S/š-Ṭ-Ḥ in Algerian Arabic

On holiday in Algeria, I've been going frequently to some coastal rocks called əṣ-Ṣṭuħ الصطوح, from the plural of ṣṭəħ صطح "flat terrace roof", and occasionally eating šṭiṭħa شطيطحة "stew". I haven't done any dancing (šṭiħ شطيح) - not really my thing. But if I go into the mountains where they speak Kabyle, I might hear someone call something aməšṭuħ (amecṭuḥ) "small"; if they ask me whether I understand, I might reply šiṭṭuħ kan (ciṭuḥ kan) "just a little."

Can all these words be connected to one another? I think they might be.

Let's start with the easiest one: ṣṭəħ is obviously from Classical Arabic saṭħ سطح "flat roof, surface", a noun corresponding to the verb saṭaħa "spread out and level out". The root has cognates elsewhere in Semitic, notably Aramaic šṭaħ ܫܛܥ "to spread something flat" and Hebrew šāṭaħ שָׁטַח "to spread, stretch", so its Semitic origin is secure. (In the usual weird quasi-biradical way, it also fits into a wider pattern of "spread" words in s-ṭ, notably Arabic basaṭa "spread out".)

However, one notes that the sibilant elsewhere in West Semitic is, regularly, š. Could the various words with šṭħ somehow reflect an early parallel borrowing from Aramaic?

Let's start with Arabic: a verb šaṭaħa can indeed be found in Classical Arabic lexicons, meaning "to go too far". In Egyptian Arabic, Hinds and Badawi give šaṭaħ as "to veer, go off course". For Levantine Arabic, Neishtadt summarises a variety of sources as giving the range of meanings "to extend, lie down; to fell, slip (on the ground); to make a big step; to move away, depart; to go out, go for a walk, have a picnic, wander", while making the point that the verb must derive from Aramaic there.

A semantic link between "go far" and "spread out flat" is easy enough to envision: to be spread out flat is to be spread across the widest area possible. (Cf. English "stretch".) A priori, the easiest way to get from "go far" to "dance" is probably via the attested Levantine sense "to make a big step".

If Arabic šṭħ comes from Aramaic, then at some point one would expect it to have retained the original sense "to spread something out flat"; attested Levantine "extend" seems sufficiently close to this. Mašṭūħ would then be the past passive participle "spread out flat". Anything in such a state has been reduced to its minimum height; it would therefore be reasonable to call something particularly short məšṭuħ, or with the Berber nominal prefix a-məšṭuħ. From "short" to "small" to "little" seems sufficiently plausible.

The biggest remaining problem, then, is šṭiṭħa - are we to exclude this delicious dish from the lexical family? (Any direct derivation from "dance" is surely to be excluded, Wikipedia's fanciful proposal notwithstanding.) Neishtadt cites a suggestive Levantine form: šṭīḥ "flat plate usually made of sheepskin (which is used to cover the bread when kneaded / on which one puts the dough and then cuts it into small pieces)". Compare likewise Iraqi šaṭiħ "flat, shallow", used among other things to describe dishes. A plausible starting hypothesis, then, is that šṭiṭħa originally described the flat-bottomed pan in which the stew was prepared. The word has the form of a regular diminutive from a noun *šəṭṭVħ-a, which might be worth hunting for.

So, on a preliminary check, I'm inclined to say that yes: all the words discussed in the first paragraph derive from the same Semitic origin, one via Arabic proper, the rest via early borrowing from Aramaic. It's quite possible that wider lexicographic research would make it possible to fill in the missing steps in the semantic shifts directly, rather than speculatively - or that it would just complicate the picture further.

Addendum: Another member of this word family in Darja is ṭəħṭaħa طحطاحة "flat expanse (of land)". Good luck simultaneously reconciling both that and basaṭa with a biradical analysis.

Comparative East Chadic B evidence on the Mubi pronominal series

Mubi is the best described member of a small "Mubic" subgroup consisting of Masmedje, Kajakse, and Zerenkel. The other members are much more poorly described, but enough can be gleaned from published materials to show that they share essentially the same system of subject pronominal suffix series with different syntactic functions.

For Masmedje, a short text is provided in Marti et al. (2007), from which much can be gleaned using Jungraithmayr's Mubi grammar. Sentence 3 is particularly illuminating:

wa	ka-sek-a	an-sual-ukun
and[?]	2-come.PFV-?	1Pl-ask-2Pl.I/td>
You (pl.) have come to ask us questions.

We also find examples of what seem to be 1Sg.III -ene (#9), 3Pl.I -ko (#6).

For Zerenkel, Ramat provides tac-ku "hit-3Sg.I", bar-kute "give-3MSg.II"; we've already seen -na "-1Sg.I", -ci "3FSg.I". Hunting through the untranslated texts might provide a fuller paradigm.

So it looks like this system is common across Mubic. What about its closest relatives, the Northern Guera languages? There, we find something reminiscent of Mubic, but a bit different: a distinction between preverbal subject proclitics (which do always seem to mark the subject, not the object) and postverbal subject suffixes, whose functions include marking subject continuity and object relativisation.

The best described of these, East Dangla (Shay 1999:125), has a distinction between preverbal subject markers (the default) and postverbal subject suffixes used when the subject is the same as in previous clauses. But the postverbal subject suffixes do not affect the interpretation of the immediate preverbal element, which remains a subject. Even if the object is fronted, the subject continues to intervene between it and the verb (ibid:130). Unfortunately, Shay does not describe the syntax of relative clauses in any detail, but every example of non-subject relativisation provided (pp. 95#24, 188#57, 213#63, 252#43) displays subject suffixes rather than subject prefixes.

For Bidiya, Alio (1986:222) makes the latter point explicit: object relativisation forces the verb of the relative clause to use the subject agreements suffixes, rather than the proclitics. The functional distinction between the two does not seem to be otherwise discussed, but examples (e.g. p. 347) again make it clear that the suffixes do not affect the interpretation of the preverbal position, which continues to be reserved for subjects.

For Migama, Jungraithmayr and Adams' (1992) sketch grammar reveals three different subject proclitic series, but no suffixed subject agreement series at all.

Further afield in the Southern Guera languages, we find even less in common with the Mubic system. As far as the materials available to me suggest, it seems that subjects there are simply preverbal irrespective of pronominal or morphological status, and irrespective of the syntactic context.

In Barain, "Subjects are primarily encoded by word order, being the only argument before the verb. Subjects pronouns appear to occur in the same syntactic position as other subjects" (Lovestrand 2012:132); pronominal elements suffixed to the verb are simply direct or indirect objects. The same structure applies in relative clauses (ibid:189). For Gali (Jungraithmayr and Peust 2019) and Saba (Jungraithmayr 2020), the brief sketches available likewise indicate that subject pronouns only precede the verb.

Putting this together, we start to get an idea of where the Mubi system might have come from. I assume that proto-Chadic basic word order was VSO, as still in some of its descendants. In proto-East Chadic B, the default word order was probably already SVO (fully generalised in Southern Guera), but VS was still preferred in object extraction contexts and for old information subjects, morphologising into a new subject suffix series in Northern Guera and Mubic. Old information subject contexts would notably have included object focus, a context for which fronting is cross-linguistically widely preferred; after the morphologisation, this would have led to the subject suffix series becoming obligatory when the object directly precedes the verb. Mubic overgeneralised this correlation, reinterpreting the subject suffix series as requiring the NP preceding the verb to be interpreted as its object.

Relativisation in Mubi and Zerenkel

The remarkable voice system described for Mubi seems to interact with relativisation, but textual data is scarce. In textual examples of subject relativisation, the same suffix shows up that Jungraithmayr describes as forming attributive adjectives; compare:

njó	tú	gì	hìtt-ìt	sàhíi	tú
person	?	M.REL	pronounce.IPFV-ADJ	truth	?
a person who speaks the truth (Jungraithmayr 2013:64, 115)

jóról	gì	mèed-ít
jackal	M.REL	die.PFV-ADJ
a dead jackal / a jackal which has died

Lukas gives phrasal examples where there is clearly no suffix, but are they really relatives?

njóo	gàŋ	húwútà	bògòs
person	M.DEM.DIST	fish[V].IPFV	fish.PL
fisherman = a person who catches fish (Lukas 1937:159)

What about when you relativise on a direct object? In the two textual examples found, the subject agreement seems to be from Series III (not I, as I would have expected); but perhaps these verbs take obliques rather than direct objects in Mubi. Thus:

lì	gì	fány-gát
thing	M.REL	want.PFV-2MSg.II/III
the thing that you want [whatever you want] (Jungraithmayr 2013:147)

lì	gì	wáa-got	wàaró	gà
thing	M.REL	call.PFV-3Pl.III	grain	DEM.M
this thing that they call grain (Jungraithmayr 2013:139)

The only other Mubic language for which data is available behaves reasonably similarly. In Zerenkel subject relativisation, we see what might be the same attributive suffix in some cases, whose -t is lost in Zerenkel:

ro	ga	guwaan-i	weyya
man	M.REL	cultivate.IPFV-ADJ?	field
the man who cultivates the field (Ramat 2017)

although the verb may instead appear with Series I for clause-internal reasons independent of relativisation:

luk	da	na	umma	ci
woman	F.REL	1Sg	see.IPFV	3FSg.I
the woman who sees me (Ramat 2017)

In object relativisation, it looks as though we see Series I as expected:

luk	da	umma	na
woman	F.REL	see.IPFV	1Sg.I
the woman whom I see (Ramat 2017)

Notes on Mubi syntax

Looking over Jungraithmayr's Mubi grammar again recently, I think I managed to make sense of the subject paradigms. Presumably all this was clear to Jungraithmayr, but it certainly wasn't clear to me, nor was it mapped to cross-linguistic typological categories; it may therefore be useful to comment.

At first sight, Mubi looks like its basic order is SVOX, as in sentences like:

màb-í	nyúbùt	rúm-jí	á	sìrèebí-jí
grandmother-1SgGen	carry_on_back.PFV	daughter-3FSgGen	LOC	rib-3FSgGen
My grandmother carries her daughter on her side. (Jungraithmayr 2013:153)

However, the reality is a little more complicated. Basically: Mubi syntax has what I suppose is a focus slot immediately before the verb, which can be filled by a noun phrase or by a short pronoun ("pronom personnel sujet préposé"). The syntactic role of an element in this slot is determined not by its position alone, but crucially also by what immediately follows the verb stem:

• If the verb has no pronominal suffixes, the preverbal element is the subject.
• If the verb has a pronominal suffix, the preverbal element is the object, whose interpretation depends on the subject suffix series:
  • J's "série I" puts the verb into object voice (cf. Remijsen on Shilluk) - that is, it promotes the direct object to this position;
  • J's "série II" puts the verb into benefactive voice - that is, it promotes the indirect object to this position;
  • J's "série III" puts the verb into applicative voice - that is, it promotes what would otherwise have been the object of a preposition (typically tí "with") to this position.

In relative clauses too, the head is treated as if it filled the preverbal slot (and, indeed, does precede the verb of the relative clause).

When arguments are not promoted, they remain postverbal; there are two distinct suffixed object pronoun series for direct and indirect objects respectively.

It is possible to have a ditransitive sentence with both objects postverbal, e.g.:

báar-á	tògò	gì	ʔjúrúk	à	sùltân
give.PFV-3MSg.0[!]	skin	Gen.M	leopard	to	Sultan
He gave the skin of the leopard to the sultan ("er gab der Fell des Leoparden dem Sultan" - Lukas 1937:177)

dì	lìî	kòrré	à	lòk
3FSg	make.PFV	beer	to	women
She made millet beer for the women. (Jungraithmayr 2013:143)

However, Jungraithmayr gives plenty of examples where one or the other is preverbal, in which case the appropriate subject pronoun series is used:

ká	háráɗí-ndé	sèerí
2MSg	tie-1Sg.II	rope
I tied the rope for you. (Jungraithmayr 2013:44)

ár	lúk	ngér-gu-dó
3MSg	woman	find.PFV-3MSg.I-1SgDat
He found a woman for me. (Jungraithmayr 2013:43)

It is also possible to use these series without any overt preverbal element, e.g.

íwín-gu	mà	cíi-gú,	ɓée-gút	á	gùrá-t
tie.PFV-3MSg.I	and	take.PFV-3MSg.I,	go-3MSg.II	to	bush-3MSg.Gen
He tied him and took him, went with him [brought him] into the bush. (Jungraithmayr 2013:43)

It looks like cases like this can most easily be interpreted elliptically, as featuring a null preverbal "pronoun".

So here we have a system that gives SVO if the verb lacks subject agreement, but SOV (one way or another) whenever it does have agreement. This is curiously reminiscent of West Nilotic. Anywaa is kind of the inverse: you get OVS if the verb lacks subject agreement, but SOV if it has it. The choice among subject agreement series is functionally very similar to what we see in Shilluk, where the preverbal argument is interpreted as a subject if the verb is in subject voice, an object if the verb is in object voice, and an instrument if the verb is in applicative voice, while all other arguments follow the verb. But Shilluk differs in that, unless the verb is in subject voice, the subject is marked by a preposition.

For completeness, here are (most of) the forms, somewhat simplified:

	free	short	I = object voice	II = benefactive voice	III = applicative voice	DO	IO	Poss
1sg	ndé	ní	-nà	-ndé	-én	-din	-dó	-̀í
2msg	kám	ká	-ká	-kát	-kát	-d(og)a	-(d)(ìg)á	-̀dá
2fsg	kín	kí	-ké	-két	-két	-dige	-(d)ìgé	-̀jí
3msg	ár	à	-kū	-kút	-kūt	-dar / -é	-(d)ìgí	- V́t
3fsg	tír	dì	-kī	-kít	-kīt	-dir	-(d)ìgí	-́jí
1exc	éné	á	-(á)nē	-nét	-ànēt	-dìné	-dìné	-́jìné
1inc	áná	á	-(á)nā	-nát	-ànāt	-dá	-dá	-́já
2pl	kéné	ká	-kún	-kún	-kún	-dùgún	-dùgún	-̀júgún
3pl	kér	kà	-kō	-kót	-kōt	-dugur / -á	-dùgó, -dógór	-́jó

We may notice that the benefactive and applicative are nearly the same, but with traces of an extra vocalic element preceding the agreement in the applicative; both show obvious traces of what must once have been a stranded preposition tí "with" at the end. Both the DO and IO series appear to show the same element before the pronominal marker rather than after it, oddly reminiscent of Hebrew ʔɛṯ. All three voice series show a curious -k- preceding the 3rd person forms.

Mabaan domestic animal loanwords

The core of the Eastern Sudanic hypothesis is the idea that Nubian - historically the largest language family of Sudan proper - is related to Nilotic, the largest language family of South Sudan. One complication for this analysis is that, as the primary language of the medieval Nubian kingdoms, Nubian was well-positioned to influence the more northerly Nilotic languages from an early date. A couple of possible cases recently came to my attention.

Mabaan (not to be confused with Maban) is spoken in South Sudan's Upper Nile province near the Yabus River, at the southern fringes of a region historicallly dominated by the Funj sultanates. It belongs to the Burun subgroup of West Nilotic. In the course of elicitation recently, I was struck some domestic animal terms that did not seem to have obvious cognates elsewhere in West Nilotic:

• "horse": mún-t̪-ʌ̂n, pl. mún-t̪án
• "pig": kúc-c-ʌ̀, pl. kwɔ́t̪-în
• "camel": kʌ́l-ŋʌ̂n, pl. kʌ́l-ŋàn
• "sheep": kál-ŋ̀, pl. kʌ́l-ìn

"Horse" immediately recalls Old Nubian and modern Nobiin murti, borrowed across a wide area of eastern Africa (see e.g. Rilly 2009:168) including into nearby Bertha as murθá (Bender 1989). The nasal is not isolated; as Andersen (2006) shows, in Mabaan most nouns that originally ended in stops have replaced them with the corresponding nasal in the singular, reflecting merger with a secondary singular suffix *-n. However, this development did not affect nouns ending in *r, which regularly becomes y; is this a regular development of the cluster *-rt-, perhaps?

"Pig" is evidently likewise to be related to Old Nubian kutunni (pl.; Browne 1996:102) and similar forms spread equally widely across the Sudan, including in Sudanese Arabic (cf. Spaulding 1989). Proto-Koman *gUd̪Um (Otero 2019) is part of the same series, but the voicing difference suggests a different borrowing path (though some Koman languages do have k- here).

"Camel" is a more widespread root, though it does not seem to be found elsewhere in West Nilotic, and most similar forms elsewhere in the region start with g/j (as in Arabic or Cushitic) rather than k. Old (and modern) Nubian kam- is a possible source, via a plural such as Kunuzi kamli; Mabaan's closest relative, Jumjum, has kalman, pl. kʌlʌmgʌ (Jumjum Language Committee 2020), suggesting historical loss of m in the Mabaan form.

For "sheep", the most immediate comparandum is found next door in Koman: Central Koman *kʰālí (Otero 2019). Jumjum kabal matches Uduk kʰāɓāl (Killian 2015). The similarity to Old Nubian tikan- (modern tiigan) is probably only suggestive; it would neither explain the l nor the b/Ø variation.

Darja notes from the past

Going through some old papers, I found some notes on Dellys dialect that I had taken years ago from Amti Khira; better to put them up (mainly so I can find them more easily) than leave them to get dusty.

A nursery rhyme, legendarily said by the swift (əl-xŭṭṭayfa) on its return from migration:

يا مولات البيت البيت
أعطيني كسيرة بالزيت
قالتلي كولي وكليت

Ya mulat əlbit, əlbit
Aʕṭini ksira bəzzit
Qalətli kuli wə-klit

"Mistress of the house, of the house,
Give me a little loaf with oil!"
She told me "Eat!" and I ate.

A few expressions, I think from kids' stories: Sidi Bəllarəj ḥməṛ əlkʷriʕat "Mr. Stork with the little red legs", Sidi ʕabbʷa lli bbʷa lbab "Mr. Abba who carried the door on his back". The verbal noun of bbʷa is mə́bbʷa.

Some traditional solar month names: məɣṛəṣ "March", yəbrir "April", mayyu "May".

A not entirely successful attempt to elicit traditional lunar month names: muħəṛṛəm, safəṛ / šiʕ əlʕašuṛ, lmulud ənnabawi, (missing), ṛabiʕ θθani, jumad ʔuwwəl, jumad θθani, ṛjəb, šaʕban, ṛəmḍan, šuwwal / ləfṭaṛ, ðulqiʕda, ðulħijja.

A curse/insult from an old story (possibly related to Qari Achour?): šixkŭm ma yətʕəqqəl, u zitunkʊm ma yətləqqəm "your elders will not grow wise, and your olive trees will not be grafted".

Miscellaneous vocabulary: đ̣əbbəħ "to shout", غلال اليهود a kind of black snail (shelled; not eaten).

Darja miscellaneous notes 2025

Every time I go to Algeria, I come back with some linguistic observations that are new to me (if not necessarily to anyone else.) Here are this year's.

Many collective nouns take plural agreement: sqit əššjəṛ əttəħtaniyyin “I irrigated the lower trees”, kanu sjəṛ “there were trees”, ənnməl haðu “these ants”. Not all do, though, or at least not all the time: nnamus bəkri kʊnna nšufuh nəqqʊtluh “mosquitoes, in the old days, if we saw them (lit. it) we’d kill them (lit. it).” A topic worth looking at in more detail.

“Have”-based expressions for “ago” are familiar from Romance languages; in Darja, however, they agree with the notional possessor, e.g. dərtu ma-ʕəndi-š bəzzaf ‘I did it not long ago’ (lit. “I did it I don’t have much”). Along similar lines, the subject of ʕla bal-i “I know” (originally “on my awareness”) was originally the theme, the fact known. Synchronically, however, utterances like ma-kʊnt-š ʕlabal-i “I didn’t know” (lit. “I was not I know”) suggest this is no longer the case.

Another example of næ̃mpoṛt (discussed here previously): u xəllih yakʊl næ̃mpoṛt ħaja ‘and let him eat anything’.

The construct state has undergone some interesting developments. Most masculine nouns have no distinct construct state, and most feminine nouns form a construct state by replacing -a with -ət. If we factor out, for the present, the stem-internal effects of schwa-zero alternations and compensatory gemination, then, for most nouns, we can speak of a single construct state used for head nouns followed by possessor NPs or by suffixed possessor pronouns alike. However, a few nouns show a different distribution. Several kinship terms in -a take the suffixes directly: yəmma-k ‘your mother’, baba-k ‘your father’, jədda-k ‘your grandmother’, even ṭaṭa-k ‘your auntie’. (These nouns have zero-marked 1Sg possession: yəmma u yəmma-k “my and your mother”.) Such nouns usually take clitic doubled possessives (yəmma-ha ntaʕ Baya ‘Baya’s mother’, lit. ‘her mother of Baya’); however, if used in the regular synthetic possessive (“iḍāfah”) construction, they take a suffix t, e.g. yəmma-t yəmma-k “your mother’s mother”. For these nouns, it seems tempting to postulate two construct states rather than one.

The noun pattern CəCCayC is not particularly productive, but I heard a new example: tərtayqat “firecrackers” (cf. tərtəq “pop”). Other examples include ħərrayqa “jellyfish” (ħrəq “burn”), xʊṭṭayəf “swallow (bird)” (xṭəf “snatch”), bu-zəllayəq “blenny (fish)” (zləq “slip”).

Feminine nouns without overt feminine marking form diminutives with overt feminine marking: yədd ‘hand’ > ydida ‘little hand’. Very few masculine nouns have apparent feminine marking, but x(a)lifa ‘caliph’ is one such; məskin əlxliyyəf haðak “poor little caliph!” shows that the converse is also true, i.e. that masculine nouns with apparent feminine marking form diminutives without it.

The verbal template CəCCəC is in generally semantically and syntactically distinct ftom its corresponding passive/middle tCəCCəC. However, the distinction is neutralised in the participles: mwəð̣ð̣i “washed for prayer” from twəð̣ð̣a, mkəṛməṣ “dried (of figs)” from tkəṛməṣ “dry (of figs, intr.)”. Some speakers, however, do say mətwəð̣ð̣i.

Passives in n usually involve a simple coda n, but I heard clear gemination in li baš yənnəqsəm ‘for it to be divided’. The question of gemination in triliteral passives would deserve a closer look.

Weak-final triliteral verbs tend to add -an- in verbal nouns: tənħaniyya “removal” from nəħħi “remove”.’

A few emotional idioms: bərrəd qəlb-u “he cooled his heart”, i.e. he satisfied his heart’s desire; ṭəyyəṛhali “he made it fly for me”, i.e. he made me lose my temper; ṭəḷḷəʕlu lgaz “he raised the gas for him”, i.e. he made him angry. A proverb: triq əlʕafya tənẓaṛ yalukan tkun bʕida “the road of safety gets visited even if it’s far away.”

The usual ‘whatchamacallit’-word in Dellys and elsewhere in Algeria is laxʊṛ, originally “the other one”, used to substitute for verbs as well as nouns. However, from a relative about 90 years old, I heard a different construction based on haðak “that”: ma-yhaðak-š “he doesn’t whatsit”. This is paralleled in Malta and Morocco, so presumably it used to be more widely used.

The usual word for “knife” in Dellys is mus, but xʊdmi (usual in Bechar) is also in use. However, I hadn’t previously heard xʊdmiša. The curious final š can perhaps be explained as a borrowing from Berber, in some varieties of which ṯaxʷəḏmiyṯ would regularly yield ṯaxʷəḏmišṯ.

French cinquante is often heard as sikõnt “fifty”. The vowel is difficult to explain – influence from another Romance language?

Some words new to me: gərziz “empty gum, empty tooth socket”; ma-ksan-š “he’d rather not”; ṣfiħa “horseshoe”.

The ʕ in the verb ‘give’ is often elided: aṭini “give me” for regular aʕṭini.

I don’t think triliteral verbs ever end in w, but quadriliterals may: yqəwqəw ‘(a chicken) cackles’ (usually yqaqi in Dellys), yčəwčwu ‘they chatter’.

HEAD = GOURD in Algeria

The metaphorical identification of heads with gourds is probably obvious enough to arise spontaneously anywhere that gourds are in regular use (even English has expressions like "stoned out of his gourd".) In Algeria, it is historically reflected in some varieties' lexicon. Kabyle has in most contexts replaced pan-Berber ixf with novel a-qəṛṛu, whose ṛ betrays its loanword origin. The immediate source seems to be dialectal Arabic qəṛṛuʕ, attested in the meaning "head" around Jijel, but originally "big gourd", imposing the augmentative template CaCCūC on the noun qarʕ (dialectal qəṛʕa) "gourd, squash". (One might also consider a role for Classical ʔaqraʕ "mangy, bald", dialectal gəṛʕa "bald".

The thing about metaphors, though, is that they appear across multiple domains, not just in language. I recently learned of a traditional Algerian treatment for migraines (reported to be very effective) that involves cutting a fragment of gourd, writing various symbols on it, and pressing it against the appropriate place on the head of the affected person. The same metaphor that produced lexical change in Kabyle has evidently inspired curative practices next door. Perhaps a wider cultural survey would yield examples in other domains as well?