Evaluative Morphology and the Syntax of Adjectives in Italian

Abstract

This paper addresses a well-known puzzle at the intersection of morphology and syntax: the categorical exclusion of adjectives modified by evaluative morphology from prenominal position in Italian. While Italian allows many adjectives to occur both pre- and postnominally, adjectives like piccolino, ‘little-dim’, are strictly postnominal (cane piccolino, lit. ‘dog little-dim’ vs. *piccolino cane, ‘little-dim dog’), a distribution not fully explained by their proposed predicative or intersective nature. Drawing on degree semantics and trope theory, we argue that this constraint arises from an incompatibility between two distinct interpretive strategies. Prenominal adjectives undergo a syntactically driven semantic shift, whereby the noun triggers a trope-based interpretation of the adjective, redefining the meaning of the A-N complex. In contrast, evaluative morphology operates through a pragmatically driven strategy, contributing speaker-oriented, context-sensitive meaning to the adjective. Crucially, these two strategies are mutually exclusive: an adjective modified by evaluative morphology has already undergone pragmatic reinterpretation and cannot simultaneously participate in the compositional syntactic process required for prenominal placement. This explains why adjectives with evaluative suffixes are excluded from prenominal contexts, despite often yielding intersective interpretations postnominally. Our proposal accounts for this distributional asymmetry without resorting to stipulations and suggests that certain interpretive procedures are not recursively applicable across syntax and pragmatics. Ultimately, this study sheds new light on a principled interface constraint linking syntactic distribution, morphological derivation and pragmatic interpretation.

1. Introduction

Evaluative morphology constitutes a long-standing puzzle at the interface between morphology and syntax, due, among other things, to its ambiguous status between derivational and inflectional morphology, and to the complexities raised by the internal ordering of the evaluative affixes within the word (Scalise, 1984, 1988; Stump, 1993). In this contribution, we aim to address and offer a solution to a specific aspect of this broader puzzle: the strong constraint observed in Italian concerning the distribution of adjectives bearing evaluative morphology with respect to the noun.

As is well known, Italian is an ANA language (Cinque, 2014, 2015; Manzini, 2025a, 2025b), allowing many adjectives to occur both prenominally and postnominally. This is shown in the following (1):

(1)	a.	Mi	ricordo	le	lezioni	noiose	del
		REFL.1SG	remember.1SG.PRS	the.F.PL	lesson.F.PL	boring.F.PL	of.the.M.SG
		mio	professore.
		my.M.SG	professor.M.SG
	b.	Mi	ricordo	le	noiose	lezioni	del
		REFL.1SG	remember.1SG.PRS	the.F.PL	boring.F.PL	lesson.F.PL	of.the.M.SG
		mio	professore.
		my.M.SG	professor.M.SG
		‘I remember my professor’s boring lessons.

However, adjectives belonging to particular classes can only occur postnominally, as shown in (2) for color adjectives:

(2)	a.	Ieri	ho	visto	molti	carlini	neri.
		yesterday	AUX.1SG	see.PTCP	many.M.PL	pug.M.PL	black.M.PL
	b.	*Ieri	ho	visto	molti	neri	carlini.
		yesterday	AUX.1SG	see.PTCP	many.M.PL	black.M.PL	pug.M.PL
		‘Yesterday I saw many black pugs.’

This restriction applies forcefully to “evaluative adjectives”1: the prenominal order in (3b) is uniformly considered as syntactically unacceptable by Italian speakers, and sharply contrasts with the grammatical postnominal order exemplified in the following (3a).

(3)	a.	Ieri	ho	visto	un	carlino	bruttino.
		yesterday	AUX.1SG	see.PTCP	a.M.SG	pug.M.SG	ugly-DIM.M.SG
		‘Yesterday I saw a rather ugly pug’
	b.	*Ieri	ho	visto	un	bruttino	carlino.
		yesterday	AUX.1SG	see.PTCP	a.M.SG	ugly-DIM.M.SG	pug.M.SG

This distributional asymmetry is also supported by corpus data. A search in the itTenTen20 Corpus (via SketchEngine) shows that evaluative adjectives such as bruttino occur almost exclusively in postnominal position, confirming the intuition that the prenominal use is virtually absent in naturalistic data. Based on this strong empirical evidence, an important issue arises at the morphology/syntax interface: Which properties of evaluative morphology, when realized on adjectives (from now on: EvA), account for this syntactic constraint, preventing evaluative adjectives from occurring in a prenominal position? In the literature, this emerges as an unsolved puzzle, though there have been some interesting hints directed to its resolution.

Savoia et al. (2017b), and more specifically Savoia et al. (2017a), propose that EvA is inherently predicative, since it expresses a property that applies to the argument R licensed by the adjectivally modified noun (on a par with gender features, see Percus, 2011). Now, in Italian, the canonical syntactic realization of predicative adjectives is the postnominal position (Cinque, 2014; Manzini, 2025b), and this arguably explains the exclusion of evaluative adjectives from the prenominal position.

This suggestion is interesting, but it faces serious problems when technically implemented. In fact, postnominal predicative adjectives are typically intersective, as confirmed by the fact that they are not subjected to internal relative ordering (Manzini, 2025b).2 However, EvA is inherently not intersective. What we mean by that is that EvA typically applies non-intersectively when it modifies, as a suffix, the original adjective to which it is attached. For instance, the meaning of piccolino in (4a) is not arrived at by conjoining the properties λx. -ino (x) and the property λx. piccolo (x); rather, it is compositionally obtained, intuitively, by considering the subset of small objects that qualify as small according to the speakers’ affective evaluation, including their subjective relativization of the notion of ‘smallness’ when it applies to elephants (subsective meaning). In other words, EvA in (4a) does not shift the non-intersective interpretation of the adjective piccolo in (4b), where it occurs prenominally. It is then difficult to explain why piccolo can also occur prenominally with a non-intersective (predicative) interpretation, as shown in (4b), whereas piccolino cannot occur in this very same position, as shown in (5). Clearly, the ban on evaluative adjectives occurring prenominally is independent from the predicative nature of EvA, for the very reason that EvA does not behave intersectively, though being realized only postnominally, i.e., in the typical position of intersective adjectives.

(4)	a.	Ho	visto	un	elefante	piccolino.
		AUX.1SG	see.PTCP	a.M.SG	elephant.M.SG	small-DIM.M.SG
	b.	Ho	visto	un	piccolo	elefante.
		AUX.1SG	see.PTCP	a.M.SG	small.M.SG	elephant.M.SG
		‘I saw a small elephant.’

(5)	*Ho	visto	un	piccolino	elefante.
	AUX.1SG	see.PTCP	a.M.SG	small-DIM.M.SG	elephant.M.SG
	‘I saw a small elephant.’

Here, we propose an alternative approach to ungrammaticality. First, we will accept and considerably extend a degree semantics for adjectives. On these grounds, we will show that prenominal adjectives in Italian tend to become independently referential, undergoing a process of meaning change based either on the property expressed by the noun they modify or on some speaker-oriented contextual properties (Delfitto, 2025). Exploiting N in this process corresponds to a syntactic strategy, while exploiting speaker-oriented properties encoded in evaluative morphology corresponds to a (morpho)pragmatic strategy. To implement the syntactic strategy, the adjective must occur pre-nominally. On the other hand, EvA essentially activates the (morpho)pragmatic strategy. Since the syntactic and pragmatic strategies are incompatible with each other, it follows that evaluative adjectives cannot occur prenominally. As we will see, this has interesting consequences for explaining the empirical pattern seen in (4) and (5) above. When occurring postnominally, as in (4a), EvA applies non-intersectively to reshape the meaning of the adjective; however, once the new meaning has been obtained, the combination of the derived adjective with the noun results in a purely intersective operation. In this way, we derive both the correctness of the postnominal position (the derived adjective interpretation is intersective) and the ungrammaticality of the prenominal position (EvA is incompatible with the syntactic strategy).3

The paper is organized as follows. In Section 2, we will introduce the proposed semantics for (Italian) prenominal adjectives. In Section 3, we will briefly review the core properties of evaluative adjectives. Finally, in Section 4, we will lay out the explanatory account for the ban on prenominal evaluative adjectives.

2. Degree Semantics, Tropes, and Associativity

Let us start by considering again the distribution of Italian adjectives relative to the nouns they modify. Here, the relevant research question is the following: How is the interpretation of an adjective redefined when it occurs prenominally instead of postnominally, as in (1b), when compared with (1a)? We propose that traditional properties such as restrictiveness and intersectivity are largely irrelevant (Manzini, 2025b; Delfitto, 2025) and that the key factor is associativity.

In the interpretation corresponding to (1a), we can build (intersectively) the set of classes given by my professor that I remember, and then further restrict this set of classes by applying the property boring. Analogously, we can first consider the set of boring classes given by my professor, and then further restrict this set to the classes that I remember. Intuitively, the result does not change; that is, the interpretation of postnominal adjectives is associative: a + (b + c) = (a + b) + c. On the contrary, the interpretation associated with (1b) conveys the sense that all classes given by my professor were boring. It is like there is no compositional step at which I may consider the set of classes given by my professor that I remember (some of which might have been non-boring), successively restricting this set by applying the property boring. In fact, the scenario envisaged by (1b) excludes the presence of non-boring lessons. This entails that we should first consider the set of boring classes given by my professor, successively restricting this set to the classes that I remember. In other words, boring is no longer associative: it must be immediately intersected with the property corresponding to the noun it modifies.

If the complex A-N behaves as an inseparable unit (contrary to what happens with N-A), the question becomes why this should be the case. Subsectivity vs. intersectivity cannot be the correct answer, since intersective interpretations are found both in postnominal and in prenominal position. For instance, the adjective bello, when applied to the noun attaccante ‘striker’, gives rise to both an intersective reading (=handsome, beautiful) and a non-intersective reading (=beautifully playing). As shown in (6a,b), both readings can be expressed in the prenominal position (see also Manzini, 2025b):

(6)		Thuram	è	proprio	un	bell’attaccante.
		Thuram	be.3.SG	really	a	beautiful-striker
	a.	‘Thuram really is a great striker.’
	b.	‘Thuram really is a handsome striker.’

We propose that the correct answer is that in an A-N complex, A undergoes meaning change, and that this semantic shift is brought about by N. In a nutshell, A-N applies to the referent ‘x’ iff ‘x is A in the measure objects satisfying N typically are’, in line with a degree semantics interpretation (Kennedy, 1999, 2007). Even better, A-N applies to ‘x’ iff ‘x satisfies the brand of A typically instantiated by N’. The latter corresponds, in our terms, to a trope-like interpretation, where tropes are particularized properties such as the boredom of my professor’s lessons, or the blackness of pugs, etc. In other words, tropes are the nominalization of the prenominal adjective at issue (see Moltmann, 2004, 2007, 2009, 2013). Under both analyses, the result is that in the A-N complex, N loses its independent referentiality and is exclusively used to modify the meaning of A. This immediately explains the loss of associativity: A and N do not define two extensionally independent sets that are intersected, but a unique set whose elements satisfy the brand of A modulated by N.

As for syntax, nothing hinges, for the purposes of this contribution, on the nature of the prenominal and postnominal word order of adjectives. The standard view, eloquently advocated by Cinque (2015), is that adjectives are universally base-generated to the left, in strict compliance with Kayne’s Antisymmetry, and that in the languages displaying ANA, the prenominal order results from moving constituents containing N(P), as an instance of roll-up movement. In independent work, we have argued for a different analysis, based on the opposite insight that adjectives are uniformly generated to the right in Italian (as in Semitic), with the prenominal order of adjectives resulting from constrained instances of rightward movement, whereby the N(P) adjoins to the adjective acting as its modifier, in agreement with the proposed semantics (Delfitto, 2025).

Let us then go back to the proposed semantics and let us see in some more detail how this works, starting with degree semantics (Kennedy, 1999, 2007). Consider first the interpretation of the prenominal A expensive in the following (7):

(7)	This expensive BMW is for sale.

In degree semantics, adjectives are degree phrases (Corver, 1997a, 1997b; Grimshaw, 2000), whose head maps objects onto degrees. When applied to the measure function corresponding to the adjective expensive, this head will return a property that holds true of objects whose degree of cost is higher than a contextually defined standard d, as indicated in the following (8):

(8)	λx. expensive(x) ≥ d

However, in an A-N complex as in (7), it is N that provides the comparison class by means of which A is averaged, as explicitly indicated in the following (9):

(9)	λx. expensive (x) ≥ Average (BMW) (expensive)

This is the property expressed by the (Italian) A-N complex costosa BMW: the set of objects that are expensive in the sense that their value (in terms of cost) is at least equal to the average value (in terms of cost) of a BMW. Still, this cannot be the whole story. If it were so, (10), drawn from (Kennedy, 2007) (ex. 16 in the original text), should be contradictory:

(10)	Kyle’s car is an expensive BMW, though it’s not expensive for a BMW. In fact, it’s the least expensive model they make.

In other words, if the role of N (BMW) were the same as the role of the PP for a BMW, we would assert and negate the same thing. Significantly for the purposes of this contribution, Kennedy’s way out consists of denying that N is necessarily the comparison class for A. On the contrary, the comparison class for A can also be contextually provided, and this is arguably the case when sentences like (10) are evaluated as non-contradictory. Delfitto (2025) develops a different analysis, based on a trope-based interpretation of prenominal adjectives, as arguably emerges from the contrast between (11) and (12):

(11)		Le	bianche	scogliere	di	Dover	sono	suggestive.
		the.F.PL	white.F.PL	cliff.F.PL	of	Dover	be.3PL.PRS	striking.F.PL
		‘The white cliffs of Dover are striking.’

(12)	a.	??Le	bianche	scogliere	sono	suggestive.
		the.F.PL	white.F.PL	cliff.F.PL	be.3PL.PRS	striking.F.PL
	b.	Le	scogliere	bianche	sono	suggestive.
		the.F.PL	cliff.F.PL	white.F.PL	be.3PL.PRS	striking.F.PL
	‘White cliffs are striking.’

If N acts as a comparison class for A, why should cliffs of Dover fare better than cliffs as a comparison class for the adjective white, as shown in (11 and 12)? We have seen that the semantic bleaching of N is essentially due to the loss of its referential independence: its semantic function simply consists of contributing to the determination of a comparison class for A. On the other hand, the contrast in (11 and 12) suggests that N does not express, by itself, the comparison class. More particularly, the hypothesis we put forward is that N does not help establish the degree to which A is A; rather, it helps establish the quality of A that is at stake. It is not so much the degree of A (based on a standard of comparison) but a brand of A, based on the numerically and qualitatively distinct instantiations of A that we find in the world. In the case of (11), we are not referring to entities that are white in the sense that their whiteness equals or exceeds the average degree to which cliffs are white; instead, we are referring to entities that are white in the sense that they instantiate the unique kind (or brand) of whiteness that is proper to a specific type of cliffs, that is, “the whiteness of the cliffs of Dover”.

Which kind of objects are these brands of A that are involved in establishing the reference of A in A-N configurations? Philosophically, these interpretive effects seem related to the notion of particularized properties, viewed as concrete manifestations of properties in concrete individuals, starting from the seminal work of Williams (1953). Reference to these kinds of entities takes place by means of de-adjectival nominalizations, of the sort of the politeness of John, the whiteness of this house, the wisdom of Plato, and so on. These particularized properties are generally referred to as tropes. In the linguistic literature, the role of tropes in the treatment of some semantic phenomena has been eloquently advocated in Moltmann (2004, 2007, 2009, 2013). If the semantic bleaching of N in A-N complexes is the final stage of a process of trope-activation, N provides the object (or kind) on which the particularized property expressed by A is instantiated. This would immediately explain why (10) is not contradictory: the PP for a BMW provides a comparison class for A (the average cost of a BMW), whereas the N BMW in expensive BMW provides the kind on which a particularized brand of expensiveness is typically instantiated. Kyle’s car in (10) can thus be expensive according to the brand of expensiveness that is typically instantiated by BMWs, while his cost may not be higher than the average cost of a BMW: no contradiction is triggered by the concomitant use of N and PP as linguistic elements affecting the original meaning of A (Delfitto, 2025).

Be that as it may, an interesting issue arises. Kennedy’s solution for the potential puzzle in (10) consists of admitting the existence of a double strategy when it comes to providing a comparison class for A: this comparison class can be provided either syntactically (by the N or by the PP) or pragmatically, based on the properties of the context of utterance. If we choose the syntactic strategy, the car will be expensive for a BMW; if we choose the pragmatic strategy, the car will be expensive according to some contextually relevant measure of cost. However, if the interpretation of A-N complexes is better captured in terms of a process of trope-activation, there is no evidence for the existence of a pragmatic strategy besides a syntactic one. The reason is that both the explanations in terms of degree semantics and trope-activation are syntactic in nature. In fact, in both explanations, the trigger for the semantic shift of the adjective is N, which provides either the standard of comparison (degree semantics) or the entity (individual or kind) on which the trope is typically instantiated. The question is thus whether there is independent evidence for a pragmatically driven process of meaning change for adjectival modification in Italian, and how this process precisely interacts with the process of syntactically driven meaning change. The proposal that we intend to put forward is that evaluative morphology realized on adjectives provides an interesting case for a pragmatically driven reinterpretation of N-A complexes, while representing, at the same time, an interesting case study for the interaction between morphosyntax and pragmatics.

3. On the Pragmatic Interpretation of Evaluative Morphology

Evaluative morphology has been traditionally assigned an ambiguous status between derivational and inflectional morphology (Scalise, 1984; Stump, 1993; Grandi & Körtvélyessy, 2015). Like inflection, evaluative morphology is typically category-preserving: for instance, in Italian, the nouns casetta ‘little/nice house’ (from casa f.sg ‘house’ + etta end.f.sg) and gattino ‘little cat’ (gatto m.sg ‘cat’+ ino dim.m.sg) are derived from nouns and retain the gender and inflection class of the base noun. This also applies to adjectives, as in bello ‘beautiful’ → bellino ‘cute/pretty’ (bello m.sg ‘beautiful’ + ino dim.m.sg). In terms of word internal distribution, evaluative suffixes appear after word formation phenomena and before gender/number inflection. For instance, consider the following example taken from Scalise (1984, p. 133):

(13)	contrabbandierucoli	‘small-time smugglers’
	compound word contrabbando (‘contraband’) + derivational suffix (-iere ag) +
	evaluative suffix (-ucol(o) pej(m.sg)) + inflectional morpheme (-i m.pl))

Despite these properties, evaluative suffixes align with derivation in that they contribute their own semantic content, while also conveying an affective or expressive nuance. Italian features a wide range of evaluative suffixes across different semantic classes, including at least diminutives (dim: -in-, -uzz-), endearing forms (end: -ett-, -ell-, -ucci-), augmentatives (aug: -on-), and pejoratives (pej: -acci-) (Merlini Barbaresi, 2004, for more details).

In the literature, the role of evaluative affixes has been modelled as akin to that of gender or classifier features, since in both cases the relevant affix expresses a property of the object filling the unsaturated position licensed by N or A (Higginbotham, 1985), as explicitly proposed in Savoia et al. (2017b). Regarding the relationship between gender and evaluative morphology, Italian exhibits alternations between evaluative suffixes that preserve the gender and class of the root and those (such as -in-/-on-) that can impose a specific gender and class. For instance, from donna (fem), one can form donnina (dim.fem) or donnona (aug.fem), which preserve the feminine gender and class of the base noun; however, forms like donn-one (aug.masc) or donn-ino (dim.masc) use evaluative suffixes that introduce their own inflectional markers, thereby shifting the noun’s gender to masculine. As Merlini Barbaresi (2004, p. 275) observes, this results in a “neutralization of gender expression”, where other elements within the noun phrase agree with the gender of the derived noun. Interestingly, certain evaluatives are restricted in their gender application: from donna, the form donnetta (using the endearing suffix -ett-) is acceptable, while its hypothetical masculine counterpart ‘donnetto’ is unacceptable. Nevertheless, donnetta may be used metaphorically or derogatorily to refer to male individuals who exhibit traits stereotypically associated with women (such as triviality and small-mindedness).

In Italian, there are important and still poorly understood restrictions on the relative ordering of the suffixes, which give rise to clear generalizations. For instance, the endearing suffixes -ett- and -ell- can only precede diminutive -in-, appearing between the root and -in-. Thus, from libro, one can form both librino (dim) and libretto (end), but only libr-ett-in-o is acceptable. The reversed order in *libr-in-ett-o sounds sharply deviant, and a similar contrast holds for voc-ett-in-a vs. *voc-in-ett-a. There have been interesting attempts to derive these distributional constraints either in terms of semantic criteria or through the hypothesis of a fixed template of functional projections hosting the various affixes (Savoia et al., 2017a), including those showing up in the adjectival domain, similar to the approach taken in mainstream cartographic work (Cinque, 2014).

Among the evaluative affixes, diminutives are the most noticeable forms. In general, they are strongly speaker-oriented: they reflect the speaker’s subjective attitude and, thus, are typically associated with the expression of affection and emotional evaluation. This layer of affective meaning typically reflects an asymmetry between the speaker and the addressee, as in the relationship between parents (or caregivers) and children, or the speaker’s intention to express tenderness by ascribing ‘childlike’ qualities to their addressee. In child-directed speech, for instance, mano often appears as man-in-a, casa as cas-in-a and crema becomes crem-in-a. In cognitive linguistics, this relation between the concept of ‘small size’ and the source concept of ‘child’ is well-documented and has been widely addressed (Prieto, 2015). This emotional or affective function of diminutives, however, has probably been overestimated, since there are discourse contexts in which the expression of affection is either secondary or even completely absent.

First, diminutive (and endearing) affixes often function as numeral classifiers, exhibiting a ‘singulative’ property: they convert the mass noun to which they apply into a count noun (Savoia et al., 2017a, 2017b). In Italian, this property becomes evident when we compare minimal pairs, such as the following (for more examples, see Savoia et al., 2017b, p. 289):

(14)	Italian	English translation
	a. zucchero → zuccherino	sugar → sugar lump (sugar cube, sweetie)
	b. cera→ cerino	wax → matchstick (wax match)
	c. paglia → pagliuzza	straw → speck of straw (small straw, straw sliver)
	d. legno → legnetto	wood → small piece of wood (twig, stick)

In all these cases, the semantic shift from mass to count is clearly prevalent with respect to the affective meaning traditionally assigned to diminutives, which is either marginal or even almost completely absent. A particularly striking example of the singulative use of diminutive affixes is provided by Dutch (Bauer, 1997), in which the suffix -ie maps every base it applies to into a referential count noun, independently of the category of the base. This provides derivations such as the following (Bauer, 1997, p. 550):

(15)	N	lepel ‘spoon’	lepeltje ‘tea spoon’
	V	zitten ‘sit’	zitje ‘seat’
	A	zoet ‘sweet’	zoetje ’sweetener’
	Num.	tien ‘ten’	tientje ‘tenner, ten-guilder note’
	Adv.	toe ‘in addition’	toetje ‘dessert, pudding’
	P	uit ‘out’	uitje ‘outing’
	Dem	dit en dat ‘this and that’	ditjes en datjes ‘odds and ends’

This derivational pattern in Dutch shows that the semantics of diminutives extends well beyond purely affective meaning. However, the most compelling evidence for this claim lies again in their strong pragmatic orientation (Dressler & Merlini Barbaresi, 1991; Merlini Barbaresi, 2015). As previously discussed, diminutives are inherently speaker-oriented, often expressing positive emotional evaluation towards the addressee. However, this speaker orientation can surface in various ways, with evaluatives playing a key role in shaping and modifying the illocutionary force of the utterance. One such function is the expression of sarcasm on the part of the speaker, as illustrated by the following examples from Merlini Barbaresi (2015, p. 32):

(16)	a.	è	il	voto di una	universitucola.
		it’s	the	vote of a	university-PEJ.FEM.SG
					(small, unimportant university)
	b.	Cosa fate voi nei vostri			salotterelli?
		What do you do in your drawing room-intf-DIM.M.PL
					(in your silly little parties?)
	c.	il	suo	è un	mestierino	che	rende.
		the	yours	is a	job-DIM.MASC.SG	which	reward. 3SG.PRES
	‘Yours is quite a lucrative job.’

Alternatively, the use of evaluatives (especially those with diminutive or endearing semantics) can serve to convey a sense of false modesty, as shown by (17), or may function as a hedging strategy, i.e., softening the force of an utterance (hedging), as illustrated by (18). Both examples are adapted from Merlini Barbaresi (2015).

(17)	Avrei	una	teorietta	da	proporvi.
	have.1SG.COND	a.F.SG	theory-END.F.SG	to	propose-INF.2PL.DAT
	‘I’d have a little theory of my own.’

(18)	Mi	presti	qualche	soldarello?
	1SG.DAT	lend.2SG.PRES	some	money-INFT-DIM.M.SG
	‘Could you lend me a bit of money?’

The specific and complex modulations of certain forms of pragmatic meaning may also attenuate the contrast between diminutives and augmentatives, as when we say, with a hedging function combined with a degree of tenderness, basically with the same meaning:

(19)	Che	stupidino/stupidone	che	sei!
	what	stupid-DIM.M.SG/stupid-AUG.M.SG	that	be.2SG.PRES
	‘What a silly little (big) fool you are!’

These effects can also be detected in the playful use of pejoratives (20a) or in the expression of understatement with endearing suffixes (20b):

(20)	a.	Ah,	Pieraccio,		me		l’hai		fatta
		Ah	Pier-PEJ-M.SG		1SG.DAT		it = have.2SG.PRES		do.PST.PTCP-F.SG
		di nuovo!
		again
		‘Ah, Pieraccio, you did it to me again!’
	b	È		una		sommetta		non		da poco!
		be.3SG.PRES		a.F.SG		sum-DIM-F.SG		NEG		of little
		‘It’s not a small amount of money.’

As we will see shortly, the core of the present proposal regarding evaluative morphology on adjectives is represented by the unity of the A-N complex, which ultimately results in the referential status of the adjective and the modifying role of the noun. This approach is supported by the flexibility in the placement of the evaluative affix within the A-N complex, as suggested by the essential equivalence of è una rispostina permalosa,‘It’s a touchy little answer’ (featuring a diminutive suffix on N), and è una risposta permalosetta, ‘It’s a bit touchy answer’ (featuring an endearing suffix on A). Both can be translated as “It’s a bit of a snappy answer”, and the overall tone can suggest mild sarcasm, criticism, or playful/ironic reproach, depending on context and intonation.

Finally, one of the most prominent features of evaluative morphology is the ambiguity of the enriched meaning it contributes to creating. As Merlini Barbaresi (2015) correctly emphasizes, evaluative morphology often generates a tension between conflicting pragmatic forces, which can yield a stylistically expressive effect, thereby enriching discourse dynamics. A case in point is a sentence like (21):

(21)	Giorgio	è	un po’	vecchietto	per	lei.
	Giorgio	be.3SG.PRES	a bit	old-DIM.M.SG	for	her
	‘Giorgio is a bit too old for her.’

In this sentence, the use of an endearing suffix on the adjective creates a subtle ambiguity between attenuation (reducing the assertiveness of the statement) on the one hand, and, on the other, sarcastic intensification of the enriched meaning (emphasizing Giorgio’s age in a teasing or critical way). This dual reading produces interpretive uncertainty, thus, potentially prompting addressee’s reactions such as “What exactly do you mean by that”?

As already seen above, evaluative suffixes are generally permitted to combine with adjectival bases, giving rise to complex adjectival forms endowed with size or other evaluative meanings, as shown by the data in

Table 1. Italian adjectives with evaluative morphology and use.

Italian Base → Diminutive	English Translation	Nuance/Use
bello → bell-ino	beautiful → cute, prettyish	attenuation, endearment
grande → grand-ino	big → kinda big, biggish	attenuation, sarcasm
simpatico → simpatic-uccio	nice/friendly → a bit nice, somewhat charming	attenuation, hedging
scemo → scem-etto	dumb → a bit silly, kinda stupid	attenuation, irony
vecchio → vecchi-etto	old → a bit old	endearment, irony/sarcasm
vecchio → vecchi(er)-ello	old → sweet old, elderly (gent)	endearment

:

As one can easily infer from these few examples, the full range of affective interpretation and pragmatic enrichment found with evaluative affixes on the noun is also detected, at least in principle, with adjectives. The semantics is also the same since evaluative morphology assigns the same classificatory properties to the argument of the adjectival root as those assigned to the argument of the nominal root.

In the next section, we will propose that a further similarity between evaluative morphology on nouns and adjectives is provided by the observation that evaluative adjectives, on a par with prenominal adjectives within the A-N complex, tend to become independently referential. We suggest that this fact is the key to a proper understanding of the cogent prohibition on evaluative adjectives from occurring prenominally.

4. Evaluative Morphology and Adjectival Roots as Standards of Comparison

As previously discussed, evaluative morphology—realized on both nouns and adjectives—is speaker-oriented and, more generally, pragmatically oriented. Moreover, we have seen that EvA is not restricted to an intersective meaning. Take, for example, the Italian word vecchierello (‘a bit old’): as a nominalized form, it typically conveys affection and tenderness toward an elderly male person. It would be misleading to propose that this interpretation is obtained through the intersection between the set of entities corresponding to the endearing suffix -ell- and the set of entities corresponding to the adjectival root ‘old’. Rather, the affective connotation introduced by the evaluative suffix is relative to the meaning of the base word, vecchio ‘old/old gent’. That is, it reflects the kind of tenderness people often feel for older individuals to whom they are emotionally attached. It follows that vecchierello refers to a subset of elderly people, exactly as piccolo ‘small’ refers to a subset of elephants when it occurs prenominally in piccolo elefante ‘small elephant’. In both cases, we are not considering all small objects and all beloved people, but specifically small elephants or beloved elderly individuals. Therefore, the relevant compositional meaning is, in both cases, inherently subsective and irreducible to an intersective compositional semantics.

In Section 2, we drew on Kennedy’s insight that with degree adjectives, the superficially modified noun acts in fact as a kind of modifier, expressing the standard of comparison required for implementing the degree semantics proper to the adjective. An A-N complex, such as expensive BMW, for example, refers to objects that are expensive in a measure that is higher than the average cost associated with BMWs. Moreover, this degree-based interpretation, whereby A becomes independently referential and N merely sets the comparison baseline, can be revised in even more radical terms by exploiting ‘trope-activation’: the adjective expresses a distinctive quality (i.e., a trope) that is typically associated with the kind denoted by the noun.

Consider the example splendida ragazza ‘splendid girl’: it does not just mean any girl who is splendid. Rather, this A-N complex refers to the set of individuals that instantiate the trope lo splendore delle ragazze ‘the splendor of girls’, a type of splendor typically associated with young ladies or, in other words, a stereotypical, culturally salient type of beauty. Quite interestingly, in this way, we derive the result that the whole A-N complex can be correctly predicated, in the right context, of individuals who are not of the kind ‘girls’. Suppose Marie is a beautiful elderly woman; we can certainly express that by asserting that Marie è (ancora) una splendida ragazza ‘Marie is (still) a gorgeous young lady’. What is commonly regarded as metaphorical usage would then be profitably reinterpreted as a trope-like reading of the adjective splendida: regardless of her age, Marie instantiates the brand of beauty that is typically (but not necessarily) associated with young ladies. In the present view, thus, this metaphorical interpretation essentially depends on a process of semantic bleaching affecting the noun. In the example above, for instance, ragazza ‘girl’ simply provides the kind necessary to single out the unique type of beauty to which the adjective ultimately refers. It is thus the adjective that does the real referential work, while the noun provides the semantic backdrop for the activation of a property frame that is culturally salient.

However, the shift from degree semantics (whereby N provides the standard of comparison) to trope-activation (whereby N provides the kind on which the particularized property is instantiated), though available in the linguistic system, cannot always be brought to its extreme consequences, whereby the derived property can be predicated of entities outside the kind expressed by N. For instance, going back to Kennedy’s original example (expensive BMW), we may notice that metaphoric shift is barred, as shown by the unacceptability of #questa Volvo è una costosa BMW ‘This Volvo is an expensive BMW’, when intended as parallel to Marie è una splendida ragazza. Unlike ragazza, the noun BMW here is too rigid in its referential scope to support a property-based reinterpretation.

At this point, the crucial observation we intend to make is that a similar process of semantic bleaching is also found with EvA. As we have seen, the evaluative affixes are pragmatically interpreted, typically manifesting speaker-oriented meanings. However, if we abstract away from the morphopragmatic vs. syntactic nature of the ingredients involved in the process, the interpretive procedure is essentially the same. To see this in some detail, consider first the tendency of evaluative morphology, when realized on adjectives, to give rise to nominalizations. In Italian, adjectives can be easily converted into nouns without formal changes, as one can observe in the following examples with scemo (A ‘dumb’ or N ‘dumb guy’) and vecchio (A ‘old’, or N ‘elderly person’).

(22)	a.	Ho	incontrato	un	ragazzo	scemo.	(scemo = A, ‘stupid’)
		have.1SG	meet.PTCP	a	boy	stupid
	b.	Ho	incontrato	uno	scemo		(scemo = N, ‘stupid guy/fool’)
		have.1SG	meet.PTCP	a	fool
		‘I met a stupid guy’

(23)	a.	Ieri	ho	aiutato	un	signore	anziano. (anziano = A, ‘elderly’)
		yesterday	have.1SG	help.PTCP	a	gentleman	old
	b.	Ieri	ho	aiutato	un	anziano.	(anziano = N, ‘elderly person’)
		yesterday	have.1SG	help.PTCP	a	senior
	‘Yesterday, I helped an elderly man’

In Italian, however, conversion from adjectives into nouns is not fully productive. For example, while it is natural to say:

(24)	Ieri	ho	incontrato	un	ragazzo	grasso/piccolo/timido/
	yesterday	have.1SG	meet.PTCP	a	boy	fat/little/shy/
	simpatico/	brutto.
	nice/	ugly
	‘Yesterday I met a fat/shy/nice/ugly boy’

It is much less acceptable to drop the noun and say:

(25)	Ieri	ho	incontrato	un	??grasso/	??piccolo/	??timido/
	yesterday	have.1SG	meet.PTCP	a	fat/	little/	shy/
	??simpatico/	??brutto.
	nice	ugly

These adjectives, in fact, cannot be straightforwardly reinterpreted as nouns. However, these same adjectives can function as nouns if they are morphologically modified by an evaluative suffix, as shown in the following (26):

(26)	Ieri	ho		incontrato	un	grass-one/	piccol-etto/	timid-one/
	yesterday	have.1SG		meet.PTCP	a	fat-AUG/	small-DIM/	shy-AUG/
	simpatico-one/		brutt-one.
	nice-AUG/		ugly-AUG
	‘Yesterday I met a big guy/little guy/really shy guy/super nice guy/big ugly guy’

On these grounds, we propose that there is a strict parallelism between A-N complexes such as brutta persona ‘ugly person’ and evaluative adjectives such as bruttone ‘ugly-AUG.MASC.SG’. In the case of brutta persona, we ultimately refer to individuals who instantiate the peculiar brand of (moral) ugliness proper to human beings. Similarly, bruttone refers to individuals who evoke, in the relevant context of utterance, the negative/repulsive feelings we culturally associate with ugliness.

In this respect, the crucial observation is that semantic bleaching applies to both cases. In the A-N complex, we have seen that N, at the end of the process, no longer has a referential status. Similarly, it is easy to see that there are cases where the adjectival root undergoing evaluative modification no longer has a referential status either. Consider the case of endearing EvA, as in vecchietto, ‘elderly-END’. Applying the speaker-oriented compositional procedure described above, we should say that vecchietto can be predicated of those individuals who evoke, in the relevant context of utterance, the feelings of affection usually associated with old age. Interestingly, the nominal vecchietto can be easily predicated of someone who is not old at all, as in the following (27):

(27)	Gianni,	nonostante	i	suoi	trent’anni,	è	proprio	un	vecchietto.
	Gianni	despite	the	his	thirty-years	be.3SG	really	a	old-DIM
	‘Gianni, despite being thirty years old, really is a little old man.’

This shows that the adjectival base in (27) has the same role as N in A-N complexes: it provides the ontological domain on which a particularized property is typically, but not necessarily, instantiated.

At this point, let us go back to our original puzzle, which constitutes the central topic of the present contribution: Why are evaluative adjectives barred from the prenominal position in Italian?

We believe that all the necessary ingredients for a principled solution to the puzzle have been satisfactorily introduced. In short, the core proposal is outlined as follows:

• In an A-N complex, N undergoes semantic bleaching, and A acquires a particularized, context-sensitive meaning;
• In evaluative adjectives, it is A that undergoes semantic bleaching and EvA acquires a peculiar, contextually driven interpretation;
• The compositional strategy in (1) is syntactically driven;
• The compositional strategy in (2) is pragmatically driven;
• If an element α has already undergone one of these two strategies, it cannot undergo the other;
• Evaluative adjectives are semantically derivative and formed through the pragmatic compositional strategy described in (2);
• Given (5), they cannot also participate in the syntactic compositional strategy described in (1);
• Evaluative adjectives are categorically excluded from the prenominal position in Italian.

We believe this account fares very well in terms of explanatory adequacy. First, it establishes an important analogy between the semantics of a certain syntactic configuration (i.e., A-N) and the semantics of a specific derivational process (i.e., evaluative affixation). Second, it explains why the two strategies, one of a morphopragmatic nature, the other of a syntactic nature, are incompatible: they are variants of the very same compositional process. Third, it shows, if we are correct, that certain interpretive processes cannot apply recursively across syntax and pragmatics.4 Last but not least, it derives the intersective interpretation to which evaluative adjectives give rise in postnominal position: though EvA is not intersective, the derived adjective interpretation that results from applying evaluative morphology to the adjectival base is, of course, permitted to intersectively combine with the preceding noun.

Importantly, this explanation avoids arbitrary stipulations. If correct, our approach to evaluative morphology provides an argument for the unity of morphosyntax and highlights real constraints on recursion in the global processing of morphosyntactic structures.