From Folk Roots to Digital Streams: The Evolution of Bangladeshi Music Genres, Listener Cultures, and Global Musical Influences

A Computational Analysis of Genre Development, Global Influences, Listener Communities, and Musical Ecosystems

Author: Md. Ismail Hossain
Affiliation: Institute of Information Technology (IIT), Jahangirnagar University, Dhaka, Bangladesh
Email: ismailhossain013@gmail.com
Profile & Links: Portfolio | GitHub | LinkedIn | Facebook Version: Second draft — first version carrying computed empirical results (2026-08-27) Keywords: Bangladesh music, genre evolution, cultural hybridity, co-appearance networks, sound evolution, concert culture, computational social science, digital humanities, global influence, local adaptation

---

Abstract

Bangladesh's popular music has been transformed since the early 1970s, from folk traditions and Adhunik Bangla through a band and metal explosion to a pluralistic digital-era ecosystem. This study builds an open, structured dataset of that ecosystem — 63 artists, 23 documented events, 169 normalised influence citations — and analyses it computationally through the Bangladesh Musical Ecosystem Model (BMEM) and its process companion, the Bangladesh Music Evolution Framework (BMEF).

Four results are reported. First, the arrival order of genre strands differs from the standard account: the catalogue's first hip-hop release (1993) predates its first alternative, folk-fusion, progressive and pop formations, placing Bengali-language rap two years after the first original Bangla metal album rather than two decades later. Second, influence citations are structured but unevenly evidenced — metal and progressive acts name specific bands (Metallica, 6 citing acts; Megadeth, 5; Pink Floyd, 4, the only vector reaching three strands), while pop and folk acts name traditions, an asymmetry that reflects who gets interviewed as much as how music transmits. Third, a multi-layer co-appearance network over the catalogue is genuinely structured (Louvain modularity 0.428 on observational layers; six communities of three or more acts) and its communities are interpretable — a mixed-strand national festival circuit, a metal genealogy joined by stated influence, a classic mainstream-rock cluster joined by shared personnel, and a founding-era cohort held together by acts citing each other. Fourth, the concert typology inherited from the project's own methodology is falsified by its data: every multi-act bill in the catalogue mixes genre strands, so large commemorative festivals function as discovery spaces rather than as the subcultural gatherings the framework predicted.

A methodological result is reported alongside these: projecting a concert hypergraph onto artist pairs without correcting for bill size lets a single twelve-act festival contribute 66 edges and depresses observed modularity from 0.428 to 0.159. Any scene network built from festival lineups needs this correction.

The study also states its boundary plainly. No streaming, audio-feature or audience-side data was obtainable, so the network reported here is a co-appearance network, not the preference network the framework specified, and every claim about sound or audience rests on documented description rather than measurement. Ten specific revisions to the BMEM are proposed on that basis.

---

1. Introduction

Music encodes social, technological and political change, and Bangladesh's case is unusually legible: a fifty-year span in which folk and patriotic traditions met Western rock, metal, hip-hop and electronic forms, while the means of distribution moved from cassette stalls to algorithmic streaming.

Existing scholarship on this transformation is thin and unevenly distributed. The Dhaka metal scene has an ethnographic literature (Quader & Redden, 2015; Quader, 2016); Bangladeshi rap acquired one only in 2021 (Hasan & Kundu, 2021, 2022); the pop strand has essentially none. Almost all of it is single-scene and qualitative. One computational precedent exists (Autul et al., 2024), comparing Bangladeshi and West Bengal rock via Spotify audio features, and it is comparative rather than ecosystem-level.

This project addresses that gap by building the dataset first and reporting what it supports. Its primary question is unchanged from the proposal:

How has Bangladesh's music culture evolved, and what factors — including global influences and local adaptation — are shaping its future?

What has changed is the standard of proof. This draft distinguishes throughout between what the data shows, what it suggests, and what the project simply cannot see. Several claims carried in the first draft are retracted here on that basis, and two dimensions of the study's own framework are corrected against its own evidence.

2. Background and scope

Bangladesh's musical substrate is deep: Baul, Bhatiali, Bhawaiya, Nazrul Geeti, Rabindra Sangeet, Adhunik Bangla. From the early 1970s, Azam Khan and Uccharon and a first wave of bands — Souls (c. 1972), Feedback (1976), Miles (1978–79), Feelings/Nagar Baul — established commercial band culture. The mid-1980s brought heavy metal (Warfaze 1984, Rockstrata 1985) and the 1990s the classic rock era (LRB and Ark, both 1991). The 2000s produced progressive metal (Artcell 1999, first album 2002), an indie expansion, and the beginnings of digital distribution. The 2010s and 2020s added streaming, hip-hop's growth in visibility, and folk-fusion platform projects.

The study works with seven genre strands: Mainstream Rock, Progressive Rock/Metal, Heavy Metal, Alternative/Indie, Pop, Hip-Hop/Rap, and Folk & Folk Fusion. Strand assignment is computed from each artist's genre list by positional weighting (analysis/lib/common.py), with two documented overrides where the project's own curated groupings disagree — the disagreements are kept visible rather than tuned away.

Five analytical dimensions are applied to each strand: musical evolution, cultural evolution, listener evolution, concert culture, and industry position.

3. Literature

The full annotated bibliography is docs/annotated-bibliography.md; the working review with retrieval status is docs/literature-review.md. In summary:

On the Bangladeshi scene. Quader & Redden (2015, Cultural Studies 29(3)) read Dhaka's metal underground as an alternative cultural space built by urban youth out of frustration with national conditions, using "scene" rather than "subculture" as the frame. Quader (2016, Metal Music Studies 2(1)) applies Bourdieu's forms of capital to the same scene. Hasan & Kundu (2021, 2022) provide the only substantial academic treatment of Bangladeshi rap, finding a generational split between politically explicit younger artists and more conformist established ones, and framing the strand as continuous with a national protest-music tradition. Mridha & Begum (2023) trace band music as a recent development inside a much longer folk continuum. Pervez (2012) applies frame analysis and Hall's theory of representation. Yoon (2019) analyses the fusion band BANGLA's reworking of Lalon repertoire as glocal cultural content. Mukherjee (2017) and Mitra (2008, 2014) supply West Bengal comparanda. Hasan (2015) is a widely cited LSE blog essay, not a peer-reviewed article — an error in this project's own earlier citation, corrected here.

Methodological precedent. Autul et al. (2024) is the sole computational study of Bangladeshi popular music located, and it is the benchmark any future audio-feature work here should be measured against.

Gaps this project addresses. No prior multi-genre, data-driven mapping of the ecosystem; no prior work linking repertoire change to concert co-appearance structure; and no academic literature at all on the pop strand, which remains the least theorised part of the ecosystem.

4. Conceptual framework

BMEM (frameworks/bmem.md) models the ecosystem as six interacting dimensions: Historical Evolution, Global Influence, Local Adaptation, Listener Preference Networks, Music Experience Culture, and Industry Transformation.

BMEF (frameworks/bmef.md), developed in this pass, is its process companion. It names four change mechanisms with distinct observable signatures — repertoire substitution (technique retained, repertoire replaced: Warfaze's 1991 pivot), personnel transmission (technique and audience carried by musicians between acts: Ayub Bachchu across Souls, LRB and Nagar Baul), programmatic exposure (audiences meeting strands they did not seek, via festival or platform programming), and format adaptation (the release unit changing, and composition following). Crucially it also states what evidence each mechanism requires, and marks sonic change and audience change as unmeasurable in this project.

The results in §7 test BMEM against data for the first time; §9 summarises the resulting corrections, set out fully in frameworks/bmem-validation.md.

5. Methodology

Design. Secondary data analysis in a computational social science / digital humanities mode. No primary surveys or interviews.

Data model. Two JSON Schemas govern the dataset (data/schemas/). Artist records carry formation and disbanding years, members with roles and years, discography summaries, and — the field that carries most of the analytical load — global_influences entries, each with the influence named, the evidence for it, a source, and a stated confidence level. Concert records carry date and date precision, venue and venue type, organiser, the billed artists with billing roles, and sources.

Pipeline. Seven scripts in analysis/scripts/, all re-runnable and deterministic:

Figures read exclusively from analysis/outputs/, so no figure can disagree with a reported number.

Network construction. Four edge layers, three observational and one inferred:

• co_billing — a documented shared concert bill;
• personnel — a shared named member;
• domestic_influence — one catalogued act naming another as an influence;
• influence_homophily — two acts citing the same named global influence. Inferred, flagged on every edge it produces, weighted at half, and excluded from the headline "observed" pass.

Co-billing edges carry a bill-size correction. Each event contributes 1/(n−1) to every pair on its bill, so an act's total contribution per event is 1 regardless of bill size. Without it, the 2020 Joy Bangla Concert's twelve-act bill alone contributes 66 edges and swamps every other signal. The weighting follows Newman's (2001) treatment of multi-author scientific papers in collaboration networks — the same dilution logic applied to concert bills instead of co-authorship.

Three analytical passes are reported for every network result: observed (the three documentary layers — the evidential result), full (all four — the exploratory result), and co_billing (the original single-layer prototype, for comparison).

Validation. Findings are treated as robust where they survive across independent evidence types. In practice this meant checking network results against the concert typology and the influence analysis, which is how the festival/discovery-space correction in §7.5 and the hip-hop documentation artefact in §7.4 were both caught.

Ethics. Public sources only; artist- and event-level records, not individual listeners.

5.1 A constraint on this pass, stated up front

The environment in which this analysis was run blocks direct page retrieval, so records added in this pass are sourced from search-result summaries of the cited articles rather than from their full text. Those records are tagged snippet-sourced in the data and should be verified against full text before publication. This affects the eight Joy Bangla Concert records and five artist records added here; it does not affect the analysis pipeline or any record predating this pass.

6. The dataset

Strand distribution: Alternative/Indie 16, Mainstream Rock 13, Heavy Metal 12, Hip-Hop/Rap 8, Pop 7, Folk & Folk Fusion 5, Progressive Rock/Metal 2.

The sample is not a census and the analysis treats it as a sample throughout. It is 21% of its artist target and 5% of its event target; it over-represents Dhaka (19 of 23 events) and rock/metal; and it was assembled outward from a well-documented metal core, which plausibly biases every strand-comparison in this paper. Where a result could be an artefact of that construction, this is stated at the point the result is given rather than deferred to §10.

7. Results

7.1 Temporal structure and strand lifecycles

(Figures 1–3; analysis/outputs/temporal-summary.json)

Catalogued formations by decade: 5 (1970s), 5 (1980s), 15 (1990s), 26 (2000s), 10 (2010s), 1 (2020s). The 2000s alone account for 42% of dated formations. The 2020s figure reflects documentation lag, not a halt in band formation.

Result 1 — the strand arrival order is not the received one. Hip-hop's first catalogued act (Ashraf Babu & Charu, Tri-Rotner Khepa, 1993, the first Bengali-language rap album) predates the catalogue's first alternative (1996), folk-fusion (1997), progressive (1999) and pop (2002) formations. It sits two years after Warfaze's 1991 original-Bangla-metal debut, not two decades. The common framing of hip-hop as Bangladeshi music's late arrival describes when the strand became visible to the press, not when it began.

Result 2 — release formats shift from albums to singles. Singles are absent from the catalogue before the 2010s and are 4 of 12 dated 2020s releases; studio albums fall from 100% of 1980s releases to 50% in the 2020s. Median gap between an act's dated releases is 5 years. Two cautions: the 2020s are incomplete, and older singles are systematically less likely to have been catalogued than older albums. The direction is well supported; the magnitude is not.

An unresolved observation. Heavy metal's latest catalogued formation is 2007, while alternative continues to 2018 and hip-hop to 2024. This may be real generational closure or an artefact of how the catalogue was assembled. The project cannot separate the two, and the observation is reported as open rather than as a finding.

7.2 Influence structure

(Figures 5–6; analysis/outputs/influence-analysis.json)

83 raw citations across 63 acts normalise to 169 citation tokens, resolving to 46 distinct named global acts and 73 tradition labels. Every raw citation carries a named source; stated confidence is medium 34, high 31, low 16, hypothesised 2.

Each canonical token's kind is resolved once across the whole corpus rather than from the string it appears in, so an influence cannot count as a named act in one cross-tab and as a tradition in another; ranks count distinct citing acts, not citation rows. Most-cited named influences: Metallica 6, Megadeth 5, Pink Floyd 4, then Sepultura, Pantera, Alice in Chains, Motörhead, Deep Purple and Iron Maiden at 3 each. Ten influences are cited across more than one strand, and Pink Floyd is the only one reaching three (progressive, mainstream rock, heavy metal).

Result 3 — influence citations are structured by strand, but so is the evidence for them.

Metal and progressive acts name specific bands; pop names one act across seven bands and folk names none at all. Two explanations are compatible with this and the catalogue cannot separate them: technical genres may transmit through identifiable models while pop and folk transmit through diffuse convention, or metal and progressive musicians are simply interviewed about influences more often — which is demonstrably true, since the metal scene has an academic literature and the pop strand has none. The asymmetry is reported here as a property of the evidence, not as a property of the music.

Citations grouped by the citing act's founding era show the expected vector shift — Clapton, Led Zeppelin, The Doors and Knopfler from founding-era acts; Metallica, Pantera, Sepultura from the metal and cassette era; Megadeth, Pink Floyd and Lamb of God from the digital transition. The counts behind that shift are small: 2, 6, 7 and 1 citing acts respectively. The streaming-era column rests on a single act and should not be interpreted at all.

7.3 Localisation mechanisms

Documented adaptation mechanisms across the catalogue: Bangla language choice 36 acts, folk or Baul motif 20, political or social content 20, literary or poetic sourcing 8, religious or spiritual 4, urban vernacular 4.

Result 4 — folk is a cross-strand resource, not a genre. Folk motifs are documented in 20 of 63 acts while only 5 sit in the folk strand by primary genre: in progressive rock (Shironamhin's Shironamhin Rabindranath engaging Tagore's repertoire), mainstream rock (Nagar Baul's folk-rock tag), pop (Habib Wahid, Minar, Nancy) and hip-hop (Jalali Set's folk-hip-hop fusion). This is the quantitative counterpart to Mridha & Begum's (2023) claim that band music is a recent development inside a longer folk continuum.

Result 5 — strands localise on different registers. Progressive on the literary register; folk fusion on the motif register; hip-hop on the vernacular register (3 of the catalogue's 4 urban_vernacular markers); heavy metal, at its 1991 pivot, on the linguistic register alone — the riffs did not change, the language did.

A negative result worth publishing. Hip-hop shows only one political_or_social marker, against 6 for alternative/indie and 5 for mainstream rock. This flatly contradicts Hasan & Kundu (2021, 2022) and the catalogue's own Hannan record, which ties a 2024 breakthrough to the mass uprising. The marker counts are measuring how the records were written — the rock profiles researched with political framing in view, the hip-hop profiles from thinner biographical sources — not what the music does. Reported as a data artefact, with re-researching the strand's records as the fix rather than adjusting the marker list.

7.4 The co-appearance network

(Figure 4; data/networks/bmpn-multilayer.json, analysis/outputs/bmpn-metrics.json)

Result 6 — the network is genuinely structured, and its communities are interpretable. Louvain modularity of 0.428 on observational layers alone is comfortably above the level at which community structure is considered meaningful. The six communities of three or more acts:

Highest eigenvector centrality on the observed largest component: Cryptic Fate (0.409), Artcell (0.388), Nemesis (0.364), Warfaze (0.349), Chirkutt (0.314). Highest betweenness: Karnival (0.080), Nova (0.069), Nemesis (0.068), Powersurge (0.065), Warfaze (0.033).

Result 7 — this is a co-appearance network, and calling it a preference network would be false. Its edges record shared bills, shared members and stated influences. The classic mainstream-rock community exists because two musicians moved between those bands. Lalon Band and Chirkutt reach degree 18 because festival programmers put a folk act on nearly every bill — programming policy, not audience overlap, and the co-billing layer cannot distinguish them. Strand assortativity of +0.05 (observed) says almost nothing: acts are barely more likely to share a bill with their own strand than with another, which is consistent with §7.5 and inconsistent with reading these communities as taste communities.

Result 8 — a methodological finding. Uncorrected hypergraph projection depresses observed modularity from 0.428 to 0.159, because a single twelve-act bill contributes 66 undifferentiated edges. Both figures come from the same pipeline run (the observed and observed_uncorrected passes), so the comparison is reproducible rather than asserted. Any study building a scene network from festival lineups needs a bill-size correction or it will primarily measure festival size.

Coverage caveat, applying to every number above. 35 of 63 acts have no documented event. Meghdol holds degree 8 and betweenness exactly 0 — not because its audience is isolated but because its live history before 2023 is not in the dataset. Centrality here is substantially a measure of festival participation, which is itself a measure of what got written about. No hip-hop act's centrality in this dataset is interpretable.

7.5 The concert ecosystem

(Figure 7; analysis/outputs/concert-ecosystem.json)

Result 9 — the study's own typology is falsified. The methodology predicted that metal and rock festivals would be low-diversity subcultural gatherings and university festivals the eclectic discovery spaces. In the catalogue, every multi-act bill mixes BMEM strands — cross-strand rate 1.00 in every measurable class. The 2020 Joy Bangla Concert placed indigenous folk (F Minor), pop (Minar), hard rock (AvoidRafa, Vikings), alternative (Shunno, Arbovirus, Conclusion, Arekta Rock Band) and metal (Cryptic Fate) on one stage.

The necessary qualification: the catalogue's festivals are disproportionately Joy Bangla Concert editions, programmed by a youth organisation for national commemoration. The corrected claim is therefore narrower than the raw number suggests — the large commemorative festival is a discovery space; the commercial metal festival remains untested.

Result 10 — an entire defined class is empty. Zero urban-hiphop-event entries across eight catalogued hip-hop acts. This is a sourcing failure: the project's event sources are news outlets that cover rock festivals, and Bangladeshi hip-hop events are promoted on platforms those outlets do not follow.

Result 11 — the live circuit is a Dhaka circuit that has begun to move. 19 of 23 events are in Dhaka. The exceptions run one way: a 2016 RockNation Sylhet date, Chattogram's Bay of Bengal opening the 2019 Joy Bangla Concert, and the 2024 edition relocating to Chattogram's MA Aziz Stadium — the first held outside the capital. Three data points, one direction.

Result 12 — the largest documented demand figure is for a free event. More than 32,000 online registrations for the 2018 Joy Bangla Concert, which charged no admission. The catalogue contains no reliable evidence about what audiences pay for live music.

8. Case studies

Developed in full in analysis/case-studies/:

• Warfaze — the legitimation arc: cover repertoire (1984–90) → original Bangla metal (1991) → Ekushey Padak (2026). Localisation here is linguistic before it is musical; the riffs did not have to change for the music to become Bangladeshi.
• Artcell — the clearest documented instance of an imported technical apparatus filled with entirely local semantic content. Highest weighted degree in the network; 10 documented events across four typology classes; a seventeen-year album gap against a catalogue median of five.
• Meghdol — literary indie arriving at the festival circuit in 2023, twenty years after forming. Degree 8, betweenness 0: the clearest illustration of what the network measures and what it does not.
• The hip-hop lineage — a thirty-year documented lineage with near-zero presence in every measurement the project can make.
• Folk fusion and platforms — folk as a cross-strand resource, carried into the mainstream by a branded studio platform and a commemorative festival rather than by labels.

9. Discussion: the framework against the data

frameworks/bmem-validation.md tests all six BMEM dimensions and proposes ten revisions. The substantive ones:

Two dimensions are corrected outright. Dimension 4's operationalisation is a co-appearance network, not a preference network, and should be renamed until audience-side evidence exists. Dimension 5's typology has the diversity prediction backwards.

Local adaptation needs sub-structure. A single "Global → Local" arrow hides four different registers of adaptation (literary, motif, vernacular, linguistic), two directions of travel (Artcell imports the form and supplies local content; Meghdol starts local and reaches for an imported idiom), and a pathway the model has no arrow for at all — internal adaptation, from marginal or classical repertoires toward the popular mainstream. F Minor arranging Garo and Marma community material and Shironamhin engaging Tagore are both this, and neither is a global-to-local movement.

Global influence needs a distinction. Named-model influence (a specific act, usually with an interview behind it) and ambient-tradition influence (a genre convention absorbed without a stated model) are documented by different means and carry different evidential weight. Treating them as one quantity produces the strand asymmetry in §7.2 and invites reading it as a fact about music.

Industry transformation needs non-commercial funders. The catalogue's most consequential recent events are funded by a state-adjacent youth platform (Joy Bangla Concert: free entry, 32,000 registrations) and a beverage brand's cultural-prestige project (Coke Studio Bangla). A dimension modelled as cassette → CD → streaming has no place for either, and in this catalogue those two carry more of the ecosystem than the labels do.

And most of that dimension is untested. The model's claims about monetisation, royalties and algorithmic visibility have no supporting data here. They are hypotheses, and this draft marks them as such rather than restating them as findings, which the first draft did.

10. Limitations

1. Sample size and construction. 63 of a target 300 artists; 23 of 500 events. The catalogue was built outward from a well-documented metal core, which plausibly biases every strand comparison in this paper.
2. No audience data of any kind. Every claim about listeners in this study is inferred from co-billing and from what artists and journalists said. The study's central construct — the preference network — is not yet built.
3. No audio features. Every claim about tempo, production or arrangement rests on documented description and discographic sequence. Blocked pending Spotify and YouTube API credentials.
4. Documentation bias is measurable and material. The hip-hop political-content result (§7.3) is a demonstrated instance of a marker count reflecting how records were written rather than what the music does. The same mechanism plausibly affects the strand influence asymmetry (§7.2) and the metal-recruitment observation (§7.1).
5. Geographic bias. 19 of 23 events in Dhaka. The ecosystem outside the capital is nearly invisible here.
6. Retrieval constraint on this pass. Records added in this pass are snippet-sourced (§5.1) and tagged accordingly.
7. Two acts in the progressive strand. Any statement about "the progressive strand" is a statement about Artcell and Karnival.
8. Recency asymmetry. Recent decades have better event coverage and worse formation coverage than older ones. Trends crossing that boundary should be read with care.

11. Future trends

Set out with evidence grades and falsification conditions in docs/future-trends-forecast.md. In brief: the single becomes the default release unit (grounded); the festival circuit decentralises out of Dhaka (indicated); cross-strand billing continues and strand boundaries soften (grounded); institutional and brand funding grows relative to label funding (indicated, with a political caveat about the Joy Bangla Concert's own funding model); hip-hop's documented presence grows sharply from a base the record understates (indicated); long-lived bands persist through personnel turnover (grounded); metal's recruitment closure (speculative, and explicitly not to be relied on).

The forecast also states what this project cannot forecast at all: streaming economics, audience size or composition, AI's role in production and recommendation, and the commercial viability of any strand.

12. Conclusion

This study set out to model Bangladeshi music as an ecosystem rather than a chronology, and it now does so on measured rather than asserted foundations. The ecosystem it describes is structured — communities are detectable at modularity 0.428, influence vectors differ systematically by strand and era, and folk functions as a resource drawn on across every strand rather than as a genre among genres. It is also more institutionally arranged than the framework anticipated: the acts that hold the network together do so through shared bills programmed by a youth organisation, through musicians moving between bands, and through a branded platform, more than through anything a label did.

The study's most useful contributions may be its negative results. Its own concert typology is falsified by its own data. Its central construct, the preference network, turns out not to be what was built. One of its localisation findings is a documentation artefact and is reported as one. Establishing those things required building the pipeline that could detect them, and they constrain what the next iteration should claim in a way that a more confident draft would not have.

What remains is largely a data problem with a known shape: audience-side signals (blocked on API credentials), the hip-hop and university-fest live circuits (blocked on nothing but collection effort), and the artist catalogue's remaining four-fifths. The framework, the pipeline and the standard of evidence are now in place to absorb them.

13. Reproducibility

pip install networkx matplotlib jsonschema scipy

python3 analysis/scripts/validate_data.py        # schema + integrity
python3 analysis/scripts/build_bmpn.py           # network layers
python3 analysis/scripts/analyze_bmpn.py         # communities + centrality
python3 analysis/scripts/temporal_analysis.py    # decade curves
python3 analysis/scripts/concert_ecosystem.py    # typology
python3 analysis/scripts/influence_analysis.py   # citations
python3 analysis/scripts/make_visualizations.py  # figures

All scripts are idempotent and overwrite their outputs. Every number in this paper comes from analysis/outputs/ or data/networks/; every figure is rendered from those files. The dataset is JSON under data/, schema-validated with zero errors at the time of writing.

Figures. Fig. 1 formations by decade; Fig. 2 releases by decade; Fig. 3 strand formation timeline; Fig. 4 the co-appearance network; Fig. 5 most-cited global influences; Fig. 6 influence vectors by founding era; Fig. 7 concert typology. All in analysis/visualizations/.

References

Full APA entries with annotations: docs/annotated-bibliography.md. A wider landscape scan — new scholarship (including on hip-hop and the July 2024 uprising), network-methodology literature, and industry sources not yet integrated into this draft — is in docs/research-landscape-extended.md.

Autul, M. R., et al. (2024). Rocking across borders. International Journal of Computer and Digital Systems. Hasan, M. (2015, December 4). Rock 'n' roll, social change and democratisation in Bangladesh. South Asia @ LSE. Hasan, M., & Kundu, P. (2021). Conformers and rebels. In Masks of authoritarianism (pp. 173–183). Hasan, M., & Kundu, P. (2022). Hip-hop music activism. In The emergence of Bangladesh (pp. 405–417). Palgrave Macmillan. Mitra, U. (2008). Image of urban youth in the lyrics of Bangla bands. West Bengal Sociological Review, 1, 112–113. Mitra, U. (2014). Exploring youth [Doctoral dissertation, Jadavpur University]. Mridha, M. A. H., & Begum, M. (2023). Continuum of folk to pop music in Bangladesh. Issues in Social Science, 11(2). Mukherjee, K. (2017). Bangla rock. International Journal of Pedagogy, Innovation and New Technologies, 4(2), 35–47. Newman, M. E. J. (2001). The structure of scientific collaboration networks. Proceedings of the National Academy of Sciences, 98(2), 404–409. Pervez, A. (2012). Music and identity. Journal of Bangladesh Studies, 137, 42–55. Quader, S. B. (2016). Forms of capital in the Dhaka metal scene. Metal Music Studies, 2(1), 5–24. Quader, S. B., & Redden, G. (2015). Approaching the underground. Cultural Studies, 29(3), 401–424. Yoon, J.-S. (2019). Popular music of Bangladesh. South Asia Studies, 25(2), 59–116.

---

Repository: https://github.com/smile-plzz/bangladesh-music-evolution Data: data/ · Analysis: analysis/ · Frameworks: frameworks/