological site in horizontal 10-cm arbitrary levels ("spits" or metric stratigraphy) assumes that all artifacts found at the same elevation below a surveyor's datum were deposited at the same time. On any real site where prehistoric or historical inhabitants terraced slopes, heaped up mounds, or dug pits, graves, and foundation trenches into earlier layers, a medieval pit fill at 2.0 meters depth can sit physically deeper than an undisturbed Bronze Age floor at 1.2 meters depth just one meter away. Mixing objects by depth mixes centuries of unrelated material, whereas a Harris matrix follows the actual physical boundaries and interfaces in the order they were formed.

What is Single-Context Recording and how does it relate to the Harris matrix?

Developed in the mid-1970s by the Department of Urban Archaeology (DUA) at the Museum of London based directly on Harris's principles, single-context recording requires the excavator to define, plan to scale (on a dedicated transparent permatrace sheet or 3D photogrammetric/total-station mesh), photograph, and fully excavate each individual stratigraphic unit in reverse order of deposition (latest first). Because every context has its own closed boundary polygon and its own row in the Harris matrix, researchers can reconstruct three-dimensional phase plans and cross-sections along any arbitrary axis after excavation is complete.

How are standing buildings analyzed with a Harris matrix?

In buildings archaeology (Bauforschung), masonry walls, blocked doorways, inserted windows, roof-line scars, and plaster renders obey the exact same four stratigraphic laws as buried soil deposits. A window cut chiseled through an existing medieval stone wall is a negative interfacial cut; the brick blocking later mortared into that window is a positive fill overlying the cut; and a coat of lime plaster covering both the original wall and the brick blocking is a superimposed layer that postdates both. Constructing a mural Harris matrix reveals every phase of construction, remodeling, and repair without dismantling the standing monument.

References

  1. Harris, E. C. (Jun 1975). The stratigraphic sequence: A question of time. World Archaeology, 7(1), 109–121. https://doi.org/10.1080/00438243.1975.9979624
  2. Harris, E. C. (Jun 1979). The laws of archaeological stratigraphy. World Archaeology, 11(1), 111–117. https://doi.org/10.1080/00438243.1979.9979753
  3. Buck, C. E., Kenworthy, J. B., Litton, C. D., & Smith, A. F. M. (Dec 1991). Combining archaeological and radiocarbon information: A Bayesian approach to calibration. Antiquity, 65(249), 808–821. https://doi.org/10.1017/S0003598X00080534
  4. Bronk Ramsey, C. (2009). Bayesian analysis of radiocarbon dates. Radiocarbon, 51(1), 337–360. https://doi.org/10.1017/S0033822200033865
  5. Bayliss, A. (2009). Rolling out revolution: Using radiocarbon dating in archaeology. Radiocarbon, 51(1), 123–147. https://doi.org/10.1017/S0033822200033750
🧬 Support Open Science
Explore interactive science games on LowKey Games or browse scientific apparel & accessories in the CRV Science Shop.
💡 Spot an inaccuracy or have primary literature to expand this article?
Submit a citation or revision note using your shared CRV Science / LowKey Games profile ID.
Suggest Revision