Redundancy

id	Language	Clausal agreement or cross-reference	Phrasal agreement	Plural concord in noun phrases containing a numeral	Tense copying
3	Bantawa	Arguments in Bantawa are cross-referenced on the predicate in accusative alignment (Doornenbal 2009: 12): their person and number properties are expressed by means of referential markers, that are hierarchical in nature, on transitive predicates (Doornenbal 2009: 144ff.). Referential markers are obligatory, but independent arguments are not, as is apparent from example (1). This means that Bantawa shows cross-reference. Doornenbal (2009: 376) (1) mɨ-poy-a 3PL-grow-PST ‘There they grew up.’	Bantawa does not have agreement in the phrasal domain, as illustrated by example (2). Doornenbal (2009: 296) (2) toppo chapkami big writer ‘big writer’	Bantawa shows optional plural concord. Plural is marked on nouns by means of a suffix -ci. This is common on human nouns, preferred on non-human animates, and optional on inanimates. Hence, a noun without this marker is not necessarily singular; for concepts and objects, number may simply be unspecified (Doornenbal 2009: 72). The numeral system in Bantawa is highly defective (Doornenbal 2009: 112), but independent numerals at least up to three. Furthermore, there are numeral prefixes, to which the so-called ‘counter’ -ka- is optionally attached, which functions as a linking element. A classifier is obligatory with independent and prefixed numerals, distinguishing between humans and various categories of non-humans (Doornenbal 2009: 113). After quantifiers, -ka- and classifiers are optional (Doornenbal 2009: 114), as shown in (4) below. With numerals and with quantifiers, the plural suffix can be expressed, as illustrated in (3) and (4). Since the expression of plurality in general is optional, plural concord is optional as well, as illustrated by example (5) in which the plural suffix is dropped even though the noun involves humans. Doornenbal (2009: 215) (3) hwa-tet gadi-ci two-CLF car-PL ‘two cars’ Doornenbal (2009: 115) (4) baddhe himsale kutiwa-ci many crazy dog-PL ‘many crazy dogs’ Doornenbal (2009: 112) (5) nəu-ka-paŋ mɨna nine-COUNT-CLF man ‘nine persons’ M. Doornenbal (personal communication, September 18, 2013) stresses that a plural suffix is only used to stress plurality. If the numeral is expressed, the expression of plurality is no longer necessary, so that plural concord is allowed, but presumably marginal.	There is no tense copying in Bantawa. Embedded clauses are obligatorily nominalised by -ʔo (Doornenbal 2009: 202). Even though nominalisations can express tense, they are not backshifted under the influence of the tense of the main clauses, as illustrated in example (6). Doornenbal (2009: 202) (6) ram-ʔa sarima-ʔa dhir-u-ʔo bakhra R.-ERG disease-ERG find-3.U-NMLZ goat syam-ʔeda in-u-ʔo iŋka kha-Ø-t. S.-LOC sell-3.U-NMLZ 1SG see-PST-1SG ‘I saw Ram sell a sick goat to Syam.’
5	Tamil	Tamil has cross-reference, as shown in (1). Finite verbs are obligatorily marked for tense, person and number (e.g. Steever 2005: 61). Expressing the argument independently is optional, as long as the referent is clear from preceding discourse (Lehmann 1989: 173). Asher (1982: 53) (1) (raaju) va-nt-aaru (R.) come-PST-3SG.HON ‘Raju came.’	There is no phrasal agreement in Tamil. There is a semantic noun classification system and nouns are marked for number, but such information is not copied to modifying units in the phrase (cf. e.g. Asher 1982: 188). An example demonstrating this is given in (2). Andronov (2004: 153) - literary (2) periya kaɳ-kaɭ large eye-PL ‘large eyes.’	Plural marking is always optional in Tamil. There is a plural suffix that one can use to individuate the noun, but this is not necessary (S. Sundaresan, personal communication, December 4, 2013). Plurality can still be expressed in combination with a numeral (Schiffman 1999: 28), as illustrated in example (3). Schiffman (1999: 28) (3) reɳɖu manevin-gaɭ=aa? two wive-PL=Q ‘Two wives?’	Tamil does not have tense copying, as shown in example (4). The present tense of the original utterance is retained when the clause is embedded and the pronoun is shifted. Lehmann (1989: 374), adapted to colloquial speech by S. Sundaresan (4) avan niiŋ-gal uur-ukku poo-r-iŋ-gal-unnu so-nn-aan he 2-PL town-DAT go-PRS-2-PL-COMP say-PST-3SG.M ‘He said that you would go out of town.’ S. Sundaresan (personal communication, December 4, 2013) states that a past tense in a sentence like (4) is also possible and results in ambiguity: one reading is that the second person has left already by the time of speaking, a second reading holds that the ‘you’ is leaving at the moment referred to in the main clause. Apparently then, backshifting of the embedded tense is semantically motivated – there is no obligatory morphosyntactic sequence-of-tense rule.
6	Bininj Gun-Wok	In Bininj, person and sometimes number of arguments are obligatorily marked on the verb by means of pronominal prefixes (Evans 2003: 318). It is possible but not obligatory to express the argument independently additionally. This happens in (1), where ‘flies’ is expressed lexically and its person feature by a verbal prefix. Evans (2003: 417) (1) bod ga-di fly 3-stand.NPST ‘There are flies there.’ Pronominal prefixes can also express the object argument, leading to cross-reference, for example in (2). The same is true for example (3), in which the noun is incorporated. In intransitive clauses, noun incorporation leads to a generic interpretation of the incorporated noun, but this is not the case in transitive clauses (Evans 2003: 330). Evans (2003: 330) (2) barri-ngune-ng gun-ganj 3PL>3.PST-eat-PST.PFV IV-meat ‘They ate the meat.’ (3) barri-ganj-ngune-ng 3PL>3.PST-meat-eat-PST.PFV ‘They ate the meat.’	In the phrasal domain, there is multiple marking of semantic class in Bininj Gun-Wok. Most Bininj Gun-Wok dialects have 4 noun classes (Evans 2003: 181ff.), the assignment of which is based on semantic principles: roughly speaking, class I contains males, II contains females, III contains plants and IV contains body parts and abstract entities. A noun’s class is in most cases marked explicitly on the noun by means of prefixes . Furthermore, modifiers of nouns obligatory receive class prefixes (Evans 2003: 181ff.), as in example (4). Evans (2003: 182) (4) a. na-rangem na-mak I-boy I-good ‘good boy’ b. ngal-kohbanj ngal-mak II-old_woman II-good ‘good old woman’ c. man-me man-mak III-food III-good ‘good food’ d. kun-wardde kun-mak IV-rock IV-good ‘good rock’ Hence, there is obligatory concord with respect to semantic class in Bininj Gun-Wok.	Plurality is hardly ever marked in Bininj Gun-Wok. Most nouns have no plural form at all; if they are plural, this becomes apparent from pronominal prefixes on the verb only (Evans 2003: 168). The few nouns that are marked for plurality, mostly denoting humans, get a reduplication prefix, e.g. daluk ‘woman’, daluh-daluk ~ dah-daluk ‘women’ (Evans 2003: 169). Numerals are infrequent in Bininj Gun-Wok (Evans 2003: 129), but there are quantifying elements, e.g. -wern ‘many’. Reduplication prefixes do not appear in combination with such suffixes, suggesting that plurality cannot be marked redundantly. Consider for instance example (5), where bininj does not get its reduplicated form binih-bininj. Evans (2003: 215) (5) kakkawar kaben-ma-ng birri-wern bininj messenger 3>3PL-bring-NPST 3PL-many person ‘The messenger will bring many people.’ In agreement with this, N. Evans (personal communication, September 20, 2012) states that plurality, if marked at all, is shown on the verbal complex, not in the NP. Number marking by means of reduplication is so marginal that it is negligible – it is unlikely that it will ever occur in combination with a numeral. Therefore, I consider Bininj Gun-Wok to be a language without plural concord.	It might seem doubtful whether subordination exists in Bininj Gun-Wok (cf. Evans 2003: 628ff.). In Bininj Gun-Wok, clauses are juxtaposed so that a bi-clausal sentence results, without morphosyntactic embedding, as illustrated in example (6). Evans (2003: 629) (6) ka-bengka-n ka-marnbu-n 3-know-NPST 3>3-make-NPST ‘He knows how to make them (=boomerangs).’ Evans (2003: 633) states, however, that sentences like (6) are formally different from ‘normally chained verbs’, since the two clauses form one intonational phrase together, and since no unit can stand in between the verbs, whereas this would be possible in a true bi-clausal sentence. Hence, this is not a ‘normal’ bi-clausal sentence, but a hierarchically organised one. Another reason to let these bi-clausal sentences qualify as cases of subordination is illustrated in example (7). Evans (2003: 634) (7) na-bene maih a-na-ng I-DEM bird 1>3-see-PST.PFV ga-m-golu-rr-en gaddum-be djohboi 3-hither-descend-REFL-NPST up-ABL poor_thing I’ve seen those birds coming down (to the waterhole) from higher up, dear little things. ‘Descending’ has a present tense, while the ‘descending’ must have occurred in the past, at the moment of perceiving it. The tense is hence a relative tense, interpretable through the deictic centre of ‘seeing’. This is a reason to argue that the verbs are syntactically linked, rather than forming two independent predicates. Note that this is not a case of tense copying, as in that case the past tense of ‘seeing’ should have been copied to ‘descending’. I will follow Evans in his judgment that (6) and (7) are cases of subordination. Since tense in such clauses is not backshifted, there is no tense copying in Bininj Gun-Wok.
7	Chukchi	Person and number are marked on the verb in Chukchi. Arguments may but need not be expressed independently as well (Dunn 1999: 80): they are only overt when they have focus (Dunn 1999: 350). If an argument is expressed both independently and as a verbal marker, as in (1), there is cross-reference. Dunn (1999: 345) (1) qənwer ʔinə piri-nin finally wolf.ABS.3SG take-3SG>3SG ‘Finally, he caught a wolf.’	Agreement in Chukchi between nouns and modifiers is not possible, since there is no attributive modification of nouns. Dunn does gloss and translate some elements as attributive adjectives, e.g. example (2). Dunn (1999: 165) (2) … ŋenku nəlɣi-n-ə-teŋ-qinet ŋewəcqet-ti wa-rkət there INTS-ADJ-ə-good-3PL woman-ABS.3PL be-PROG.3PL ‘… there are really good women there.’ However, the modifying element teŋ ‘good’ is inflected by –qinet, the usual verbal suffix for third person plural. The prefix n- is also used on verbs to mark habitual aspect. Hence, we can see the modifier in (2) as a relative clause containing a verb with regular verbal inflection, literally translating as “women that are really good”. This analysis involves no agreement or concord between a noun and its modifier, but between a predicate and its argument. Thus, there is no attributive modification in Chukchi and this feature does not apply.	So-called common nouns in Chukchi are usually unmarked for number, in which case they denote a non-plural entity (Dunn 1999: 63), but sometimes they do receive an explicit singular or plural marker. Consequently, a noun stem can have three different forms, e.g. mane-t ‘money-PL, several coins’, mane-man ‘money-RDP, money in general’, mane-lɣ-ən ‘money-SG-ə3SG, one coin’ (Dunn 1999: 64). On higher animate nouns, plural marking is obligatory (1999: 64). Plurality can be expressed on nouns modified by a quantifier or numeral, as illustrated in (3), resulting in plural concord. Dunn (1999: 131) (3) orw-ət əməlʔo … sled-əABS.3PL all.ABS.3 ‘all the sleds’ However, both elements can be seen as independent constituents (cf. Section 4.2.1), so that this is not genuine modification, but rather apposition, translating as ‘The sleds, all of them, …’. Therefore, this does not qualify as numeral concord – the feature does not apply.	There is no syntactic embedding in Chukchi (Dunn 1999: 84) and therefore, there cannot be a sequence of tense rule – this feature does not apply. Dunn (1999: 85) does mention a tendency for multi-clausal sentences to have the same TAM marking in the different clauses, but this is a narrative preference rather than a grammatical rule.
8	Dutch	In Dutch, person and number properties of the subject are obligatorily marked on the verb. Arguments have to be expressed independently as well; pro-drop is highly restricted, cf. example (1). Therefore, Dutch is classified as a language with clausal agreement. (1) *(hij) zoen-t krzysztof 3SG kiss-3SG K. ‘He is kissing Krzysztof.’	Attributive modifiers are obligatorily marked for the number, gender and definiteness of the noun they modify. This is true for at least articles, adjectives, demonstratives and relative pronouns, as apparent from example (2). (2) a. het stoer-e zeepaardje, dat … DEF.SG.N tough-DEF.SG.N sea_horse REL.SG.N ‘The tough sea horse, that…’ b. de stoer-e zeekoe, die … DEF.COMM tough-DEF.SG.COMM manatee REL.COMM ‘The tough manatee, that…’	Plurality is expressed on nouns by means of a suffix -en or -s. Plural marking is obligatory on nouns modified by numerals, e.g. vijf olifant-en ‘five elephant-PL’. For one group of nouns denoting measures, e.g. measures of time, length or weight, plural concord is optional, e.g. twee jaar-Ø ‘two year-SG’.	Tense is obligatorily marked on verbs by means of suffixes or stem alternations. If a verb is embedded under a past tense verb, it is backshifted, cf. example (3). If the embedded proposition is applicable in the present, backshifting is optional (cf. Leufkens 2013b). (3) mare zag dat het goed was M. see.PST COMP it good be.PST ‘Mare saw that it was good.’
9	Egyptian Arabic	There is cross-reference in Egyptian Arabic. Predicates are inflected for person, number and gender of the subject argument (Gary & Gamal-Eldin 1981: 60). This is also the case when an independent argument is expressed, resulting in the double expression of reference, for instance in (1). The independent argument is optional. Gary & Gamal-Eldin (1981: 60) (1) (s[ā]mi) giri-ø (S.) run.PST.PFV-3SG.M.A ‘Sāmi ran.’	There is agreement in number, gender and definiteness between nouns and modifying adjectives, as illustrated in example (2). Gary & Gamal-Eldin (1981: 107) (2) a. walad [ʃ]aatʕir-ø boy smart-SG.M ‘a smart boy’ b. ʔil-ʔawlaad ʔiʃ-ʃa[tʕ]r-iin DEF-boy.PL DEF-smart-PL.M ‘the smart boys’ c. ʔil-bint ʔiʃ-ʃa[tʕ]r-a DEF-girl DEF-smart-SG.F ‘the smart girl’	Marking of plurality on nouns is obligatory after numerals 2 to 10, but not on numerals 11 or higher, as demonstrated in example (3) (Gary & Gamal-Eldin 1981: 111). Hence, in numeral phrases with numerals ‘two’ up to ‘ten’, there is plural concord. Gary & Gamal-Eldin (1981: 111) (3) a. talat kutub three book.PL ‘three books’ b. ħidaaʃar kitaab eleven book.SG ‘eleven books’	There is no tense copying in Egyptian Arabic (M. Hegazy, personal communication, July 22, 2013), as illustrated by example (4). M. Hegazy (personal communication, July 22, 2013) (4) ʔaal-ø inn-u ha-yi-ʕmil-u bukra say.PST.PFV-3SG.M.A that - 3SG.M FUT-3SG.M.A-do.IPFV-3SG.M.U tomorrow ‘He said that he would (lit.: “will”) do it tomorrow.’
10	Fongbe	There is no predicate marking of any property of arguments in Fongbe (e.g. Lefebvre & Brousseau 2002: 240ff.), which means that arguments can only expressed by independent NPs. Hence, there cannot be cross-reference or agreement; this feature does not apply.	There is no agreement or concord in the phrasal domain; no features are copied from nouns to attributive modifiers. This is illustrated in example (1). Lefebvre & Brousseau (2002: 51) (1) ɖìɖè ɖàgbè kɔ̀kú tɔ̀n ɔ́ lɛ́ sketch good K. GEN DET PL ‘Koku’s good sketches.’	Bare nouns are unspecified for number in Fongbe; they can refer to single and plural entities. There is a marker lɛ́ that is glossed as plural by Lefebvre & Brousseau (2002), and is argued by different authors (cf. Bobyleva 2013: 219-262 for a complete analysis of this type of determiners both in Gungbe and in related creoles) to trigger a definite reading as well. In combination with a numeral, this marker has to be used, thus creating plural concord, as shown in (2). Lefebvre & Brousseau (2002: 54) (2) ɖìɖè kɔ̀kú tɔ̀n wɛ̀ lɛ́ sketch K. GEN two PL ‘Koku’s two sketches.’	In Fongbe, temporal reference is expressed by means of phonologically independent markers (Lefebvre & Brousseau 2002: 85). Aboh (2004) argues that in Gungbe, which is closely related to Fongbe in this respect, these markers distinguish future tense, expressed by explicit markers, from non-future tense, signalled by the absence of a visible TMA marker in combination with the TMA semantics of the verb. Thus, a zero-marker in combination with the lexical aspect of the verb together determine the temporal semantics of the clause, i.e. whether this is factitive (with dynamic verbs) or present-state (with state verbs), as shown by example (3). Gungbe - Enoch Aboh (personal communication, July 9, 2014) (3) a. ùn Ø yi 1SG NFUT go ‘I went’ b. ùn ná yi 1SG FUT go ‘I will go.’ Thus, tense marking is most visible in case of future tense, which means that to test whether there is copying of tense marking from a main to a complement clause it is best to find an example with a future tense in the main clause. Example (4) is such an example and shows that there is no sequence of tenses in Gungbe and Fongbe. Gungbe – E. Aboh (personal communication, July 10, 2014) (4) jó kèkè lɔ́ dó, é má nyín mɔ́n leave bike DET at if NEG COP that_way súrù ná ɖɔ̀ mí Ø hɛ̀n kɛ̀kɛ̀ émìtɔ̀n glé S. FUT say we NFUT hold bike his break ‘Don’t play with the bike, otherwise Suru will say that we caused the bike to break down.’
11	Georgian	Person and number of arguments are expressed by pronominal affixes on verbs; at least one such affix is obligatory (Hewitt 1995: 128). Since arguments can be expressed independently as well (Vamling 1989: 16), there is cross-reference in Georgian, e.g. in example (1). Vamling (1989: 19) (1) is xat’-av-s surat-s NOM.3SG paint-THEMSUF-PRS.3 picture-DAT ‘He paints a picture.’	There is agreement in case between nouns and modifying adjectives ending in a consonant (Vamling 1989: 16, Hewitt 1995: 45), as demonstrated in (2). Vamling (1989: 16) (2) a. did-i kalak-i big-NOM city-NOM ‘big city’ b. did-ma kalak-ma big-ERG city-ERG ‘big city’ When the attributive adjective does not end in a consonant but in a vowel, there is no agreement with the head noun, cf. (3). Vamling (1989: 16) (3) a. p’at’ara kalak-i small city-NOM ‘small city’ b. p’at’ara kalak-ma small city-ERG ‘small city’	Georgian nouns can be marked for plurality by means of the suffix -eb (cf. Vamling 1989: 16), but remain singular after all numerals and quantifiers (Hewitt 1995: 55), as illustrated in (4). Hewitt (1995: 55). (4) or-i kal-i hundred_two-NOM woman-NOM ‘hundred two women’ Hence, there is no plural concord in Georgian.	When quoting speech, a direct speech construction is preferred in Georgian. However, a ‘semi-indirect’ as well as a fully indirect speech construction exist (Hewitt 1987: 215). In the semi-indirect construction in (5), the embedded clause features a complementiser and shifting of pronouns, but no tense shift. Hewitt (1987: 215) (5) tkv-a, rom ga-a-k’et-eb-s say-AOR.3 COMP PFV-NV-do-THEMSUF-PRS.3 ‘He said that he would (lit.: will) do it.’ In the fully indirect speech construction, tense is shifted, as in example (6) in which the embedded tense is inflected according to the so-called conditional screeve (Hewitt 1995: 238), ‘screeve’ being the term used by Kartvelologists for particular combinations of tense, mood and aspect, e.g. ‘present indicative’ (Hewitt 1995: 122). Hewitt (1987: 216) (6) tkv-a, rom ga-a-k’et-eb-d-a say-AOR.3 COMP PFV-NV-do-THEMSUF-COND.IPFV-AOR.3 ‘He said that he would do it.’
12	Huallaga Quechua	Number and person of the subject are marked on the predicate by means of suffixes (Weber 1989: 10). The independent expression of arguments is not obligatory, as can be derived from example (1). Weber (1989: 11). (1) hwan-ta kuchi-ta rantiku-sha J.-ACC pig-ACC sell-PFV.3 ‘He sold a/the pig to John.’	There is no phrasal agreement between nouns and their modifiers (cf. Weber 1989: 36), as shown in (2). Weber (1989: 17) (2) chay ishkay hatun wasi-kuna DEM two big house-PL ‘Those two big houses’	There is plural concord in Huallaga Quechua, as shown in (2), repeated here as (3). I have not been able to find out whether this is optional or obligatory. Weber (1989: 17) (3) chay ishkay hatun wasi-kuna DEM two big house-PL ‘Those two big houses’	In Huallaga Quechua, embedded clauses are nominalisations on which tense cannot be expressed, as in (4). Therefore, this feature does not apply. Weber (1989: 21) (4) rura-na-:-paq ni-sha do-NMLZ-1.POSS-PURP say-PFV.3 ‘He said that I should do it.’
13	Japanese	No properties of referents are marked on the Japanese predicate (Hinds 1986: 323). Therefore, agreement and cross-reference in the clausal domain are impossible in Japanese, even if an argument is explicitly expressed, which is quite rare (Hinds 1986: 74). This feature does not apply.	No properties of the referential head are expressed on its modifiers, hence, there is no noun-attributive agreement or concord in Japanese (Hinds 1986: 82, 347). This is illustrated in example (1). Hinds (1986: 346) (1) kirei na heya clean LINK room ‘a clean room’	Plural inflection on nouns is optional in Japanese (Hinds 1986: 223). Humans and higher animals are optionally marked for plural by the suffix -tachi. Some other nouns are pluralised by means of reduplication (Hinds 1986: 225). Two native speaker informants (according to R. van Munster, personal communication, September 22, 2013) claim that -tachi can be used in combination with numerals, especially with higher ones, e.g. futa-ri no hito-tachi ‘two-CLF LINK person-PL’ is acceptable. This means that there is optional plural concord in Japanese.	Plural inflection on nouns is optional in Japanese (Hinds 1986: 223). Humans and higher animals are optionally marked for plural by the suffix -tachi. Some other nouns are pluralised by means of reduplication (Hinds 1986: 225). Two native speaker informants (according to R. van Munster, personal communication, September 22, 2013) claim that -tachi can be used in combination with numerals, especially with higher ones, e.g. futa-ri no hito-tachi ‘two-CLF LINK person-PL’ is acceptable. This means that there is optional plural concord in Japanese.