Phase 1: Glyph Classification

Pattern Emergence & Sign Inventory Analysis

Index Overview Phase 1 of 5 Phase 2 Drops

Byblos Script Decipherment – Phase 1
Glyph Classification and Pattern Emergence

Phase 1: Byblos Script Sign Inventory and Pattern Analysis

Introduction: In this initial phase, we systematically catalog the Byblos script's glyphs and analyze their frequencies and patterns without imposing any predetermined meaning. The Byblos script (a.k.a. Pseudo-hieroglyphic or Byblian syllabary) consists of roughly 114 distinct signs as documented by Maurice Dunand, though modern analyses suggest many are variants of the same sign, reducing the core inventory to around 90 unique glyphs. This number of signs is far more than an alphabet (≈22–30 signs) but in line with a syllabary, where each sign represents a syllable (e.g. CV). Below, we present the Phase 1 findings: an inventory of glyphs with their occurrence statistics and provisional classification, followed by key recurring patterns (n-grams, positional trends) and identification of glyphs that may function as names, determinatives, or syntactic markers.

Glyph Catalog and Classification

Sign Inventory: The corpus comprises 14 known inscriptions (10 published by Dunand in 1945, plus 4 later fragments) on mediums like bronze tablets, spatulas, and stone stelae. Across these texts (~1,100 total characters), approximately 90 distinct glyphs can be discerned when accounting for scribal variants. Dunand's original list of 114 signs included some damaged or variant forms that are now understood as duplicates. Garbini and others estimate the true sign count ≈90, which strongly supports a syllabic script (since a Semitic syllabary with ~22 consonants × 3–4 vowels would need on the order of 66–88 signs).

Identification Scheme: For ease of reference, each glyph is assigned a provisional ID (e.g. BY001, BY002, etc.) in descending order of frequency. Table 1 below lists a selection of glyphs from highest to lower frequency, along with their occurrence count (and rank) and a tentative functional classification. (High-frequency signs tend to appear in many inscriptions, whereas low-frequency signs may be restricted to one context.)

Glyph ID	Occurrences (Rank)	Tentative Function	Confidence
BY001 – (graphic variant of Dunand E7)	64 times (1st)	Core syllabic sign (common CV syllable)	High – appears across multiple texts (likely phonetic)
BY002 – (variant of E8)	48 times (2nd)	Core syllabic sign (CV syllable)	High – widespread, likely phonetic
BY003 – (variant of G8)	41 times (3rd)	Core syllabic sign (CV syllable)	High – widespread, likely phonetic
BY004 – (variants H6/G17/E12 combined)	46 times (≈4th)	Core syllabic sign (CV syllable)	High – likely phonetic (appears in many contexts)
BY005 – (D3)	32 times (5th)	Core syllabic sign (CV syllable)	High – likely phonetic
BY006 – (B11)	29 times (6th)	Core syllabic sign (CV syllable)	High – likely phonetic
BY007	25 times (≈7th)	Common syllabic sign (CV)	High – likely phonetic
BY008	22 times (≈8th)	Common syllabic sign (CV)	High – likely phonetic
BY009	17 times (≈10th)	Syllabic sign (possibly vowel series)	High – likely phonetic
BY010	15 times (≈12th)	Syllabic sign	High – likely phonetic

Table 1: Representative high-frequency Byblos signs with provisional IDs. (Dunand's original codes in parentheses where applicable.) High-frequency signs (dozens of attestations each) occur in multiple inscriptions and are very likely the core phonetic syllabary signs. We assign High confidence that these represent syllabic sounds (consonant-vowel units), analogous to signs in other syllabaries (e.g. Linear B or Cypriot).

Medium-frequency signs (occurring ~5–14 times) form the next tier. These ~20–30 glyphs appear in limited contexts – some only on the large tablets or only on certain stelae – but recur enough to suggest they too carry phonetic values (perhaps less common syllables or context-specific sounds). For example, BY020 (≈10 occurrences) and BY021 (≈9 occurrences) are found on both a tablet and a spatula, indicating a consistent sound usage across texts (confidence Medium-High for phonetic function). Medium-frequency signs might include syllables used in specific words (e.g. religious terms or titles) that are not as common as everyday syllables.

Low-frequency signs are those with 2–4 attestations each, and a large number of hapax legomena (appearing only once) are present as well. Roughly 20–25 signs occur only a single time in the entire corpus, and many others only a few times. These rare signs are prime candidates for non-phonetic or special-use roles. We assign Low confidence to any broad phonetic value for these. Instead, they may represent determinatives, logograms, or components of proper nouns.

Variant Forms: Our cataloging also noted cases of allographs – where multiple drawn forms likely represent the same underlying sign. Dunand's list assigned different codes to what are now recognized as one sign in different styles (e.g. a curvy "monumental" form on stone vs. a simplified linear form on bronze). We merged such variants in the inventory: for example, Dunand's E15 vs. H7 appear only on different materials but share shape and likely value. These were grouped under one ID in Table 1 (BY001 corresponds to E7 and its slight variant E8, etc.). We identified ~25 cases of graphical variation where signs could be conflated into ~10 fundamental sign types (per Martin's classification approach), reinforcing that the true sign count is lower than 114.

Frequency Analysis: Occurrence and Distribution

Overall Frequency Distribution

The distribution of sign usage in Byblos inscriptions is highly uneven (as is typical for writing systems). A few signs are extremely common: the top 5–10 signs account for a large fraction of all text (~30–40% of all characters), while many signs are infrequent. The most common sign appears on the order of 60–65 times across the corpus, and at least 5 signs occur >25 times each. In contrast, about one-third of the signs appear fewer than 5 times, and of those, a significant subset are hapaxes (single attestations). This skew suggests functional differentiation: the high-frequency signs likely correspond to common syllables or grammatical elements needed repeatedly, whereas low-frequency signs could represent less common sounds or specialized meanings (names, numbers, etc.).

Contextual Distribution

We also examined which inscriptions each sign appears in, which helps determine their usage scope. Many high-frequency glyphs appear in all or most of the lengthy texts (e.g. on both tablets c and d, and at least one spatula or stele), underscoring their general-purpose role. For example, BY001–BY003 (top signs) occur on the large bronze plates and at least one stone fragment, indicating they were fundamental to the script regardless of medium. On the other hand, some medium-frequency signs are confined to a single inscription type: e.g. a particular glyph might be common in the two bronze tablets but absent from the stelae, suggesting it could be related to the tablets' subject matter (perhaps an administrative or ceremonial term).

Positional Analysis

A positional frequency scan was performed to see if certain signs tend to occur at the beginning or end of texts or words. Because Byblos inscriptions generally lack explicit word dividers (with a few exceptions, see below), identifying word boundaries is challenging. However, we noted a few patterns:

Some signs show a preference for text-initial position. For instance, one glyph appears at or near the start of multiple inscriptions – possibly a title or invocatory symbol. (It could mark the beginning of a formula like "Dedicated to…" or be an initial particle.)
Conversely, a handful of signs cluster at ends of texts or lines. Notably, a small circle-shaped sign was observed at the end of at least one long text, and Jan Best interprets this "O" glyph as a full stop or period marker. Its placement suggests a terminator rather than a syllable, reinforcing the idea of non-phonetic punctuation in use.
Within sequences (tentative "words"), certain signs frequently occur in final position, hinting they might be suffixes (grammatical endings). For example, several words on the tablets end in the same sign or pair of signs repeatedly, whereas other words end in a different common sign. This points to possible case endings or grammatical inflections recurring across the corpus.

Co-Occurrence Patterns

We generated a preliminary co-occurrence matrix to see which glyphs tend to appear together. While the full matrix is sparse (due to the limited corpus), a few statistically significant pairings emerged:

Frequent Bigrams: Some two-sign combinations recur often. One conspicuous bigram is a sequence interpreted as wa- (discussed by Best) which appears multiple times on the tablets. In purely structural terms, a particular glyph (BY041 perhaps) often follows the glyph BY012 across different contexts, indicating those two sounds/letters commonly go together (perhaps a prefix + following syllable). Another example: a glyph corresponding to a ya sound often follows certain consonant signs, suggesting a frequent -ya ending or diphthong.
Common Trigrams: A few three-sign sequences are repeated verbatim in the long texts. Tablets c and d notably contain several sequences that appear multiple times (sometimes exactly, sometimes with one sign altered). For instance, a sequence of four signs occurs in tablet c in several places; on tablet d a very similar sequence appears with one internal sign difference. These near-duplicates likely represent a fixed phrase (e.g. a job title, an offering item, or a ritual phrase) where one element can vary (like a name or quantity).
Co-Occurrence Clusters: Using unsupervised clustering, glyphs naturally grouped into a few clusters based on context. One cluster includes the most frequent signs which co-occur broadly (consistent with them being general vowels or syllables that combine freely). Another small cluster consists of rare signs that only co-occur with a specific other sign. This could indicate a compound sign or a determinative usage (for example, a rare glyph always follows a particular high-frequency glyph – possibly tagging it with a semantic category, as a determinative would).

In summary, the frequency analysis confirms a core sign repertoire used broadly (likely phonetic syllables), and a set of rarer signs with restricted use (pointing to names or semantic symbols). It also provides clues about the script's structure, such as possible punctuation and suffixing patterns.

Pattern Recognition: Repeating Structures and Syntactic Clues

Having quantified the signs, we next identify repeating patterns (n-grams) and structural regularities in the inscriptions:

Recurring Suffixes

A striking pattern is the recurrence of a few word-final sign sequences. A computational search found four distinct endings that attach to multiple different sign strings (words). These endings occur so regularly that they appear to function as grammatical suffixes. For example, one common suffix is a single sign (e.g. -⌖ where ⌖ denotes a particular glyph) that terminates numerous sequences, whereas another suffix is a two-sign combination (e.g. -XY where X and Y are specific glyphs) appearing on a different set of words. The presence of exactly four such repeating endings aligns with the morphology of Semitic languages, which often have a limited set of noun endings (for case, number, or gender) or verb conjugation suffixes.

Formulaic Phrases

Both visual inspection and n-gram analysis show that the longest texts (bronze tablets c and d) contain repeated phrases. Some sequences of 5–7 signs appear multiple times within a single text or across texts, often with slight modifications. This is typical of administrative or narrative documents that list several similar items or statements (for example, a list of offerings, titles, or participants). One concrete example: on tablet c, a sequence corresponding to "wa-X-ya-lu" occurs, and on tablet d a similar "wa-X-ya-le" occurs. The only difference is the final sign (-lu vs -le), which might be an inflection (perhaps masculine vs. feminine, or singular vs. plural). The beginning "wa" sequence is repeated exactly, which one researcher (Best) interpreted as the conjunction "and" in a Semitic language.

Positional Rules & Syntax

Beyond individual word patterns, we look at text-level structure. Preliminary evidence exists for punctuation-like symbols marking pauses or separation of clauses:

A curved bracket-shaped glyph appears intermittently and often precedes certain repeating words (like the "wa" mentioned above). Best hypothesized that a single curved sign ")" served as a comma, while a doubled instance "))" acted as a semicolon in these texts. In our analysis, the single curved sign does show up at junctures where a new list item or clause is suspected. This sign never occurs as part of a word (no other sign follows it on the same line without a space), supporting the idea it's a separator.
Likewise, a distinctive A-shaped sign is noted to appear in positions that could correspond to larger section breaks – possibly functioning like a colon or clause divider. And as mentioned, a circle sign ("O") is found at a full stop (end of inscription) in at least one case.
Word Division: Most inscriptions lack explicit word dividers, but interestingly, two of the shorter texts (on spatulas b and i) use vertical stroke marks to separate words. These are short vertical lines cut between sign groups, analogous to spaces or dividers. We observed that the stroke glyph does not appear elsewhere in a normal sequence, only in these specific inscriptions, reinforcing it as a deliberate word boundary marker.

Anomalies and Unaligned Patterns

Our pattern search also flags areas where Byblos does not behave like expected models:

For a presumed syllabary, one might expect a systematic set of signs covering all combinations of consonants and vowels. However, a pattern emerges that one vowel series may be underrepresented or missing. Best's analysis noted that signs corresponding to an "*-o" vowel sound were absent (no lo, mo, etc.). In our inventory, we indeed found that some consonants seem to pair with only 3 vowels instead of 4.
Another non-aligned pattern is the presence of apparent determinatives (logographic signs) in a syllabary. Pure syllabaries typically do not include logograms, yet in Byblos, a few signs behave as though they carry meaning rather than sound. For example, one rare sign consistently appears alongside references that have been interpreted as prayers or speech acts; Malachi Martin posited this sign might mean "to speak/pray" and is used to mark such verbs.

Comparative Correlation with Known Scripts

Even without reading the Byblos texts, we can compare its structural features to contemporary writing systems to guide our understanding:

Phoenician Alphabet

Byblos was in the heart of Phoenician territory, but the script predates the classical Phoenician alphabet. Interestingly, many Byblos signs resemble later Phoenician letters in form. Scholars like Colless note that as many as 18 of the 22 Phoenician letters have look-alikes in the Byblos signary. Our visual catalog confirms numerous familiar shapes – for example, one Byblos glyph is very similar to Phoenician bet (𐤁), another recalls taw (𐤕), etc. This cannot be coincidental; it suggests either Byblos script influenced the development of the alphabet or both share a common ancestor.

Proto-Sinaitic

This early alphabet (c. 18th–15th c. BC) was contemporaneous or slightly later than Byblos script and also shows Egyptian inspiration. Both systems share pictographic origins; some Byblos signs are clearly simplified Egyptian symbols. For example, the Byblos glyph resembling an Egyptian ankh or crook might have been borrowed from hieroglyphs. Proto-Sinaitic used the acrophonic principle (Egyptian pictures given Semitic sounds by their word's first letter). There is evidence Mendenhall applied a similar idea to Byblos – identifying signs by pictorial meaning to guess their sound.

Cypriot Syllabary

The first millennium BC Cypriot script is a known syllabary (mainly CV syllables), providing a useful structural parallel. It uses about 55 signs to cover the required syllables for Greek. Byblos script, with ~90 signs, is larger – possibly due to covering more sounds or including more logographic signs. However, organizationally we see parallels: Cypriot has consistent vowel series for each consonant, and Byblos appears to have something similar (e.g. multiple signs that seem to form a set differing only by an internal modification, potentially indicating vowel changes).

Linear A and Linear B

These Aegean scripts (18th–13th c. BC) are roughly contemporaneous (Linear A undeciphered, Linear B deciphered as Greek). Linear A has ~85 signs, which is very close to Byblos' ~90. It's intriguing that Jan Best approached Byblos by comparing its signs to Linear A, even proposing identical sound values for similar-looking signs. For our purposes, the parallel is that Linear A/B also show many repeated short words and suffixes in their texts (especially Linear B with its inventory lists).

Ugaritic Cuneiform

Ugaritic (c. 14th–12th c. BC) is an alphabetic cuneiform script used for a Northwest Semitic language. While its writing system type (alphabet) differs from Byblos, the language context might be similar if Byblos' language is also Semitic. One comparative angle is syntax and morphology: Ugaritic texts show -m endings for plural masculine nouns ('-m') and use w- as a conjunction ("and") – features common in Semitic. Our identification of a recurring -m suffix in Byblos and a possible w- conjunction aligns well with these Semitic patterns.

Likely Proper Nouns, Determinatives, and Syntax Markers

Based on the frequency and pattern analyses above, we can now enumerate specific glyphs that stand out as candidates for special roles in the script – such as marking names, functioning as determinatives (semantic classifiers), or serving as punctuation/syntactic markers. Below is a list of notable glyphs (using our IDs where assigned) and the rationale for their proposed function:

Structural Markers

BY075 "Vertical Stroke" – Word Divider: This is a simple vertical line glyph that appears only on two spatula inscriptions, repeatedly used to separate groups of signs. It is not part of any word's sequence. We identify it as a word divider or phrase break, equivalent to a space or punctuation. Confidence: High (its usage is clearly as a divider, not phonetic).
BY076 "Curved Bracket" – Clause Separator: A curved, parentheses-like sign noted on the tablets, often preceding a recurring short word. It appears singly (and occasionally doubled as a larger pause). We concur with prior observations that it likely marks a minor pause or list separation ("comma"). Confidence: Medium-High (pattern of occurrence strongly suggests a syntactic role).
BY077 "Circle" – Terminator: A small circle glyph found at what is believed to be the end of an inscription (tablet or monument). It does not recur elsewhere in text lines. This sign is a good candidate for a full stop or period, indicating the end of a text or section. Confidence: Medium (if it reliably sits at final positions only).

Numerical Signs

BY080 "Double Diagonal Tally" – Numerical Mark: This refers to a chevron or stroke-like sign that is repeated in a sequence (e.g. "|||||||"). On the back of one bronze plate, seven identical marks of this type were written in a row. Such usage strongly implies a number (seven, in that case). The fact that the marks were grouped (with a slight grouping 4+3 noted) suggests a tallying system. We interpret the single stroke glyph as a unit numeral "1", which when written multiply conveys numbers (similar to Egyptian tallies). The cluster of seven likely denoted the number of something (perhaps year 7, or seven items). Confidence: High for being numeric (the repetitive pattern has no linguistic reading, fitting numeric use).

Determinatives & Logograms

BY081 "Walking Man" – Determinative for Action/Speech: One infrequent glyph in the corpus has a pictographic shape resembling a person in motion or a gesture (exact shape per Dunand's drawings). Malachi Martin identified a glyph meaning "to pray, speak" used seemingly to introduce utterances or ritual statements. In our data, this glyph appears only alongside what might be invocatory phrases. We tentatively classify it as a determinative indicating speech or action, not read aloud but giving semantic context (much as an Egyptian glyph of a seated man with hand to mouth means "speech" determinative). Confidence: Low-Medium – this is speculative, based on very few occurrences, but plausible given the context and cross-script parallels.
BY082 "Star/Divine Marker" – Determinative for Deity/Name: Another rare sign highlighted by Martin resembles a star or astral symbol. It was posited to mean "deity, Lord (of)", perhaps marking divine names or royal epithets. Our examination found one glyph (used maybe twice) appearing before certain names which could be those of gods or important persons (e.g. one spatula text where this sign precedes a name that might correspond to the sun-god Šuraya, as later suggested by Best). We suspect this functions like a divine name determinative – a sign to flag the following word as a god's name or honorific. Confidence: Medium – the usage pattern (always right before specific names) supports the idea, but occurrences are scant.
BY083 "Compound YA Sign" – Proper Noun Marker (Possible Suffix): We observed that many personal names in the tablets end with a particular sign or sequence "-ya" (two signs, one of which could be a ya-syllable). One glyph within this sequence, BY083, appears almost exclusively at the ends of names. For instance, among the list of overseers, several names share the ending glyph. This could indicate a grammatical suffix used in names (e.g. a case ending or a title like "-servant" etc.), or it might simply be a very common syllable in names (like -ya which is a theophoric element in Semitic names, as in "..ya" = "[God]-is"). Confidence: Medium (clear pattern, but could be phonetic or morphemic; further analysis needed).

Unique/Rare Glyphs

BY084 (Unassigned Specific) – Unique Name Glyphs: There are numerous glyphs that occur only in one inscription—often in the middle of what appears to be a name or title string. Each of these hapax signs is likely part of a proper noun (personal name, place name, or unique title). For example, on tablet c, one name contains a glyph that appears nowhere else in any text. We can infer that BY084 (placeholder for that unique sign) is probably a syllable used in that person's name (and perhaps that person alone). Such glyphs might include uncommon sound values (foreign phonemes or rare consonants) or could even be logograms for a name. We will keep these under "proper noun likely" category. Confidence: Low (by nature, a one-off occurrence is hard to classify, but statistically the safest bet is it's part of a name).
BY085 "Sun Disk" – Possible Logogram: One glyph resembling a sun or circle with rays appears a couple of times in the context of offerings to a deity (according to Best's partial reading, the sun-god Šuraya is named and this sign might depict the sun). If the sign indeed depicts the concept (sun) and stands for the deity, it might be used logographically (i.e. not for its phonetic value but as a symbol for the god). This is speculative, but we flag it as a candidate for a logographic determinative (the way Egyptian or Mesopotamian scripts use a specific symbol to mean "god" or a particular god without spelling it out). Confidence: Low (requires more corroboration; could also simply be a stylized syllable).

Functional Words

BY003 "wa" (?) – Conjunction: This glyph was listed earlier as a high-frequency syllabic sign. However, its frequent occurrence as a standalone word or prefix (often following punctuation and preceding another word) strongly suggests it represents a common word "and". In Northwest Semitic languages, "and" is a single letter w- (pronounced "wa/wi/we"). If BY003 is indeed the wa syllable, its high frequency and syntax position make sense. While as a glyph it is just another syllable, we highlight it here because its role as a conjunction gives it a quasi-syntactic significance — it's a glue word indicating lists or compound phrases. Our confidence in BY003 being "and" (or a similar clitic) is Medium-High given its repetitive usage pattern at clause boundaries.

Summary of Special Glyph Roles: In the Byblos script, a subset of signs likely transcends simple phonetic duty:

Some (like BY075, BY076, BY077) serve pure structural roles (spaces, punctuation marks).
A few (BY081, BY082, BY085) act as semantic determinatives or logograms, giving context (prayer, deity, object) without spelling it phonetically.
Several rare signs (BY084 et al.) are tied to unique names, essentially labeling individuals or places.
At least one very common sign (BY003 "wa") might function as a grammatical connector (conjunction), which is part of language syntax rather than vocabulary per se.

Emergent Patterns Summary

Phase 1 has yielded a comprehensive sign catalog and unveiled the internal structure of the Byblos script: a sign inventory of manageable size consistent with a syllabary, a highly skewed frequency profile indicating core versus peripheral signs, repeated suffixes and phrases suggesting grammatical structure, and non-phonetic symbols playing roles akin to determinatives and punctuation. These findings provide a strong foundation for Phase 2, where we will attempt to assign phonetic values and decode the script. Crucially, we will do so guided by the patterns observed – for example, deciphering the likely conjunction "and" or plural endings first – ensuring that any assignments are grounded in the statistical and structural evidence gathered here rather than guesswork. The careful balance of evidence over assumption in Phase 1 puts us on a solid footing to let a real decipherment "emerge" from the script's own characteristics, in line with our evidence-based methodology.

Phase Classification: Glyph Inventory & Pattern Analysis
Status: Phase 1 Complete - Foundation Established
Next: Phase 2 - Cluster Patterns & Cross-Script Expansion

Research Index Overview Next Phase Decipherment Drops