https://orcid.org/0000-0001-9154-6161
ABSTRACT
The great majority of lines in Eugenio Montale’s published work are written in recognizable metres, mostly hendecasyllables. There are, however, approximately 300 lines that are slightly longer than the hendecasyllable, yet shorter than other known metres, typically counting 13 metrical syllables. These lines have traditionally been referred to as ‘tridecasyllables’, or ‘hypermetrical hendecasyllables’, simply acknowledging their length. This article, after a review of the literature and of the structural features of the tridecasyllable, and on the basis of a scansion of all such lines, argues that they can be given a definite description by adopting the formalism known as Bracketed Grid Theory. Thus, this article develops an account of these forms, including both 13- and 12-syllable lines, and explains both their structural features and their similarities with the dodecasyllable and, most importantly, the hendecasyllable.
SOMMARIO
L’opera edita di Montale contiene in grande maggioranza versi scritti in metri riconoscibili, soprattutto endecasillabi. Vi si trovano tuttavia circa 300 versi di 13 sillabe metriche, di lunghezza cioè leggermente maggiore rispetto all’endecasillabo, e al tempo stesso più brevi di altri metri tradizionali. Questi versi sono sempre stati tradizionalmente denominati ‘tredecasillabi’, o ‘endecasillabi ipermetri’, riferendosi semplicemente alla loro lunghezza. Questo articolo, dopo una revisione della letteratura sul verso tredecasillabo e delle sue principali caratteristiche strutturali, e, sulla base di una scansione di tutti i versi siffatti, argomenta che è possibile dare una descrizione definita di questo metro adottando il formalismo noto come ‘Bracketed Grid Theory’. L’articolo ne sviluppa dunque un’applicazione che spiega la forma del verso tredecasillabo, la sua relazione con il dodecasillabo, e, soprattutto, con l’endecasillabo.
Introduction
Eugenio Montale’s poetry consists of about 11,000 lines, half of which are written in the most common metre of the Italian tradition, the endecasillabo. The endecasillabo can be understood as a line of verse containing at least 10 syllables, although, more often than not, a metrical convention known as synaloepha may be applied, by which two or more adjacent vowels belonging to different syllables may be counted as one single metrical unit. The 10th metrical syllable must be stressed; most Italian words carry lexical stress on the penultimate syllable, so that most endecasillabo lines comprise 11 metrical syllables, as the name of the verse suggests. The other half of Montale’s poetical works display diverse kinds of verse design: some of them are shorter than the endecasillabo, whereas others are longer and can be typically scanned as compound metres. Among the non-hendecasyllabic metres there are, however, approximately 300 lines that do not fit either category: longer than a canonical endecasillabo, but shorter than any expected compound metre, these lines lack a clear status and denomination. They have been referred to as tredecasillabi and dodecasillabi (13- and 12-syllable lines, respectively) or as endecasillabi ipermetri, or, in some cases, as endecasillabi caudati (with a coda). In each of these interpretations literary critics have focused on syllable count, and on these lines’ similarity with the endecasillabo. While tredecasillabo lines cannot be seen as a regular hendecasyllabic form, they are not free verse in the proper sense either. In fact they instantiate a form, which stands in some relationship (be it formal or contextual) with that of the endecasillabo, as they retain some of its formal characteristics. This intuitive relationship has attracted the attention of several notable scholars, yet agreement on a precise definition has never been reached.
The goal of this article is to provide a clear scansion procedure for these lines, one that can resolve the problem of what they are. The leading assumption here is that the main tasks of poetic metrics, as a branch of the study of literature, are first to assign a definite description to the form of lines of verse and, secondly, to characterize as precisely as possible the specific language use of a poet (or of a poetic school, or tradition) in metred texts. This article aims to perform the first task: to seek the above-mentioned description with regard to these lines of verse. In other words, the article provides a non-arbitrary methodology to discern the metrical form of the tredecasillabo lines, and to understand where exactly their similarity with the endecasillabo lies. The focus of the present proposal is thus not on Montale’s language use in metre, but rather on what kinds of metrical structure surface in his poetry.
By the standard Italian taxonomy, tredecasillabo is the metre of lines whereby the rightmost stress occupies the 12th metrically relevant syllable. The lines in (1) exemplify this verse form and the numbers indicate the stressed syllables:Footnote1
These lines have an overall number of syllables varying between 12 and 19, depending on how many unstressed syllables follow the last stress of the line, and on how many synaloephas are applied between syllables 1 and 12. (1a) displays, for example, 14 phonological syllables, with the rightmost stress of the line corresponding to the 13th; (1b) has 13 syllables, with a last lexical stress on the 10th syllable, but an intonationally relevant stress on the 13th; (1c) comprises 19 syllables, if no synaloepha is applied, and so on. If we instead assume that synaloepha always applies wherever a vowel sequence is found, in all the above examples the last stressed syllable can be counted as the 12th position in the line, which can thus receive a scansion as an instance of tredecasillabo.
We also see that these lines do not exhibit any definite stress pattern, but instead a seemingly unconstrained variety thereof, as shown by the parsing next to examples above. Finally, these lines lack a clear metrical context, since they always appear in polymetrical poems – as far as Montale’s poetry is concernedFootnote2 – that is, poems that alternate different kinds of metre, often without any recognizable schema. The metrical context thus has a bearing on whether the reader decides to apply synaloepha. This (it should be stressed again) is by no means obligatory; it just represents the relation of a line to a supposed metrical pattern. It is a fact, however, that Montale’s metrical forms are traditional for the most part, if not as a whole, and that even the most peculiar lines are in some relation, either formal or contextual, with the metres of the Italian poetic tradition.Footnote3 All the examples in (1) can thus be considered tredecasillabi, but, following the observations above, their metrical status is somewhat unstable.
My argument unfolds as follows. In the second section I discuss the theoretical problems raised by the tredecasillabo metre, and more generally by hypermetrical forms of endecasillabo lines, by reviewing some of the most important previous contributions to the issue at hand. The third section provides details on the corpus of tredecasillabi at the basis of the present study, and brings out the most relevant structural facts of this verse with reference to sample lines. The subsequent section develops an explicit scansion procedure for the tredecasillabo based on the Bracketed Grid Theory of poetic metre; I then outline the fundamental similarities with the 11-syllable line, and show that the majority of lines in the corpus should be seen as a closely related variant of the common source endecasillabo. Finally I draw the conclusions of my argument and highlight its relevance for further research, and discuss possible influences on this metre from other authors’ loose metrical versification practice.
Previous Approaches to the Tredecasillabo Metre
Several critics have discussed the issue of how the tredecasillabo should be treated. These approaches essentially debate whether this metre (i) has an independent status, (ii) should be analysed as a compound meter, or (iii) should be conceptualized as a relaxation of the rules underlying the Italian endecasillabo. Note that if the latter is true, the tredecasillabo should be treated as a dependent form. Also, modern Italian metrics is particularly concerned with the dichotomy between isosyllabism and anisosyllabism. That is, it often refers to a divide between metrical systems requiring all lines in a given metre to have a constant number of syllables as opposed to others which do not (or not strictly). Costanzo Di Girolamo defines an anisosyllabic versification system as a system whereby all lines in a set (e.g. a given metre, or a poem) belong to the set even though they are not strictly bound (they are bound together by a ‘non costante identità metrica’).Footnote4 In other words, in these systems the identity between two metrical lines is not constant, while the Italian poetic tradition since the establishment of the canon with the works of Dante and Petrarca is isosyllabic for the most part. I provide below a cursory review of the most important contributions to the understanding of the tredecasillabo metre.
Giuseppe Sansone focuses on examples of tredecasillabo and of what he calls dodecasillabo and endecasillabo con sillabe soprannumerarie.Footnote5 He analyses the lines by counting syllables, making lexical stresses explicit, and assuming that synaloepha normally applies throughout each line. He sees these longer lines as constituted by two hemistichs, of variable length. An overview of the examples he lists shows that he places these mid-line breaks at the edges of phonological phrases. Sansone concludes his study by citing an observation drawn from Antonio Da Tempo’s Summa Artis Rithimici Vulgaris Dictaminis, in which the medieval scholar states that hendecasyllables may sound right to the hearer, even when they to do not respect the right number of syllables (‘si [. . .] non contingeret ad rectum numero syllabarum, numquam bene sonaret auribus audientibus’).Footnote6 Sansone’s arguments and examples, as well as this final citation, all tend to support the position that both tredecasillabo and dodecasillabo are a variant of the regular hendecasyllabic form, whereby syllables are not counted correctly in favour of an equivalence that, however, sounds right.
A similar line of reasoning is adopted by Beltrami,Footnote7 who considers modern examples of anisosyllabic versification. In a paragraph dedicated to Montale, Beltrami does not refer to the tredecasillabo as such, but sees the verse form as one possible alteration of the strict syllable-counting rules of canonical endecasillabo. While not making direct reference to the specific form of the tredecasillabo line, he does address Montale’s frequent ‘clear alterations of the endecasillabo’, and also notes that such lines always appear in polymetric contexts, alternating with the canonical hendecasyllable, thus contributing to a general hendecasyllabic rhythm. Beltrami discusses the first strophe of the poem, Notizie dall’Amiata, and highlights those parts of the dodecasillabo and tredecasillabo lines that, if omitted, would render the line a perfectly canonical endecasillabo (e.g. ‘E tu seguissi le fragili architetture’). In general, Beltrami does not see these longer lines as a compound metre, but instead focuses on the issue of how they are basically hendecasyllables with added parts, which he mostly identifies with words rather than with syllables or feet. He finally suggests that these lines are a revival of the anisosyllabic versification that was widespread in Italy before the establishment of the canon, that is, before the thirteenth century.
Massimo Antonello provides a list of the metres used in Montale’s first three poetry collections, thereby noting all the tredecasillabi.Footnote8 While not directly discussing the structural issues at stake here, Antonello marks these lines as tredecasillabo. He thus appears to assign an independent status to such lines, by focusing on the number of syllables in each line.
Antonio Pinchera, although not specifically discussing Montale or the tredecasillabo,Footnote9 sees these lines, together with the dodecasillabo, as a case of anisosyllabic forms of endecasillabo, thus not diverging from Beltrami on that matter. He does this, for instance, for a number of 12-syllable lines appearing in D’Annunzio; and he also, interestingly, suggests that the dodecasillabo lines are clear-cut double 6-syllable lines, regardless of the specific linguistic form of the lines.
The two scholars who worked most extensively on the role – both historical and structural – of the tredecasillabo in contemporary Italian poetry are Paolo Giovannetti and Pier Vincenzo Mengaldo. Giovannetti provides several analyses of the role of tredecasillabo as a kind of free versification in Italian modern literature, most notably in Corrado Govoni. In particular, Giovannetti looks into several tredecasillabo lines with an identical verse design to those in Montale’s corpus. Giovannetti, framing his analyses within the discussion of the nature of Italian free versification, shows how Govoni’s verse makes use of tredecasillabo both in polymetrical contexts and in verse entirely composed in that metre, thus confirming its key, autonomous role in the departure from canonical versification.Footnote10 Giovannetti then furthers this line of inquiry by providing a definition of the verse in point as an ‘autentico endecasillabo alternativo’.Footnote11 In a perspective arguably based on pragmatics and Rezeptionästhetik, he emphasizes Montale’s modernist attitude, actively adopting measures such as the dodecasillabo and the tredecasillabo to substitute for the canonical endecasillabo. On the basis of comparisons with other poets, both before and after Montale, Giovannetti thus sees these two metres as structurally independent but clearly connected to the endecasillabo.
Mengaldo expressed his positions on several related issues in a number of essays.Footnote12 Analysing some lines of D’Annunzio, a predecessor and an influence to both Govoni and Montale, Mengaldo notes that in many cases, lines of 13 syllables would actually constitute canonical endecasillabi, with the addition of a disyllabic metrical foot, and calls them ‘endecasillabi caudati’ (with coda).Footnote13 In most other cases, he treats tredecasillabi as manifestations of influences from the French alexandrin libéré, and as ‘misure variamente esametriche’ (variable hexametric measures).Footnote14 Most importantly, to explain the possible influences on the form of the tredecasillabo, he sees all the instances of verse forms that, despite being 14 syllables long, cannot be scanned as a straightforward doppio settenario (i.e. the Italian version of the French metre with a compulsory mid-line caesura) as variably longer lines and, most important, as traces of a shift from the traditionally isosyllabic Italian system:
A una versificazione fondata – in buona sostanza – sull’isosillabismo, ma allo stesso tempo sulla varietà e posizione degli ictus, può tendere a subentrare una versificazione […] basata su ictus di numero o posizione fissa, vale a dire su ‘piedi’. Ne vanno distinti morfologicamente due tipi fondamentali: a) versi con numero variabile di piedi e di ictus, i cui piedi hanno però sempre la medesima misura sillabica […]; b) variabili sono sia la lunghezza dei versi, che la consistenza sillabica dei piedi, ma costante è il numero dei piedi, cioè degli ictus, di ogni verso.Footnote15
The last insight touches upon another crucial point in the perspective this article intends to develop. Mengaldo’s stance is close to the idea proposed by another renowned scholar, Mikhail Gasparov. In an appendix to A History of European Versification, Gasparov expresses the view that the Italian poetic tradition develops a syllabic metrical system – in which only the number of syllables matter – with a tendency to a syllabotonic one – in which both syllables and position of stresses matter.Footnote16 From this perspective, Italian versification cannot but differ from the purely syllabic French versification.Footnote17 If Gasparov’s point is taken further, it appears plausible that Italian, at the point of expanding the boundaries of traditional versification rules, does so by allowing more syllables and, at the same time, by keeping the number of feet – thus stresses – constant. Yet, by this combined reading it is unclear which syllables in a line should constitute longer feet, and which would not; or whether these should coincide with lexical units or with other linguistic units, and if this choice would hold for each and every tredecasillabo line.
In sum, current accounts of these verse forms do not yet provide a definite way to distinguish the size of the metrical feet, which syllables could be disregarded in the scansion, or, finally, why these lines can still be seen – if not perceived – as close enough to the form of the endecasillabo. Along the same lines, there should be no reason why the dodecasillabo should present any difference: it should be understood as an endecasillabo with an extra syllable. The perspectives discussed in this section find agreement on the fact that tredecasillabo and dodecasillabo lines stand in some relation to the endecasillabo. This relation may be perceptual, formal, or both. There exists no consensus on whether these lines should be understood as a single unit or as a two-part line, as it is for example in the case of the doppio settenario. The underlying assumption – entailed in the definition of anisosyllabism provided above, and explicit in Mengaldo’s reference to French influences on Italian literature – is that syllable-counting prevails in the Italian metrical tradition as a formal principle.
It is clear that there exist lines, longer than the endecasillabo, that do not express a specific rhythmic pattern. The extent to which such lines are longer than a hendecasyllable depends on the reader’s choice whether or not to apply synaloepha in the line. In these cases, the reader is left with a few unanswered questions. How should these lines be conceptualized, and represented? Is it possible to provide a clear analysis of the tredecasillabo, such that it uniquely yields a definite description for this metre? Would that analysis cover also the dodecasillabo? And, finally, how can we determine the relation of these longer verse forms with respect to the canonical endecasillabo?
Corpus and Analysis
This section provides further details on the corpus under consideration, as well as an in-depth analysis of two sample lines from those quoted in the first section. The analysis strictly follows the questions raised in the previous section, so as to provide the reader with a precise description of the issue at stake. I suggest that, given the facts highlighted here, it is possible to give a definite description for these lines, which I develop in the following section.
The features used to define and extract all potential tredecasillabo lines in this corpus are their being longer than the endecasillabo and their having a stress on the 12th metrical syllable. Yet this allows very variable configurations in both final and medial position in the line. Three examples are given in (2), with lexical stresses marked:
As expected, given the phonology of the Italian language, the 12th metrical syllable may be the penultimate (2a), the last (2b), or the antepenultimate syllable of the line (2c). This variability normally does not affect the denomination of the verse. The tredecasillabo line is thus defined as having a stress on the 12th metrical syllable, with no added constraint on the number or location of stresses. If we dig deeper, however, it is possible to make some observations on the specific linguistic aspects that define a tredecasillabo. Examples (3a) and (3b) provide the text and the phonological phrasing for one such line,Footnote18 and (3c) offers a more detailed analysis of its structure, while (3d) provides the same parsing for the line quoted in (2b):
The line in (3a) can be split into two phonological phrases, indicated by square brackets and φ in (3b). In the latter example, lexical stresses are marked, and underlining indicates the syllables’ nuclei carrying each phrase’s major stress (see Nespor and Vogel 1986). (3c) provides a first metrical parsing, with the supposed metre indicated below the text. Note that the categorization of the line as a tredecasillabo implies the application of synaloepha on the two adjacent syllable nuclei /i/ and /a/ corresponding to position 9 in the example; and this observation, in turn, relies on the assumption that synaloepha is preferred wherever possible unless otherwise required by the metre or by the poet’s idiolect. If the synaloepha on position 9 is accepted, the 14 phonological syllables can be analysed as a line with 12 metrical positions, with 3 lexical stresses in polysyllabic words, in addition to the pronoun ‘tu’, which can arguably receive prosodic stress;Footnote19 all these elements are underlined in (3b). A synaloepha such as the one in the example is a common assumption among literary critics, albeit a never fully explicit one. Yet, there is no rule that prescribes that this line should be interpreted as a legitimate instance of tredecasillabo, because there is no specific metrical context for it in Montale’s poetry.
These considerations correspond to some facts about the form of the tredecasillabo. How can these facts be combined with the insight, as seen in the previous section, that there is some kind of similarity relation between this verse and the endecasillabo? While this similarity might intuitively be self-evident, it is based (i) on an approximate estimation of these lines’ syllabic length and (ii) on their apparent lack of a definite form, including rhythm (e.g. iambic alternation of stresses). The only formal aspect that a first investigation could ascertain is their length, which deviates by just two syllables from the mandatory ten metrical syllables strictly defining an endecasillabo. This amounts to saying that the surface linguistic form of the line is different from that of a canonical endecasillabo. But nothing prevents us from thinking that a metre can encode, within a surface linguistic form, an underlying metrical structure. If this structure can be precisely and consistently recognized in all occurrences of tredecasillabo, it means that the intuitive similarity emerging in the literature reviewed in the previous section may be the correlate of a structural fact. Bearing this in mind, I now lay out in example (4) three possible hypotheses that would support the similarity between the tredecasillabo and the endecasillabo:
The first hypothesis derives from tagging two syllables within each line as extrametrical. At best, as in (4a), these stresses could then become the expected strong positions in the endecasillabo: here, positions 5 and 8 (in brackets in the metre) would need to be considered extrametrical. The second hypothesis, in (4b), marks an entire foot as extrametrical, thus developing a proposal made by Beltrami:Footnote20 here, positions 9 and 10 (in brackets in the metre) would be considered extrametrical. The fundamental iambic rhythm of the endecasillabo, and the correct syllable count, would then emerge. Finally, in (4c) the Underlying Structure Hypothesis formalizes the view expressed by Trissino that the canonical endecasillabo is a full iambic trimetre.Footnote21 In his terminology, a metron amounts to two metrical feet; thus, the full hendecasyllabic form (‘trimetro giambico pieno’) would realize 12 metrical syllables, two of which would instantiate a metrical foot, completing the third metron, that follows the 10th position. These 12-syllable line would then constitute a hyper-metrical variant of the canonical endecasillabo (thus appearing to corroborate the points made by Sansone and Mengaldo, as seen in the previous section). This variant would be subject only to the condition of having at least one stressed position for each metron, namely that located in each 2nd foot (these positions are underlined in the metre in (4c)).
All three hypotheses nonetheless share the same simple problem: how does one decide in a non-arbitrary way what is ‘added’, and in what cases, to the structure of the endecasillabo? In other words, it is left to the interpreter to decide which part of the line matters to its structure and which does not. In addition, the variability of stress patterns in the tredecasillabo, as seen at the beginning of this article, by no means ensures that all lines can allow a neat scansion to highlight an underlying hendecasillabic structure, as in (4). If there is something meaningful in the idea that the structure of these lines is dependent on the endecasillabo, or at least similar to it, however, there must be a more consistent way to uncover this structure. This idea is loosely related to the computational theory of mind, which grounds much current work in linguistics and psychology. As the linguist Ray Jackendoff puts it, for example, this theory holds that whenever a psychological constancy exists, there must be a representation encoding that constancy.Footnote22
I suggest that the lines at stake in this article call for a unitary formal account, which I develop in the following section. Before that, I summarize the metrical facts on which that an account must be based:
No formal account of this metre can ignore these points. It should also serve as a discovery procedure, in the sense that a scansion – as a definite description in the terms highlighted throughout this article – must define exactly what is and is not included in that description. If the Underlying Structure Hypothesis holds, the tredecasillabo might well not be a metre in its own right – at least at some level of analysis – but some kind of variation on the endecasillabo; should this prove true, the dodecasillabo too might represent just another variant of the same hendecasyllabic measure. If an explicit, formal account can shed light on these points, it should allow us to discern both the nature of these lines and the extent to which they are or are not similar to the endecasillabo; and whether all the lines included in the present corpus are an instantiation of the same form.
The Tredecasillabo (and Dodecasillabo) as a Kind of Loose Metre
Bracketed Grid Theory
The noted philologist and theorist of literature Costanzo Di Girolamo has written that each and every philosophy of metre has consequences for the treatment of texts.Footnote23 This is particularly true for the perspective developed in this section. Here I develop an account of the metrical structure of the tredecasillabo as a ‘loose metre’, namely a loose kind of the iambic metre endecasillabo, in the terms of the Bracketed Grid Theory of poetic metre proposed by Nigel Fabb and Morris Halle in 2008.Footnote24 From this perspective, the tredecasillabo as well as the dodecasillabo can be scanned by the specific computation developed by Fabb and Halle to analyse metres that do not strictly control all the syllables within a line. In the ‘metrical philosophy’ of Fabb and Halle, poetic metre is a syllable-counting system that relates the syllables to the line as a whole unit. This has a bearing on how not only canonical metres but also less traditional forms of versification are treated, such as the ones at stake here. I shall briefly introduce this theory, and the concept of ‘loose metre’, before moving on to develop the algorithm for the loose endecasillabo.
The term ‘Bracketed Grid Theory’ derives from its particular formalism, which essentially is an automatic procedure that counts, or groups by twos or threes, the units constituting the verse-line by inserting brackets among them. According to the authors, this procedure takes as input the metrical syllables (or, in other metres, other phonological units), and represents them as asterisks; if the line is metrical, it gives as an output a well-formed grid. Each metre is thus a set of ordered rules that outputs a bracketed grid built from G(rid)L(ine) 0 up by following a strictly ordered procedure. The brackets create the groups (traditionally referred to as ‘metrical feet’) by iteratively scanning the syllables according to the algorithm – i.e. according to the rules and the parameters specified by at the outset of each scansion.
Since Bracketed Grid Theory appeared in its definitive form, despite its aims to offer a potentially universal coverage of metrical forms in the poetries of the world, it has attracted only a few applications.Footnote25 Its potential for the discovery and analysis of the metres of the world remain remains largely untested, as do its fundamental theoretical claims. In this section, I show how the tredecasillabo lines discussed in the previous sections can receive a correct scansion – i.e. a definite description – if analysed as a loose metre. Additionally, I show how most dodecasillabo lines can also be subsumed under the same category, and scanned accordingly, thus providing formal justification of the claim made by Sansone and Di Girolamo (for example) that these verse forms are related to one another and to the canonical endecasillabo.
Loose Metres in Bracketed Grid Theory
One of the most interesting aspects of Fabb and Halle’s theory is its approach to verse forms of variable length. Fabb and Halle account for such forms by a variation in the way metrical rules build the structure of the verse-lines, the bracketed grid. While grids for strict metres control all available metrical syllables, a loose metre excludes some syllables from the computation, under very specific conditions and with no particular reference to the phonology of the line. As I will show below, the computation still controls the length of the lines by grouping syllables, but the grouping itself is influenced by the presence of metrically relevant stresses, which the theory defines as ‘stress maxima’ (henceforth, SM), each of which forces the grouping to restart anew. In other words, in these forms the number of groups in the metrical structure does not directly correspond to the actual linguistic form of the line, but, as I show below in more detail, the grouping procedure automatically excludes some syllables from this structure. Following an intuition by Robert Frost, Fabb and Halle name these poetic forms ‘loose metres’, and analyse them with a dedicated formalism.Footnote26 The recognition of a loose metrical form, in the terms discussed in the second section, in the Italian poetic tradition – strictly isosyllabic for the most part – could pave the way to a deeper understanding of how Italian poetic forms gradually shifted towards free versification, and how intermediate steps in the development of a poetic culture contributed to changes in the poetic metrical system. This article proposes that the structure of the tredecasillabi in Montale’s corpus can be successfully analysed as a loose variant of the main strict metre of the Italian tradition, the endecasillabo.
For the reader’s convenience, I first describe the basic workings of the Bracketed Grid Theory, then move on to a more detailed exposition of how the scansion procedure operates for loose metrical forms, and in the case of Montale’s tredecasillabi in particular. Fabb and Halle define a metre as a set of rules and conditions, and this holds for both strict and loose metres. These rules run over the input – each line of verse – and, if the line is metrical, automatically produce the output – the bracketed grid. Each kind of metre involves the setting of five parameters for each gridline constructed by the algorithm to specify how the rules operate:Footnote27
The above parameters correspond to rules that iteratively run over the input – each verse-line – and automatically produce the output – namely, a set of successive grid lines (GL). The parameters in (6i) and (6ii) specify where parenthesis insertion begins. (6iii) refers to the kind of parenthesis inserted – Left, grouping the asterisks at their right, or Right, grouping asterisks at their left. Then, depending on the setting of the parameter (6iv), the algorithm groups together either two or three asterisks by inserting parentheses. Parameter (6v) then attributes the role of head to one syllable within each group. Each head gets projected to the following GL, until there is only one asterisk left. This is termed the Head of the Verse (HoV). The final result of this algorithm is a bracketed grid; a varying number of specific conditions may then check the grid and constrain its form.
A crucial aspect of Bracketed Grid Theory is the underdetermination of the input: the computation always differentiates between only two entities among the wider range of phonological possibilities available in a language to mark prominence (e.g. stressed/unstressed syllable). Along the same lines, the concept of SM also varies from each poetic tradition to another, or even individual poetic practice. This aspect is particularly relevant in the case of loose metres, because Fabb and Halle’s theory specifically captures such forms by adding to the above-mentioned set of parameters the following rule:
The rule in (7) applies after all relevant syllables have been projected onto GL0, and before the iterative rules start grouping these projections. This marks an important difference between loose and strict metres: in the latter the counting applies without taking into account the actual phonological rhythm of the line, whereas in a loose metre the presence of designated metrically relevant stresses directly influences the structure of the grid already at GL0. This happens because such stresses force the insertion of non-iterative parentheses into the grid before the iterative rules apply. Coupled with rule (7), a final, loose-metre-specific rule applies, as stated in (8):
Before moving on to the specific algorithm for Montale’s tredecasillabo, I use an example drawn from Fabb and Halle (65–93) to detail how the scansion of a loose metre proceeds. Note that SM in the case of the metre of the example (a loose iambic tetrameter) are defined as: ‘the syllable carrying lexical stress in a word if it is preceded and followed by syllables carrying lesser stress’.
Bearing in mind the parameters listed in (6), example (9a) shows the first steps of grid construction. In (9a) we see the projection of the syllables as an asterisk on GL0. Three asterisks correspond to SM by definition, and are marked by a square R bracket (also indicated by the arrows). In (9b), the iterative rule starts to scan the line, in (9a) from R to L, building binary groups. When the iterative rules, however, reach the asterisk signalled by the arrow, a binary group cannot be constructed because the rules encounter a R-parenthesis. The rules thus automatically skip the syllable, which is left ungrouped. (9c) shows how the iterative rules scan through all asterisks at GL0 and construct GL0. In (9d), we see how the heads of each GL0 group are promoted to GL1, and how the iterative rules start grouping again the four available asterisks. This procedure is repeated at each successive GL, until it leaves a single asterisk ungrouped, the HoV. The procedure is fully automatic; it is the grouping by iterative rules – not any phonological characteristics of the syllable, except from not being a SM – that leaves ungrouped some syllables within the line.
An Algorithm for Montale’s Tredecasillabi (and Dodecasillabi)
The tredecasillabo lines discussed here can also be analysed as a kind of loose metre – more specifically, a loose endecasillabo, i.e. a loose variant of the 5-feet structure of the endecasillabo, thus a loose iambic pentameter. In what follows, I propose a full set of rules for the loose endecasillabo, and I show how the same set of rules scans the form of the dodecasillabo as well, thus providing a formal explanation for the idea that both terms should be subsumed under the category endecasillabo ipermetro. The final result will be a single algorithm for Italian loose endecasillabi, covering both forms. This algorithm will exclude some syllables from the metrical structure for no other reason than the metre itself (e.g. no specific phonological feature except that of being a SM); and it will also be capable of distinguishing between these lines and others which, despite having the same number of syllables, actually instantiate different metres. I now follow the steps to generate a bracketed grid from a sample tredecasillabo line. To develop the algorithm, I propose that SM should be defined as in (10) below:
If the definition above is applied to example (11), we see that, after the application of synaloepha and the projection of the metrical syllables as asterisks onto GL0, the line has 13 metrical syllables including the syllable carrying the rightmost stress – corresponding to the 12th position – and three SM, marked below by square brackets:
Then a rule, specific to this metre, deletes the projection of the rightmost asterisk, unless it corresponds to a stressed syllable. The deleted asterisk is thus marked with Δ:
Note that the Δ syllable is invisible to iterative rules. These now begin scanning the line, starting just at the R-edge (parameter 6i,ii), inserting L brackets (6iii), and building binary groups (6iv) until they scan the whole GL0 from the R to the L edge. Two syllables, marked below by arrows, are skipped and left ungrouped:
In (11c) we see how the application of the iterative rules has created five right-headed binary groups (parameter 6v). The five heads get promoted to GL1, as in (11d), where parameters are set anew and the iterative rules apply again. Parenthesis insertion now begins just at the R edge, inserting L brackets, building R-headed binary groups:
The heads project onto GL2, and then the iterative rules once more build a single R-headed binary group (because there are just two asterisks left):
The rules building GLs 1–3 are less complex, in comparison to those required for GL0, which is in turn particularly relevant in loose metres. As we have seen in detail for (11), in these metrical forms SM intervene in the very first steps of grid construction; rule (8) allows these metres to leave syllables ungrouped, thus yielding a simpler (iambic in this case) structure than in strict metres.Footnote28 For this reason, in (12) I state the full set of rules that generate the bracketed grid for Montale’s loose endecasillabo:
The grid thus constructed is the output of ordered rules, and represents the structure of the tredecasillabo line. This analysis has two immediate advantages. First, it scans the line by automatically leaving out of its structure two syllables. This happens with no arbitrary decision concerning which linguistic constituent in the line should be part of the metre and which should not. Second, it brings out the structure of an iambic pentameter, with two prominent positions, one in the middle and one at line-end. This strikingly mirrors the structure of the endecasillabo. If we consider again what I have termed the Underlying Structure Hypotheses (example (4) above), it is easy to see how the present analysis directly confirms the hypothesis (4a), that of the ‘extrametrical orphan syllables’, which I reproduce in (13):
The present theory has thus provided a formal justification for considering the syllables corresponding to P5 and P8 as different from the others. While being part of the line’s linguistic representation, they do not belong to its underlying metrical representation – in other words, to its metre. It should be stressed again that this exclusion does not follow from any phonological property of the two syllables. Strictly determined by the counting rules of the metre, it only expresses the relative distance from SM.
The line analysed above is a straightforward case of a tredecasillabo line, and its scansion represents what the Bracketed Grid Theory achieves on the greater part of the corpus. I now discuss three other cases whereby the application of the algorithm is less obvious: first, a line which has two unstressed syllables following the 12th; second, a twofold case where the scansion does not yield the desired output, because of the specific stress patterns of the lines; third, I discuss the dodecasillabi, lines in which the last stress of the line falls on the 11th position. In all three examples (14–16) I only provide the metrical representation up to GL0, and highlight with arrows the points under discussion.
In the line above we find a rather common configuration in Montale’s language use: the choice of proparoxytonic words in both line-medial and line-final position. In this case, the algorithm – after deletion of the R-most syllable – scans through the line, correctly starting the grouping from the R-most stress because of rule (12iv): parenthesis insertion, that is, starts just at the R-edge, but is forced to skip the asterisk immediately to its left because of the SM on position 12. This is the effect of the interaction between loose-metre-specific formalism and rule (12vi), which deletes the R-most asterisk. The GL0 output is the desired 5-feet structure, and such examples motivate the specific setting of the parameters.
In (15a) and (15b) we see instead that the scansion procedure outputs a GL0 representation with a 6- and a 4-feet structure, respectively:
Such cases are a minority in the corpus, but not insignificant. In (15a) the line has a regularly alternating stress pattern (including potential stresses at the level of phonological feet, such as that on /nòn/). On the other hand, (15b) is one foot short, again because of the regular distribution of stresses. These two cases show an important property of the proposed algorithm and of a definite description in general: because the algorithm scans the structure of the line taking into account both SM and their location with respect to the whole line, it provides a disambiguation procedure capable of distinguishing which lines are in loose endecasillabi and which ones are not. This is not possible if one relies on the mere number of syllables in the representation of the line. (15a) could thus be analysed as a longer iambic metre, or as an Italian alexandrine metre if no synaloepha in line-medial position is applied; (15b), on the other hand, does not instantiate an underlying iambic metre, but should be analysed as an anapaestic metre, which would not be uncommon in the Italian tradition of the first half of 1900.Footnote29 This would also be a verse design that is indeed rare, but not impossible in Montale’s poetry (cf. also the line quoted in (1b), ‘vále l’áltro ma quésto mi attíra di più’, among other occurrences). (16) shows just what GL0 for the anapaestic metre would look like, but I provide no details on the relevant rules. In both cases, the proposed scansion excludes such configurations from the corpus of tredecasillabi. This alternative scansion, represented in (15b), thus confirms the intuition that such lines are in a strict anapaestic tetrameter:
We should also consider the other case that the Italian metrical tradition has regarded as an instance of anisosyllabic versification: the dodecasillabo. An example is analysed in (17):
The algorithm proposed here for the loose endecasillabi scans the line above, producing exactly the same well-formed grid as in the case of tredecasillabi. The output is different from tredecasillabi in that the same rules in (12) in this case end up skipping only one syllable, as shown by the arrow in (17). This scansion proceeds in the same way for all other dodecasillabo lines except for lines that share with (15b) the ternary alternation in the stress pattern, which should then be considered anapaestic and not iambic metrical forms.
None of the three examples is an exception to the argument of this article. On the contrary, they specify the contribution that the Bracketed Grid Theory of metre can make to the analysis of these forms. A definite description not only characterizes precisely the form at stake, but also separates other forms that may be similar on the surface (by mere syllable-counting) but do not share the same underlying structure. In this article, I have shown how an account of these lines as a loose variant of endecasillabo can accomplish this task.
Conclusions
The German philosopher Leibniz would arguably have said that if two things can be distinguished by some property, they cannot be identical. The form of the tredecasillabo is not defined by its length only, but by its very structure; a precise identification of this form uncovers the differences from other lines of the same length and, as a correlate, its relation to the endecasillabo. This article has developed a formal argument concerning the structure of the tredecasillabo in order to shed light on these formal relationships. This argument has highlighted where the similarities actually lie by analyzing the tredecasillabo as a kind of loose endecasillabo metre; and it has excluded some specific lines which, though apparently similar in length to the tredecasillabo, should not be regarded as members of the same set. On the basis of the structural observation that the tredecasillabo and the dodecasillabo share with the endecasillabo an identical underlying (5-feet) structure, the present theoretical proposal provides a correct scansion for the c. 200 tredecasillabo lines in Montale’s poetry within the framework of Fabb and Halle’s theory of metre; the dodecasillabo lines do not differ in any substantial regard under this analysis, and should also be considered instances of the metre I have termed ‘loose endecasillabo’. I have shown how previous approaches to these metrical forms all contribute to the idea that the 5-feet structure of the endecasillabo lies at the basis of these metrical forms, but that an exact, non-arbitrary analysis needs the Bracketed Grid Theory to provide a definite description for these lines.
It is a fact that Montale adopted in his poetry other forms that can be analysed as loose metres, including the dactylic hexameters of the Epigramma dedicated to Camillo Sbarbaro, and other measures longer than the endecasillabo. These have been traditionally considered compound metres, but they can be more precisely characterized as a kind of loose metre. And a number of other cases could potentially receive a similar explanation, such as the long lines of the poem La primavera hitleriana. It is also a fact that, when translating poets that experiment with metres variable in length, such as Gerard Manley Hopkins, T. S. Eliot, or Dylan Thomas, he tends to regularize his own renditions, while adopting a looser kind of versification in his own poetry. The scansion procedure developed by Fabb and Halle for loose metres thus may have further promising applications in the study of the Italian metrical tradition,Footnote30 and of neighbouring traditions. In particular, the rule proposed here in (12iii) closely resembles the rule adopted by Fabb and Halle to scan Hopkins’s sprung rhythm. Although it is not the objective here to argue for a direct relation between the two analyses, there is a potential structural similarity between Montale’s versification and the verse of an author he knew and translated.
The argument put forth here has a general consequence as well as one specific to Montale’s poetry. On the one hand – regardless of which formalism one chooses to adopt – an exact definition of a metre’s properties is the prerequisite for understanding both historical and formal changes in versification systems, such as the one that induced the Italian poetic tradition to expand the boundaries of its strictly isosyllabic versification – up at least to the second half of the nineteenth century. This probably happened under the influence of other European traditions. The issue of how that change came about, and interacted with the Italian language, still needs to be fully investigated. On the other hand, Montale certainly absorbed influences from a number of European literary traditions; and was himself a talented translator. When it comes to versification, he explored most possibilities available within the Italian tradition, and rarely declared his influences. However, if the present argument is correct, it represents a basis from which we can start asking questions as to what these ‘loose endecasillabi’ actually are: for example, whether they are a return to pre-canon anisosyllabic tendencies, or whether they are an attempt to mimic northern European modern poetic traditions, and if so, which ones. The formal account developed here thus sets out to be, rather than a conclusion, a starting point and a contribution to these broader questions.
Acknowledgements
I am greatly indebted to several scholars who dedicated their time to commenting on this article. The first acknowledgement goes to Nigel Fabb, whom I have bothered with the tredecasillabi since the time of my PhD thesis. More recently, he gave me precious advice on the technical-metrical part of this work. I am also grateful to Stefano Colangelo, Costanzo di Girolamo, and Paolo Giovannetti for having commented on the article at various stages of its preparation. Insightful remarks from an anonymous reviewer, as well as by Sara Sullam and Vassilina Avramidi, helped me shape its final form. Not least, this article cannot but be dedicated to the memory of Morris Halle, who back in 2009, after listening to the core of the present proposal, exclaimed: ‘Loose metres in Italian! That’s a discovery!’
Disclosure statement
No potential conflict of interest was reported by the author(s).
Notes
1 All Montale’s lines are quoted from Eugenio Montale, Tutte le poesie (Milan: Mondadori, 1984).
2 This is not the case for another noted author, Corrado Govoni, who had some influence on Montale (e.g. according to Pier Vincenzo Mengaldo, La tradizione del Novecento: Terza serie (Turin: Einaudi, 1991), pp. 37–8. Govoni’s tredecasillabo lines may appear more consistently throughout entire poems, and not occasionally, as in Montale’s poetry.
3 See e.g. Antonio Pinchera, La metrica (Milan: Mondadori, 1999).
4 See Costanzo Di Girolamo, Teoria e prassi della versificazione (Bologna: Il Mulino, 1976), p. 119.
5 Giuseppe Sansone, ‘Appunti sul tredecasillabo e sull’endecasillabo ipermetro’, Giornale storico della letteratura italiana, 128 (1951), pp. 176–83 (pp. 178–9).
6 Ibid., p. 183.
7 See Pier Giorgio Beltrami, Gli strumenti della poesia (Bologna: Il Mulino, 1996), pp. 187–8.
8 Massimo Antonello, La metrica del primo Montale: 1915–1927 (Lucca: Marina Pacini Fazzi, 1991).
9 See Pinchera.
10 Paolo Giovannetti, Metrica del verso libero italiano (1888–1916) (Milan: Marcos y Marcos, 1994), p. 231.
11 Paolo Giovannetti, Modi della poesia italiana contemporanea: Forme e tecniche dal 1950 ad oggi (Rome: Carocci, 2005), pp. 114–16.
12 Pier Vincenzo Mengaldo, La tradizione del Novecento: Seconda serie (Turin: Einaudi, 2003), and La tradizione del Novecento:Terza serie.
13 See Pier Vincenzo Mengaldo, La tradizione del Novecento: Da D’Annunzio a Montale (Milan: Feltrinelli, 1975).
14 See Pier Vincenzo Mengaldo, ‘L’opera in versi di Eugenio Montale’, in Letteratura italiana Einaudi: Le opere, iv.1, ed. by Alberto Asor Rosa (Turin: Einaudi, 1995), pp. 625–8 (p. 16).
15 In Mengaldo, La tradizione del Novecento: Terza serie, pp. 37–8.
16 Mikhail L. Gasparov, A History of European Versification (Oxford: Clarendon Press, 1996).
17 See Stefano Versace, ‘A Bracketed Grid Account of the Italian Endecasillabo Meter’, Lingua, 143 (2014), 1–19, for the development of Gasparov’s perspective into a complete generative account of the Italian endecasillabo.
18 I use the same examples cited in the analyses by Beltrami and Antonello.
19 Throughout this article, I refer to the specific notion of prosody developed in Marina Nespor and Irene Vogel, Prosodic Phonology (Dordrecht: Foris, 1986).
20 See again Beltrami, pp. 187–8.
21 Gian Giorgio Trissino, ‘Poetica’, in Trattati di poetica e retorica del Cinquecento, ed. by Bernhard Weinberg, iii.1 (Bari: Laterza, 1970), pp. 24–158 (p. 49).
22 Ray Jackendoff, Languages of the Mind (Cambridge, MA: MIT Press, 1992).
23 Costanzo Di Girolamo, ‘Gli endecasillabi dei siciliani’, in Bollettino di studi filologici e linguistici siciliani, 24 (2013), 289–312.
24 See Nigel Fabb and Morris Halle, Meter in Poetry: A New Theory (Cambridge: Cambridge University Press, 2008).
25 A few exceptions to this are: the above-mentioned article by Versace, as well as Nina Topintzi and Stefano Versace, ‘A Linguistic Analysis of Modern Greek Dekapentasyllavo Meter’, Journal of Greek Linguistics, 15 (2015), 235–69, and Laura C. Dilley and J. Devin McAuley, ‘The Fabb–Halle approach to Metrical Stress Theory as a Window to Commonalities between Music and Language’, in Language and Music as Cognitive Systems, ed. by Patrick Rebuschat, Martin Rohrmeier, John A. Hawkins, and Ian Cross (Oxford: Oxford University Press, 2012), pp. 22–31.
26 See Fabb and Halle, pp. 67–93.
27 Ibid., p. 13.
28 The reader may refer to Fabb and Halle for a more detailed comparison between strict and loose metrical forms, and to Versace for an exhaustive discussion of the application of Bracketed Grid Theory to the Italian endecasillabo metre.
29 Cf., e.g., the metre of Cesare Pavese’s Lavorare stanca. A similar parallel is highlighted by Mengaldo (in La tradizione del Novecento: Terza serie), although he considers Pavese’s lines to be loosely hexametrical.
30 As an example of such approaches, in particular one that aims to discover metricality in apparently free verse texts, could be a dissertation I recently co-supervised: Alessia Giordano, Un’analisi metrica del verso di Umana gloria: L’anapesto e la Teoria della Griglia con Parentesi (tesi di laurea magistrale, Università degli Studi di Bologna, 2024).