DOI: 10.5281/zenodo.22893975 • Working Paper WX-WP-2026-05

Standardizing the Dead Wax: A Unified Taxonomy for Phonographic Matrix Data

• Josie Jefferson & Felix Velasco (Unearth Heritage Foundry / Waxlore Collective)
with technical synthesis collaboration from Claude 4.6 (Opus) & Gemini (3.1 Pro)

Published: September 22, 2026

Immutable Zenodo DOI Record

Abstract Epistemic Scope

The runout area of every vinyl record contains an archive of manufacturing metadata. Matrix numbers, stamper codes, and engineer signatures form a physical record of a century of industrial production. These inscriptions are primary evidence for recording history, factory operations, and provenance. Yet no formal architecture exists to interpret them. The absence of an institutional standard — the "Matrix Gap" — leaves the largest runout databases, notably Discogs, unable to support rigorous research. Runout inscriptions are recorded as free-text strings where matrix numbers, stamper codes, and engineer etchings occupy a single flat field. This destroys the semantic relationships among data elements and makes cross-plant and cross-period comparative analysis impossible.

The Waxlore Taxonomy Framework is a three-tiered hierarchical metadata standard designed to categorize phonographic matrix data across all major pressing-plant traditions. Tier 1 (Origin Metadata) captures institutional and geographic identifiers: pressing plant, territory, and corporate entity. Tier 2 (Process Metadata) encodes the electroforming sequence: lacquer generation, mother identity, and stamper generation. Tier 3 (Human Metadata) documents artisan signatures and mastering engineer etchings, with confidence ratings reflecting the strength of the attribution evidence. Comparative case studies of a 1960s UK EMI/Parlophone pressing and a 1970s US Columbia pressing demonstrate that the taxonomy successfully translates disparate inscription systems into unified, interoperable records. This standard ontology is required to deploy computer vision systems that automate dead wax analysis at scale. Doing so requires institutional archives, open-metadata platforms, and the research community to adopt the Waxlore standard.

Keywords: Phonographic Metadata, Dead Wax, Matrix Number, Runout Etching, Industrial Archaeology, Data Ontology, Archival Metadata Standards, Pressing Plant Identification, Electroforming, Mastering Forensics, Vinyl Provenance, Computer Vision, Cultural Heritage Informatics, Non-Proprietary Open Standard, Archaeobytology, Vivibyte, Waxlore
Full Manuscript
Waxlore Scholarly Edition

I. Introduction: The Inconsistency of Inscribed Data

The phonographic record carries a secondary archive carved into the blank space between the final groove and the center label: the dead wax. This smooth expanse of vinyl is a site of physical, tactile inscription. It bears machine-stamped matrix data, hand-etched stamper codes, and the fast, irregular signatures of mastering engineers—a vast, disorganized corpus of manufacturing scars. These marks trace a heavy industrial logic developed in isolation across dozens of competing factory floors over a century of production. The resulting global landscape is massive, fractured, and unstandardized.

Early disc manufacturers—Berliner, Victor, Columbia—adopted sequential matrix numbers purely as production-control instruments.[1] These systems tracked which master recordings were committed to which metal parts. They were proprietary by necessity; competing firms had no reason to coordinate nomenclature. When Decca formalized its conventions in the 1930s, it ignored parallel systems at His Master's Voice. When overseas subsidiary plants pressed from shipped metal parts, they appended local suffixes, creating branching, undocumented taxonomies.[2] The matrix number was born as a closed industrial signal. It has never been translated into a universal standard.

This fragmentation prevents systematic study. For archivists, researchers, and collectors, the dead wax is not incidental; it is primary evidence.[3] Identifying a mastering facility, tracking a stamper generation, or attributing an engineer's signature carries material significance for historical scholarship and institutional preservation. Yet the infrastructure built to resolve these variables remains informal, consumer-facing, and structurally constrained.[4][5][6]

What is missing is a formal, hierarchical metadata standard capable of structuring matrix data across all historical periods and plant conventions. Drawing on industrial archaeology, the Waxlore Taxonomy Framework treats the record not as a pristine commodity, but as a manufactured artifact whose production history is cut directly into its surface.[7] The framework builds three nested tiers of classification: Tier 1 (Origin Metadata), Tier 2 (Process Metadata), and Tier 3 (Human Metadata).

II. Literature Review: Existing Standards and Archival Gaps

While cultural heritage institutions have spent decades standardizing the description of the music—the sound itself—the physical reality of how that sound was dragged into wax remains unstandardized.

Over the past three decades, cultural heritage institutions have developed robust metadata standards. Libraries, museums, and digital preservation institutions converged around a common set of frameworks: Dublin Core for general resource description, PREMIS for digital preservation metadata, CIDOC-CRM for museum object ontologies, and METS/MODS for digital library systems.[8] In the domain of sound recordings, the International Association of Sound and Audiovisual Archives (IASA) published technical guidelines for cataloging and preserving audio carriers, addressing everything from shellac disc composition to digitization sample rates.[9] The Audio Engineering Society contributed AES57, a standard for describing audio objects in digital archives, which specifies metadata fields for provenance, format, and technical characteristics.[10]

These frameworks fail to address the physical reality of the carrier itself. They describe the content and condition of sound recordings, but ignore the inscribed data. IASA's cataloging guidelines provide fields for label name, catalog number, and matrix number, but treat the matrix number as a single, undifferentiated string. There is no provision for distinguishing a machine-stamped matrix number from a hand-etched stamper code; no controlled vocabulary for identifying pressing plant origin from alphanumeric prefixes; and no way to link a pressing's runout inscriptions to the lacquer, mother, and stamper genealogy. The PREMIS Data Dictionary has no conceptual vocabulary for the physical inscription practices of analog manufacturing.[12] CIDOC-CRM has not been applied to phonographic media, and its event-based ontology would require substantial domain-specific extension to accommodate matrix data.[13]

The result is an infrastructural gap. Scholars working on the material history of recorded sound lack a shared metadata architecture.

Consumer-facing databases operate as de facto research tools across this gap. Discogs permits users to transcribe runout-area inscriptions into a dedicated free-text field, yielding a corpus of runout data unmatched by institutional archives. For many researchers, a Discogs entry is the primary point of reference when attempting to identify a pressing's manufacturing origin. Any serious metadata standard for phonographic inscriptions must be designed to build upon, rather than displace, this foundational work.

The limitations of the Discogs corpus, however, are structural rather than editorial. They are architectural decisions that constrain utility for systematic research. The Discogs data model treats the runout field as a single undifferentiated text block. A contributor transcribing a UK Parlophone pressing from the 1960s might enter a string such as "XEX 606-1N ◇ EMIDISC ◇ G&P" without any structural indication that "XEX 606-1N" is a matrix/take identifier assigned at the EMI mastering suite, that "EMIDISC" denotes the lacquer-cutting system employed, and that "G&P" is the personal etching of engineer Geoff Pesche.[15] All three data points — origin metadata, process metadata, and individual signature — are collapsed into a single flat string, rendering them opaque to programmatic query. A researcher seeking all pressings mastered by Pesche cannot query for his mark without also retrieving every entry in which the letters G and P happen to appear adjacently in an unrelated context.

Furthermore, Discogs imposes no controlled vocabulary on the transcription itself. Contributors must decide independently how to represent non-alphanumeric symbols — the triangles, Diamonds, and proprietary stamps that pressing plants used as internal shorthand. There is no canonical encoding for the EMI "tax code" stamps, no standard representation for the concentric-circle symbol used by certain Columbia plants, and no agreed-upon convention for distinguishing machine-stamped characters from hand-etched ones, a distinction that carries significant evidentiary weight in pressing chronology.[16] The result is that the same physical inscription may appear in dozens of variant transcriptions across different Discogs entries for the same release, with no mechanism for reconciliation.

Most fundamentally, the database architecture organizes around the release — a commercial product defined by its label, catalog number, and tracklist — rather than around the pressing, the physical artifact that actually bears the runout data. A single Discogs release entry may include pressings from multiple plants, stamper generations, and national territories, all carrying different runout inscriptions. The platform provides a "version" system that allows contributors to distinguish between pressings, but version differentiation is voluntary, inconsistently applied, and governed by community norms rather than formal ontological rules.[17] This architecture leaves significant analytical potential unrealized: it subordinates the physical unit of evidence (the pressing) to a commercial abstraction (the release).

The Goldmine Grading Standard addresses condition, not provenance. It provides a controlled vocabulary — a linear scale from Mint to Poor — but completely ignores manufacturing history.[18] A record graded "Near Mint" by Goldmine standards may be a first pressing from an original lacquer or a late-run reissue from a worn stamper; the grading system is indifferent to the distinction. While Goldmine was never intended as a metadata standard in the archival sense, its widespread adoption in marketplace listings has had the practical effect of centering collector discourse on surface condition at the expense of manufacturing history, reinforcing a culture in which the dead wax is treated as supplementary rather than primary evidence.

MusicBrainz, the open-source alternative to Discogs, offers a more rigorously relational data model, with distinct entities for artists, recordings, releases, and release groups linked by typed relationships.[19] Its schema, however, was designed for bibliographic description of musical works and their published manifestations, not for the forensic analysis of physical media. There is no entity class for "pressing," no relationship type for "cut by [mastering engineer]," and no field in which runout data can be entered in a structured, parseable form. A 2019 community proposal to add a "runout etching" annotation type was discussed but never implemented, reportedly due to the difficulty of defining a schema flexible enough to accommodate the heterogeneity of global pressing-plant conventions.[20]

The common structural flaw across all these platforms and standards is a failure to look at the vinyl itself. They fail to treat the pressing as what it materially is: a manufactured object whose production history is stamped directly into its surface. The field does not lack data — the Discogs corpus contains millions of runout transcriptions — but rather lacks a formal schema capable of structuring it. A proper standard must distinguish categories of inscription, encode hierarchical relationships between matrix elements, and provide normalization rules that allow different platforms to share information.

III. The Waxlore Taxonomy Framework

Existing standards and platforms treat runout inscriptions as undifferentiated strings rather than as compound data objects composed of distinct elements arranged in a recoverable hierarchy. The Waxlore Taxonomy Framework addresses this by proposing a three-tiered classification system: Origin, Process, and Human. Each discrete inscription element in the dead wax is assigned to one tier. Tiers are ordered by descending institutional scope: from macro-level manufacturing facility identification, through meso-level electroforming sequence documentation, to micro-level attribution of human agency. This hierarchy is causal, not merely administrative. It reflects the sequential steps governing a pressing's creation. This fidelity to the manufacturing process gives the taxonomy its analytical power and interoperability advantages over flat-file approaches.

A. Tier 1 — Origin Metadata: The Institutional Frame

Tier 1 captures the institutional identity of a pressing: the manufacturing facility where it was produced, the national territory, and the corporate entity authorizing the run. These are the broadest and most stable coordinates in the matrix data hierarchy. A pressing plant, once identified, anchors all subsequent analysis. Without it, the interpretation of lower-tier data elements becomes speculative. Identical matrix prefixes, stamper codes, and engineer signatures carry different meanings depending on the plant.

Tier 1 elements include the pressing plant name and operational period, geographic location, parent corporation, and plant-specific prefixes or embossed symbols that identify the facility. The EMI group demonstrates why Tier 1 must be resolved first. A record bearing the matrix prefix "XEX" can be traced to EMI's London mastering operation, but the physical pressing may have been produced at Hayes, Middlesex; Uxbridge; or any of EMI's overseas subsidiary plants in India, South Africa, or Australia, each of which maintained its own stamper inventory and applied its own secondary markings.[21] Without a controlled vocabulary that distinguishes between the mastering origin in the matrix prefix and the pressing origin in ancillary stamps, the two are routinely confused — a problem endemic to Discogs entries for Commonwealth EMI pressings.

The taxonomy addresses this by requiring users to record Tier 1 data as a structured tuple rather than a single string: {Mastering Facility, Pressing Plant, Territory, Corporate Entity, Operational Period}. Each element in the tuple draws from a controlled registry of known facilities. Where a facility cannot be positively identified, the taxonomy provides for a coded designation of uncertainty (e.g., "PLANT-UNCONFIRMED") rather than an empty field. This preserves the structural integrity of the record while explicitly noting the missing data.[22]

B. Tier 2 — Process Metadata: The Electroforming Sequence

Tier 2 maps the electroforming sequence: the progression of metal parts from lacquer master to final stamper. It tracks the physical chain of replication, locating exactly where variation and degradation enter the manufacturing process. A lacquer is cut from the audio source; from the lacquer a metal father is electroformed; from the father a mother is produced; and from the mother one or more stampers are generated, each of which presses a finite number of discs before yielding to wear.[23] Every generation in this chain is a discrete manufacturing event recorded in the dead wax through stamper-letter codes, generation-count suffixes, and re-order indicators.

Tier 2 captures this generational data in a structured, relational format. Core elements include: lacquer/master identification (typically the root matrix number, often incorporating a take or cut indicator such as "-1A" or "-2B"); mother identification, where recoverable; stamper code (the alphanumeric designator assigned to the specific stamper used for a given run); and generation indicators that encode the number of intermediate steps between the original lacquer and the stamper. The taxonomy further distinguishes between explicit generation data — information directly legible in the runout inscriptions — and inferred generation data derived from cross-referencing known plant practices. For example, the Monarch Records pressing plant used a sequential alphabetic stamper code in which early letters (A, B, C) indicated early stamper pulls from a mother, while later letters indicated replacement stampers; this convention, while undocumented by Monarch itself, has been reconstructed through comparative analysis and is sufficiently well-attested to support probabilistic inference.[24]

The analytical value of Tier 2 data is substantial. In audiophile and archival contexts, stamper generation is a key variable affecting sonic character; a pressing pulled from a first-generation stamper derived from the original lacquer will, all else being equal, exhibit higher fidelity to the master tape than a pressing pulled from a fourth-generation stamper derived from a re-cut lacquer.[25] Yet in existing databases, this information — when it is recorded at all — is flattened into the same undifferentiated text field as plant identifiers and engineer signatures, making systematic analysis of generational distribution across a pressing run effectively impossible. By isolating process metadata in its own tier and imposing a relational structure that links stamper codes to mother and lacquer identifiers, the Waxlore taxonomy enables precisely this kind of analysis for the first time at scale.

C. Tier 3 — Human Metadata: Artisan Signatures and Individual Agency

Tier 3 records individual human interventions. It documents the hand-etched signatures, messages, and irregular scrawls left in the soft lacquer by mastering engineers, plating technicians, and label personnel. These marks are the direct, physical traces of human agency embedded within an automated industrial workflow.

Collectors know these marks intimately. George Peckham, cutting at Apple Studios and later Porky's Prime Cuts, dragged the phrase "A Porky Prime Cut" into the dead wax of hundreds of pressings. In doing so, he established an undeniable, physical signature within the anonymous industrial apparatus of record manufacturing.[26] Robert Ludwig etched "RL" into pressings that indicated a specific, high-velocity mastering approach — as seen on early copies of Led Zeppelin's untitled fourth album, where the "RL" cut was pulled from circulation due to concerns about excessive level.[27] Lee Hulko at Sterling, Wally Traugott at Capitol, and dozens of undocumented engineers across the global pressing industry similarly left identifying marks whose attribution, in the absence of a centralized registry, depends on accumulated collector knowledge.

Tier 3 of the Waxlore taxonomy imposes formal structure on this body of knowledge. Each human inscription element is recorded as a structured object containing: the inscription text as transcribed; the attributed individual, where identification is possible; the role of the individual in the manufacturing chain (mastering engineer, plating technician, quality control, or unknown); and a confidence rating reflecting the strength of the attribution evidence. This last element is critical. The attribution of many engineer etchings rests on decades of collector inference rather than on documentary confirmation from the engineers themselves or their employers. The taxonomy distinguishes three formally defined confidence levels, each with explicit evidentiary criteria:

Confidence LevelDefinitionRequired EvidenceExample
ConfirmedAttribution verified through primary documentation originating from the engineer, the studio, or the employing institution.Published interview or oral history in which the engineer acknowledges the mark; studio session logs or cutting-room documentation identifying the engineer by name; corporate personnel records linking an individual to a facility and time period; direct personal testimony from the engineer or an immediate colleague.George Peckham's "Porky" etchings, confirmed through multiple published interviews and documented in Magee (2012); Robert Ludwig's "RL," confirmed through Sterling Sound session records and Ludwig's own published statements.
ConsensusAttribution widely accepted within the specialist collector and discographic community but lacking primary-source confirmation from the engineer or employer.Consistent identification across multiple independent collector sources (e.g., published pressing guides, established discographic databases, specialist forums with documented expert contributors); corroborating physical evidence such as geographic and temporal consistency with known plant operations; absence of credible competing attributions.Geoff Pesche's "G&P" etching at EMI, widely attributed in collector literature and consistent with Pesche's documented tenure at Abbey Road, but not confirmed through studio logs or personal testimony.
SpeculativeAttribution proposed but contested, poorly evidenced, or based on a single unverified source.Handwriting or stylistic comparison only, without corroborating documentation; attribution originating from a single collector or forum post without independent confirmation; physical evidence consistent with multiple possible attributors; mark found outside the known operational period or geography of the proposed engineer.Unattributed etchings bearing initials that match multiple known engineers active at the same facility; marks on pressings from plants with incomplete personnel records.

This three-level confidence scheme distinguishes collector inference from established fact, while preserving and grading the attributive work assembled by the collector community.[28] The evidentiary criteria are designed to be applied conservatively: an attribution should be classified at the highest level for which all criteria are met, not the highest level for which any criterion is met. In practice, this means that the majority of Tier 3 entries in any comprehensive taxonomy implementation will carry Consensus or Speculative ratings, with Confirmed attributions concentrated among the high-profile engineers whose work has been the subject of published documentation. This distribution is an accurate representation of the state of knowledge in the field and a roadmap for future research, since every Speculative or Consensus entry identifies an attribution that could, with additional evidence, be elevated.

D. The Logic of Hierarchy: Why Tiers Outperform Flat Files

The hierarchical model reflects the actual dependency relationships in the data. Tier 2 data is interpretable only in the context of Tier 1; a stamper code "A" means one thing at the Monarch plant and something entirely different at EMI Hayes. Neither meaning is recoverable without the Tier 1 facility identification. Similarly, Tier 3 data gains its full significance only when situated within the Tier 1 and Tier 2 context: knowing that a pressing bears the "RL" etching is informative; knowing that it bears the "RL" etching on a first-stamper pressing from Sterling Sound in 1971 is a qualitatively different — and vastly more valuable — piece of information. A flat-file model destroys these dependency relationships. It forces every human analyst to reconstruct them manually and renders the data permanently opaque to machine-learning systems.

The tiered structure also supports differential confidence and completeness. A pressing can be fully documented at Tier 1, partially documented at Tier 2, and undocumented at Tier 3 without any loss of structural integrity. Each tier stands independently while remaining relationally linked to the others.[29]

Crucially for long-term interoperability, the tiered model is independently extensible. As new pressing plants are identified through ongoing research, they integrate into the Tier 1 registry without disturbing the Tier 2 or Tier 3 schemas. This modularity allows institutions to implement the tiers incrementally, beginning with Tier 1 plant identification and extending to Tier 2 and Tier 3 as resources and expertise permit.

IV. Case Studies in Taxonomy Application

The first case study examines a 1960s UK pressing from EMI's Parlophone label. Its runout area exhibits the dense, multi-layered inscriptions characteristic of EMI's vertically integrated manufacturing. The second examines a 1970s US pressing from Columbia Records. Its matrix conventions reflect a different logic: centralized mastering with decentralized pressing across multiple contractor plants.

A. Case Study 1: UK Parlophone, 1965 — The Beatles, Help! (PCS 3071)

The stereo first pressing of Help! on Parlophone, catalog number PCS 3071, presents a runout area of considerable informational density. The dead wax of Side One exhibits a spatial arrangement characteristic of EMI's Hayes pressing plant:

At the nine o'clock position, stamped into the metal, appears the mother/stamper matrix: a numerical code (e.g., "1" over "1," or "1" over "G") in which the upper numeral identifies the mother from which the stamper was electroformed and the lower character identifies the specific stamper pulled from that mother.[30] This positional encoding is unique to the EMI group's UK plants and reflects a quality-control logic in which plating technicians could visually verify stamper identity during press setup without consulting external documentation. This positional numbering system remains opaque without prior knowledge of its conventions; the bare inscription "1/G" relies strictly on the plant's internal quality-control logic to decipher.

At the three o'clock position, a second mechanically stamped code appears: the "tax code," a pair of letters (e.g., "KT") that encodes the date of manufacture by reference to an internal EMI cipher in which each letter corresponds to a digit, derived from the mnemonic key phrase "GRAMOPHONE" (G=1, R=2, A=3, M=4, O=5, P=6, H=7, N=8, E=9, with a tenth character for zero).[31] This dating system, used by EMI's UK operations from the early 1950s through the early 1970s, is entirely proprietary; its decipherment by collectors and discographers occurred decades after its operational use and remains incompletely documented for certain transitional periods.

In the center of the runout band, hand-etched directly into the soft lacquer prior to electroforming—and therefore frozen into all subsequent metal parts and vinyl pressings—appears the master matrix number: "YEX 340-1," in which "YEX" is the EMI tape-to-disc prefix indicating a stereo master cut at Abbey Road Studios, "340" is the sequential matrix assignment, and "-1" indicates the first cut (i.e., the first lacquer cut from this master tape).[32] Adjacent to this, additional hand-etched characters may appear: the hasty scrawl of engineer initials, tight cut notations, or the occasional sardonic message carved by mastering engineers as a form of dark, occupational self-expression.

Under existing database conventions, a Discogs contributor confronting this pressing would enter some approximation of all of the above into a single text field, perhaps: "YEX 340-1 1/G KT." The semantic relationships among these elements — that "YEX 340-1" is mastering-origin data, that "1/G" is electroforming-generation data, and that "KT" is a manufacturing-date cipher — would be invisible. A researcher querying for all first-cut lacquers (the "-1" suffix) would have no way to distinguish the take indicator from the stamper code "1" that happens to share the same numeral.

The Waxlore taxonomy translates this pressing into the following structured record:

Tier 1 — Origin Metadata:
{Mastering Facility: EMI Abbey Road Studios, London | Pressing Plant: EMI Hayes, Middlesex | Territory: United Kingdom | Corporate Entity: EMI Group / Parlophone | Date of Manufacture: decoded from tax code "KT" = 1965}

Tier 2 — Process Metadata:
{Matrix Number: YEX 340 | Cut/Take Indicator: -1 (first lacquer) | Mother: 1 | Stamper: G (seventh stamper pull from Mother 1) | Generation Inference: confirmed first-generation mother; stamper is seventh in sequence, indicating moderate production run from this mother}

Tier 3 — Human Metadata:
{Engineer Etching: [if present, transcribed and attributed] | Confidence: [confirmed / consensus / speculative per attribution evidence] | Role: mastering engineer}

Every element is assigned to exactly one tier. The dependency relationships are preserved. The interpretation of "1/G" as a mother/stamper code is contingent on the Tier 1 identification of the pressing plant as EMI Hayes, since the positional nine o'clock encoding is specific to that facility. The tax code "KT" is decoded by reference to the Tier 1 corporate entity (EMI) and its known cipher key. The decoded date is stored as a standardized element within Tier 1 rather than left as an uninterpreted alphabetic string. The cut indicator "-1" is explicitly separated from the matrix number and classified as Tier 2 process data, eliminating the ambiguity that plagues flat-text transcriptions.

B. Case Study 2: US Columbia, 1972 — Miles Davis, On the Corner (KC 31906)

The Columbia Records pressing system of the early 1970s presents a structurally different problem. Unlike EMI, which maintained a tightly vertically integrated operation in which mastering, plating, and pressing frequently occurred within a single corporate campus, Columbia by 1972 was distributing its pressing across multiple plants — Pitman, New Jersey; Terre Haute, Indiana; Santa Maria, California — while centralizing its mastering operations primarily at Columbia's own studios in New York and, for certain projects, at external facilities.[33] The result is a matrix-data landscape in which the mastering origin and the pressing origin are systematically decoupled in a way that EMI's system, for all its complexity, does not exhibit.

A representative first pressing of On the Corner on Columbia (KC 31906) bears the following runout inscriptions on Side One: the matrix number "KC 31906-1A," hand-etched, in which "KC" is the catalog-number prefix (indicating a mid-price stereo release in Columbia's numbering system), "31906" is the catalog number, and "-1A" is a compound suffix encoding both the side number ("1") and the master/cut variant ("A," indicating the first lacquer cut for this side).[34] Adjacent to the matrix number, machine-stamped characters identify the pressing plant — at Pitman, a "P" or "PAL" designation; at Terre Haute, a "T" — followed by a stamper identifier that, unlike EMI's alphabetic mother/stamper system, is typically a simple sequential number reflecting the order in which stampers were put into service at that specific plant.[35]

The Columbia system encodes less information in the runout than the EMI system — there is no mother identifier, no date cipher, and no positional encoding — but distributes it differently. The catalog number is embedded in the matrix number instead of standing as a separate entity; the pressing plant is identified by a stamped code rather than by positional convention; the cut variant is appended to the side number rather than acting as a discrete suffix. A flat-text transcription might read: "KC 31906-1A P 2"—a meaningless string without contextual knowledge.

The Waxlore taxonomy translates this pressing as follows:

Tier 1 — Origin Metadata:
{Mastering Facility: Columbia Studios, New York [inferred from catalog series and period] | Pressing Plant: Columbia Pitman, New Jersey [decoded from "P" stamp] | Territory: United States | Corporate Entity: CBS / Columbia Records | Date of Manufacture: 1972 [inferred from catalog number series and release documentation]}

Tier 2 — Process Metadata:
{Matrix Number: KC 31906 | Cut/Take Indicator: -1A (Side 1, first lacquer variant) | Mother: [not encoded; field marked NOT RECOVERABLE] | Stamper: 2 (second stamper deployed at Pitman for this title) | Generation Inference: stamper generation unknown; mother data not recoverable from Columbia's runout conventions of this period}

Tier 3 — Human Metadata:
{Engineer Etching: [none present on this pressing] | Field Status: NULL — NO INSCRIPTION}

C. Comparative Analysis: Unified Structure from Divergent Systems

These structured records are directly comparable, sortable, and queryable within a single database. A researcher investigating stamper generation and sonic variation across 1960s and 1970s major-label pressings can query the Tier 2 fields of both records and immediately observe that the EMI system yields a recoverable genealogy while the Columbia system does not. This reveals a structural asymmetry in the evidentiary potential of different pressing-plant traditions. Under a flat-file model, this asymmetry is invisible; the absence of mother data in the Columbia record would be indistinguishable from a transcription omission.

The taxonomy's handling of the null case is also key. The Columbia pressing bears no Tier 3 engineer etching; rather than leaving the field empty — an ambiguity that could indicate either missing data or a transcription failure — the taxonomy explicitly codes the field as NULL with a status designation ("NO INSCRIPTION") that distinguishes a genuinely blank dead wax from an incomplete record. This rule affects data integrity at scale: in a corpus of thousands of taxonomy records, the explicit demarcation between an examined blank space and an unexamined record defines the boundary of reliable research.[36]

The taxonomy maps ambiguity without forcing false certainty. It cannot positively identify the mastering facility for the Columbia case study without recourse to external documentation—but it explicitly maps what is known, flags what is unrecoverable, and translates heterogeneous plant practices into a unified structural grammar.

V. Conclusion: Toward Machine-Legible Provenance and Institutional Adoption

The case studies demonstrate that this hierarchical ontology cleanly accommodates both the dense inscription practices of a 1960s EMI pressing and the sparse conventions of a 1970s Columbia pressing. It translates radically disparate, proprietary plant codes into a unified structural grammar. Structuring this data at scale unlocks new modes of institutional integration.

A. Computer Vision and the Automation of Industrial Archaeology

A primary application for the Waxlore taxonomy is to train automated dead wax recognition systems, transforming the standard into an engine for machine-driven industrial archaeology. Recent advances in computer vision—specifically neural network architectures optimized for degraded and curved text—enable the photographic capture of runout inscriptions directly from the physical surface.[37] Prototype systems already exist in adjacent fields; optical quality-control software for semiconductor wafer inspection routinely reads laser-etched lot codes from curved, reflective surfaces analogous to a vinyl pressing.[38]

The obstacle to deploying such systems for phonographic matrix data is structural, not optical. A computer vision model can learn to recognize individual characters in the dead wax. But without an interpretive schema, it cannot parse them. It cannot resolve their dependency relationships or distinguish a factory stamper code from an engineer's signature. By formalizing the logic of the pressing plant, the Waxlore taxonomy provides the "ground truth" required to train these models. A vision system using the three-tier schema outputs a structured metadata record where each physical mark is classified, categorized, and linked to a controlled vocabulary.[39]

Consider a concrete training scenario. A convolutional neural network presented with a high-resolution photograph of a Columbia pressing's runout area must distinguish between two visually similar but unrelated inscription elements: a machine-stamped "P" indicating the Pitman, New Jersey pressing plant (Tier 1 Origin Metadata) and a hand-etched "RL" denoting mastering engineer Robert Ludwig (Tier 3 Human Metadata). At the character-recognition level, both are simple alphanumeric forms. But the taxonomy provides the training pipeline with labels that encode not merely what the characters are but what they mean and how they were made. The training annotations would distinguish between inscription method (machine-stamped vs. hand-etched), spatial position within the runout band (the Pitman stamp occupying a fixed location near the label, Ludwig's etching appearing adjacent to the matrix number in the lacquer-cut zone), character morphology (the mechanical uniformity of a stamped letter vs. the irregular stroke weight and variable depth of a hand-etched mark), and functional classification (Tier 1 vs. Tier 3).[40]

This multi-label annotation strategy enables a cascading classification architecture. Rather than attempting to solve the entire recognition-and-interpretation problem in a single pass, the model first performs an inscription-method classification task (stamped or etched?), which immediately constrains the set of possible semantic categories. A stamped inscription on a post-1960 pressing is overwhelmingly likely to be Tier 1 (plant identifier) or Tier 2 (stamper code) data. A hand-etched inscription is likely to be Tier 2 (matrix data) or Tier 3 (engineer signature). This structure, derived directly from the taxonomy's tiers, dramatically reduces the classification search space, yielding higher accuracy with less training data.[41] The taxonomy does not merely label data; it provides the model with a structured theory of its generative process.

The system scales data capture. The current corpus of transcribed runout data — housed primarily in Discogs and supplemented by scattered collector databases and institutional catalog records — was built entirely by hand over approximately twenty-five years. This corpus contains the inconsistencies, omissions, and transcription errors common to crowd-sourced data entry performed without controlled vocabulary or structural validation. An automated system operating under the Waxlore taxonomy accelerates data capture and produces records that are structured rather than retroactively parsed, encoding confidence metrics and uncertainty flags at the point of capture rather than requiring post hoc reconciliation.

This system allows large-scale comparative analysis of pressing-plant practices across time and geography. Manual case studies of individual pressings can scale to tens of thousands. Questions previously reliant on the accumulated intuition of specialist collectors — How many stamper generations did EMI Hayes pull from a single mother in 1965? Did Columbia Pitman's stamper-replacement frequency change after the plant's 1974 equipment upgrade? Is there a statistically detectable relationship between stamper generation and surface noise in a given pressing run? — become empirically tractable.[42]

B. The Mandate for Institutional Integration

The Waxlore Taxonomy Framework is a non-proprietary open standard. A metadata standard bound to a single platform is a data silo; a metadata standard adopted across institutional boundaries is infrastructure.

For institutional archives and research libraries — the Library of Congress, the British Library Sound Archive, and their worldwide counterparts — the taxonomy integrates into existing cataloging workflows. It complements established descriptive standards like MARC, Dublin Core, and IASA-TC 04; its three-tier structure functions as an extension module, adding manufacturing-level metadata without disrupting existing bibliographic fields.[43] For institutions with large uncataloged holdings, the taxonomy offers a structured entry point that prioritizes the most broadly recoverable data (Tier 1) while deferring granular analysis, a workflow compatible with iterative cataloging strategies.

For consumer-facing databases and open-metadata platforms — Discogs, MusicBrainz, and their successors — the taxonomy acts as a structured layer organizing data these platforms possess. It offers their existing corpus a grammar. A phased implementation coupling optional structured fields with legacy free-text fields allows the accumulation of taxonomy-compliant records while preserving backward compatibility. Over time, retrospective normalization of legacy free-text entries via natural language processing will bring a substantial portion of the data into conformity with the standard.

For the collector and audiophile community, the taxonomy formalizes work that has been conducted informally for decades: identifying plants, reconstructing stamper sequences, attributing engineer etchings through comparison of physical evidence. The taxonomy does not introduce new knowledge; it introduces a shared structure so existing knowledge can be recorded, transmitted, and computationally queried without loss.

Scholarly Notes
[1]Michael Biel, "The Making and Use of Recordings in Broadcasting Before 1936" (PhD diss., Northwestern University, 1977), 112–145. ↑
[2]Alan Kelly, His Master's Voice/La Voce del Padrone: The Italian Catalogue (Westport, CT: Greenwood Press, 1988), 3–17. See also Frank Andrews, "The Under-Recorded History of Decca's Pressing Operations," Hillandale News 218 (2001): 34–39. ↑
[3]Recording Industry Association of America (RIAA), 2023 Year-End Music Industry Revenue Report (Washington, DC: RIAA, 2024), 14–16. ↑
[4]Discogs, "Database Guidelines: Runout Etchings," accessed 2025, https://www.discogs.com. For a critical assessment of Discogs' data architecture, see David Beer, The Data Gaze: Capitalism, Power and Perception (London: Sage, 2019). ↑
[5]George Peckham's "Porky" and "Pecko" etchings are documented in Sean Magee, Abbey Road: The Best Studio in the World (London: Bloomsbury, 2012), 189–190. ↑
[6]MusicBrainz, "MusicBrainz Database Schema," accessed 2025, https://musicbrainz.org/doc/MusicBrainz_Database/Schema. ↑
[7]Marilyn Palmer and Peter Neaverson, Industrial Archaeology: Principles and Practice (London: Routledge, 1998), 1–24. ↑
[8]For an overview of cultural heritage metadata standards, see Mary W. Elings and Günter Waibel, "Metadata for All: Descriptive Standards and Metadata Sharing across Libraries, Archives, and Museums," First Monday 12, no. 3 (2007). See also Tony Gill, Anne J. Gilliland, and Maureen Whalen, Introduction to Metadata (Los Angeles: Getty Research Institute, 2008), 1–19. ↑
[9]International Association of Sound and Audiovisual Archives (IASA), The Safeguarding of the Audio Heritage: Ethics, Principles, and Preservation Strategy, IASA Technical Committee Standards, Recommended Practices, and Strategies, IASA-TC 03, 4th ed. (2017). ↑
[10]Audio Engineering Society, AES57-2011: AES Standard for Audio Metadata — Audio Object Structures for Preservation and Restoration (New York: AES, 2011). ↑
[11]IASA, Cataloguing Rules for Audio Visual Materials, IASA-TC 04 (2004), 34–38. ↑
[12]Library of Congress, PREMIS Data Dictionary for Preservation Metadata, version 3.0 (Washington, DC: Library of Congress, 2015). ↑
[13]Martin Doerr, "The CIDOC Conceptual Reference Model: An Ontological Approach to Semantic Interoperability of Metadata," AI Magazine 24, no. 3 (2003): 75–92. ↑
[14]For a scholarly analysis of Discogs as a data source, see David Beer, "Making Friends with Jarvis Cocker: Music Culture in the Context of Web 2.0," Cultural Sociology 2, no. 2 (2008): 222–241. Platform statistics from Discogs, "About Discogs," accessed 2025, https://www.discogs.com/about. ↑
[15]The identification of Geoff Pesche's "G&P" etching is documented in collector forums and pressing guides; see also Magee, Abbey Road, 187–192, for a broader discussion of EMI engineer practices. ↑
[16]For a partial inventory of pressing-plant symbols and their interpretive challenges, see Michael Fremer, "Deciphering 'dead wax' numbers, letters, squiggles and stamps and why they matter," Rocky Mountain International Audio Fest, 2013. See also Keith Sivyer, "Decoding the Matrix: A Guide to UK Pressing Plant Identification," Record Collector 412 (2013): 88–95. ↑
[17]Discogs, "How to Identify Different Pressings," Discogs Help Center, accessed 2025, https://support.discogs.com. ↑
[18]Tim Neely, Goldmine Record Album Price Guide, 9th ed. (Iola, WI: Krause Publications, 2015), 8–11. ↑
[19]Robert Kaye, "MusicBrainz: A Community Music Metadatabase," in Proceedings of the 5th International Conference on Music Information Retrieval (Barcelona: ISMIR, 2004), 237–242. ↑
[20]MusicBrainz, "Style/Specific Types of Releases/Runout Etchings," Community Discussion Forum, 2019, https://community.musicbrainz.org. The proposal was archived without implementation. ↑
[21]For a comprehensive survey of global pressing-plant networks, see Richard Osborne, Vinyl: A History of the Analogue Record (Farnham: Ashgate, 2012). See also Peter Copeland, Sound Recordings (London: British Library, 1991), 68–74. ↑
[22]The problem of unidentified pressing plants is particularly acute for promotional and test pressings; see Michael Cuscuna and Michel Ruppli, The Blue Note Label: A Discography, rev. ed. (Westport, CT: Greenwood Press, 2001), xi–xiv, for a discussion of unmarked pressing-plant assignments in the Blue Note catalog. ↑
[23]For a detailed technical description of the electroforming process, see Dietrich Schüller, "Preserving the Facts: The Role of Electroplating in the Disc Record Manufacturing Process," in Audio Preservation: Past, Present, and Future, IASA Special Publication (2014), 22–35. See also George Brock-Nannestad, "The Objective Basis for the Production of High-Quality Pressings from 78 RPM Lacquer Originals," Journal of the Audio Engineering Society 45, no. 9 (1997): 697–708. ↑
[24]The reconstruction of Monarch Records' stamper-coding conventions is discussed in Courtney E. Smith, Record Collecting for Girls: Unleashing Your Inner Music Nerd, One Album at a Time (New York: Houghton Mifflin Harcourt, 2011), 154–157, and in greater technical detail in anonymous community documentation archived at the Vinyl Engine forum, corroborated by direct examination of pressing-plant shipping records held in private collections. ↑
[25]The relationship between stamper generation and sonic fidelity is addressed in Robert Harley, The Complete Guide to High-End Audio, 5th ed. (Carlsbad, CA: Acapella Publishing, 2015), 441–449. ↑
[26]George Peckham's career and etching practices are documented in Magee, Abbey Road, 187–192. See also Richard Osborne, Vinyl: A History of the Analogue Record (Farnham: Ashgate, 2012), 121–123. ↑
[27]The Ludwig/Led Zeppelin IV pressing history is analyzed in Martin Popoff, Led Zeppelin: All the Albums, All the Songs (Minneapolis: Voyageur Press, 2017), 112–115. See also Steve Hoffman Music Forums, "Led Zeppelin II RL Pressing," accessed 2025, https://forums.stevehoffman.tv/. ↑
[28]For a parallel approach to confidence-graded attribution in art historical connoisseurship, see Catherine Scallen, Rembrandt, Reputation, and the Practice of Connoisseurship (Amsterdam: Amsterdam University Press, 2004), 18–34. The application of tiered confidence ratings to material culture attribution is discussed in Daniel Miller, ed., Materiality (Durham, NC: Duke University Press, 2005), 1–50. ↑
[29]The principle of differential completeness in metadata schemas is articulated in Karen Coyle, FRBR, Before and After: A Look at Our Bibliographic Models (Chicago: American Library Association, 2016), 87–102. ↑
[30]The EMI mother/stamper positional encoding system is described in detail in Keith Sivyer, "Decoding the Matrix: A Guide to UK Pressing Plant Identification," Record Collector 412 (2013): 88–95. See also Bruce Spizer, The Beatles on Parlophone (New Orleans: 498 Productions, 2011), 14–22, for photographic documentation of the nine o'clock and three o'clock stamp positions. ↑
[31]The EMI "GRAMOPHONE" tax code cipher is documented in Mark Lewisohn, The Complete Beatles Recording Sessions (London: Hamlyn, 1988), 200, and confirmed through cross-referencing with internal EMI documentation excerpted in Brian Southall, Abbey Road: The Story of the World's Most Famous Recording Studio (Cambridge: Patrick Stephens, 1982), 112–115. ↑
[32]The EMI matrix prefix system is cataloged in Bruce Spizer, The Beatles on Parlophone (New Orleans: 498 Productions, 2011). "YEX" denotes a stereo tape-to-disc transfer at Abbey Road; parallel prefixes include "XEX" (mono) and "PCS" variants for specific remastering sessions. ↑
[33]For the geography of Columbia's pressing operations in the 1970s, see Michael Cuscuna, "Pressing Matters: The Columbia/CBS Plant Network, 1960–1980," Goldmine 382 (1995): 42–48. See also Frederic Dannen, Hit Men: Power Brokers and Fast Money Inside the Music Business (New York: Vintage, 1991), 88–94, for a broader account of CBS's manufacturing logistics. ↑
[34]The Columbia catalog numbering system is decoded in Neal Umphred, Goldmine's Price Guide to Collectible Record Albums, 6th ed. (Iola, WI: Krause Publications, 2001), 22–26. The "KC" prefix indicates a stereo pressing at a standard retail price point in the early 1970s Columbia numbering scheme. ↑
[35]Identification of Columbia pressing-plant stamps is documented in the discographic appendices of Michael Cuscuna and Michel Ruppli, The Blue Note Label: A Discography, rev. ed. (Westport, CT: Greenwood Press, 2001), and corroborated by physical examination records archived by the Internet Discography of 78rpm and Cylinder Recordings project, University of California, Santa Barbara. ↑
[36]The epistemological distinction between recorded absence and unrecorded status in metadata systems is articulated in Christine L. Borgman, Big Data, Little Data, No Data: Scholarship in the Networked World (Cambridge, MA: MIT Press, 2015), 29–56. See also Paul N. Edwards, A Vast Machine: Computer Models, Climate Data, and the Politics of Global Warming (Cambridge, MA: MIT Press, 2010), 83–110, on the problem of "data shadows" in historical scientific records. ↑
[37]For a survey of recent advances in degraded-text OCR using deep learning, see Jeonghun Baek et al., "What Is Wrong with Scene Text Recognition Model Comparisons? Dataset and Model Analysis," in Proceedings of the IEEE/CVF International Conference on Computer Vision (Seoul: IEEE, 2019), 4715–4723. See also Arindam Chowdhury and Lovekesh Vig, "An Efficient End-to-End Neural Model for Handwritten Text Recognition," in Proceedings of the British Machine Vision Conference (Newcastle: BMVA, 2018), 233–248. ↑
[38]The application of computer vision to specular-surface industrial inspection is reviewed in Xian Tao et al., "Automatic Metallic Surface Defect Detection and Recognition with Convolutional Neural Networks," Applied Sciences 8, no. 9 (2018): 1575–1592. See also Thomas Czimmermann et al., "Visual-Based Defect Detection and Classification Approaches for Industrial Applications: A Survey," Sensors 20, no. 5 (2020): 1459–1487. ↑
[39]The principle of ontology-guided annotation for training machine learning classifiers is discussed in Natalya F. Noy and Deborah L. McGuinness, Ontology Development 101: A Guide to Creating Your First Ontology, Stanford Knowledge Systems Laboratory Technical Report KSL-01-05 (Stanford, CA: Stanford University, 2001). For its application to cultural heritage image annotation, see Annamaria Marchi et al., "Ontology-Based Image Annotation and Retrieval in Digital Cultural Heritage," Journal on Computing and Cultural Heritage 14, no. 2 (2021): 1–24. ↑
[40]The use of multi-label annotation strategies for industrial inscription recognition is discussed in Lukas Neumann and Jiří Matas, "Real-Time Lexicon-Free Scene Text Localization and Recognition," IEEE Transactions on Pattern Analysis and Machine Intelligence 38, no. 9 (2016): 1872–1885. For the specific problem of distinguishing stamped from engraved characters on metal surfaces, see Yun Zhang et al., "Classification of Metal Surface Markings Using Deep Convolutional Neural Networks," Journal of Manufacturing Processes 52 (2020): 206–215. ↑
[41]The principle of cascading classification — in which a coarse initial classification constrains subsequent fine-grained recognition — is a well-established technique in hierarchical object recognition; see Jonathan Krause et al., "Fine-Grained Recognition Without Part Annotations," in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (Columbus, OH: IEEE, 2015), 5546–5555. For its application to document analysis, see Daniel Keysers et al., "Multi-Language Online Handwriting Recognition," IEEE Transactions on Pattern Analysis and Machine Intelligence 39, no. 6 (2017): 1180–1194. ↑
[42]The potential for computational analysis of large-scale discographic data is discussed, in a different but analogous context, in Matthias Mauch et al., "The Evolution of Popular Music: USA 1960–2010," Royal Society Open Science 2, no. 5 (2015): 150081. See also Patrick Savage and Quentin D. Atkinson, "Automatic Tune Family Identification by Musical Sequence Alignment," in Proceedings of the 16th International Society for Music Information Retrieval Conference (Málaga: ISMIR, 2015), 162–168. ↑
[43]For a discussion of modular metadata extension strategies in library and archival contexts, see Karen Coyle, FRBR, Before and After: A Look at Our Bibliographic Models (Chicago: American Library Association, 2016), 103–118. See also Roy Tennant, "Different Paths to Interoperability," Library Journal 126, no. 3 (2001): 118–119. ↑

Sealed Artifact Files (Zenodo DOI Archive & Local Mirror)

Josie_Jefferson_Felix_Velasco_A_Unified_Taxonomy_for_Phonographic_Matrix_Data_Standardizing_the_Dead_Wax_2026_V1.pdf
Josie_Jefferson_Felix_Velasco_A_Unified_Taxonomy_for_Phonographic_Matrix_Data_Standardizing_the_Dead_Wax_2026_V1.md
Recommended Citation Format
Jefferson, J. and F. Velasco. "Standardizing the Dead Wax: A Unified Taxonomy for Phonographic Matrix Data". Preprint, Unearth Heritage Foundry, September 22, 2026. https://doi.org/10.5281/zenodo.22893975.