Is the Foot a Prosodic Domain in European Portuguese?

Abstract

It is widely accepted that languages organize speech material into prosodic domains, which are hierarchically arranged. However, it is still a matter of debate whether this prosodic hierarchy is composed of a small number of universal categories, or whether these prosodic categories are language-particular and emergent. In this article, we concentrate on one of these categories, the foot, and we investigate its role in European Portuguese (EP). Whereas research on EP has shown that other prosodic domains commonly found crosslinguistically are active in the language, it seems that EP may lack this prosodic constituent. Therefore, the goal of this article is to systematically investigate the role of the foot in a number of areas within EP grammar. In our study, we defend some new approaches to several long-standing issues in EP phonology and we conclude that many facts of the language can be, in fact, better understood resorting to the foot domain. Namely, the distribution of stress in the most frequent morphological classes (thematic non-verbs and present tense verbs), -inh-/-zinh- diminutive formation, the domain for regular vowel reduction, obligatory glide formation, the stress window, and poetic rhyme seem to benefit from a metrically-governed account. By contrast, other facts, such as vowel lowering, word clipping, minimal words, secondary stress, and schwa deletion are shown not to be conditioned by the foot in EP. Importantly, the evidence we found for metrical structure only cues the prominent foot of the word, suggesting that pretonic material may not be footed. All in all, in addition to shedding light on facts previously poorly understood, and exposing some noteworthy specificities of EP in the realm of Romance languages and other varieties of Portuguese, we draw implications for the universal characterization of the prosodic hierarchy and its acquisition.

1. Introduction

It is widely accepted that all languages organize speech material into prosodic domains, which are arranged hierarchically, forming a structure that is regulated by a number of constraints (Selkirk 1980, 1984, 1996; Nespor and Vogel 1986/2007; Pierrehumbert and Beckman 1988; Hayes 1989; Inkelas 1990; Itô and Mester 1992/2003 inter alia). There are, nevertheless, major unsettled issues, including whether the prosodic hierarchy is composed of a small set of categories common to all languages (see, among others, Schiering et al. 2010; Hyman 2011; Selkirk 2011; Labrune 2012; Ito and Mester 2012; Frota and Vigário 2018; Bennett and Elfner 2019; Vogel 2019; Downing and Kadenge 2020). In this paper, we will address this question by discussing the status of one prosodic constituent, the foot, in one particular (Romance) language, European Portuguese.

The foot (Ʃ) is one of the domains present in most work considering low-level prosodic phonology in the languages of the world. Within the prosodic tree, scholars agree that it dominates the syllable (σ) and is dominated by the prosodic word (ω). It is usually considered a rhythmic category, consisting of a strong syllable (or mora), which constitutes the head of the foot—and is generally realized with overt stress1—and a weak syllable (or mora), the dependent of the foot, as represented in (1) (throughout the paper we indicate headedness with vertical lines and foot boundaries with parentheses):

(1)

The original motivation for the foot arose in the late seventies based on the typology of stress systems, and the observation that an overwhelming majority of the languages of the world with rhythmic stress display a binary alternation between unstressed and stressed syllables. The assumption that syllables were grouped into a binary constituent with a head-dependent relationship allowed for a simpler and more elegant account of the distribution of stressed syllables (e.g., Liberman and Prince 1977; Hayes 1980, 1995; Selkirk 1980).

Beyond rhythmic stress, independent evidence for the foot category comes from a wide range of phonological and morphophonological phenomena. On the one hand, it has been observed that the domain of application of various segmental processes is often the foot. In particular, many strengthening phenomena (e.g., consonant fortition, vowel lengthening, diphthongization, and the epenthesis of a coda consonant) have been claimed to target metrically strong positions, such as the head of the foot and the foot-initial position (Kiparsky 1979; Selkirk 1980; Nespor and Vogel 1986/2007; Bennett 2012; Martínez-Paricio 2013a, inter alia), whereas weakening processes (e.g., vowel reduction, vowel deletion, and the weakening and deletion of consonants) tend to be circumscribed to the weak branch of a foot, i.e., its dependent (de Lacy 2002; Gouskova 2003; McCarthy 2008). On the other hand, it has been argued that morphophonological patterns such as truncation and reduplication often need to refer to the category of the foot, either because the form of the reduplicative or the truncated morpheme matches that of a foot, or because the size and shape of the derived form corresponds with that of a foot (McCarthy and Prince 1986/1996; Inkelas 2008; Alber and Arndt-Lappe 2012 inter alia). Furthermore, in a number of languages, the particular selection of a phonologically-conditioned allomorph involves the improvement or maintenance of the preferred foot in the language (e.g., Kager 1996; Booij 1998; Nevins 2011). The foot is also relevant for tonal distribution in many languages (e.g., de Lacy 2002; Green 2015; Pearce 2013; Breteler 2017), as well as for phonotactic constraints and the definition of poetic rhyme (e.g., Nespor and Vogel 1986/2007, and references therein). The arguments in favor of the foot are further supported by research carried out in the domain of language acquisition (e.g., Demuth 2015; Fikkert et al. 2021), language processing, and speech planning and production (e.g., Cutler and Norris 1988; Perry et al. 2006; Tilsen 2011; Domahs et al. 2013). Although the foot is often assumed to be universal, there is also work suggesting that some languages, like French, Turkish, or Brazilian Portuguese, lack this domain (Özçelik 2017; Garcia and Guzzo 2022 and references therein).

We believe EP stands out as an optimal testing ground for re-examining the universality of the category of the foot and, indirectly, the universality of the prosodic hierarchy. The prosodic phonology of the language has been extensively studied for the last decades (Frota 2000, 2014; Vigário 2003, 2010, 2022; Veloso 2007, 2012; Fonseca 2012; Cruz 2013; Barros 2014; Frota et al. 2015; Vigário et al. 2019; Paulino et al. 2022, a.o.) and, while phonological facts clearly signal prosodic domains that are commonly found across languages—the syllable, the prosodic word, the prosodic word group (or composite group), the phonological phrase, and the intonational phrase—the foot seems to play only a very limited role in the language, if any at all.

A review of the literature on EP phonology and morphophonology suggests that typical phenomena cuing this domain in other languages are absent, or at least not clearly identified, in EP, e.g., rhythmic-based secondary stress, strengthening phenomena, word clipping, restrictions on vowel reduction and deletion, and word minimality constraints (e.g., d’Andrade and Viana 1989; Frota and Vigário 2000; Mateus and d’Andrade 2000; Vigário 2003; Castelo 2006; Veloso 2017). The foot is mentioned in only a few (controversial) areas, such as mid vowel lowering, which in Brazilian Portuguese (BP) has been argued to depend on particular foot types (Wetzels 1992, 2007)—although it possibly displays a more limited manifestation in current EP—or word primary stress (Pereira 1999; Wetzels 2007; Magalhães 2016). However, with respect to the distribution of primary stress, there have been alternative non-foot-based accounts, which assume that what is relevant for stress assignment in EP is morphological information, or that it is based on lexical storage and probability, without referring to the foot (e.g., Mateus 1983; Mateus and d’Andrade 2000; Wetzels 2007; Garcia 2017; Garcia and Guzzo 2022).

With the aim of answering the question stated in the title of this paper (Is the foot a prosodic category in EP?), in this work we take a closer look at the potential evidence for the foot in the language. We will systematically reexamine the role of the foot in word-level stress assignment, the selection of the diminutive formatives -inh-/-zinh-, vowel reduction, glide formation, the size of the stress window, poetic rhyme, vowel lowering, word clipping and hypocoristic formation, minimal words, secondary stress, and schwa deletion, most of these being long-standing issues in the literature on EP (e.g., Mateus 1983; Pereira 1999; Mateus and d’Andrade 2000; Vigário 2003; Wetzels 2007; Villalva 2008; Rio-Torto 2013).

2. Evidence for the Foot in European Portuguese

2.1. Word-Level Stress Assignment

Like in other Romance languages, word stress assignment is one of the topics most discussed in the literature on Portuguese phonology (see Magalhães 2016 for an overview). In general, it is assumed that Brazilian and European varieties of Portuguese have the same grammar of stress (e.g., Lee 1995; Mateus and d’Andrade 2000; Garcia 2017). Here, we focus specifically on EP. We note that there are important differences in the phonology of the two varieties, some of them potentially bearing on the grammar of stress assignment. In addition, there are also many obvious similarities in the distribution of stress in EP and in other Romance languages, which we will only expose episodically, in the interest of readability.

Before we proceed, we give below basic elements on Portuguese morphology in a nutshell, whilst skipping details that are orthogonal to the main point of this section (see Mateus and d’Andrade 2000; Villalva 2008; Rio-Torto 2013; Villalva and Gonçalves 2016, and references therein):

(i)

Inflected verbs instantiate the following strictly ordered morphological components: (a) a derivational root, containing a root, and possible derivational affixes (Rt)—obligatory; (b) a theme vowel or thematic constituent (TC)—obligatory; the TC -a is the most frequent and the only productive TC; -e and -i are also quite frequent TC; the TC is deleted when followed by an affix starting with a vowel, except when the TC and the following vowel are both non-open, non-back, in which case they undergo fusion (see also note 2); (c) a tense–mood–aspect morpheme (TMA)—a single morpheme instantiating the categories of Tense (past, present, and future/conditional), Aspect (perfect and imperfect), and Mood (indicative and subjunctive); in present indicative, TMA is not visible (this may be analysed as lack of a TMA morpheme or presence of a Ø morpheme); TMA is always present elsewhere; and a person–number morpheme (PN)—usually with no phonetic content in the 3Sg, and also in the 1Sg (again, this may be analysed as lack of PN or presence of a Ø morpheme), but obligatory elsewhere.2 Examples are given in (2a), below.

(ii)

Non-verbs may be (a) thematic, in which case they are composed of the derivational root (Rt), plus the thematic indexes (or TC) -a, -u (spelt as <o>), or -e, or (b) athematic, in which case the Rt also corresponds to the stem; both thematic and athematic non-verbs may inflect for plural (Pl), by adding a partially underspecified coronal fricative (/S/).3^,4^,5 Examples are given in (2b), below. Hereafter, boldface signals the syllable bearing word stress in orthographic forms.

(2)	a.	Inflected verb forms
		fala	falRt aTC	‘speakPres.Ind.3Sg’
		falavas	falRt aTC vaTMA sPN	‘speakPast.Imp.Ind.2Sg’
		falemos	falRt aTC eTMA mosPN	‘speakPres. Subj.1Pl’

	b.	Thematic non-verbs (Sg, Pl)
		menino/meninos	meninRt oTC sPl	‘boy(s)’
		batata/batatas	batatRt aTC sPl	‘potato(s)’
		envelope/envelopes	envelopRt eTC sPl	‘envelop(s)’

		Athematic non-verbs
		papel/papéis	papelRt,St/papejRt,St sPl	‘paper(s)’
		andar/andares	andarRt sPl	‘flat(s)’
		rapaz/rapazes	rapazRt sPl	‘boy(s)’
		café/cafés	caféRt sPl	‘coffee(s)’
		menu/menus	menuRt sPl	‘menu(s)’

(iii)

The root (Rt) and the thematic constituent (TC) form a morphological constituent, the stem (St), both in verbs and non-verbs (Villalva and Gonçalves 2016). In athematic non-verbs, which lack a TC, the root and the stem coincide, as in café and cafés ‘coffee(s)’. Inflectional affixes are morphosyntactic specifiers of the stem (MSS) (Villalva and Gonçalves 2016). According to Villalva (2008), inflectional morphemes form a morphological domain: in verbs the inflection domain contains TMA and PN suffixes; in non-verbs, it contains the Pl suffix. If the word lacks MSS, the St and morphological word (MW) coincide, as in café ‘coffee’ (but not in cafés ‘coffees’). Non-compound content words in EP display the structure in (3), where MSS may be null in verbs and MSS and TC may be null in non-verbs (4).

(3)	[[[X]Rt [Y]TC]St [Z]MSS ]MW (adapted from Villalva and Gonçalves 2016: 167)

(4)	[[[fal]Rt [a]TC]St [va s]MSS ]MW	‘speakPast.Imp.Ind.2Sg’
	[[[fal]Rt [a]TC]St [ ]MSS ]MW	‘speakPres.Ind.3Sg’
	[[[menin]Rt [o]TC]St [s]MSS ]MW	‘boys’
	[[[café]Rt [ ]TC]St [ ]MSS ]MW	‘coffee’

Ignoring morphological information, EP words are typically stressed in the penultimate syllable (about 54%) and the final one (about 43%, of which around 28% are monosyllabic words, which pattern like stressed final syllables of polysyllabic words), whereas antepenult stress is residual (less than 2%) (token frequency—cf. Vigário et al. 2010; Vigário 2022; see also Viana et al. 1996). If we consider the most frequent ω format, with less than three syllables (corresponding to 71% of adult data), the proportion of penult and final stress in adult speech is very similar, with a small advantage to final stress (Vigário et al. 2010).6

Despite many areas of controversy in the literature on the topic, currently, there seems to be agreement on two major points about stress assignment (e.g., Lee 1995, 2007; Pereira 1999; Mateus and d’Andrade 2000; Vigário 2003, 2022; Wetzels 2007; Magalhães 2016; Garcia 2017): (i) verbs and non-verbs are distinct subsystems, stress-wise, and (ii) morphological structure is relevant for stress assignment in verbs. In this article, we will challenge the different treatment of non-verbal and verbal stress in traditional analyses. Furthermore, we will assume that morphology is sometimes relevant in stress assignment, but also in purely rhythmical phonological constraints. Illustrative examples of the position of stress in verbs and non-verbs appear in (5): in the present tense (indicative and subjunctive), in most forms, though not all, stress is on the final syllable of the root; in the past tense (indicative and subjunctive), stress always falls on the syllable containing the theme vowel; in future and conditional, stress falls on the first syllable of the TMA morpheme, which also corresponds to the syllable immediately following the theme vowel (see 5a); and in non-verbs, stress tends to occur in the syllable containing the last syllable of Rt—in thematic words, that syllable is the penultimate, and in athematic words, that syllable is often closed by C(onsonant) or G(lide) (see 5b), although there are exceptions to that pattern.

(5)	a.	Stress distribution in inflected forms (falar ‘to speak’)
	Present			1Sg	falo
				2Sg	falas
				3Pl	falam
	Past			1/3Sg	falava
				1Pl	falávamos
				3Pl	falavam
	Future/Conditional			1/3Sg	falaria
				1Pl	falaríamos
				3Pl	falariam
	b.	Stress distribution in non-verbs
			sapato	‘shoe’
			batata	‘potato’
			animal	‘animal’
			chapéu	‘hat’

There are different views on other areas, however, such as the following:

(i)

The role of syllable weight—see in particular the discussion in Pereira (1999), who defends that in EP what matters is number of syllables at the right edge in present tense verbs, and the morphological structure and lexical marking of elements that cannot bear stress elsewhere, and Wetzels (2007), who extensively argues for the role of syllable quantity in (B)P non-verbs only (see also the review of Magalhães 2016).

(ii)

The relevance of metrical foot construction in computing stress location—e.g., for Pereira (1999) present tense forms are assigned stress via the formation of trochaic feet, whereas for Wetzels (2007), in the present tense, what accounts for stress location is morphological position—stress falls on the last vowel of the root, and 1/2Pl is subject to a specific constraint that ensures that stress falls on the theme vowel; by contrast, in non-verbs, the general rule in Pereira assigns stress to the final syllable of the (derivational) root, whereas for Wetzels, stress is assigned via the formation of quantity-sensitive trochaic feet (like Bisol 1992, and others).

(iii)

The extent to which there is a general pattern for stress assignment in non-verbs, with some exceptions marked in the lexicon (as assumed in most current literature on the topic), or stress is simply lexically marked for every non-verb (Garcia 2017).

(iv)

The domain for stress assignment, whether this is the (derivational) root, the stem, or the word, either morphological or prosodic—e.g., for Pereira (1999), the domain for stress in non-verbs is the (derivational) root, whereas in verbs it is the word.

(v)

The role of foot structure in stress assignment—most frequently, under the more widespread approaches, where stress is morphologically conditioned but there is also metrical stress, the foot is relevant for determining stress location in some areas of the lexicon, such as non-verbs, but not in verbs, where stress location is obtained as a function of morphological information, and foot construction takes stress location into account, via alignment constraints, as in Wetzels’ (2007) approach.

Purely metrical accounts of EP stress (e.g., Bisol 1992) need to make extensive use of extrametricality, or similar devices. Lexical listing, on the other hand, as in Garcia’s (2017) approach to stress in non-verbs, disregards the effect of the regularities in the system in plausibly leading a child acquiring EP to extracting very general rules or constraints for stress assignment. Finally, the various approaches that assign a relevant part to both prosodic (syllable or mora count; foot formation) and morphological information (such as morphological position or tense information) attribute different roles to metrical feet in verbs and non-verbs (see ii, above). In fact, if one adopts Wetzels’ (2007) approach to stress assignment in verbs (entirely dependent on morphological and lexical information) and Pereira’s (1999) views on stress assignment in non-verbs (the assignment of stress to the last syllable of the derivational root and lexical information), the result is such that there is in reality no need to assume that the foot is relevant for stress distribution, since the foot itself does not contribute to the computation of stress. A systematic evaluation of the role of the foot in EP stress assignment is therefore necessary.

In what follows, we will defend that, in order to uncover the relevant regularities of the EP stress system, it is indeed necessary to recognize the relevance of both phonological and morphological information for stress assignment, like, for instance, in Lee (1995, 2007), Pereira (1999), Mateus and d’Andrade (2000), and Wetzels (2007). However, although using a similar sort of morphophonological information that has been considered relevant for stress distribution in the literature on Portuguese, in this work we propose a new approach to the location of stress in EP, which gives the metrical foot a crucial role. We hope to show that this is the most general and less stipulative account of EP stress, and hence, potentially the one that is most likely to be learned by children acquiring the language. Importantly, we must keep in mind that the basic goal of this section is to determine if the foot plays a role in stress placement in the language. Therefore, we will make an effort to stick to the fundamental ideas concerning the topic, leaving a fully-fledged account of stress in EP to future work.

Like in most literature on Portuguese stress, we shall assume that there are general principles for stress assignment, and only exceptions to these principles need to be listed in speakers’ mental lexicon. By contrast, we depart from other authors in defending that there is no difference in the mechanism of stress assignment in verbs and non-verbs. We propose that verbs and non-verbs are subject to the exact same general constraints on stress assignment, where metrical stress, and the assignment of stress with reference to a designated morphophonological position, are the key features. We also assume that EP is not quantity-sensitive (see Pereira 1999, and some additional evidence below).

We propose a general mechanism for stress assignment involving the following:

(a)

Phonological metrics: metrical stress springs from the formation of syllabic trochees—the same foot type for designated subcategories of verbs and non-verbs.

(b)

Reference to a morphological position: the location of stress is obtained with reference to a morphological position—also the same position for designated subcategories of verbs and non-verbs.

Stress on unpredictable positions, i.e., not predicted by (a) or (b), is lexically specified. In the next paragraphs, we briefly describe how the system works.

In the absence of any relevant morphological/lexical marking, words build a left headed binary foot (i.e., a trochee). The subcategories that undergo metrical stress, are present tense verb forms (6a) and thematic non-verbs (6b), which are the most frequent subcategories of EP verbs and non-verbs, respectively.

(6)	a.	(fa.la)	‘speakPres.Ind.3Sg’	(fa.le)	‘speakPres.Subj.1/3Sg’
		(fa.las)	‘speakPres.Ind.2Sg’	(fa.les)	‘speakPres.Subj.2Sg’
		fa(la.mos)	‘speakPres.Ind.1Pl’	fa(le.mos)	‘speakPres.Subj.1Pl’
		(fa.lam)	‘speakPres.Ind.3Pl’	(fa.lem)	‘speakPres.Subj.3Pl’
	b.	sa(pa.to)	‘shoe’
		ba(ta.ta)	‘potato’
		car(pe.te)	‘carpet’

As pointed out in most current literature, stress in verbs shows a fixed position in all forms in the past tense. In this subclass, morphology is important for stress location. Although we will not develop here a formal analysis of stress assignment in EP, a potential Optimality Theory (OT) account of the facts would need to assume that there are morphologically indexed constraints that regulate the assignment of stress in these forms: instead of assigning stress to the head of the trochee, what is crucial in these subcategories is the morphological position. This type of stress assignment is not purely phonological in the sense that it needs to “see” morphological information. We propose that what accounts for the fixed position of stress in these forms is a general constraint imposing that stress falls on the syllable containing the last vowel of the St. Importantly, the same constraint accounts for stress in athematic non-verbs.7 This is illustrated in (7).8

(7)	a. Verbs—Past (falar ‘speak’)
		Indicative Imperfect			Subjunctive Imperfect
	1/3Sg	falava	fal a)St va		falasse	fal a)St sse
	2Sg	falavas	fal a)St va s		falasses	fal a)St sse s
	1Pl	falávamos	fal a)St va mos		falássemos	fal a)St sse mos
	3Pl	falavam	fal a)St va m		falassem	fal a)St sse m
	b. Non-verbs—Athematic
		papel	papel)St	‘paper’
		andar	andar)St	‘flat’
		Inglês	Inglês)St	‘English’
		chapéu	chapéu)St	‘hat’
		herói	herói)St	‘heroe’
		bisturi	bisturi)St	‘scalpel’
		menu	menu)St	‘menu’
		você	você)St	‘you’
		avô	avô)St	‘grandpa’
		café	café)St	‘coffee’
		avó	avó)St	‘grandma’
		sofá	sofá)St	‘sofa’

In the future and the conditional, stress is not stem final, but it always surfaces on a fixed morphological position, i.e., the first syllable of a TMA affix, irrespective of number of syllables or syllable composition at the right edge of the word, see (8). We adopt here the general idea defended at length in Vigário (1999, 2003, chp. 4), which departs from others, such as Pereira’s (1999), Mateus and d’Andrade (2000), and Villalva (2008): in current EP, future and conditional inflected forms are built like the inflected forms in the rest of the verbal paradigm, and are composed of the (derivational) root, the TC, and TMA (ra/re, ria) and PN suffixes.9

(8)		Future—Indicative	Conditional
	1Sg	falarei	falaria
	2Sg	falarás	falarias
	3Sg	falará	falaria
	1Pl	falaremos	falaríamos
	3Pl	falarão	falariam

Given this pattern, within our general approach, there are at least two possible motivations for the fixed stress location in future and conditional: (a) stress may be required to fall on the syllable occupying a specific morphological position—the initial position of the inflection domain, or (b) given the historical derivation of the future and conditional, it is also conceivable that stress is a lexical property of the TMA morpheme.10 We will briefly discuss each option in the following paragraphs.

An argument in favor of (a) would be the fact that, under our approach, regular stress in verbs and non-verbs is either metrical or sensitive to a morphological position, whereas lexical marking is left for unpredictable exceptions; (a) captures the fact that stress location in future and conditional forms is regular, in the sense that it always appears in the same fixed position; by contrast, admitting (b) would allow a system whereby morphologically sensitive stress would always refer to the same morphological position, in verbs and non-verbs, i.e., St., and not St and the inflection domain.

In fact, data from non-standard varieties of EP, indicate that stress in verbs is closer to becoming assigned with reference only to different morphological positions. In these non-standard varieties, in subjunctive present (though not in indicative present), stress is always on the syllable containing the last vowel of the verb Rt, irrespective of the number of syllables from the right margin of the word (Pereira 1999; Vigário 2022), as illustrated in (9), with inflected forms of supor ‘suppose’ and perceber ‘understand’.11

(9)	Non-standard present subjunctive inflection
	1/3Sg	suponha
	2Sg	suponhas
	1Pl	supônhamos (suponhamos, in Standard EP)
	3Pl	suponham
	1Sg	perceba
	2Sg	percebas
	1Pl	percêbamos (percebamos, in Standard EP)
	3Pl	percebam

Interestingly enough, even in non-standard subvarieties of EP, stress in the indicative present is still purely prosodic, falling on the penult syllable in all inflected forms. The resistance of the indicative present to be assigned stress with reference to Rt may be explained by the high frequency of indicative present, as suggested in Vigário (2022) (see also Bybee and Hopper 2001 for other areas where high frequency has inhibited regularization). In any event, at least for non-standard subvarieties of EP, the hypothesis in (a) is to be favored, since, to the exception of the present indicative, stress location can be fully obtained with reference to morphological position (last syllable of the Rt, last syllable of the St, and first syllable of the inflection domain).

As said above, exceptions to the general mechanism are marked in the lexicon, as in any other framework. Assuming that verbs display only metrically-based and morphological sensitive stress, exceptions are only found in non-verbs: antepenult stress in thematic non-verbs (10a), and penult and antepenult stress in athematic non-verbs, as in (10b):

(10)	a. Exceptional antepenult stress in thematic non-verbs
		genético	(instead of *genetico)	‘genetic’
		calêndula	(instead of *calendula)	‘calendula’
		ênfase	(instead of *enfase)	‘emphasis’
	b. Exceptional penult and antepenult stress in athematic non-verbs
		líder	(instead of *lider)	‘leader’
		fácil	(instead of *facil)	‘easy’
		Lúcifer	(instead of *Lucifer)	‘id.’
		Júpiter	(instead of *Jupiter)	‘id.‘

In short, EP facts indicate that foot formation is the source of stress assignment in thematic non-verbs and in present tense verbs, that is, stress is obtained via trochaic foot formation, operating at the level of the word, since PN inflectional affixes contribute to the computation of stress. By contrast, in athematic non-verbs, in past tense verb forms, in the future and conditional, and in words with unpredictable stress, stress is assigned on the basis of morphologic position or lexical specification—syllables bearing this type of stress are marked to be the head of a foot. This approach can be formalised in OT terms by making use of well-established constraints, appropriately ranked, including (i) constraints on foot form, ensuring for instance that feet are bisyllabic, (ii) alignment constraints, enforcing initial foot headedness and the alignment of designated aspects of prosodic and morphological structure, some of which are indexed to particular morphological subcategories, and (iii) faithfulness constraints, ensuring that, when stress is lexically marked, this is reflected in the output.12

The system we propose is highly consistent, and more so than competing models: for instance, morphological information relevant for stress is reduced to structural positions, not specific morphemes for verbs and morphological position for non-verbs, as in Pereira (1999), and it avoids rather ad hoc stipulations, as in the accounts of the present tense that rely on morphological position, such as Wetzels (2007), or in purely phonological approaches, requiring an extensive use of extrametricality or similar devices. In particular, for both verbs and non-verbs, stress involves either the formation of a syllabic trochee or stress assignment on the syllable containing the final vowel of the stem (in addition to the lexical specification of exceptions, which must be assumed in any approach to EP stress).

Our analysis relies on the assumption that stress assignment in EP is not sensitive to syllable quantity, a position shared with, for instance, Pereira (1999) and Mateus and d’Andrade (2000). A number of arguments support our view:

(i)

In inflected verbs, in the present tense, where stress is prosodically-based, syllables closed by nasals or consonants at the right edge of the word do not attract stress (batem_Pres.Ind.3Pl, batemos_Pres.Ind.1Pl, and bates_Pres.Ind.2Sg).

(ii)

In non-verbs, syllables closed by /S/ at the right edge of the word pattern variably with respect to stress: an inflectional /S/ in word final position does not trigger final stress (janelas ‘windows’), whereas a non-inflectional final /S/ triggers final stress (rapaz ‘boy’). Under an approach whereby stress is quantity-sensitive, the reason for this asymmetry is not clear; in our account, the difference is expected, since metrical stress is not sensitive to quantity (and hence the presence or absence of /S/ yields the same result, that is, a syllabic trochee) and stress assigned to the syllable containing the final vowel of the stem simply ignores the syllable (number and/or composition).

(iii)

Under our account, the tendency for final stress in words ending in a diphthong or non-inflectional consonant, which is an important argument suggesting the relevance of syllable weight (chapéu ‘hat’), springs from the fact that in non-verbs such endings appear in the stem final position of athematic classes; crucially, the fact that the same behavior is found in athematic words with stems ending in vowel (café ‘coffee’) strongly suggests that what is relevant is morphological position, instead of syllable quantity.

(iv)

Along the lines of Pereira (1999), we may account for the fact that in the Portuguese lexicon stress does not appear to the left of closed syllables in penult position (*σHσ) by admitting that a different phonological grammar shaped the lexicon of the language, which however is no longer active in the present day; loanwords seem to support this view, since stress is preserved in the antepenult syllable in words like Washington, and Parkinson.

Importantly, within an integrated approach to stress in EP verbs and non-verbs, it is expected that stress distribution plays no special role in distinguishing among major word classes. This is indeed the case. Despite the fact that the examples of the contrastive role of stress found in the literature most often involve nouns vs. verbs, like in English, contrast based on stress alone may involve non-verbs vs. verbs, as shown in (11a), as well as words belonging to the same major word class, as illustrated in (11b); in addition, homophony with verbs and non-verbs showing the same stress pattern is not uncommon, as exemplified in (11c).

(11)	a.	música (N)	‘music’	/	musica (V)	‘musicPres.Ind.3Sg’
		crítica (N)	‘criticism’	/	critica (V)	‘criticizePres.Ind.3Sg’
		furão (N)	‘ferret’	/	furam (V)	‘drillPres.Ind.3Pl’
		zangão (N)	‘bumblebee’	/	zangam (V)	‘get madPres.Ind.3Pl’
		lícito (A)	‘legal’	/	licito (V)	‘betPres.Ind.1Sg’
		vívido (A)	‘vivid’	/	vivido (V)	‘livePast.Part’
	b.	bambo (A)	‘wobbly’	/	bambu (N)	‘baboo’
		têmpora (N)	‘temple’	/	tempura (N)	‘tempura’
		partiram (V)	‘leavePast.Perf.3Pl’	/	partirão (V)	‘leaveFut.Ind.3Pl’
		incidiram (V)	‘fallPast.Perf.3Pl’	/	incidirão (V)	‘fallFut.Ind.3Pl’
		contem (V)	‘tellPres.Subj.3Sg’	/	contém (V)	‘includePres.Ind.3Sg’
	c.	canto (N/V)	‘corner; chant’, ‘singPres.Ind.1Sg’
		manifesto (N/V)	‘manifest’, ‘manifestPres.Ind.1Sg’
		conquistas (N/V)	‘conquests’, ‘conquer Pres.Ind.2Sg’
		verão (N/V)	‘Summer’, ‘seeFut.Ind.3Pl’
		arranque (N/V)	‘start’, ‘startSubj.Pres.1/3PSg’
		acórdão (N) ‘court decision’ / acordam (V) ‘wake upPres.Ind.3Pl’

As the examples show, contrasts may involve all sorts of stress assignment rules, and absence of contrast may spring from the same rule (e.g., metrical stress, in both verbs and non-verbs, as in canto, manifesto, conquistas, and arranque) or different stress rules (e.g., stress on the last syllable of the stem vs. stress on the first syllable of the inflection domain/lexical stress, as in verão, or lexical stress vs. metrical stress, as in acórdão/acordam).

Other scholars have defended morphologically-based stress assignment for non-verbs of both classes of words: thematic and athematic (e.g., Lee 1995; Pereira 1999; Mateus et al. 2003). We have not followed this approach for the following main reasons: (i) within our system, there is a perfect parallelism between verbs and non-verbs in the domain that is relevant for morphologically-based stress (the stem), as well as in the type of subclass that is subject to prosodically-based stress (in both cases, the most frequent subclasses of verbs and non-verbs, respectively, i.e., inflected verbs (present) and thematic non-verbs; by contrast, the generalization that stress falls on the syllable containing the last vowel of the root is only true for non-verbs, but not for verbs; (ii) there is a morphophonological process where the distinction between prosodically-based stress and morphologically-based and lexical stress seems to be relevant, which is diminutive/evaluative formation with the suffix -inh- (see Section 2.2).

It is, nevertheless, observationally true that, in regular non-verbs, the syllable containing the last vowel of the root bears stress, both in thematic and athematic non-verbs (cf. batatas: batat)_Rt a)_St s)_MW ‘potatos’; cafés: café)_Rt )_St s)_MW ‘coffees’). We believe this may indeed be a constraint of the language. However, we put forward the hypothesis that in languages with word stress, the default mechanism for stress assignment is prosody-based (Selkirk 1978/1981; Nespor and Vogel 1986/2007; Hayes 1995), and when data are compatible with phonological assignment, that should be the one operating, unless there is compelling evidence against it. Stress in the most frequent subclasses of verbs and non-verbs may be obtained by a default process of trochaic foot formation, the result of which satisfies both the phonological constraints, and the constraint requiring that syllables containing the last vowel of the root are stressed. Redundancy is part of the system, with benefits for word processing.

2.2. Diminutive/Evaluative Formation with -inh-

Diminutive/evaluative (hereafter, DIM) formation with -inh- and -zinh- is extremely productive in Portuguese. In general, any base of the category N(oun) and A(djective) can undergo DIM with at least one of these forms.13 Furthermore, DIM formation with -inh-/-zinh- is extremely frequent in familiar contexts and, in particular, in Child Directed Speech. Even in a corpus of rather formal Adult Directed Speech, analysed in Vigário and Garcia (2012), as much as 3.9% of total tokens and 8.9% of total types show (-z)inh- affixation.

DIM suffixation with -inh- and -zinh- is a complex matter, investigated by many scholars, with different views on several issues, some of which dependent on the variety of Portuguese considered (EP or BP), such as (i) whether -inh- and -zinh- are different morphemes or allomorphs of the same morpheme, (ii) what type of morphological base each of these elements attaches to a root, a stem, or a word, (iii) whether the base must be an independent ω, and (iv) what factors account for the occurrence of -inh- and -zinh- (see, among others, Rio-Torto 1993, 2013; Menuzzi 1993; Lee 1995, 2013; Villalva 2008; Bisol 2010; Ulrich and Schwindt 2018, and the references therein). Like before, we will try to avoid aspects of the discussion that are not central to our purpose, including important differences in this area between EP and BP. In the next lines, we will focus our attention on the role of stress patterns in DIM suffixation involving -inh- and -zinh- in EP.

In EP, there is sound evidence that the base -inh- attaches to is either a root or a stem, but not a word, whereas -zinh- attaches to full words (Rio-Torto 1993, 2013; Mateus et al. 2003; Villalva 2008, and references therein). The base -inh- attaches to a pattern very much like other typical bases that undergo suffixation: they have no independent stress (as indicated, for instance, by vowel reduction facts14), and do not show thematic index or inflectional suffixation (12); by contrast, the base -zinh- attaches to forms an independent stress domain (13i), when the base is thematic, the thematic index surfaces in the base (13ii), and, if the resulting word is plural, plural suffixation is found both in the base and in the resulting word (13iii).15^,16

(12)	patos - p[á]tos ‘duckMasc.Pl’ / patinhos - p[ɐ]tinhos, *p[a]tinhos	‘duckDIM.Masc.Pl’
	peras - p[é]ras ‘pearDIM.Fem’ / perinhas - p[ɨ]rinhas, *p[e]rinhas	‘pearDIM.Fem.Pl’
	gordos - g[ó]rdos ‘fatMasc.Pl’ / gordinhos - g[u]rdinhos, *g[o]rdinhos	‘fatDIM.Masc.Pl’

(13)	(i)	amável - am[á]vel ‘gentle’ / amavelzinho - am[á]velz[í]nho, *am[ɐ]velz[í]nho
		‘gentle(DIM)’
		avô - av[ó] / avozinho - av[ó]zinho, *av[u]z[í]nho			‘grandpa(DIM)’
	(ii)	pêndulo - pendulozinho			‘pendulum(DIM)’
		carapaça - carapaçazinha			‘shell(DIM)’
	(iii)	amável / amavelzinho ‘gentleSg(DIM)’ vs. amáveis / amaveizinhos ‘gentlePl(DIM)’
		leão / leãozinho   ‘lionSg(DIM)’ vs. leões / leõezinhos   ‘lionPl(DIM)’

In general, if semantically appropriate, -zinh- can be added to any word of the categories N and A, although, when -inh- is also possible, -inh- is usually preferred over -zinh-. What is of interest to us is the fact that -inh-, in contrast with -zinh-, cannot be added to any base of these categories. Importantly, -zinh- may be added to morphological bases containing the roots that allow -inh- (see 14a), and even to bases with -inh- affixation, as illustrated in (14b), whereas the converse is not true (14c–d).17

(14)	a.	livrinho/livrozinho	‘bookDIM’
		almofadinha/almofadazinha	‘pillowDIM’
	b.	livrinhozinho	‘bookDIM’
		almofadinhazinha	‘pillowDIM’
	c.	sofazinho/*sofainho	‘sofaDIM’
		europeuzinho/*europeuinho	‘EuropeanDIM’
		seculozinho/*seculinho	‘centuryDIM’
	d.	*livrozinhinho	‘bookDIM’
		*almofadazinhinha	‘pillowDIM’

Given these observations, we will explore here the contexts where -inh- is forbidden, rather than where -zinh- occurs. In the literature, in addition to the morphological subclass (thematic/athematic), several phonological properties of the base are typically mentioned as relevant, e.g., -inh- cannot be added to bases ending in nasal vowels and diphthongs (whether thematic or athematic), as in (15)18, nor to bases ending in a vowel and diphthong with final stress, as in (16) (Rio-Torto 2013); there is also a very strong tendency not to have -inh- attached to athematic roots ending in C (with some additional factors possibly playing a role), as in (17). For BP, Bisol (2010) also observes that -zinh- tends to be preferred with bases with antepenult stress. We believe that in EP too, -inh- tends not to occur with thematic bases with lexically marked stress, as in (18) (note that there are exceptions to the patterns in (17) and (18), which we believe are simply lexically stored). In addition, any athematic base with exceptional stress blocks DIM formation with -inh-, as illustrated in (19).

(15)	irmã	irmãzinha/*irmãinha	‘sister(DIM)’
	irmão	irmãozinho/*irmãinho, *irmãoinho	‘brother(DIM)’
	jardim	jardinzinho/*jardinho, *jard[ı̃]inho	‘garden(DIM) ’
(16)	café	cafezinho/*cafeinho	‘coffee(DIM)’
	europeu	europeuzinho/*europeuinho	‘european(DIM)’
(17)	pomar	pomarzinho/*pomarinho	‘orchard(DIM)’
	tropical	tropicalzinho/*tropicalinho	‘tropical(DIM)’
(18)	século	seculozinho/*seculinho	‘century(DIM)’
	frívolo	frivolozinho/*frivolinho	‘frivolous(DIM)’
	câmara	camarazinha/*camarinha	‘camera(DIM)’
	frigorífico	frigorificozinho/*frigorifiquinho	‘fridge(DIM)’
	pálida	palidazinha/*palidinha	‘pale(DIM)’
(19)	Lúcifer	Luciferzinho/*Luciferinho	‘Lucifer(DIM)’
	júnior	juniorzinho/*juniorinho	‘junior(DIM)’
	álcool	alcoozinho/*alcolinho	‘alcohol(DIM)’
	líder	liderzinho/*liderinho	‘leader(DIM)’
	átlas	atlazinho/*atlasinho	‘atlas(DIM)’
	abdómen	abdomenzinho/*abdomeninho	‘abdomen(DIM)’

Even though there are other phonological constraints regulating the use of -inh-, we would like to suggest that some of the most important ones are related to stress. Specifically, we propose that (i) -inh- attaches to morphological stems,19 and (ii) the stems -inh- attaches to are not marked for stress. The latter restriction excludes as potential bases for -inh- (i) stems that are marked with morphologically-based stress, that is, athematic stems, irrespective of the type of segment in final position, (ii) thematic bases where the thematic marker and the final vowel of the stem underwent fusion, which pattern, stress-wise, as athematic bases, and (iii) bases with lexically marked stress, whether thematic or athematic. The bases to which -inh- may be added are therefore limited to thematic stems that are not marked for stress. Where this pattern does not occur, we believe other factors may be at play (such as lexicalization, or further phonological restriction on -inh- and/or -zinh- attachment).20

Importantly, to account for the data, it is insufficient to say that -inh- attaches to thematic bases, since -inh- is usually blocked if the thematic base is marked for stress and it is impossible if there is fusion between the phonological material of the base and thematic index. What the lexical marking of stress and morphology-sensitive stress have in common is the fact that, in both cases, the presence of stress on the base is available at the level of the stem, whereas in thematic words, stress is metrically-based and assigned only when the word is formed, including inflection suffixation.

-inh- formation does not provide new evidence for the foot in EP, but it indicates that the distinction we proposed in Section 2.1 within the class of non-verbs is adequate. In particular, in thematic non-verbs, stress assignment is metrically-conditioned, i.e., dependent on trochaic foot formation, like in present tense verbs. In other words, stress in thematic non-verbs is assigned to the penult syllable of the word, even though the result is compatible with the constraint that, in this subclass only, the syllable containing the last vowel of the root is stressed.

2.3. Vowel Reduction

Vowel reduction is quite a general process that affects unstressed vowels in EP, reducing a phonological system of seven vowels (/i, e, ɛ, a, ɔ, o, u/) to four ([i, ɨ, ɐ, u]), as illustrated in (20). Within Romance languages, EP stands out for the extension to which unstressed oral vowels are reduced.

(20)	fala [ˈfalɐ]	‘(he) speaks’
	falava [fɐˈlavɐ]	‘(he) spoke’
	falaria [fɐlɐˈɾiɐ]	‘(he) would speak’

There are contexts where VR regularly does not apply. Unstressed vowels do not reduce in the following:

(1)

Syllables with post-vocalic glides, nasals, and laterals (Mateus 1975/1982; Mateus and d’Andrade 2000), that is, syllables with a branching nucleus, assuming partial nuclearization of /l/ (Vigário 2022, and references therein)—vaidade [vajˈdad] ‘vanity’ and beldade [bɛlˈdad] ‘beauty’.

(2)

The ω-initial position (non-central V) (Mateus 1975/1982; Mateus and d’Andrade 2000); here, there is a partial reduction (Vigário 2003)—erguer [eɾˈɡeɾ]~[iɾˈɡeɾ] ‘rise’ and ocupar[okuˈpaɾ]~[ɔkuˈpaɾ] ‘occupy’.

(3)

The final position of a ω containing non-thematic material: athematic non-verbs, non-final roots within a morphological compound, clipped words resulting from morphological compounds, and stressed prefixes—líder [ˈlidɛɾ] ‘leader’ and monogâmico [ˈmɔnɔˈɡɐmiku] ‘monogamous’ (Vigário 2003, 2016, 2022).21

Despite the fact that, most often, non-high vowels reduce, there are also unpredictable exceptions to VR, as shown in (21) (Mateus 1975/1982; Mateus and d’Andrade 2000; Bisol and Veloso 2016):22

(21)	protocular	[pɾɔtɔkuˈlaɾ]/*[pɾutukuˈlaɾ]	‘protocol’
	preleção	[pɾɨlɛˈsɐ̃w̃]/*[pɾɨlɨˈsɐ̃w̃]	‘lecture’
	afetividade	[ɐfɛtiviˈdad]/ *[ɐfɨtiviˈdad]	‘affectivity’
	retórica	[ʀɛˈtɔɾikɐ]/*[ʀɨˈtɔɾikɐ]	‘rhetoric’
	Aveiro	[aˈvɐjɾu]/*[ɐˈvɐjɾu]	‘Id.’

It seems clear that exceptions are not dependent on the length of words and their potential foot composition. For example, in a long word like in (22), vowel reduction applies to all unstressed vowels.

(22)	perceberás (/peɾsebeɾaS/) [pɨɾsɨbɨˈɾaʃ] (>[pɾsbˈɾaʃ]) ‘understandFut.2sg’

Crucially, however, there are no unpredictable exceptions in posttonic position (Vigário 2003; see also Marquilhas 2003, for word final position).23 In addition, there is no distinction among posttonic non-final versus final unstressed syllables in EP (unlike, for instance, in Brazilian Portuguese).

These generalizations can be perfectly captured if we assume that EP ω are headed by a right-aligned, left-headed foot and regular vowel reduction is metrically governed, in the sense that it targets the weak branch(es) of the foot. There is ample cross-linguistic evidence that the weak branch of a foot is the target of multiple weakening processes such as vowel reduction, vowel syncope, vowel deletion, and consonant deletion (Booij 1977; Gouskova 2003; McCarthy 2008; among many others). Even though VR in EP can target other positions as well (as seen above), the fact that there are no unpredictable exceptions in posttonic position does point to the presence of a prosodic domain (i.e., a foot), containing a head and one or two potential dependents, based on the foot type, as in prata [(ˈpɾa.tɐ)_Ʃ] ‘silver’, with a regular trochee, and pétala [((ˈpɛ.tɐ)_Ʃ lɐ)_Ʃ] ‘petal’, with an internally layered foot consisting of a trochee and an adjunct, as proposed in some metrical models (see Martínez-Paricio and Kager 2015, and Section 2.5 for additional relevant details).

It is well known that in the closely related variety of Portuguese, BP, unstressed vowels’ reduction applies differently in posttonic and in pretonic position (Câmara 1970), indicating distinct domains for vowel reduction (Wetzels 1992; Lee 1995; Bisol and Veloso 2016). Given the observations above, we also propose that there is a specific domain for regular VR in EP. We suggest that VR applies regularly within the (ω-head) foot domain, equally affecting all weak branches of the foot.

According to this view, aspects of vowel reduction, a process that plays a determinant role in the definition of EP phonological profile, provide evidence for the word prominent foot.

2.4. Glide Formation—VV > GV

Glide formation creating rising diphthongs is a regular process in EP. Rising diphthongs are in general non-lexical, and alternation between VV and GV is optional, as illustrated in (23) (d’Andrade and Viana 1994; Mateus and d’Andrade 2000), with varying probabilities of occurrence across dialects (Paulino et al. 2022).24 This is a pure phonological, intonational phrase (IP) span process (Frota 2000; Vigário 2003; Paulino et al. 2022).

(23)	criatura	[i]/[j]	‘creature’
	canoagem	[u]/[w]	‘canoeing’

In one particular context, however, gliding is obligatory. Specifically, gliding is mandatory in posttonic position, as exemplified in (24) (Mateus 1975/1982; Vigário 2003, 2016).

(24)	mobília	mobíl[j]a / *mobíl[i]a	‘furniture’
	nódoa	nód[w]a / *nód[u]a	‘stain’

Like with VR, we may straightforwardly explain the data if we assume that gliding is a mandatory process within the foot domain, affecting a vowel in the weak branch of the (ω-head) foot, and is optional elsewhere, within the IP-domain.

Note that gliding is also triggered by enclitics (25).

(25)	bebe-o	beb[j]o / *beb[i]o		‘drink-it’
	pede-a	ped[j]a / *ped[i]a		‘ask-itF’

This is expected, since there is sound evidence that enclitic pronouns are postlexically incorporated into the host PW (Vigário 2003), and thus they undergo the postlexical processes that operate in that prosodic position. Since this gliding process is postlexical, enclitics may create the conditions for mandatory gliding within the foot domain (cf. (beb[j]o)_Ʃ, (ped[j]a)_Ʃ).

Notice that the result of gliding is a trochaic foot. In fact, it is plausible to think that the process becomes mandatory in this particular position in order to avoid a marked foot type (a dactyl) and a lexically marked, exceptional antepenultimate stress pattern (see also Wetzels 2007, for PB). In fact, thanks to gliding, EP avoids a dactyl and displays a trochee. Without reference to the foot, given that the process is optional in other positions, both the motivation and location for its mandatory application in posttonic position remains a puzzle.

We conclude that gliding creating rising diphthongs is a pure phonological process that is obligatory within the same prosodic domain as VR, i.e., the (word prominent) foot.

2.5. Three Syllables Stress Window

Notwithstanding the distinct stress patterns reported in Section 2.1, a fundamental observation prevails in the lexicon concerning stress in EP: the location of regular and exceptional stress is invariably confined within a three-syllable span from the rightmost boundary of the ω. This limitation ensures that primary stress is always located on either the final, penultimate, or antepenultimate syllable of each ω (26). Importantly, three syllables is the maximum documented cross-linguistically for stress windows in accentual and stress systems (Kager 2012).

(26)  Possible locations of primary stress
a. Final syllable:	amará ‘loveFut3Sg’, colosal ‘colossal’, jacaré ‘alligator’
b. Penultimate syllable:	comemos ‘eatPres.3Sg’, amigo ‘friend’, fácil ‘easy’
c. Antepenultimate syllable:	amávamos ‘lovePast.Imp.1Pl’, câmara ‘camera’, Júpiter ‘Jupiter’25

To account for the typological restriction on the maximal size of stress windows, Kager (2012) has recently proposed to resort to an Internally Layered Ternary (ILT) foot, a maximally trisyllabic foot that consists of a binary foot with an adjoined syllable. If feet are maximally trisyllabic, and in window’s systems the prosodic words must be aligned with a foot, at its left or right edge, it is clear why the stressed syllable cannot be located further away than three syllables from a word edge.26 Kager (2012) shows that alternative analyses that do not rely on internally layered ternary feet to account for the maximal size of the stress window cannot account for all of the typological facts, or they make incorrect predictions. Furthermore, since there is ample independent (stress, segmental, and tonal) evidence for rehabilitating ternary feet in metrical representations (see, among others, Bennett 2012; Martínez-Paricio 2013a; Martínez-Paricio and Kager 2015), we propose that the stress window in EP too is limited by the foot domain. By the same token, the EP stress window constitutes powerful evidence for the foot, since every content word contains a foot head plus, at most, two dependent syllables.

Importantly, morphophonological approaches to stress are compatible with this view, since the morphological positions relevant for stress location are always close to word final position, at most three syllables away from the right edge.

Before we conclude, it is important to note that in EP—as in several other Romance languages and varieties, like Spanish, Roca 1999, a.o.)—the three-syllable window constraint has an apparent exception in sequences of verbs and enclitics (e.g., dávamo-no-la ‘give_Past.Imp.1Pl-us-it_F’ and encontrávamo-nos ‘meet_PastImp.1Pl-us’). This can be straightforwardly accounted for by assuming the basic distinction between lexical and postlexical phonology, for which there is ample independent evidence in EP (Vigário 2003, 2022). Post-verbal clitics are combined with the host verb post-lexically (as shown, for instance, by the fact that clitic pronouns can appear before or after the verb, depending on syntactic factors); since word stress is a lexical phenomenon (e.g., morphological and lexical information is relevant for stress location), constraints on stress location are not active when enclitics are incorporated into the host ω. This analysis explains why host-enclitic combinations undergo processes that are postlexical and apply within or at the edge of ω, but do not contribute to lexical stress location (Vigário 1999, 2003, 2022).27

2.6. Poetic Rhyme

Like in many other languages (e.g., van der Hulst 1984; Nespor and Vogel 1986/2007), in EP poetic rhyme typically involves the repetition of the last portion of words, from the stressed vowel to the end of the word, as shown in the extract of the poem “A Mensagem”, by Fernando Pessoa, given in (27) (rhyming stretch underlined, stressed syllable in boldface, rhyming words identified by superscript numbers).

(27)	Outros haverão de ter1
	O que houvermos de perder1.
	Outros poderão achar2
	O que, no nosso encontrar2,
	Foi achado, ou não achado3,
	Segundo o destino dado3.
	Mas o que a eles não toca4
	É a Magia que evoca4
	O Longe e faz dele história5.
	E por isso a sua glória5
	É justa auréola dada6
	Por uma luz emprestada6.

What is considered a perfect poetic rhyme in EP may be captured in a simple way by assuming that it consists of the repetition of a word final foot minus the onset of its initial syllable, like in other languages, such as Dutch (van der Hulst 1984). Note that identifying or forming pairs of words that rhyme are in general not difficult tasks for EP speakers, showing that the unit for poetic rhyme is not merely a common literary construct.

We take this as strong evidence that prosodic words contain a prominent foot in EP. Importantly, the pattern described is obtained with any type of foot.

2.7. Vowel Lowering Processes

Assuming that syllable weight in non-verbs is relevant in BP phonology, Wetzels (1992, 2007) proposes that BP mid stressed vowels in non-verbs are lowered when they are the heads of particular foot types: dactyl or spondaic feet (see also Nevins and Costa 2019). Because there are similarities in this respect between BP and EP (Bisol and Veloso 2016; Vigário 2016), and since foot (type) is considered relevant in Wetzels’ analysis, we investigate here the extent to which there is compelling evidence that the foot (namely, foot type) is indeed involved in this type of lowering in EP as well. The alternations in (28), common to both BP and EP, illustrate the patterns observed:28

(28)	Lowering of stressed mid vowels
	a. So called Dactylic Lowering
		esquel[ɛ́]tico ‘skeletal’	(cf. esquel[é]to ‘skeleton’)
	b. So called Spondaic Lowering
		d[ɔ́]cil ‘gentle’	(cf. d[ó]ce ‘sweet’)

Although EP shares examples of this sort of alternation with BP, we think that this is in fact not a general process dependent on foot type in this variety of Portuguese. First, we should note that lowering is by no means a very general process in this variety, as there are many exceptions to it (29) (a fact also observed for BP in Wetzels 2007, in any case).29

(29)	f[ó]lego ‘breath’	est[ó]mago ‘stomach’	ser[ó]dio ‘old’
	s[ó]frego ‘lickerish’	b[é]bedo ‘drunk’	l[é]vedo ‘fermented’
	pul[ó]ver ‘pullover’	Almod[ó]var ‘Id.’	Est[é]vão ‘Id.’

Second, Wetzels’ analysis of lowering presupposes a weight-sensitive system, since words like dócil in (28b) above can only be considered to display a spondaic foot under the view that the last closed syllable is heavy. However, weight seems not to play a role in EP phonology (see Section 2.1). Finally, there is another process of lowering that is specific to EP, where foot type is (also) irrelevant and which we believe may bear some relation to this one, as explained further below.

Hence, we propose a different approach for stressed vowels’ lowering, relying on the following observations: (i) in EP non-verbs, both regular metrical or morphology-sensitive stress assignment results in stress on the syllable containing the last vowel of the root; (ii) lowering never affects vowels with (regular) metrical or morphology-sensitive stress (30); and (iii) when lowering applies, it targets mid vowels with lexically marked stress, which are always root non-final (31). In our view, this type of lowering can be seen as part of a mechanism to enhance stress, since lowering makes the vowel more sonorous, in the contexts where morphological position (prominence by position = root edge) does not contribute to this end.

(30)	No lowering of mid stressed vowels in root final position (regular)
	Thematic	car[é]t)Rt a)St)MW	‘grimace’
		r[é]d)Rt e)St)MW	‘net’
		m[ó]rr)Rt o)St)MW	‘hill’
	Athematic	voc[é])Rt)St)MW	‘you’
		av[ó])Rt)St)MW	‘grandpa’
		corred[ó]r)Rt)St)MW	‘runner’

(31)	Lowering of mid stressed vowels not root final
	Thematic	esquel[ɛ́]tic)Root o)St)MW	‘skeletal’ (cf. esquel[é]to ‘skeleton’)
	Ahematic	d[ɔ́]cil)Root)St)MW	‘gentle’ (cf. d[ó]ce ‘sweet’)

Let us now turn to the other lowering processes that are specific to EP, mentioned above, which are not related to foot types, but, as we will see, depend on the right edge of a (ω-level) stress domain. These lowering processes are in fact much more general, as they are exceptionless, affecting ω-final vowels under particular conditions (Vigário 2003). One of the processes applies to all unstressed non-high vowels in syllables closed by a consonant at the right edge of a stress domain not coinciding with the morphological word (MW) (32)—this includes athematic stems and stressed prefixes. The other process applies to all unstressed mid vowels at the edge of a stress domain not coinciding with an MW—this includes roots that form independent ω within morphological compounds, roots+linking vowel forming an independent ω within morphological compounds, and clipped words, as illustrated in (33), below.

(32) Lowering of unstressed non-high vowel in closed final syllables at the edge of a stress domain not coinciding with the MW
	Athematic stems
	amáv[ɛ]l	‘kind’		líd[ɛ]r	‘leader’
	Vít[ɔ]r	‘id.’		híf[ɛ]n	‘hyphen’
	âmb[a]r	‘amber’		ímp[a]r	‘odd’
	fórc[ɛ]ps	‘forceps		síl[ɛ]x	‘flint’
	Prefixes that form independent stress domains
	sup[ɛ]r-interessante		‘super interesting’
	híp[ɛ]r-giro		‘mega cute’
	int[ɛ]r-geracional		‘intergenerational’

(33) Lowering of unstressed mid vowel at the edge of a stress domain not coinciding with the MW
	Roots that form independent ω within morphological compounds
	aut[ɔ]-correção		‘self correction’
	bi[ɔ]-degradável		‘biodegradable’
	tel[ɛ]-chamada		‘phone call’
	Roots+linking vowel forming an independent w within morphological compounds
	israel[ɔ]-palestiniano		‘Israeli-Palestinian’
	sóci[ɔ]-cultural		‘sociocultural’
	polític[ɔ]-económico		‘politic-economical’
	Clipped words
	eur[ɔ]	‘euro’
	fot[ɔ]	‘photo’
	exp[ɔ]	‘exhibit’

Note that the unstressed central vowel (/a/) surfaces as low in closed final syllables, but not in absolute final position within a stress domain (cf. ímp[a]r ‘odd’ vs. extr[ɐ]-fino ‘extra thin’). Whether this indicates that there are two distinct lowering processes, or a single one, with the central low vowel patterning slightly differently, is a matter for future research. We may assume, in any case, that for mid vowels there is a single general process of lowering.

In order to define the context for lowering, it seems crucial that the vowel is final within ω but not within the morphological word: the unstressed vowel is lowered when it is the last vowel of an athematic stem, which also corresponds to a stress domain (recall from Section 2.1 that in this subcategory of non-verbs, stress is assigned to the syllable containing the last vowel of the stem), a root forming an independent stress domain within a morphological compound, a composite formed by a stress bearing root + a linking vowel, or a stress bearing prefix. By contrast, the process never applies in the complementary environment, that is, when the vowel is in the final position of both the ω and the morphological word; the latter condition is obtained only with thematic non-verbs, where the stress domain coincides with the morphological word, whether or not showing a final consonant (i.e., plural marker /S/) (casa, casas ‘house_(Pl)’ ponte, pontes ‘bridge_(Pl)’, trono, tronos ‘throne_(Pl)’). An elegant way of accounting for the exact context for the application of lowering is to admit that it operates when a stress domain (ω) coincides with a morphological domain below the morphological word level (e.g., in the sense of Bermúdez-Otero 2017).30

Importantly, the lowering of unstressed final vowels provides evidence for the prosodification of all of the above listed morphological domains as ω and/or as footed domains, prior to reaching MW level in EP.

These sets of data indicate that lowering processes are part of an edge-marking phenomenon in non-thematic non-verbs: stressless non-high or mid vowels surface as low to signal the right edge of an independent stress domain (a ω and/or a Ʃ) that does not coincide with the edge of a morphological word, when stress fails to mark that edge, irrespective of syllable weight or foot structure considerations. In other words, in the absence of final stress in the relevant categories, sonority marks the edge of the unit that forms an independent stress domain but does not correspond to a morphological word.

Interestingly enough, this seems to be an innovation specific to EP, since it is not found in closely related Romance languages and varieties (such as BP, Galician, and Spanish). We would like to entertain the hypothesis that all processes of lowering (lowering in stressed position, and the lowering of unstressed vowels in closed and open final syllables) may reflect a change in EP, possibly related to a change from a quantity-sensitive metrical system to a partially morphologically-based stress system. The lowering process that targets vowels in stressed position currently has many exceptions; it affects mid vowels with lexically marked stress, i.e., not conforming to the general rules of stress assignment, both in thematic and non-thematic non-verbs; the new pattern involves a shift in the target of lowering, from stressed position to unstressed final position; and in addition, the new process affects only ω containing non-thematic morphological material, contributing to enhancing the distinction between thematic and non-thematic subclasses of non-verbs, which we have seen in Section 2.1 is also very important for stress assignment.

The innovative processes of lowering apply to non-thematic (prosodic) words that are exceptional, stress-wise, in that they fail to show domain final stress. There seems to exist a general requirement in the language that a ω containing non-thematic morphological material should have a prominent element at the right edge. In the regular cases, morphologically-based stress assignment ensures that stress falls on the syllable containing the last vowel of the stem, hence, satisfying this constraint. In non-thematic words with non-final stress, the constraint is satisfied by lowering the vowel (i.e., increasing sonority).

This view allows for an integrated analysis of lowering in stressed position and in ω final unstressed position. It also avoids the problem of having to postulate that EP is quantity-sensitive (since there is evidence suggesting otherwise—Section 2.1). Finally, this approach straightforwardly explains why an inflectional final /S/ does not trigger lowering, as in fontes [ˈfõtʃ]/*[ˈfõtɛʃ] ‘sources’, a puzzle under Vigário’s (2003) approach.

In the interest of readability, we will leave the details of this proposal to future work. For the present purpose, we can conclude that there is no sufficient evidence to assume that so-called dactylic and spondaic lowering in stressed position depend on particular foot types in current EP. By contrast, the prosodic processes of lowering of unstressed vowels that operate at the right margin of ω, and that are specific to this variety of Portuguese, further demonstrate the importance in the language of the distinction between thematic and athematic subclasses of non-verbs. As we have seen in Section 2.1, the same two subclasses of non-verbs are subject to two distinct stress assignment mechanisms: one prosodically-based (crucially dependent on foot formation at the word level) and the other morphology-sensitive (referring to stem final position).

2.8. Other Areas in Phonology Potentially Relating to the Foot

2.8.1. Hypocoristic Formation and Word Clipping

In many languages, the foot has been shown to play a role in word clipping (or truncation) and hypocoristic formation, e.g., the output often corresponds to a single foot, either present in the input or formed of material from the input copied to a template imposing a particular foot format, and size constraints at the foot level explain the limited size of shortened forms (e.g., McCarthy and Prince 1986/1996; Poser 1990; Mester 1990; Cabré and Kenstowicz 1995; Thornton 1996; Colina 1996; Piñeros 2000; Bat-El 2005; Lappe 2007; Alber 2010; Alber and Arndt-Lappe 2012; Krämer 2018; Martínez-Paricio and Torres-Tamarit 2019, and many others). In this subsection, we briefly assess the potential role of the foot in EP hypocoristic formation and word clipping in EP.

We will start by looking at the role of the foot in hypocoristic formation. To the best of our knowledge, the topic has not been subject to systematic investigation in EP before. Our analysis therefore builds on and expands the observations in Pereira (2013), who describes some general tendencies of hypocoristic formation in Portuguese, signaling a few noteworthy differences between EP and PB.

First of all, we need to acknowledge that hypocoristic formation in EP seems much less productive than in other Romance languages, and many of the existing forms seem to have been in the language for a long time or have been imported from other languages.

Typical hypocoristic formation in EP includes the types below (A–D), illustrated in (34)–(38), and displays the following general properties: syllables in the output are simple or simplified to a CV format, i.e., input codas and the second C of a branching onset do not surface in the output, but all nuclei that are possible in EP are preserved (e.g., containing a nasal vowel or a falling diphthong); foot format at the input seems to matter in types (A–B); in all types, there is a maximal size to the output form—two syllables; and if hypocoristic names are formed of two ω (as in Cb, with reduplication, and in D), the whole noun corresponds to a Prosodic Word Group (PWG) (Vigário 2010), also maximally composed of two ω.31^,32

• Thematic non-verbs (regular stress pattern): output may correspond to the input trochee, minus coda consonants, if any (ˈCVCV).

(34)	Helena > Lena	Filomena > Mena
	Palmira > Mira	Alcino > Cino
	Albertina > Tina	(Manuela>) Manela > Nela
	Francisco > Chico	Alexandra > Xana
	Fernando > Nando	Andreia > Neia
	Alberta > Berta

B.

Athematic non-verbs (regular stress pattern): output may correspond to what may be analysed as the input (degenerate) foot, minus the coda consonant, if any.33

(35)	José > Zé
	Isabel > Bé
	Joaquim > Quim
	Conceição > São

C.

Various stress patterns in the input and an output template:

(a)

Output consists of the first two syllables of the input noun, minus the second C of a branching onset and/or a coda consonant, if any (regular and irregular stress patterns in the input)—output template: CVˈCV (stress final).34

(36)	Gabriela > Gabi
	Patrícia > Pati
	Fátima > Fati
	Fernando > Fanan
	Filomena > Filó

(b)

Output consists of the input stressed syllable (minus coda consonant, if any), reduplicated—output template: a branching, rightheaded PWG with the format ((CV)_ω ˈ(CV)_ω).35

(37)	Pedro > ((Pê)ω (Pê)ω)PWG
	Ricardo > ((Cá)ω (Cá)ω)PWG
	Jorge > ((Jó)ω (Jó)ω)PWG
	Leonor > ((Nô)ω (Nô)ω)PWG
	José > ((Zé)ω (Zé)ω)PWG

D.

Proper names composed of two w in the input: output corresponds to the prominent syllable (minus coda consonant, if any), or the (prominent) foot of each ω in the input, filling in a right-headed, branching PWG template ((CV)_ω ˈ(CV)_ω)_PWG:36

(38)	Carlos Jorge > Cá Jó
	António José > Tó Zé
	José António > Zé Tó

Type (A) is intuitively the most frequent, which may be a consequence of the fact that thematic non-verbs are also the most frequent sub-class for proper nouns. If we admit that in EP every foot is head initial (Section 2.1), the output in types (A–B) consists in preserving the input foot (a trochee and a degenerate foot, respectively). The fact that in all types of patterns, ω are maximally disyllabic strongly suggests that the output of hypocoristic formation corresponds to a foot.

Note that, in some of these cases, the output foot contains a single open syllable, and syllables tend to be simple/simplified (e.g., CV), even if the resulting word forms a degenerate foot. This type of subminimal hypocoristic name is also found in some other Romance languages, including BP and Italian, and is not necessarily seen to invalidate the role of the foot in the formation of hypocoristics—see, for example, Alber (2010) for an account of this type of violation of Foot Binarity in Italian, where the foot also appears to play a crucial role in hypocoristic formation.

Looking at the productivity of the process in other closely related Romance languages, and even in other varieties of Portuguese, EP stands out for its paucity of hypocoristic nouns, at least in adult vocabulary. Despite the existence of hypocoristic forms in the EP lexicon, it is unclear to us at this point whether this is a productive phenomenon in current adult language. For that reason, in our view, hypocoristic names do not necessarily constitute clear-cut evidence for the foot domain in present-day EP, unlike what seems to be the case, for example, in other Romance languages, such as Spanish (Piñeros 2000; Martínez-Paricio and Torres-Tamarit 2019), Brazilian Portuguese (Thami da Silva 2013), Italian (Alber 2010; Krämer 2018), Catalan (Cabré and Kenstowicz 1995), or Sardinian (Cabré et al. 2021).

Let us now consider the potential role of the foot in word truncation. In contrast with hypocoristic formation, word truncation is highly productive in EP. However, unlike what happens in many other (Romance) languages, the foot cannot be considered relevant for clipping in this variety of Portuguese. In fact, for a word to undergo truncation, what is crucial is that the target word is composed of two or more ω. In addition, as the data in (39a) illustrates, the deleted material always corresponds to one or more ω in the input and the output always matches one or more of the ω in the input (Vigário 2003, 2022); by contrast, clippings like those in (39b), attested in other Romance languages (e.g., Thornton 1996; Krämer 2018; Vilela et al. 2006), are not possible in EP, because the non-truncated words form a single ω.

(39)	a.	supermercado (super)ω (mercado)ω > super	‘supermarket’
		hipermercado (hiper)ω (mercado)ω > híper		‘megastore’
		heterossexual (hétero)ω (sexual)ω > hétero		‘heterosexual’
		telemóvel (tele)ω (móvel)ω > móvel		mobile phone’
		BMW (bê)ω (eme)ω (dâblio)ω > BM (bê)ω (eme)ω		‘id.’
		JSD (jota)ω (esse)ω (dê)ω > (jota)ω     ‘Social Democratic Youth’(abbreviation)
		computador portátil (computador)ω (portátil)ω > (portátil)ω ‘portable computer’
		paragem cardíaca (paragem)ω (cardíaca)ω >(paragem)ω		‘heart arrest’
		pré-primária (pré)ω -(primária)ω > (pré)ω		‘kindergarden’
	b.	sindicato (sindicato)ω > *(sindi)ω, *(cato)ω		‘union’
		frigorífico (frigorífico)ω > *(frigo)ω, *(rífico)ω		‘fridge’
		bicicleta (bicicleta)ω > *(bici)ω, *(cleta)ω		‘bicycle’
		televisão (televisão)ω >*(tele)ω, *(visão)ω		‘automobile’
		polícia (polícia)ω > *(poli)ω, *(cia)ω		‘police’
		meteorologia (meteorologia)ω > *(meteo)ω (rologia)ω		‘weather forecast’

In short, although clipping depends on the prosodic composition of the input word, and there are prosodic constraints on the output as well, in EP this process is ω-based, not foot-based, and therefore does not provide evidence for the foot.

2.8.2. Minimality Constraints

It is well established that in many languages of the world, words have a minimal size, and formats shorter than two syllables or moras are disallowed (McCarthy and Prince 1986/1996; Hayes 1995). This is usually attributed to the fact that words must contain a foot, which in turn must obey a general constraint requiring feet to be disyllabic or bimoraic (FootBin). In languages that show minimality effects, those very effects are interpretable as evidence for the foot. In Portuguese, however, monosyllabic words, and in particular, monosyllabic words with open syllables are allowed (Bisol 2000; Vigário 2003; Vigário et al. 2006), and in EP, for instance, monosyllabic prosodic words with open syllables are in fact quite frequent (ca. 8%, token frequency in corpora of various sizes, cf. Vigário et al. 2006, 2010). As the examples below illustrate, many ω, including highly frequent words, contain a single open syllable (and this is also true for most names of letters and musical notes):

(40)	vi	‘seePast.Per.1PSg’	li ‘readPast.Per.1PSg’	si ‘him/herdat’	ri ‘laughPast.Per.1PSg’
	dê	‘givePres.Subj.1/3PSg’	lê ‘readPres.Ind.3PSg’	sê ‘beImper’	vê ‘see Pres.Ind.3PSg’
	pé	‘foot’	ré ‘defendentFem’	sé ‘cathedral’	fé ‘faith’
	nu	‘nakedMasc’	cu ‘ass’	tu ‘you’	Ju ‘id.’ (hypocoristic)
	pó	‘dust’	dó ‘pain’	nó ‘knot’	mó ‘millstone’
	cá	‘here’	dá ‘givePres.Ind.3PSg’	já ‘already’	má ‘meanF’

Current words like é ‘is’, and the names of letters á, é, i, ó, and u, which correspond to a single vowel, indicate that in Portuguese a ω minimally contains a vowel (i.e., the minimal unit to have a syllable).37

That minimal words do not have to form a disyllabic or a bimoraic foot is also suggested by the way abbreviations and acronyms are oralized. According to Veloso (2017), abbreviation and acronym formation needs to respect a functional constraint which imposes ω to be minimally composed of three segments (for Spanish, see also Torres-Tamarit and Martínez-Paricio 2024), instead of a bimoraic or a disyllabic foot. In addition, as we have seen above, hypocoristic nouns also indicate that subminimal words are well tolerated in EP, since they often take the shape of a single CV, and if the input’s stressed syllable contains a consonant in coda, that consonant is deleted (Isabel > Bé), adding to the number of monosyllabic words with an open syllable.

EP therefore patterns like a rather small number of languages in not requiring words to be minimally disyllabic or bimoraic (see, e.g., Hayes 1995). However, in our view, the existence of so-called subminimal words does not constitute an argument against the foot. Instead, we suggest that in EP, a ω minimally contains one syllable, which is the head of a degenerate foot. All monosyllabic words containing an open syllable are assigned morphologically-based stress. In polysyllabic words, just like in monosyllables, morphologically-based stress often induces violations of FootBin. We suggest that FootBin is violable in languages where stress is not (solely) metrically-based. In a language like EP, violations of FootBin may be due to higher ranked constraints, for instance, those enforcing that the syllable containing the last vowel of an athematic stem bears stress and imposing initial foot prominence.

2.8.3. Secondary Stress

Much work on stress distribution across languages has established a sound relation between rhythmic stress and the foot (e.g., Hayes 1995; Kager 2007; Bennett 2012, and references therein). This is therefore an obvious area to look for evidence for the foot in this variety of Portuguese.

Scholars agree that in EP secondary stress often appears in ω initial position, including proclitics, which adjoin at the level of the ω, displaying a recursive ω (d’Andrade and Viana 1989; Frota and Vigário 2000; Vigário 2003)—see (41).

(41)	a inteligência or a inteligência vs. *a inteligência	‘the intelligence’
	(Frota and Vigário 2000; Castelo 2006; proclitics: adjuncts to the ω, Vigário 2003)

Some work also suggests that alternating secondary stresses, distributed according to rhythmic principles, are also possible (d’Andrade and Viana 1989; Pereira 1999, footnote 71, and references therein). Nevertheless, rhythmic stresses are in general not an observable pattern in current speech. According to d’Andrade and Viana (1999, p. 88), for instance, non-initial secondary (rhythmic) stresses are only likely to emerge in careful speech at the level of the citation form. In addition, long sequences of syllables are perceived with no internal secondary stress, as in the word gramaticalidade ‘grammaticality’ (Frota and Vigário 2000). Note that the perception of secondary stress is possibly also affected by factors like vowel quality, syllable structure, and particular morphemes (Castelo 2006), but seemingly not foot-related factors.

We conclude that, in EP, secondary stress does not provide evidence for the foot in the pretonic region.38 We should nevertheless point out that, in contrast with pretonic position, literature on EP converges in that secondary stresses do not occur after word stress. A total lack of secondary stresses in posttonic position finds straightforward explanation if syllables bearing word stress are the head of a foot domain that also includes posttonic syllables.

2.8.4. Schwa Deletion

Relative prominence at the foot level has been shown to be relevant for strengthening and weakening processes (e.g., de Lacy 2002; Nespor and Vogel 1986/2007; Bennett 2012, and references therein). Schwa deletion is a frequent process in EP, which could potentially be sensitive to metrical relations within feet, and cue the foot domain. The matter is inspected in the following paragraphs.

Schwas in EP always result from vowel reduction, and there is no evidence for underlying schwas (Mateus 1975/1982; Mateus and d’Andrade 2000; Massini-Cagliari et al. 2016; Vigário 2022). Because of constraints on vowel reduction, schwa cannot appear in ω absolute initial position, nor be part of syllables with a branching nucleus (see Section 2.3).

Schwa deletion is a very general process in EP, and has the properties of a pure prosodic process: it is exceptionless, sensitive to speech rate, and conditioned by prosodic structure (e.g., Mateus and d’Andrade 2000; Vigário 2003, 2022). It is mandatory in ω final position (Vigário 2003), except under tone crowding conditions (Vigário et al. 2019) and to avoid stress clash, within a PWG (Vigário 2010); it is optional elsewhere, depending on speech rate (faster speech rates tend to trigger more deletion).

Since, like other prosodic domains, the foot has a prominent element, and since prominence can in principle inhibit vowel deletion, schwa deletion may give insights into the organization of pretonic syllables in the ω, i.e., it can help in determining whether they are organized in secondary feet or not. Examples like those in (42), similar to those in Mateus and d’Andrade (2000), show that schwa deletion may affect all pretonic syllables of the word, and any schwa seems equally likely to delete, to the possible exception of ω initial position.

(42)	desprevenidos	[dʃpɾvˈnidwʃ]	‘unawarepl’
	restremecerás	[ʀʃtɾmsˈɾaʃ]	‘tremble againFut.2Sg’
	descreverá	[dʃkɾvˈɾa]	‘describe Fut.3Sg’

Notice that some of these examples have four pretonic syllables. Given that it is possible to have words like these with no vowel surfacing in pretonic position, we believe this constitutes a strong indication that, at least in this type of word, pretonic syllables are not footed.

3. Discussion and Conclusions

We started this work raising a fundamental question regarding the prosodic hierarchy (PH): is the PH composed of a fixed set of universal domains? By systematically investigating the role of the foot in EP phonology, we expected to contribute to this debate, either (i) by finding positive evidence for the foot, in which case EP would enlarge the immense number of languages where the foot plays a role—hence, strengthening the hypothesis that PH universally includes the foot domain—or (ii) by not finding such evidence, weakening the universality of the PH.

We have looked at a number of areas of EP phonology and morphology where the foot could be relevant. Interestingly, while several kinds of phenomena that typically reflect the foot domain in other languages do not seem to cue the foot in this language, we have seen that a number of facts find a straightforward explanation if one assumes the foot is a phonological category in the language.

In particular, lowering processes, clipping, secondary stress, and schwa deletion do not seem to be metrically-governed, or, at least, alternative non-foot-based accounts are possible and, in fact, simpler. On the one hand, lowering processes of the type proposed for BP, which have been argued to depend on foot type (Wetzels 2007), seem to be best viewed in EP as part of a system for marking the right margin of ω that exhibit exceptional stress in non-thematic non-verbs. On the other hand, we showed that clipping is not based on the foot either, but rather on the ω. Additionally, lack of pretonic feet may in fact explain the choice of the ω rather than the foot as a domain for clipping in this language. The same reason may account for the absence of rhythmic secondary stresses—pretonic syllables not being grouped into feet readily accounts for the absence of rhythmic alternations in pretonic position in current language. Finally, the lack of feet in pretonic position may also explain why schwas in pretonic position may delete with no apparent restrictions.

By contrast, there are a number of facts that, in our view, do reflect the foot domain in the language:

(i)

Stress assignment—we have proposed that stress is best analysed as regularly assigned via two mechanisms, one that refers to morphological position (final in the stem in athematic non-verbs and in past tense verbs), and another that is prosodically-based, involving the construction of syllabic trochaic feet at the right edge of the word, accounting for stress placement in the most frequent subclasses of non-verbs and verbs (thematic non-verbs and present tense verbs, respectively). Independent evidence for the two types of regular stress assignment was also found, coming from -inh-/-zinh- DIM formation—while -zinh- attaches to morphological bases that bear (any kind of) stress (full words with regular stress, morphologically-based or metrically based, or words with lexically marked stress), -inh- attaches to stems not-marked for stress (with no lexical stress and no stem final stress).

(ii)

The domain of regular vowel reduction—non-high unstressed vowels usually undergo vowel reduction in EP, but there are unpredictable lexical exceptions; however, in a number of prosodic contexts, both VR and the absence of VR are fully predictable. We have shown that one of these contexts is the foot domain, since there are no unpredictable exceptions to VR within the prominent foot of a word (i.e., in posttonic position). When VR does not apply within the foot, that is entirely predictable, as VR systematically fails to apply to the last vowel of a stress domain containing non-thematic material. This means that, in EP, VR is sensitive to the position of the vowel within the syllable (no VR in a branching syllabic nucleus), the foot (regular VR within the prominent foot; unstressed vowels ending a stress domain (ω) coinciding with a morphological domain below the morphological word-level systematically escape VR), and the prosodic word (no complete VR of non-high, non-central vowels in ω initial position).

(iii)

Glide formation, creating rising diphthongs, at the right edge of ω—this pure prosodic process, which is optional elsewhere, is mandatory in posttonic position; the fact that it is obligatory only in this position finds a straightforward explanation if we admit that its motivation is to create a trochee, the most frequent foot in EP, and avoid a dactyl, which is a rare and marked ternary foot. Note that a purely non-structural stress-based account of gliding in EP would correctly spot the location of gliding (i.e., post-stressed position), but would be unable to account for its double underlying motivation (i.e., the construction of the unmarked foot in the language, the avoidance of a marked foot in the language).

(iv)

The maximal size of the stress window—in the language, stress always falls in one of the last three syllables; as defended by Kager (2012) and others, such a limited window for stress clearly suggests that the foot, maximally consisting of a binary foot with an adjoined syllable, plays a role in restricting the possible locations of ω stress;

(v)

Poetic rhyme—in EP, like in other languages, perfect poetic rhymes involve the repetition of a subdomain of the word; the definition of this domain, which is intuitive for EP speakers, is straightforward if it is considered the final foot of the word, minus the onset of its first syllable.

Formats of hypocoristic nouns could also be thought to add to the list of processes that reflect the foot domain in EP, since hypocoristics often consist of the (prominent) foot of the input word, and the result also corresponds to a foot, often with the typical formats found in the language. However, we believe that hypocoristic formation is not a productive process in the adult grammar of EP, and existing hypocoristics may have sources that are not informative of present-day adult phonology, reflecting, for instance, (i) child language constraints and lexicalization, a possibility entertained, for instance in Bennett (2012, pp. 319–20); (ii) products formed in previous stages of the language stored in the lexicon; and (iii) loans.

The data reviewed here raise questions on the relation between minimal words and the foot. It is in general thought that subminimal words are not allowed because they violate a constraint assumed to hold for the foot, FootBin (McCarthy and Prince 1986/1996; Prince and Smolensky 1993/2004), requiring a foot to contain exactly two syllables or moras. However, it is well established that binarity is not a necessary property of prosodic constituency, at least above the foot (Nespor and Vogel 1986/2007, and much subsequent work). We would like to suggest that in languages like EP, FootBin is a crucial constraint in the formation of (prosodically-based) metrical feet, since it ensures that prosodically-based stress is assigned via the formation of a head initial binary foot (i.e., a trochee), and there is independent evidence for that binary constituent. By contrast, morphology-sensitive and lexical stresses may originate violations of FootBin, due to the need to satisfy higher ranked constraints, such as interface constraints, involving the alignment of morphological and prosodic material, as well as faithfulness constraints. In our view, subminimal words do not indicate that the foot is absent in EP, since FootBin may be violated (Hayes 1995). Instead, subminimal words in the language strengthen the idea developed in this work that two major subclasses of words (athematic non-verbs and past tense verbs vs. thematic non-verbs and present tense verbs) are subject to different stress assignment mechanisms, with impact on foot format: while in both subclasses, the foot is head initial, in thematic non-verbs and present tense verbs, feet are the result of metrical (trochaic) foot construction, whereas in athematic classes and in past tense verbs, the foot needs to obey constraints aligning a stem final stressed syllable with the left boundary of the foot and the ω right boundary with the right boundary of the foot, even if that may induce violations of FootBin.

In a nutshell, a subset of the phenomena inspected in this study provides evidence for the foot in EP. However, it seems that this domain is only phonologically cued in the main foot of the prosodic word, since we found no evidence for additional rhythmic feet. For instance, vowel reduction and unstressed vowel deletion do not distinguish pretonic syllables other than ω initially; gliding is optional in pretonic position (e.g., Mateus and d’Andrade 2000; Chitoran and Hualde 2007; Vigário 2022; Paulino et al. 2022), but does not seem to distinguish among pretonic material; the productive process of clipping does not target words with a single stress, even if they are long (Vigário 2003; Pereira 2013); among hypocoristic types, there is no output based on a non-final potential foot in the input (see the data in Pereira 2013, and Section 2.8.1); there are no rhythmic stresses in current adult language; and secondary stresses are usually only found at the left margin of ω (d’Andrade and Viana 1989; Frota and Vigário 2000; Vigário 2003, 2022). This raises questions on how pretonic syllables are prosodically parsed. Several hypotheses may be entertained.

Pretonic syllables may be conjectured, for instance, to form a single foot (43a), to be adjoined to the ω head foot, under an unbalanced recursive structure (43b) (van der Hulst 2010), or to be attached directly to the ω (43c). Possibility (43a) predicts that a non-head foot could function as a domain relevant for operations such as clipping and hypocoristic formation, which is not the case; in addition, it violates Foot Binarity, and it would open the way for the emergence of feet larger than three syllables (e.g., as in a word like gramaticalidade, which does not tend to bear ω internal secondary stresses, Frota and Vigário 2000). Furthermore, note that unbounded feet have been independently shown to make incorrect predictions (see Bennett 2012 and Martínez-Paricio 2013a and references therein), and currently, there seems to be only positive evidence for “minimal recursion”, which gives rise to maximally trisyllabic layered feet (Kager 2012; Martínez-Paricio 2013a). Alternatives (43a) and (43b) violate constraints on well formedness of prosodic structure (Recursivity_(Ʃ) and Exhaustivity_(σ), respectively—e.g., Selkirk 1996), and (43b) would also entail the possibility of adjunction of multiple syllables to the ω head foot, potentially creating a maximal foot with more than three syllables, at odds with work showing that feet are maximally (internally layered) ternary. At this point, given the available evidence, and because proclitics, which are highly frequent in the language, also systematically violate Exhaustivity_(σ) in EP (Vigário 2003, 2022), we suggest that the representation in (43c), which is also the one displaying less structure, should be preferred.39

(43)

Given our observations, it seems clear that the foot is weakly cued in EP. Following Hyman (2011), we defend here that EP phonology does not invalidate the universality of the foot domain, but it reflects a low level of phonological activation of this prosodic constituent. In this respect, whereas there is sound evidence that the prosodic word, the prosodic word group, and the intonational phrase are extensively active in EP (Frota 2000, 2014; Vigário 2003, 2010, 2022; Fonseca 2012; Veloso 2007, 2012; Cruz 2013; Barros 2014; Frota et al. 2015; Vigário et al. 2019, a.o.), we have seen that the foot is only marginally activated. Yet, reference to the foot allows (i) a more simplified and unified account of the location of stress (Section 2.1), also relevant for -inh-/-zinh- DIM formation (Section 2.2); (ii) defining the domain for predictable vowel reduction, as well as for mandatory gliding, and poetic rhyme (Section 2.3, Section 2.4 and Section 2.6); and (iii) a straightforward account of the restrictions on the size of the stress-window in EP (Section 2.5).

It would seem fair to conclude that only languages with no apparent trace of the activity of the foot at any level of their phonology or their morpho-phonology could potentially stand as counterarguments of the universality of the foot. We should also acknowledge, like others before us for different prosodic domains, that even those languages would not strongly invalidate the universal status of the foot, since it would still be possible that future work would spot some area that would potentially benefit from a foot-based account. In this respect, it is important to remark that even in languages where primary stress can be described with reference to word edges alone and with no traces of secondary stresses, evidence for foot structure has been found, coming from domains other than stress (see Bennett 2012, pp. 266–67 for examples). This is the case, for instance, in Irish and Uspanteko, two unrelated languages which generally locate stress on one of the last two syllables of the prosodic word. Still, in Irish, reference to a trochaic foot allows for a simpler explanation of some cases of phonologically-conditioned allomorph selection (Bennett 2017), whereas in Uspanteko, reference to a right-aligned foot facilitates a unified and simpler account of the interactions between tone and stress (Bennett and Henderson 2013).

We believe that the account proposed for the various phenomena addressed in this paper is the most economical, in the sense it captures very important generalizations of EP phonology with rather simple phonological apparatus, and it is less stipulative than other accounts, for instance, of word stress assignment, -inh-/-zinh- DIM formation, and vowel lowering at the right edge of designated stress domains, while resorting to phonological and morphological information that is in general required in any framework and approach, including the separation between morphological sub-classes for the treatment of stress assignment and vowel lowering in unstressed position at the right edge of a stress domain. This account also allows for an integrated analysis of facts that were previously unconnected: (i) the same sub-classes reflected by the process of stress assignment (thematic, athematic, and lexically marked words) are crucial for -inh-/-zinh- diminutive formation and for lowering processes; (ii) the types of feet formed in the general vocabulary are also observable in hypocoristic nouns; and (iii) degenerate feet that are found in the general vocabulary (resulting from morphologically conditioned stress assignment) are also allowed in subminimal words and in hypocoristic formation.

Our approach to word stress presupposes a major separation between trochaic (metrical) stress and final, edge-based stress (morphology sensitive). Under this view, stress location in EP has the major function of signaling morphological distinctions: major morphological sub-classes, although not those usually considered in the literature, but rather thematic non-verbs and present tense verbs vs. athematic non-verbs and past tense verbs. In addition, word stress also signals word boundary, since in any case, stress falls on the last foot of ω. Independent evidence for the relevance of the morphological distinctions that we suggest, is the fact that, if exceptions inhibit final stress in ω containing non-thematic material, final vowel lowering ensures that the final vowel of the ω is still prominent, via lowering (i.e., sonority increase). Interestingly enough, unlike what is usually considered in EP, we have shown that the two operating mechanisms for stress location do not distinguish nouns from verbs.

The facts analysed also indicate that, contrary to what is often assumed in the literature for the foot, in EP the foot is not primarily a rhythmic domain.40 In particular, metrical feet only account for morphological subclasses of verbs and non-verbs (thematic and present tense, respectively), which are usually considered to be the most frequent subclasses, whereas in a large portion of word forms, foot construction crucially depends on morphologically conditioned stress, which determines the location of the foot’s head, and right edge alignment between foot and ω. In our view, this creates the grounds for another interesting feature of EP: in addition to the crucial distinction between head and dependent, internal organization within the foot domain seems irrelevant in the language. All foot types in EP are initially stressed, minimally composed of a syllable and maximally containing three syllables, conforming to the three syllables stress window. These properties of the language may contribute to the difficulty of EP speakers in perceiving stress, as shown in Correia et al. (2015), in a sequency recall task (Dupoux et al. 2001): in the absence of vowel quality cues to stress, EP speakers show a stress ‘deafness’ effect, patterning like speakers of a language like French, which has fixed stress, rather than of languages with variable stress, like English or Spanish. And this despite the fact that in production, EP speakers do not show any signs of inconsistency in identifying stressed syllables, as shown, for example, by pitch accent alignment with the stressed syllable (Frota 2014, and references therein).

Importantly, although the foot in EP seems to play little role as a rhythmic unit within the word, it does seem to contribute to ω-level rhythm. A ω in the language contains (i) a prominent syllable, which is the head of the final foot, (ii) often, there is a secondary stress in ω-initial position, and (iii) any additional existing syllables tend to show vowel reduction (and frequently, vowel deletion); however, (a) VR is not complete in ω absolute initial position and, crucially, (b) non-initial vowels that occur in a non-branching syllable nucleus always reduce within the foot domain, unless they occupy a very specific prosodic position, i.e., when they are final in ω containing non-thematic material. Prominence by stress and lowering (i.e., increased sonority), as well as vowel reduction and deletion, in part defined with reference to the foot domain, contributes to what may be classified as word-based rhythm. This characterization of the language distinguishes EP from other well studied languages, including both Romance and Germanic languages.

Other areas where a higher-level phonological profile proves most relevant involve the intonational grammar: (i) only higher-level prosodic domains are tonally marked, and (ii) when tune-text conflicts emerge and there are insufficient syllable/segmental anchors for the realization of complex tonal material, and unlike in the other languages studied in this respect, EP resolves the conflict by inserting a vowel (and consequently, a syllable), instead of truncating the tune (Frota et al. 2015; Vigário et al. 2019). This has been attributed to the high phonological profile of the language: higher-level phonology is more relevant in EP than low-level phonology, which comprises segments and syllables, and, we may now add, feet as well.

We have shown that morphological information is relevant for capturing the necessary generalizations about stress distribution and foot formats. If our analysis is on the right track, we predict that the foot is not among the initial prosodic categories to unfold in language acquisition (Frota et al. 2016, 2024). This prediction seems indeed correct, as perception studies looking at event-related potential (ERP) responses in infants learning EP with a passive odd ball paradigm have shown an initial processing advantage for final stress (Lu et al. 2023), reflecting, we may conjecture, higher-level prosodic prominence.

In several parts of this work, we have signaled some important differences between EP and other Romance languages. Although there are obvious similarities in phonological and morphological grammar between EP and other Romance languages, for instance BP and Spanish, the areas of divergence are noteworthy. Broadly speaking, the foot seems a much weaker domain in EP that in most other Romance languages. Several related matters explored in this study certainly call for more systematic comparative research involving EP and the other languages of the same close family.