Data Dictionary & Codebook
Hip-Hop Periodic Table: what the variables mean, how scores were assigned, and who is in the data
This page documents every variable in the Hip-Hop Periodic Table dataset, the constructs behind the five lyricism scores, how scores were assigned, and the sampling frame the dataset draws from. The same content ships as the Data Dictionary sheet inside the Excel workbook, which is the single source of truth.
The working premise of this page: clarity is kindness. A reader should never have to guess what a number means or where a judgment lives.
The Five Constructs
Each act is scored 1–10 on five dimensions of lyricism. Scores are drawn from published criticism and hip-hop scholarship; no automated text analysis was run. Each dimension has a named anchor at 10 so the top of the scale means the same thing across scorers.
| Construct | What it measures | Anchor at 10 |
|---|---|---|
| Rhyme Density | Frequency and complexity of rhyme: internal rhyme, multisyllabic chains, rhymes per bar, scheme inventiveness | Rakim: pioneered internal and multisyllabic rhyme as standard practice |
| Vocab Breadth | Lexical diversity: range of unique words and registers across the catalog | Aesop Rock: per published studies counting unique words in rap |
| Storytelling | Narrative construction: character, plot, perspective shifts, arc within songs and across albums | Scarface |
| Metaphor/Imagery | Density and originality of figurative language: extended metaphor, concrete imagery, double meanings | Lil Wayne: simile machine; metaphor-stacking as his defining critical reputation |
| Conceptual Depth | Thematic ambition and coherence: album-level concepts, social and philosophical substance | Kendrick Lamar: 10 is reserved for carrying a concept across a full album at the highest level |
The Composite Score is the plain average of the five dimensions, rounded to one decimal. Counting all five equally is itself a choice, and one of the module’s core discussion prompts.
Scoring Method
- Basis: published music criticism, academic hip-hop scholarship (Bradley, Chang, Kajikawa, Rabaka), and documented critical consensus. Scores are judgments, not calculations.
- Scale: integers 1–10 per dimension; anchored at the top of each scale (table above).
- How anchors create scores: a score here is a comparison, not an absolute rating. Each dimension pins 10 to a named exemplar, so scoring an act means asking how far below the anchor it sits: how far below Rakim’s rhyme craft, how far below Lil Wayne’s simile-stacking. That comparison does three jobs. It calibrates scorers to each other, since disagreements become arguments about distance from a shared ceiling rather than about what an 8 means. It keeps the five scales comparable enough to average into the Composite, because each column’s 10 is equally rare and equally extreme. And it keeps future entries comparable to past ones, since every act, whenever it is added, is measured against the same fixed ceiling. Two limits travel with the method: an anchor pins the top of the scale, not the gaps between scores, so this is ordinal measurement with a calibrated ceiling; and the choice of anchor quietly defines the construct itself (Rakim makes rhyme density mean innovation and internal complexity rather than speed). The scoring ladders below extend the same idea down the scale: named rungs at 8, 6, and 4, and a defined floor. The judgment does not disappear. It moves to one visible place.
- Confidence flag: every act carries an H/M/L rating of how much to trust its scores, based on the depth of its critical record (rubric below).
The Scoring Ladders
An anchor alone defines only the top of a scale. Each dimension therefore carries a ladder: the anchor at 10, calibration rungs at 8, 6, and 4 naming acts from this table whose stored score equals the rung, and a written floor at 1. Scoring an act means placing it on the ladder, between rungs when the record says between. The floor is defined but empty on purpose: the sampling frame requires a national critical record, which excludes each construct’s true bottom, so the lowest occupied scores sit near 2 and 3, not 1.
Confidence Rubric
| Flag | Meaning | Criteria |
|---|---|---|
| H (High) | Extensive critical record | Sustained critical/academic attention; multiple long-form analyses of the act’s lyrical craft; consistent reception across sources; a body of work large enough that scores are stable across albums |
| M (Medium) | Moderate documentation | Meets 1–2 of the High criteria: historically important but under-analyzed, newer with an incomplete body of work, mixed critical reception, or regional/underground importance without mainstream scholarly attention. Treat as a working estimate |
| L (Low) | Thin or disputed record | Little formal analysis, a very recent debut, single-source scores, or reception clouded by controversy unrelated to the music. Treat as first estimates that need checking |
Good places to double-check M and L scores: RateYourMusic critical consensus, Pitchfork / The Wire discography reviews, Google Scholar (artist + “lyricism”), and genre publications (The Source, XXL retrospectives, Passion of the Weiss).
Sampling Frame & Scope Rules
Two scope rules are worth calling out:
- Scene = scene, not birthplace. Acts are assigned to the US scene they are professionally rooted in: label home and scene ties during their defining run (Da Brat is South via So So Def, though born in Chicago). Region rolls scenes up to the four coast-level categories the culture itself uses: East Coast (NYC + Philly), West Coast (LA + Bay Area), South, and Midwest.
- Era = the signature-year bracket. Era is derived, not judged: it is the bracket holding the Signature Work year (Debut Year only when no signature exists, i.e. Kool Herc). Brackets: Old School through 1985, Golden Age 1986–1994, Late 90s 1995–1999, then calendar decades. Debut years still overlap era boundaries by design; the judgment lives in the Signature Work pick, which has its own documented procedure (see Variable definitions).
The Eight Styles
Style is one value per act: the dominant mode of the act’s defining run, judged from critical descriptions of the catalog. The axis separating the two left-field categories is where the strangeness lives: in the words, it is Abstract; in the sound or form, it is Experimental. Each definition names an anchor act, the same way the five constructs anchor their scales.
Variable Definitions
Contested Calls
The Signature Work column drives Era, so it carries real weight, and for a handful of acts the critical record genuinely splits between two candidate releases. Those calls are not hidden: the 2026-07 audit reviewed every pick against the documented procedure, changed ten where the consensus clearly pointed elsewhere, and resolved the genuine splits by maintainer ruling. New splits get the same treatment as acts are added. The rulings are logged here so the judgment is inspectable.
| Act | The split | Ruling | Why |
|---|---|---|---|
| Common | Resurrection (1994) vs Like Water for Chocolate (2000) vs Be (2005) | Resurrection | The album 90s critics call his best work, and the earliest of a three-way split |
| 2Pac | Me Against the World (1995) vs All Eyez on Me (1996) | All Eyez on Me | The cultural monument outweighs the critics’ cohesion pick |
| Busta Rhymes | The Coming (1996) vs When Disaster Strikes (1997) | When Disaster Strikes | The peak-era statement over the classic debut |
| Royce da 5’9” | Death Is Certain (2004) vs Book of Ryan (2018) | Book of Ryan | The critical high point of the catalog |
| 8Ball & MJG | Comin’ Out Hard (1993) vs On Top of the World (1995) | On Top of the World | Their highest-rated album over the influence landmark |
| Roc Marciano | Marcberg (2010) vs Reloaded (2012) | Reloaded | The refined peak edges the blueprint on critic scores |
| Little Brother | The Listening (2003) vs The Minstrel Show (2005) | The Minstrel Show | The fan-canon favorite |
| Xzibit | At the Speed of Life (1996) vs 40 Dayz & 40 Nightz (1998) | 40 Dayz & 40 Nightz | The album most often called his best |
| 21 Savage | i am > i was (2018) vs Savage Mode II (2020) | Savage Mode II | His best-reviewed work; the joint billing with Metro Boomin is fine by the Madvillainy precedent |
| Playboi Carti | Die Lit (2018) vs Whole Lotta Red (2020) | Whole Lotta Red | Divisive at release, canon now; the record the rage sound grew from |
| Ice Spice | Like..? EP (2023) vs Y2K! (2024) | Y2K! | The EP is the consensus moment, but the candidate pool is full-length projects: the same album-over-EP ruling as GloRilla |
| Doja Cat | In the frame at all, then Hot Pink (2019) vs Planet Her (2021) vs Scarlet (2023) | In; Planet Her | She raps on most songs with a national critical record, so she is in; the defining run beats the bars-forward flex |
If you would rule differently on any of these, that disagreement is the point: the pick is documented, so the argument has something to grab.
Known Biases
The dataset over-represents critically acclaimed acts (vs. commercially dominant ones), male acts, Golden Age NYC, and English-language rap. International and non-English rap is excluded from the frame entirely, a documented scope decision. The dashboard section “Who Is In the Data?” measures these gaps and turns them into teaching prompts on sampling bias.