A sharp surge in hearing notices served under the Special Intensive Revision (SIR) of electoral rolls in West Bengal has triggered allegations of selective scrutiny. Field data reveals a strong correlation between minority-dominated districts and the volume of voters called for verification under the process. Data compiled from across the state indicates that the SIR’s “discrepancy” detection system is disproportionately flagging voters in Muslim-majority areas. There is growing concern that software-driven processes may be generating false positives in Bengali names, risking the silent disenfranchisement of legitimate voters before the upcoming Assembly election.The concern has sharpened after initial SIR data failed to consolidate support for the Bharatiya Janata Party’s (BJP) earlier polarisation attempts around Rohingya and Bangladeshi “infiltration”. Instead, the release of the Draft Roll and Absentee, Shifted, Suspected Dead (ASSD) list sparked resentment in Matua and other Hindu Scheduled Caste localities. These communities were hit heavily by deletions and no-mapping, due to which the narrative backfired. In recent weeks, however, the administrative focus appears to have shifted, through the SIR hearing process, towards minority-heavy Polling Stations (PS).In The Wire’s earlier constituency-level reading of the SIR draft rolls, “No Mapping” emerged as a key pressure point. Voters who could not be linked to the 2002 roll, the legacy anchor in the SIR 2025 framework, were effectively pushed into a higher-risk verification pathway. But the analysis also produced an inconvenient statistical outcome for BJP’s “infiltration” propaganda. Muslim-majority constituencies appeared among the most consistently mapped and, therefore, among the most documented. In Murshidabad, for instance, Domkal (77.67% minority population) showed just 0.42% No Mapping, while Raninagar (75.40% minority) recorded 0.91%, Hariharpara (74.96% minority) 0.60%, and Lalgola (77.29% minority) 1.1%.Muslim-majority districts see the most SIR hearingsYet the district-wise hearing numbers reviewed by The Wire tell a sharply different story. Across districts, the share of voters called for hearings climbs steeply with the Muslim population share. Statistical analysis of the dataset shows a strong positive correlation between a district’s Muslim population percentage and the percentage of electors called for hearing per constituency, with a correlation coefficient of 0.79. In practical terms, this implies that approximately 62% of the variation in hearing-call rates across districts can be explained by Muslim population density alone.The implication is stark: even where Muslims appear highly “mapped” (low No Mapping), Muslim-heavy districts still show the highest hearing intensity, suggesting the system driving voters into hearings is not primarily No Mapping, but other triggers such as “logical discrepancies”, duplicates and name-related mismatches, especially those amplified by script conversion and rigid matching rules.In Murshidabad, with a Muslim population of roughly 66%, 30.20% of electors were called for hearings, the highest intensity in the dataset. Similarly, Uttar Dinajpur (49.9% Muslim) and Malda (51.3% Muslim) recorded hearing rates of 29.75% and 28.42%, respectively.Crucially, these specific districts recorded very low No Mapping, a technical indicator that means a vast majority of voter records in the current electoral roll have been successfully connected (“mapped”) to a verified legacy record, specifically the 2002 electoral roll, which is being used as the anchor database in SIR 2025. In a neutral administrative exercise, districts with high linkage success should logically require fewer hearings.Speaking to the Wire, Subrata Gupta, appointed the Special Roll Observer for West Bengal by the Election Commission of India (ECI) has admitted there are challenges, but denied community profiling.“There are errors – we admit it. It could be around 15-20% in case of No Mapping. The software flags discrepancies in case of spelling mismatch. However, the mistakes can be rectified during hearings. Our idea is to have as clean a list as possible,” Gupta said.In sharp contrast, districts with low Muslim populations, such as Bankura (8% Muslim) and Purulia (7.8% Muslim), showed high mapping success rates, at 99.16% and 98.45%, respectively. Consequently, only roughly 10-13% of their voters were called for hearings.Statistically, a voter in Murshidabad (roughly 66% Muslim population) is approximately three times more likely to receive a hearing notice than a voter in Bankura. The fact that districts with high linkage success as well as high shares of Muslims in the population are facing hearing rates as high as 30% implies that the ECI’s much touted “logical-discrepancy” method introduced in West Bengal is being used to call up vast numbers of mapped and verified voters.Bengali name variations trip up SIR softwareThis shifts the focus to the software used by the ECI, reportedly developed by state-owned Centre for Development of Advanced Computing (C-DAC), and how it has handled Bengali names. In multiple instances, the system’s approach to name comparison has failed to adequately accommodate common variations in Bengali naming conventions, especially their transliteration between scripts.The Wire can confirm that in at least one case, a woman bearing a Muslim name as per the draft roll, whose father has a Hindu surname and whose first name is a common female name that is also often used for men in hilly regions of the state, was flagged by the system and issued a hearing notice.Also read: Verification or Disenfranchisement? People in Bengal, Tamil Nadu Struggle With SIRSince the voters are “linked” (meaning their existence in a family tree has been established), the notices are likely being triggered by name spelling, age mismatches or transliteration errors rather than identity gaps. Technical experts consulted by The Wire attribute this to a probable lack of robust fuzzy-matching algorithms capable of navigating Bengali naming conventions. This is particularly problematic for Muslim voters, where variations in surname spellings or a different surname from one’s father are common. Consequently, even a voter who has successfully provided documents and established linkage risks being flagged as a “discrepancy” merely because the software fails to reconcile a minor spelling variation between databases.“In Malda district’s Harishchandrapur Assembly Constituency, multiple minority-dominated polling stations recorded unusually high numbers, with PS 52 receiving 540 notices, PS 53 receiving 530, and PS 63 receiving 503. Correcting errors in the voter list is a constitutional responsibility of the Election Commission. But instead of doing that work, the commission is now emphasising exclusion,” alleged Communist Party of India (Marxist) state secretary Mohd Salim.The intensity of this scrutiny is visible in granular snapshots from other constituencies, which confirm that the high hearing volumes are a heavy burden on specific localities. The Wire has accessed data for over twenty constituencies, spread across five districts. All of them show a similar pattern.In Manikchak Assembly Constituency, also in Malda district, the hearing burden is being reported from minority-dominated booths, where notice counts run into several hundreds. Field data shows PS 24 alone receiving around 600 SIR hearing notices, while PS 23 and 25 each recorded around 300 notices. Other minority-heavy booths in the same cluster also show significant volumes: PS 28 reportedly has around 220 notices, and PS 56 around 130 notices. Taken together, these five minority-dominated booths account for roughly 1,550 hearing notices, indicating a concentrated and unusually high hearing load within a single constituency.Similarly, in Murshidabad’s Lalgola Assembly Constituency, the burden was heavy across multiple polling areas, with PS 165 recording approximately 535 notices and PS 163 recording 390, both with a high Muslim concentration.Hooghly offers a revealing contrast. At the district level, the average share of voters called for hearings is lower than the statewide mean, about 13.9% per constituency. Yet, field reports indicate that specific pockets such as Chanditala are witnessing unusually dense clusters of notices. In Chanditala, the concentration of Muslims sharpens at the panchayat level: Nawabpur panchayat (65% Muslim) reportedly recorded over 5,000 notices and Bhagabatipur (63% Muslim) over 4,500. By comparison, in Krishnarampur panchayat (10% Muslim) a little over 1,000 people were reportedly summoned.This divergence between a modest district-wide average and intense local spikes points to substantial within-district variation, reinforcing calls for transparency in the booth- and panchayat-level criteria, not just district aggregates.When verification turns into exclusionThis pattern raises serious questions about the safeguards – or the lack of them – built into the SIR process. Hearings triggered under the broad label of “logical discrepancies” appear to be turning into mass harassment, compelling thousands of daily-wage earners and economically vulnerable voters to miss work and spend money on travel simply to re-establish details that, by the mapping statistics, are already linked.While the ECI maintains that the hearings are a neutral mechanism to rectify errors and ensure a clean electoral roll, the data suggests that the cost of this cleanliness is being extracted unequally. The burden of correcting state-generated errors is falling overwhelmingly on the shoulders of minority voters.In effect, the SIR has procedurally prioritised software flags over established mapping data, creating a scenario where a voter in a minority district faces a triple penalty: getting documented, yet being flagged by the software, then being forced to bear the economic cost of the state’s technological limitations. Without immediate intervention to audit the “logical discrepancy” criteria for transliteration bias, West Bengal’s verification exercise risks becoming, in practice, a targeted exclusion of the very citizens it claims to verify.