Dotawo/content/article/starostin.md

---
title: "Restoring “Nile-Nubian”: How To Balance Lexicostatistics and Etymology in Historical Research on Nubian Languages"
author: "George Starostin"
abstract: ""
keywords: "comparative linguistics, Nilo-Saharan, glottochronology, lexicostatistics, Nubian"
---

# Introduction

Although there has never been any serious disagreement on which languages constitute the Nubian family, its internal classification has been continuously refined and revised, due to such factors as the overall complexity of the processes of linguistic divergence and convergence in the "Sudanic" area of Africa; constant influx of new data that forces scholars to reevaluate former assumptions; and lack of scholarly agreement on what types of data provide the best arguments for language classification.

Traditionally, four main units have been recognized within Nubian[^1]:

* Nile-Nubian, consisting of the closely related Kenuzi-Dongolawi (Mattokki–Andaandi) dialect cluster and the somewhat more distant Nobiin (= Fadidja-Mahas) cluster;
* Kordofan Nubian, or Hill Nubian, consisting of numerous (and generally poorly studied, although the situation has significantly improved in the past decade) languages such as Dilling, Karko, Wali, Kadaru, etc.;
* Birgid (Birked, Birged), now-extinct , formerly spoken in Darfur;
* Midob (Meidob), also in Darfur.

This is, for instance, the default classification model adopted in Joseph Greenbergʼs general classification of the languages of Africa,[^2] and for a long time it was accepted in almost every piece of research on the history of Nubian languages.

[^1]: Bechhaus-Gerst, “Nile-Nubian Reconsidered,” p. 85.
[^2]: Greenberg, *The Languages of Africa,* p. 84.

More recently, however, an important and challenging hypothesis on a re-classification of Nubian has been advanced by Marianne Bechhaus-Gerst.[^3]  Having conducted a detailed lexicostatistical study of a representative batch of Nubian lects, she made the important observation that, while the percentage of common matches between the two main components of Nile-Nubian is indeed very high (70%), Kenuzi-Dongolawi consistently shows a much higher percentage in common with the other three branches of Nubian than Nobiin (**table 1**).

[^3]: Bechhaus-Gerst, “Nile-Nubian Reconsidered”; Bechhaus-Gerst, *Sprachwandel durch Sprachkontakt am Beispiel des Nubischen im Niltal*; Bechhaus-Gerst, *The (Hi)story of Nobiin*.


|            | Meidob | Birgid | Kadaru | Debri | Dilling | K/D |
| ---        | ---    | ---    | ---    | ---   | ---     | --- |
| **K/D**    | 54%     | 48%     | 58%     | 57%    | 58%      | |   
| **Nobiin** | 40%     | 37%     | 43%     | 41%    | 43%      | 70% |

**Table 1. Part of the lexicostatistical matrix for Nubian[^t1]**

[^t1]: Bechhaus-Gerst, *Sprachwandel durch Sprachkontakt am Beispiel des Nubischen im Niltal,* p. 88.

In Bechhaus-Gerstʼs view, such a discrepancy could only be interpreted as evidence of Kenuzi-Dongolawi and Nobiin not sharing an intermediate common "Nile-Nubian" ancestor (if they did share one, its modern descendants should be expected to have more or less the same percentages of matches with the other Nubian subgroups). Instead, she proposed independent lines of development for the two dialect clusters, positioning Nobiin as not just a separate branch of Nubian, but actually the earliest segregating branch of Nubian. Consequently, in her standard historical scenario described at length in two monographs, there was not one, but two separate migrations into the Nile Valley from the original Nubian homeland (somewhere in South Kordofan/Darfur) — one approximately around 1,500 BCE (the ancestors of modern Nobiin-speaking people), and one around the beginning of the Common Era (speakers of Kenuzi-Dongolawi). As for the multiple exclusive similarities between Nobiin and Kenuzi-Dongolawi, these were explained away as results of "intensive language contact."[^4] The lexicostatistical evidence was further supported by the analysis of certain phonetic and grammatical peculiarities of Nobiin that separate it from Kenuzi-Dongolawi; however, as of today it is the lexical specificity of Nobiin that remains at the core of the argument.

[^4]:  Bechhaus-Gerst, *The (Hi)story of Nobiin,* p. 22.

Bechhaus-Gerstʼs classificatory model, with its important implications not only for the history of Nubian peoples, but also for the theoretical and methodological development of historical and areal linguistics in general, remains somewhat controversial. While it has been embraced in the recent editions of such influential online language catalogs as [Ethnologue](https://www.ethnologue.com/subgroups/nubian) and [Glottolog](https://glottolog.org/resource/languoid/id/nubi1251) and is often quoted as an important example of convergent linguistic processes in Africa,[^5] specialists in the field often remain undecided,[^6] and it is concluded in the most recent handbook on African linguistics that "the internal classification of Nubian remains unclear."[^7] One of the most vocal opponents of the new model is Claude Rilly, whose research on the reconstruction of Proto-Nubian (in conjunction with his work on the historical relations and genetic affiliation of Meroitic) and investigation into Bechhaus-Gerstʼs evidence has led him to an even stronger endorsement of the Nile-Nubian hypothesis than ever before.[^8]

[^5]:  E.g., Heine & Kuteva, "Convergence and Divergence in the Development of African Languages."
[^6]:  E.g., Jakobi, "The Loss of Syllable-final Proto-Nu­bian Consonants."
[^7]:  Güldemann, "Historical Linguistics and Genealogical Language Classification in Africa," p. 283.
[^8]:  Rilly, *Le méroïtique et sa famille linguistique,* pp. 211–288; Rilly, "Language and Ethnicity in Ancient Sudan," pp. 1180–1183. We will return to Rillyʼs arguments in the final section of this paper.

While in theory there is nothing impossible about the historical scenario suggested by Bechhaus-Gerst, in practice the idea that language A, rather distantly related to language B, could undergo a serious convergent development over an approximately 1,000-year long period (from the supposed migration of Kenuzi–Dongolawi into the Nile Valley and up to the attestation of the first texts in Old Nubian, which already share most of the important features of modern Nobiin), to the point where language A can easily be misclassified even by specialists as belonging to the same group as language B, seems rather far-fetched. At the very least, it would seem to make perfect sense, before adopting it wholeheartedly, to look for alternate solutions that might yield a more satisfactory explanation to the odd deviations found in the data.

Let us look again more closely (table 2) at the lexicostatistical evidence, reducing it, for the sake of simple clarity, to percentages of matches observed in a "triangle" consisting of Kenuzi–Dongolawi, Nobiin, and one other Nubian language that is universally recognized as belonging to a very distinct and specific subbranch of the family — Midob. Comparative data are given from the older study by Bechhaus-Gerst and ,my own, more recent examination of the basic lexicon evidence.[^9]

[^9]:  Starostin, *Jazyki Afriki,* pp. 24–95.

|            | Nobiin | Midob |
| ---        | ---    | --- |
| **K/D**    | 70%    | 54% |
| **Nobiin** |        | 40% |

**Table 2a. Lexicostatistical relations between Nile-Nubian and Midob \(Bechhaus-Gerst\)[^t2a]**

|            | Nobiin | Midob |
| ---        | ---    | --- |
| **K/D**    | 66%    | 57% |
| **Nobiin** |        | 51% |

**Table 2b. Lexicostatistical relations between Nile-Nubian and Midob \(Starostin\)[^t2b]**

[^t2a]: Bechhaus-Gerst, “Nile-Nubian Reconsidered”
[^t2b]: Storostin, *Jazyki Afriki*.

The significant differences in figures between two instances of lexicostatistical calculations are explained by a number of factors (slightly divergent Swadesh-type lists; different etymologizations of several items on the list; exclusion of transparent recent loans from Arabic in Starostinʼs model). Nevertheless, the obvious problem does not go away in the second model: Midob clearly shares a significantly larger number of cognates with K/D than with Nobiin — a fact that directly contradicts the K/D–Nobiin proximity on the Nubian phylogenetic tree. The situation remains the same if we substitute Midob with any other non-Nile-Nubian language, such as Birgid or any of the multiple Hill Nubian idioms.

The important thing is that there are actually two possible	reasons for this discrepancy in the lexicostatistical matrix. One, endorsed by Bechhaus-Gerst, is that the K/D–Nobiin number is incorrectly increased by the addition of a large number of items that have not been inherited from a common ancestor, but actually borrowed from Nobiin into K/D. An alternate scenario, however, is that the active recipient was Nobiin, except that the donor was not K/D — rather, a certain percentage of Nobiin basic lexicon could have been borrowed from a third, possibly unidentified source, over a relatively short period of time, which resulted in lowering the percentage of Nobiin matches with *all* other Nubian languages.

Thus, for instance, if we assume (or, better still, somehow manage to prove) that Nobiin borrowed 6% of the Swadesh wordlist (i.e., 6 words on the 100-item list) from this third source, exclusion of these words from lexicostatistical calculation would generally normalize the matrix, increasing the overall percentage for the K/D–Nobiin and Nobiin–Midob pairs, but not for the K/D–Midob pair.

The tricky part in investigating this situation is determining the status of those Nobiin words on the Swadesh list that it does not share with K/D. If the phylogenetic structure of the entire Nubian group is such that Nobiin represents the very first branch to be split off from the main body of the tree, as in Bechhaus-Gerstʼs model (**table 3**), then we would expect a certain portion of the Swadesh wordlist in Nobiin to be represented by the following two groups of words:

* archaic Nobiin retentions that have been preserved in their original meaning in that subgroup only, replaced by innovations in the intermediate common ancestor of Midob, Birgid, K/D, and Hill Nubian;
* conversely, more recent Nobiin innovations that took place after the original separation of Nobiin; in this case, the Nobiin equivalent of the Swadesh meaning would also be opposed to the form reconstructible for the common ancestor of the remaining four branches, but would not reflect the original Proto-Nubian situation.

INSERT TABLE

**Table 3. The revised classification of Nubian according to Bechhaus-Gerst**

Indeed, we have a large share of Nobiin basic words that set it apart from every other Nubian languages (see the more than 30 items in section III of the list below), but how can we distinguish retentions from innovations? If the word in question has no etymological cognates in any other Nubian language, then in most cases such a distinction is impossible.[^10] However, if the retention or innovation in question was not accompanied by the total elimination of the root morpheme, but rather involved a semantic shift, then investigating the situation from an etymo­logical point of view may shed some significant light on the matter. In general, the more lexico­statistical discrepancies we find between Nobiin and the rest of Nubian where the Nobiin item has a Common Nubian etymology, the better the case for the "early separation of Nobiin" hypothesis; the more "strange" words we find in Nobiin whose etymological parallels in the other Nubian languages are highly questionable or non-existent, the stronger the case for the "pre-Nobiin substrate" hypothesis.

[^10]:  One possible argument in this case would be to rely on data from external comparison. Thus, if we agree that Nubian belongs to the Northern branch of the Eastern Sudanic family, with the Nara language and the Taman group as its closest relatives (Rilly, *Le méroïtique et sa famille linguistique*; Starostin, *Jazyki Afriki*), then, in those cases where Nobiin data is opposed to the data of all other Nubian languages, it is the word that finds better etymological parallels in Nara and Tama that shouud be logically regarded as the Proto-Nubian equivalent. However, in order to avoid circularity or the additional problems that one runs into while investigating chronologically distant language relationship, I intentionally restrict the subject matter of this paper to internal Nubian data only.

In order to resolve this issue, below I offer a concise and slightly condensed etymological analysis of the entire 100-item Swadesh wordlist for modern Nobiin.[^11] The lexical items are classified into three groups:

[^11]:  Reasons of volume, unfortunately, do not allow to go into sufficient details on many of the more complicated cases. A subset of 50 words, representing the most stable (on average) Swadesh items, has been analyzed in detail and published (in Russian) in Starostin, *Jazyki Afriki,* pp. 224–95. A complete 100-item wordlist reconstructed for Proto-Nubian, with detailed notes on phonetics, semantics, and distribution, is scheduled to be added to the already available annotated 100-item wordlists for ten Nubian languages, published as part of [The Global Lexicostatistical Database](http://starling.rinet.ru/new100).

* 	I. Lexicostatistical matches (i.e., cases where the exact same lexical root is preserved in the exact same Swadesh meaning, without semantic shifts) between Nobiin and K/D. These are further divided into subcategories I.1: common Nubian roots, also found in the same meaning in all or some other branches of Nubian beyond Nile-Nubian; and I.2: exclusive isoglosses between Nobiin and K/D that may be either retentions from Proto-Nubian, lost in all other branches, or Nile-Nubian innovations replacing more archaic words. In either case, these data have no bearing on the issue of Nobiinʼs uniqueness (although isoglosses in I.2 may be used to strengthen the case for Nile-Nubian).
* 	II. Lexicostatistical matches between Nobiin and other Nubian branches (Midob, Birgid, Hill Nubian) that exclude K/D. Upon first sight, such isoglosses might seem to weaken the Nile-Nubian connection, but in reality they are not highly significant, as the K/D equivalents of the respective meanings may simply represent recent lexical innovations that took place already after the split of Nile-Nubian.
* 	III. Nobiin-exclusive lexicostatistical items that have a common Nubian etymology (III.1) or do not have any parallels in any of the other attested Nubian languages (III.2). This is the most significant group of cases, with items in subgroup III.1 testifying in favor of the early separation hypothesis (particularly if the lexicostatistical meaning in Nobiin can be shown to be archaic), and items in subgroup III.2 favoring the substrate explanation. Needless to say, it is the items in this group that will be receiving the most extensive commentary.[^12]

[^12]:  Note on the data sources: for reasons of volume, I do not include all available data in the etymologies. Nobiin (N) forms are quoted based on Werner's *Grammatik des Nobiin*; if the word is missing from Wernerʼs relatively short glossary, additional forms may be drawn upon from either older sources, such as Lepsius's *Nubische Grammatik*, or newer ones, e.g., Khalil's *Wörterbuch der nubischen Sprache* (unfortunately, Khalilʼs dictionary is unusable as a lexicostatistical source due to its unwarranted omission of Arabic borrowings and conflation of various early sources). The ancient forms of Old Nubian (ON) are taken from Gerald Browneʼs *Old Nubian Dictionary.*

  	Data on the other languages are taken from the most comprehensive published dictionaries, vocabularies, and/or wordlists and are quoted as follows: Kenuzi (K) — Hofmann, *Nubisches Wörterverzeichnis*; Dongolawi (D) — Armbruster, *Dongolese Nubian*; Midob (M) — Werner, *Tìdn-áal*; Birgid (B) — Thelwall, "A Birgid Vocabulary List"; Dilling (Dl) — Kauczor, *Die Bergnubische Sprache*. Hill Nubian data other than Dilling are used sparingly, only when it is necessary to specify the distribution of a given item; occasional forms from such languages as Kadaru, Debri, Karko, and Wali are quoted from wordlists published in Thelwall, "Lexicostatistical Relations between Nubian, Daju and Dinka" and Krell, *Rapid Appraisal Sociolinguisyic Survey among Ama, Karko, and Wali Language Groups*.

    Proto-Nubian forms are largely based on the system of correspondences that was originally laid out in Marianne Bechhaus-Gerstʼs reconstruction of Proto-Nubian phonology in "Sprachliche und his­torische Rekonstruktionen im Bereich des Nubischen unter beson­de­rer Berücksichtigung des Nilnubischen," but with a number of emendations introduced in Starostin, *Jazyki Afriki*. Since this study is more concerned with issues of cognate distribution than those of phonological reconstruction and phonetic interpretation, I will refrain from reproducing full tables of phonetic correspondences, but brief notes on peculiarities of reflexes of certain PN phonemes in certain Nubian languages will be given for those cases where etymological cognacy is not obvious or is disputable from the standard viewpoint of the neogrammarian paradigm.

# 100-Item Swadesh List for Nubian: The Data

I. Nobiin/Kenuzi-Dongolawi Isoglosses

I.1. General Nubian Isoglosses
* “ashes”: N *ùbúr-tí*, K/D *ubur-ti* (= M *úfù-dì*, B *ubur-ti*, etc.).
*	“belly”: N *tùː*, K/D *tuː* (= M *tə̀ː*, B *tuː*, etc.).
* “bird”: N *kawar-ti*, K *kawir-te*, D *kawɪr-tɛ* (= M *àːbéd-dí*, B *kwar-ti*, etc.).
*	“bite”: N *àc-*, K/D *acc-* (= M *àcc-*, Dl *aɟ*, etc.).
*	“black”: N *úrúm*, K/D *urumm*- (= M *údí*, B *úːdè*, Dl *uri*, etc.). ◊ The Nile-Nubian form is an original nominal derivate (*\*ur-um* “darkness”) from the adjectival stem *\*ur*- “black.”
* “bone”: N *gìsìr*, K *kiːd*, D *kɪhɪːd* (= M *ə̀ːdí*, B *kìzídì*, etc.). ◊ Voiced *g-* in Nile-Nubian is irregular, possibly as a result of assimilation (← *\*kizir*) or contamination.[^13]
* “breast”: N *óg*, K/D *og* (= M *ə́ː*, B *ogi*, Dl *ɔki*, etc.).
* “claw/nail”: N *sun-ti*, K *sutti*, D *sun-tɪ* (= M *súŋún-dí*, B *sun-di*, etc.).
* “cold”: N *ór-kí*, K *oroːke-l*, D *oroːfɛ-l* (= Wali *ór-kō*, Debri *worr-uŋ*, etc.).
* “die”: N *dí-*, K/D *diː* (= M *tíː-*, B *ti-n-*, Dl *ti*, etc.).
* “drink”: N *ní-*, K/D *niː* (= M *tìː-*, B *ɲiː*, Dl *di*, etc.). ◊ From PN *\*ni-* with regular denasalization in M and Hill Nubian languages.
* “ear”: N *úkkí*, K/D *ulug* (= Dl *ulɟe*, M *úlgí*, etc.). ◊ From PN *\*ulg-i*. The Nobiin form goes back to ON *ul(u)g-* and shows a specific phonetic development (*\*-lg-* → *-kk-*); the latter, however, can in no way be construed as an archaism.
* (?) “eye”: N *máːɲ* (= ON *maɲ-*), D *mɪssɪ*, K *missi* (= M *pì-dì*). ◊ A complicated case. The K/D forms perfectly correspond to M *pì-dì*, going back to PN *\*miC-ti*, where *-C-* is one of several consonants capable of triggering the lenition *\*-t- → -s-* in K/D. If *\*-C- = \*-ɲ-*, then the forms are further comparable with N *máːɲ*, and we are either dealing with a one-time vocalic dissimilation *\*miɲ → \*maɲ* in N or two independent assimilations (*\*maɲ → miɲ-*) in M and K/D, respectively. Alternately, the N form may be completely unrelated to the K/D–M isogloss, in which case the word should be moved to group III.2, since a separate form like *\*maɲ* “eye” would have no Common Nubian etymology of its own.
* “fire”: N *íːg*, K *iːg*, D *ɪːg* (= Dl *ike*, Debri *ika*; probably also B *uzug*, M *ússí*). ◊ The forms in B and M are comparable if the original stem is to be reconstructed as *\*usi-gi*, with regular elimination of intervocalic *\*-s-* in Hill and Nile Nubian. The vocalism is still problematic, but even without the B and M forms, parallels in Hill Nubian clearly show that the Nile-Nubian items represent an inherited archaism.
* “foot”: N *óːy*, K *ossi*, D *oss(ɪ)* (= B *ose*, M *òttì*). ◊ All forms go back to PN *\*oy(-ti)*.
* “give”: N *tèː-r*, K *ti-r*, D *tɪ-r* (= M *tì-*, B *teː-n*, Dl *ti*, etc.).
* “green”: N *déssí*, K *desse ~ dosse*, D *dɛssɛ* (= M *tèssé*, B *teːze*, Dl *teɟe*).
* “hand”: N *èd-dì* (= ON *ey-*), K *iː*, D *ɪː* (= M *ə̀ssì*, B *essi*, Dl *iši*, etc.). ◊ All forms go back to PN *\*əsi ~ \*əsi-ti*.
* “head”: N *ùr*, K/D *ur* (= M *òr*, B *úr*, Dl *or*, etc.).
* “heart”: N *áy* (= ON *ai-l-*), K/D *aː* (= B *ai-di*, Dl *a-l*, etc.).
* “horn”: N *nìːšì*, K *nišši*, D *nɪšši* (= M *kə́ːcí*, B *ŋis-ti*, D *dɔ-ti*). ◊ All forms go back to PN *\*ŋəɟi*.[^horn]
* “I”: N *ày*, K/D *ay* (= M *ə́y*, Dl *ɛ*, etc.).
* “kill”: N *fáːy-èːr*, K *beː*, D *bɛː* (= M *pé-r-*, B *fi-laːle*).
* “knee”: N *kúr-tí*, K *kur-ti*, D *kur-tɪ* (= M *ùrú-d*, B *kur-ti*, etc.).
* “know”: N *ìrbé-èr* (= ON *i- ~ ia-r- ~ ie-r-*), K *iy-ir* (= M *ìːyá-*, D *i-er-*). ◊ The stem in modern Nobiin seems to be an extended form of the original stem, though the nature of the extension is not quite clear.
* “long”: N *nàssí*, K *nosso*, D *noso* (= M *tàssè*, B *nizze*, Dl *dɔɟi*, etc.). ◊ Goes back to PN *\*nossi*, although vocalic correspondences are somewhat irregular.
* “louse”: N *issi*, K *issi*, D *ɪssɪ* (= M *ìːdì*, Dl *iti-d*, etc.).
* “moon”: N *ùn-áttí*, K *un-atti ~ an-atti*, D *un-attɪ* (= Dl *nɔn-ti*, Wali *ūm-tù* etc.). ◊ The Nile-Nubian root is *\*un-*; there are some problems with Hill Nubian forms, such as explaining the initial *n-* in Dl, but overall, there is no reason to doubt the common origin of all these items.
* “neck”: N *íyyí*, K *eyye*, D *ɛyyɛ* (= Kadaru *eː*). ◊ Not clear if M *éːr* “neck” also belongs here (with a suffix?), but the Kadaru form is sufficient by itself to trace the word back to PN *\*eyi*.
* “not”: N *-mùːn*, K/D *-mun-* (= Dl *-min*, B *-m-*, etc.). ◊ A common Nubian negative verbal stem (interestingly, not attested in M, which instead uses the suffixal morpheme *-áː-* for negation, something that could be construed as an archaism and used as a serious argument against early separation of Nobiin).
* “one”: N *wèːr ~ wèːl*, K *weːr*, D *wɛːr* (= M *pàr-*, B *meːl-*, Dl *be*, etc.).
* “person”: N *íd* (= M *ír*, Dl *id*, etc.). ◊ The old Nubian root is largely replaced by Arabisms in K/D (K *zoːl*, D *adɛm*), but the word *ɪd* is still used in D as an archaism or in various idiomatic formations.
* “rain”: N *áwwí*, K *a-nn-essi* (← *\*aru-n-essi* “rain-waterʼ), D *aru* (= M *áré*, B *aːle*, Dl *are*, etc.). ◊ The development *\*-r- → -w-* in N is regular before *\*-i*.[^14]
* “red”: N *géːl*, K *geːle*, D *gɛːlɛ* (= M *kéːlé*, B *keːle*, Dl *kele*, etc.).
* “sand”: N *síw*, K *siːw*, D *sɪu* (= Dl *šu-d*, Debri *šu-du*, etc.).
* “see”: N *nèːl*, K/D *nal* (= M *kə̀l-*, B *ell-*, Dl *gel*, Kadaru *ŋeli*, etc.). ◊ All forms go back to PN *\*ŋali-*.
* “sit”: (a) N *àːg-*, K/D *aːg* (= M *àːg-*, Dl *ak-i*, etc.); (b) N *tìːg-*, K *teːg*, D *tɛːg* (= M *tə́g-*). ◊ Two roots with very close semantics, both easily reconstructible back to PN.
* “sleep”: N *nèːr-*, K *neːr*, D *nɛːr* (= M *kèrà-*, B *neːri*, Dl *ɟer*, etc.). ◊ All forms go back to PN *\*ɲɛːr-*.
* “star”: N *wìnɟì*, K *wissi*, D *wɪssɪ* (= M *òɲè-dì*, B *waːɲ-di*, Kadaru *wonɔ-ntu*, etc.). ◊ There are some problems with the reconstruction, but it is possible that all forms go back to PN *\*wiɲ- ~ \*waɲ-*; at the very least, *\*wiɲ-ti* “star” is definitely reconstructible for Proto-Nile-Nubian.
* “sun”: N *màšà* (= ON *mašal-*), K *masil*, D *masɪl* (= M *pàssàr*). ◊ The isogloss with M confirms PN status, although some phonetic peculiarities (such as the irregular *-š-* in N) as well as the attestation of the term *maša ~ masa* in Meroitic, where it denotes a supreme deity[^sun] indirectly suggest a possible areal isogloss; if so, an alternate candidate for PN “sun” would be *\*eːs-* → B *iːzi*, Dl *eɟ* “sun,” further related to M *èːsì* “heat; midday,” K *eːs* id., D *ɛːs* “afternoon.” In either case, N still aligns with K/D rather than anything else.
* you (sg.): N *ì-r*, K *e-r*, D *ɛ-r* (= M *íː-n*, B *e-di*, Dl *a*, Karko *yā*, etc.). ◊ Although all the forms are related (going back to PN *\*i-*), N is noticeably closer to K/D in terms of morphological structure (with the direct stem marker *\*-r*).
* “tongue”: N *nàr*, K *ned*, D *nɛd* (= M *kàda-ŋì*, B *nat-ti*, Dl *ɟale*, Debri *ɲal-do*, etc.). ◊ All forms go back to PN *\*ɲal(T)*-.[^tongue] Interestingly, the ON equivalent tame- (no parallels in other languages) is completely different — the only case on the list where ON differs not only from N, but from all other Nubian languages as well.
* “tooth”: N *nìːd*, K *nel*, D *nɛl* (= M *kə̀d-dì*, B *ɲil-di*, Dl *ɟili*, etc.). ◊ All forms reflect PN *\*ɲəl-*.
* “two”: N *úwwó*, K *owwi*, D *owwɪ* (= M *ə́d-dí*, B *ul-ug*, Dl *ore*, etc.). ◊ All forms go back to PN *\*awri*; the unusual cluster *\*-wr-* is responsible for the unusual development *\*-r- → *-w-* already in Proto-Nile-Nubian (rather than just in N), and is actually seen explicitly in the extinct and very poorly attested Haraza Nubian: *auri-yah* “two.”[^15]
* “walk (go)”: N *ɟúù-*, K/D *ɟuː* (= M *sə́-r-*, Dl *šu*, etc.). ◊ All forms go back to PN *\*cuː-*.
* “warm (hot)”: N *ɟùg*, K/D *ɟug-ri* (= M *sùːw*). ◊ From PN *\*cug-*.
* “who”: N *nàːy*, K *niː*, D *nɪː* (= M *kə̀ː-rén*, B *neː-ta*, Dl *de*, etc.). ◊ All forms go back to PN *\*ŋə(y)*.

[^13]: Bechhaus-Gerst, "Nile-Nubian Reconsidered," p. 94 lists this as one of two examples illustrating the alleged archaicity of Old Nubian and Nobiin in retaining original PN *\*g-*, together with ON *gouwi* "shield.” However, in both of these cases K/D also show *k-* (cf. K/D *karu* "shield"), which goes against regular correspondences for PN *\*g-* (which should yield K/D *g-*, see "red"), meaning that it is Nobiin and not the other languages that actually have an innovation here.
[^horn]: Reconstruction somewhat uncertain, but initial *\*ŋ-* is fairly clearly indicated by the correspondences; see detailed discussion in Starostin, *Jazyki Afriki,* pp. 56–57.
[^14]:  Bechhaus-Gerst, "Nile-Nubian Reconsidered," p. 93 counts this as an additional slice of evidence for early separation of N, but since this is an innovation rather than an archaism, there are no arguments to assert that the innovation did not take place recently (already after the separation of N from K/D).
[^sun]: Hofmann, *Material für eine Meroitische Gram­ma­tik,* 86.
[^tongue]: See the detailed discussion on this phonetically unusual root in Starostin, *Jazyki Afriki,* p. 80.
[^15]: Bell, "Documentary Evidence on the Haraza Nubian Language," p. 10.

I.2. Exclusive Nile-Nubian Isoglosses

* “all”: N *màlléː*, K *malleː*, D *mallɛ*.
*	“big”: N *dàwwí*, K/D *duː-l*.
*	“burn”: N *ɟùgé-èr*, K/D *ɟug*.
*	“egg”: N *kúmbúː*, D *kumbu*. ◊ Replaced in K by the recent compound innovation *gas-katti* (where the first root probably = *gaːsi* “heavy, hard, rough"), but clearly reconstructible for Nile-Nubian on the whole.
*	“feather”: N *šipir*,[^feather] D *sɪbɪr*.
*	“leaf”: N *úkkí*, K/D *ulug*. ◊ Same word as “ear.”
*	“man”: N *ògɟí-l*, K *ogiɟ*, D *ogɪɟ*.
*	“many”: N *díyyí*, K *dig-riː*, D *díyyí*. ◊ In ON usually attested as *diː-*, once as *dig-* (reflecting dialectal differences between N and K/D).
*	“nose”: N *sòrìŋ*, K *sorin*, D *sorɪɲ*.
* “smoke”: N *túllí*, D *tulla*. ◊ This may be a recent innovation in both languages; cf. the morphological discrepancy, the fact that the stem in N is a better match for K *tulli* “chewing tobacco,” and the lack of attestation in ON. Obvious similarity with Nuer *toːl*, Dinka *tol* “smoke” suggests an old areal isogloss.
* “that”: N *mán*, K/D *man*.
* “this”: N *in*, K *in*, D *ɪn*. ◊ The subsystems of deictic pronouns in M, B, and Hill Nubian are much less cohesive than in Nile-Nubian and do not allow for reliable reconstructions of any PN items that would be different from Nile-Nubian.
* “what”: N *mìn*, K *min*, D *mɪn*. ◊ It is quite possible that the Nile-Nubian situation here is innovative, since all other branches agree on *\*na(i)-* as a better equivalent for PN “what?”: M *nèː-n*, B *na-ta*, Dl *na*, Karko *nái*, etc.[^16]
* “woman”: N *ìd-éːn*, K *eːn*, D *ɛːn*. ◊ Technically, this is not a fully exclusive Nile-Nubian isogloss — cf. B *eːn* “woman.” However, the main root for “woman” in Nubian is *\*il-* (ON *il-*, M *ìd-dì ← *il-ti*, Dl *eli*, Karko *îl*, etc.); *\*eːn* is the common Nubian word for “mother,” which has, most likely, independently shifted to “woman” in general in modern Nile-Nubian languages and in B. N is particularly innovative in that respect, since it uses a compound formation: *ìd* “person” + *éːn* “mother.”

[^feather]: Khalil, *Wörterbuch der nubischen Spra­che,* 124.
[^16]: In Starostin, *Jazyki Afriki,* p. 92 I suggest that, since the regular reflex of PN *\*n-* in Hill Nubian is *d-*, both Nile-Nubian *\*min* and all the *na(i)*-like forms may go back to a unique PN stem *\*nwV-*; if so, the word should be moved to section I.1, but in any case this is still a common Nile-Nubian isogloss.