It is clear that the Punic personal name (BAAL-HANNONI) corresponds to the LA personal name (B-HNH), and the Punic parental name (IEPDAT) corresponds to the LA parental name (YFDT). The translation of the Punic word (BAAL) is (B) in LA; both words might mean “the owner of.” The second parts of the personal names are (HANNONI) in Punic and (HANNO) in LA according to Chabot’s vocalization (1941, p. 12). In this personal name, the Punic sound (H) corresponds to the vertical tribar, that is, the vertical tribar represents the voiceless laryngeal /h/. The presence of the Punic letter (He) in RIL31 alongside other letters such as (Aleph, Nun, and Taw) suggests that this inscription belongs to the Later Punic script. This version differs from the one used in Dougga inscriptions (RIL1 and RIL2).

I assume that the phonological adaptation of foreign names in LA or in Punic is the reason behind the Punic transcription of the vertical tribar as /h/. Therefore, the Punic letter (H) might be a phonological adaptation of the sound (z1) in LA. I believe that this argument is an appropriate explanation to the diverse Punic transcriptions given to the vertical tribar as well. Why does the Punic (H) correspond to the LA (Z1)?

In fact, one of the phonological shifts that is observed in Pan-Amazigh is the change of the sound /z/ to /h/, /ʒ/, or /ʃ/ in Touareg, or the other way around (cf. Chaker, 2015, p. 9). Chaker (2015) states that the palatalization of /z/ would be almost a spontaneous palatalization because in the majority of cases there is no detectable contextual conditioning (azal, “day” > Touareg: ahǝl, ažǝl, ašǝl…) (Chaker, 2015, p. 9). The Zenaga dialect, which belongs to a different zone, also encounters this phonological shift. In Zenaga, the sounds /ʒ/ and /ʃ/ change to /z/ and /s/ in the plural form, for example:

afouš “hand”      ofessan “hands”

amouj “will”       mouzzen “wills”

However, the sound /h/ is absent in the common Amazigh phonological system (Chaker, 2015). Therefore, I am skeptical about the use of this sound in the LA inscriptions. I argue that the              (z1 ~ h) correspondence between LA and Punic in RIL31 is due to phonological adaptation of foreign names.

RIL85 is a Libyco-Latin (LL) biscript where some names correlate with each other, L: [F]AVSTVS and LA:  FWST(H). The personal name FAUSTUS is definitely of Latin origin because it is familiar in Latin literature, and it means “lucky, fortunate”. The Latin suffix (TVS) is transcribed as (TH) in the LA personal name FWST(H); the transliteration of the Latin suffix[6] in LA is also a good reason for its borrowing from Latin (cf. RIL252: CHINID-IAL/ KND-YL). Thus, the Latin letter (S) is transcribed as a vertical tribar in LA. Again, the vertical tribar is transcribed as a sibilant, likely pronounced in LA as /s1/ or /z1/.

In comparison, the equivalent of the Latin name SACTUT is the LA name ZKTT in RIL151; the LA sign − /z/ corresponds to the Latin letter S /s/ in this inscription. This sound change is another instance where phonological adaptation occurs in foreign names.  The table below demonstrates that phonological adaptations might be confusing; 

Latin                  LA
FAUSTUS          FWSTZ1
SACTUT            ZKTT

Analogically, I argue that the vertical tribar is transcribed as /z1/ in RIL85 because of the phonological adaptation which occurred to the foreign name FAUSTUS / FWSTZ1. Therefore, the Latin sound /s/ might be realized as /z1/ in LA.

Fournet (2014) argues that most Amazigh dialects do not have laryngeal, pharyngeal, and uvular phonemes in their core, authentically Amazigh vocabulary. He states that in Nigerian Touareg dialect, where the lexical influence of Arabic is weak, there are no laryngeal or pharyngeal phonemes in the phonological inventory. He adds that it should be also noted that the Ancient Tifinagh alphabets and the LP alphabet, used some two thousand years ago to denote what was certainly an Amazigh dialect, do not have the equivalents of the glottal stop /ʔ/, the voiceless pharyngeal /ḥ/, the voiced pharyngeal /ʕ/ (Fournet, 2014, p. 33). Perhaps, he does not mention the voiceless glottal /h/ because it could be represented by the tribars.

In Tashelhit, the voiced laryngeal sound /h/ is produced with vocal fold vibration, and the geminate counterpart of this glottal fricative is not frequent in the lexicon (Ridouane, 2014, p. 210). This is a fact in the majority of Amazigh dialects, except in Touareg where this sound is frequent, bearing in mind that it corresponds with the sibilants /z/, /ʃ/, /ʒ/ in the rest of the dialects.

Certainly, the vertical tribar might be either the sibilant /z1/ or the laryngeal sound /h/ in LA inscriptions. This assumption includes one problematic issue that is the presence of three variants of the sound /z/ in RIL inscriptions, (Z1, Z2, and Z3).  Kossmann (2020, p. 877) makes a comparison between the phonetic values given to sibilants, namely the variants of (Z), in RIL inscriptions in the following table;

Table 10: The Variants of the Sound Z in RIL LP Inscriptions

Certainly, the vertical tribar is a variant of the voiced sibilant /z/. Therefore, this sibilant might be the sound /z/, /ẓ/ or /ʒ/. Yet, a conditioned distribution pattern might explain the use of different variants for (Z) if it existed.

In the following table, Kossmann (2020, p. 882) observes that the variants of (Z) are never used in the final position of words in RIL. Therefore, I assume that the vertical tribar might be a variant of (Z) that is mostly used in the final position. 

Table 11: Distribution of the Variants of Z in RIL

In fact, all these variants of (Z) are related to the laryngeal sound /h/ in Pan-Amazigh. This explains why it is transcribed as /h/ in RIL31. The following table provides some example of the well-attested /z/ ~ /h/ or /ʒ/ alternation in Pan-Amazigh (Haddadou, 2007).

Table 12: The /z/ ~ /h/ or /ʒ/ Alternation of Sibilants in Pan-Amazigh

In Amazigh, the prefix (S-) indicating causality changes to /z/ when it coexists with the sounds /z/, /ḍ/, and /ẓ/. This phonological shift is called assimilation because the sound /s/ changes to become similar to the neighboring sound /z/, for example: azel “to run” - zuzel “to make run” (Touareg), azzel “to run” - ssizzel, zzizzel “to make run” (Ouargla), and azzel “to run” - zizzel “to make run” (Tamazight) (Haddadou, 2007, 957 ZL). Moreover, it might also change to (Ž) /ʒ/ or (Š) /ʃ/ when it occurs adjacently with the sound /ʒ/ in Zenaga, for example, eneh “to be inclined forward downward”- zeneh “to tilt forward downward, to cause incline downward” (Touareg) […] enji “to be sold”- jenj, ecenj “to sell” - enji, pl. enjan “sale”  (Zenaga) (Haddadou, 2007, 604 NZ).

Philologically, the parental name (W-ZWŽ-SN) includes three parts: (W-) meaning “son of”, the frequent suffix (-SN) meaning “their” (cf. MSNSN, MKWSN, NBDDSN, etc.), and the main root (Z-WŽ) bearing in mind that (Z-) is a prefix. Based on a lexical study, two roots might be related to (WŽ). Firstly, the derivatives of the root 890 WZ that share a common meaning of “a resilient plant” (thorny broom, wild asparagus, barely beard, etc.), that is, “their resilient person”. Secondly, the derivatives of the root 891 WZ that share the meaning of “tingling, shivering and numbing”, which could be a positive meaning in a personal name as well.

The following table sums up the findings observed and analyzed in this section; this analysis is based on the study of three bilingual inscriptions; two of them associate the vertical tribar with sibilants /z/ and /s/ (RIL 145, 85). One of them assigns it the voiced laryngeal fricative /h/ (RIL31).

Table 13: Correspondences among the Vertical Tribar and Sibilants in LA

In conclusion, these transcriptions are not accurate because they are not identical to each other. This diversity is due to the phonological adaptations that LA names encountered in Target languages, namely Latin and Punic (cf. RIL151: (LA) ZKTT- (L) SACTUT). Certainly, the vertical tribar cannot be the voiced sibilant /z/ or the emphatic /ẓ/ because these sounds have their own signs in LA. Yet, it should represent another sibilant, likely /ʒ/ due to its phonetic characteristics.

The Punic (H) corresponds to the vertical tribar in RIL31, which might suggest that the tribar represents /h/. However, phonological evidence from Pan-Amazigh indicates a well-attested /z/ ~ /h/ alternation (or the other way around), particularly in some Touareg dialects (cf. Ridouane 2003). Given that the voiced laryngeal fricative /h/ is rare in native Amazigh vocabulary and often appears in Arabic borrowings, it is possible that the Punic transcription reflects a dialectal or phonological shift rather than a distinct phoneme. Therefore, the correspondence between Punic (H) and the vertical tribar does not necessarily contradict the hypothesis that the tribar originally represents a sibilant, likely /ʒ/. In addition, the potential root of the frequent name (W-ZWH-SN) is ZWH; the latter suggests that the prefix Z- might be the assimilated prefix S- indicating that the sound (H) is a sibilant as well, likely /ʒ/.

The distinction between the vertical and horizontal tribars is based on data obtained from multiple biscripts. I observe that the horizontal tribar consistently appears in final positions and  is absent in the roots of the potential words associated with this sign. Therefore, I assume that it may function as a final vowel. In contrast, I suggest that the vertical tribar might be a sibilant possibly a variant of /z/, likely /ʒ/. Its phonological correspondence with the laryngeal sound /h/ can be explained by the well-attested /z/ ~ /h/ alternation in Pan-Amazigh, particularly in certain Touareg dialects.

3-    Discussion and Limitations

     We now understand why some researchers argue that the tribars might represent a laryngeal sound, while others suggest they might indicate vowel sounds (see Table 4). These conflicting phonetic values arise because scholars do not systematically distinguish between vertical and horizontal tribars. Moreover, those who propose a velar sound seem to be influenced by the modern phonetic value assigned to the tri-dot sign ⵗ in Tifinagh (cf. Sudlow, 2011, p. 33). In this section, I will tackle other important issues that support the distinctiveness of the tribar hypothesis, mainly the morphological and epigraphic features of the tribars and some of the limitations of those features.

a-     Feminine Prefix (T-) as Evidence for the Verbal Character of the Tribars

The occurrence of certain Segment-Hs suggests these segments might have a specific grammatical or semantic function, potentially related to verbs or predicates. This hypothesis could be consolidated by the presence of the feminine marker (T) in some Segment-Hs, which provides strong evidence for their verbal character, indicating gender agreement with the deceased’s name. The verbal character of at least some Segment-Hs is indicated by their incorporation of a verbal feminine marker, specifically an initial T-. Evidence for this can be found in at least the following inscriptions: RIL (80, 104, 212, 276, 965, 968, and 969, 1125).

Table 14: Feminine Segment-Hs in RIL Inscriptions 

In these inscriptions, not only segment-Hs, which have the initial feminine marker (T) but also tribars, exhibit initial feminine markers (T). Both segment-Hs and tribars agree in gender with the personal name of the deceased, indicating that these segments function as verbal predicates.

RIL 80 includes the common segment MSW in its feminine form, TMSW-TH; the personal name of the deceased might be BNT[?].  The puzzling BNS is recurrent but cannot be a personal name; Chabot (1941) assumes that it means “his tomb”. Additionally, in RIL 968, the recurrent segment appears to be appended with a feminine marker, (T). The isolated segment TYL, featured in the first line of this inscription and probably pronounced /tɑjlɑ/ might be a feminine personal name. RIL104, which is a BNS inscription as well, features the feminine personal name TBRN, and the second segment TBKNTH could be a Segment-H containing two verbs TBKN and TH indicating gender agreement.

In RIL 965, the parental name of the deceased begins with WLT, which can only mean “the daughter of” in Pan-Amazigh. The second segment, TDMS[7], starts with (T), indicating a likely agreement marker. RIL 969 also contains a parental name starting with WLT, specifically WLT-MRT, and both the second and third lines, TRN-TH and TBRNG-H, respectively, begin with (T), again suggesting a feminine agreement marker.

In contrast, RIL 211 and RIL 1082 provide evidence for an independent use of WLT, likely to link a personal name with a parental name: ZRTLN and MZKL in RIL 211, and MBTR WLT MGN[I] in RIL 1082. One potential argument against this interpretation of the independent WLT is the existence of at least one RIL inscription, namely RIL 222, where WLT is used independently in the leftmost line, the parental name, prefixed with a masculine (W-), is clearly in the rightmost line, and the non-onomastic segment in the middle does not feature any letter T that might be interpreted as a feminine marker. We acknowledge that this is strong evidence against this analysis. However, there is no principled reason to disregard the possibility that the two versions of WLT are homographs, with one meaning “the daughter of” and the other being a personal name.

Moreover, WLT in (RIL211 and 1082) might be misrepresentations of the recurrent personal name YLT as well (RIL159, 220, 223, 225, 227).  RIL216 is another challenging inscription because the two segments T[G]RN and [TH] feature initial markers T unless the deceased’s name is WMRT, and NMMN is a parental name lacking WLT (cf. HLT: RIL 1110, 1111; WHLT: RIL 161). 

However, if one assumes that these Segment-Hs are feminine nouns rather than verbs, the use of the initial (T-) and the final (-T) indicates that these words might be feminine titles (cf. Chaker, 1986). This assumption suggests that the tribars might be unknown suffixes that are coupled with titles in funerary LA contexts. Further research is needed to explain this important argument as well.

To sum up, some inscriptions show that Segment-Hs might be verbs depending on their morphological features, namely the use of the initial feminine marker (T). Interestingly, some tribars also exhibit this morphological feature, that is, they are independent of Segment-Hs. Hence, Segment-Hs might be verbal collocations with certain funerary meanings and functions. The presence of the word WLT meaning “the daughter of” in some inscriptions supports this argument as well. Yet, this argument could be valid only if feminine Segment-Hs are not titles.

b-    Interchangeability vs. Distinctiveness

The apparent interchangeability of the vertical and horizontal tribars, especially in Segment-Hs, presents challenges in assigning them distinct phonetic values. If we assume that the horizontal tribar represents an unknown suffix or a final vowel while the vertical tribar corresponds to a variant of (Z), likely (Ž), inconsistencies arise in the interpretation of these segments.

The following table demonstrates the potential interchangeable use of the horizontal and vertical tribars in Segment-Hs. For instance, the recurrent Segment-Hs MSKRH and MSWH exhibit both tribars interchangeably; I use the letter (I) to represent the horizontal tribar only for a descriptive reason.

Table 15: MSKRH and MSWH’s Occurrences in RIL

Apparent interchangeability explains why many scholars regard both tribars as variants of the same phoneme rather than distinct sounds. Yet, the distribution of the tribars in Segment-Hs also demonstrates that these tribars are not part of the roots of the Segment-Hs because these segments sometimes occur without any tribars, for example MSKR and MSW.  Additional evidence supporting the apparent interchangeability of the tribars includes:

-       Both tribars coexisting with different Segment-Hs; for example both of them are found in MSKR and MSW (i.e. RIL 144, 158, 185, 182, 324, 326, 724).
-       The same tribar (Ž) or (I) occurring in Segment-Hs and names (i.e. RIL 72, 105, 325).
-       Segment-Hs and names being marked by different tribars (Ž) or (I) (i.e. RIL 85).
-       Both Segment-Hs and personal names being marked with the same tribars (i.e. RIL 146, 325).

Apparent interchangeability, therefore, is a counterargument against the previous findings analyzed in earlier sections, namely; the horizontal tribar might be an unknown suffix or a final vowel, and the vertical tribar might be a variant of (Z), likely (Ž).

c-     Tribar Clusters as Evidence for the Tribars’ Distinctiveness

However, this apparent interchangeability might be conditioned by certain comparative linguistic and epigraphic patterns. On the first hand, the vertical and horizontal tribars co-occur in different lines of the same inscription, which suggests that they might have different phonetic values. Yet, I acknowledge that some graphemes like ⵎ (open to the right), its variant ⵎ (open to the left) are employed interchangeably to represent the sound /d/ in RIL inscriptions. For example, the variants of this grapheme are found in MNKD as follows;

Table 16: The Variants of the Grapheme D in MNKD

This observation suggests that co-occurrence of variants of the same graphemes is well attested in RIL inscriptions, and it might not be enough to distinguish between different graphemes depending on co-occurrence technique. Analogically, I argue that co-occurrence is not a good argument for the tribar distinctiveness hypothesis.

On the other hand, both tribars cluster if they co-exist one next to the other at the end of Segment-Hs or names; particularly, the horizontal tribar tails the vertical one. The table below exhibits some examples of clustered tribars; 

Table 17: Clustered Tribars in RIL Inscriptions

There are at least 17 examples of clustered tribars in RIL inscriptions. These clustered tribars may be categorized into four types: Vertical + Horizontal (VH), Horizontal + Vertical (HV), Vertical + Vertical (VV), and Horizontal + Horizontal (HH). While the common cluster (VH) is found in 13 RIL inscriptions, the other clusters (HV, HH, and VV) are employed only once in RIL inscriptions. The prevalence of the cluster (VH) suggests that the distribution of the tribars might be conditioned by a certain linguistic pattern or writing convention because the systematicity of the (VH) cluster cannot be a random occurrence.

Therefore, I assume that these clusters, mainly (VH and HV), could be good arguments for the distinctiveness tribar hypothesis because employing two different clustered graphemes might be a good reason for their distinct phonetic values. Previously, I have argued that the initial feminine marker (T-) is used with some tribars[8] as a morphological prefix featuring the verbal character of the tribars. Comparatively, the common cluster (VH) also suggests that the clustered tribars might be different verbs, that is, two clauses.

To sum up, the tribars are different from the variants of the grapheme (D) because they     are clustered one next to the other, namely the prevalent cluster (VH). Therefore, the tribars might be different graphemes used in hundreds of RIL inscriptions.

Conclusion

In conclusion, the horizontal tribar in RIL inscriptions, lacking Punic or Latin transliterations, consistently appears in final positions, both in non-onomastic (RIL1, RIL2, RIL451) and onomastic contexts (RIL72). Philological analysis of the non-onomastic words indicates it likely represents either an unknown suffix or a final vowel. Conversely, the vertical tribar, present in both LL and LP bilingual inscriptions, might be a variant of /z/, likely /ʒ/. Its correspondence with Punic (H) and Latin (S) might be attributed to the phonological adaptation of foreign names and explained by a common phonological shift in Pan-Amazigh, namely the /z/ ~ /h/ alternation.

The distinction between these tribars is further supported by their clustered occurrences, particularly the prevalent VH, and the presence of the initial feminine marker T-, indicating potential verbal functions. Despite phonetic inconsistencies due to apparent interchangeability of the tribars, the weight of evidence, including biscript and Pan-Amazigh analysis and distribution patterns, supports the distinctiveness of these tribars.

The tribars used in Segment-Hs, therefore, are likely recurrent verbs or suffixes within the RIL inscriptions. The analysis of these tribars sheds light on the morphology and syntax of the Libyco-Amazigh language, providing valuable insights into the funerary practices and LA writing conventions of the Amazigh people.

References

Boukous, A. 2009. Phonologie de l'Amazighe. Rabat: Publications of IRCAM.

Camps, G. 1993. “A la recherche des Misiciri: cartographie et inscriptions libyques.” à la croisée des études libyco-berbères : mélanges offerts à Paulette Galand-Pernet et Lionel Galand. Drouin J., Roth A. ed. Geuthner: 113-126

Chabot, J. B. 1941. Recueil des inscriptions libyques. Paris: Imprimerie nationale.

Chaker, S. 1986. “A propos de la terminologie libyque des titres et fonctions.” Annali del’Istituto Universitario Orientale: 541-562.

____________. 1996 a. “Manuel de linguistique berbère II, syntaxe et diachronie.” Algeria: ENAG.

____________. 2015. “Phonologie et Phonétique.” Encyclopédie Berbère 33. doi: hal-01780808

Cid Kaoui, S. 1907. Dictionnaire français-Tachelh't et Tamazir't (dialectes berbères du Maroc). Paris: Ernest Leroux.

Dallet, J. M. 1982. Dictionnaire Kabyle français, parler des Al Mangellat, Algérie. Paris : Société d'études linguistiques et anthropologiques de France.

Faidherbe, L. G. 1877. Le zenaga des tribus sénégalaises. Paris: Ernest Leroux.

Février, J. 1956. “Que savons-nous du libyque ?” Revue africaine : 263-273.

Fournet, A. 2014. “Les consonnes du chamito-sémitique et le proto-phonème *H du berbère.” P. d. Proche-Orient, ed. Bulletin d’études orientales : 23-50. doi:10.4000/beo.1257

Galand, L. 1973. “L'alphabet libyque de Dougga.” L'Occident musulman et de la Méditerranée 13-14 : 361-368.

Haddadou, A. M. 2007. Dictionnaire des Racines Berbères Communes. Algeria: Haut Commissariat à l’Amazighité.

Justinard, C. 1914. Manuel berbère marocain (dialecte Chleuh). E. Guilmoto, ed. Librairie orientale et Américaine.

Kossmann, M. G. 2020. “Sibilants in Libyco-Berber.” The American Oriental Society 140 : 875-888. doi: 10.7817/jameroriesoci.140.4.0875

Krahmalkov, C. R. 2001. “A Phoenician Punic Grammar.” Leiden: Koninklijke Brill NV.

Motylinski, A. d. 1898. Le djebel nefousa. Paris: Ernest Leroux.

Pichler, W. 2007. Origin and Development of the Libyco-Berber Script. Cologne: Rudiger Koppe Verlag.

Ridouane, R. 2014. “Tashlhiyt Berber.” Journal of the International Phonetic Association. doi:10.1017/S0025100313000388

Rössler, O. 1958. “Die Sprache Numidiens.” Sybaris. Festschrift H. Krahe : 94-120. Wiesbaden.

Sfaxi, I. 2014. “L'onomastique libyque : son intérêt - état des recherches.” Revue des Etudes Berbères (REB) (9) : 565-575.

____________. 2022. “La Valeur Sémantiques du Libyque YSK et ses Variantes.” Études et Documents Berbères: 105-112.

Sudlow, D. 2011. “The Tamasheq of North-East Burkina Faso. 2nd ed.” Cologne: Rudiger Koppe Verlag.

Vaan, M. d. 2008. Etymological Dictionary of Latin and the Other Italic Languages (Vol. 7). A. Lubotsky, ed. Leiden: Brill.

[1] Recueil des inscriptions libyques (RIL).

[2] Some researchers argue that some Segment-Hs could be ethnonyms, that is, tribal names (Camps, 1993; Chabot, 1941; Sfaxi, 2014). Yet the ethnonym interpretations do not provide any explanations to the significance of the tribars (H). However, my concern in this article is not to explain the meaning of Segment-Hs but to determine the phonetic value of the tribars in RIL.

[3] Mater lectionis is a Latin word meaning a consonant letter used to indicate a vowel sound.

[4] See (H) in Appendice, table des noms propres. I could not find these inscriptions (RIL737, 798, 807) in the available online version of RIL (Chabot, 1941).

[5] Author’s translation from French.

[6] The Latin suffix (US) is kept in one Latin loanword in Tashelhit and Moroccan darija; (afulus) or (fellus) meaning “chick, chicken” comes from (pullus) “foal, chick, young of an animal” (cf. Vaan, 2008, p. 502).

[7] Chabot (1941) comments that this segment could be either DMS or RMS (see RIL965).

[8] See (RIL 80, 104, 965, 968, 969, 216).