Introduction: linguistic diversity. In Gīlān there are three major Iranian language groups, namely Gīlakī, Rūdbārī, and Ṭālešī, and pockets of two other groups, Tātī and Kurdish. The non-Iranian languages include Azeri Turkish and some speakers of Gypsy (Romany, of Indic origin). Gīlakī is spoken by possibly three million people as a first or second language, and has had a budding literature and fledgling prose publications, including newspapers, but both Gīlakī and Ṭālešī are rapidly losing ground in many cities of Tavāleš due to heavy immigration of people from Azerbaijan.

Dialectology. The five Iranian languages in Gīlān belong to the Northwestern branch of Iranian. Gīlakī, which has two main dialect types, eastern and western, with the Safidrūd River as the general border, is a member of the Caspian subgroup. Tātī and Ṭālešī (Talyshi) together make up the larger dialect chains which together make up the larger Tatic family (not to be confused with Tat-Persian spoken in pockets north of the Baku area). Among these, the two Tātī pockets in Gīlān, Kalāsī and Kabataʾī, have their closest relatives in Upper Tārom in Zanjān province. Tālešī is a dialect chain of three main types, southern, central, and northern; and southern Ṭālešī is closer in type and mutual comprehension to some forms of Tātī than it is to central or northern Ṭālešī. Rūdbārī may originally have been a subgroup of Tatic that has largely adapted structurally to Gīlakī.

In the citations below, Gīl. = common Gīlakī/all dialects; WG and EG = Western and Eastern Gīlakī; Lāh. = Lāhījānī; Gāl./Lang. = Gālešī/Langarūdī as in Pāyanda; Māč. = Māčīānī (unlabeled forms cited below are WG/Raštī; EG refers to Lāhījānī unless otherwise stated).

Diachronic developments. Due to heavy influence from Persian and SW Iranian, the typical NW Iranian have been retained only sporadically: IE * ǵ > Proto-Ir. *dz > z: zama “son-in-law, bridegroom, wife’s brother” but SWI d, as in Gīl. dan- “to know,” dil “heart” diruz “yesterday.” IE *ḱṷ > *tsv > sp: səbəj “louse,” but SWI s, as in Gīl. səg “dog.” IE *tr > Ir. *θr > hr- > r-: dare “sickle,” but SWI s-, as in Gīl. se “three,” pəsər “son.” Initial Proto-Ir. *dṷ > b: Gāl. bər “door,” but SWI d-, as in WG dər “door,” Gīl. de, digər “else, other.” Ir. -rt > -rd/-rt, as in purd/t “bridge” (cf. Pers. pol). Ir. -šm > -m in čum “eye” besides čušm. Retention of initial IE *l in Gāl. luas “fox” (cf. Skt. lopāśá, Pers. rūbāh).

Later changes include: Initial Ir. *fr- > WG: f-, EG h-, as in the preverb *frā- > WG fa-, EG ha-, e.g., WG fa-gift-, Lang. ha-git- “to take from, to buy,” but SWI fr, as in Gīl. furuš- “to sell,” fərma- “to command.” Initial Ir. *xr- > h-: hin- “to buy,” but SWI xr-, as in Gil. xurus “rooster,” xərəd “wisdom.” Initial Ir. *w- > v- versus SWI g, as in WG. višta “hungry,” Gāl. vəšnə “hungry,” vuruj- “to flee,” vərg “wolf,” but also WG guriz- “to flee,” gurg “wolf,” and Gīl. gul “flower,” guzər- “pass”; Gīl. v-versus SWI b- as in vásti “must,” var|varəst “to precipitate (rain, snow),” və́stə “enough!,” və-gərd- “turn, return,” vini- “nose” (in vinizək “snot”), Lang. va “wind,” vəlg “leaf,” vərf “snow,” Māč. and Gāl. vi- “willow,” but also Gīl. bad “wind,” and WG bərg “leaf,” bərf “snow,” bid “willow,” badam “almond” (note Mid. Ir. vyāg- “place” > Gīl. jiga, cf., Solaymānī/Mokrī Kurd. jēga, South Tātī yaga, Ḵᵛānsārī yaγa. Ir. *-ft, -xt > -(h)t > -(t)t, exemplified in EG past stems: (a) ft > t in git- “took,” kət- “fell,” gut- “said,” xot- “slept,” but SWI ft,in WG: gift-, kəft-, guft-; xuft-; (b) xt > t, in sut “burned,” pet “cooked,” dut “sewed,” sat “built,” but WG. soxt-, pəxt-, doxt-, saxt-; note also Gāl./Lang. dətər “daughter.” Initial Ir. *xw > x,as in xab “sleep,” xast- “wanted,” xand- “read,” xiš “relative” (probably all borrowed words), and xaxur “sister.” Ir. *č >j,as in suj- “to burn,” Māč. vürüj-, Gāl. vuruj, “to flee,” but SWI z as in WG pəz “to cook,” guriz “to flee”; note alternates je ~ az “from,” duj ~ duz- “to sew”; but č is retained in WG alternate: pəč- and Gāl./Lang.: puč- “to cook”).

More recent changes include weakening or loss in intervocalic and final position: Loss of d as in Gīl. mar “mother,” per “father,” Lang. məar, pier,Gīl. bərar “brother,” du “smoke,” zama “son-in-law.” Loss of γ, as in Lang. duro “lie,” du “buttermilk,” but WG: doq, duroq.

Other miscellaneous consonant changes: nd > d, as in WG: du-xan-, du-xad “call, invite” (cf. Pers. xān/xānd-), də-vəd- “to close, to tie.” p > b > v, as in va-vurs- “ask” (cf. Pers. pors-); va-vixt-, Lang. fi-vit “tied, wrapped,” but di-pext- “idem,” luas “fox,” aseyə, aseyow, “mill,” but WG asiab, ruba; note devoicing in Māč. juraf “socks,” jif “pocket” vs. WG. jurab, jib. Ir. b- > v- in bərd- “carry, take away” vs. fə-vərd- “swallow,” va-bin ~ va-vin “to cut,” də-bəd- ~ də-vəd- “to close, tie.” r may be dropped in clusters, as in kud “made” < kurd, gift- “to take” < grift, bin- “to cut” < brin-, bij- “to roast, fry” < brēj- (cf. Sangesarī berin-, beriž-; Vafsī birin-, biriz). Other typical Northwestern Iranian features not found in Gīlakī, such as the retention of initial y- vs. SWI j-, have been totally lost probably due to heavy borrowing from Southwestern Iranian.

In some cases, the Southwestern Iranian borrowed root has become the primary word in the modern language, but the original Northwestern Iranian equivalent, or a Gīlakī innovation, is found as a doublet in frozen form, often in compounds, e.g., WG: duz- “to sew,” but fuduj kudən “to darn,” az ~ “from,” damad, zama “son-in-law, bridegroom.”

Vowels. The distinction of length is lost in both the inherited and borrowed words, and the resulting merger often includes original majhūl vowels as well. Thus, long ī > i, as in fil “elephant,” bil “spade,” and short i > i, as in dil “heart.” Long ū > u in mur “ant” and short u > u in pur “full,” note gul “flower” (< gul), gul “deceit” (< gūl); occasional u > o in Arabic loan words, as in sob “morning,” roxsət “permission, leave,” šoql “occupation,” sol “peace,” but also buxar “steam,” fursət “opportunity” hukumət “government.” Due to heavy influence from Persian, there is considerable variation. Earlier long ē, ō are retained as mid vowels e, o, or merge with i, u.; thus ē > e, as in der “late,” seb “apple,” bex “root,” teγ “blade,” but > i in tiz ~ tij “sharp,” bi “without,” bivə “widow.” Similarly ō > o, as in doxt- “sewed,” soxt- “burned,” xob “good,” kob- “to pound,” but > u in dust “friend,” ruz “day,” mum “wax,” zur “strength.” There are doublets such as mex ~ mix “nail,” gor ~ gur “grave,” including alternation in present and past stems, as in soxt ~ suz- “to burn,” doxt ~ duz- “to sew,” and contrasts such as sir “garlic” vs. ser “satiated.” The retention of o < ō and o < u may be conditioned by back consonants (k, g, x, g, h), e.g., present stem duz-, but past stem doxt, sob < subh, but it is inconsistent, as in buxar, hukumət. The diphthong ow merges with o, as in nobet “turn,” julo “forward.”

In terms of historical morphology, the most significant features are two tense formants in Eastern Gīlakī (see Table 6a): (1) the present tense formant, -ən-, which originates in Indo-European *-ent-, the formant of the active present participle. This feature unites Eastern Gīlakī with Māzandarānī, the dialects in the Semnān area (including Šahmīrzādī, Sangesarī, Aftarī, Sorḵaʾī, Lāsgerdī, but not Semnānī itself), Northern Tātī, and Zāzākī/Dimlī (q.v.) in eastern Turkey (immigrated from former Deylam abutting on Gīlakī-speaking areas). (2) The past conditional formant, -èn- (see Table 6a ), which parallels the conditional -èn- in Gōrānī/Gūrānī (q.v.; found north of Kermānšāh and in the area of Mosul) and in Baluchi past subjunctive -ēn- (Rastorgueva, pp. 336-37), is suffixed to the past root and used in irrealis conditional sentences in all three languages. In both of these features, Western Gīlakī behaves more like its western and southern neighbors, since the present and imperfect tenses are formed the same way they are in Southern Ṭālešī, in Kolūrī Tātī, and in Rūdbārī.

Phonology. Consonants are similar to Persian, except that uvular q has only voiced fricative pronunciation in all positions, e.g., γurban [γurbšn] “sacrifice.” The Gīlakī vowel system sounds radically different from other Iranian languages and seems quite elusive. Eastern Gīlakī vowel phones have never been clarified, and transcription in this article (ə, š, I, U, etc.) is not meant to imply phonemic status. While Arthur Christensen (1930) distinguishes twenty-one vowel phones and Rastorgueva (1971) posits nine phonemes for western Gīlakī, there are probably not more than six or seven phonemes: i, u, e, o, ə, a1 (a2?). Russian sources transcribe a1 and a2 as a and å respectively and distinguish between kari “you plant” and kåri “working (adj.)” or mašin “machine, automobile” and målik “landowner.” The present author has not been able to find any such distinction. The symbol å usually represents a low, back, rounded a, but this sound only occurs in the heavily Persian-influenced speech of some bilinguals; a1 and a2 seem to be pronounced exactly the same in most situations and contrast only marginally. That is, while a2 always remains constant, a1 has a wider allophonic distribution, alternating between central [a] and [æ>](dášti “you had” is often pronounced [dæˊ>šti], almost identical to Pers. dæˊšti “a field”). Only in cases of potential ambiguity are a1 and a2 clearly distinguished, e.g., gila1n [gilan] “Gīlān” vs. gila2n [gilæn] “mud (pl.).” The pronunciation [œ] is heard more commonly in Lāhījānī. Further confusion is caused by the reflexes of original proto-Ir. *a¡ in Gīlakī as ə in all positions except initial, where it is also realized as a, e.g., atraf “sides,” asban “horses,” adəs “lentil.” The phoneme ə alternates between [ə], [e ], [e],and final [I]:xəstə [xəstI] “tired.” Gīlakī e and o are seen in: der “late,” bihem “I bought,” vapéxtəm “I wrapped (it), wound (it) around,” and kor “girl,” sob “morning,” and xob “good.”

Stress. Stress is mostly syllable-final, but some morphemes, particularly in verb forms, require stress shifts: xayə́m “I want,” nə́xayəm “I don’t want,” bəxástidi “you/they wanted,” bəxastə́-bid “you/they had wanted,” facukəstə́n-dərimi “we are climbing,” duṹcəkəstə-bu “it hadn’t stuck.” More Eastern Gīlakī verb forms take initial stress than western Gīlakī: EG pórsənəm “I ask,” bə́porsəm “that I ask (subjunctive),” bə́porsem “I asked,” bə́porse-bum “I had asked,” hæˊdyenæ-bum “I would have given,” but other stress rules are common to all dialects. Note the following stress contrasts: Gīl. xaná “legible,” xána “the khan (dir. obj.)”; ruzí “daily sustenance,” rúzi “some day, per day”; bidarí “wakefulness,” bidári “you are awake”; WG vavursə́m “I ask,” vávursəm “that I ask”; bídin “see!” bidín “irreligious”; EG xánəm “I want,” xanə́m “lady.”

Noun phrase: Number. One plural, -an (Lang. -ən, Lah., Gāl. -on ), for all noun types: seb-an “apples,” kor-an “girls”; Lang. kərk/kərkən “chicken/s”; Lah., Gāl. kərk/kərkon “chicken/chickens.” Unstressed suffixed indicates indefiniteness (singular or plural): xob duktúri-ə "He/She is a good doctor,” xob cizáni aya naha "There are good things here.”

Object marking. Unstressed -(y)a/-ra (-ra with ki “who” and či/čə “what” and with most pronouns in Western Gīlakī [see Table 5 for EG pronouns], -ya after nouns ending in vowels, -a after other nouns) marks both direct and indirect objects: WG mən dastán-a həsə́n-a/tə́-ra bugúftəm, EG mu dastán-œ həsə́n-œ /tæ bútəm “I told the story (dir. obj.) to Hasan/to you (ind. obj.).” WG ána ána fadám, EG únæ únæ hádam “I gave it (dir. obj.) to him (ind. obj.).”

Modifiers. Most possessives and adjectives precede the head noun, with a “reverse- eżāfa-like” connector, e.g., noun-noun possessives: WG məhin-ə zakan "Mahīn’s children,” baγ-ə gulan “garden flowers” (not “flower garden” as in Persian); EG (Lang.) xærs-ə kutə “bear cub,” kœrk-ə owlə “chicken pox”; adjectival modification: WG pilla-yə zakan “big children,” surx-ə gul “red flower”; EG (Lang.) særd-ə aw “cold water,” kul-ə caqu “sharp knife.”

Eżāfa. In addition to modifier + head noun, the Persian eżāfa construction with modifiers following the head noun also abound, probably exclusively as borrowings: rúz-ə təvəllud “birthday,” ijazə-yə xuruj “exit permit,” kár-ə šəbanə “night work,” hə́rf-ə muft “worthless talk, nonsense.”

Deixis (words that point to specific items, places, manners). Demonstratives are: (adj.) a “this,” u “that,” ha/hu “this/that very (same)”; (pron.) an “this one,” un “that one,” (plurals: ašan, ušan), han/hun “this/that very (same) one”; uy, uy-danə, uy-ta “that (other) one” (pron., adj.) (< u + i “one”): kitáb-a úyta-ya fadám “I gave the book to that one,” uy-ta xiaban “that other street.” Other deictics are: áya “here,” úya “there,” háya “right here,” húya “right there,” ára/úra “this/that way (direction),” ato/uto “so, such a, this/that way, this/that kind of,” hato/huto “this/that very way, just like this/that, this/that very kind of.” In Māč. hin and hun are not intensives as in the other dialects, but the usual demonstratives, cf., words derived from them: hišon “they,” hi(n)tor “this way,” etc.

Personal pronouns. A special set of possessive pronouns exists preceding the head noun as other possessives do. Generally pronominal suffixes do not occur, but are increasing due to Persian influence: bə gəmán-əm “I guess,” šni šoon/šoón-əš “his going,” xúdəš “himself”; EG: nevištɛ̄nèš sœxt-ə “It is hard to write (lit: its writing),” œmu xówemun š´ne "We are sleepy” (alternate forms with full pronouns probably exist as a more common occurrence). When possessives are used independently, a meaningless šin (EG ši/še ) is required: mi šin “mine,” həsən-ə šin "Hassan’s,” etc. (see Table 5, above).

Reflexives. Gīlakī has two reflexive forms: xu (xo) and xud- + pronominal suffixes(the latter probably borrowed from Persian). Xu only occurs in the third person, with an obligatory pronominal suffix in the plural, i.e., xušan:xo-ra bide “he saw himself,” xu per “his own father,” but pl. xušan-a biden, xušan-ə per. Other persons use the regular personal pronouns, sometimes interchangeably with xud-:tu be ti per bušói "you take after your father,” ti-məra/xudət-ə məræ bə́bər “take it with you.” Xud- (any person) occurs as either a reflexive, Lang. xúdœšə cakun-vakun bugud “she gussied herself up,” or as an emphatic, especially when alone: xúdəm bugúftəm “I said it myself” (xu may also occur here: Lang.: mu xu buššom “I went myself”). Xud- may also be used with suffixes after the head noun or before the head noun with an eżāfa (and no personal suffixes, as in Persian): həsən xúdeš, xúd-ə həsən "Hasan himself.”

Adpositions. Gīlakī probably originally had only postpositions, but prepositions borrowed from Persian are increasing. Postpos. -jə(n), -ja “from,” -re “for,” durun, mian “in, within,” -məra, am(a)ra “with,” bija, virja (EG værje) “near (person),” jir “under,” sər, ru “on,” jor “above”; Lg. muson “like,” vas(t)i (Lg. visin ) “for, for sake of,” etc. Examples are: baγ-ə ja “from the garden,” ti-re “for you,” WG/EG: ti məra, də́st-ə məra “with you, with the hand,” WG áb-ə durun/áb-ə mian “in the water”; mi bija/mi virja “near me,”(EG equivalents: áb-ə mian, mi værje), Māč. mi-ji “from me.” Prepositions are: az/jə “from,” ta “until, up to,” “to,” etc. Postpositions and prepositions often alternate: ti per-jə /az ti per “from your father,” or are used slightly differently: EG áb-ə mian “in the water”; mian-ə dərd-ə sər “in trouble.” Occasionally both occur together: az ún-ə ja “from it.” Postpositions occasionally take the objective form of pronouns: ti-ja or tə́ra-ja, Ga.: tə́rə-ji “from you.”

Verbs. The major differences in grammar between Western and Eastern Gīlakī are found in the verb (see Table 6, Table 6a, and Table 7).

Verb stems. Gīlakī generally retains a clear distinction between present and past roots of verbs, with the same types of connections between the two roots as in most other Iranian languages, e.g., Pres. root + -əst: jav/javəst “chew/chewed”; +-d/-t: man/mand “stay/stayed,” kəf/kəft “fall/fell”; + -e: kəš/kəše “pull/pulled,” (and all causatives, see below); + -a: is/isa “stand/stood”; -n dropped from Pres. root: hin/he “buy/bought”; Pres. root + various changes: šor/šost “wash/washed,” xus/xuft “sleep/slept,” guriz/guroxt “flee/fled” (and many others). Some verbs have alternate past roots formed with different processes: gərd/gərdəst~gəšt “go/went around, look/looked for,” kob/kobəst~koft “pound/pounded.” A certain number of verb roots have been converted to the invariable element of a compound verb: xəndə kud- “to laugh,” amuj da- “to teach” (still a simple verb root in Māč. amuj), etc.

Preverbs. Preverbsfurther expand or specify the root lexically and have proliferated in form and use, especially in Western Gīlakī: də-/di-, du-, fa-/fə-/fi-, fu-, jə-/ji-, ju-, va-/vi-; rare forms: a-, i-, u-, ca-, cu-, ta-. Note the lexical function of preverbs: kəf- “to fall,” va-kəf- “to attack,” də-kəf- “to fall (into),” jə-kəf- “to fall down”; xus- “to sleep,” u-xus- “to attack”; fu-xus- “to attack,” ju-xus- “to hide”; cin- “to pick (fruit), pile up,” u-cin- “to gather,” di-cin- “to pile up,” fu-cin- “to peck”; gift- “to take, catch,” fa-gift- “to take from, buy,” jə-gift- “to wean”; and the latter verb in Lang.: git- “to take, catch,” ha-git- “to take from, buy,” vi-git- “to pick up,” dš-git- “to start (raining, snowing),” va-git- “to escape (from someone’s clutches).” Other contrasts in Lang.: kutane “to pound (generally),” fu-kutane- “to pound body against something,” dš-kutane “to punch,” and gərdəst- “to go around, wander,” va-gərdəst- “to turn around (partially), turn back, return,” ju-gərdəst- “to snap out (of place),” də-gərdəst- “to turn over, turn around (completely).” Eastern Gīlakī has fewer forms: də-, u-, ha-, va-, vi-, rare: ca-, to-. WG fa- is EG ha-. Some Western/Eastern Gīlakī differences: WG va-vurs- “to ask,” va-vin- “to cut,” vs. EG: pors-, bin (but, va-pors- “to investigate”) and WG va-kəft- “to attack” vs. Lang. va-kət- “to fall off one’s feet (from fatigue).” There does not seem to be support for the claim (Rastorgueva, p. 127) that alternate forms of preverbs are determined by the following vowel, e.g., WG fa-: fa-da- “to give,” fa-cukəst- “to climb,” fa-gift- “to get, buy,” fa-kəš- “to pull out”; fi-: fi-caləst- “to wring,” fi-bišt- “to roast”; fu-: fu-rad- “to chase,” fu-duš “to milk,” fu-bost- “to spill (intrans.),” fu-cin- “to peck.”

Negation. Negationis expressed by an obligatorily stressed, prefixed nV´-, which has the same four alternations as bV- (see below): nə́-xayəm “I don’t want,” nú-goftəm “I didn’t say,” ní-dinəm “I don’t see,” n-amo “he/she didn’t come.” Western Gīlakī present and subjunctive fall together in their negative forms: dinə́m “I see,” bídinəm “that I see,” nídinəm “I don’t see, that I not see”; usanə́m “I pick up,” úsanəm “that I pick up,” uṹsanəm “I don’t pick up, that I not pick up”; nV´- always follows preverbs, and is then expressed in two variant forms: (1) preverb + -nV´-, (2) preverb + vowel of preverb repeated, nasalized, and stressed: dunə́cəkə, duṹcəkə “doesn’t stick,” dənə́kəfəm, dəkəfəm “I don’t fall” (latter forms probably best analyzed du-ún-cəkə, də-ə́n-kəfəm). Eastern Gīlakī has more vowel alternates to the negative but, as mentioned, phonemic status of the vowels is not clear: nə́šənəm “I don’t go,” nóxonəm “I don’t want,” nÚdonəstəm “I didn’t know,” nɛˊporsənəm “I don’t ask.” Here the negative form with preverbs is -n-:vítəm, víntəm “I picked up/didn’t pick up,” hádyenəm, hándyenəm “I give/don’t give.”

Non-finite forms. There are two non-finite forms each derived from the present stem (present participle 1, present participle 2) and the past stem (infinitive, past participle). Present participle 1 (present stem + -əndə ): bər-əndə “winner, winning”; present participle 2 (present stem + -an ): xəndə kun-an “laughing.” Infinitive (past stem + -ən or -n after vowels): WG/EG: kəft-ən/kət-ən “to fall,” amo-n/œmš-n, “to come,” butmonosyllabic stems repeat vowel: ze-en “to strike” (EG: zə-ən ), šo-on “to go.” Past participle (bV-/preverb- + past stem + -ə́ or Ø after vowel): (WG/EG) bə-kəft-ə́/bə-kət-ə́ “fallen,” fa-gift-ə́/ha-git-ə́ “taken,” bamo/bemma “come” (= b-amo-Ø/b-emma-Ø ).

Personal endings. There is no distinction in past intransitive and transitive conjugations (as opposed to other NWI; see Table 5, Table 6, and Table 6a). In Western Gīlakī (a) 3rd singular present differs from subjunctive/past; (b) 2nd and 3rd plural are identical; (c) and the final vowel is optional in plural. In Eastern Gīlakī the 3rd person singular subjunctive ending of those present stems that consist of one consonant is -un, -on: bÚ-g-on, bÍ-š-un, bÚ-b-on “that he/she says, goes, becomes,”e.g., Lang. va bazar bəšun “he/she must go to the bazaar.”

Tenses. (a) Present. The present has no marker (Ø) in Western Gīlakī and Gālešī, but a suffixed -(ə)n- in most other dialects (see Diachronics, above): WG , dəvədə́, xayə́; Gāl. gue, də-bəsə, xay; Lāh./Lang. gÚnə, də́vədənə, xánə “says, closes, wants.” The sounds r- and n- generally drop before the -(ə)n- suffix: xor-, (vi-)gir-, dan-, din- > EG xónəm, víginəm, dónəm, dínəm,but WG: xorə́m, girə́m, danə́m, dinə́m “I eat, pick up (take), know, see.”

(b) The tense/aspect marker bV- is used in the formation of the common Gīlakī subjunctive, imperative, past and perfect tenses (see past participles above) and is omitted when a preverb or the negative particle is present. Three Western Gīlakī alternates,, bi-, and bu-,are determined by the vowel of the following syllable and a fourth variant, b-, occurs directly before a vowel, e.g., (subj., past) bídinəm, bidíem “I should see, I saw,” búxurəm, buxúrdəm “I should eat, I ate,” bə́šəm, bušóm “I should go, I went,” b-ávərəm, b-avə́rdəm “I should bring, I brought.” In verbs with no vowel in the present stem (-š- “go,” -g- “say,” -b- “become,” etc.), Western Gīlakī bV- changes according to the vowel of the ending: bə́šəm, bíši, bə́šə, bíšim, bušóm (1st-3rd sg. and 1st pl. subjunctive, 1st sg. past of “go”). Western Gīlakī bV- is stressed in the subjunctive and imperative, and unstressed in the past, and perfect tenses. In Eastern Gīlakī bV- is stressed in all forms and the alternates are somewhat different, e.g., (subj., past) bə́binəm, bə́beəm “I should cut, I cut,” bóxorəm, bóxordəm “I should eat, I ate,” bə́zənəm, bə́zyeəm “I should strike, I struck.”

(c) Imperfect marker is an unstressed, suffixed -i in Western Gīlakī, as in xórdim “I used to eat,” kə́ftim “I used to fall,” danə́stid "You knew,” gúftid "They used to say,”but no marker (Ø) in Eastern Gīlakī, as in xórdəm, kə́təm, donə́stən, gútən “I used to eat, fall, know, say.” Since the imperfect 3rd sg. ending is Ø in both dialects, the 2nd sg. and 3rd sg. fall together in Western Gīlakī, as in xórdi “you, he/she used to eat” (< 2nd sg. xord-i-i, 3rd sg. xord-i-Ø),but remain separate in Eastern Gīlakī, as in xórdi,“you used to eat” (< xord-Ø-i) vs. xórd, “he/she used to eat” (< xord-Ø-Ø). Eastern Gīlakī past and imperfect fall together in their negative forms, e.g., bóxordəm “I ate,” xórdəm “I used to eat,” nóxordəm “I didn’t eat (simple or habitual),” as do all verbs with preverbs (affirmative and negative), e.g., váporsiem/vámporsiem “I investigated/didn’t investigate” (both forms are simple or habitual).

(d) Progressive (all examples mean “I am/was taking”) has three present/past forms in Western Gīlakī: (1) infinitive plus be4 “be in,” as in giftə́n-dərəm/giftə́n-dubum, (2) invariable kə́ra (< proto-Ir. *kā/ăr- “work, doing”) preceding a fully conjugated present or imperfect, as in kəra girə́m/kəra gíftim, or (3) both types combined, as in kəra giftə́n-dərəm/kəra giftə́n-dubum. Eastern Gīlakī forms only parallel Western Gīlakī types (1) and (3) with: (1) the infinitive (minus final -n) plus be4, as in gité-dərəm/gité-dəbum and (3) by an invariable ka inserted between the elements of type (1), as in gite-ká-dərəm/gite-ká-dəbum.

(e) Future. Western Gīlakī future is formed with the conjugated present of xastə́n “to want” plus the infinitive. The process is reversed in Eastern Gīlakī, in which an invariable present form of xastə́n (devoid of personal ending) is followed by the conjugated subjunctive, e.g., WG xayə́m guftə́n “I shall say,” xayí guftə́n "You will say,”etc. versus EG xan bú-gom, xan bégi, etc.

(f) Perfects. Most dialects of Gīlakī have lost the present perfect, merging most functions with the past: mən fərda ta a moqe vagərdə́stəm “I will have returned by this time tomorrow” (but see Dialects below). Other perfect tenses are formed with buon “be1,” as in mən γəblən buguftə́-bum “I had already said it.” Perfect subjunctive forms are buguftə́-bim, buguftə́-bi, buguftə́-be, buguftə́-bim, buguftə́-bid, buguftə́-bid, e.g., núkune bugúftə́-be “I hope he hasn’t said (anything),” mən xayəm ta fərda in moγe bərəsé-bim “I want to have arrived by this time tomorrow.”

(g) Past subjunctive and conditionals. Past subjunctive and conditionals exhibit two completely different situations in Western and Eastern Gīlakī. The past subjunctive in Western Gīlakī has coalesced with the perfect subjunctive, except for an optional 3rd sg. buguftə́-bi, which is interchangeable with buguftə́-be (see above sec. f). The past subjunctive is used in various present and/or past contrary-to-fact (irrealis) senses both in if-clauses (in the protasis) and other situations: agər zudtər bamó-bim, həsən núšoi “If I had come sooner, Hasan wouldn’t have left” (present or past situation), mi dil xasti, mi pəsər kučik bu¦, giləki yad bəgiftí-be~ bəgiftí-bi “I wish (that) when my son was little he had learned (would have learned) Gīlakī” (¦= sustained subordinate clause intonation, implying “when”). In some cases, however, the past subjunctive is interchangeable with the past perfect with no difference in meaning: vássi zudtər buguftú-bum (past perf.) /bugoftí-bim (past. perf./past subj.) “I should have said something sooner.”

Eastern Gīlakī, while it has no tense like the Western Gīlakī past subjunctive, has instead three additional formal tenses, conditional1-3, used for various expressions of conditionality. The first two are formed by the addition of a conditional marker -en- to the past root (hence not the same as the present -ən-, but parallel to forms in Gūrānī, Dimli/Zāzākī and Baluchi, qq.v.), and are distinguished from each other by the presence or absence of the perfective marker, bV- (there is no distinction with preverbs): bóxordɛnæ-bum vs. xórdɛnæ-bum. While these two forms sometimes contrast with each other, they are often interchangeable. The three conditionals are used in the following situations: (1) in either clause of present contrary-to-fact sentences (less commonly): æger mo xassæm, xótènæ-bum “If I wanted to (right now), I would sleep”; æger hæva rošèn-æ bona-bu, mu èrè núbom “If the weather were better (would get better), I wouldn’t be here.” The imperfect tense is most common in both clauses of this type of sentence: æger mo donəstəm, tæ gúttam “If I knew (now), I would tell you.” (2) in either clause of past contrary-to-fact sentences (more commonly): æger mu donəstəm, diruz tæ gÚttɛnæ-bum “If I had known, I would have told you yesterday”; æger tu mš búttɛnæ-bi, i ettefaqdə́nkətɛnæ-bu (or past perfect: də́nkətə-bu) “If you had told me, this would not have happened.” Note the following contrast between imperfect and conditional2 in the apodosis in the following contrary-to-fact situations: š´ge mu donəstəm, ta gÚttam “If I knew, I would tell you (present situation),” æˊge mu donəstəm, ta gUttena bum “If I had known, I would have told you (past situation).” (3) After kaški/ey kaš! “I wish, would that,” conditional2 may be used, but the past perfect is more common (conditional1 may not be used): kaški/ey kaš mi dəvš hœr ruz bóxordɛnæ-bum/bóxordə-bum “I wish I had taken my medicine every day.” (4) Occurring alone in the sense of “should have”: zudtær gÚttènæ-bi "You should have said (something) sooner!” (5) In a past subjunctive usage: be:tær bu tæ búttɛnæ-bum or gÚttɛnæ-bum “It would have been better to tell you/It would have been better had I told you.” The third additional tense, conditional3, is simply the past perfect, minus the perfective marker (bV- ): xórdə-bum. The present author has collected only two examples of this tense, and both were interchangeable with the past perfect, the other two conditional tenses presented in this section and even the imperfect: š´gš bəxástə-bum (protasis), mo xórdə-bum, bóxordə-bum, xórdɛnæ-bum, bóxordɛnæ-bum “If I had wanted (to), I would have eaten it"(four interchangeable forms in the apodosis, the last of which is possible but not common), bəyÉsti zudtær gÚtti-bi, bÚtti-bi, bÚttènæ-bi, gÚttènæ-bi, gÚtti “You should have said (something) sooner” (5 possibilities, “all very acceptable” according the native speaker).

The verbsto be. Gīlakī distinguishes six verbs “to be” for combinations of equation, existence, animateness, humanness, containment, and emphasis (cited here in 3rd sg., affirmative forms; EG forms differ only in be5; it is not known if be6 exists in EG): (1) (neg: níyə ), a general, neutral copula(enclitic in the affirmative), as in həsən duktur-ə “Hasan is a doctor,” EG mi nom irej-I, “my name is Iraj”; (2) íssə (transcribed isə in Rastorgueva), used only in the present affirmative and generally in those places the enclitic copula cannot occur: (a) independently, as in Gīl. íssə? “Is he?” or (b) stressed, as in Gīl. həsən duktur íssə "Hasan is a doctor,” but may also be used in the same places as the copula (no special emphasis), as in a məγåzə ci məγåzə íssə? “What kind of store is that (store)?” (Kerimova, Mamedzade, and Rastorgueva, p. 121); (3) isá (neg: né:sa), location (with human subjects), as in həsən aya isá, EG həsən e isá "Hasan is here”; (4) də́r-ə (neg: dəníyə, diínnə ) “be in” (usually with inanimate subjects, but full distribution is quite complex): čay γurí-ə durun də́r-ə, EG čay γurí-ə miɛn də́r-I “The tea is in the teapot”; (5) nahá (neg: nə́na), EG: hánna (neg: nə́nna), inanimate existence “there is,” hence only occurring in the 3rd person: pəla nahá búxurim? “Is there any rice to eat?”; EG kisey miɛn pul hæˊnnæ? “Is there any money in the bag?” (6) mane occasionally replaces nahá (in dependent forms only): ita cəšmə vásti u γár-ə mian bə́manə~nahá-bi “There must be a spring in that cave”; isa, nahá,and mane still function as “to stand,” “to put,” and “to stay” respectively, but lose their original senses when used as “to be,” as in həsən utaγ-ə xab-ə durun isá "Hasan is in the bedroom” (even when sitting or sleeping). Note that a contrast in verbs may have implications for the noun in both dialects, as in EG pəley miɛn jujš isœ “There is a chick in the rice” vs. pəley miɛn jujš də́r-ə “There is chicken in the rice.” The first sentence indicates that the chicken is alive, whereas the second sentence means that it is cooked chicken.

Statives (a past participle used as an adjective with no tense or action implied; the action is completed and its results are in a fixed state, e.g., “a broken glass,” “the glass is broken”). As is the case in some Iranian languages (cf. Sangesarī, Windfuhr and Azami), Vafsī, and others, a special stative form exists sporadically for certain Gīlakī verb forms, and examples are not easily uncovered, due to the general tendency to avoid the grammatical situations that require statives. These forms consist of the past participle without the addition of bV-. When bV- is present, it forms the past participle as part of the perfect tenses that contrast with the stative forms: Stative (no action, a description of a state): səg xuftə́-bu "The dog was asleep, lying” (Kerimova, Mamedzade, and Rastorgueva, p. 268) vs. past perfect (an action) səg buxuftə́-bu "The dog had gone to sleep, had lain down.” The stative forms, however, are indistinguishable from the perfect tenses either when the verb root takes a preverb (since preverbs remain in all verb forms) or in the negative forms, e.g., vavostə́ bu “it was open” and “he/she had opened (it)” < va-v/bostən “to open.” The stative may also occur in attributive uses, as in EG vapitə čušm “crossed eye (lit: twisted eye).”

Modals. Modals are Gi. xa/xast “to want,” WG tan/tanəst, EG: ton/tonəst “can,” and WG va/vas(t)i,EG: všne~bayæd/bayésti “must,” WG ša/šasti,EG: šane/? “must, should,” Gīl. b/bost “be possible, acceptable (= become),” e.g., WG xástim bə́gəm və́li núguftəm “I wanted to say (something), but I didn’t,” be tu a kára búkuni? “Is it possible (i.e., would you mind?) for you to do this (work, favor)?”; EG mašin všne rošen-æ bon "The car must start (i.e., turn on).” “Can” occurs in a personal sense either with the subjunctive or with a full infinitive (less commonly, frequency depending on dialect): kaški tanəstim bə́šəm/kaški tanəstim šoon “I wish I could go.” The last three modals listed may also be used impersonally with a full infinitive, e.g., WG nə́ša guftən, EG nə́šane gutə́n “it shouldn’t be said”; WG/EG: nəva dəs zeen/zəən “one must not touch it”; šasti šoon “one should go (there).” The past forms of both verbs for “must” given here, WG/EG vas(t)i /bayésti and šasti/??, have essentially lost any sense of tense and are used interchangeably with the present forms. The sense of time is conveyed by the present subjunctive, perfect subjunctive, or past subjunctive of the following verb, e.g., WG mašin nə́na, piyadə va/vasi bíšid "There is no car, you will have to walk”; EG bayæd/bayésti hæˊdyenæ-bom və́li hæˊndam “I should have given it, but I didn’t.”

Change of state. The verbs kudən “to make, do” and bostən, boon “to become” with adjectives form causatives and inchoatives respectively. In both cases an unstressed -(v)à is optionally added between the adjective and the verb. In tenses formed with bV, either -a or bV- is usually eliminated: mən livə́n-a púr-a kúdəm, mən livə́n-a pur bukúdəm “I filled the glass” and livan púr-a bo, livan pur bu“the glass became full” vs. livan pur bu “the glass was full.” The alternate -va occurs after a (and occasionally elsewhere), as in sia-va bo(st) “it turned black,” sia-va-m-bo(st) “It didn’t turn black.” When an -ə vowel is followed by the particle, the former vowel is then lost, as in kar-ə duruzə > kar-ə duruza-bo(st) “It turned into a two-day affair.” This particle may also occur with nouns, as in áb-a bo(st), áb-va-bo(st) “it turned to water, it melted,” zən-o šohər-a bóstid “they became husband and wife.”

Causatives and passives. The causative marker is -an, added to the present stem, but the present is marked with the infix-an- and the past with the infix -ane- : pər/pərəst “to fly” > pəran/pərane “to make fly” and gərdan/gərdane “to make turn.” The passive is formed analytically, as in Persian, but occasionally the suffix -(v)a, which is an additional formant to indicat e a change of state, is added to the past participle, as in buxurdə́ bubo(st), buxurd-á bo(st) “it was eaten.”


There are many subdialects of Gilaki, and, progressing to the east, it gradually blends into Māzandarāni. The intermediate dialects of the area between Tonokābon and Kalārdašt serve as a transition between Gilaki and Māzandarāni. The differences in forms and vocabulary lead to a low mutual intelligibility with either Gīlakī or Māzandarānī, and so these dialects should probably be considered a third separate language group of the Caspian area. Some additional Eastern/Western Gīlakī differences are the following:

Phonology. A medial d is lost in the negatives of two WG verbs: danə´m and nánəm “I know” and “I don’t know” (EG dónəm and núdonəm), and darə́m and nárəm “I have” and “I don’t have” (EG dánəm and nə́danəm), but EG dínəm and báynəm “I see” and “that I see” (WG dinə́m and bídinəm). Langarudi seems to have lost this medial d altogether in the verb di-, as in inə “he or she sees.” Medial g is sometimes lost in eastern Gilaki (e.g., vítəm, bútəm ), though not in western Gilaki (e.g., fa-gíftəm, bugúftəm) and Gāleši (e.g., vi-gítəm, bəgut). An original a: before a nasal is raised to o (e.g., da:- > dónəm “I know” vs. WG danə́m), but only if the nasal is not part of the stem (e.g., EG xa- > xánəm “I want” and dar- > danə́m “I have”).

Grammar. The main areas where eastern Gilaki differs from western Gilaki concern the verbal system (see Table 6, Table 6a, and Table 7): different formation of present and future tenses and imperfect; three eastern Gilaki conditionals correspond to one western Gilaki past subjunctive; different stress patterns for past and present tense; different plural personal suffixes; and two versus three progressives. A unique negative form of tonəstən “can” exists in eastern Gilaki: mányem “I can’t” and mányəssəm “I couldn’t.” Gālešī, and possibly Langarudi, seem to distinguish a present perfect: Gāl. bəmurdi “you died” and bəkət “he fell” vs. bəmurdəy “you have died” and bəkəti “he has fallen.” The western Gilaki preverb fa- corresponds to the eastern Gilaki ha-: WG fa-dám and EG há-dam “I gave.” Even though Āstāna is on the east bank of the Safidrud river only 9 km from Lāhijān, in Āstānaʾi the present tense has no marker and if formed as in western Gilaki, while its other conjugations follow the rules of Lāhijāni. Gāleši, or at least the dialect of Deylamān as described by Maḥmūd Pāyanda Langarudi, is located well within the limits of eastern Gilān, yet its present tense has also no marker. In general, however, Gāleši contains many features of both western and eastern Gilaki.


Since the time depth between southwestern Iranian and northwestern Iranian is greater than that of, for example, English and Swedish within the Germanic languages, Gilaki and Persian differ on almost all grammar points mentioned above. Time depth within western Iranian, however, is not an absolute measure of distance, since northwestern Iranian and south¬western Iranian have coexisted within the same cultural zone for millennia, during which Persian has consistently been by far culturally dominant. All Caspian languages contain many lexical items (e.g., dan- “to know,” xast “to want,” guft- “to say,” tanest “can”) and certain grammatical features (the loss of the conjugation of transitive verbs and the use of ra) that most likely show quite early influence of Persian.

More recently, however, due to both the economic importance of the Caspian and the Gilān’s proximity to Tehran, Gilaki has been under¬going a massive, indelible Persian imprint: heavy influx of vocabulary (e.g., Pers. pəsər, duxtər, damad, negah kudən have replaced the native rey, kor, zama, fəndərəstən), significant syntactic interference (e.g., eżāfa), changes in vowel pronunciation, and even morpheme borrowings. One thus gets the erroneous impression that Gilaki is merely a dialect of Persian. Yet it is a mixed language, and is becoming even more mixed. Virtual one-to-one correspondences between Gilaki and Persian are commonplace, and often unavoidable: Gil. məšγul-ə taayi kudən durust kudə́n-əšåm-u γəzå bid (Rastorgueva, 1971, p. 140) and Pers. mašḡūl e tahīya kardan dorost kardan e šām o ḡaḏā būdand “they were busy providing and making dinner.”

Not all native elements of Gilaki are lost. Gilaki verbs have been particularly resistant: e.g., WG u san/sad “to pick up,” də gan/gad “to throw,” bu-bux/əst “to rot,” va vin/ve “to cut,” məj/məxt “to crawl, wander around,” va məj/məxt “to look for;” Lang. də rgen/e “to hang up,” varjin/rje “to slice, mince,” va viškan/e “to kindle, light,” də var/əst “to pass by,” fu čurusan/e “to hold back tears,” hæ klašt “to scrape” (WG fa-kəlašt), fu rus/æs “to scratch”; Lang. dæ raγen/e “to stuff (esp. food),” gæn/æss “to bump into”; and Gāl. jur/æs “to look for,” pir/æs “to look at.” Other elements have also sporadi¬cally resisted Persianization: e.g., WG čičini, Lang. mæl(i)jə, Āstānaʾī čušnək “sparrow;” WG hæsə, Gāl./Lang. isə “now;” WG sukule, EG tæla “rooster;” Gāl./Lang. hænde “again;” Lang. kæsæni “each other;” xuræm “good;” lako(y) “girl;” rika “boy;” burmə “weeping.” Note also the long list of features, in which the Gilaki verb system differs from the Persian one. In addition to the examples, quoted above, one can find contrasts such as this: Gīl. kə́fš-ə tə́xt-a in γədər bəsavane, de nə́-ša dukudən (Kerimova, Memedzade, and Rastor¬gueva, p. 205) and Pers. taḵt e kafš-rā īn qadr sābid, dīgar namišavad pušid “he or she wore down the soles of his or her shoes so badly that they can no longer be worn.”



Lāhijāni and Āstānaʾi materials and most examples of Rašti are from the author’s own field notes. The following works were also consulted: H. ʿAbbāsi, Šāʿerān e Gilak o šeʿr e gilaki: Taḏkera-ye šāʿerān e gilakisarā, Rašt, 1997.

M.-ʿA. Afrāšta, Šeʿrhā-ye gilaki, tr. M. Pāyanda Langarudi, Rašt, 1954.

M. Bošrā, Ilah jar: Dujah gilakišeʾran jah salana (1344-1356 Š) , Rašt, 1989; Gilaki poems with Pers. translation. A. Christensen, Contributions à la dialectologie iranienne: Vol. I – Dialecte guiläki de Recht, dialectes de Färizänd, de Yaran et de Natanz, avec un supplément contenant quelques textes dans le persan vulgaire de Téhéran, Kongelige Danske Videnskabernes Selskab: Historisk-filologisk Meddelelser 17/2, Copen¬hagen, 1930; in part tr. J. Ḵomāmizāda as Guyeš-e gilaki-ye Rašt, Tehran, 1995.

ʿA. ʿEmād, “Por-nemunatarin pišvandhā-ye afʿāl e sāda-ye deylami,” in M. ʿA. Ṣādeqiān and M. Ḥ. Iskandari, eds., Proceedings of the Fourth Congress of Iranian Studies, 3 vols., Shiraz, 1974-75, I, pp. 322-37; in Persian.

ʿA. K. Golšani, “Gilaki,” ibid., III, pp. 131-52; in Persian. T. Gurgin, “Šeʿr e gilaki wa šāʿerān e gilakisarā,” in E. Eṣlāḥ ʿArabāni, ed., Ketāb e Gilān, 3 vols., Tehran, 1989-95, II, pp. 515-61.

A. A. Kerimova, A. K. Mamedzade, and V. S. Rastor¬gueva, Gilyansko-Russkiĭ slovar’ (Gilaki-Russian dictionary), Moscow, 1980.

M. F. Māčiāni, “Zabān o farhang e Māčiān,” NDAT 16, 1964, pp. 277-96, 451-70; 17, 1965, pp. 109-28, 261-84.

D. N. MacKenzie, The Dialect of Awroman (Hawrāmān-i Luhon): Grammatical Sketch, Texts, and Vocabulary, Kongelige Danske Videnskabernes Selskab: Historisk-filosofiske Skrifter 4/3, Copenhagen, 1966.

A. Marʿaši, Vāža-nāma-ye gūyeš e gilaki, Rašt, 1984.

ʿA.-A. Morādiān Garrusi, Tarānahā-ye rustāʾi e Gilak, Tehran, 1968.

M. Pāyanda Langarudi, Maṯalhā o eṣṭelāḥāt e Gil o Deylam, Tehran, 1973.

Idem, Farhang e Gil o Deylam: Fārsi be gilaki, Tehran, 1987.

L. A. Pireiko, Talyshsko-russkiĭ slovar’ (Ṭāleši-Russian dictionary), Moscow, 1976.

M. Rahmani, “Ethnography of Language Change: An Ethnolinguistic Survey of the Gilaki Language,” Ph.D. diss., University of Oklahoma, 1985.

V. S. Rastorgueva, “Beludzhskiĭ yazyk” (Baluchi language), in Yazyki narodov SSSR, 5 vols., Leningrad, 1966-83, I, pp. 323-41.

Idem, Gilyanskiĭ yazyk (Gilaki language), Moscow, 1971.

V. S. Rastorgueva and D. I. Edel’man, “Prikaspiĭskie yazyki” (Caspian languages), in V. S. Rastorgueva, et al., eds, Osnovy iranskogo yazykoznaniya (Foundation of Iranian linguistisc), 3 vols., Moscow, 1979-82, III, pp. 447-554.

J. Sartippur, Vižegihā-ye dasturi o farhang e vāžahā-ye gilaki, Rašt, 1990. M. Sotuda, Farhang e gilaki, Tehran, 1953a.

Idem, “Tarānahā-ye Gilān,” FIZ 1, 1953b, pp. 41-56, 271-86.

Idem, “Amṯāl e gilaki,” ibid., pp. 363-74.

D. Stilo, “The Tati Language Group in the Sociolinguistic Context of Northwestern Iran and Transcaucasia,” Iranian Studies 14, 1981, pp. 137-87.

Idem, “Ambipositions as an Areal Response: The Case-Study of the Iranian Zone,” in E. Bashir, et al., eds., Select Papers from South Asian Languages Analysis 7: Roundtable Conference Held in Ann Arbor, Mich., May 17–19, 1985, Bloomington, Ind., 1987.

M. A. Ṭabāṭabāʾi, “Guyeš e gilaki,” in E. Eṣlāḥ ʿArabāni, ed., Ketāb e Gilān, 3 vols., Tehran, 1989-95, II, pp. 505-14.

Y. Ṯamara, “Taḥlil e sāḵtāri ye feʿl dar guyeš e gilaki ye Kalārdašt,” MDAM 24, 1988, pp. 169-87.

G. Windfuhr, “A Sketch on Persian and Parthian, Kurds and Medes,” in Monumentum H. S. Nyberg, 2 vols., Acta Iranica 4-5, Liège and Tehran, 1975, II, pp. 457-72.

G. Windfuhr and C. A. Azami, A Dictionary of Sangesari with a Grammatical Outline, Tehran, 1972.

V. I. Zavyalova et al., “Gilyanskiĭ i Mazanderanskiĭ yazyki” (The languages of Gilān and Māzandarān), in Novye svedeniya po fonetike iranskikh yazykov, Trudy Instituta Yazkyoznanija 6, 1956, pp. 92-112.

(Donald Stilo)

Originally Published: December 15, 2001

Last Updated: February 9, 2012

This article is available in print.
Vol. X, Fasc. 6, pp. 660-668