2 Polarity (anti-agreement) and impoverishment

Early Arab grammarians have described gender polarity in the numeral phrase using the term muxaalafa, which means ‘opposite’ and is the equivalent of polarity. With negation, the word polarity (or negative polarity) is used to refer to a dependency between a certain word and the negative marker, such that the word can occur in a domain containing negation. This dependency is also related to what is known in the literature as negative agreement or concord (cf. Zeijlstra Reference Zeijlstra2004 regarding negative concord). However, ‘polarity’ in the term ‘gender polarity’ refers to the opposite, i.e., to anti-agreement. For example, the feminine ending of the numeral occurs only when the enumerated noun is masculine.

The first step in formalizing an analysis that captures the anti-agreement or gender polarity is to determine whether the feminine ending of the numeral is added when the enumerated noun is masculine or whether it is deleted when the enumerated noun is feminine. These are the two logical possibilities available, and I argue in favour of the second. I propose that the numerals 3–10 are underlyingly feminine (the default). Evidence for this comes from the pronunciation of these numerals when counting, as shown in (6), and from numeral phrases that have a partitive interpretation, as in (7). The fact that these numerals are feminine when used without an enumerated noun, as in (6), suggests that the feminine morpheme they have is not added as a result of syntactic agreement. I therefore assume that the numeral is specified with fem before it enters the syntactic derivation.

1. (6) θalaaθ-at,  ʔarbaʕ-at,  χams-at,  sitt-at,  sabʕ-at,  θamaaniy-at,  tisʕ-at,  ʕašr-at

   3-f.sg,  4-f.sg,  5-f.sg,  6-f.sg,  7-f.sg,  8-f.sg,  9-f.sg,  10-f.sg

   ‘Three, four, five, six, seven, eight, nine, ten’

2. (7)

   1. a. ʕašar-aat  al-mudarrɪs-i:n

      ten-f.pl  the-teacher-m.pl

      ‘Tens of male teachers’

   2. b. mɪʔ-aat  al-mudarrɪs-i:nFootnote ²

      hundred-f.pl  the-teacher-m.pl

      ‘Hundreds of male teachers’

Similarly, for number polarity, I contend that the enumerated noun is plural by default. There is distributional and cross-linguistic evidence for this. First, recall that the enumerated nouns are always plural when the numeral follows the noun, as in (8).

1. (8)

   1. a. al-tˁullaab-u  al-χams

      the-student.m.pl-nom  the-five

      ‘The five male students’

   2. b. al-tˁullaab-u  al-ʕɪšr-uun

      the-student.m.pl-nom  the-ten-m.pl

      ‘The twenty male students’

   3. c. al-tˁullaab-u  al-χams-a  ʕašar-a

      the-student.m.pl-nom  five-acc  ten-acc

      ‘The twenty male students’

Second, there is cross-linguistic evidence from several languages representing different language families that the enumerated noun is plural (English, French, Russian and other Slavic languages, Dravidian languages, and Yoruba, among others, as discussed by Corbett Reference Corbett1978a, Reference Corbett1978b).

Therefore, the enumerated noun enters the syntactic derivation fully specified for the number feature: singular, dual or plural. This is consistent with the fact that the enumerated noun in numeral phrases with numerals three and above has a plural referent. It is also consistent with the fact that the enumerated noun does not even co-occur with the dual numerals in (9) below. In other words, the number feature of the enumerated noun is spelled out as the dual suffix –aan. Put differently, the fact that a noun specified for a dual number feature is the only specification of quantity in the examples below, and that no dual numeral is used, suggests that the number feature of the enumerated noun is essential.

1. (9)

   1. a. (*ʔɪθn-aan)  tˁaalɪb-aan

      (*two-m)  student-m.du

      ‘Two male students’

   2. b. (*ʔɪθn-taan)  tˁaalɪb-taan

      (*two-f)  student-f.du

      ‘Two female students’

Further support for the second approach comes from the theoretical economy that deletion exhibits. As Sadiqi (Reference Sadiqi2003: 103) points out, under the generative approach, deriving the masculine from the feminine is theoretically superior, since it is “less costly” and is explanatorily adequate given that the outcome of deletion as opposed to addition is predictable (cf. Chomsky Reference Chomsky1965; Lightfoot Reference Lightfoot1979). Moreover, in analysing gender in Arabic from a sociolinguistic perspective, Sadiqi (Reference Sadiqi2006) claims that the feminine is the basic form in Arabic and the masculine was derived by reducing the feminine.

I adopt a distributed morphology (DM) view on this topic following proposals by Embick and Noyer (Reference Embick and Noyer2001, Reference Embick, Noyer, Ramchand and Reiss2007) on impoverishment in the morphological structure at PF. In DM, impoverishment involves the deletion of certain morphosyntactic features from morphemes in certain morphological environments. For example, see Embick and Noyer (Reference Embick, Noyer, Ramchand and Reiss2007) on impoverishment in the Standard Arabic case system. This deletion process applies at PF, and guarantees blocking of Vocabulary Insertion for the deleted morphemes.

The idea that anti-agreement can be explained by impoverishment is not new. Fuß (Reference Fuß2005: 51) suggests that ‘impoverishment rules may perhaps also be used to account for a number of apparently syntactic anti-agreement effects. For example, the absence of verbal agreement in the context of wh-subject extraction, which can be observed in a number of Northern Italian dialects (cf., e.g., Brandi and Cordin Reference Brandi, Cordin, Jaeggli and Safir1989) may result from an impoverishment rule that deletes agreement features in the presence of a wh-feature, leading to the insertion of the default 3sg ending’. Since gender and number polarity is basically an anti-agreement phenomenon, the deletion of the feminine and the plural morphemes leads to the insertion of the default, i.e., the masculine and the singular respectively. Figure 1 shows the order of operations on the PF branch, as articulated by Embick and Noyer (Reference Embick and Noyer2001). Impoverishment takes place immediately before vocabulary insertion.

Figure 1: Order of operations on the PF branch

To reiterate, this article is concerned with numeral phrases in their canonical form, where the numeral precedes the enumerated noun, which exhibit number and gender polarity. The syntactic structure of these numeral phrases is based on three things. First, the numeral assigns case to the enumerated noun. The numerals 3–10, ‘hundred’, and ‘thousand’ all assign genitive case, as in (10a) and (10b), while other numerals above 10 assign accusative case, as in (10c). While there is no direct relation between these case facts and polarity, it is worth mentioning them here to show that the numeral is the head of the numeral phrase.

1. (10)

   1. a. χams-u  tˁaalɪb-aat-ɪn

      five-nom  student-f.pl-gen

      ‘Five female students’

   2. b. mɪʔat-u / ʔalf-u  mu'allɪm-ɪn

      hundred-nom / thousand-nom  teacher-gen

      ‘One hundred / thousand teachers’

   3. c. χams-a  ʕašr-at-a  tˁaalɪb-at-an

      five-acc  ten-f-acc  student-f.sg-acc.indef

      ‘Fifteen female students’

Second, the numeral enters the derivation specified for gender. Third, the enumerated noun enters the derivation specified for number. The syntactic structure of these numeral phrases has the cardinal numeral as the head of the phrase CardP and the enumerated noun as its complement. This structure follows Shlonsky's (Reference Shlonsky2004) syntactic analysis of the Arabic numeral phrase:

1. (11)

For the purposes of this article, the syntactic terminals of the numeral and the enumerated noun are specified for gender and number. Taking the numerals 3–10 as an example, the numeral is Feminine [f] and the enumerated noun is Plural [pl]. The representation is as follows:

1. (12)

The impoverishment rules apply at this post-syntactic level, where the syntactic terminals of the numeral and the enumerated noun are specified for gender and number features, as will be explained in sections 3 and 4, respectively.

3 Gender polarity

In line with our conception of impoverishment, the numeral in (13a) loses its feminine morpheme when followed by a feminine enumerated noun, as stated in rule (14).

1. (13)

   1. a. χams  tˁaalɪb-aat-ɪn

      five  student-f.pl-gen

      ‘Five female students’

   2. b. χams-at  tˁullaab-ɪn

      five-f  student.m.pl-gen

      ‘Five male students’

2. (14) Digit 1 fem Deletion:

   [f] → $\emptyset$ / [numeral] + ___ + [enumerated noun] + [f]

Table 1 gives the derivations for examples (1)a and (1)b.

Table 1: Derivations for examples (1)a, b

Interestingly, the deletion of the feminine morpheme carried by single-digit numerals is also attested in double-digit numerals, specifically in numeral phrases with the numbers 13–19, as in (15a–b), and numerals like 23–29, 33–39, etc., as in (15c–d). The trigger for this deletion, as with the single-digit numerals, is the presence of the feminine morpheme on the enumerated noun. Therefore, I revise the rule in (14) as in (16) below.

1. (15)

   1. a. χams-at-a  ʕašar-a  tˁaalɪb-an

      five-f-acc  ten-acc  student-acc.indef

      ‘Fifteen male students’

   2. b. χams-a  ʕašr-at-a  tˁaalɪb-at-an

      five-acc  ten-f-acc  student-f.sg-acc.indef

      ‘Fifteen female students’

   3. c. χams-at  wa-ʕɪšr-uun  tˁaalɪb-an

      five.f-acc and-ten-m.pl  student-acc.indef

      ‘Twenty-five male students’

   4. d. χams  wa-ʕɪšr-uun  tˁaalɪb-at-an

      five  and-ten-m.pl  student-f.sg-acc.indef

      ‘Twenty-five female students’

2. (16) Digit 1 fem Deletion:

   [f] → $\emptyset$ / [numeral] ___ ([numeral])+[enumerated noun]+[fem]

Table 2 gives the derivations for the examples in (15).

Table 2: Derivations for examples (15)a–d

However, because the numerals 3–10 are underlyingly feminine, both digits in the numbers 13–19 are expected to carry a feminine suffix. In other words, the derivations for (15a–b) should have a feminine morpheme in the underlying representation for the digit ʕašar. Moreover, the derivation should explain why this feminine morpheme is deleted in example (15a), but maintained in (15b). If the feminine enumerated noun is the trigger for the deletion of the feminine morpheme carried by the first digit, it then makes sense to expect the underlying feminine morpheme of the second digit to survive. This is borne out in example (15b). But now the question is why the feminine morpheme of the second digit does not survive in (15a).

To answer this question, we must first verify the assumption that the second digit is underlyingly feminine. Remember that our analysis of the digit ʕašar-at as feminine is based on the observation that it is feminine when used without an enumerated noun. However, we find the digit ʕašar only in its masculine form in the double-digit numerals 13–19 when they are used separately without an enumerated noun, as in (17). This suggests that the second digit in the numerals 13–19 is underlyingly masculine. It also suggests that the second digit in (15b) carries the feminine morpheme as a result of agreement in gender with the enumerated noun:

1. (17)

   1. a. χams-at-a  ʕašar  b.  χams-at-a  ʕašr-(*at-a)

      five.f-acc  ten          five-f-acc  ten-(*f-acc)

      ‘Fifteen’            ‘Fifteen’

The numerals 11 and 12 are another case where we find agreement in gender between the numeral and the enumerated noun. Both digits making up each of these two numerals agree with the enumerated noun in gender as in (18).

1. (18)

   1. a. ʔahad-a  ʕašar-a  tˁaalɪb-an

      one-acc  ten.m-acc  student-acc.indef

      ‘Eleven male students’

   2. b. ʔɪħda  ʕašr-at-a  tˁaalɪb-at-an

      one.f ten-f-acc  student-f-acc.indef

      ‘Eleven female students’

   3. c. ʔɪθna  ʕašar-a  tˁaalɪb-an

      two  ten.m-acc  student-acc.indef

      ‘Twelve male students’

   4. d. ʔɪθna-ta  ʕašr-at-a  tˁaalɪb-at-an

      two-f  ten-f-acc  student-f-acc.indef

      ‘Twelve female students’

Once again, this pattern of agreement can be correlated with the observation that each digit of the two numerals 11 and 12 can be in the masculine form only when these numerals occur separately without the enumerated noun, as in (19) and (20) below.

1. (19)

   1. a. ʔahad-a  ʕašar-a    b.  ʔɪθna  ʕašar-a

      one-acc  ten-acc            two    ten-acc

      ‘Eleven’                ‘Twelve’

2. (20)

   1. a. *ʔɪħda  ʕašar-a      b.  *ʔɪθna-ta  ʕašar-a

      one.f  ten-acc                two-f      ten-acc

      ‘Eleven’                  ‘Twelve’

   2. c. *ʔahad-a  ʕašr-at-a  d.  *ʔɪθna  ʕašr-at-a

      one-acc  ten-f-acc          two    ten-f-acc

      ‘Eleven’              ‘Twelve’

In generative syntax, an element that is feminine in its underlying representation, as in feminine nouns in Arabic, is classified as inherently feminine. In other words, these nouns do not receive the feature [fem] as a result of entering into a dependency relation with some other element. Adjectives, on the other hand, are not inherently feminine. Their feminine feature arises from an agreement relation that adjectives have with the feminine noun they modify. The feminine noun in this case carries an interpretable feminine feature [ifem] and the adjective an uninterpretable feminine feature [ufem] (Chomsky Reference Chomsky, Martin, Michaels and Uriagereka2000).

Accordingly, the fact that the feminine form of the second digit (ʕašar-at) in the double digit numerals in (17) and (20) can only occur with a feminine enumerated noun can be captured if these numerals have an uninterpretable feminine formal feature [ufem], licensed by an interpretable feminine feature [ifem] carried by the feminine enumerated noun. By the same token, the fact that the feminine numerals 3–10 can occur without a feminine enumerated noun suggests that they enter the syntactic derivation specified with an interpretable feminine feature [ifem].

In compound numerals such as (21a), the numerals χams- [ i fem ] and ʕashar- [ ufem ] are sister nodes under the Cardinal head and the feminine enumerated noun tˁaalɪb [ ifem ] is in the NP complement of the cardinal phrase:

1. (21)

   1. a. χams-a  ʕašr-at-a  tˁaalɪb-at-an

      five-acc  ten-f-acc  student-f.sg-acc.indef

      ‘Fifteen female students’

   2. b. χams-at-a  ʕašar-a  tˁaalɪb-an

      five.f-acc  ten-acc  student-acc.indef

      ‘Fifteen male students’

2. (22)

The Agree relation between the feminine feature of the enumerated noun and ʕašar can value the ufem feature of ʕašar. When the enumerated noun is masculine, as in (21b), the second digit ʕašar does not have a feminine feature.

The way to capture the contrast between numerals that are subject to the deletion rule and those that are not is by limiting the deletion rule to interpretable feminine features [ifem], as in the rule in (23), the revised version of the rule in (16).

1. (23) Digit1 fem Deletion:

   [ifem]→$\emptyset$ / [numeral] ___ ([numeral])+[enumerated noun]+[ifem]

Table 3 gives the derivations for the examples in (21):

Table 3: Derivations for examples (21)a, b

Table 4 gives the derivations for the four examples in (18).

Table 4: Derivations for examples (18)a–d

Now moving to the higher numerals m ɪʔat ‘hundred’ and ʔalf ‘thousand’, we find an apparent challenge to rule (23). Rule (23) underapplies in (24) in Standard Arabic. The numeral ‘hundred’ carries a feminine morpheme, which is not deleted when the enumerated noun is feminine, contrary to what rule (23) predicts. Moreover, it seems that rule (23) overapplies in (25) and underapplies in (26).

1. (24) mɪʔat  mu'allɪm-in/ mu'allɪm-at

   hundred  teacher-m.sg/ teacher-f.sg

   ‘a hundred male/teachers/ female teachers’

2. (25) χams  mɪʔat  tˁaalɪb/ tˁaalɪb-at

   five.f.sg  hundred  student.m.sg/ student-f.sg

   ‘Five hundred male students/ female students’

3. (26) χams-at  ʔaalaaf  tˁaalɪb/ tˁaalɪb-at

   five.f.sg  thousands  student.m.sg/ student-f.sg

   ‘Five thousand male students/ female students’

However, the rule in (23) makes correct predictions for all of these examples if we consider the numerals ‘hundred’ and ‘thousand’ to be enumerated nouns, with ‘hundred’ specified as inherently feminine and ‘thousand’ as inherently masculine. The feminine morpheme therefore does not undergo deletion, but rather triggers deletion of the feminine morpheme carried by the first digit. Because ‘hundred’ is feminine, the numeral is always masculine in (25), and because ‘thousand’ is always masculine the numeral is always feminine in (26).

This means that these high numerals act as enumerated nouns in that they trigger the deletion of feminine morpheme of the other numerals, and the gender status they have is never influenced by the rules. They pattern exactly as enumerated nouns do when following the numerals 3–10: exactly what we would expect if they are nouns. Corbett (Reference Corbett1978a) argues that high numerals in Slavic languages act like nouns. Under this assumption, rule (23) will not apply in (24) because there is no numeral. Both ‘hundred’ and ‘teacher’ are nouns. By the same token, the feminine morpheme carried by ‘hundred’ triggers the deletion of the feminine morpheme carried by χams-at in (25). Table 5 gives the derivations for (24), (25), and (26).

Table 5: Derivations for examples (24)–(26)

Before ending this section, I would like to give further evidence that gender polarity involves the deletion of the feminine morpheme of the numeral rather than the addition of a feminine morpheme when the enumerated noun is masculine. By looking at the gender in the numeral phrase in Jordanian Arabic (JA), we find that the numerals 3–10 are similar to Standard Arabic in being feminine underlyingly.Footnote ³ In other words, they have an [ifem] morpheme that is expected to be deleted if the deletion rule in (23) is operative in JA. Crucially, if the process that produces gender polarity in MSA is deletion and if JA overgeneralizes that process, we would expect the [ifem] of the numeral to be deleted, not only when the enumerated noun is feminine but also when it is masculine. By the same logic, if the process behind gender polarity were the addition of a feminine gender, and if this process were to overgeneralize, the outcome would be that the feminine morpheme would be added to the numerals not only when the enumerated noun is masculine but also when it is feminine. In fact, Jordanian Arabic exhibits the first of these patterns, as shown in (27).

1. (27) χams-(*at)  tˁaalɪb-aat/ tˁullaab

   Five-(*f) student-f.pl/ student.m.pl

   ‘Five female students’

Rule (23) thus overapplies in Jordanian Arabic, deleting fem even when the enumerated noun is masculine. The example in (27) can be produced by the rule in (28), according to which the feminine morpheme of the numeral is deleted whenever it occurs with an enumerated noun. Interestingly, there are cases where the MSA deletion rule as formulated in (23) seems to be respected in JA. This takes place in certain collocations where the feminine morpheme of the first numeral is kept when expected under (23) (i.e., when the enumerated noun is masculine, as in (29a–e), or when the numerals 13–19 are used separately). These cases provide evidence that JA once had the deletion rule in (23), and these cases are a residue of the earlier system. The feminine morpheme has been reanalyzed as a liaison in these very frequently used combinations. In other words, the t– segment has undergone resyllabification as in the examples in (29).

1. (28) Digit1 FEM Deletion: [ifem]→$\emptyset$ / [numeral] ___ ([numeral])+[enumerated noun]

2. (29)

   1. a. χams  t-ɪrbaaʕ  b.  χams  t-iyyaam  c.χams  t-aalaaf

      five  t~quarters    five    t~days    five  t~thousands

      ‘Five quarters.’      ‘Five days.’    ‘Five thousands.’

   2. d. χams  t-ʊšhʊr  e.  χams  t-ʊstˁʊr  f.χams  tˁ-aašFootnote ⁴

      five  t~months/    five    t~lines    five  t~ten

      ‘Five months.’      ‘Five lines.’    ‘Fifteen.’

Finally, an anonymous reviewer points out that hallmark examples of impoverishment produce unmarked forms. The deletion rules in this article seem to run counter to this by producing polarity, a marked surface phenomenon. However, while it is true that Polarity produces marked ‘surface structures’, I have shown here that gender polarity is a grammatical relation that produces unmarked gender forms (the masculine). In the next section, I will show that number polarity also produces unmarked forms (the singular).

4 Number polarity

As for number polarity, recall that the enumerated noun is plural by default. I propose a deletion rule that deletes the plural morpheme of the enumerated noun (EN), as in (30). It is motivated by economy and applies when the numeral carries a plural morpheme. Recall that the enumerated noun must be plural when the numeral is one of the cardinals 3–10, as in (31a). If the rule in (30) only applies to nouns enumerated by numerals that carry a plural morpheme, then the rule does not apply in (31a).

But with the compound numerals 11–19 in (31b), the numeral triggers the deletion of the plural morpheme of the enumerated noun. This suggests that these compound numerals carry a plural feature. There is evidence that the second digit ʕašar in these compounds has a grammatical function rather than a lexical meaning corresponding to the numeral ‘ten’. This numeral in the compound numerals 11–19 is reduced to a suffix in many Arabic dialects such as Jordanian, illustrated in (32). But even in Standard Arabic, we saw earlier in section 3 that this numeral is underlyingly masculine ʕašar, and therefore distinct from the numeral ʕašar-at ‘ten’. Moreover, this numeral is not interpreted as ‘ten’. The whole compound is interpreted the same way the English teen numerals are. So the compound numerals 11–19 are marked with pl underlyingly by virtue of being compound. Now even more explicitly, with numerals indicating decades, as in (31c), the numeral carries the plural suffix –uun.

1. (30) EN PL Deletion: [+pl] → $\emptyset$ / [numeral]+[pl]+[enumerated noun] ___

2. (31)

   1. a. χams  tˁaalɪb-aat-ɪn

      five  student-f.pl-gen

      ‘Five female students’

   2. b. ʔahad-a  ʕašar-a  tˁaalɪb-an

      one-acc  ten.m-acc  student-acc.indef

      ‘Eleven male students’

   3. c. ʕɪšr-uun  tˁaalɪb-an

      ten-m.pl  student-acc.indef

      ‘Twenty male students’

3. (32) χams-tˁaaš

   five-ten

   ‘Fifteen’

If rule (30) applies in compound numerals and decade numerals, the plural morpheme of the enumerated noun is deleted, giving the singular enumerated noun. Table 6 gives derivations for the forms in (31).

Table 6: Derivations for examples (31)a–c

Like the compound numerals 11–19, the compound numeral consisting of one of the numerals 3–9 and the numeral ‘hundred’ is marked with pl underlyingly, hence the singular enumerated noun, as in (33). On the other hand, in (34), the numeral χams does not form a compound with the numeral ʔaalaaf. Therefore, ʔaalaaf is pluralized as if it were an enumerated noun as in (31a) above. So, although the numerals in (34) do not form a compound marked with pl, the numeral ʔaalaaf is the plural of ʔalf and therefore carries a plural morpheme, unlike mɪʔat in (33). This plural morpheme is the trigger for having a singular enumerated noun.

1. (33) χams  mɪʔat  tˁaalɪb/ tˁaalɪb-at

   five.f.sg  hundred  student.m.sg/ student-f.sg

   ‘Five hundred male students/ female students’

2. (34) χams-at  ʔaalaaf  tˁaalɪb/ tˁaalɪb-at

   five-f.sg  thousands  student.m.sg/ student-f.sg

   ‘Five thousand male students/ female students’

Interestingly, the singular is ungrammatical as in (35a) below. An anonymous reviewer points out that if ʔaalaaf in (34) is an enumerated noun, it should take the singular form when it follows numerals greater than 10. Indeed, this prediction is borne out in (35b), where the numeral ‘eleven’ is followed by the singular form of ‘thousand’.

1. (35)

   1. a. *χams-at  ʔalf  tˁaalɪb

      five-f.sg  thousand  student.m.sg

      ‘Five thousand students’

   2. b. ʔahad-a  ʕašar-a  ʔalf

      one-acc  ten.m-acc  thousand

      ‘Eleven thousand’

Crucially, the reason the numeral ʔaalaaf must be plural in (34) is that it is enumerated by χams-at, and thus interpreted as plural. In the derivation of (34) the string composed of the two words χams-at ʔaalaaf is subject to rule (30), which clearly does not apply since χams-at does not carry a pl morpheme. Then, the rule applies to the string made up of χams-at ʔaalaaf as the numeral and tˁaalɪb as the enumerated noun. This suggests that mɪʔat in the compound numeral χams-mɪʔat gets its obligatory plural interpretation from being in the compound construction. In other words, being in the compound construction gives it the pl abstract morpheme. Tables 7 and 8 show the derivations for (33) and (34) respectively.

Tables 7: Derivation for example (33)

Tables 8: Derivation for example (34)

For the numerals ‘hundred’ and ‘thousand’ in (36), it is not clear why the enumerated noun is singular, since these numerals do not carry a plural morpheme. That is, mɪʔat is not part of a compound numeral, and ʔalf is not pluralized. The use of a singular enumerated noun in this context suggests that these higher numerals are treated as plurals. Interestingly, while neither the numeral nor the enumerated noun is marked as plural, the verb in (37) displays plural number agreement with the numeral phrase in preverbal subject position. This plural number inflection on the verb can be explained if the numeral phrase exhibited an abstract plural morpheme in the syntax proper, which was deleted later in the derivation. Suppose that, consistent with the analysis of number polarity in this article, the higher numerals have an abstract plural morpheme that triggers the deletion of the plural morpheme of the enumerated noun.

1. (36) mɪʔat/ʔalf  mu'allɪm-ɪn

   hundred/thousand  teacher-gen

   ‘One hundred/thousand teachers’

2. (37) mɪʔat/ʔal  mu'allɪm-ɪn  šaarak-u  fi-l-ʔɪdˁraab

   hundred/thousand  teacher-gen  participated-3.m.pl  in-the-strike

   ‘One hundred/thousand teachers participated in the strike’

Plural agreement on the verb is evidence that the numerals mɪʔat and ʔalf are plural despite the fact that they do not carry a plural morpheme. This would be similar to collective nouns such as al-naas ‘people’ which is plural although it does not carry a plural morpheme. But this raises the question of why these numerals behave like collective nouns. Interestingly, there is cross-linguistic evidence that higher numerals are ‘nounier’ than lower numerals (Corbett Reference Corbett1978a, Reference Corbett1978b).

Abbreviations

3m.pl

third person masculine plural

acc

accusative

f/fem

feminine

gen

genitive

indef

indefinite

m

masculine

Nmrl

numeral

pl

plural

sg

singular

du

dual