Benjamin Schwartz — Mathematical and Structural Analysis

The Scholar and His Approach

Benjamin Schwartz (died 1981) was an American linguist who taught as Associate Professor of Classics at Lincoln University in Pennsylvania from 1956 to 1975. His principal research interests spanned Indo-European languages, Anatolian texts — particularly Hittite and Luwian — and Judaeo-Greek hymns. In 1959 he published two articles on the Phaistos Disc in the Journal of Near Eastern Studies: "The Phaistos Disk" (vol. 18, no. 2, pp. 105–112) and "The Phaistos Disk II" (vol. 18, no. 3, pp. 222–226). The timing was significant: Michael Ventris had deciphered Linear B only seven years earlier, in 1952, and the scholarly world was newly alert to what structural and statistical methods could extract from undeciphered Aegean scripts.

Schwartz's contribution was distinctive in its priorities. Where most of his predecessors had begun by guessing a target language and then forcing phonetic values onto the disc's signs, Schwartz reversed the order. He treated the disc first as a mathematical object — a data set of symbols, positions, and frequencies — and asked what kind of writing system could produce the patterns he observed. P. Jackson Macdonald, in a 1999 statistical study published in Kadmos, described the approach as "concise and trenchant and at least partly right," and noted that Schwartz's work stood out among scores of decipherment attempts. David Diringer considered it important enough to include in his monumental survey The Alphabet.

Main Claims

A Syllabary, Not an Alphabet or Logography

Schwartz's central argument rested on the compositional ratio of the disc. The text contains 241 sign tokens composed from 45 distinct sign types — a ratio of roughly 5.4 to 1. He placed this figure beside known benchmarks: alphabetic scripts typically produce ratios between 10 and 20 to 1, while logographic systems like Chinese yield ratios closer to 1 or 2 to 1. Syllabaries — Linear B among them, with its roughly 4-to-6-to-1 range — fall in between. The disc's ratio of 5.4 to 1 sat squarely in the syllabic zone. Schwartz took this as quantitative evidence that the disc records a syllabic script, an argument that has since become the majority scholarly position.

The Disc Encodes Real Language

Beyond classifying the script type, Schwartz argued the disc's content was genuinely linguistic — not decorative, not game-related, not calendrical. He examined the frequency distribution of individual signs and found it broadly consistent with Zipf's law, the empirical observation that in any natural-language text a small number of words account for a disproportionate share of total occurrences, while the majority appear only once or twice. The disc's frequency curve followed this pattern: a handful of signs — notably sign 02, the plumed head 𐇑, appearing 19 times — dominated the count, while more than a third of the 45 sign types occurred fewer than three times each. Schwartz regarded this as strong evidence against non-linguistic interpretations.

A Genetic Relationship with Cretan Scripts

Schwartz further asserted a genetic relationship between the disc's script and the Cretan linear scripts. He compared distributional properties of disc signs with those of Linear B, using the recently deciphered syllabary as a structural reference point. He proposed the disc recorded Mycenaean (Achaean) Greek in syllabic writing, reading Side A first, spiralling inward. Whether or not his specific language identification proves correct, his insistence that the disc belongs within the broader Aegean scribal tradition — rather than being an isolated import — shaped subsequent research, including later comparisons with Linear A by Torsten Timm and others.

Methodology

Frequency Mapping and Group-Length Statistics

Schwartz began by cataloguing every sign on both faces — 45 distinct types across 241 tokens — then ranked them by frequency. This table enabled comparison with the frequency profiles of known scripts. He also measured the disc's 61 sign-groups (31 on Side A, 30 on Side B): group lengths ranged from two to seven signs, with a mean of roughly three to four, closely matching word-length profiles in Linear A and Linear B — consistent with syllabic scripts, where each sign encodes a consonant-vowel pair and words tend to be short.

Positional and Bigram Analysis

Schwartz mapped where each sign type appeared within its group — first position, middle, or final — and found pronounced positional bias. Certain signs appeared almost exclusively at the beginning of groups, others only at the end. He also examined sign-pair co-occurrences across the disc and found the resulting matrix to be highly non-random: specific pairs appeared together far more often than chance predicts, while others never co-occurred at all. Both patterns are hallmarks of phonological and morphological rules in real writing systems.

Repeated-Sequence Identification

Schwartz identified approximately seven to nine two- or three-sign strings recurring in multiple groups across both faces — too regular to attribute to coincidence. He argued these functioned as grammatical elements comparable to the recurring sign clusters found in Linear B administrative texts.

Examples Illustrating Claims and Methodology

The Tripartite Sign Partition

Perhaps the most consequential finding to emerge from Schwartz's positional analysis is that the disc's 45 signs divide into three non-overlapping functional classes. Four signs — including 02 𐇑 (PLUMED HEAD) — appear exclusively in the first position of groups and never anywhere else. Twenty-six signs — including 12 𐇛 (SHIELD), the third most frequent on the disc at 17 occurrences — never once open a group; they appear only in non-initial positions. The remaining fifteen signs move between both domains.

The significance of this three-way partition is not that some signs are common and others rare — that is true of any script. It is that position and identity are locked together by absolute rules: sign 02 𐇑, the single most frequent sign on the entire artifact at 19 occurrences, does not stray into a non-initial slot even once in 19 chances. Sign 12 𐇛, the third most frequent, does not stray into an initial slot even once in 17 chances. These are not tendencies. They are categorical constraints — zero-exception rules that hold across both faces of the disc. In natural languages, such rigid positional partitioning maps onto the distinction between closed-class function words (articles, prepositions, determinatives — a small fixed set that marks grammatical roles) and open-class content words (nouns, verbs — a large set that carries lexical meaning). The disc's initial-only set of four signs is small and closed, exactly as one would expect of a class of grammatical markers. The non-initial set of twenty-six signs is large and open, as expected of a content vocabulary. Any valid interpretation of the disc — whether linguistic, administrative, or otherwise — must account for this partition: why four signs are structurally forbidden from appearing after the opening slot, and why twenty-six signs are structurally forbidden from occupying it.

The 02–12 Morphological Lock

Schwartz's bigram analysis uncovered internal structure within the sign-groups themselves — evidence that the groups are not mere strings of independent units but contain sub-sequences bonded by grammatical rules. The case in point is the pair 02–12 (𐇑𐇛, PLUMED HEAD–SHIELD). Of sign 12's 𐇛 17 occurrences on the disc, thirteen immediately follow sign 02 𐇑. The conditional probability runs strongly in both directions: if 𐇛 appears, there is roughly a 76 percent chance it was preceded by 𐇑; if 𐇑 appears, there is roughly a 68 percent chance 𐇛 follows. No other bigram on the disc approaches this level of mutual dependency. Meanwhile, the two signs occupy rigidly different positional slots — 𐇑 is locked into first position, 𐇛 never occupies it — so the pair forms a fixed two-sign opening unit: a group-initial marker followed by a second-position element that is functionally bound to it.

This pattern constrains interpretation. The pair 𐇑𐇛 behaves like a bound morpheme sequence — a determiner-plus-noun or prefix-plus-stem combination — whose internal order is governed by a rule. In Sumerian cuneiform, determinatives (semantic classifiers like DINGIR for "divine" or KI for "place") occupy fixed positions before the words they modify. In Egyptian hieroglyphic, royal names are enclosed in cartouches that begin with fixed-form markers. The 𐇑𐇛 pair on the Phaistos Disc exhibits the same structural logic: a classifier or grammatical marker in first position, followed by a dependent element in second position. Any proposed decipherment that assigns phonetic values to these signs must explain why they form a rigid sequential pair rather than appearing independently — and why sign 12 𐇛, despite being one of the most common signs on the disc, is never free to stand on its own at the head of a group.

Group-Final Signs and the Linear A Parallel

At the opposite end of the sign-groups, Schwartz found a complementary pattern: a small cluster of signs that appear disproportionately at or near the end of groups. Later work by Torsten Timm and others sharpened this finding. Signs 18 𐇡 (BOOMERANG), 23 𐇦 (COLUMN), and 35 𐇲 (PLANE TREE) — each occurring 11 or 12 times — cluster heavily in the penultimate or final position of their groups. The pattern is the mirror image of the prefix cluster at the front: where a closed set of signs marks group beginnings, a different closed set marks group endings.

The deeper significance emerged from cross-script comparison. When Giulio Facchetti published a systematic positional analysis of Linear A in Kadmos in 1999, the three most common word-final characters in Linear A turned out to match — in positional behaviour, if not in visual form — the three most common group-final signs on the disc. The overlap is structural rather than pictographic: the disc and Linear A share the same statistical fingerprint for how their endings are constructed. Timm calculated, using a hypergeometric distribution model, that the probability of this parallel arising by chance was roughly 1 in 40,000. This result gave quantitative weight to a claim Schwartz had made on broader grounds in 1959: that the disc's script belongs to the same scribal family as the Cretan linear scripts. The evidence lay not in whether individual signs looked alike, but in whether they occupied the same structural slots — a far more robust criterion, because positional behaviour reflects the underlying language's grammar, not the arbitrary shapes a scribe chose for individual characters.