Ancient writing systems in the Mediterranean

A critical guide to electronic resources

Linear B

- 14th - 13th century BC

edited by: Maurizio Del Freo     DOI: 10.25429/     (translation by Melanie Rockenhaus)
Last update: 4/2022

  • Introduction
  • Index
  • Further information

Tablet An 657 from Pylos, Messenia, end 13th century BC. (National Archaeological Museum of Athens). Courtesy of the Palace of Nestor Excavations, Department of Classics, University of Cincinnati.

Linear B is a script documented on Crete and on the Greek Mainland from the 14th to the 13th century BC and used to write an archaic form of Greek. The oldest group of texts, found at Knossos in the so-called “Room of Chariot Tablets”, can be dated from the beginning of the 14th century BC (or from the end of the 15th century BC). The other texts from Knossos were produced later, in the first half of the 14th century BC according to the majority of researchers, or at the end of the 13th century BC according to others. On the other hand, the texts found on the Greek Mainland can be dated with a few exceptions from the second half of the 13th century BC.

Go to the online resources.

Online resources


Origins of the writing system

Linear B has more than 70% of its syllabograms in common with linear A There is therefore no doubt that it was born from this latter writing system. During the process of adaptation, some signs from Linear A were suppressed, others were created ex novo, almost certainly because Linear A was at least in part inadequate for Greek phonetics. The signs of Linear B are often more elaborate than the homomorphic Linear A signs. Linear B would thus seem to derive from a form of Linear A that is different from that found in the documents which have arrived to us. Moreover, Linear B was used only for bookkeeping purposes by officials working in the Palaces. Archaeological evidence suggests that in Greece the conditions necessary for the rise of a Palace civilization were reached only around the beginning of the 15th century BC. Linear B must consequently have been born around the 15th century. The 1994 discovery in Kafkania, near Olympia, of a  pebble engraved in Linear B in a non-palace context, datable to the 17th century BC, contrasts sharply with this scenario and raises serious doubts about the authenticity of the inscription.
Given the current state of knowledge, it is impossible to say whether Linear B was born on Crete or on the Greek Mainland.

Types of inscriptions

In total about 6,000 inscriptions in Linear B have arrived to us, 98% of them engraved on clay archival documents: tablets (in the form of a page or a leaf), labels and nodules of various types. The rest are painted on terracotta vases, mostly large “stirrup jars” in coarse clay, destined for the trade of olive oil and wine. To these must be added an inscription on an ivory seal and one on a possible weight of stone. Serious doubts have been expressed about the circumstances of finding of a stone pebble at Kafkania, near Olympia and of an amber seal at Bernstorf near Munich.
The inscriptions on archival documents register economic transactions (lists of commodities, entry and exit of goods). The page and leaf tablets contain respectively final and preliminary registrations. The labels have on the back the impression of a container (usually a wicker basket) and on the front a brief text referring to the content of the tablets archived in the containers. The nodules almost always have the impression of a seal and brief texts about the transaction certified by the seal. The inscriptions on the stirrup jars refer to places and people most probably involved in the production of the goods contained in the vases. It is therefore probable that the inscriptions on these jars were also economic-administrative in nature.
The tablets, labels, nodules and stirrup jars mostly came to light in the Palaces. When they come from buildings outside the Palaces, the contents of the texts and in some cases the relationships between these texts and the texts found in the Palaces have led researchers to identify these buildings as Palace annexes. It therefore seems that there was no “private” bookkeeping in Linear B.

Area where inscriptions are found

Linear B inscriptions have been found on  Crete (Armeni, Khania, Knossos, Mallia) [clickable map] and in Mainland Greece (in Argolis: at Mycenae, Tiryns and Midea; in Messenia: at Pylos and Iklaina; in Laconia: at Hagios Vassilios; in Boeotia: at Thebes, Kreusis, Orkhomenos and Gla; in Attica: at Eleusis; in Phocis: at Medeon; in Thessaly: at Dimini and Volos).The archival documents have been found at Knossos, Khania, Mycenae, Tiryns, Midea, Pylos,Iklaina, Haghios Vassileios, Thebes and Volos, while the inscriptions on vases have been found everywhere, except at Pylos, Iklaina, Haghios Vassileios, Volos and Medeon in Phocis. This last place has given us the only certain example of a Linear B inscription on a seal, while a possible inscribed stone weight comes from Dimini.


Linear B was deciphered by the English architect Michael Ventris in 1952. In reality, Ventris had already begun to study the problem when he was still very young, in the 1930s and ‘40s, but he returned to his study only after the end of World War II and after completing his degree in architecture.

When he began to systematically study the problem again in 1950, Ventris knew that Linear B was a writing system formed by logograms, metrograms, arithmograms and syllabograms, that the numeric system used by the scribes was decimal (Evans 1935), and that the metrograms were organized in systems of multiples and submultiples (Bennett 1950). He also knew that the syllabograms, counting less than 100, had to correspond to open syllables, that is, syllables formed by either a vowel or a consonant + vowel. Lastly, thanks to the observations made by Alice Kober on the alternations of signs and groups of signs at the end of words (1943), he knew that the language Linear B represented had to be an inflecting language.

Ventris’ method was statistical and combinatorial, centered on the study of the frequency and alternations of syllabograms, whatever their phonetic values. To this end, Ventris used the first reliable list of Linear B syllabograms, published by Emmett Bennet in 1951. Based on a statistical property typical of writing systems that only note down open syllables, Ventris hypothesized that the syllabograms most frequent at the beginning of words and less frequent within and at the end of words corresponded to vowels. Moreover, since the regular alternation among the syllabograms at the end of words seemed to serve an inflective function and express distinct grammatical cases, he hypothesized that the alternating syllabograms had an identical consonant and a different vowel (as in Latin ni and no, respectively in the genitive singular domi-ni and in the dative/ablative singular domi-no). Vice versa, in the cases in which the alternation expressed distinct grammatical genders, as in the lists of men and women, Ventris hypothesized that the syllabograms utilized for the same gender had an identical vowel and a different consonant (like Latin ni and li in the nominative plural masculine domi-ni and famu-li as compared to na and la in the nominative singular feminine domi-na and famu-la). Based on these and other similar observations, towards the end of 1951 Ventris began to organize the syllabograms on a grid, placing those with the same vowel in one column and those with the same consonant in a single row, whatever their phonetic value. The grid underwent various reorganizations and corrections over the following months as the relationships among the syllabograms became more and more precise. The publication of the first edition of the texts from Pylos by Bennett at the end of 1951 played an important role in this process.

The turning point came on 1 June 1952, when Ventris – in his Work Note No. 20 titled “Are the Knossos and Pylos tablets written in Greek?” – formulated some hypotheses, which he called “a frivolous digression”, about how to read Linear B. The understatement was in large part due to the fact that at that time archeologists and historians categorically excluded the possibility of a Greek presence at Knossos in the Late Bronze Age.

The hypotheses at the base of the “digression” were the following:
1) that each of the groups of three alternating syllabic sequences, identified by Alice Kober in the Knossos epigraphic material (the so-called “Kober’s triplets”), were composed of a toponym and by two ethnic designations (analogous to the cities and corporations registered in the Ugarit tablets);
2) that some of these toponyms and ethnic indications had survived into the classical period;
3) that among these there were Amnisos (the port of Knossos according to Str. X, 4, 8);
4) that the syllabogram *08, given its very high initial frequency, corresponded to the vowel a (which is in general the most frequently used vowel);
5) that the syllabograms *06 and *37 had the same phonetic value as the homomorphic syllabograms of the Classical Cypriot Syllabary, i.e.,  na and ti;
6) that the phonetic value of the syllabogram *30 of the grid, located at the intersection of the row of na and the column of ti, was ni;
7) that in Linear B, for the notation of groups of consonants, the rule of the "silent vowel" was used, as in the Classical Cypriot Syllabary, and so the toponym Amnisos would graphically be rendered as a-mi-ni-so (with the i of mi written but not pronounced), i.e., based on the preceding hypotheses *08-?-*30-?. 
Now, the only sequence among “Kober’s tiplets” which was compatible with *08-?-*30-? was *08-*73-*30-*12. Ventris therefore assigned *73 the phonetic value mi and *12 that of so. This was encouraging, since in the grid *73 (mi) was in the same column as *30 (ni). As for *12, the phonetic value so automatically caused the vowel o to be attributed to all of the syllabograms situated in the same column, among which there was also *52. This latter, which was on the same row as *30 (ni), could be read as no. The fact that among Kober’s triplets there was also the sequence *70-*52-*12, and that – based on the preceding observations – this sequence could be read as ?o-no-so (in the grid *70 was in the same column as *12), was a further encouragement. The sequence ?o-no-so could, in fact, be read as ko-no-so, and identified with the toponym Knossos, which would allow the attribution of the phonetic value ko to the syllabogram *70. In this way, a chain reaction was set off, thanks to which Ventris could propose the readings a-mi-ni-si-jo and a-mi-ni-si-ja for the ethnic adjectives from a-mi-ni-so, the readings to-so and to-sa for the formulas for total, and so on.

Despite these encouraging results, Ventris remained skeptical about the possibility that documents in Linear B were written in Greek when he was reaching his conclusions. What led him to doubt the correctness of his decipherment were especially some anomalous and unexpected scripts like ko-wo and ko-wa for ‘boy’ and ‘girl’ (instead of ko-ro and ko-ra, cf. Attic koros, kore) or the use of the syllabogram *78 for the enclitic conjunction te ‘and’ (instead of the syllabogram *04 te).

The collaboration with John Chadwik, Greek scholar from the University of Cambridge, begun in July 1952, helped dissipate these doubts. Most of the anomalous or unexpected scripts could in fact be explained by the archaic nature of the Greek used in the tablets. Thus, for example, ko-wo and ko-wa turned out to be the scripts for korwos e korwa, archaic forms of koros and kore, while the use of syllabogram *78 for the enclitic conjunction became understandable as soon as it became clear that the Greek used on the tablets still retained the occlusive labiovelar sounds inherited from Indo-European (te in fact comes from *kwe, cf. Latin -que).
In 1953, when the majority of the uncertainties were dissipated, Ventris and Chadwik published the results of the decipherment in the article “Evidence for Greek Dialect in the Mycenaean Archives”.

The definitive confirmation of the correctness of the decipherment arrived the same year, when Carl Blegen, the archeologist responsible for the excavation of the Palace of Pylos, communicated to Ventris and Chadwick that by applying the phonetic values they proposed to some hitherto unpublished tablets, it was possible to read, in front of an unequivocal drawing of a vase with three feet, the word ti-ri-po-de, ‘tripod’.

Type of writing system

Linear B is a logo-syllabic writing system, i.e., it is formed by logograms and syllabograms. Logograms are signs which correspond to words, while syllabograms are signs which correspond to syllables. The Linear B syllabary is composed of 87 signs (some are used as logograms as well). The syllabograms can be divided into simple, complex and doublet (also called homophone) syllabograms. The first correspond to syllables formed by a vowel (a, e, etc.) or by a consonant + a vowel (pa, ro, etc.), the second to syllables formed by a consonant + a consonant + a vowel (nwa, pte, etc.), and the third are phonetically similar to other syllabograms (a2 similar to a, pu2 similar to pu, etc.) but specialized in the graphic representation of certain sounds (e.g., a can graphically render both [a] and [ha], whereas a2 is specialized in graphically rendering only [ha]). There are some rarely-occurring syllabograms whose phonetic value is still unknown. For this reason they are not transliterated, but simply transnumerated (*18, *19, *22, etc.).

As regards orthographic rules, Linear B does not distinguish long vowels from short vowels (po-ro, pōlos ‘colt, foal’; po-de, podei ‘foot’), l from r (a-pi-po-re-we, amphiphorēwes ‘amphorae’; e-re-u-te-ro, eleutheros ‘free’), voiceless from voiced consonants (ke-ra, geras ‘privilege’; ke-ra, keras ‘horn’) [except for dentals (de-so-mo, desmos ‘bond’; te-me-no, temenos ‘reserved land’)], non aspirated from aspirated consonants (te-o, theos ‘god’; te-ko-to-ne, tektones ‘carpenters’). Final consonants of words are not noted down (pa-te, patēr ‘father’; ka-ke-we, khalkēwes ‘bronze smiths”; po-me, poimēn ‘shepherd’). The same is true for the final consonants of syllables (pa-te, pantes ‘all, every’) and for the s placed before a consonant (pa-i-to, Phaistos ‘Phaistos’). Consonant groups are graphically rendered using “silent vowels” (a-mi-ni-so, Amnisos; wi-ri-no, wrīnos ‘leather’). They may be rendered using a doublet syllabogram (di-pte-ra3, diphtherai ‘hides’). The second part of a diphthong may or may not be noted down (qo-u-qo-ta or qo-qo-ta, gwougwotās ‘oxherd’; wo-i-ko or wo-ko, woikos ‘house’). They may be rendered using a doublet syllabogram (e-ra3-wo, elaiwon ‘olive oil’; a3-ku-pi-ti-jo, Aiguptios ‘Egyptian’). Vice versa, the transitional semi-vowel sounds y and w which occur between i and u + vowel are regularly noted down (ku-pi-ri-jo, Kuprios ‘Cypriot”; ku-wa-no, kuanos ‘blue glass paste’).
As regards the logograms, about 170 have been identified to date and they represent men, animals, objects and foodstuffs. There are also five symbols for numbers (1, 10, 100, 1000 and 10000) and nine metrograms (five for weight measurements and four for volume measurements). The logograms can be divided into simple logograms (VIR ‘man’, MUL ‘woman’, VIN ‘wine’, SUS ‘pig’, etc.) and “ligatures” (combinations of logograms and syllabograms with an acrophonic function, e.g. SUS+SI, where SI is an abbreviation for si-a2-ro, i.e., sihalos ‘fattening pig’).

Categories of signs similar to the logograms are the abbreviations (isolated syllabograms with an acrophonic function; e.g. KU, abbreviation of ku-mi-no, i.e., kuminon ‘cumin’) and the monograms (resulting from the fusion of several syllabograms in one single outline; e.g., ME+RI, i.e., meli ‘honey’). Logograms are often preceded by isolated syllabograms with an acrophonic function. These syllabograms, which serve to specify the meaning of the logograms, take the name of adjuncts (e.g. pa OVISm means ‘old ram’; pa in fact is the abbreviation of  pa-ra-jo, i.e., palaios ‘old’).
Lastly, as for the direction of writing, Linear B inscriptions are written left to right.