2.1 The case for post-syntactic feature insertion in Nimboran: Noyer (Reference Noyer, Lapointe, Brentari and Farrell1998)

Post-syntactic feature insertion was first proposed in Distributed Morphology by Noyer (Reference Noyer, Lapointe, Brentari and Farrell1998) to account for the verbal morphology of Nimboran, a Papuan language of New Guinea (Anceaux Reference Anceaux1965, Inkelas Reference Inkelas1993). The structure of the Nimboran verb, inasmuch as it concerns us here, is given in Figure 3.

Figure 3.

The Nimboran verbal structure.

Figure 3 Long description

The diagram is a syntactic tree with a diagonal orientation from bottom left to top right. It begins with a node labeled 'root' at the bottom left, connected to a branch that splits into two. The left branch leads to a node labeled 'ASP' in parentheses. The right branch continues upward, splitting into three more branches. The first branch leads to a node labeled with a hash symbol. The second branch leads to a node labeled 'T'. The final branch splits into two, with the left leading to a node labeled with the Greek letter pi and the right leading to a node with male and female symbols combined. Each node represents a different syntactic or morphological element in the structure.

The verbal root is immediately followed by an optional node, which I have called asp, that can host certain particles, masculine object agreement, or the durative aspect marker. Then comes number agreement information. Above this is a node that hosts Tense information, and finally a node that hosts person and gender (masculine or non-masculine) agreement information (Inkelas Reference Inkelas1993).

Long description

(6) Two Nimboran sentences a. ŋgedóu- ke- d- ú draw- DL- FUT- 1EXC ‘We will draw (here).’ b. ŋgedóu- tam- en- t- ím draw- DUR- SG- PRES- 1INC ‘We (two) are drawing.’.

‘We (two) are drawing.’ (Annceaux Reference Anceaux1965: 186, 234)

Nimboran verbs obligatorily agree with their subject, and obligatorily bear tense (Inkelas Reference Inkelas1993). The structure in Figure 3 is pre-phonological; the morphemes can change order at PF to satisfy phonological constraints in the language that do not concern us here (Noyer Reference Noyer, Lapointe, Brentari and Farrell1998). Throughout the rest of this article, I give the pre-phonological morphemes of Nimboran, based on the reconstructions in Anceaux (Reference Anceaux1965) and Inkelas (Reference Inkelas1993). Nimboran roots show alternations, examples of which are shown in Table 4.

Table 4.

Nimboran verbal root alternations (Noyer Reference Noyer, Lapointe, Brentari and Farrell1998)

Table 4 Long description

The table is intended to summarize patterns of verbal root alternations in Nimboran, likely listing root forms alongside their alternated forms across grammatical contexts. However, no rows, columns, labels, or values were provided, so specific alternation types, frequencies, or comparisons cannot be described. Without the actual entries, it is not possible to identify which roots alternate, what triggers the alternations, or whether any trends dominate. If you share the table data or an image of the table, I can produce accurate short and long that covers key patterns and contrasts.

We can see the morphology of Nimboran verbs when the optional node asp is absent (6a) illustrated in Table 5 in the future tense, -d.Footnote ¹²

Table 5.

Subject agreement affixes of Nimboran when ASP is absent

Table 5 Long description

The table is intended to report Nimboran subject agreement affixes in contexts where aspect marking is absent. However, no rows, columns, or values were provided, so specific affixes, categories, and contrasts cannot be described. As a result, it is not possible to identify patterns across person, number, tense, or gender, or to compare forms across grammatical contexts. If you share the table entries, I can produce that summarizes the key forms and any notable regularities or exceptions.

Morpheme denotations: Root -number -tense -person$+$gender.

Noyer (Reference Noyer, Lapointe, Brentari and Farrell1998) argues on both phonological and morphological grounds that the B allomorphs are the elsewhere exponents of the verbal root. Phonologically, the A allomorphs show metathesis of B’s final syllable, and the C allomorphs are B’s ablauted counterparts. Noyer adds that Inkelas notes that the B allomorphs appear in the passive, suggesting that they have the most heterogeneous distribution (Noyer Reference Noyer, Lapointe, Brentari and Farrell1998: 274).

Since the allomorphy of the verbal root is conditioned by the number of its subject, I will represent this with the rules in (7).

Long description

A typeset academic figure numbered (7). The title reads: Noyer's Rules of Exponence for Nimboran verbal roots. Three rules are listed below the title, each on a separate line. Rule one: square root symbol double arrow A divided by [ plus atomic plus minimal]. Rule two: square root symbol double arrow C divided by [ minus minimal]. Rule three: square root symbol double arrow B.

There are three number affixes in Nimboran. The affix -maN only appears in the 1inc.sg; the affix -$^{i}$ (an autosegmental entity whose effect is to induce a palatalisation of nearby segments (Noyer Reference Noyer, Lapointe, Brentari and Farrell1998: 272)) is restricted to plural subjects specified for [$-$participant]; and the affix -k appears everywhere else in the dual and plural for all persons. Noyer’s rules of exponence for the number features are in (8).

Long description

(8) Noyer's Rules of Exponence for Nimboran's number agreement affixes (version 1) [+atomic +minimal] ⇔ -maN divided by [ +participant +author] [-atomic -minimal] ⇔ -i divided by [ -participant] [-atomic] ⇔ -k.

The future tense is exponed by the affix -d. This is followed by the person-gender affix, which corresponds regularly to the person features, except for the affix -um. This includes exponence of [$-$masculine], which is not realised in the plural. Noyer’s remaining rules of exponence are given in (9).

Long description

(9) Noyer's remaining Rules of Exponence (version 1) [-participant -author -masculine] ⇔ -um / [+minimal ] [-participant -author] ⇔ -am [+participant -author] ⇔ -e [+participant +author] ⇔ -ám [-participant +author] ⇔ -u [FUTURE] ⇔ -d.

Noyer’s argument for post-syntactic impoverishment and feature insertion is motivated by the morphology of the verbal roots, number affixes, and person-gender affixes, which surface when certain particles, the plural object morpheme dar, or the durative affix -tam appear in the verbal structure (6b) (Noyer Reference Noyer, Lapointe, Brentari and Farrell1998). Following Noyer, I illustrate this in Table 6 with the durative affix -tam, in cases where the tense is present, -t.Footnote ¹³

Table 6.

Subject agreement affixes of Nimboran when asp is present

Table 6 Long description

The table is intended to list Nimboran subject agreement affixes used when an aspect marker is present, likely organized by grammatical features such as person, number, tense, and gender. However, no rows, columns, or values are included in the provided table data. Because the actual affix forms and their distribution across categories are missing, it is not possible to identify key patterns, contrasts, or exceptions. Any conclusions about which affixes occur with particular persons, numbers, tenses, or genders would be speculative. A complete transcription of the table entries is needed to produce accurate, WCAG-compliant.

Morpheme denotations: Root -aspect -number -tense -person$+$gender.

The verbal morphology that appears in the durative is very informative. First, we can see that the affix -um, which realises the 3$^{\text{rd}}$ person non-masculine, is not restricted to the singular and dual, but can appear in the plural too. Noyer observes that this indicates that it expones the feature bundle [$-$participant $-$author $-$masculine] (10a) with no conditioning environment (cf. (9)), and that the feature [$-$masculine] is impoverished when the feature [$-$minimal] is present (10b), but that this rule of impoverishment is blocked in non-neutral environments.Footnote ¹⁴

Long description

(10) Noyer’s Rules of Exponence and Impoverishment for [−masculine] (final version) a. [−participant −author −masculine] ⇔ -um b. [−masculine] →/[−minimal ].

Second, the appearance of the B root when the subject is singular indicates that the feature [$+$minimal] has been deleted in the presence of the durative aspect (11a) (given as asp, so that the rule applies whenever the Aspect node is present in the structure). This prevents the A root from being inserted because the conditioning environment for the A root is not met, and the B root $-$ now the most highly-specified root $-$ replaces it. The feature [$+$atomic] cannot have been deleted, because it is exponed in the 1inc.sg as -maN (11b) (cf. (8a)).

Long description

(11) Noyer’s Rules of Impoverishment and Exponence for the singular number features (final version) a. [plusminimal] → ɨ divided by [ASP ] b. [plusatomic] ⇔ -maN divided by [ plusparticipant plusauthor].

Third, the appearance of the C root when the subject is dual, indicates that the feature [$+$minimal] has been deleted (11a) and replaced by its polar counterpart [$-$minimal] in the presence of the durative (12).Footnote ¹⁵ This is necessary because the conditioning environment for the C root is [$-$minimal] (7), not [$+$minimal].

This is the crucial data supporting Noyer’s claim for post-syntactic feature insertion: an allomorph known to be conditioned by a feature g appears in an environment where the feature g is not provided by the syntax. It follows that g is inserted post-syntactically. Noyer’s analysis is dependent on the assumption that the C roots are conditioned by the presence of [$-$minimal]. Since the B roots are the default verbal roots, the C roots will only appear in a conditioned environment. If that environment was specified for [$-$atomic], the C roots would appear when the subject was dual in the neutral environments, contrary to fact. Thus, the conditioning environment for the C roots must be [$-$minimal]. Since the C roots then appear with dual subjects when the durative affix is present, Noyer concludes that the [$+$minimal] feature of the dual subject is impoverished and replaced by its polar counterpart [$-$minimal], feeding the presence of the C roots.

Noyer restricts the power of post-syntactic feature insertion by stating that only unmarked values of features can be inserted. Cross-linguistically, dual is known to be marked with respect to plural, which is in turn marked with respect to singular. Given this, [$+$minimal] is the marked value of [$\pm$minimal] in the context of [$-$atomic] (just as [$-$atomic] is the marked value of [$\pm$atomic] in the absence of [$\pm$minimal]). Thus, post-syntactic feature insertion in Nimboran is restricted to the unmarked value [$-$minimal] being post-syntactically inserted in the presence of [$-$atomic].

Support for the post-syntactic insertion of [$-$minimal] comes from the fourth morphologically significant difference found in (6). This is the presence of the -$^{i}$ affix in the dual and plural for all persons. That this affix appears in the 1inc.pl and 2pl when the durative affix is present indicates that -$^{i}$ expones the feature bundle [$-$atomic $-$minimal] (13a) with no conditioning environment (cf. (8b)), and that the feature [$-$minimal] is deleted when the feature [$+$participant] is present (13b), but that this rule of impoverishment is blocked in non-neutral environments.

Long description

(13) Noyer’s Rules of Exponence and Impoverishment for [−minimal] (final version) a. [−atomic −minimal] ⇔ − superscript i b. [−minimal] → ∅ / [ plus participant].

That this affix appears in the dual when the durative is present is further evidence that the feature [$+$minimal] is deleted (11a) and replaced by its polar counterpart in the presence of the durative (12).

2.2. Accounting for the Nimboran data without post-syntactic feature insertion

Suppose we look at the Nimboran verbal morphemes in both environments simultaneously (see Table 7) to deduce the Nimboran rules of exponence.

Table 7.

The subject agreement affixes of Nimboran

Table 7 Long description

The table is intended to list Nimboran subject agreement affixes and how each affix corresponds to grammatical categories such as aspect, number, tense, person, and gender. However, no rows, columns, or values are included in the provided table data. Because the actual affix forms and their category mappings are missing, it is not possible to summarize key entries, compare forms across categories, or describe any trends or exceptions. If the table content is supplied, the can describe which affixes occur for each combination of categories and note any gaps, overlaps, or irregular patterns.

Morpheme denotations: Root -aspect -number -tense -person$+$gender.

Let us retain the B roots as the elsewhere verbal exponents (14c), and retain the conditioning environment for the exponence of the A roots (14a). As Noyer argues, the C roots must be restricted to appearing in a given environment, so that they “beat” the B roots. But here, the decision to consider the Nimboran morphemes in both environments simultaneously results in a crucial difference between the present analysis and the analysis in Noyer (Reference Noyer, Lapointe, Brentari and Farrell1998), because the common feature of the environments in which the C root appears in (Table 7) is [$-$atomic] (14b).

Long description

Figure (14) Revised rules of exponence for the Nimboran verbal roots: a. square root symbol is equivalent to A / [ (ASP) plus atomic plus minimal] b. square root symbol is equivalent to C / [ (ASP) minus atomic] c. square root symbol is equivalent to B. The square root symbol represents the root and the double-arrow symbol indicates equivalence. The rules are listed with lettered points a to c, using bracketed feature contexts and plus/minus notation for features.

Since the B root appears in the neutral environments with a dual subject, we can conclude that there is a rule of impoverishment for [$-$atomic] that applies when the subject is dual, but is blocked from applying when the durative aspect appears in the verbal structure. The rule required is given in (15).

Long description

Item 15: Rule of Impoverishment for negative atomic. The rule is stated as follows. Left side: negative atomic. Right side: negative atomic implies lightning bolt symbol, read as not derivable, followed by a forward slash, then a nested bracketed expression. The nested expression reads: left square bracket, left square bracket, square root of an omitted term, right square bracket, underscore metavariable, right square bracket, plus minimal. The full rule reads: negative atomic implies not derivable from double square brackets containing square root of an omitted term, followed by an underscore metavariable, closed in double square brackets, plus minimal.

Crucially, the node that hosts the durative aspect is not given as part of the conditioning environment. This ensures that when the aspect node is included in the verbal structure, the rule of impoverishment cannot apply.

We also need a rule of impoverishment that applies to either [$+$atomic] or [$+$minimal] when the subject is singular and the durative is present, to explain the presence of the B roots in the structures with the asp node and singular subjects. If this rule applies to [$+$atomic], we cannot explain the presence of the -maN affix marking 1inc.sg subjects, so the rule must target [$+$minimal] (16).

Long description

Rule 16 is titled Rule of Impoverishment for left bracket plus minimal right bracket. The expression below reads: left bracket plus minimal right bracket implies not A S P right bracket plus atomic underscore right bracket. The implies operator is represented by a rightward arrow. The not operator appears before A S P. The terms plus atomic and a literal underscore character follow A S P, all enclosed within a right bracket.

Since [$+$minimal] is deleted in singular environments with the asp node, the affix -maN must expone [$+$atomic], rather than the bundle [$+$atomic $+$minimal], and its conditioning environment is 1inc (17a). [$+$atomic] is exponed by a phonologically null exponent elsewhere (17b).

Long description

Numbered linguistics example (17). Title line in italics: Rules of Exponence for [plus atomic]. Two subparts follow. Subpart (a): [plus atomic] bi-conditional -maN, in the context: right bracket, ellipsis indicating omitted material, right bracket, plus participant plus author right bracket. Subpart (b): [plus atomic] bi-conditional null divided by zero exponence. The rules state that [plus atomic] is realized as -maN in a specified morphosyntactic conditioning environment and otherwise has zero exponence represented by the empty set symbol.

This leaves the puzzle of the remaining number exponents. The affix -$^{i}$ appears when asp is absent and the subject is plural, and when asp is present and the subject is dual or plural. Recall that [$-$atomic] is deleted, asp is absent and the subject is dual (15); since this is the feature that characterises the environments where -$^{i}$ appears, we already have an explanation for its absence here. But a puzzle remains: -$^{i}$ does not appear when asp is absent and the subject is 1inc.pl or 2pl. It follows that there is a rule of impoverishment deleting a feature, because -$^{i}$ does appear with these subjects when asp is present. This rule of impoverishment cannot target [$-$atomic] because the C root appears when asp is absent and the subject is 1inc.pl or 2pl; thus, -$^{i}$ must realise another feature too. The only feature available is [minimal] (18b) (I collapse the optionality of [$+$minimal] or [$-$minimal] to [$\pm$minimal] (Harbour Reference Harbour2003a, Despić and Murray Reference Despić and Murray2018)).Footnote ¹⁶

Long description

Title: Rules of Exponence for the remaining number features. a. Minus minimal maps to a special character, divided by a complex structure involving square root, minus atomic and participant features. b. Minus atomic plus/minus minimal is realized as suffix subscript i. c. Elsewhere is realized as suffix minus k. Italicized text includes 'Rules of Exponence' and 'elsewhere'.

The rule of impoverishment that captures the fact that the affix -$^{i}$ does not appear when asp is absent and the subject is 1inc.pl or 2pl thus targets [$-$minimal] (18a). It follows that -k is the elsewhere number exponent (18c).

Finally we come to the rules of exponence for the person features. These do not need to change from those proposed by Noyer (Reference Noyer, Lapointe, Brentari and Farrell1998) (19), though the rule of impoverishment for the [$-$masculine] feature has to be specified so that it cannot apply when the durative aspect is present (20).

Long description

(19) The remaining Rules of Exponence a. [-participant -author -masculine] DELDEL -um b. [-participant -author] DELDEL -am c. [plusparticipant -author] DELDEL -e d. [plusparticipant plusauthor] DELDEL -\u001b\u001bm e. [-participant plusauthor] DELDEL -u f. [FUTURE] DELDEL -d g. [PRESENT] DELDEL -t.

Long description

(20) Rule of Impoverishment for [−masculine]. [−masculine] → ∅ / [[[√] −atomic −minimal] T] .

By considering the morphemes in all environments simultaneously, it was possible to identify the features they exponed and the rules of impoverishment that must precede exponence, without invoking post-syntactic feature insertion. I now turn to Kiowa.

3.1 The case for post-syntactic feature insertion in Kiowa: Harbour (Reference Harbour2003a) Kiowa is a Kiowa-Tanoan language indigenous to Oklahoma, USA

3.1.1 The structure of the Kiowa agreement prefix

Kiowa agreement prefixes subsume argument information into a single pre-verbal prefix. Kiowa verbs agree for up to three arguments: an agent, a goal, and an object (21).

While these prefixes are famously difficult to decompose morphologically (since many are portmanteaux), there is sufficient morphological evidence to support the structure in Figure 4, which forms a phonological island distinct from the verb (Harbour Reference Harbour2003a, Reference Harbour2008).

Figure 4.

The structure of the Kiowa agreement prefix.

Each node in this tree hosts copies of the person and number features of the argument it agrees with. This is straightforward for the person features, already detailed in section 1.3. The number features of the Kiowa agreement prefixes require a little more explanation. The features that underlie number are again those detailed in section 1.3. Kiowa nouns each belong to one of four Classes, each of which has an inherent number. The inherent number of Class I is “one or two”; the inherent number of Class II is “two or more"; the inherent number of Class III is “two"; and the inherent number of Class IV is “any number" (Harbour Reference Harbour2003a: 551). Harbour notes that these inherent numbers correspond to particular features that underlie number systems more generally. Specifically, Class I bears the inherent feature [$+$minimal], Class II bears the inherent feature [$-$atomic], Class III bears the inherent features [$-$atomic $+$minimal], and Class IV bears no inherent number features at all. The nominal structure in Kiowa is given in Figure 5.

Figure 5.

The nominal structure of Kiowa DPs.

Figure 5 Long description

A plain line diagram showing a syntactic tree with angled connecting lines and text labels. The top node splits into two branches. The left branch splits again into two leaves. The right branch ends at a single leaf. The leftmost leaf shows a root sign placed over the word noun. Below it are two bracketed feature lines: left bracket plus or minus atomic right bracket and left bracket plus or minus minimal right bracket. The middle leaf shows the number sign. Below it are two bracketed feature lines: left bracket plus or minus atomic right bracket and left bracket plus or minus minimal right bracket. The rightmost leaf shows the letter D. Below it are two bracketed feature lines: left bracket u atomic right bracket and left bracket u minimal right bracket.

Figure 6.

The feature bundles in the syntax.

Figure 6 Long description

[ you CL 2π ACC DAT −masculine +animate ] [ him CL 3π ACC +masculine +animate ].

Harbour proposes that “inverse is the realisation of uninterpretable number features on D [...] Uninterpretable number on D is valued by a computation over number and Noun. If [the inherent number] and the [natural] number are consistent, then the values are copied onto D. If they clash, however, copying of all values is impossible and D is valued instead as [inverse]” (Harbour Reference Harbour2003a: 553). The verbal agreement prefixes Agree with the features on D, so inverse marking is carried over into the verbal domain. The number agreement values for the prefixes are given in the final column of Table 8.

Table 8.

Prefix number agreement in Harbour (Reference Harbour2003a)

Table 8 Long description

The table is intended to report results about prefix number agreement from a study attributed to Harbour. However, no rows, columns, labels, or numeric entries are included in the provided table data. Because the underlying values are missing, it is not possible to identify key data points, compare categories, or describe trends. Any interpretation about agreement patterns, frequencies, or differences would be speculative without the actual table contents. Provide the table’s headings and cell values to generate accurate, WCAG-compliant.

The features hosted by the nodes x, y and z in Figure 4 are thus those in Figure 1 and those in the final column of Table 8. We are now ready to turn to Harbour’s (Reference Harbour2003a) argument for post-syntactic feature insertion in Kiowa.

3.1.2 The argument for grouping dual, plural and inverse as a natural class

Harbour’s (Reference Harbour2003a) argument for post-syntactic feature insertion hinges on the need to group dual, plural and inverse marked arguments together as a natural class. This presents a problem, as dual, plural and inverse marked arguments do not have any features in common that would underlie this natural class; the only thing they have in common is the absence of [$+$atomic] (the distinguishing feature of singular arguments) from their feature bundles. Harbour presents three pieces of evidence showing that inverse-marked arguments form a natural class with plural and dual-marked arguments.

[$+$nasal]. The feature [$+$nasal] is an exponent of dual. Consider first the agreement prefixes in Table 9, where either the agent or the goal is dual.

Table 9.

Examples of agreement prefixes where the agent, x, or goal, y, is dual

Table 9 Long description

The table is intended to list examples of agreement prefixes used when either the agent or the goal is dual. However, no rows, columns, or values were provided, so specific prefixes, categories, counts, or comparisons cannot be described. If you share the table entries, I can summarize which prefixes occur for dual agents versus dual goals and note any patterns or exceptions.

3a denotes a 3$^{\text{rd}}$ person plural animate argument, and the asterisks indicate that the prefix lowers the tone of the subsequent verb.

In all of these agreement prefixes, [$+$nasal] has spread to all phonemes (following Harbour Reference Harbour2003a, I have not marked nasality on the vowels). However, when the object is dual, but neither the agent nor goal are likewise, the prefix is only exponed as [$+$nasal] if the agent (in transitives) or agent and goal (in ditransitives) are singular, as in Table 10 (cf. Table 11).

Table 10.

Prefixes with a 3dl object and singular agents and goals

Table 10 Long description

The table’s data is missing, so its subject, variables, and values cannot be described. Without rows, columns, headers, or units, it is not possible to identify key points, comparisons, or trends. Please provide the table contents, including column headings and any totals or notes, to generate accurate WCAG-compliant.

Table 11.

Prefixes with a 3dl object and at least one other non-singular argument

Table 11 Long description

The table’s data is missing, so its subject, variables, and values cannot be described. Without rows, columns, headers, or units, it is not possible to identify key points, comparisons, or trends. Please provide the table contents, including column headings and any totals or notes, to generate accurate WCAG-compliant .

Harbour concludes that plural and inverse agents and goals “block’" prefixes with 3dl objects from being exponed as [$+$nasal]. This is the first piece of evidence that the plural and inverse form a natural class.

Dental codas. The second piece of evidence for grouping dual, plural, and inverse together as a natural class comes from codas. Consider the transitive agreement prefixes in Table 12.

Table 12.

Transitive agreement prefixes

Table 12 Long description

The table is intended to present information about transitive agreement prefixes, but no rows, columns, or values were included in the provided data. Because the table body is missing, it is not possible to describe what variables are listed, how categories are organized, or what patterns or comparisons appear. No key data points, trends, or differences can be identified without the actual entries. If you share the table contents, an accessible summary can describe the categories, any notable contrasts across prefixes, and any exceptions or gaps in the data.

Harbour observes that the agreement prefixes with singular objects uniformly have no codas. The agreement prefixes with non-singular objects, on the other hand, uniformly have codas, except for those with a 3inv object; these have a coda only when the agent is non-singular (inspection of the agreement prefixes shows that the 1exc and 1inc have no morphological number). We therefore have a second piece of evidence where dual, plural and inverse group together as a natural class.

The vowel ɔ. The vowel ɔ only appears in ditransitive agreement prefixes. It almost uniformly denotes 3inv objects, which do not concern us here, and is also seen when the prefix has a 3sg or ‘advanced’ object (the z = {∅} column in Table 13).Footnote ¹⁷ Consider the agreement prefixes in Table 13, which denote ditransitive prefixes whose agent has been obliterated (Harbour Reference Harbour2008: 165).

Table 13.

Agreement prefixes with obliterated agents

Table 13 Long description

The table is intended to report agreement prefixes in relation to agents that are described as obliterated. However, no rows, columns, labels, or numeric values are included in the provided table data. Because the dataset is missing, it is not possible to identify key counts, proportions, rankings, or trends, or to compare categories. Any interpretation about how agreement prefixes vary with obliterated agents would be speculative without the actual entries. Provide the table’s headers and cell values to enable an accurate summary and comparisons.

Harbour notes that when the goal is non-singular and the object is either advanced or 3sg, the vowel ɔ appears. He concludes from this that the conditioning environment for ɔ is a non-singular agent or goal, again supporting the need to group dual, plural and inverse together as a natural class.

Summary. The evidence for grouping dual, plural and inverse together as a natural class is summarised in Table 14 (Harbour Reference Harbour2003a: 559).

Table 14.

Summary of the evidence that dual, plural and inverse form a natural class

Table 14 Long description

The table is intended to summarize evidence about whether the categories dual, plural, and inverse behave as a single natural class. No rows, columns, labels, or values are provided, so the measures being compared and the strength or direction of any evidence cannot be determined. Because the dataset is missing, no key data points, contrasts, or trends can be reported. Any interpretation about support, mixed results, or lack of support would be speculative. Provide the table entries to enable an accessible summary of the main findings and comparisons.

3.1.3 The argument for post-syntactic feature insertion in Kiowa

The data that clinches the case for post-syntactic feature insertion for Harbour is given in Table 15, where the goals of the agreement prefixes are 2sg.

Table 15.

Agreement prefixes where the goal is 2sg

Table 15 Long description

The table is intended to report agreement prefixes in cases where the goal argument is second-person singular. No rows, columns, or values were provided, so specific prefixes, counts, or proportions cannot be described. Because the dataset is missing, it is not possible to compare forms, identify the most frequent prefix, or note any trends across categories. Provide the table entries, including headers and any totals, to enable an accurate accessibility description.

The agreement prefixes in the first row of (15) are used when the agent is absent or 1sg. The prefixes in the second row are used the rest of the time. Harbour argues that it is this second row of prefixes that presents the evidence for post-syntactic feature insertion. The morphological evidence here shows that these prefixes meet the requirement in (14) that either the agent or goal is non-singular: the prefix with the 3dl object is not nasalised, the prefix with the 3inv object has a coda, and the prefixes with the advanced and 3sg objects have the vowel ɔ. The goal in the other:2sg:z prefixes is indisputably [$+$atomic], and thus cannot be the non-singular argument conditioning the morphology.Footnote ¹⁸ It follows from this that it is the agent that is non-singular. The question that now arises is how we group the dual, plural and inverse together as a natural class.

Harbour argues that since non-singularity is the shared property of dual, plural and inverse, they must each bear the feature [$-$atomic]. This is straightforward for dual and plural, which already bear this feature, but not for inverse, which does not. A more complex explanation is required. Harbour proposes the following. First, to capture the fact that the {∅}:2sg:z prefixes are syncretic with the 1sg:2sg:z ones, Harbour argues that when the agent is 1sg, the entire x node is obliterated (22).

Long description

(22) Rule of Obliteration when the agent is 1SG. The notation is structured with a title at the top and a rule line below. The rule is expressed as: open bracket subscript x 1SG close bracket, arrow, special symbol, slash, underscore, open bracket subscript y 2SG close bracket.

Harbour observes that the syncretism between the non-singular agents in the other:2sg:z prefixes cannot be captured by obliteration of the x node here, as this would cause them to syncretise with the {∅}/1sg:2sgz ones. Consequently, he argues that when the agent is non-singular, all the features of the x node, but not the node itself, are impoverished (23).

Long description

Numbered example (23) titled 'Six Rules of Impoverishment when the goal is 2SG', presented as a vertical list of six lettered sub-rules in a serif font. Each rule follows the same format: a bracketed feature label maps to a deletion symbol via an arrow, in the environment of [x blank] [y 2SG]. The six rules are: a. [ANIMATE] → deleted divided by [x blank] [y 2SG]. b. [INVERSE] → deleted divided by [x blank] [y 2SG]. c. [plusminus minimal] → deleted divided by [x blank] [y 2SG]. d. [plusminus atomic] → deleted divided by [x blank] [y 2SG]. e. [plusminus author] → deleted divided by [x blank] [y 2SG]. f. [plusminus participant] → deleted divided by [x blank] [y 2SG]. The title line is set in italics. Feature labels in rules a and b are printed in small capitals. Rules c through f use a plusminus sign before the feature name. The deletion symbol used throughout is a slashed zero-like glyph representing morphological impoverishment.

Once these features have been deleted, the x node is empty, and crucially free to host any feature that one might happen to want to insert. This was precisely what Harbour wished, and he proposed the rule of post-syntactic feature insertion directly (24).

Long description

Numbered example (24), titled in italics: Rule of post-syntactic feature insertion. Below the title, the formal rule reads: null symbol (∅) arrow (→) [minus atomic] divided by [subscript x blank ( ) ], meaning the null element is inserted with the feature [minus atomic] in the environment of a bracket containing a subscripted x followed by a blank position.

Harbour demonstrates that the default number in Kiowa is plural. He shows firstly that sentences with expletive subjects in English have plural (inanimate) agreement in Kiowa, secondly that Kiowa predicates with dummy (i.e., null and adverbial) objects take plural agreement, and thirdly that Kiowa reflexives are systematically realised as 3$^{\text{rd}}$ animate plural objects. That plural is the default number in Kiowa is important for Harbour because (as we have already seen) Noyer (Reference Noyer, Lapointe, Brentari and Farrell1998) argued that when features are inserted post-syntactically, only the default value of the feature is inserted. This restricts post-syntactic feature insertion so that it cannot happen rampantly. Harbour thus argues that [$-$atomic] and [$-$minimal] are the default values of [$\pm$atomic] and [$\pm$minimal], and are therefore the values that participate in post-syntactic feature insertion in Kiowa.

The happy consequence of the post-syntactic insertion of [$-$atomic] is that Harbour can now explain why dual, plural and inverse marked arguments behave as a natural class, which he previously claimed was impossible.

3.2 Accounting for the Kiowa data without post-syntactic feature insertion

Unbeknownst to him, Harbour eliminated the need for post-syntactic feature insertion in Kiowa within five years of publishing his paper. In his thesis, he wrote that “After further thought, I am disinclined to take the position just outlined [...] The introduction of a feature violates inclusivity. Within Minimalism, the issue of whether syntax obeys inclusivity $-$ that is, whether it does anything more than stick things together, move them around, and duplicate them $-$ is major as it speaks to the issue of design optimality. If syntax is not subject to inclusivity, if, instead, it can produce things from nowhere, then the theory of syntax becomes potentially far less constrained” (Harbour Reference Harbour2003b: 76, typographical errors corrected).

In his (Reference Harbour2008) book, Harbour provides a new analysis of the valuation of uninterpretable number features on D (25), such that D is able to host both values of a single feature.

(Harbour Reference Harbour2008: 74)

Harbour writes that:

The consequence of this is that inverse marking now appears whenever D hosts both values of one number feature, and one value of the other. Harbour captures the uniformity of inverse marking with an abstract rule of exponence for inverse (26) (where inverse is a placeholder for a number of phonologically-conditioned exponents).Footnote ¹⁹

Long description

Printed text from a formal academic source. Title line in italics: (26) Rule of Exponence for inverse in Kiowa. Formula line below: [ plus F minus F plus minus G ] double left right arrow INVERSE. The word INVERSE appears in full capital letters. The expression is centered beneath the title line.

As before, inverse marking is carried over into the verbal domain, but we must update the features to reflect the overspecification that triggers inverse. The number agreement values for the prefixes are given in the final column of Table 16, with the instances of inverse in bold.

Table 16.

Prefix number agreement in Harbour (Reference Harbour2008)

Table 16 Long description

The table is intended to report results about prefix number agreement, but no rows, columns, or values were included in the provided data. Because the numerical entries and labels are missing, it is not possible to describe what was measured, identify key data points, compare categories, or summarize trends. If you share the table headings and cell values, I can produce WCAG-compliant short and long that highlights the main findings and any notable contrasts.

Crucially, the four clashes between inherent and natural number that generate inverse marking all involve the valuation of D with [$-$atomic], rendering inverse a member of the natural class that also includes dual and plural, and allowing Harbour to account for the data discussed above without post-syntactic feature insertion.Footnote ²⁰

4.1 The case for post-syntactic feature insertion in Leísta Spanish: Nevins (Reference Nevins2007)

Nevins (Reference Nevins2007) presents an analysis of the spurious se effect (Perlmutter Reference Perlmutter1971, Bonet Reference Bonet1991) found in many Spanish dialects, including Leísta Spanish. The spurious se effect arises as a consequence of the ban on *le lo in clitic doubling sequences (27a). Violation of this ban is avoided by the mysterious appearance of the highly underspecified clitic se replacing le (27b).

Long description

(27) The spurious se effect in Leísta Spanish a. *A Pedro, el premio, le lo dieron ayer. to Pedro, the prize, 3CL.DAT 3CL.ACC gave.PL yesterday ‘To Pedro, the prize, they gave him it yesterday.’ b. A Pedro, el premio, se lo dieron ayer. to Pedro, the prize, CL 3CL.ACC gave.PL yesterday ‘To Pedro, the prize, they gave him it yesterday.’.

(Nevins Reference Nevins2007: 282)

Nevins argues that the spurious se effect arises as a consequence of dissimilation, and accounts for it with a rule of impoverishment that deletes the 3$^{\text{rd}}$ person features from the dative clitic feature bundle when it immediately precedes another 3$^{\text{rd}}$ person clitic (28) (Nevins Reference Nevins2007: 310).

Long description

(28) Rule of Impoverishment for Leísta Spanish: Italicized title followed by a rule/formula. The rule is bracket minus participant, minus author, maps to a slashed symbol, in the context of bracket CL, underscore, ACC, DAT, ellipsis, bracket CL, 3 pi, ellipsis. The title is italicized and the example number is at the left.

There are three 3sg clitics in Standard Spanish, shown in Table 17: the masculine accusative lo, the feminine accusative la, and the genderless dative le.

Table 17.

Standard spanish 3sg clitics

Table 17 Long description

Table values or headers were provided, so the specific categories and measurements for Spanish third-person singular clitics cannot be described. With the data missing, it is not possible to report key entries, compare forms, or identify patterns such as frequency differences or distribution across contexts. If you share the table rows and column headings, I can summarize which clitic forms appear, how they are grouped, and any notable contrasts. Until then, any description would be speculative and may misrepresent the intended information.

In the Leísta Spanish dialect, the genderless dative clitic le is additionally used as a direct object clitic when the referent is animate (Table 18, Bleam Reference Bleam1999: 1). To account for the different distributions of the clitics, Nevins (Reference Nevins2007) proposes the rules of exponence for Leísta Spanish shown in (29). Note that none of the clitics expone the Dative case, which structurally includes the Accusative: [[[nom] acc] dat] (Smith et al. Reference Smith, Beata Moskal, Kang and Bobaljik2019, Zompi Reference Zompì2019).

Table 18.

Leísta 3sg clitics

Table 18 Long description

The table is intended to report information about third person singular clitics in Leísta usage, but no rows, columns, or values were included. Because the data are missing, it is not possible to describe what was measured, identify key figures, compare categories, or note trends. If you share the table entries, I can summarize the main patterns, highlight notable contrasts, and mention any limitations based on the table structure.

Long description

(29) Nevins' partial list of the Rules of Exponence for Leísta Spanish clitics a. [CL 3π ACC +masc −animate] ⇔ lo b. [CL 3π ACC −masc −animate] ⇔ la c. [CL 3π ACC] ⇔ le d. [CL 2π ACC] ⇔ te e. [CL 1π ACC] ⇔ me f. [CL] ⇔ se.

While these rules of exponence capture the direct object 3$^{\text{rd}}$ person clitics and the animate 3$^{\text{rd}}$ person clitics, they incorrectly predict that masculine inanimate indirect objects will be doubled with the clitic lo (as demonstrated in (30)), while feminine inanimate indirect objects will be doubled with the clitic la.

Long description

(30) Clitic doubling of an inanimate Indirect Object in Leísta Spanish Le pusé azucar a-l pastel. 3CL put.PAST.1SG sugar A-the cake.MASC ‘(On it,) I put sugar on the cake.’.

(Nevins Reference Nevins2007: 282)

It follows that Leísta Spanish has a rule of impoverishment that deletes gender features in the Dative (31). Nevins does not discuss this rule, but it is necessary for his account to work.

Long description

Printed text figure from a linguistics document. Numbered item (31) appears at the left. The italicised title on the first line reads: Rule of Impoverishment for gender features in the Dative (Version 1). On the second line, the rule notation reads: female plus male symbol arrow divided by [CL 3π ACC DAT minus animate]. The notation uses a combined female and male gender symbol on the left of the arrow, followed by a null or empty symbol, then a forward slash, then a bracketed environment containing: CL, 3π, ACC, DAT, a blank underscore slot and minus animate.

Once gender has been deleted from the feature bundle, the exponents lo and la are overspecified, so inanimate Indirect Objects will be doubled with the less-specified clitic le. Now consider the data that prompts Nevins to invoke post-syntactic feature insertion. Like the spurious se cases, this is another instance of dissimilation, triggered by neighbouring animate clitics (32).

Long description

(32) Another case of dissimilation in Leísta Spanish a. *Te le di. 2CL 3CL.A gave.1SG ‘I gave him to you.’ b. Te lo di. 2CL 3CL.MASC.I gave.1SG ‘I gave him to you.’ or ‘I gave it to you.’.

(Nevins Reference Nevins2007: 282)

The unexpected interpretation of (32b) is I gave him to you, which should, according to the rules of exponence in (29), be rendered with the clitics te le. The clitic feature bundles of (32a) as they appear in the syntax are given in Figure 6, and the clitic feature bundles as they are when they reach PF are given in Figure 7.

Figure 7.

The feature bundles at PF.

Figure 7 Long description

The image contains two boxed matrices side by side. The first box is labeled 'you' and includes the following features: CL 2\u0003c0, ACC, DAT, -masculine, +animate. The second box is labeled 'him' and includes: CL 3\u0003c0, ACC, +masculine, -animate. The layout visually contrasts feature specifications for the pronouns 'you' and 'him'.

The key difference is that the [$+$animate] feature in the direct object feature bundle for him has been deleted and replaced by [$-$animate]. Nevins argues that there must be an additional animacy dissimilation rule that applies to the second clitic of such sequences (33), followed by a post-syntactic rule of feature insertion for [$-$animate] (34) (Nevins Reference Nevins2007: 282).Footnote ²²

Long description

Title line: (33) Nevins' Rule of Impoverishment for [plus animate]. Rule line: [plus animate] arrow impoverishment symbol (struck-through zero) divided by [CL ... plus animate] [CL ... blank ].

Long description

(34) Nevins' Rule of Feature Insertion for [−animate]. ∅ → [−animate] / [CL … +animate] [CL … ]. The title is above and the rule/formula is below.

4.2 Accounting for the Leísta Spanish data without post-syntactic feature insertion

Post-syntactic feature insertion can be avoided completely with a simple reassessment of the rules of exponence for Leísta Spanish. Consider the alternative rules of exponence in (35), which no longer mention the feature [$\pm$animate] at all.

Long description

(35) Part ial list of the Rules of Exponence for Leísta Spanish clitics (Final) a. [CL 3π ACC plus masculine] ⇔ lo b. [CL 3π ACC minus masculine] ⇔ la c. [CL 3π ACC] ⇔ le d. [CL 2π ACC] ⇔ te e. [CL 1π ACC] ⇔ me f. [CL] ⇔ se.

These rules alone predict that lo and la will always expone 3$^{\text{rd}}$ person clitic feature bundles, so the rule of impoverishment in (31) must be modified so that gender features are deleted when the clitic denotes an animate (36a) or Dative argument (36b).

Long description

(36) Rules of Impoverishment for gender features (Final Version) a. ♀♂ → divided by [CL 3π ACC plusanimate] b. ♀♂ → divided by [CL 3π ACC DAT].

However, if these rules of impoverishment apply when two animate clitics form a doubling sequence, the ungrammatical (32a) will surface, instead of the attested (32b). It follows, then, that the rules of impoverishment in (36) apply after the rule of impoverishment in (33), which deletes the feature [$+$animate] in the second feature bundle. This bleeds the rule in (36), deriving the data in (32). Hence, with the simple removal of [$-$animate] from the rules of exponence, and a slight readjustment to the rules of impoverishment for gender features, it is possible to account for the Leísta Spanish data without recourse to post-syntactic feature insertion.

5.1 The case for post-syntactic feature insertion in Romanian: Farkas (Reference Farkas1990)

The morphological puzzle in Romanian presented in Farkas (Reference Farkas1990) can also be accounted for without post-syntactic feature insertion if the feature inventory and rules of exponence are suitably adjusted. The data that concerns us here is gender agreement. Masculine and feminine nouns in Romanian trigger masculine and feminine agreement with determiners, adjectives and predicative adjectives (37).Footnote ²³

Long description

(37) Masculine and feminine agreement in Romanian a. Un trandafir alb e scump. a.MASC.SG rose.MASC white.MASC.SG is expensive.MASC.SG 'A white rose is expensive.' b. NiÈte trandafiri albi sunt scumpi. some rose.MASC.PL white.MASC.PL are expensive.MASC.PL 'Some white roses are expensive.' c. O garoafă albă e scumpă. a.FEM.SG carnation.FEM.SG white.FEM.SG is expensive.FEM.SG 'A white carnation is expensive.' d. NiÈte garoafe albe sunt scumpe. some carnation.FEM.PL white.FEM.PL are expensive.FEM.PL 'Some white carnations are expensive.'.

(Farkas Reference Farkas1990: 539)

Neuter nouns, however, trigger masculine agreement in the singular, and feminine agreement in the plural (38).

Long description

(38) Neuter agreement in Romanian a. Un scaun confortabil e folositor. a.MASC.SG chair.NEUT comfortable.MASC.SG is useful.MASC.SG ‘A comfortable chair is useful.’ b. NiÈte scaune confortabile sînt folositoare. some chairs.NEUT.PL comfortable. FEM.PL are useful.FEM.PL ‘Some comfortable chairs are useful.’.

(Farkas Reference Farkas1990: 540)

Romanian demonstrative pronouns display similar agreement behaviour. Romanian has four demonstrative pronouns, given in (39).

Long description

(39) The four Romanian demonstrative pronouns a. acesta (ăsta) ‘this.masc.sg’ b. aceÈtia (ăÈtia) ‘this.masc.pl’ c. aceasta (asta) ‘this.fem.sg’ d. acestea (astea) ‘this.fem.pl’.

(Farkas Reference Farkas1990: 540)

When Romanian demonstrative pronouns refer deictically, they agree in gender and number with their referent (40) (Dowty and Jacobson Reference Dowty, Jacobson, Powers and de Jong1988, Farkas Reference Farkas1990: 540).

Long description

(40) Demonstrative pronouns in Romanian a. ăsta e confortabil. this.superscript MASC.SG is comfortable.superscript MASC.SG ‘This is comfortable.’ b. ăÈtia sînt confortabili. these.superscript MASC.PL are comfortable.superscript MASC.PL ‘These are comfortable.’ c. asta e confortabilă. this.superscript FEM.SG is comfortable.superscript FEM.SG ‘This is comfortable.’ d. astea sînt confortabile. these.superscript FEM.PL are comfortable.superscript FEM.PL ‘These are comfortable.’.

As was the case with determiner, adjective and predicative adjective agreement, demonstrative pronouns denoting neuter nouns trigger masculine agreement in the singular (40a), and feminine agreement in the plural (40d) (Farkas Reference Farkas1990: 540). Farkas captures the Romanian data with the exponents in Table 19.

Table 19.

Farkas’ Romanian agreement and demonstrative pronoun exponents

Table 19 Long description

The table is intended to report data on Romanian agreement patterns and the forms used for demonstrative pronouns. However, no rows, columns, headings, or values were provided, so specific measurements, key data points, and comparisons cannot be described. Without the table entries, it is not possible to identify trends, differences across categories, or any notable exceptions. If you share the table data or an image of the table, I can produce accurate short and long that reflects the actual contents.

Table 20.

The copula BE and its subject in a hypothetical relative of English

Table 20 Long description

The table is intended to report how the copula verb “be” patterns with its subject in a hypothetical English-like relative. However, no rows, columns, headings, or values were provided. Because the dataset is missing, it is not possible to identify frequencies, compare subject types, or describe any trends. Any interpretation about agreement, word order, or subject marking would be speculative without the actual entries. Provide the table’s column headers and cell values to generate a complete, WCAG-compliant description.

Farkas assumes that there is one gender feature in Romanian, [$\pm$feminine]. Feminine nouns are [$+$feminine], masculine nouns are [$-$feminine], and neuter nouns bear no gender feature at all. Farkas states that the default value of [$\pm$feminine] in Romanian is negative. This is consistent with the fact that when there is no noun for an adjective to agree with, the adjective is realised with masculine agreement (41b).

Long description

(41) Masculine agreement when there is no noun a. Vrema e noroasă. weather.FEM.SG is cloudy.FEM.SG ‘The weather is cloudy.’ b. E noros afară. is cloudy.MASC.SG outside ‘It is cloudy outside.’.

(Farkas Reference Farkas1990: 543)

To account for this and the neuter agreement patterns, Farkas proposes two post-syntactic feature insertion rules (42).

Long description

(42) Farkas’ Rules of Feature Insertion for [±feminine] a. ∅ → [+feminine] / [N−atomic ] b. ∅ → [−feminine].

(Farkas Reference Farkas1990: 542, presented here in the DM style)

These rules of feature insertion ensure that when agreement takes place, neuter nouns will trigger masculine agreement in the singular, and feminine agreement in the plural. They also ensure that when there is no noun for the adjective to agree with, its gender will be masculine.

However, when there is no noun for the demonstrative pronoun to take its gender from, such as when the speaker doesn’t know what the referent is (43), or when the pronoun denotes an event (44), the pronominal form used is feminine.

Long description

(43) Feminine demonstrative pronouns when the referent is unknown a. Ce e asta? what is this superscript FEM.SG 'What is this?' b. Ce sînt astea? what are these superscript FEM.PL 'What are these?'.

(Farkas Reference Farkas1990: 540)

Long description

(44) A feminine demonstrative pronoun when the referent is an event Petru e acasă, dar asta nu o Ètie decât Maria. Peter is home but this.FEM.SG not CL knows only Mary ‘Peter is home, but this is known only to Mary.’ (idem).

To account for this fact, Farkas proposes another post-syntactic rule of feature insertion, for demonstrative pronouns (45):

Long description

(45) Farkas’ Final Rule of Feature Insertion for [+fem]: Ø → [+feminine] / [Dem ].

(Farkas Reference Farkas1990: 543)

Finally, Farkas demonstrates that her rules of feature insertion apply before agreement takes place. The key data involves a demonstrative pronoun referring to an event that triggers agreement on a predicative adjective (46).

Long description

(46) Mismatched agreement in Romanian a. Petru a acasă. Asta e uluitor / uluitoare. Peter is home this.FEM.SG is amazing.MASC.SG / amazing.FEM.SG ‘Peter is home. This is amazing.’.

(Farkas Reference Farkas1990: 541)

In this example, the demonstrative pronoun is marked as feminine, but masculine agreement is triggered on the predicative adjective. This is only possible in Farkas’ system if the rules of feature insertion apply before agreement takes place; otherwise, we would see the predicative adjective in (46) agreeing in gender with the demonstrative pronoun.

5.2 Accounting for the Romanian data without post-syntactic feature insertionFootnote ²⁴

It is not necessary to post-syntactically insert gender features in Romanian if two gender features are recognised instead of one: [$\pm$feminine] and [$\pm$masculine] (see Matushansky Reference Matushansky, Kim, Kurtz, Hewett and Kasper2020 for a similar proposal with slightly different exponents, or Kramer Reference Kramer2015 for one which involves only one binary feature).Footnote ²⁵ In this system, feminine is denoted by the features [$+$feminine $-$masculine], masculine by [$-$feminine $+$masculine], and neuter by [$-$feminine $-$masculine]. We can see the exponents for the agreement affixes and demonstrative pronouns with the relevant features in the charts in Figures 8 and 9 respectively.

Figure 8.

Agreement affixes.

Figure 8 Long description

The table has two main columns labeled with male and female symbols. The male column is divided into two sub-columns: minus FEM and plus MASC. The female column is divided into two sub-columns: plus FEM and minus MASC. Below these, there are two rows labeled plus A and minus A. Row plus A has three cells: a blank cell under plus MASC, a blank cell under minus MASC and a cell with the marker 'ă' under plus FEM. Row minus A has three cells: a cell with the marker '-i' under plus MASC, a cell with the marker '-e' under minus MASC and a cell with the marker '-e' under plus FEM.

Figure 9.

Demonstrative pronouns.

Figure 9 Long description

The table categorizes Romanian demonstrative forms based on gender and feature values. It has two main columns labeled with gender symbols: masculine (male symbol) and feminine (female symbol). Each gender column is further divided into subcategories: -FEM and +FEM for feminine, +MASC and -MASC for masculine. The rows are labeled +A and -A, representing feature values. The table contains four demonstrative forms: 'ăsta', 'ăÈtia', 'asta' and 'astea'. Under +A, 'ăsta' appears in both masculine categories, while 'asta' appears in the feminine category. Under -A, 'ăÈtia' is found in the masculine category and 'astea' appears in both feminine categories. The table has 2 rows and 3 columns of data excluding headers, illustrating how demonstrative forms vary by gender and feature values.

The syncretism between the masculine and neuter singular is captured by their shared feature [$-$feminine], and the syncretism between the feminine and neuter plural is captured by their shared feature [$-$masculine].

Since these feature bundles do not account for the fact that the default agreement gender is masculine and the default demonstrative pronoun gender is feminine, the exponents cannot be specified for the full feature bundles. In the case of the agreement affixes, the features that are mentioned in the rules of exponence are [$-$masculine] and [$+$feminine] (47), while the features mentioned in the rules of exponence for the demonstrative pronouns are [$+$masculine] and [$-$feminine] (48).

Long description

(47) RoEs for the agreement affixes: [−masculine −atomic] maps to -e, [+feminine +atomic] maps to -ă, [−atomic] maps to -i, [+atomic] maps to ∅. (48) RoEs for the demonstrative pronouns: [+masculine −atomic] maps to ăÈtia, [−feminine +atomic] maps to ăsta, [−atomic] maps to astea, [+atomic] maps to asta.

When there is no noun for the predicative adjective to agree with, the agreement affix will be -{∅}, the masculine singular. When there is no noun for the demonstrative pronoun to agree with, the form will be asta, the feminine singular. This apparent mismatch in agreement will surface when both appear together in the absence of a noun, as we saw in (46).

A reviewer observes that these exponents alone do not capture the fact that neuter is always syncretic with either masculine or feminine; on the contrary, these exponents allow for neuter to be exponed without necessitating syncretism. We need two rules of impoverishment to ensure that neuter is always syncretic with another gender. The first will delete all masculine features in the presence of [$+$atomic] (49a); this accounts for the fact that neuter and masculine are always syncretic in the singular. The second deletes all feminine features in the presence of [$-$atomic] (49b); this accounts for the fact that neuter and feminine are always syncretic in the plural.

Long description

(49) Rules of Impoverishment for Romanian gender features (title in italics) a. [plusminusmascuine] divided by [plusatomic ] b. [plusminusfeminine] divided by [minusatomic ] — a two-item list where each rule shows a feature specification followed by an arrow and a zeroing/impoverishment symbol, within a bracketed environment containing an atomic feature and a blank slot.

By adding another gender feature to the inventory, it has been possible to capture the Romanian data and its interesting mismatches in agreement, without recourse to post-syntactic feature insertion.