In the nomina mobilia the feminine is in –a (1st decl.) and the masculine usually is in –us (2nd decl): serva ~ servus. It seems that the gender assignment depends on the inflexional class but there are masculine nouns in –a and feminine nouns in –us. Moreover, the same meaninig can be represented by nouns of different genders and inflexional class, as almond that is either amygdala (Plin.) or amygdalum (Ovid.).

The Latin grammars ask about a noun: Cuius generis est? (e.g. Prob. in Keil IV: 52-3¹). We can ask ourselves how the gender assignment worked in the minds of speakers and how much the irregularities were important. In order to solve these problems, we thought it was better to look at the statistics and to get a corpus through the CD-ROM of IL².

As a first approach, we examined all of the common nouns of the 1st, 2nd, 4th e 5th inflexional class that begins with the letter a-; then, the nouns were divided into groups on the ground of gender. The corpus is both so large to permit a comprehensive view and so small to allow an analysis of the irregularities per word. We obtained the following data:

• I 333: f 285, m 48
• II 486: m 203, f 16, n 267
• IV 96: m 94, f 2
• V 6: f 6

We do not consider here the nouns, about 10, that give rise to interpretative doubts: they shall be examined afterwards. We included in a separate list those nouns that have different genders (and different inflexional class):

• I m / f 3
• II m / f 3
• II m / n 10
• II f / n 4
• I f / II n 8
• I f / III m 1
• II n / III f 2
• II n / IV m 1
• II m / f / n 1
• I f / II f / n 1
• I f / II m / n 2

As we can see, these nouns are only 36, whereas the ones with one gender are 921.

The nouns were divided into groups on the ground of word-formation and, particularly, on the ground of suffixation. For instance, out of the 94 masculine nouns of the 4th inflectional class, 86 are verbal nouns, that is suffixed in -tus, whereas only 3 are radical ones, like algus; the 2 feminine nouns, like anus, are radical, too.

With regard to the nouns in –o, it is surprising that the neuter ones are the most numerous. Moreover, the feminine nouns are much fewer than the masculine ones (less than 10%). Is the femminine in –o an irregularity?

The masculine nouns of the 1st inflectional class are 48 but, since those ones in in –as (4) and in –es (14) can be considered marked, only the 30 in –a are important for us. Although they are not so many, they are interesting and we must devote our best attention to their analysis.

The nouns with more than one gender (and inflexional class) are neither so many nor always of everyday use; nevertheless, it is important to note the trends they display. For instance, the neuter armentum and arvum in Cicero’s Latin prevail over the feminine armenta and arvae by Plautus, Naevius and Pacuvius.

The aim of present study is to establish that linguistics and psychology can provide the appropriate theoretical constrains in order to explain phenomena, such as the presence of ending-parts that are having the form a suffix but without carrying the semantic instruction of a suffix. These elements are defined as categorical markers in the framework of D. Corbin that we adopt here. These categorical markers appear at the end of loanwords and constructed words, especially when these words are prefixed. The role of categorical marker is synchronically and diachronically important. Synchronically, because the categorical marker introduces loanwords and prefixed words either into a grammatical category or into a referential class, i.e. the class of trees. Diachronically, because the categorical marker smoothed the inflexional system of Modern Greek, especially for verbs and adjectives.

Also, empirical evidence from the field of psychology, support such a view, as they underline the crucial role of the last syllable of a word for memory storage. Furthermore, recent  models that describe organization of knowledge in memory, emphasize the need for induction in a semantic and phonological category for beneficial storage of words in memory.

With present work we support the notion that mental lexicon is a well organized place and the tendency of the cognitive system is to regularize the irregularities. In Modern Greek this is possible by adding categorical markers at the end of words. We tested this hypothesis experimentally and we present preliminary data that are on line with this notion.

Catalan, one of Europe's most vigorous minorized languages, appears to be a very interesting subject when it comes to study verb morphology. Its dialectal variation used to be such that it could help identify the geographical origin of the speaker. Verb morphology in Catalan is quite varied and complex as is shown in the survey Antoni Alcover, one of the fathers of Catalan linguistic studies, conducted with Bernhard Schädel from 1906 to 1928 in 149 points of the Catalan-speaking territory. Its results, La flexió verbal en els dialectes catalans, were published a few decades later (from 1929 to 1932).

Hence, the first-person singular (for example  'I sing') can reveal if the variety used is Central Catalan (canto), North Catalan (canti), Valencian (cante) or Balearic (cant).

Our attention will focus on the Catalan variety spoken on French territory, more precisely on Rossellonese which exhibits a wide range of specific solutions in verb morphology, particularly on the first and second-person singular. Here are a few examples of this particular process: for the first-person sg. crec ('I believe, I think') > creui, ric ('I laugh') > riui, and for the second-person sg. tens ('you have') > tenes and dius ('you say') > diues.

The evolution of this dialect seems to privilege the regularity of the system while trying to get rid of the "irregularities", seeking to attain some kind of equilibrium with new "regular" solutions.

This reconstruction process is similar to what has happened in Occitan where the phenomenon has generated a nearly general homogeneity. Nonetheless North Catalan has not undergone the same drastic transformation and has retained some of its genuine "irregular" solutions.

This paper will aim to examine the diverse regularizations in the verb morphology of this Catalan variety compared to its Occitan neighbour, to determine their geographical extent and their current acceptation, or use, in modern Rossellonese.

Defective verbs are irregular in that native speakers exhibit the inability to produce certain forms otherwise commonly used in the language. For example, there are about 60 Russian verbs traditionally cited as being defective in the 1sg. non-past. Thus, Russian speakers are unable to generate the 1sg. non-past of the verb pobedit’ ‘win’ replacing it by the periphrastic ‘obtain a victory’. In languages like French, Spanish and English such verbs are rarer, but do occur (the imperfect of French frire, the 1sg present of Spanish abolir, occasionally the past participle of English stride). Such verbs pose a particular difficulty for current theories of morphology, especially those relying on defaults, because the default will always generate a form for each verb (Albright 2003, Baronian 2009).

Recent studies by Daland, Sims & Pierrehumbert (2007) and Baerman (2008) on this phenomenon argue against the existence of synchronic motivations for defectiveness, while Albright (2003) equates the phenomenon with hesitation and lack of familiarity. Our paper argues for two points.

First, an original field survey with native Russian speakers shows the synchronic productivity of defectiveness (so to speak). We found that speakers have difficulty conjugating computer technology borrowings from English, following the same conditions as traditional defective verbs. The latter are second conjugation verbs associated with a morphophonemic alternation: the final dental of the stem undergoes palatalization as in, for example, the non-defective verb vodit’ ‘conduct’, whose 1sg. non-past is vožu, to be contrasted with defective pobedit' cited above. In our survey, technological borrowings like upgrade-it' also show defectiveness. This is not to deny the diachronic source of defectiveness, but to argue for synchronic factors maintaining the gaps other than prescriptive ones.

Second, after reviewing the difficulty of accounting for such facts in theories heavily relying on defaults, such as Distributed Morphology and Paradigm-Function Morphology, we offer a synchronic analysis couched in Whole Word Morphology (WWM, Ford, Singh & Martohodjono 1997) explaining the defective cases at hand in Russian, French, Spanish and English. WWM is a word-based theory of morphology relating word forms through Word-Formation Strategies (WFS), which allow speakers to project new words, as needed, from lexicalized ones. Our crucial argument involves detailing independently motivated insertion conditions, which guide speakers in choosing between WFS. In the case of defective verbs, we show that insertion conditions do not allow speakers to choose any WFS as appropriate in precisely the cases where a form is missing, thus leaving gaps in the generated paradigms.

We conclude in two parts. First, we list a certain number of defective verb types, which, we argue, are of a different nature and may be explained by different factors. Second, we offer morphologists an opportunity to reflect on the notion of irregularity in morphology: here, a theory which puts all word-formations at par, giving no different status to regular/irregular morphology, is able, better than default-reliant theories, to account for the irregularity of the paradigm gaps of defective verbs.

In the generative tradition, all irregularity in grammar is relegated to the lexicon (cf. e.g. Borer 1984 and much subsequent work). As a result, all syntactic (and semantic) processes are regarded as highly regular, following universal principles, which allow for variation only within the narrow margin of well defined parameters. Morphology, in this conception, straddles the line between regular and irregular processes, the dividing line being the distinction between inflectional (=regular) and derivational (=irregular) morphology (cf. e.g. Chomsky 1970, 1995 and many others). More recently, however, many of the idiosyncrasies that initially have been assigned to the lexicon have been reanalyzed as regular syntactic or post-syntactic phenomena (cf. to varying degrees e.g. Hale & Keyser 2002, Halle & Marantz 1993, Harley 1995 and subsequent work, Marantz 2001, Borer 2005).

This paper follows the trend of reanalysis and offers a new account for the description of the distinction between nominal root compounds in Germanic vs. Romance languages, thus sharpening the line between regular and irregular processes of word formation. The claim that Germanic languages allow for productive, recursive and compositional root compounding whereas Romance languages do not (cf. among others Roeper, Snyder & Hiramatsu 2002; Delfitto, Fábregas & Melloni 2008) will be addressed and it will be shown that a more careful distinction needs to be made. The German examples in (1) illustrate this:

(1a)	Landkarte	country + map	‘map’
(1b)	Landsmann	country + gen + man	‘compatriot’ or ‘man who loves the countryside’, ‘man who lives in the countryside’, ‘man who advocates for the conservation of the country side’ etc.
(1c)	Landeskirche	state + gen + church	‘national church’ or ‘church that is associated with the country’, ‘church that shows the country’s typical architecture’ etc.
(1d)	Länderspiel	country + pl + match	‘match between two national teams’ or‘game that involves knowledge about certain countries’, ‘game that is typically played in certain countries’ etc.

While the stem Land is the first element in all of the compounds in (1), in (1b) -(1d) but not in (1a) this stem is inflectionally marked for either plural or genitive. Also, the compound in (1a) has only one fixed reading, while the compounds in (1b) -(1d) may have a fixed and drifted interpretation in their preferred reading, but they alternatively also allow for a compositional interpretation. Hence, the compound in (1a) shows the typical properties of Romance compounds and the compounds in (1b) -(1d) show the characteristics of Germanic compounds.

These differences result from two radically different word-formation processes. The compound in (1a) is the result of a lexical process, the compounds in (1b) -(1d) are formed by a regular syntactic process by first merging a stem with a categorizing nominal head (cf. Marantz 2007) that carries  inflectional features and subsequently inserting this piece of (non-maximal) structure into an abstract  clitic position of the compound head (cf. Keyser & Roeper 1992 and Roeper & Snyder 2005). The  difference between lexical vs. syntactic derivational processes of compound formation also explains the  interpretational differences and the differences in recursivity. Combining two stems in (1a) is an  idiosyncratic lexical process the product of which is stored in the lexicon as one word with a fixed and  (conceivably) drifted meaning. Potential recursivity in this case is limited to combining the stem of  such a compound with another stem. This, however, is likewise a lexical process and not a syntactic  one, yielding the same interpretational effects as the compound in (1a). Combining a stem with a  category determining head that carries inflectional features and inserting this structure into the abstract  clitic position of the compound head are two syntactic processes. Each of these is an instance of set merge of a head and a non-maximal projection (cf. Roeper; Snyder & Hiramatsu 2002). Following the  argumentation in Boeckx (2008, 2009) that every second instance of merge constitutes a phase, I will  argue that set-merge of a head and a non-maximal projection below the word-level follows the same  principles as set-merge of a head and its complement above the word-level. In consequence, the  compounds in (1b) -(1d) involve a phase-level at which the non-head, which corresponds to the first  inflected stem is transferred to the interpretational component and is spelled-out. Here the interpretable  features [plural] or [lexical genitive case] (cf. Chomsky 1995) are computed, which leads to the  compositional interpretation of the whole compound. Subsequent lexical drift of a syntactically formed  compound is what is expected of syntactic operations that operate below the word-level, because any  syntactically derived novel compound will be stored in the lexicon thus being immediately available for  drift. As long as inflectional marking is recoverable, however, a compositional interpretation is still  available, which allows for a unified account of a syntactic derivation that proceeds in cyclic phases  above and below the word-level and thus allows for a more narrowly defined distinction between  regular syntactic and irregular lexical processes of word formation.

In this paper, we discuss how, in morphology, notions like 'regularity' and 'irregularity' are problematic in the light of the superlative formation in Dutch. Our suggestion is that the rather rigid notions of 'rule' and 'exception' best be replaced with a more parsimonious model involving only analogical reasoning over a set of stored and categorized linguistic experiences. The formation of superlatives in Dutch is typically seen as the application of the derivational rule 'adjective + -st', as in (1) (cf. Booij and Van Santen, 1998). In a pilot study on a 1.5 million words corpus, we found that almost half of the types consisted of the periphrastic (and stigmatized) superlative, as in (2).

In certain cases we can even find minimal pairs within types, like in (3-4):

Despite its high productivity, the periphrastic superlative is considered to be the exception in the few studies that describe it (Gehlen, 1989; Booij, 2005). Phonological features of the adjective that are typically thought of as determinants for the periphrastic superlative in English (stress pattern, coda of the final syllable, number of syllables, cf. Wick (2005)), only predict 77 percent of the Dutch outcomes correctly. We will discuss several factors above the level of phonology that increase the predictive power of our model, such as syntactic environment, level of conventionalization and register.

Given that the formation of the superlative can be predicted only by a multitude of interacting variables on different levels of the language, and that both patterns are productive, we can doubt how suitable a rule-based account is. Rather, we propose, superlatives are formed analogically; a speaker uses the outcome of the exemplar that shows the most overlap with the item she tries to produce: its nearest neighbour (cf. Daelemans and Van den Bosch, 2005). Rules, on this account, can best be seen as epiphenoma of highly productive clusters of exemplars. Irregularities are simply less productive clusters of exemplars. This paper follows the recent line of research in memory-based models (Skousen, 1989; Bod, 1998; Daelemans and Van den Bosch, 2005), in which a speaker's linguistic knowledge is seen as 'a statistical ensemble of language experiences that changes slightly every time a new utterance is processed' (Bod, 1998).

Irregularity is intuitively obvious but theoretically difficult. If a phenomenon lies outside our generalizations, it may be that we have simply given up too soon. And yet, examples like Russian čelovek ‘person’, plural ljudi ‘people’ are reasonably analysed as being irregular in terms of form. We can establish the extremes: in this instance suppletion shows irregularity, as opposed to regular stems. But we can go beyond and treat irregularity as a gradient notion, which means that we should examine the intervening cases. To establish this full range of possibilities, and to disengage properties which happen to be found together in a given language as opposed to those that must co-occur, a canonical approach proves helpful. In this approach to typology, we first delineate the theoretical space, and only then ask how it is populated with examples. Earlier research has set up the basic criteria for canonical inflection, and there has been more detailed investigation of suppletion, syncretism, deponency and inflectional classes. These different types of non-canonical behaviour are of interest independently, and the descriptive clarity of the canonical approach gives an interesting perspective on their (ir)regularity. However, more complex instances are found when these phenomena interact, for instance, when the same cells in a paradigm are targeted by overdifferentiation and suppletion, as with East Norwegian liten/vesle/små ‘small’. This adjective is evidently shows suppletion, and is overdifferentiated, since it distinguishes definite singular from plural, unlike other adjectives in this dialect.

Returning to our example of suppletion, then, a first way of approaching irregularity is to calibrate from the point where the criteria for canonical inflection converge, and to contrast these differing degrees of irregularity based on the form. A second way to look at irregularity is to turn from issues of form to meaning; in this respect our suppletive example is fully regular. In the canonical case, the meaning of an inflected form is predictable from the combination of the lexical and grammatical meaning, as with our example. Yet the feature number, apparently so simple, shows many instances of irregularity according to this criterion, as with English (coffee) grounds. A third way to approach irregularity is to consider the distribution of these non-canonical behaviours across the lexicon: we ask how many eligible lexical items are involved in potential irregularities, and whether these items form principled groups (such as fitting as segments of the Animacy Hierarchy, for instance). This third way recognizes the intuition that irregularity is contingent, or locally defined. These are in principle three distinct dimensions, yet our terminology does not draw fully consistent distinctions. Separating out these dimensions of irregularity throws into sharp relief the particular challenge of the most complex examples.

The standard assumptions about inflectional morphemes are that:

1. They can be added to specific classes of lexical categories
2. They carry a specific semantic function or a set of semantically related functions

Our aim in the present study is to demonstrate that there exists a class of morphemes that:

1. can be added to any free lexical or grammatical morphemes and
2. that does not carry any specific semantic function.

The broad function of those morphemes is to allow the listener to parse an utterance into constituents. We demonstrate the existence of morphemes that do not carry any specific semantic function in two languages, Wandala (Central Chadic) and Kabyle (Berber). We demonstrate the existence of morphemes that can be added to any lexical or grammatical category in Wandala.

In Wandala, a lexical item or a grammatical morpheme may have three forms: root, root + a, and root + high tone. The last two morphemes viz. a and the high tone can be applied at the same time resulting in the form á added to the root. The root form, indicates that the ensuing  morpheme constitutes a natural follow-up of the preceding morpheme. The root + a indicates that the ensuing lexical item is a less expected follow-up for the given lexical item. The root + á indicates that the ensuing lexical item is an unexpected follow-up but should nevertheless be interpreted in connection with the preceding morpheme. The three types of connection entail narrower semantic functions determined by the lexical category and semantic class of the preceding and ensuing lexical item.

The root form of the noun preceding another noun entails kinship relation; the root form of the verb followed by a noun entails direct object function of the noun; any lexical category has to have a root form before a preposition.

The root + a form of the verb preceding a noun phrase entails nominal subjects of intransitive and transitive verb with no object in the proposition; root + a form of any morpheme followed by a noun phrase entails the subject role of the noun phrase.

The  root + á of the verb entails controlling subjects of inherently intransitive verbs whose inherent argument is affected. Root + á form on the noun entails  genitive non-kinship relationship with the ensuing noun.

In Kabyle, the forms in question relate to the properties of nouns. The unmarked form of the noun, labeled ‘annexed’ in traditional grammars, indicates that the noun should be interpreted in connection with the preceding lexical item, be it a verb, another noun, or a preposition. The unmarked form of the noun may entail: genitive relationship with the preceding noun; subject relationship with the preceding verb; object relationship with the verb that has object pronouns; modifier of a preceding numeral ‘x number of NP’. The marked form of the noun, labeled ‘absolute’ in traditional grammars, allows an interpretation of the noun independently of the preceding lexical item. The noun in the absolute state can be either the subject or the object, among other functions. Berber languages that lost the formal distinction between the absolute and the annexed states, grammaticalized the linear order and  prepositional marking as compensatory means.

The existence of morphemes whose function is to code a type of syntactic connection raises a question whether the notion of regularity applies to all classes of morphemes or only to those that occur with only one class of lexical items and which indeed carry specific meaning.

This paper examines the degree of regularity-irregularity and the complex morphological type of the personal pronouns, in particular in the Germanic languages, but also in other languages such as Japanese.

The morphology of the pronouns with considerable irregularity is not accounted for by standard morphology which generally speaking concentrates on regularity. This paper, as well as examining personal pronoun morphology in detail, attempts to account for this complex morphology.

A fundamental characteristic of the personal pronouns in the Germanic languages – and a central one for personal pronoun morphology – is that they are short. A main function of personal pronouns and other pro-forms is to abbreviate – i.e. it is the rationale of the personal pronouns to be (relatively) short. In addition, the personal pronouns are generally among the most frequent words in the Germanic languages – virtually all the subjective and objective pronouns in English for example occur in the first one hundred most frequent words in speech. One of the most obvious consequences of high frequency is that a frequent form is more likely to be short rather than long. However, it is important to distinguish between shortness and ambiguity – the relevant factor is not how short a personal pronoun is, rather whether or not it is ambiguous. Ambiguity as a factor in change in the personal pronouns will also be taken up in the paper.

Theoretically, the morphology of the personal pronouns is analysed as representing two different systematic types: either systematic in terms of marking property connections or systematic in terms of marking property differences. Either on the one hand representing properties by morphological patterning, i.e. shared properties are indicated by shared formatives; or, on the other hand, marking differences in property by suppletion – i.e. a personal pronoun is morphologically distinct from other pronouns with which it shares a property or properties.

The paper also discusses how accented and unaccented forms of the same pronoun can vary in their connection to one another. Not only can the personal pronouns show suppletive or suppletive-like distinctions between separate pronouns, i.e. not derivable by general synchronic rule, but also non-synchronically-derivable variants of the same pronoun may occur.

The morphology of personal pronouns is in many cases grammatically, semantically and formally complex. The personal pronouns are primarily representative, portmanteau forms rather than active indicators of each category/property: one personal pronoun cannot usually be derived from another just as one lexeme cannot usually be predicted from another. The personal pronouns are generally (co)referring terms, both grammatically and semantically to the external world – (in their core meaning) I = the speaker, we = a group to which I belong, du = the addressee, he = the male person, er = the grammatically masculine referent etc. – and therefore it is perhaps not surprising that also formally the personal pronouns show similarities both with inflectional morphology and with the lexicon. This duality will be further discussed in the paper.

Finally, the author will examine to what extent the factors discussed can be applied to comparable paradigms in other languages – the irregularity of the personal pronouns and of comparable forms in many languages suggesting some common factors.

The contribution will first aim aim at discussing some basic knowledge around the term irregularity, mostly in relation to the different German counterparts or concepts like Irregularität, Unregelmäßigkeit, Regellosigkeit, Ausnahmslosigkeit, Wahrscheinlichkeit etc.; it will moreover evaluate the question on what basis we classify systems as 'irregular' or just 'different'. It will mostly deal with the impossibility of irregularity in phonology (unless in the sense of 'Regellosigkeit' [lack/infringement/violation] of rules) based on the fact that phonologies must guarantee pronounceability and perceptibility of language.

In addition, the presentation will discuss the manifestations of a regularity-irregularity scale in the formation of grammar and resulting principles (cf. well introduced concepts like markedness, naturalness, complexity, prototypicality etc.). Such structural (typological) principles in phonology create preferences. I will advocate that we rather need a principled understanding of infringement of principles in order to understand the system itself.

A dialectic view on phonology will show that the typological perspective might simply offer a different position. In prosody, e.g., we do have well-established preferences for syllable structures, but on the other hand the principles violating these preferences do follow a proper Gesetzmäßigkeit (regularity, probability). It will turn out that infringements are not infringements but regularities, and not only that we cannot think irregularity (in the sense of Kant) as any reflection on irregularity is a search of / reflection on regularity, but that, moreover, the apparent irregularity itself may be type-constituting.

The goal of this paper is to analyze the data on suspended affixation (SA) in Turkish that has been previously ignored in generative literature. SA is a situation in coordinated constructions when two (or more) members (conjuncts) can potentially be affixed (in 1a), but only the rightmost is (in 1b). The affix appended on the last conjunct has scope over all the preceding conjuncts:

Although very prominent in Turkish, this phenomenon has only recently brought up to light in the generative framework (Kornfilt 1996, Orgun 1996, Kabak 2007, Hankamer 2008). Kabak proposes that there is a well-formedness constraint for SA to be realized : non-final conjunct has to be morphologically a well-formed word; while Kornfilt and Hankamer relate SA to coordination of complements of a functional head. However, none of these analyses treats specifically nominal compounds and none of them takes into account significant variation among speakers in their acceptability judgments. Not all the native speakers accept SA in all coordinated nominals: compounds (in 1b) versus possessive compounds (in 2).

It is often assumed (van Schaaik 2002) that in possessive compounds CM is deleted when the possessive suffix is present. Therefore, when possessive suffix is suspended (in 2), nothing appears on the bare conjunct ‘yer’. However, by bringing new data collected from a range of Turkish native speakers (interviews and written questionnaires) we show that this is not the most common and regular pattern. In fact, there are two types of SA for possessive compounds : total (in 2) and partial (in 3).

Partial SA occurs when CM stays on the first conjunct when the possessive suffix is suspended. Contrary to the common belief in generative theory (Orgun 1996, Kabak 2007), Turkish speakers use partial SA (in 3) as a default SA pattern, and not total SA (in 2).

Results suggest that social and stylistic factors do not appear to condition the use of SA, but there exists an implicational hierarchy of grammatical conditioning such that for total SA in possessive compounds (in 2) to be allowed SA in compounds (in 1b) needs to be accepted. Speakers with no SA in compounds (in 1a) will accept only partial SA in possessive compounds (in 3). Therefore, since most speakers accept partial SA, but not total SA, partial SA is argued to be an unmarked SA pattern while  the total SA pattern is more marked.

1. Introduction:

Based on original fieldwork, this study examines unexpected double person agreement in Thompson Salish clefts. Double agreement results in a mismatch between syntax and focus semantics, but a match between syntax and the ordinary (non-focus) semantic denotation. As such, it points to a post-syntactic LF level of cognitive processing (Heycock and Kroch 1999).

2. Clefts:

Thompson Salish is a severely endangered language in southwestern Canada; only a few hundred elderly speakers remain. Clefts mark focus on DP arguments, and are biclausal (Kroeber 1999). In (1), the focus Bill is base-generated as the head of the cleft, and is followed by a cleft residue clause containing the backgrounded information that ate the food.

(1)	c’é	[he Bíll]focus	[e	ʔúpi-t-mus	tx	e	sɬaʔxáns]background.
	CLEFT	DET Bill	COMP	eat-TRANS-SUBJ.GAP	tx	DET	food
	‘It was [Bill]focus [that tx ate the food]background.’

Special -mus morphology indicates a subject gap txin the residue clause; standard lambda abstraction gives us its semantic representation λx.x ate the food(Heim and Kratzer 1998). This clause refers to a semantic set of focus alternatives where x corresponds to different possible persons that ate the food (2) (e.g. Rooth 1992, Krifka 2006). Crucially, x has no person features. The syntactic residue clause thus straightforwardly maps onto the focus semantic set of focus alternatives in (2).

(2) Focus alternatives of (1): {Pam ate the food, Bill ate the food, Sue ate the food, ...}

Under the Structured Meaning approach to focus semantics (e.g. Krifka 2006.), Focus and Background are represented as <F, B>. In (1), the syntax straightforwardly gives us the focus semantic pair < Bill, λx.x ate the food>.

3. Problem:

When the focus is 1st or 2nd person, person agreement is optionally marked twice. In (3), the focus is the clefted 1sg pronoun ncéwe÷, as expected. But, the backgrounded clause on the right also carries 1sg subject agreement -ne on the verb ‘eat.’ A residue clause syntactically marked for 1sg subject cannot map straightforwardly into the set of focus semantic alternatives ranging over any subject person (2).

(3)	c’é	[ncéweʔ]focus	[e	ʔúpi-t-ne	tx e sɬaʔxáns]background.
	CLEFT	1SG.EMPH	COMP	eat-TRANS-1SG.SUBJECT	tx DET food
	‘It was [me]focus [that tx ate the food]background.’
	(more literally: ‘It was me that I.ate the food.’)

4. Proposal:

Structured Meaning proposes a function that maps Focus into Background, B(F). This produces the ordinary meaning of the sentence in (1), Bill ate the food. This level of meaning is not straightforwardly represented by the syntax of the cleft, but is a further level of LF processing. I propose that the same level of LF processing enables the syntax to “look ahead” and realize 1sg person agreement marking inside the cleft residue clause in (3), since the ordinary semantic meaning is processed as I ate the food. The syntax thus interfaces well with the ordinary semantic meaning, but not with the focus semantic alternatives.

Other explanations seem inadequate, as double agreement occurs at higher rates than expected for speech errors. Furthermore, this processing is fully grammaticalized in neighboring Halkomelem Salish (Gerdts 1989).

This talk sheds light on the the correlation between defectiveness and valence changing operations. I examine the productivity of valence changing manifested in Hebrew by relations among prosodically distinct configurations, called binyanim (e.g. niCCaC, hitCaCeC). Decausativization is an operation that derives decausative verbs (e.g. hirtiv ‘make wet’ à nirtav ‘become wet’), by eliminating an external theta role of cause (Reinhart 2002). Following Reinhart & Siloni (2005) I assume that decausativization applies in the lexicon. I address the productivity of decausativization with regard to morpho-phonology. I argue that the latter restricts the application of valence changing, thereby creating pradigmatic defectiveness.

In contrast to other operations (e.g. reflexivization), decausativization is relatively productive and its morphology can be predicted. However, some transitive verbs demonstrate an intruiging morphological behavior with regard to decausativization. I argue that this results from their irregular morpho-phonology, as most of them of them are stems with only two consonants. These verbs demonstrate three distinct patterns.

1. Defectiveness: Some verbs simply do not have decausative counterparts (he'ik ‘oppres’). I argue that their irregular morphology blocks the application of decausativization. Examining their thematic grids does not exaplain why they do not undergo this operation, as there is no observed difference with other verbs that undergo this operation. This results in defectivness of the transitive-decausative derivational paradigm.

2. Melodic overwriting: Verbs can undergo decausativization that is manifested by melodic overwriting (e.g. hetiš ‘exhaust’ à hutaš ‘become exhausted’). The vocalic pattern of the verbs change into u-a, similarly to the passivization (e.g. siper ‘tell’ à supar ‘be told’). This results in the unification of the form some passive and decausative verbs. The verb huvax ‘be/become embarrassed’, for instance, can be interpreted as both passive and decausative.

3. Lexicalization: Other decaussative verb are formed by reduplication of the second stem consonanat (e.g. hit'orer ‘wake up’), while others occur in irregular templates (e.g. nizok ‘get damaged’). Their morphology is cosidered exceptional and unproductive in terms of innovation. I assume such forms are lexicalized and their formation is not a part of the morphological component.

The morphology of decausativization demonstrates three distinct patterns with regard to defective verbs. These patterns of irregular morphology and morphological blocking reveal the role of morphology as part of an active lexicon and its interaction with thematic operations. It provides an account for the gaps and irregularities within derivational paradigms. The analysis also gives rise to a surface-based account, in which forms are derived from actually occurring words (Aronoff 1976), rather than a system in which forms are derived by relating to an entity that never occurs in isolation on the surface (Bat-El 1994, Ussishkin 2000). If we assumed that such decausative verbs are derived on the base of roots, there would be no reason for their relatively low productivity. Time permitting, I will discuss defective paradigms of other valence changing operations.

Among the Slavic languages, Polish has one of the most complex inflectional systems. Almost all the desinences inherited from Common Slavic have been preserved in the diachronic development up to now, whereas in other Slavic languages fundamental simplifications have been taking place. In order to manage the abundance of formally different markers, paradigms were restructured according to formal and semantical criteria: gender and animacy; morphological structure of the nominative singular (word based / stem based inflection, cf. Wurzel 1984); phonological features of stem final consonants. In some cases, it has been suggested that certain markers are assigned in accordance to other (non-nominative) forms: e.g. the vocative sg. is nauczycielu ‘teacher’ since the locative sg. is also o nauczycielu ‘about the teacher’ – this is a very rare paradigm structure condition in the Slavic languages. As these criteria intersect, the number of different paradigms is simply impressive. This paper will try to cast some light on principles and restrictions of the assignment of markers based on phonological criteria. Especially we will discuss the supposed assignment of markers in reliance on other (non-nominative) desinences (cf. W. U. Wurzel 1984: Flexionsmorphologie und Natürlichkeit. Berlin, p. 122).

In Preparation

This paper proposes an analysis of Italian verb conjugation within a realizational (Word and Paradigm) model of morphology. One of the main claims of our approach is that irregular verbs are not a more complex case of regular verbs, but rather that the latter are a simplification of the former. This means that a verb (and in general any inflected lexeme in a language) is planned to reach a certain degree of complexity, but no verb actually displays the maximal complexity. In fact, the majority of verbs (the regular ones) display the minimal complexity. This conclusion is based on the simple observation that all verbs only display a small subset of the conceivable configurations of allomorphy distribution. Another important claim is that no lexemes (not even the regular ones) have a unique basic underlying form from which all the other forms can be derived. Rather, an inflectional paradigm can be viewed as a network of connected word forms. The way in which suppletion distributes allows to identify some zones within paradigms in systematical covariation. These zones correspond to stems in traditional morphological analysis, and their distribution is called a “Stem Space”. Stems are linked one to the other and to word forms, which are the actual input for syntax, by functions. These functions may correspond to a simple identity relation or to more or less motivated phonological processes. Each relation between two stems or between a stem and a form may be expressed by one or by a series of functions. In this case, the functions may be either mutually exclusive or hierarchically organized: the application of a general function may be blocked by a more restrictive function.

Since stems are expected to vary according to constant functions, the spreading of the same stem throughout the whole paradigm of a lexeme, often considered as an important property of regular lexemes, is just a possibility which may not correspond to a regular lexeme. Accordingly, the identification of inflectional classes, seen as a diacritic assigned to individual lexemes is no more necessary. All information relative to conjugation can be inferred by the phonological form of the stem in those cases of the paradigm which display non-ambiguous forms. This conclusion is consistent with the idea that the explicit marking of inflectional classes is necessary only when variation in the inflection of a lexeme cannot be entirely predicted from its phonological form, and with the observation that in Italian (and in other Romance languages), unlike in classic examples of inflectional class variation (such as e.g. nouns in Latin), verbs display a large amount of redundancy in their endings.

The notion of regularity in language and the interrelationship with irregularity as its counterpart deserves to be investigated within the framework of second language acquisition: SLA studies have proved to have a high degree of effectiveness to test language theory  since a learner variety can be conceived as a sub-system, sharing features with the source and the target language and manifesting preferences to be interpreted according to naturalness (see studies by Dressler) and typological principles (such as the Markedness Differential Hypothesis, Eckman, 1977 and 1984).

Phenomena attested in L2 varieties, though irregular if compared to the standard variety of the target language, present an inner regularity: regular variants occur in the production of non-native speakers at different stages (pre-basic, basic, post-basic variety)  along the process of acquisition (see The structure of learner varieties in European Science Foundation Project, in Perdue, 2000).

In this view, irregularity is but a relative notion: irregularities in non-native speech are such if compared to the standard L2 variety, whereas if considered as peculiar features of a non-standard variety they might be interpreted as regular markers.

The acquisition of Italian as L2  has been investigated in different research studies within the ‘Pavia project’ during the last decades (see Giacalone Ramat,2003) at different level - morphology, syntax and text organization –revealing common sequences in the acquisition of Italian as L1 and as L2 (the speakers’ source language not being relevant). Phonetics of Italian as L2 of Moroccan-Arabic speakers has been described in terms of presence of common phonetic variants: a core of features in interlanguage phonology shapes a non-native variety (Moroccan-Arabic accented Italian) which might be considered a sociolect within the Italian diasystem. (Mori, 2007).

As for Italian, phonology is a quite unexplored area in the field of SLA; interface of prosody with other linguistic levels of analysis is almost unknown. Therefore, our aim is to investigate in a functional perspective marked morphosyntactical structures being used by non-natives to focalize a cataphoric element. In order to substantiate our hypothesis of focalization strategy, the morphosyntax-intonation interface of some peculiar structures (such as ‘averci’ + direct object pronoun + O) will be analized:  suprasegmental features (especially intonation contours) provide an additional informative load, especially when the pragmatic mode is still predominant in interlanguage and the lexical and grammatical competences have not yet been developed (a continuum from pragmatics to syntax, see Gìvon, 1979).

Such non-native (ir)regularities feature the interlanguage at different levels (i.e. phonetic, morphological, syntactical, markers) and can be considered a result of either L1 interference, input language pressure (i.e. colloquial, popular regional, dialectal varieties of Italian) or phonological/morphological naturalness and typological markedness.

In order to lead our research study, original data will be extracted from a corpus of Italian as L2 (ArabIt) consisting of semi-guided recorded interviews with 28 Moroccan speakers belonging to a different socio-cultural background, immigrated in Italy (Lazio region), having acquired the Italian language in a natural context without experiencing any kind of formal foreign language learning.

In Preparation

Introduction

Inflectional morphology is an area of language that is frequently affected in acquired or developmental language disorders. Typically, such deficits do not affect all inflected forms in uniform, but operate quite selectively. A distinction that has proven to be of relevance is the distinction between regular and irregular inflectional morphology. According to dualistic approaches to inflection, regular and irregular inflection are fundamentally different: regular inflected forms are built by application of a mental symbolic rule, whereas irregular forms are stored as fully inflected word forms in the mental lexicon (cf. Pinker 1997). Ullman (2005) has suggested that regular and irregular inflectional morphology are subserved by different brain systems. According to his procedural/declarative model, regular inflection is part of the rule component of grammar, critically subserved by left frontal brain regions (most importantly Broca’s area), whereas irregular inflected forms are stored in temporo-parietal memory systems. Based on this model, Ullman has claimed that Broca’s aphasia that is caused by brain lesions affecting left frontal brain regions leads to deficits with regular inflection, whereas Wernicke’s aphasia, typically associated with temporo-parietal brain regions, affects irregular inflection.

Method

We report data on the production of regular and irregular inflected past participles and noun plurals that come from eight German subjects with Wernicke’s aphasia and eleven German subjects with Broca’s aphasia. In the past participle task, subjects were asked to transform a given 1st person singular verb form into a participle. We elicited 39 regular and 39 irregular participles carefully matched for frequency (lemma and participle frequency). In the plural task, we tested regular and irregular German -n-plurals for which a similar difference between regular -nfem-plurals (-n-plural on feminine nouns ending in schwa) and lexically stored irregular -nnonfem-plurals (-n-plural on masculine/neuter nouns not ending in schwa) holds (Penke & Krause 2002). Here, the task was to transform a given singular noun into the respective plural form. We elicited 40 noun plurals with matched plural and lemma frequency, 20 -nfem-plurals and 20 -nnonfem-plurals.

Results

Figure 1 presents the correctness scores for regular and irregular participles and noun plurals obtained from the two groups of aphasic subjects. The figure reveals that both subject groups produced significantly more suffixation errors with irregular than with regular inflected participles and noun plurals.

Participle error rates in %	-n-plural error rates in %

Fig. 1: Error rates for participles and –n-plurals

Moreover, we observed that in both groups of subjects and in both inflectional systems error rates for irregular inflected forms were affected by word form frequency. Both subject groups produced significantly more errors with infrequent irregular inflected forms than with frequent irregular inflected forms. In contrast, error rates for regular inflected participles and noun plurals were not affected by word form frequency (cf. fig. 2).

Frequency distribution of errors with participle inflection

Frequency distribution of errors with –n-plurals

Fig. 2: Frequency distribution of error rates

Discussion

We observed deficits with accessing irregular inflected German word forms that are independent of the type of aphasic syndrome and the underlying brain lesion. The observation that error rates are influenced by the frequency of the irregular inflected form suggests that difficulties in both aphasic syndromes are due to problems in accessing irregular infrequent forms stored in the mental lexicon. This finding is incompatible with the procedural/declarative model according to which the different types of aphasic syndromes should lead to different impairments with regular and irregular  inflectional morphology (cf. Ullman et al. 2005). We will discuss the importance of this and related findings for Ullman’s model on the neuronal foundation of inflectional morphology.

Tagalog, a language of the Austronesian language family under the Malayo Polynesian subgroup is one of the major languages spoken in the Philippines based on the number of speakers. This language has a rich morphology, particularly with affixes that indicate the focus and aspects of the verbs yet behind this richness, irregularity befalls. Aspects, as Comrie (1981) defines are the different ways of viewing the internal temporal constituency of a situation while focus, as defined by Motus (1971) is the grammatical relationship between a verb with its affixes. This paper will deal basically with familiar irregularities in some common Tagalog verbal affixes.

Like any other language, Tagalog uses various affixes in order to show the focus and aspects of verbs which later on leads to the formation of the verbal conjugations. However, an irregularity that can be seen in Tagalog is the use of more than one affix to indicate the actor focus of the verb. These affixes may either be the prefix mag- or the affix –um-, which may function as prefix for verbs beginning with vowels (e.g. umawit ‘to sing or sang’, umalis ‘to leave or left’, umikot ‘to turn or turned’, etc.), and as infix for verbs beginning with consonants (e.g. kumuha ‘to get or got’, sumulat ‘to write or wrote’, kumain ‘to eat or ate’, etc.). Another complexity is noticeable with the examples given for the affix um-, the same form of the verb is used to show the infinitive, imperative and completed aspects.

Moreover, there is a difference with the way of writing the verbs with the prefix mag-; for verbs beginning with consonants, the prefix is simply connected with the root (e.g. magdala ‘to bring’, magluto ‘to cook’, maglinis ‘to clean’, etc.), while for verbs beginning with vowels, a hyphen is placed after the prefix, (e.g. mag-alay ‘to offer’,mag-isip ‘to think’, mag-usap ‘to talk with someone’, etc.) which gives a slight pause in pronouncing such words. And in connection with the aspects of the verbs, mag- denotes the infinitive, imperative as well as to indicate the proposed aspect of verbs with the initial CV reduplication for the third.

Surprisingly, both of these actor focus affixes may be used interchangeably with some verbs without altering their meanings and aspects, (e.g. magsulat or sumulat ‘to write’, magbasa or bumasa ‘to read’, magpunta or pumunta ‘to go’). Nevertheless, there are cases wherein some semantic changes happen when both of the affixes are used with certain verbs, (e.g. bumili ‘to buy’ and magbili ‘to sell’, umuwi ‘to go home’ and mag-uwi ‘to bring home’, tumayo ‘to stand’ and magtayo ‘to build’, etc.).

Another verbal affix to be discussed is the Tagalog object focus affix –in¸ which has a similarity with the affix –um that functions as prefix for verbs with initial vowel (e.g. inamoy ‘smelled’¸ iniba ‘changed’, inalis ‘removed’, etc.), and as infix for verbs with initial consonant (e.g. binuhat ‘lifted’, ginamit ‘used’, kinain ‘ate’, etc.). These examples show the completed aspects of verbs but if the infinitive, imperative and proposed aspects are to be formed, -in becomes a suffix, (e.g. buhatin‘to carry’, alisin‘to remove’, gamitin‘to use’, etc.). Furthermore, there are irregular verbs in Tagalog, those beginning with l and y wherein this affix is metathesized when used as prefix, (e.g. nilipad ‘flew’, niyakap ‘embraced’, niluto ‘cooked’).

Irregularities in grammar, perhaps, are “regular” with most of the languages, just like in Tagalog morphology, sometimes considered as complexities yet distinct characteristics that make each and every language unique.

I will consider in this talk the many causes which may lead to irregularity in language, from analogy  to  paraetymology and so on. I have chosen as starting point the analysis of the so-called  Sturtevant’s paradox, since it implies more levels than its bare formulation seems to refer to. The  well-known paradox says that sound-change is regular but creates irregularity, while analogical change is irregular but creates regularity. Sturtevant’s formulation implies  two concepts located at different levels, the phonological level and the morphological one —and, above all, it entails two (or more) different chronological levels. A new rule B can act on the outputs of rule A (see, e.g., the Sanskrit palatalization of velar consonants). Remnants of the phase A, untouched by the phase B, become irregular exceptions.

Irregularity, markedness, frequency, and complexity.

Irregularity and analogical change, which is typically local. “This is because analogical innovations are abductive: the relation and the result are given but the specific case is inferred” (Bynon 1994:111). Examples of analogical intrusion which, contrary to Sturtevant’s paradox, alterate the paradigmatic regularity (e.g. the introduction in the Germanic languages of  strong forms in originally weak verbs). Sturtevant’s paradox concerns just a part of  language change and the irregularities which are bound to language change. Analogy, on the contrary, is the ability to perceive similar features in  a set of objects and to extract  them from the set (see Anttila 1977). Is analogy predictable?

Other possible causes of irregularity:

1. Antonymy (magister ~ minister).
2. (Systematic) Suppletivism (Nübling 2000).
3. False segmentation, Phonetic attrition (Seiler 2008)  and paraetymology (‘Volksetymologie’).
4. Hypercorrection.
5. anguage contact.

Time permitting, examples will be given for 1. to 5.

In Latin epigraphic sources, a morpheme -e(i)s can occur as the plural nominative of some masculine -o stems: the forms Valerieis, Turpilieis, fabres, vireis,etc.… (instead of Valerii, Turpilii, fabri, viri, etc.…) are in fact well attested (82 occurrences: data from Rovai 2008) in texts from III to I century A.D.

Currently labelled as “irregular”, this morphological class actually developed as a prototypically-structured category, whose members share both semantic and phonological features, although at different degrees. Starting from a set of core members, which can be identified in proper nouns ending in -rj- or -lj- (such as Sertorieis, Valerieis, Atilies, Turpilieis), more marginal members (such as Herennieis, Septumieis, feilieis, thurarieis) have gradually been associated according to a “family resemblance” relation (Rosch & Mervis 1975): there is not a strict set of features necessarily shared by all members, but each member has some feature in common with at least one other member. Such a development is supported by diachronic evidence.

In the framework of “lexical connectionism” the structure of this morphological class can be characterized as a “schema”.  Following Bybee & Moder (1983: 267), “schemas should be thought of as associations among lexical items. There are associations on many levels: on the phonological level, words may be associated by initial segment, by rhyme, by stress pattern, or by number of syllables; […] on the semantic level, they are associated by being similar or opposite in meaning, or by belonging in the same semantic field”.

Similar constraints have been recognized by Lazzeroni (2005-2006) as governing the distribution of a morpheme -us, which can sometimes replace -is as the singular genitive of third declension nouns (Venerus, Capitonus, nominus, etc… instead of Veneris, Capitonis, nominis, etc…).

Significantly, neither plural nominatives in -e(i)s, nor singular genitives in -us are predictable according to any productive Latin morphological rule: the former are in fact unexpected innovations (probably borrowed from Oscan), the latter archaic relics.

However, the principle underlying the structure of these morphological classes seems to be the same: in front of a set of unpredictable (lexicalized) forms, speakers tend to associate them into “schemas”, restoring a certain degree of organization and predictability. As noticed above, plural nominative in -e(i)s became in fact a partially productive morphological class, despite its short life and before its complete disappearance.

These diachronic data add significant evidence for a scalar representation of morphological (ir)regularity (Ramat 2005): between fully lexicalized information and fully productive rules, there appear to be other ways to organize linguistic items for a more efficient accessing.

This study presents data on the acquisition of French and Spanish structures: past tense and adjectives. Spontaneous speech samples for French pre-school children reveal few linguistic errors in these structures (Thordardottir, 2005). However, speech data can overestimate linguistic production and comprehension abilities in children (de Villiers, 1992; Royle & Thordardottir, 2008). Two experiments probe the influence of frequency (token and morphological pattern) and regularity on acquisition patterns.

A first experiment explores the processing of French verbs. These fall into three conjugation groups 1. default, 2. sub-regular and 3. irregular. The regular pattern is the most frequent in types and tokens (Nicoladis et al., 2007), irregular verbs have high token frequencies, and sub-regular and irregular patterns have low type but high token frequencies (Royle, 2007). We probe production of group 1 and 3 verbs in French children aged 2;11 – 4;6. Production abilities varied according to regularity and frequency. Regular verbs surpassed irregular ones. High token frequency verbs showed better production than low ones. In addition, these factors interacted, as low token frequency irregulars were often regularized into regular and sub-regular patterns, but not irregular patterns (also seen in Nicoladis et al., 2007).

A second experiment probes the question of regularity in adjectives. French and Spanish adjectives agree in gender with the noun they modify (e.g., La maison blanche / la casa blanca ‘the.f house white.f’). French (irregular) variable adjectives have word-final consonants in their feminine forms, but these are not predictable via a rule (Royle & Valois, submitted), while regular variable Spanish adjectives have predictable –o and –a endings for the masculine and feminine. Experiments elicited the production of these adjectives in children aged 3:0 to 4:11 in both languages. French variable feminine adjectives were harder to master than other forms and gender errors occurred on feminine structures (in elicited data and spontaneous speech). Comparable data from Spanish shows equal performance on feminine and masculine targets.

These data show that the acquisition of morphosyntax is strongly mediated by two factors. On the one hand, pattern regularity is a salient factor.  In particular, the default status of a pattern can modulate its use by children. However, low type-frequency patterns can emerge, if they are regular (e.g., second conjugation verbs). Finally, even though a pattern is frequent, it might not be useful for acquisition: French third conjugation verbs and feminine variable adjectives do not seem to be integrated into any relevant morphological system, even though they are common corpus types or tokens. These data are discussed adapting Bertram et al. (2000) by integrating both pattern frequency and morphological regularity.

Tamil has a bimoraic minimality requirement. Besides this, verbs are also subject to what I term the regularity criterion that calls for all verbs to be minimally disyllabic. A monosyllabic verb, by definition, does not meet this requirement, and is invariably irregular. There are disyllabic irregular verbs also, but in many of them, the irregularity mainly manifests in the participle / past stems, like in English.

Unlike English, irregularity in monosyllabic verbs is not always seen in any one fixed stem. It has an interesting distribution, and is better explained with stem dependency (Bonami & Boye 2002). Semantically very different stems are built from the same underlying stem (Aronoff 1994). Tense stems (past, present and future) are built upon aspectual stems (participle, infinitive and imperative) in the derivation of the paradigm of a given lexeme; their form (but not function / semantic meaning) is dependent on the form of the parent stems. These stem dependencies hold within each of the three inflectional classes (consonant-final, vowel-final and approximant-final) and are also evident in irregular lexemes, which do not randomly project unpredictable inflectional realizations, but surface with conjugations associated with another inflectional class. For instance, the (partial) paradigms of two irregular homophonous approximant-final stems [vej] are compared to the partial paradigm of a  regular approximant-final verb [vərəj] are given below. The first [vej] follows the inflectional  pattern of a  well-behaved vowel-final verb while the second [vej] projects the paradigm of a regular consonant-final verb. They are both supposed to pattern like [vərəj].

IMP	FUTURE	PTCPL	PAST	INF	PRESENT
vej keep	ve-pp-	ve-čč-	ve-čč-	ve-kkə	ve-kkə-r-
vej scold	ve-jj-ɨv-	ve-øǰ-	ve-øǰ-	ve-jj-ə	ve-jj-ə-r-
vərəj draw	vərəj-ɨv-	vərə-øǰ-	vərə-øǰ-	vərəj-ə	vərəj-ə-r-

The stem dependency tree, which gives the dependency relationships between the different tense and aspectual stems, is part of the grammar. For regular verbs, only the lexemic stem needs to be stated; all the other cells in the lexemic paradigm are filled in by the tree. For irregular lexemes like [vej], the lexemic stem and the specific conjugation class that it illegally copies need to be stated. Once the inflectional class is known, the tree predicts the other verbal stems. Thus seemingly random conjugations of irregular verbs are also predictable if we assume that irregular lexemes merely express solidarity with an inflectional class other than the one it is expected to be assigned to.

Despite the traditional conception of Old Armenian nominal inflection as a perfect stage in the evolution of the language (for a challenge of this topic, see Djahoukian 1965), this language displays a great amount of irregularity in nominal morphology: in modern descriptions are illustrated ten declensions, with many sub-classes and lexical exceptions (see for ex. Jensen 1959; Schmitt 1981; for a different approach Matzinger 2002); moreover many nouns follow more than one declension in their inflection (see Olsen 1999; Belardi 2004).

This study concerns one of the main causes of morphological opacity in Old Armenian: stems alternation in nominal inflection. Our aim is to investigate partition classes of paradigm as concrete objects organizing the distribution of morphomes (on this point of view, see Stump 1998; Pirrelli – Battista 2000; Thornton 2005).

Armenian nominal inflection is traditionally divided in two main sections: variable vs. invariable (see Marr 1903; Meillet 1913). The latter includes nouns with stems ending in consonants: nasals, laterals or rhotics; the variation concerns the vowels preceding the ending consonants, and is not predictable on the basis of synchronic phonology.

We are going to verify the following hypothesis concerning the variable group: paradigms involving alternation of two or three stems are derived from the more complex one, in which there are four different stems (ex. matn ‘finger’, with stems matn, matan, matin-, matown‑: vowel alternation is a reflex of Indo-European apophony, see Hamp 1988). Less complex patterns are derived from more complex ones by deletion of boundaries among partition classes of the paradigm, and without introduce new partitions in it. For example, the different paradigms with three stems are derived from the four-stems-paradigm, with different combination of the four partition classes of it; the same phenomenon is observable in the derivation of the paradigms with two inflection stems from the three-stems-paradigms. Such morphological motivation of stem variation can explain its historical spreading beyond its original range.

A large number of intransitive and transitive verbs in Hopi exhibit root suppletion conditioned by the number of the subject or object.  In this paper, I address the morphological operations underlying suppletion in Hopi.  I propose a novel, phase-theoretic analysis of verbal root suppletion within the framework of Localist Theory of contextual allomorphy (Embick 2008).  Contra to Embick & Halle (2005), I argue that verbal roots supplete and their suppletion is constrained by linear adjacency and cyclic spell-out.  My approach correctly predicts that when linear adjacency and cyclic spell-out are disrupted, suppletion cannot occur.

I propose a unified analysis of transitive (1a-b) and intransitive (2a-b) verbal root suppletion.  My main assumption is that cyclic and non-cyclic nodes can interact for allomorphic purposes as long as they are spelled-out in the same PF cycle.  I argue that suppletion of a transitive suppletive Root √ QÖYA (1b) is triggered by the [+plural] feature of a non-cyclic Num when i) Num and the root are linearly adjacent, and ii) spelled-out in the same v cycle.  I assume that intransitive suppletive roots behave syntactically as unaccusatives, and their suppletion is triggered by an underlying object (Hale & Jeanne 1991; Harley et al 2006).  Suppletion of an intransitive suppletive Root √ YUɁTU (2b) is triggered by the [+plural] feature of a non-cyclic Num when i) Num and the root are linearly adjacent, and ii) spelled-out in the same v cycle.

(1 a.)

(1 b.)

(2 a.)

(2 b.)

I present novel data which show that when linear adjacency and cyclic spell-out are violated, suppletion is blocked by default morphology.  When a suppletive verbal root √YU’TU has been elided in the main clause (3), the AGR head cannot be sensitive to the suppletive identity of the root and has to default to the suffix /-ya/.

(3)

In this paper, I have provided a more constrained account of Hopi verbal root suppletion constrained by the architecture of grammar with local constituent relations and cyclic derivation.

Greek diminutives formed with the super-productive suffix –/aki/ (moro>moraki, ‘baby> little, cute baby’, kukla>kuklaki, ‘doll>little doll’, naftis>naftaki ‘sailor> young sailor’) are notorious for lacking a genitive form, both in singular and plural (*morakju/morakjon,*kuklakju/kuklakjon etc). There has been a number of different proposals seeking to explain this irregularity: traditional Greek grammarians (Triandafyllidis 1963) have put forth an explanation in line with the historical fact of analytical formations replacing synthetic inflectional types and the pressure exerted upon the paradigm by this process, while recent research (Sims 2006, 2008) considers genitive plural avoidance in general in MGR in the light of the speakers’ attempt to deal with competing stress patterns and paradigmatic structure, as well as with their socially motivated ‘insecurity’.

Nevertheless, there are some additional facts that need to be taken into consideration: diminutives in –/aki/ form a subclass (of considerable volume due to their high productivity) within the inflectional class of neuter nouns in /–i/, which are not defective in genitive. Moreover, once lexicalised, items in –/aki/ behave like ordinary nouns in /–i/, i.e. they do form a genitive form. In addition diminutives in –/aki/ align partly with other classes of diminutives and non-diminutives, which also lack the genitive plural form.

It is argued that frequency plays a major part in accounting for the observed absence of the genitive in these diminutives, if it is related to the different strength of types within the paradigm (Thomadaki 2008). The notion of relative vs. absolute frequency (Corbett et al. 2001), as well as the impact of productivity (understood in this context as possibility of forming a diminutive form) on this kind of irregularity will be further discussed.

In the canonical approach to morphological description (Corbett 2005, 2007a, 2007b), a canonical paradigm is expected to exhibit completeness, i.e., “for any given lexeme, every cell of its paradigm will be filled by the inflectional system” (Corbett 2005:33). A further expectation, not (yet) named by Corbett, could be called uniqueness of realization, and stated in the following way: “for any given lexeme, every cell of its paradigm will be filled in a unique way”. Deviations from this canonical situation are represented by cases in which a cell is filled by two synonymous forms which realize the same set of morphosyntactic properties.

With respect to the properties of canonical inflection (Corbett 2007a: 9), double forms can represent deviations, i.e., irregularities, of different kinds. Table I shows cases of deviations from uniqueness of realization of cells, exemplifying from Italian verb morphology.

Table I. Deviations from uniqueness of realization of cells in Italian verb morphology

	Deviations instantiated by double forms in the same cell of the same lexeme	Examples
composition/structure	forms following two different conjugations in the same cell	eseguo / eseguisco ‘perform:1.prs.ind’
lexical material (≈ shape of stem)	forms built on two different stems in the same cell	devo / debbo ‘must:1.prs.ind’ vado / vo, faccio / fo ‘go:1.prs.ind’, ‘do:1.prs.ind’ doliamo / dogliamo ‘suffer:1pl.prs.ind’ giacciamo /giaciamo‘lie_down:1pl.prs.ind’ parvi /parsi ‘seem:1.prf.ind’ sederò / siederò ‘sit:1.fut’ morrò / morirò ‘die:1.fut’ udirò / udrò ‘hear:1.fut’ visto / veduto (?) ‘see:pst.ptcp’
inflectional material (≈ shape of inflection)	forms with two different inflectional endings in the same cell	temei / temetti (?) ‘fear:1.prf.ind’ va’ / vai, fa’ / fai, sta’ / stai ‘go:2.imp’, ‘do:2.imp’, ‘stay:2.imp’ udente / udiente ‘hear:prs.ptcp’

As the examples in Table I show, double forms can be claimed to represent deviations from canonicity in all areas, but doubling due to the employment of two different stems seems prevalent in Italian. The classification of some examples depends on the analysis: is temei / temetti an instance of different inflectional endings realizing the same cell (analysis: teme-i / teme-tti)  or an instance of two different stems employed in the same cell (analysis: teme-i / temett-i)?

The presentation will aim at a description of the areas in which double forms occur in Italian verb inflection. The relations between double forms and other kinds of deviations from canonicity, such as heteroclisis, suppletion, and morphomic stem alternation, will be addressed. The relative strength of different kinds of double forms will be assessed through a corpus-based study of their frequency of occurrence in contemporary Italian (based on the la Repubblica corpus, containing 16 years of issues (1985-2000) of the daily newspaper la Repubblica (about 330M tokens): http://sslmitdev-online.sslmit.unibo.it/corpora/corpus.php?path=&name=Repubblica). In conclusion, some criteria will be established to recognize canonical doublets; these will be related and compared to criteria of canonicity for other irregular morphological phenomena, such as suppletion (Corbett 2007a).

Under negation, the existential subject and direct object in Finnish are usually in the partitive case. In affirmative sentences (1), the direct object case involves aspect and it is thus variable to a certain extent, but under negation the partitive is called for (2):

1.	Anna osti auton.	Anna bought a car.
	Anna-NOM bought car-GEN
2.	Anna ei ostanut autoa.	Anna didn't buy a car.
	Anna-NOM not bought car-PAR

The same can be observed in existential clauses, affirmation in (3) and negation in (4):

3.	Puussa on linnunpesä.	There is a bird's nest in the tree.
	tree-INE is bird's-nest-NOM
4.	Puussa ei ole linnunpesää.	There is no bird's nest in the tree.
	tree-INE not-3sg be bird's-nest-PAR

However, in certain word orders, the presence of pragmatic particles changes the speech act from negation to an affirmation. In these instances, even in the presence of the negative verb, the typical partitive co-occurring with the negative verb disappears, and the case of the affirmative sentences surfaces (5) and (6).

5.	Eikö-kö Anna ostanut-kin auton.	Anna bought car, didn't she.
	not-particle Anna-NOM bought-particle car-GEN.
6.	Ei-kö-hän puussa ole linnunpesä.	There's a bird's nest in the tree, dont you think
	not-Q-particle tree-INE be bird's-nest-NOM.

I will argue that the choice of these case suffixes under negation is a syntactic choice only by default. Ultimately the decision depends on the illocutionary force of the speaker. There is thus an apparent irregularity in the case marking in the presence of clitic particles, but when we take into account the speech act, everything is perfectly regular after all.

This paper deals with a rather problematic issue and status of lexical inflections. The problem of lexical plurals has been studied by Acquaviava (2008). However, for the topic of this paper I choose to discuss –ed adjectives (among others studied by Hirtle, 1970; Hudson, 1975; Beard, 1976) and gerunds (Borer, 1990; Blevins, 2005; Vujic, 2006) in English. In order to study the status of -ing and –ed affixes in English WF, as a starting point I use the concept advocated in modern morphological studies that gerunds

and “–ed formations” are seen as dual representations, mixed category structures or dual lexical categories (Yoon, 2005). It is a well-known fact that, occasionally, administering a certain inflectional suffix may lead to a change in a part of speech category (class) of the base or Head (Yoon, 2005). I use this notion and on the examples of various –ed formations and gerundial –ing formations in English, which I perceive as inflected forms I am trying to test Yoon’s claim that all inflectional forms represent mixed categories in the respect that it is the closing affix that determines the properties of root+affix combinations, while the root determines the marking on its dependents (selectional rules of categorical type, as suggested by Fabb, 1988; and later discussed by Plag, 1996). But what happens when a selectional rule of categorical type becomes violated such as in combinations Nominal root +-ed or Nominal root + -ing? I argue that such forms are the products of word-formation processes which incorporate both inflectional and derivational morphologic and semantic features. Furthermore, I advocate the point that modern WF in English is, to a great extent, semantically governed. Consequently, such semantic point of view leaves very little room for the traditional notions of inflection/derivation dichotomy.

I argue that in cases such as  Nominal root +-ed or Nominal root + -ing happens the process of crossing the borders between components during which categorical borders are being  disregarded  in a most unpredictable way. Speakers when lacking an appropriate lexical item in their lexicon (or the lexicon of one particular language as a whole) use the available existing elements with which they are familiar. In this process they are governed primarily by the meaning or semantic component of the particular element in question (Aronoff and Anshen, 1988).

I suggest that in cases of lexical inflections the limited and regulated scope of inflection is used to complement the imperfections and deficiencies of WF processes. Following the works of Lakeoff, Anshen and Aronoff (1988) I suggest that it is the semantic component that primarily governs contemporary WF, and that inflectional affixes are stored in our mental lexicon carrying multiple meanings and serving multiple functions. One of them is to form new words. If it is so, can such formations be considered dual lexical categories since they bear inflectional marks? What is certain is that they tend to show dual nature in their internal and external properties; more precisely they have external distribution of one POS and internal syntax of another POS (Laponte, 1999).