Prenominal possessives in Yiddish: mayn khaver versus mayner a khaver

2.1 Constituency

We start with some preliminary considerations (for very useful background information on Yiddish, see Jacobs [2005]; Hoge [2018], and many others). Note first that these two types of possessive construction form constituents. As this is straightforward for the first construction, I will focus on the second construction. Constituency can be demonstrated by the Verb-Second constraint (4a); by (asymmetric) coordination, that is, coordination of different elements (4b) (for symmetric coordination, see Section 3.3.3); and by the construction occurring as the complement of a preposition (4c) (note that all prepositions in Standard Yiddish take dative case).

In addition, there is also language-specific evidence for constituency. Note first that eyner can also precede the indefinite article (5a). Crucially, the possessive can be sandwiched between eyner and the indefinite article (5b). Following Lockwood (1995: 66), I translate the string eyner a as ‘a certain’.^[2]

This constituency is confirmed by agreement facts. To begin, elements of regular DPs agree in case, number, and gender; that is, they participate in concord sharing the same features.

The same holds for the second construction where the possessive pronoun, the indefinite article, and the head noun must all agree in case, number, and gender.^[3] This is illustrated with elements in the masculine gender in (7a)–(7b) (see also (5) above).

Combinations of elements in the singular and plural are not possible.^[4]

It is clear then that all these elements form a constituent, and that the possessive pronoun must agree with the rest of the nominal. Possessors involving regular DPs also form constituents with the lower nominal. However, as shown in the next subsection, they do not share concord features with the rest of the nominal.

2.2 Possessors involving regular DPs are dative

As seen above, possessive pronouns (mayner) vary with regard to case, number, and gender. As is well known though, possessive determiners (mayn) only change as regards plural (mayne) but not case or gender. Recall also that the singular indefinite article is invariant as regards case and gender. Thus, if an adjective is missing, morphological case cannot easily be inspected on DPs with possessive determiners or indefinite articles (e.g., mayn/a lerers ‘my/a teacher’s’).^[5] The same holds for proper name possessors, which do not change their case form in the possessive (cf. (2b) and (3b)).^[6] This is different for DPs involving definite articles where case can easily be identified. Above, I followed the traditional analysis of glossing the possessors in (2c) and (3c) as dative case.

In a recent article, Hoge (2018) argues that possessors involving DPs with a definite article should be analyzed as genitives. Discussing the first construction, she points out that if the possessor DP were indeed dative, then certain noun phrases would appear with unexpected endings on the head noun; for instance, while der tate ‘the father’ appears with the expected dative suffix -n in (9a), di mame ‘the mother’ does not exhibit this suffix in (9b).

Hoge (2018: 262) argues that these cases should be analyzed as genitive case positing suffixes different from traditional assumptions (10a). This analysis seems to be consistent with the fact that group genitive constructions involving PPs (or relative clauses) are not possible (10b).

Hoge proposes that unlike in English, the possessive marker -s in Yiddish is not a clitic. Since (prenominal) possessors with postnominal modifiers as in (10b) are not possible, Hoge (2018: 252) concludes that Yiddish -s attaches to the head noun of the possessor DP, provided the head noun is at the right edge. Thus, -s is part of a genitive suffix.^[7]

However, there is clear evidence that the possessive -s can be separated from the possessor head noun. This includes an inflected head noun in the first construction where the suffix -ns is split into two separate parts (11a). The possessive -s can also be in a position separated from the head noun in the second construction (11b). As explicitly stated in Falkovitsh (1940: 181), only the last noun receives -s. As for (11c), note already here that a possessor can also follow its possessum head noun (see Section 5.4) and that this entire noun phrase can form a possessor to the possessum head noun of the larger DP. Again, the possessive -s is not on the relevant noun.^[8]

Furthermore, the possessive marker can also appear on elements higher in the noun phrase, on adjectives (12a) and determiners (12b). This is also possible with quantificational elements (12c) including indefinite pronouns (12d) (for more relevant examples, see Jacobs [2005: 187]).

Recall that Hoge assumes that dem is (also) a genitive form (10a). From this perspective, it is unclear why dem in (12b) has -s. Interestingly, Hoge (2018: 248) provides an example similar to (12b): (ot) dems / yenems bukh ‘this/that one’s book. In order for her generalization to hold, she must assume that dem and yenem are head nouns. This, however, is unlikely. Among others, note that dem in Hoge’s example is preceded by the optional deictic reinforcer ot ‘here.’ However, this element cannot precede regular nouns without a determiner. Furthermore, all other elements marked by -s in (12) would also have to be analyzed as head nouns. Again, this is unlikely for adjectives (12a) and the quantificational elements in (12c)–(12d).

To rescue Hoge’s proposal, one might claim that this -s licenses the ellipsis of the lower overt material of the possessor DP (13a). However, indefinite pronouns as in (12d) never have an overt noun (13b), and yet they appear with -s in possessive contexts. This means that indefinite pronouns cannot involve ellipsis (strikethrough marks ellipsis).

Furthermore, in the discussion of the second construction (Section 3), we will see that inflected elements (e.g., zayner) can license ellipsis. From that perspective, it is unexpected that inflected adjectives as in (12a) still exhibit -s. Thus, the general function of -s in the first construction is not the licensing of ellipsis but rather the indication of possession.

As far as I have been able to establish, the cases in (11) and (12) are possible quite generally. Now, given the fact that the possessive marker can appear on an element other than the head noun, that is, either to the left or right of it, I will continue assuming that the traditional analysis (i.e., DP_DAT + -s) is correct (for most dialects). This means that the unexpected possessor forms discussed in Hoge (2018) are in need of a different explanation. Currently, I have no solid insights to offer.^[9] In what follows, I will focus on possessive pronouns as they exhibit concord with the lower nominal and show a clear difference to the first construction (inflection).

2.3 The second construction is indefinite

In the introduction, I illustrated two morpho-syntactic differences between the first and the second construction. While the former has an (uninflected) possessive determiner and no article, the latter has an (inflected) possessive pronoun and an indefinite article. Furthermore, I stated that unlike the first construction, the second construction is indefinite in interpretation. Indeed, in reference books, text books, and survey articles, mayner a khaver is rendered as ‘a friend of mine’ or ‘one of my friends’ (Lockwood 1995: 53–54; also, Birnbaum 1979: 297; Jacobs 2005: 184; Schaechter 2003: 268; Wiener 1893: 66). These are indefinite readings. Using mostly possessive pronominals, I will provide evidence that the first construction is definite and that the second one is indefinite. Making this point in different ways, the following discussion will add more distinctions between the two possessive constructions showing that they are systematically different.

First, in a running text, new information, the rheme, appears in an indefinite form, but old information, the theme, surfaces in a definite form. This is exactly what we find with the possessives under discussion here where the second construction introduces new information, but the first construction is used to continue the discourse (for a similar case, see Boris Sandler’s Keynemsdorf, p. 50).

Note that both the possessor and the entire nominal change from an indefinite to a definite form. I take this to mean that the (in-)definite forms reflect the corresponding interpretations.

Second, both types of possessive behave differently in presentational (/existential) sentences where the DP with the possessive determiner is marked, but the one with the possessive pronoun and indefinite article is not (cf. Jacobs 2005: 225; ^m = marked).

Third, it is clear that Peter is married, and Peter is not married is contradictory if Peter is the same person. There is a difference in Yiddish when mayn khaver replaces both names versus when mayner a khaver replaces both names. The first leads to a contradictory statement, but the second is non-contradictory.

Presumably, this follows from the observation that mayn khaver implies uniqueness, but mayner a khaver does not. Note now that uniqueness is part of a definiteness interpretation (Lyons 1999: 7–12).

Fourth, khaver can have different denotations: it means both ‘boy-friend’ and ‘friend’ in the first construction but only ‘friend’ in the second construction. This is indicated by the respective translations in (17a) and (17b).

Again, this distinction follows from the fact that dayn khaver has a uniqueness implication, which dayner a khaver lacks. Given the above differences, I conclude that the first construction is definite and that the second one is indeed indefinite. If this is so, then we might expect the second construction to have two types of indefinite interpretation as regards specificity. This is borne out.

Considering the data below, it is clear that (18a) describes a specific (known) friend. In contrast, contexts such as (18b) describe a more general fact. This case is unspecific as to the identity of the friend; that is, this statement could apply to any of my friends.

These two types of interpretation are typical of indefinite articles or indefiniteness in general. In other words, the presence of the possessive pronoun makes no difference for the (in-)definiteness of the noun phrase as a whole.

To summarize, the second construction forms a constituent, its possessor involving a regular DP is in the dative, and it is indeed indefinite in interpretation.

3.1 Structure of the first construction

Beginning with the first construction, I propose that the (uninflected) possessive determiner is in Spec,DP:

(21)

The same holds for possessors involving proper names or regular DPs and makes Yiddish structurally similar to English.^[10]

Falkovitsh (1940: 181) and Jacobs (2005: 240) point out that the possessive in the first construction cannot co-occur with a definite article. Articles, definite or indefinite, are commonly assumed to reside in D. Now, the DP-level in West Germanic usually contains only one overt element, either in Spec,DP or in D. This restriction (whatever its deeper explanation) immediately accounts for the complementary distribution between the possessive and the article in (21) (for the discussion of definite elements in this regard, see Roehrs [2019a]; also, Haspelmath [1999]).

Turning to Definiteness Spread, I make the standard assumption that the (in-)definiteness of the entire noun phrase is determined at the DP-level (Lyons 1999). Following Alexiadou (2005), an element in D or, by Spec-head agreement, in Spec,DP will yield the relevant interpretation of the DP. With the possessive in Spec,DP in (21), zayn will bring about definiteness by Spec-head agreement. Consequently, the first construction adheres to Definiteness Spread as regards definite possessives. This makes the predictions that indefinite possessives should lead to indefinite DPs. This is borne out.

A reviewer reminds me that indefinite possessives are indeed possible (22a) and are not marked in presentational contexts (22b).

This then is similar to certain cases of the second construction (e.g., a lerers a khaver ‘a friend of a teacher’s’). Note that this option should not come as a surprise. As in Yiddish, in English there are several ways to express indefinite DPs involving possessives (e.g., a teacher’s friend, a friend of a teacher[’s]).

3.2 Structure of the second construction

If this is on the right track, then we expect the presence of an indefinite article to lead to indefiniteness. We have seen in Section 2.3 that the second construction is indeed indefinite in interpretation. The question arises as to why the definite possessives do not “overwrite” the indefiniteness in the second construction.

Starting with inflected zayner, Jacobs (2005: 242) proposes that these cases involve two separate noun phrases. Note though that he does not provide a detailed analysis. Basically following Jacobs, I take (20b) to hold for all possessive pronouns. In other words, I argue that the possessive in the second construction also involves an elliptical construction where the null noun is licensed by the inflection on the pronominal. As such, I propose that zayner involves its own nominal. As argued further below, I claim that this possessive is lower inside its own nominal indicated by Ø in (23a). Furthermore, I propose that the structure in (23a) is part of the matrix nominal (23b). This yields one complex nominal with two separate subparts (I turn to the identification of XP below).

(23)

a. [dp Ø zayner e_N]

[DP Ø zayner eN ]

Before we proceed, note that the presence of the indefinite article immediately shows that the possessive marker -s, present on proper names and regular DPs, is not in D.

Recall that I assume that (23a) involves an elliptical construction. There is evidence that possessives are in a lower position in elliptical contexts as they can follow a determiner in such contexts (24a)–(24b). Note that a lower position of a possessive is also possible (albeit rare) in non-elliptical contexts (24c) (observe that Perets is followed by a non-standard dative inflection).^[11]

If the possessive is indeed lower inside the embedded DP as in (23a), several facts follow: zayner does not “overwrite” the indefiniteness of the larger noun phrase (in fact, there is no clash in definiteness). Furthermore, there is no violation of the Backward Anaphora Constraint.

With the possessive pronoun lower inside the embedded DP in (23b), it precedes but does not c-command the lower matrix nominal, its antecedent.

Thus far, it appears that Definiteness Spread can be maintained for Yiddish. In view of the structure in (21), definite possessors of the first construction straightforwardly yield definite DPs. Given the structure in (23b), possessive pronouns of the second construction are compatible with indefiniteness. As for proper name and regular DP possessors, I assume that they are also lower inside the embedded DP where the possessive marker licenses the null noun. In other words, while -s indicates possession, it can also license ellipsis similar to the inflection on the possessive pronoun. That such possessives can indeed license ellipsis is clear from the following cases (examples are taken from Falkovitsh [1940: 181]).

Other elements with possessive -s can also license ellipsis.

Note again that the possessive marker can appear on an element other than the head noun. This is particularly clear in (27c), which involves a postnominal possessive.

The above possessors are phrasal in structure and thus must be in a phrasal position. Without going into too much detail, I assume they are in a Possessive Phrase (PossP), which in turn is in a specifier position above the null noun. The entire structure of the embedded possessive makes up a DP (28a). The structure of the possessive pronoun in (23a) is fleshed out in (28b).^[12]

Next, I turn to the second question from above, specifically to the question of the exact structure of the second construction. After fleshing out the details of XP in (23b), I will briefly return to the issue of Definiteness Spread.

3.3 Structure of the second construction in more detail

There are four options of where the embedded DPs in (28) could be located. Focusing on the possessive pronoun, the latter and the indefinite article could form a doubly-filled DP (29a) where the pronoun is in Spec,DP and the article in D. Both of these elements could also form a complex specifier in Spec,DP (29b). Third, the possessive pronoun could be in a specifier above the DP-level (29c), where XP is interpreted as the Left Periphery Phrase (see Giusti and Iovino 2016). Finally, the possessive pronoun could be adjoined to DP (29d).

In what follows, I will discuss these four options in more detail showing that (29c) or (29d) are most likely correct.

4.1 Eyner and epes are similar to the inflected possessive pronominal

As seen in Section 2.1, eyner can precede an indefinite article. Like mayner, eyner also agrees with the lower nominal in concord features (39a). The word epes can also precede an indefinite article, but this element is invariant (39b).^[13]

The words eyner and epes behave like the possessive in the second construction in other respects. First, eyner can be separated from the indefinite article by the degree word zeyer ‘very’ (see also Section 5.1).

As might be expected, epes can also be separated from the indefinite article, for example by the word nokh ‘still’ (41a) or the degree word zeyer ‘very’ (41b).^[14]

Second, as with the second possessive construction, the lower indefinite DP can occur as eyner. Consider examples involving epes.

As far as I know, this is not possible when eyner is the first element, presumably because the substitution of the lower indefinite DP would yield two identical elements (eyner eyner).

To sum up thus far, eyner and epes are similar to the possessive pronoun in the second construction: these elements can precede the indefinite article, and they can be separated from it by other words. Furthermore, and to the extent possible, they exhibit concord in features with the lower nominal (eyner a Noun), and the lower nominal can be substituted by eyner (epes a Noun).

4.2 Some arguments for adjunction

There is evidence that the possessive pronoun, eyner, and epes are adjoined to the lower DP (Option 4). There are two facts that indicate adjunction: the above elements can be stacked, and they can switch positions. First, these predeterminer elements can be stacked; for instance, eyner can precede the possessive (both examples from Zaretski [1929: 172]).

Syntactic context and agreement facts indicate clearly that these complex structures form constituents.

As might be expected, epes can also precede a possessive.^[15]

Second, a very strong argument in favor of adjunction comes from the fact that the two elements above the indefinite article, epes and the possessive, can actually switch positions. Compare (44) to (45).^[16]

Again, syntactic context and agreement facts show that these are constituents. Importantly, it is unlikely that the possessive in (45a)–(45b) is adjoined to epes itself. I assume that adjunction is to the entire lower nominal (more on this below).

Turning to the interaction between eyner and epes, the former can precede the latter:

As might be expected by now, both of these elements can also switch positions.^[17]

Again, it is clear from the context that these structures involve constituents. Similar to above, I assume that eyner in (46) and epes in (47) are adjoined to the entire noun phrase. Finally, at the beginning of this subsection, I provided examples like eyner mayner a Noun. I have not found any examples of the type mayner eyner a Noun. I briefly return to this in Section 5.1.

Above, I observed that it is unlikely that the possessive pronoun is adjoined to epes itself. There is some good indication that the first/higher element of the stacking is indeed adjoined to the entire nominal (rather than to the second/lower element of the stacking). Consider the following two examples. Note again that eyner shows concord in features with the following possessive element, dative masculine in (48a) and nominative masculine in (48b).

In Section 2.2, we have seen that definite DP possessors are in the dative case. While not overtly marked, I assume that this is also true of indefinite DP possessors as in (48a). Assuming that adjunction involves the sharing of agreement features, the dative case on eynem in (48a) indicates that the possessor nominal a yeshuvnik is in the dative and that eynem belongs to it (and not to nominative a zun). Similarly, eyner could be in construction with the possessive pronominal in (48b). However, while I have seen hundreds of examples where eyner or zayner form a constituent with a lower indefinite DP (e.g., a khaver), I have never seen an example where eyner zayner occurs by itself.

I believe these cases do not exist.^[18] It appears then that eyner zayner is only possible if an indefinite DP follows the two elements. I take this to mean that eyner is adjoined to the whole phrase zayner a khaver in (48b) (rather than to zayner itself). In other words, the bracketed structures in (48a–b) have the same analysis, with the qualification that zayner in (48b) separates the two elements. If zayner itself is also adjoined to the indefinite DP as proposed above, then its presence should not raise any issues.^[19]

Given this discussion of adjunction, the examples in (46a) and (47a) can be illustrated in a simplified way as follows.

(49)

The same holds for the cases involving possessives. If this is on the right track, then the data discussed in this article provide more evidence that there are syntactic positions above the DP proper.

To explain these stackings and their different distributional options in a Split-DP (Option 3), we would have to assume that LPP is recursive.

(50)

This, however, raises other questions: for instance, what is the nature of this recursive LPP, and why can the elements in the specifiers of this recursive LPP be switched? These types of questions do not arise in an account involving adjunction.

That adjunction might be on the right track is indicated by the fact that Yiddish is a language that tolerates adjunction quite freely. This can be seen with elements such as ot ‘here’ and take ‘really’ in the left periphery of the noun phrase in that they can also be stacked and switch positions.

Like above, these elements can be analyzed as left adjoined to the DP. Postnominal modifiers also indicate that Yiddish allows adjunction quite generally (also Section 5.3):

However, these types of modifiers are presumably right adjoined to the preceding DP.

In this section, I showed that eyner, epes, and the possessive pronoun of the second possessive construction behave in the same way. These elements seem to be left adjoined to the lower DP.

5.1 Restrictions in Yiddish

There are some issues left unaddressed thus far. First, a possessive pronoun cannot be followed by a definite article (53a). In fact, these constructions can only be followed by an indefinite article (53b). As pointed out to me by Dovid Katz (p.c.), cases containing the singularity numeral can only involve a 19th-century spelling, where eyn, however, was already pronounced as a(n) or close to it. A relevant example is given in (53c). As noted by a reviewer, the indefinite element, romanized as <ajn> in the original source, is a Southeastern or Central Yiddish form, presumably pronounced as [ain], and might present a transitional instance:^[20]

Given current assumptions, one may wonder how we can formalize the fact that a possessive pronoun can only be followed by a DP involving an indefinite article.

Second, it is well known that the Germanic languages have no overt indefinite articles in the plural. Let us assume that this is a null article. Intriguingly, the second construction cannot be in the plural at all. As explicitly confirmed by Jean Lowenstamm (p.c.), mayne (dray) khaveyrim ‘my (three) friends’ does not feel at all like the plural of mayner a khaver ‘a friend of mine’. In other words, there is an interpretative restriction such that possessive constructions in the plural can only be interpreted as definite (54a). This means that (54a) cannot be analyzed as (54b) (Ø_INDEF = null indefinite article in the plural):^[21]

To repeat, the second construction can only occur in the singular. This is different for postnominal possessives, which can be both indefinite and plural:

Similar to the first question, one may wonder how we can formalize the fact that these prenominal possessives only occur in the singular.

Although we stated above that an analysis involving LPP is unlikely, let me start with this option. Selection could be invoked to explain why only an indefinite DP occurs in this construction. Selection is only statable if a higher head is present. Thus, we could stipulate that the LP head selects a DP involving the feature [indef]. Furthermore, we could exclude plurals by stipulating again that LP only selects a singular DP. As shown above though, all nominal elements inside the larger noun phrase participate in concord of case and phi-features. While these agreeing elements involve phrases (including heads), they show no restrictions as regards plural (e.g., ale mayne khaveyrim ‘all my friends’). So contrary to fact, we would presumably expect plurals to be possible with the second construction under selection.

Above, I argued for adjunction. We could suggest then that adjunction in the plural is blocked by the more economical derivation of plural mayne khaveyrim structurally on par with singular mayn khaver. In other words, only a simple DP as in (21) would be projected in the plural. Depending on the internal structure of the possessive itself, this might also explain why an indefinite interpretation in the plural is not available (cf. Footnote 18). However, if economy of derivation holds more generally as one would expect, then one may wonder how the singular cases of the second construction are allowed back in. These should be blocked by the simpler derivation in (21) as well. Thus, this does not resolve the issue as regards plural either.^[22]

We have seen that eyner and epes can precede an indefinite article (see, e.g., [39]). Interestingly, degree words such as zeyer / gor / graylekh ‘very’ can also precede the indefinite article but not the definite one (Jacobs 2005: 240; Lockwood 1995: 61). Consider (56a)–(56b).^[23] Importantly, these degree words can follow both the indefinite and the definite article (56c)–(56d):

As far as I can see, we can add to this list of preceding elements take ‘really’ (30b) and nokh ‘still’ (41a). Furthermore, a sakh ‘many/much’ seems to behave in the same way (57a). This complex element can also follow an indefinite and a definite article (57b)–(57c):

In fact, unlike what is often reported in the literature, there are more elements that can precede the indefinite article (58a)–(58d). This includes non-standard words like etvos ‘somewhat’ (58e).

With this in mind, we can state that there are restrictions in Yiddish on elements in the DP-level if something precedes them. Specifically, this is true of the indefinite article. It seems that typically only two elements can precede this article (for preceding PP possessives, see next section):^[24]

This restriction has the looks of a template (cf. Culicover and Jackendoff [2005: 29]’s notion of ‘grammatical frame’) that accommodates two predeterminer elements. The structure in (59a) is only a first approximation, and I will discuss it in a little bit more detail in the next subsection. However, I will not investigate what all the possible and impossible combinations of X and Y are. Let me just repeat here that while I have found examples of the type in (60), I have not come across any examples of the kind in (61). This includes a fairly comprehensive search of these kinds of examples in the Modern Yiddish Corpus and on the Internet:

Interestingly, like epes and take, the degree word zeyer ‘very’ can also intervene between the possessive pronominal and the indefinite article:^[25]

Thus, let us assume that (61) is indeed not possible. Note that epes, take, and zeyer are invariant forms. In contrast, both eyner and mayner share concord features. Let us also assume that epes and eyner are base-generated above the DP-level but that mayner moves from a lower base position (see Section 5.3).^[26] If so, one could speculate that mayner cannot cross eyner (perhaps because these two elements share the same features and involve a similar structural makeup).

5.2 Restrictions in German

As pointed out in Roehrs (2019b), elements in the German DP-level are also subject to restrictions if something precedes them. Among others, this can be seen with preposed PPs where determiners are possible, but Saxon Genitives are not (data are taken from Haider [1992: 315]; for (63c), see also Fortmann [1996: 22]).^[27]

The generalization is as follows:

Thus, in German there are also restrictions on elements in the DP-level if something precedes them. Roehrs (2020) argues that preposed PPs are in Spec,LPP.

In order to account for the German restriction in (64), Roehrs (2019b) argues that this is a selectional requirement imposed on the DP by the head LP. Now, preposed PPs are also possible in Yiddish (65a). Similar to German, Yiddish also allows a definite article here (65b)–(65c).

Given the possibility of a definite article, the restriction to the indefinite article in the second construction does not seem to be relatable to the restriction seen with German preposed PPs. Furthermore, preposed PPs are stressed, but possessives in the second construction are not. As such, adjunction as discussed above seems to be more plausible for the Yiddish possessives. Given my current understanding of these issues, I would like to suggest that a template as in (59a) appears to be needed for Yiddish.

Interestingly, left-adjoined elements and preposed PPs can co-occur. Specifically, the example (66a) involves a preposed fun-possessive and epes and eyner to the left of it (note that ire ‘her/its’ refers back to feminine di shtot ‘the town’); the case in (66b) arguably involves a preposed fun-adverbial in the second construction:

If the above discussion is on the right track, this yields a fairly complex structure with the fun-phrase in Spec,LPP, and epes, eyner, and the possessive adjoined to LPP (recall that preposed PPs are in Spec,LPP; Roehrs [2020]):^[28]

(67)

Note that epes, eyner, and the possessive are not directly adjoined to the indefinite DP (although such a DP is lower in the structure). This means that the template in (59a) must be stated in a general way, that is, without any labels on the left.

5.3 An alternative analysis

In the present proposal, possessives in the second constructions are left-adjoined to the lower DP. Prince (2001: 273–276) makes a different proposal for the similar eyner-a-Noun construction in (68a) relating it to (68b). She states that both of these cases are intoned as a single constituent with no break between the two nominals. Furthermore, the second nominal forms a restrictive modifier to the first:

Prince does not flesh out the syntactic details of these constructions. At the end of Section 4.2, I briefly suggested that post-modificational elements are right adjoined to the preceding DP. Thus, similar to a sheyne in (68b), a yid in (68a) could presumably also be taken to be right adjoined to its preceding element.

Above, we have documented that epes and possessive pronouns have properties similar to eyner. Thus, it seems desirable to extend this proposal to these elements as well:

Employing right adjunction, the three cases under discussion here could be analyzed as follows:

(70)

All elements in (70) can occur by themselves. One advantage of this analysis is that right adjunction captures the fact that the element on the right is optional. I return to this point below.

Prince (2001: 273) makes the discussion even more general. Again, as briefly mentioned at the end of Section 4.2, Yiddish also allows structures of the type a-Noun-a-Noun, which are very frequent in occurrence. They can be emotive (71a)–(71b) or more matter of fact in tone (71c)–(71d):

Many of these examples do not have a comma ((71a), (71c)), but quite a few do ((71b), (71d)).^[29] Although they appear with or without a comma, they also form one intonational phrase. Semantically, nouns are sometimes treated as predicates (Heim and Kratzer 1998). As such, they are similar to adjectives. Consequently, one could follow Prince and relate the string of a-Noun-a-Noun to the one of a-Noun-a-Adjective as in (68b).

However, there is one obvious difference between (68a) and (69a)–(69b) versus (68b) and (71): the overt head noun is part of the second nominal in (68a) and (69a)–(69b), but it is in the first nominal in (68b), and the examples in (71) have even two overt nouns. In my view, this makes it unlikely that all these cases receive the same analysis; that is, that right adjunction is also involved in both (68a) and (69a)–(69b).^[30] Besides this difference in the occurrence of the noun, consider some other arguments against this alternative proposal.

As pointed out above, right-adjoined elements are typically (syntactically) optional. This means that epes occurs as a separate, “complete” nominal in the alternative account. As an individual nominal element, epes functions as a syntactic argument meaning ‘something’ (72a). This is an element of neuter gender as can be gleaned from (72b):

Now, while this indefinite pronoun is a neuter non-animate element, the second nominal in (70) is a masculine animate one. On the alternative account, this would lead to a morphological and semantic mismatch (see Roehrs [2012] for detailed discussion and more data). As a dependent element, as in my account, epes presumably has different properties.

Turning to another issue for the alternative proposal, Prince (2001: 273) points out that these post-modificational structures can also involve definite elements (73a). Again, some authors separate the two nominals by a comma (73b):

However, as mentioned above, a definite DP cannot follow mayner, and as far as I know, this is also impossible with eyner (and presumably epes). This indicates that all these structures should not receive the same type of account. Finally, the alternative account does not shed any light on the restriction of the second construction to be in the singular only (in fact, some of the proposed related constructions can be in the plural as seen in [73a]). I assume then that unlike the predeterminers, cases like Article–Noun–Article–Noun and Article–Noun–Article–Adjective involve right adjunction of the second nominal to the first (cf. the structure in [70]).

5.4 Postnominal possessives

As briefly mentioned above, possessives can also occur in postnominal position. Falkovitsh (1940: 181, 244) states that possessives are put in this position if one wants to “accentuate” them more (my translation, DR). This positional option is fairly frequent with possessive pronouns (74a) and is occasionally found with proper name and definite DP possessors (74b)–(74c) (for more examples of the latter two cases, see Falkovitsh [1940: 181]):^[31]

Jacobs (2005: 242) argues that these possessives involve their own separate nominals, and they are taken to be in apposition like postnominal modifiers of the type a-Noun-a-Adjective. It is unlikely though that these possessives are right adjoined to their preceding DP. As possessives, they receive a theta-role, and they must be in an A-position when they are base-generated.

Given that these possessives are on the right side of the possessum head noun, it is more likely that they are in a low specifier position and that the head noun underwent partial N-raising (see, e.g., Julien [2005]). If this is on the right track, one could speculate further that this is the position where prenominal possessives move from.

I briefly return to the possessive marker -s, discussed in the first part of the article. It was pointed out there that -s is not possible on postnominal modifiers (75a), but it is possible on postnominal possessive pronouns, both in the second (75b) and in the first construction (75c) (note that the latter involves elided material in the plural):

Recall that we cannot claim that -s can only occur on the head noun of the possessive as that would immediately exclude (75b)–(75c). With the above discussion in mind, we can state that postnominal PPs are embedded below the possessor head noun; that is, they are outside of the possessor DP. The same holds for relative clauses under the Kaynian view. In contrast, postnominal possessives are in a low specifier of the possessor head noun; that is, they are inside the possessor DP. Now, if the possessive marker is sensitive to structure, we can suggest that it cliticizes to the rightmost element of the possessor DP but not material embedded below it.^[32]

To sum up, the second construction can only involve a lower indefinite DP in the singular. As tentatively suggested, this restriction follows from a template. This peculiarity could not be related to some data from German, which involve definite DPs. Furthermore, eyner a Noun, epes a Noun, and the second construction should not be related to structures containing post-modifiers: the former three involve eyner, epes, and the possessive pronoun left adjoined to an indefinite DP, but the latter involve right adjunction of the second nominal to the first. Finally, postnominal possessives are most likely in a low specifier position.