Open research platform
A new evidence-led approach to the Indus script
IndusLanguage.com brings together inscriptions, archaeological context, sequence analysis and competing hypotheses to investigate what can responsibly be inferred about one of the ancient world’s great undeciphered writing systems.
The Indus script has no universally accepted decipherment. Any language or semantic reading shown here is a labelled hypothesis, not established fact.
Unique pictograms · locked corpus · no claimed readings
Induscript · Structural Atlas v1.0
Geometry first. Speech later, if ever.
The atlas will not speak for the corpus. It weights provenance, partitions direction, discounts casts, and quarantines narrative ontology so the signs can be measured without being ventriloquised.
Ideation stack · labelled hypotheses
Possible functions and meanings — tested, not claimed
Every card is a candidate: a structural role, a functional guess, or a published semantic proposal from Mahadevan, Parpola, Wells–Fuls, Farmer–Sproat–Witzel, Rao–Yadav–Vahia or Kenoyer/Vidale. None is a decipherment. Ranked on the locked compilation (2543 inscriptions · 592 unique pictograms).
031
Anthropomorph as agent / office
Human outlines are read as person, agent, office, or deity classifiers — a semantic class, not a personal name.
Why it is useful: Frequent enough to test. Use it as a structural anchor before any phonetic guess. Counter: Classifiers are borrowed from other scripts. Posture variants may be different signs.
740
Possessive / titular ending
Read as a morphological ending on names or titles (e.g. Parpola’s -an via ‘pot’ rebus; Mahadevan’s genitive-like closer).
Why it is useful: Frequent enough to test. Use it as a structural anchor before any phonetic guess. Counter: No bilingual; rebus from pictogram to phonetics is an extra untested step.
520
Lattice / grid
A six-cell lattice. Administrative partition, field sign, or ‘town’ pictogram have all been floated.
Why it is useful: Frequent enough to test. Use it as a structural anchor before any phonetic guess. Counter: Parpola-style ‘city’ readings and Farmer-style emblem readings are both extra steps.
003
Graphic class · stroke / numeral class
ICIT 003 is drawn in the stroke / numeral class. Class is a graphemic convenience used so relatives can be compared — not a word, not a translation.
Why it is useful: Frequent enough to test. Use it as a structural anchor before any phonetic guess. Counter: Wells over-splits; Parpola and Mahadevan merge. The family of this page is a working cluster.
235
Graphic class · geometric frame class
ICIT 235 is drawn in the geometric frame class. Class is a graphemic convenience used so relatives can be compared — not a word, not a translation.
Why it is useful: Frequent enough to test. Use it as a structural anchor before any phonetic guess. Counter: Wells over-splits; Parpola and Mahadevan merge. The family of this page is a working cluster.
861
Graphic class · built / architectural class
ICIT 861 is drawn in the built / architectural class. Class is a graphemic convenience used so relatives can be compared — not a word, not a translation.
Why it is useful: Frequent enough to test. Use it as a structural anchor before any phonetic guess. Counter: Wells over-splits; Parpola and Mahadevan merge. The family of this page is a working cluster.
820
Fish class / mīn rebus
Inside the Proto-Dravidian model the fish class is mīn ‘fish’ / ‘star’, used in names of deities or titles. Knorozov’s school reached a similar Dravidian fish reading independently in spirit.
Why it is useful: Frequent enough to test. Use it as a structural anchor before any phonetic guess. Counter: Language ID, rebus convention, and allograph grouping must all be true at once.
176
Graphic class · vessel / jar class
ICIT 176 is drawn in the vessel / jar class. Class is a graphemic convenience used so relatives can be compared — not a word, not a translation.
Why it is useful: Frequent enough to test. Use it as a structural anchor before any phonetic guess. Counter: Wells over-splits; Parpola and Mahadevan merge. The family of this page is a working cluster.
Every unique pictogram
592 unique catalogue numbers attested in this compilation, each with its own drawing. Fuls 2023 lists 709–713 ICIT signs; signs not in this digitization are not invented here. Click any tile for a full source-stacked dossier. Graphic aids only — not plate tracings, not translations.
592 pictograms in this view · each unique · each clickable
- 740
- 002
- 400
- 220
- 032
- 240
- 520
- 033
- 390
- 235
- 060
- 861
- 003
- 820
- 741
- 001
- 590
- 176
- 000
- 031
- 817
- 233
- 705
- 920
- 415
- 090
- 803
- 806
- 407
- 798
- 100
- 405
- 760
- 255
- 840
- 156
- 900
- 550
- 368
- 140
- 700
- 061
- 690
- 151
- 706
- 503
- 004
- 231
- 617
- 615
- 017
- 142
- 692
- 575
- 892
- 527
- 125
- 435
- 752
- 055
- 460
- 845
- 005
- 016
- 154
- 923
- 416
- 717
- 904
- 440
- 388
- 585
- 350
- 501
- 850
- 790
- 455
- 595
- 742
- 158
- 070
- 226
- 413
- 335
- 630
- 482
- 013
- 236
- 165
- 171
- 384
- 621
- 636
- 832
- 034
- 091
- 104
- 555
- 745
- 097
- 222
- 495
- 500
- 632
- 035
- 175
- 480
- 048
- 137
- 320
- 436
- 491
- 645
- 831
- 877
- 095
- 321
- 679
- 297
- 318
- 526
- 605
- 711
- 838
- 921
- 065
- 072
- 337
- 382
- 565
- 890
- 905
- 056
- 161
- 190
- 241
- 243
- 484
- 519
- 530
- 554
- 821
- 853
- 063
- 111
- 347
- 374
- 406
- 504
- 792
- 794
- 130
- 232
- 250
- 260
- 269
- 315
- 336
- 371
- 417
- 468
- 510
- 545
- 597
- 772
- 773
- 804
- 812
- 844
- 880
- 906
- 927
- 015
- 093
- 098
- 101
- 215
- 298
- 317
- 360
- 365
- 422
- 467
- 540
- 556
- 592
- 703
- 018
- 064
- 066
- 219
- 234
- 252
- 354
- 370
- 391
- 423
- 465
- 518
- 521
- 603
- 626
- 702
- 776
- 777
- 824
- 924
- 027
- 136
- 155
- 159
- 201
- 221
- 257
- 263
- 352
- 401
- 408
- 426
- 456
- 472
- 511
- 531
- 564
- 637
- 791
- 808
- 869
- 028
- 049
- 067
- 132
- 153
- 167
- 202
- 204
- 244
- 267
- 272
- 323
- 353
- 380
- 409
- 421
- 443
- 502
- 525
- 572
- 611
- 625
- 704
- 712
- 736
- 746
- 749
- 842
- 868
- 871
- 903
- 910
- 007
- 014
- 019
- 036
- 081
- 103
- 146
- 150
- 160
- 168
- 177
- 193
- 216
- 268
- 278
- 279
- 285
- 292
- 322
- 345
- 346
- 392
- 420
- 470
- 483
- 543
- 552
- 561
- 600
- 634
- 639
- 710
- 744
- 805
- 829
- 856
- 879
- 893
- 912
- 930
- 006
- 012
- 025
- 029
- 041
- 042
- 046
- 059
- 062
- 071
- 082
- 092
- 096
- 112
- 118
- 127
- 143
- 157
- 178
- 191
- 192
- 195
- 200
- 205
- 229
- 242
- 262
- 286
- 287
- 299
- 300
- 301
- 302
- 304
- 308
- 316
- 319
- 324
- 326
- 341
- 351
- 361
- 367
- 373
- 381
- 410
- 411
- 424
- 450
- 475
- 479
- 494
- 513
- 516
- 528
- 533
- 534
- 544
- 560
- 579
- 610
- 680
- 683
- 698
- 720
- 721
- 732
- 747
- 767
- 770
- 775
- 783
- 785
- 786
- 795
- 797
- 799
- 813
- 814
- 825
- 828
- 836
- 865
- 891
- 896
- 898
- 901
- 908
- 919
- 922
- 933
- 940
- 950
- 045
- 050
- 051
- 057
- 058
- 069
- 073
- 083
- 084
- 094
- 099
- 105
- 106
- 107
- 110
- 117
- 119
- 121
- 122
- 123
- 126
- 128
- 131
- 133
- 138
- 141
- 144
- 145
- 152
- 166
- 170
- 172
- 173
- 179
- 180
- 194
- 203
- 207
- 208
- 209
- 224
- 228
- 230
- 253
- 256
- 264
- 266
- 270
- 275
- 276
- 277
- 280
- 281
- 291
- 293
- 294
- 295
- 296
- 305
- 306
- 307
- 309
- 311
- 325
- 328
- 329
- 330
- 338
- 340
- 342
- 343
- 344
- 362
- 363
- 372
- 375
- 376
- 383
- 385
- 386
- 389
- 393
- 412
- 427
- 428
- 429
- 430
- 441
- 442
- 444
- 445
- 452
- 457
- 466
- 469
- 471
- 474
- 481
- 490
- 505
- 506
- 514
- 515
- 535
- 541
- 542
- 546
- 547
- 563
- 570
- 571
- 573
- 576
- 577
- 578
- 586
- 596
- 604
- 606
- 620
- 627
- 629
- 633
- 635
- 640
- 647
- 678
- 681
- 682
- 684
- 685
- 686
- 688
- 689
- 693
- 694
- 697
- 707
- 713
- 716
- 723
- 727
- 728
- 729
- 731
- 734
- 748
- 751
- 753
- 761
- 762
- 764
- 765
- 768
- 778
- 781
- 793
- 796
- 807
- 809
- 810
- 811
- 818
- 822
- 826
- 830
- 833
- 843
- 849
- 852
- 855
- 858
- 859
- 862
- 863
- 864
- 866
- 870
- 872
- 874
- 876
- 878
- 894
- 895
- 897
- 902
- 909
- 914
- 918
- 931
- 932
- 942
- 944
- 947
- 952
- 953
- 956
- 958
- 999
What can we know?
Signs have structure
Positions, neighbours and templates can be measured without inventing a language.
Objects have context
Site, material, motif and object type constrain function — not captions.
Meaning requires evidence
A semantic candidate is labelled, sourced and falsifiable, or it does not appear as a reading.
Explore the corpus
CompiledResearch compilation from a published ICIT-style digitization — not a complete or licensed photographic corpus
How a possible meaning is tested
- 01
Sign form
- 02
Position
- 03
Fragment
- 04
Template
- 05
Object context
- 06
Competing models
- 07
Prediction test
- 08
Provisional hypothesis
Research Case 001
Does the sign pair 033–705 form a recurring directional compound?
Illustrative workflow example. Requires verified corpus import for historical conclusions. No translation is offered.
Open the full caseExplore a sign
740
Sign 740
Three-pronged upright. Abstract stand-in for ICIT/G catalogue 740 — not a plate tracing.. Visual aid only — not a meaning.
Tokens 1267 · independent objects 1250 · role opener-biased
No established reading.
Open SignLabWhy the script is not yet deciphered
Short inscriptions
Most texts are only a handful of signs, which under-constrains any lexicon.
No bilingual anchor
There is no Rosetta Stone equivalent with a known language.
Uncertain sign inventory
Counts shift with how variants and ligatures are merged.
Unknown underlying language
Family identification is a hypothesis, not a starting fact.
Archaeological context gaps
Many objects lack fine findspot or rights-cleared images.
Competing models, not one answer
under investigation
Proto-Dravidian logo-syllabic
The most widely discussed linguistic hypothesis: a logo-syllabic system encoding an early Dravidian language, including rebus proposals such as mīn ‘fish / star’.
inconclusive
Indo-Aryan / Indo-Iranian-related
Proposals that the inscriptions encode an early Indo-Aryan or related Indo-Iranian language.
inconclusive
Munda / Para-Munda / Austroasiatic-related
The idea that an Austroasiatic-related language, sometimes called Para-Munda, underlies at least some inscriptions.
under investigation
Language isolate
The underlying language may be unrelated to any surviving family — a lost isolate or family.
stronger
Administrative coding system
Inscriptions may encode offices, commodities, lots or affiliations more like a structured code than running speech.
under investigation
Numeral / classifier system
Some signs, especially stroke groups, may encode quantities, measures or class marks rather than names.
weak
Non-linguistic emblem system
The signs might be heraldic or emblematic marks with little or no linguistic encoding.
under investigation
Mixed-register system
Different object classes may mix linguistic, numerical and emblematic registers rather than one uniform script.
Latest notes
18 September 2026
Welcome to the evidence, not the myths
Welcome to the evidence, not the myths. Platform journal notes are editorial, not excavation reports.
18 September 2026
Notes on the demonstration corpus
Notes on the demonstration corpus. Platform journal notes are editorial, not excavation reports.
18 September 2026
Preview: the 033–705 workflow
Preview: the 033–705 workflow. Platform journal notes are editorial, not excavation reports.
Follow the evidence, not the myths
We separate what is observed on an object from what is statistically inferred, functionally hypothesised and semantically proposed.
No spam. Methodology and case updates only.
Featured sequence from the locked corpus
Fish-like 033 as a Proto-Dravidian mīn rebus is a semantic candidate only. See the hypothesis card.
592 catalogue-style signs in the locked signary. Established meaning remains empty for every sign.