Juan de Mena’s arte mayor verse in Laberinto de 
fortuna: Identifying and describing a Spanish 
dolnik

Jesús M. Saavedra Carballido*

Abstract: The metre used by the Spanish writer Juan de Mena in his long allegorical 
work Laberinto de fortuna has been puzzling metrists for over five centuries. Usually 
identified as a variety of the verse called arte mayor, this metre has been analysed, 
among other things, as a divided line – one systematically allowing internal extra-
metricality – and as an amphibrachic line. Here it will be argued that the arte mayor 
exemplified by Mena’s Laberinto does not allow internal extrametricality and that it 
is a dolnik, a syllable-stress form not usually discussed in connection with the Span-
ish language. At the same time as it introduces the main problems encountered by 
traditional approaches, this article analyses the constant and necessary characteristics 
of the metre used in Mena’s composition, advocating the reliance on the notion of 
the metrical maximum. Next come some comments on other phenomena that are 
constant in Laberinto but not necessary in other kinds of arte mayor, and on several 
characteristics displayed by those other kinds but absent from Mena’s work. The Con-
clusion offers a series of suggestions for further research into the limits of this metre.

Keywords: Spanish versification, Juan de Mena, arte mayor, dolnik, maximum

In his classic study on Spanish verso de arte mayor (higher-artistry verse), 
Le Gentil states that this is “une versification […] assez compliqué, dont le 
secret a longtemps échappé aux critiques” = “a rather difficult versification 
[…] whose secret has escaped critics for a long time” (1953: 363).1 Despite 
the best efforts of the metrists coming after him, the nature of arte mayor has 
remained “obscure” (Roubaud 1971: 372).

* Author’s address: Jesús M. Saavedra Carballido, Universidade de Santiago de Com-
postela, Rúa das Carretas 22, 3, Santiago de Compostela, 15705, A Coruña, Spain, email:  
jesus.saavedra@usc.es.
1 The author wishes to thank the editors – Maria-Kristiina Lotman in particular – for their 
patience, and the anonymous reviewers for their insightful comments and suggestions. In 
responding to them, every effort has been made to keep the argument as easy to follow as 
possible. To avoid essay-like notes, the background information concerning Spanish metrics, 
phonology and previous theories of arte mayor has been confined – together with all statistics – 
to the Addenda.

Studia Metrica et Poetica 7.1, 2020, 47–96

https://doi.org/10.12697/smp.2020.7.1.03

https://doi.org/10.12697/smp.2020.7.1.03


48 Jesús M. Saavedra Carballido

Arte mayor dominated learned Spanish verse throughout the 15th century, 
such that “no other long verse was ever able to establish itself ” (Clarke 1964: 
212). Here I shall focus on its most frequently studied variety, the one used by 
Juan de Mena in his allegorical work Laberinto de fortuna = Labyrinth of Fortune 
(finished in 1444). Mena has long been regarded the indisputable master of arte 
mayor, Laberinto has been called “the best poem” in this form, and, though per-
haps “not the typical” example, its verse “traditionally has been considered the 
norm” (Clarke 1964: 167); not for nothing is arte mayor also known as verso de 
Juan de Mena = Juan de Mena’s verse. Most conveniently, Mena’s 2376-line work 
represents “a consistent – and still sizeable – corpus” (Piera 1981: 102). For all 
of these reasons, it is a good starting point in the study of this form.

Domínguez Caparrós classifies arte mayor as an irregular form (2014: 
159ff.), i.e. one that cannot be properly called free verse but whose lines, despite 
being metrically equivalent, display some oscillation in terms of syllable count. 
Though he describes arte mayor as accentual, from his account it is obvious 
that he does not refer to a metre regulating stresses alone, but both the syllable 
count and the distribution of stresses;2 this view is shared by most scholars 
(e.g. Duffell 1999: 60– 61, 73–74).

In an oft-quoted passage, which has served as the basis of later descriptions 
such as Domínguez Caparrós’s, Clarke states that the “basic pattern” of arte 
mayor can be “roughly” represented as:

(w) s w w S (w) # (w) s w w S (w)3

As she explains, w corresponds to a nonstress, S to a stress and s to a second-
ary – i.e. optional – stress;4 the hash sign represents a midline boundary rule 
or caesura (1964: 51). She adds:

2 Domínguez Caparrós makes it clear that the metre that he is dealing with tends to be aniso-
syllabic rather than isosyllabic (2014: 45, 159). His reference to arte mayor as accentual coincides 
with that of previous authors like Henríquez Ureña (2003: 171, 175, 182, 192–193, 241), who dis-
tinguishes two Spanish metres in which the number of syllables is irregular or fluctuating (within 
limits): one he calls amétrico (ametrical); the other, to which are mayor would belong, acentual 
(accentual). Actually, both are types of metrical verse: neither is what we would call free. The differ-
ence between them seems to lie in the (relative) absence or presence of an insistent rhythm based 
on the placement of stresses, which is less noticeable in “ametrical” verse than in “accentual” verse.
3 Clarke’s notation has been adapted to the one employed in the rest of this article.
4 In Spanish, syntactic words are prosodically divided into tónicas and átonas, that is, tonic 
(always stressed) and atonic (usually unstressed). For more details on Spanish stress, see Section 
2c of the Addenda.



49Juan de Mena’s arte mayor verse in Laberinto de fortuna ...

The caesura is movable between [the second and third] stresses, the second-
ary stresses are not absolutely fixed in required presence or position, and the 
unstressed syllables in parentheses are not absolutely fixed except that at the 
caesura at least one of the unstressed syllables is usually present (Clarke 1964: 
51; my italics).

Clarke assumes the line to be unified (verso simple) rather than divided (verso 
compuesto), and this despite the caesura (1964: 57, 60, 169, 210). Her opinion 
is in line with other descriptions of Spanish metrical verse (for its general 
characteristics, see Section 2a of the Addenda). Among the “common […] 
modifications of the pattern”, she mentions “the use of two unaccented syl-
lables preceding the [first] secondary stress”, instead of one or none (1964: 
168). Excluding the optional nonstress at line end, arte mayor lines have from 
eight – usually nine – to twelve syllables.5

Clarke’s description is quite detailed, yet it fails to explain the syllable count, 
the exact distribution of stresses, the precise nature of the midline boundary 
and the relation between the two parts of the line. The uncertainties which 
surround Clarke’s analysis are common to all existing accounts of arte mayor. 
In the course of time, to begin with, its template has been described as falling, 
or – more usually – as amphibrachic or rising (for previous theories of arte 
mayor, see Section 3a of the Addenda). To start dispelling these uncertainties, 
I shall take my cue from Ker and García Calvo. Ker pointed out over a century 
ago the affinities between Spanish arte mayor and the English metre known 
as tumbling verse (1899: 119 ff.). So did García Calvo more recently (2006: 
1627–1628, 1640–1662). Arguably, the metre with which these two authors 
associate arte mayor is a dolnik – a metre that regulates the syllable count and 
the distribution of stresses so that strong positions will be usually realized 
by a stressed syllable and usually separated from each other by two or one 

5 For the theoretical assumptions behind the metrical approach adopted in this article, see 
Section 1a of the Addenda. I shall always refer to the length of verse lines counting up to the 
syllable realising the last strong position, and disregarding as extrametrical the material that may 
follow. By doing this, I avoid the traditional Spanish terminology according to which, because 
of the predominance of paroxytonic lines (with one extrametrical syllable), the total number 
of syllables in a line is always calculated by counting up to rightmost stress (or equivalent) and 
adding one. If we applied this traditional system to the English iambic pentameter (a 10-position 
line, according to most current accounts), we would have to describe it as an 11-syllable line 
(cf. the Spanish term endecasílabo). The traditional system is overcomplicated, because it relies 
on counting and adding up when only counting is necessary; it has been criticized, but never 
replaced once and for all, by a number of Spanish scholars (cf. Piera 1981: 5–7; Domínguez 
Caparrós 2014: 62 n. 41).



50 Jesús M. Saavedra Carballido

unstressed syllables – with a predominant but not constant ternary rhythm. 
Unfortunately, these two authors do not go into any details concerning the 
exact characteristics of the metre and its realization. What I propose to do is 
analyse the constant and necessary features of the verse of Laberinto by treating 
it as a rising dolnik made up of lines with four strong positions.6

To substantiate the dolnik interpretation, I shall examine the arte mayor 
used in Laberinto,7 comparing it with compositions by other Spanish authors 
from just before Mena’s times (Alfonso Álvarez de Villasandino and Micer 
Francisco Imperial) to contemporary times (Ramón María del Valle-Inclán); 
scholars have identified some of the verses used by these authors as varieties 
of arte mayor, and I accept and take advantage of this identification.8

6 Dolnik is the Russian name – lately adopted in English – for the metrical form to which I 
am assimilating arte mayor (cf. Tarlinskaja 1993: 12; Gasparov 1996: 18–26, 175–200, 274–281; 
Attridge 2012; 2013: 103–111, 147–177). In English dolnik has received other names, among 
them loose iamb (Fabb and Halle 2008); but dolnik and loose iamb have sometimes been dis-
tinguished on statistical grounds (for example by Tarlinskaja 1992: 3, 5; 1995: 505; Hanson 
2008: 380). Tarlinskaja characterizes dolnik as follows (her ictus is a strong position; ictic and 
nonictic positions are strong and weak, respectively; nonictic syllables are those realizing weak 
positions): “The regularity of stress alternation […] is fairly consistent, so the ictic positions 
are, as a rule, easy to identify. However, the number of non-ictic syllables (usually unstressed 
or, less often, bearing lighter stresses) between adjacent ictuses may vary between one and 
two, and the number of syllables […] preceding the first ictus, may be either constant, or vary 
from 0 to 2” (Tarlinskaja 1993: 12). As Attridge does for English dolnik (2012: 8–9 n. 15; 2013: 
107–108, 154–155 n. 11), Tarlinskaja admits that a weak position between strong positions may 
be realized by “zero unstressed syllables”, though she thinks that this only happens “in rare cases” 
(Tarlinskaja 1993: 13); Duffell concurs (2008: 34). In Spanish, such unrealized or void weak 
positions do not occur. Anisosyllabic dolnik is to be distinguished from another metre often 
referred to in English, again by its Russian name, as taktovik. This is an even “looser” form in 
which “the intervals between adjacent ictuses vary from one to three syllables” (Tarlinskaja 1993: 
14; cf. Gasparov 1996: 18–26, 274–275). In English taktovik has also been identified as a kind of 
loose anapaest (a metre allowing 4-syllable intervals, according to Fabb and Halle 2008: 80-82); 
but when Piera analyses arte mayor as loose anapaest (2008), he only contemplates 3-syllable 
intervals.
7 I rely on the text established in Maxim P. A. Kerkhof ’s authoritative edition (Mena 1995). 
Though the text that it offers does not differ substantially from that of other editions, such as 
John G. Cummins’s (Mena 1979), Kerkhof ’s is the only one that collates the variants from all 
known sources.
8 Navarro Tomás traces the whole history of Spanish arte mayor from the Middle Ages to the 
20th century (1974: 95–98, 115–124, 225–226, 277–278, 325, 443–444, 524). Saavedra Molina 
(1945: 9, 18–20, 31–32, 45–46, 54–57, 65–66, 115–116), Navarro Tomás (1974: 115–117) and 
Clarke (1964: 69–72) agree that Villasandino was a practitioner of arte mayor. According to 
Saavedra Molina (1945: 18, 32, 62–63), Navarro Tomás (1974: 117), Clarke (1964: 99–151) and 
García Calvo (2006: 1656), so was Imperial. The basic identity between the metre of Laberinto 



51Juan de Mena’s arte mayor verse in Laberinto de fortuna ...

Constant and necessary characteristics of arte mayor in 
Laberinto

It has been repeatedly said – and can be easily noticed – that the predominant 
rhythm in Laberinto is usually ternary (e.g. Navarro Tomás 1974: 115–116; 
Duffell 1999: 56; for my own figures, see Sections 4b and 4c of the Addenda). 
To begin with, therefore, a convenient way of approaching the arte mayor of 
Laberinto is to analyse it as a rising ternary metre with four strong positions, 
each realized by a stressed syllable and separated by two unstressed syllables; 
all syllables to the right of the rightmost stress are considered extrametrical 
(they do not correspond to any of the metrical positions that make up the 
template). The only relevant variations that this metre allows are in the number 
of nonstresses before the first strong position (from zero to two) and of final 
extrametrical nonstresses (either zero or one). The template would be:

(w) (w) s w w s w w s w w s {x}

Here s signals a strong position, w signals a realized weak position, (w) indi-
cates an unrealized (void) weak position, and {x} represents the optional 
extrametrical material.

In Laberinto, the following 12-syllable line with an oxytonic ending, i.e. 
without extrametrical nonstresses, realizes the ternary template quite faithfully 
(X marks a stress and x a nonstress):9

and that of some Valle-Inclán folk-inspired compositions has been suggested by Saavedra 
Molina (1945: 10, 91–92) and Navarro Tomás (1974: 444), the latter of whom mentions their 
predominant binary rhythm. (In English there is a similar division between folk or folk-inspired 
dolnik, which is predominantly binary, and dolnik of learned origin, as shown by Tarlinskaja 
1993: 39.)
9 I accept the way in which Kerkhof stresses words in his edition of Laberinto. This means 
assuming, with Piera (1981: 109), that Mena assigned stresses in much the same way as in 
present-day Spanish. There are some possible exceptions, mostly in proper names and cultisms, 
some of them adapted to fit the rhyme (cf. Lázaro Carreter 1972: 355–356, 360; Piera 1981: 
141–144). In this respect, it must be borne in mind that, back then, many words of classical ori-
gin were stressed differently from now, and that the stressing of others may have been unstable, 
so Mena’s tinkering is much less frequent than it might seem. This tallies with the idea that in 
Spanish verse stress wrenching is generally confined to song, to burlesque compositions and to 
folk-song imitations intended for recitation (Henríquez Ureña 2003: 183 n. 18, 257–258 n. 26). 
As Mena’s address to the “lectores”= “readers” implies (l. 1910), and as Lázaro Carreter admits 
(1972: 354), Laberinto was meant to be read not listened to.



52 Jesús M. Saavedra Carballido

(1) Aristótiles cerca del padre Platón (l. 939)10

 x  x  X x x   X  x   x    X   x    x X

Since, in Spanish, paroxytonic words are far more numerous than oxytonic 
ones (cf. Quilis 1993: 403), few Laberinto lines have this abrupt ending; in even 
fewer is the stressing so regular.11 Similarly rare are 11-syllable lines (with a 
void weak position at the beginning) without extrametricality, like this one:

(2) Agora, respuse, conosco mejor (l. 1881)
 x  X x    x   X  x   x  X   x     x X

Lines with a paroxytonic ending, i.e. with one extrametrical nonstress at the 
end, are more abundant. Here is an 11-syllable example:

(3) tomando castillos, ganando lugares (l. 1181)
 x   X    x   x  X  x      x  X    x  x  X {x}

Such 11-syllable lines with extrametricality are extremely frequent.12 So much 
so that many commentators have considered them to reveal an ideal arte mayor 

10 In addition to initial and final punctuation, I remove quotation marks and editorial diaer-
esis – used by editors to indicate that two contiguous vowels cannot be treated as one single 
syllable – from all the examples.
11 My scansions reflect all stresses corresponding to the prosodic heads of tonic words, includ-
ing the head of each part of compounds (primary stresses), yet disregards all other possible 
stresses not corresponding to the heads of tonic words (secondary stresses). I assume that 
these secondary stresses are inexistent or irrelevant for Laberinto. For the rationale behind this 
assumption, see Section 3b of the Addenda.
12 Navarro Tomás has not hesitated to scan a whole stanza of Laberinto as belonging to this 
type (1974: 119): “Con dos quarentenas e más de millares / le vimos de gentes armadas a punto, / 
sin otro más pueblo inherme allí junto, / entrar por la vega talando olivares, / tomando castillos, 
ganando lugares, / faziendo por miedo de tanta mesnada / con toda su tierra temblar a Granada, 
/ temblar las arenas fondón de los mares” (stanza 148, ll. 1177–1184). Given the presence of 
contiguous vowels in most lines, other scansions are of course possible. According to the figures 
furnished by Foulché-Delbosc (1902: 99–102), which encompass all lines in Laberinto, this 
kind of line represents 56.8 % of the total. His analyses are not entirely reliable, as they do not 
have clear criteria concerning how to count syllables and what counts as the rhythm-marking 
element (sometimes it is stresses corresponding to the heads of tonic words; sometimes, other 
alleged stresses whose nature is not specified). Nevertheless, these figures suffice to direct our 
attention to unmistakable tendencies – in terms of syllabic length and of stressing – in Mena’s 
arte mayor. My own calculations of line types in Laberinto can be consulted in Section 4a of 
the Addenda. In this case, they are largely congruent with Foulché-Delbosc’s (55.2 %).



53Juan de Mena’s arte mayor verse in Laberinto de fortuna ...

form corresponding to a 4-amphibrach template (the vertical bar signals the 
metrical boundary between consecutive feet):

w s w | w s w | w s w | w s w

Also frequent are 10-syllable lines without nonstresses at the beginning and 
with one extrametrical syllable at the end:

(4) sirve metales, metales adora (l. 789)
 X   x    x X x      x  X  x   xX {x}

The last three are, by far, the most usual kinds of line in Laberinto.13 Another 
source of variation from isosyllabic ternary verse comes not from the addition 
and ommission of nonstresses at the extremes but from the fact that stresses 
are not confined to strong positions or nonstresses to weak positions. If we 
resist the temptation to suppress and add stresses in order to be more faithful 
to the alleged template, we realize how frequent these kinds of departure are. 
Witness the second stress in this line (I provisionally underline the syllables 
realizing strong positions in such a way that a ternary rhythm is somehow 
preserved):

(5) Dame tú, Palas, favor ministrante (l. 1125)
   X   x  X    X x     x  X     x  x   X  {x}

The occurrence of unexpected stresses in weak positions is a well-known and 
accepted phenomenon in Spanish and in other versifications, where it receives 
different names (in English, for instance, Attridge calls it demotion). The 
inverse phenomenon, an unexpected nonstress in a strong position (Attridge’s 
promotion), is likewise permitted. In the following line, for example, if we 
want to keep the four strong positions that most commentators agree on, but 
without adding stresses at will, we have to admit that the first strong position 
is realized by a nonstress:

13 According to Foulché-Delbosc (1902: 99–102), amphibrachic lines, including those without 
a final nonstress, amount to 63.4 %. Adding headless amphibrachic lines (31.4 %), without and 
with a final nonstress, we have over 94 % of lines in Laberinto. Finally, 12-syllable ternary lines 
like example (1) amount to 0.1 %. My figures are similar (97.3 % and 0.2 % respectively). Obvi-
ously, one of the difficulties of arte mayor analysis lies in accounting for the minority of lines 
that remains.



54 Jesús M. Saavedra Carballido

(6) introduzido por público pro (l. 1844)
 x    x  x X x   x     X   x x    X

In this other example, the affected strong position is the third:

(7) así como príncipe legislator (l. 643)
 x X  x    x    X   x  x  x x  x X

From ternary verse to dolnik

The variations observed until now are among the most frequent in Laberinto. 
They are not the only ones though. More troubling for the ternary and amphi-
brachic theory are cases like the following, in which stresses are separated by 
one nonstress instead of two:

(8) en Ras con aquel señor de Charní (l. 1590)
 x     X    x   x    X   x  X    x      x   X

One way of explaining this phenomenon, in particular within the amphibrach 
framework, has been to speak of a missing nonstress in the middle of the line 
(Piera 1981: 130; Dufell 1999: 68). From the perspective of traditional Spanish 
metrics, this solution has an ad hoc air. Moreover, it cannot explain the fol-
lowing case, where a nonstress does not seem to be missing midline (between 
the second and third stresses) but elsewhere (between the third and fourth 
stresses):

(9) aquel coraçón que si no querer (l. 894)
 x    X   x  x  X       x  x  X    x  X

From the perspective of a rising ternary template, it can be argued that lines 
without two initial nonstresses are headless, and that final nonstresses are 
extrametrical. From the perspective of an amphibrachic template, it has been 
argued that lines with an oxytonic ending are truncated, that lines beginning 
with a stress are headless, and that lines beginning with two nonstresses have 
1-syllable initial extrametricality. The possibility of extrametricality and/or of 
void weak positions is thus assumed at the beginning and at the end of lines. 
Internally, void positions must also be allowed.



55Juan de Mena’s arte mayor verse in Laberinto de fortuna ...

The assumption of so many void positions in arte mayor is problematic, 
because in Spanish final truncation and initial headlessness are not strictly 
symmetrical: truncation is considered a permissible exception to the usual 
paroxytonic ending (cf. Domínguez Caparrós 2014: 60ff.); by contrast, head-
lessness is not habitually acknowledged, at least in long lines. The same can be 
said of initial extrametricality, which is traditionally permitted in short lines 
but not in long ones. As regards internal voids, the truth is that, outside song, 
Spanish verse does not usually permit them at all.

The dolnik solution proposed here is unusual too, yet simpler. What I sug-
gest is treating the verse of Laberinto as a rising, 8-position form in which 
internal weak positions may be realized by two or one syllables and initial weak 
positions may be realized by two, one or – in the case of headlessness – zero 
syllables. Its template would be:14

  (w) s | w s | w s | w s {x}

The dolnik hypothesis makes it possible to forget about void weak positions 
between consecutive stresses. (Later we shall see that, in addition to being sim-
pler, this interpretation of arte mayor may also be more accurate, as it predicts 
lines that are absent from Laberinto but present in other compositions.) The 
downside is that it may complicate the identification of some strong positions 
in certain lines:

(10) oras silvando como dragón (l. 1965)
  X x    x  X   x    x    x     x X
  X x    x  X   x    x    x     x X

This difficulty can be solved, in some cases, thanks to the notion of the maxi-
mum. For the time being, I shall just suggest that, in order to account for 
the demotion of odd-numbered strong positions without manipulating the 
stresses and thus violating the givens of the language, the metrical template 
of arte mayor should be conceived of as containing two cola and two kinds of 

14 The idea that weak positions may be realized by two or one syllables is taken from such 
disparate authors as Attridge (1982; 2012; 2013), Chisholm (1995) and García Calvo (2006: 
1640–1662). Alternatively, it would be possible to posit the template: (w) (w) s | w (w) s | w (w) 
s | w (w) s {x}, where every s is realized by one syllable, every w after s is also realized by one 
syllable, and any other w is optionally realized by one syllable or left unrealized. My notation 
has the advantage of less cumbersome and of requiring not two but only one type of w when 
dealing with dolnik without headlessness.



56 Jesús M. Saavedra Carballido

strong positions, only some of which (the second and fourth) function as the 
head and rightmost limit of a colon (the hash sign marks the frontier between 
initial and final cola):

(w) s | w s # w s | w s {x}

To test the dolnik hypothesis, I shall apply it to lines whose prosodic analysis is 
relatively straightforward. The presence of contiguous vowels in Spanish verse 
can result in considerable variation in the syllable count and in the determi-
nation of the rhythm of the line, because in these cases all vowels or just one 
may be taken into account (cf. Piera 1981: 8; 2008: 95ff.). For this reason, I try 
to avoid instances with vocalic sequences (as advised by Clarke 1964: 105). 
Further, in Kerkhof ’s edition of Laberinto (Mena 1995), there are words writ-
ten with and without h, such as (h)avía = there was/were and (h)ermanos =  
brothers, a variation that may or may not affect the syllable count. In view 
of this, I also leave out of my analysis lines with h between vowels. For good 
measure, I also exclude lines with elided vowels (marked with an apostrophe 
’ by Kerkhof ). These exclusions should make it possible to determine more 
accurately the permissible line lengths. (The complete list of lines meeting 
these conditions can be found in Section 4a of the Addenda.)15

On the face of it, in the arte mayor of Laberinto colon-final strong posi-
tions are realized by a stress, while other strong positions are less strictly 
constrained: the syllables realizing them can be stressed or not. Weak posi-
tions are even less constrained: they can be realized by two or one syllables, 
whether stressed or not, or even – at the beginning – by zero syllables. Each 
colon is realized by between four and six syllables. Extrametrical syllables 
cannot be stressed.

This template and these realization rules suffice to explain all the examples 
given so far. In practice, the verse of Laberinto seems to consist in long unified 
lines; colon-final strong positions must be realized by a stress. The syllable 
count and the distribution of compulsory stresses are as follows: after a stretch 
of between three and five syllables, every line contains a stress realizing a 
colon-final strong position; then, after another interval of between three and 
five syllables, there is one more stress realizing the rightmost strong posi-
tion. In its reliance on colon-final anchor-points, coupled with the frequent 

15 I assume that the treatment of contiguous vowels in 15th-century verse was as it is today, 
and that this treatment is the same throughout the line (pace Piera 1981: 9, 134), but not across 
lines (cf. Piera 1981: 34). For the problem of contiguous vowels in Spanish metrics, see Section 
2b of the Addenda.



57Juan de Mena’s arte mayor verse in Laberinto de fortuna ...

disregard of stresses in other parts of the line, arte mayor would not differ from 
other Spanish long metres.16

For the sake of convenience, all instances of arte mayor can be labelled 
through a two-figure formula representing the number of syllables in each 
colon. The examples quoted so far belong to the following types:

6#6: example (1)
5#6: examples (2), (3), (7)
4#6: examples (4), (5), (6)
5#5: examples (8), (9)
4#5: example (10)

The frequencies of line types observed in this group are largely congruent with 
those found in Laberinto as a whole (see Section 4a of the Addenda). Judging 
from the examples, it can be surmised that the final colon cannot be realized 
by fewer syllables than the initial one, which itself cannot be realized by fewer 
than four syllables (even when there is headlessness). Laberinto excludes a 
construct like the following, which can only be scanned as 6#5 and yet reads 
quite well:

(a) *esperaba lograr mejor resultado
  x    x X x  x   X #   x X   x  x  X {x}

As examples (7) and (10) show, between the stress realizing the first colon-final 
strong position and the next stress sometimes there can be three or more syl-
lables instead of the two or one that Kerkhof stipulates (Mena 1995: 31, 110). 
The following 5#6 lines display similar departures from a ternary rhythm:

16 Colon-end (including line-end) stresses are thus the foundation of whatever stress rhythm 
arte mayor lines may have. It must be noted that, in Laberinto, it is precisely at colon-end 
(particularly at line-end, where there is rhyme) that most of the (few) unusual stressings that 
can be determined with certainty occur. An example: “tú, Penelope, la qual en la tela” (l. 510), 
where the proper noun Penelope must be pronounced as a paroxytone (X x x X # x…) not as a 
proparoxytone, as is done today, in order to prevent the two cola from being incorrectly realized 
by three and seven syllables, respectively. Compare it with “nueva Penélope aquésta por suerte” 
(l. 622), where editors treat the noun as proparoxytonic.



58 Jesús M. Saavedra Carballido

(11) vi despojadores e vi despojados (l. 1235)17

     X   x    x x  X#x  x X   x    x X {x}

(12) en las baxas artes que le da Minerva (l. 474)
     x    x    X x   X # x   x   x  X    x  X {x}

On this occasion I have only underlined the stresses realizing colon-final 
strong positions, because it is not always easy to determine which syllables 
correspond to other strong positions.18 Both these examples contain a stressed 
monosyllable that appears earlier or later than expected (in the initial and 
final colon, respectively) and that cannot be interpreted as realizing a strong 
position, but only as being displaced (i.e. omitted at some point and added at 
some other).

As shown by the final cola in these other 5#6 lines, the displaced stresses 
can also belong to polysyllables:

(13) primero fallaron, gentes amonitas (l. 284)
       x   X x  x  X#x     X   x      x  x  X {x}

17 If the metre that I am positing were not rising but falling, the template would be {x} s w | s 
w # s w | s (w), the initial colon would be realized by between three and four syllables up to the 
rightmost stress, and the final colon by between four and six syllables, i.e. the whole line would 
be realized by between seven and ten syllables up to the rightmost stress; truncation and 1- or 
2-syllable initial extrametricality would be necessary to account for many lines in Laberinto. 
This approach only works if extrametrical stresses are accepted at the beginning of 5- and 6-syl-
lable cola, as at the beginning of example (11); that is why the falling (in particular dactylic) 
hypothesis has been criticized by Piera (1981: 122–127). As it requires final extrametricality (as 
in other Spanish metres) instead of initial extrametricality, the rising version is more congruent 
with counting metrical positions from left to right up to the rightmost stress, precisely what 
should be generally done – as already suggested – with Spanish metres.
18 In cola realized by six syllables, the placement of the internal strong position is clear, and 
a ternary rhythm realizes the metre more faithfully than a binary one. If consecutive strong 
positions are ruled out, in cola realized by four syllables the placement of the internal strong 
position in final cola is also clear: it cannot be at the beginning, so in this case a binary rhythm 
is more faithful to the template; in initial cola, if we accept headlessness, the placement of the 
first strong position is clear when the first two syllables are a stress and a nonstress (s w s #, with 
headlessness) or a nonstress and a stress (w s w s #), but less so when both are stresses or non-
stresses. Finally, in cola realized by five syllables, the placement of the internal strong position 
cannot always be known with certainty beyond the fact that it must two or three syllables before 
the colon-final strong position. If we marked this internal strong position always according to a 
ternary rhythmic norm, we would be artificially reinforcing what is only a statistical tendency, 
making it more normative than it really is.



59Juan de Mena’s arte mayor verse in Laberinto de fortuna ...

(14) e lo que queremos menos acabamos (l. 232)
    x  x     x     x  X # x      X  x   x  x  X  {x}

What these two lines have in common is that a nonstress belonging to a poly-
syllable realizes a strong position while a stress in the same word appears in 
a weak position. This is a phenomenon that, from the perspective of other 
metrical traditions, can be taken to invalidate the dolnik explanation of Mena’s 
arte mayor (and also, for that matter, the ternary and amphibrachic explana-
tions – though not the taktovik explanation, which the syllable count makes 
nevertheless unlikely). It must be borne in mind, however, that the practice 
of misplacing disyllabic and longer words is common currency in Spanish 
verse.19 Such cases are relatively frequent in Laberinto (though not at colon 
end): a nonstress realizing a strong position can be next to a stress belonging 
not to a monosyllable (e.g. l. 474, example (12)) but to a polysyllable (e.g. the 
paroxytone mesmo = same in l. 815: “e madre del fi#jo de su mesmo pa{dre}”, 
whose final colon scans as …# x x x X x X {x}), and both the nonstress and the 
stress can belong to the same word, which therefore thwarts metrical expecta-
tions not only on one occasion, but on two or more, and not only within one 
foot (e.g. l. 284, example (13); l. 232, example (14)), but even across feet (e.g. 

19 Misplaced polysyllables do not detract from the metrical correctness, for example, of the fol-
lowing lines by the 16th-century author Garcilaso de la Vega, which realize a binary 10-position 
template of the a maiore type (6#4): “Cómo deste lugar hice mudanza” (X x X x x X # X x x X 
{x}), in which all stresses save those realizing colon-final strong positions realize weak positions 
and correspond to paroxytonic disyllables, while all nonstresses in the misplaced words occupy 
strong positions; “él olvidará presto la braveza” (X x x x X X # x x x X {x}), with the oxytonic 
tetrasyllable olvidará = will forget realizing the sequence s w s w; “recogido le llevan, alegrando” 
(x x X x x X # x x x X {x}), with the paroxytonic tetrasyllable recogido = gathered realizing the 
sequence w s w s (Vega 2001: 171, 195, 244). Despite these (permissible) sources of variation, 
in Vega’s hands the rhythm of the 10-position metre is overwhelmingly binary (cf. Duffell 1999: 
38–39). In English dolnik, misplaced polysyllables are rare but do not result automatically in 
incorrect lines. As in other English metres, initial and final stresses in polysyllables can appear 
after a metrical, syntactic or prosodic boundary (cf. Jakobson 1966: 363–364), but also in other 
contexts (for examples outside dolnik, cf. Kiparsky 1977; Attridge 1982: 265–275; Wright 1988: 
198–199). Take this dolnik line by Yeats: “The days chatter and scream” (from “The Tower”; 
qtd. Tarlinskaja 1993: 102); it realizes the template w s w s w s #. The paroxytonic disyllable 
chatter realizes either the second w s (Tarlinskaja’s interpretation: x X X x x X) or the second s 
w (in which case the preceding w is left unrealized; this would probably be Attridge’s preferred 
scansion, with the void marked as [x]: x X [x] X x x X). The same can be said of the paroxytonic 
disyllable started in “That was how the grief started to melt” (…x X X x x X {x} or …x X [x] X 
x x X {x}, from Frost’s “The Thatch”; qtd. Tarlinskaja 1993: 102). In Spanish, the assumption 
that voids play no role leaves us with Tarlinskaja’s reading, which does not differ much from 
my own interpretation of similar realizations in Laberinto.



60 Jesús M. Saavedra Carballido

the oxytone derribar = knock down in l. 2119: “Basta que pu#do derribar al 
u{no}”, whose final colon scans as …# x x x X x X {x}). (For the specific figures 
of misplaced polysyllables, see Section 4d of the Addenda; cf. the discussion 
and Appendixes in Duffell 1999: 74–78, 89–90.)

In the following case there are zero syllables between stresses belonging to 
different cola (the relative lengths of cola observed until now serves to disam-
biguate the line as 5#5 rather than 6#4):

(15) e sobre partir  tales discordanças (l. 2034)
   x  x    x   x  X # X x    x | x   X  {x}

What sets this example apart from the preceding ones is that the displaced 
disyllable tales = such can be interpreted as fitting snugly in one single weak 
position, hence as disrupting the rhythm just once  – through demotion. 
(By contrast, the strong position would be realized by the first nonstress of 
discordanças = discords – a case of promotion.) Leaving aside such cases of 
comfortable fit (another example would be the oxytone primer = first in the 
initial colon of l. 371: “que del primer fran#co que tovo corona”, whose initial 
colon can be scanned X X | x X X #…), in Laberinto all the displaced polysyl-
lables appear in final cola; there is even the possibility of misplacing several 
polysyllables (as in l. 1239: “e vi dos estre#mos fechos una par{te}”, whose final 
colon scans as …# x X x X x X {x}, where the paroxytone fechos = made realizes 
a weak and a strong position while the paroxytone una = one fits in a weak 
position). A possible explanation is that final cola are often longer, affording 
greater possibilities to deviate from the template, and that authors simply take 
advantage of them; to this it can be added that in the final colon the need to 
be strictly regular is not so pressing after the metre has been fairly faithfully 
indicated by the rhythm in the initial part of the line.

All this indicates that in arte mayor a ternary rhythm may prevail, but it is 
far from constant or necessary: stresses other than the ones at colon end are 
frequent, but they are not required; when they do occur, they often go against 
the predominant ternary rhythm, even if they belong to polysyllables.20 As a 
result, we even find lines with a binary rhythm throughout:

(16) Echate, dixo, non te fazen guerra (l. 1995)
    x   X x   X#x   X    x  X x       X {x} (4#6)

20 The pressure of the metre – and of the rhyme – over the author’s language, highlighted by 
Lázaro Carreter (1972: 356), occasionally shows in the form of wrenched stresses; but it must 
be remembered that some of the linguistic peculiarities of Laberinto that this scholar attributes 



61Juan de Mena’s arte mayor verse in Laberinto de fortuna ...

Introducing the maximum

The rules given so far cover most lines in Laberinto. However, they cannot 
explain some instances in which colon-final strong positions are realized by 
nonstresses (given the scarcity of lines without contiguous vowels, in some of the 
following examples I put that restriction aside, scanning only the relevant part):

(17) el ánima contra la falsa fortuna (l. 1554)
    x  X  x  x   x #  x  x  X  x  x  X {x} (5#6)

(18) Ca un condestable armado, que sobre (l. 2113)
                                   …    X  x      x   x {x}

In these two lines (in which the underlined x indicates a nonstress realizing a 
strong position) the relevant atonic words are disyllabic. This is not a neces-
sary condition: the distribution of cola in the following example is ambiguous, 
but whether we scan the initial colon as ending with los = those or with que = 
whom its rightmost word is an atonic monosyllable:

(19) a fijos de los que libró del desierto (l. 1244)
   x X  x  x   x #   x     … (5#…)
   x X  x  x   x      x #  … (6#…)

When analysing them, these nonstresses should not receive any emphasis, 
lest we go back to the old practice of stress manipulation. It is clear, then, 
that in Laberinto some colon-final strong positions are realized by nonstresses 
which can belong to both monosyllables and disyllables (a relatively rare yet 
important phenomenon that recurs throughout the history of Spanish metri-
cal verse). It is simply not true that all colon-final nonstresses in Laberinto 
correspond to a “monosyllable”, as Piera states (1981: 140).

to the influence of the metre also appear in Mena’s prose (cf. Cummins in Mena 1979: 43). The 
fact that Mena does not always insist on a regular ternary rhythm does not prevent performers 
from regularizing it as they please. Obviously, nothing precludes a performance of Mena’s arte 
mayor with four perceptible stresses, as evenly spaced as possible, in all lines. However, this 
imposed regularity cannot form the basis of metrical analysis: as soon as it has more than two 
syllables, any line – of prose and of verse, whether metrical or free – can be performed with a 
perfect ternary rhythm (but then it can also be performed with a binary movement). The differ-
ence between strict ternary metres and other kinds of verse – dolnik included – lies elsewhere: 
not in the performance but in the rhythm that is phonologically determined.



62 Jesús M. Saavedra Carballido

Either we declare the last three examples to be incorrect, or we find a way 
to accommodate colon-final atonic words within our rules. I shall try to do 
the latter by adopting the notion of the metrical maximum, a kind of syllable 
that metrical rules recognize as particularly salient (i.e. a metrified stress), 
banning it from (certain) weak positions and/or requiring it in (certain) strong 
positions (for the history of the notion of the maximum, see Section 1b of the 
Addenda). I suggest that in Laberinto colon-final strong positions are realized 
by maxima which need not be stressed.21

The first step towards a fitting definition of the maximum is to recognize 
that in Laberinto some unstressed syllables can perform this role, but not all. 
Judging from what we have just seen of Mena’s arte mayor, the only nonstresses 
that count as maxima belong to what Spanish phonologists consider atonic 
words (palabras átonas), never to tonic words (palabras tónicas). This excludes 
constructs like the following, in which a colon-final strong position is realized 
by the last nonstress  (indicated by an underlined x) of a tonic proparoxytone:22

(b) *los otros estaban pensándoselo
   x   X   x  x  X #x     x   X    x  x x (5#6)

(c) *las grandes hazañas históricas
   x      X    x    x X #x    x  X x x (5#5)

The phonological reason for this exclusion lies in the syntactic relations of 
dependence into which Spanish words enter when combined with each other, 
affecting their prosody in such a way that atonic words do not receive stress in 
most contexts. Hualde states that these words, “like all other words, have a syl-
lable that is designated as head of the word” (Hualde 2009: 202). Monosyllabic 
words, whether tonic or atonic, only have one syllable, which functions as its 
prosodic head; in simple polysyllabic words, tonic as well as atonic, only one 

21 More generally, I would suggest that the rightmost (strong) positions of all short unified 
lines in Spanish have to be realized by maxima. The same rule applies to the rightmost (strong) 
positions of the cola that make up long unified lines and to the rightmost (strong) positions of 
the hemistichs making up divided lines (cf. Domínguez Caparrós 2014: 49).
22 This kind of realization at line end is permitted in languages such as English (cf. Attridge 
1982: 166–167; Tarlinskaja 1987: 66, 80, 178, 181 ff.). Notice that, if the basis for the metrical 
analysis of Laberinto were the performance, it would not be possible to find a reason for the 
absence of nonheads in colon-final strong positions: if the mere addition of postlexical stresses 
sufficed to render nonstresses of tonic polysyllables fit for these positions, their failure to appear 
there would be impossible to explain.



63Juan de Mena’s arte mayor verse in Laberinto de fortuna ...

of the syllables can be the head. In tonic words, the head corresponds to a 
stress, but in atonic words the head usually corresponds to a nonstress: it only 
“receives stress when the [atonic] word is cited, nominalized or focalized”, to 
name but a few suitable contexts (202).

In this article, it will be assumed that every syntactic word and every half of 
a compound have one head, which in speech may correspond to a stress or not. 
It will be further assumed that in Laberinto only some heads can be maxima, 
hence not all heads can realize colon-final strong positions. For the arte mayor 
of Laberinto, I would suggest that every head of a tonic word qualifies as a 
maximum, and that so does every head of an atonic word not immediately fol-
lowed by the head of a tonic word within the same line (in divided lines, within 
the same half-line). This is one of the reasons why example (6), repeated here 
for the sake of convenience, cannot be scanned as 6#4, but only as 4#6 (from 
now on, the signs X and x will indicate maxima and nonmaxima):

(6’)  introduzido por público pro (l. 1844)
 *x   x  x X  x    x #  X  x  x    X
   x   x  x X#x    x     X  x  x    X

For the same reason, the following construct cannot be a Laberinto line of the 
5#5 kind, as no amount of stress manipulation can make the atonic conjunc-
tion que = that realize the strong position at the end of the initial colon:

(d) *fenómenos que  no comprenderemos
    x  X   x  x      x #  X   x        x     x X  {x}

The explanation is not the unusual placement of the adverb no = not in a weak 
position – a phenomenon that Spanish metres permit, as we have seen – but 
the fact that (the head of ) que is somehow overshadowed by the (head of ) the 
tonic word that follows it: unlike no, the monossylable que is not a maximum.23

23 My definition of maxima takes into account all primary stresses (stresses corresponding to 
the heads of words, including the head of each part of compounds) and disregards all second-
ary stresses. That said, the proposed definition is still pretty generous. Notice, to begin with, 
that a compound like vilmente = vilely in l. 809 of Laberinto has two maxima: “Vimos en uno 
vilmente abrazados” (X x x X # x X X…). Also, a stressed maximum can be immediately followed 
by another one belonging to a different word in the same line, as in “lavandera que no llore de 
pena” (x x X x x X # X x x X {x}, a 10-position line of the a maiore, 6#4 type, in Góngora 1985: 
220). Likewise, an unstressed maximum can be immediately followed by another unstressed 
maximum: witness the realization of the position at the end of the initial colon of l. 1437 of 
Laberinto, scanned as either “tanto que los # que de allí pelea{van}” (X x X X #…, 4#5 or 4#6 



64 Jesús M. Saavedra Carballido

In the following 5#6 example the first head of the compound malamente =  
badly acts as a maximum, but the head of the preceding atonic word que does 
not:

(20) e los matemáticos que malamente (l. 1031)
    X X      x  x  X#x x      x     X x   X {x}

As in this line, usually there are more maxima than stresses. This happens even 
in some of the lines discussed before, including those in which colon-final 
positions are realized by (syllables of ) atonic words; all of those lines can be 
rescanned accordingly (extrametrical syllables are never heads so they cannot 
be maxima):

(7’)  así como príncipe legislator (l. 643)
  x X X   x    X #x  x  x x  x X

(8’)  en Ras con aquel señor de Charní (l. 1590)
  x     X   X   x   X # x  X    X      x   X

(9’)  aquel coraçón que si no querer (l. 894)
  x    X   x  x  X  # X   x  X     x  X

(10’)  oras silvando como dragón (l. 1965)
     X x    x  X # x   X   x     x  X

(12’)  en las baxas artes que le da Minerva (l. 474)
    X    x    X  x  X # x     X   x  X    x  X {x}

type) or “tanto que los que # de allí pelea{van}” (X x X X X #…, 5#5 or 5#6 type). Further, an 
unstressed maximum can be preceded by a stressed maximum, as in “Allí vi de # Pigmalión el 
herma{na}” (a 4#6 arte mayor line, in Santillana 2003: 346). Notice also that, if we admitted that 
Laberinto contains divided lines, construct (d) would create no difficulty: x X x x X + X x x x 
X {x} (the + sign indicates an epic caesura). Finally, notice that my account does not make the 
presence of maxima depend on punctuation. To be sure, punctuation can be a clue to prosody 
and it often reminds us of such important yet often overlooked prosodic phenomena as the 
preservation of stress in atonic words placed at the end of intonational phrases (e.g. before a 
parenthetical break). However, punctuation depends on writing conventions and on editorial 
practices, which, even within the same language, vary from period to period, from area to area, 
and sometimes from person to person. This is enough reason to ignore it in analysis.



65Juan de Mena’s arte mayor verse in Laberinto de fortuna ...

(14’)  e lo que queremos menos acabamos (l. 232)
    X X    X      x X # x      X  x  x  x  X {x}

(15’)  e sobre partir tales discordanças (l. 2034)
     X  X x   x   X #X x    x   x   X  {x}

(17’)  el ánima contra la falsa fortuna (l. 1554)
     x  X  x   x   X # x  x  X  x  x   X{x}

(18’)  Ca un condestable armado, que sobre (l. 2113)
                                    …   X  x     X   X {x}

(19’)  a fijos de los que libró del desierto (l. 1244)
    x  X x    X X   # X  x   X   …
    x  X x    X X      X#x   X   …

Signalling maxima makes it easier to appreciate why, in contrast to (d) above, 
the following construct is correct (here the atonic que is a maximum):

(e) fenómenos que jamás comprendimos
  x  X   x  x      X #x   X    x        x    X {x} (5#5)

The syllable count, the realization of the colon-final strong positions, and the 
syllabic relation between cola – all of these elements comply with the rules 
given so far.

Midline metrical boundary (caesura)

Now that the syllable count and the realization of strong and weak positions 
in Laberinto are clearer, it is necessary to elucidate the relation of linguistic 
boundaries to the template.

When confronted with lines whose rightmost word is atonic, such as (18’), 
the notion of the maximum and the acceptance of unstressed maxima make 
it unnecessary to assume, as Piera does (1981: 211–216), that line ends in the 
domain of metre necessarily translate into phrase boundaries in the realiza-
tion. What makes enjambment possible is the fact that, contrary to Piera’s 
argument, metre does not automatically introduce a systematic “line–phrase 
correspondence” at the end of all lines (Piera 1981: 216; cf. Piera 2003a); and 



66 Jesús M. Saavedra Carballido

neither is there, of course, any line-internal colon–phrase correspondence, as 
he further claims (Piera 2003b). Unless we are willing to violate the givens of 
the language, we must recognize that the lowest line boundary in Laberinto is 
above the syllable but below the phrase: it operates between syntactic words, 
whether tonic or atonic. Examples (17’) and (19’) suggest that the same hap-
pens with colon boundaries inside lines.

In view of the lines scanned so far, we can therefore add one more element 
to our account of arte mayor: between the syllables realising the second and 
third strong positions, as between lines, there must be some kind of linguistic 
boundary. The existence of a metrical rule explicitly regulating such bounda-
ries is rejected by Piera, at least inside lines (Piera 1981: 119–121). However, 
the examples analysed indicate that the syllable realizing the last strong posi-
tion of the initial colon and the syllable realizing the first strong position of 
the final colon cannot belong to the same syntactic word.

Due to the exclusion of 8-syllable lines, Laberinto does not present lines 
in which the first strong position of the final colon is unequivocally realized 
by the last nonstress of a proparoxytone (as in this 4#4 construct: *“Muchos 
fenó#menos | extra{ños}”, scanned as X x x X # x x | x  X {x}). Neither does 
Mena’s work contain lines in which the last word of the initial colon presents 
more than two nonstresses at the end (as in this 5#6 construct: *“estaban 
lleván#doselos | regala{dos}”, scanned as x X x x X # x x x | x x X {x}). The for-
mer absence might be just a stylistic preference of Mena’s, and the latter might 
be due to the scarcity of such words in the language. Yet the fact is that in lines 
without contiguous vowels most straddling proparoxytones end in 6-syllable 
cola (see Section 4a of the Addenda); thus they necessarily fall just short of 
realizing a strong position in the final colon, as in this example:

(7’’) así como príncipe legislator (l. 643)
  x X X    x    X #x  x  x| x  x X

In lines with contiguous vowels the situation is similar: the colon in which 
straddling proparoxytones end always allows – and frequently demands – 
a 6-syllable interpretation. Finally, while in other Spanish verse compounds 
can straddle two cola and two lines,24 Laberinto contains no such cases 

24 The two stressed maxima in a compound can appear in different lines and in different 
cola within the same line. The possibility has been exploited, for example, by Luis de León 
and Luis de Góngora. As regards the former, “Y mientras miserable- / mente se están los otros 
abrasan{do}”, are a 6-position line and a 10-position line straddled by the 2-maximum adverb 
miserablemente = wretchedly, whose first maximum realizes the rightmost strong position of 



67Juan de Mena’s arte mayor verse in Laberinto de fortuna ...

(cf. Duffell 1999: 71–72). All things considered, I tentatively include a com-
pulsory midline boundary rule, of the kind explained, among the realization 
rules of arte mayor.

Beyond metrical correctness?

The characteristics of Mena’s arte mayor in Laberinto that I have identified as 
constant and necessary are the following: two rising cola, each with two strong 
positions; extrametricality aside, each colon is realized by four to six syllables; as 
in other dolnik, every strong position is realized by one syllable and every inter-
nal weak position by one or two; colon-final strong positions must be realized by 
maxima, whether stressed or not; both cola are kept distinct by a midline bound-
ary rule, such that the syllables realizing the last strong position of the initial 
colon and the first strong position of the final colon cannot belong to the same 
syntactic word, in spite of which the lines can be said to be unified; unusually for 
Spanish metres, headlessness is permitted (but only as long as the initial colon 
is realized by at least four syllables); the final colon cannot be shorter than the 
initial one, so it must be realized by at least five syllables; extrametricality only 
appears at line end, and it is confined to one nonmaximum, so proparoxytonic 
line ends and extrametrical atonic words are not tolerated.

In Kerkhof ’s edition of Laberinto there are only two lines without contigu-
ous vowels that fail to comply with these rules:

(21) *par en el ánimo, non en la fortuna (l. 1563)
        X  X   x  X#x  x     X   X  X  x   X {x}
        X  X   x  X  x  x     X #X  X  x   X {x}

(22) *ramo ninguno non   avrá menester (l. 224)
      X   x   x    X#x   X      x   X    x  x   X
      X   x   x    X  x   X #   x   X    x  x   X

Line 1563 is of the hypothetical 4#7 or 7#4 type. In either case, it is incor-
rect because there can be no 7-syllable cola. Line 224 also has two possible 
scansions: as either 4#7 or 6#5. The former is incorrect for the same reason. 

the first line (León 2011: 74). As regards the latter, “y al fin ambos igual#mente ayuda{dos}” is a 
10-position line (of the a maiore, 6#4 type) in which the 2-maxima adverb igualmente = equally, 
whose first maximum realizes the rightmost strong position of the initial colon, straddles both 
cola (Góngora 1985: 53).



68 Jesús M. Saavedra Carballido

According to the latter scansion, the division into cola goes against the phrasal 
movement of the line. Nevertheless, this does not make the scansion incor-
rect (we have already seen that the only boundary required between cola is 
placed between two syntactic words, not necessarily tonic or phrase-final). It is 
incorrect simply because the final colon cannot be shorter than the initial one.

Both lines are correct if we accept the occurrence of internal extrametrical-
ity (of two and one syllables, respectively); this means accepting the occasional 
combination in Laberinto of unified lines (the norm) and divided lines (with 
possible internal extrametricality, i.e. made up of different hemistichs, here sep-
arated by the + sign that indicates epic caesura): “par en el á{nimo}, + non en 
la fortu{na}” (of the 4+5 line type) and “ramo ningu{no} + non avrá menester” 
(4+6). It could be argued that, if such combinations were possible, it would be 
strange for Mena to employ them so rarely. However, there is also line 939 – 
example (1) – of a type which is equally rare (6#6). The difference is that this 
line does not present any textual alternatives turning it into something more 
usual, whereas lines 1563 and 224 can be amended into metrical correctness 
with a different selection of manuscript variants. Thus we can get, for example, 
the lines “par en el á#nimo, non la fortu{na}” (4#6) or “par en el á#nimo, no en 
la fortu{na}” (4#6), and “ramo ningu#no no avrá menester” (4#6).25

Some lines with contiguous vowels are also dubious. The correctness of the 
following example depends on treating fue as a disyllable; all other scansions are 
incorrect (the o indicates a vowel that does not count metrically inside the line):26

(23) desque Juba  les   fue   prepotente (l. 400)
      X     x  X  x   X   #xX       x  x X {x} (5#5)
 *    X     x  X#x   X     xX      x  x X {x} (3#7)
 *    X     x  X#x   X     oX      x  x X {x} (3#6)
 *    X     x  X  x   X     oX #   x  x X {x} (6#3)
 *    X     x  X  x   X     xX #   x  x X {x} (7#3)

25 The two lines would also be metrical with other variants. The point here is not so much to 
choose the right version from the point of view of the manuscript tradition as to show that the 
scribes themselves may also have had doubts about the correctness of the readings eventually 
chosen by Kerkhof, opting for something more regular instead.
26 In the case of contiguous vowels, the ones marked as metrically invisible are the higher  
(cf. Navarro Tomás 1967: 27, 66–73, 156, 161 ff.; but cf. 151); in the case of vowels that are 
identical in this respect, I mark as invisible the weaker ones (unstressed as opposed to stressed); 
finally, in the case of identical vowels in this respect too (stressed or unstressed), I mark as 
invisible the ones to the left. Invisible vowels do not count for the determination of maxima; in 
other words, maxima are determined after marking invisible vowels, as if they were not there.



69Juan de Mena’s arte mayor verse in Laberinto de fortuna ...

An alternative version, present in some sources and whose metrical correct-
ness is beyond doubt, reads: “desde que Ju#ba les fue prepoten{te}” (4#6).

This other line is incorrect unless we interpret quieres as a trisyllable:

(24) de los  passados,   si quieres ver espanto (l. 483)
     X X      x   X  x       x # x Xx     X  x   X {x} (6#6)
 *   X X      x   X#x       X    o Xx    X  x   X {x} (3#7)
 *   X X      x   X{x} +  X    o Xx    X  x   X {x} (4+6)

The third scansion shows a divided line with internal extrametricality. 
Examples (21) and (22) above might be similarly divided; the addition of 
(24) makes three lines out of 2376 in Laberinto. However, with different vari-
ants we can arrive at a correct unified version of this line too; this is only one 
of the existing possibilities: “de los passa#dos, si quiés ver espan{to}” (4#6).

The following line is equally problematic, now due to the fact that the initial 
colon is perhaps realized by three syllables and to the relative length of the cola:

(25) *Femonoe,  por   orden la sesta (l. 969)
        x   x  X#x   x      X    x   x   X {x} (3#6)
        x   x  Xo#   x      X    x   x   X {x} (3#5)
        x   x  Xx     x      X # x   x   X {x} (6#3)
        x   x  Xo     x      X # x   x   X {x} (5#3)

With other variants the line would read, for example, “Femona por or#den era 
la ses{ta}” (5#5), whose correctness is beyond dispute.

This line is divided or else ends in a 7-syllable colon:

(26) *de los que demuestran    y de los demostrados (l. 1028)
               … +{x}   X   X  X    x   x    X {x}
                              … #  x    X   X  X    x   x    X {x}

Some manuscripts read “de los que demues#tran e los demostra{dos}”, with a 
correct 6-syllable colon at the end.

In the following case, the final colon remains correct, in terms of length, 
only if the two vowels separated by the letter h are treated as one (which proves 
that this letter is no obstacle to doing so):

(27) que fe non se guarden hermanos a hermanos (l. 2031)
              … #x     x    X  x    x  o     X {x}



70 Jesús M. Saavedra Carballido

Though Piera’s amphibrachic account does not find any incorrect lines in 
Laberinto (1981: 144), one of the lines that I find problematic – example (22) – 
is also considered incorrect in Piera’s second account (2008: 125). Other lines 
that he deems incorrect can be considered well-formed examples of the 5#6 
type: line 791 (“que de lo gana#do sufre mengua tan{ta}”, scanned as X X X x 
X # x X x X o x X {x}) and line 354 (“de parte del Aus#tro vimos toda Gre{cia}”, 
scanned as x X x x X o # x X x X x X {? x}, where the ? mark indicates that the 
metrical status of a vowel is unclear).

As regards the constructs that Piera presents as possibly correct (Piera 
2008: 123), one is obviously incorrect by my rules:

(f ) *do veredes los grandes   señores e los reyes (based on l. 1779)
    X   x X  x   x      X # x      x  X x  X  x    X{? x} (6#7)
    X   x X  x   x      X {x} +   x  X x  X  x    X{? x} (6+6)

Another construct that Piera provisionally allows I find incorrect, at least if 
we accept his disyllabic interpretation of the first word (by virtue of which the 
line begins with a 3-syllable colon):

(g) vieras   otro d’essa nación (based on l. 558)
 *oXx     X #x …

This line is correct if we treat vieras as a trisyllable; with the scansion x X x X 
# the line is of the 4#5 or 4#6 type.

In this other construct that Piera judges problematic but possibly correct, 
it is the final colon that causes difficulties:

(h) oras silvando tanto drago (based on l. 1965, quoted as example (10’) above)
 X x    x  X # x  X   x     X {x}

This example belongs to the 4#4 type, which is incorrect only if we take arte 
mayor – qua long unified dolnik with four strong positions – to exclude the 
4-syllable realization of the final colon. It is precisely to this and other related 
issues that I turn now.



71Juan de Mena’s arte mayor verse in Laberinto de fortuna ...

Other constant but not necessary characteristics of Laberinto

One major problem with the dolnik account of arte mayor offered so far is that 
in Laberinto all final cola are realized by at least five syllables. The absence of 
final cola featuring four syllables means that Mena shunned 8-syllable lines. 
This peculiarity lends weight to the amphibrachic hypothesis. In Piera’s ver-
sion of it (1981), 1-syllable extrametricality is acceptable at the beginning of 
lines; both cola can be headless, and the line can be truncated; between the 
syllables realizing two strong positions within the same colon there are always 
two syllables. The corresponding template can be represented as:

{w} (w) s w | w s w # (w) s w | w s (w)27

One possible charge against the amphibrachic account is that 4-syllable final 
cola (with one syllable alone between the syllables realizing the strong posi-
tions), and hence 8-syllable lines, do appear in other authors’ arte mayor.

In his first approach to Laberinto, Piera notes that some nonamphibrachic 
“descriptions predict eight-syllable lines, whose conspicuous absence from 
all major poems must then be explained away” (Piera 1981: 102). Even if in 
Mena’s composition final cola are always realized by five or more syllables, this 
constant characteristic may not be necessary for other kinds of arte mayor. 
The best way to determine whether arte mayor admits of 4#4 lines is to find 
them in compositions using what appears to be the same metre. Actually, this 
line type is not unheard of. This is an instance by Imperial (Dutton, González 
Cuenca eds. 1993: 261):

(28) Tan grant amor nunca mostraron
     X      X    x   X  #X    x     x    X{x}28

27 According to Duffell (1999: 68), the third foot cannot be headless, while the second can be 
truncated: {w} (w) s w | w s (w) # w s w | w s (w).
28 The initial colon is oxytonic, and, despite all appearances, the internal weak position of the 
final colon cannot be filled by two syllables, as this would mean positing two consecutive strong 
positions (one at the end of the initial colon and another at the beginning of the final one, as 
in some divided lines). The paroxytonic disyllable nunca contains a maximum followed by a 
nonmaximum realizing a strong position, both within the same foot; in this respect, this line 
is similar to several others of Laberinto already mentioned: l. 284, scanned as example (13);  
l. 232, example (14’).



72 Jesús M. Saavedra Carballido

In addition to Imperial, there are other authors that accept 8-syllable lines.  
A case in point is the 20th-century writer Valle-Inclán, arguably one of a hand-
ful of versifiers that kept the Spanish arte mayor tradition alive. In his “Los 
pobres de Dios” = “God’s Poor” (Valle-Inclán 1995: 67), written in what seems 
to be an 8-position dolnik, we find 5#6 and 4#5 lines with a strong resemblance 
to those of in Laberinto:

(29) Va la   caravana   de los desvalidos
      X X    x x X#x     X  X     x   x X {x}

(30) El polvo quema  sus llagas rojas
     x   X   x    X # x   x     X  x   X{x}

However, there are also 4#4 lines:

(31) Por los  caminos  florecidos
     X   X     x   X #x     x  x X {x}

(32) En donde  cantan  las vaqueras
    X    X    x   X #x     X    x   X {x}

All four examples, like the other lines in the composition from which they 
are taken, display the word boundary between cola mentioned in connection 
with Laberinto. Indeed, this is one of the reasons to conclude that Valle-Inclán 
“Los pobres de Dios” is, like Mena’s work, an example of arte mayor. Therefore 
Piera’s construct (h), scanned above, can be tentatively labelled as a correct 
4#4 line.29

29 Apart from a dubious case with contiguous vowels like “a Rute e Prie#go e a Carcabuey” 
(l. 2036), which can be scanned in several other ways (e.g. as 4#5), the 4#4 line type is absent 
from Laberinto. By contrast, Mena uses 6#6 lines, which are avoided by Valle-Inclán. It could be 
hypothesized that, while Mena wants to keep its distance from popular arte menor by avoiding 
initial cola of three syllables and final cola of four (cf. Piera 2008: 123), Valle-Inclán does aim 
at a popular flavour. In this regard, we cannot forget that Valle-Inclán was of Galician origin, 
and that “Los pobres de Dios” is closed by a stanza made up of four isosyllabic 7-position lines 
(romance), borrowed from a well-known folk song in the Galician-Portuguese language. Among 
the possible origins of arte mayor, the influence of Galician-Portuguese versification on medi-
aeval Spanish verse has been mentioned since Santillana’s Proemio = Preface (Santillana 2003: 
653). This explanation has been accepted by many modern scholars (e.g. Le Gentil 1953: 409; 
Clarke 1964: 53, 219; Lázaro Carreter 1972: 344–345; Navarro Tomás 1974: 121–122; Gasparov 
1996: 138; García Calvo 2006: 1650–1651; Domínguez Caparrós 2014: 166). While Spanish 



73Juan de Mena’s arte mayor verse in Laberinto de fortuna ...

If final cola realized by four syllables are accepted, then arte mayor is less 
likely to be an amphibrachic form than a dolnik. Further confirmation of this 
view would come in the form of headless lines with initial cola realized by 
three syllables (corresponding to the metrical sequence s w s #). Because of the 
presence of contiguous vowels, on the one hand, and of the existence of manu-
script variants, on the other, there are no certain instances of this phenomenon 
in Laberinto;30 but two possible cases can be found in “Son de muñeira” = “The 
Sound of the Mill-Song”, another composition by Valle-Inclán (1995: 89).31 
In my scansion I have assumed that contiguous vowels in the final colon do 
not count:

(33) ¡Toc! ¡Toc! ¡Toc!… Bate la espadela
       X       X       X   #     X x X o    x  X{x} (3#5)

(34) ¡Toc! ¡Toc! ¡Toc!… Da vueltas la muela
       X       X       X   #      X   oX x    x    o X{x} (3#5)

The truth, however, is that both lines can be scanned differently: the initial cola 
can be continued up to the fourth maximum, and there is the possibility of 
counting a greater number of contiguous vowels. For this reason, I provision-
ally dismiss initial cola with fewer than four syllables, here and in Laberinto.

As regards extrametricality at the end of lines, in Laberinto it is always 
limited to one syllable. However, in Spanish this limitation is not usually 

arte mayor was abandoned when the Italianate 10-position line was introduced for learned 
verse, in Galician-Portuguese it remained (under the name of verso de moinheira = mill-song 
verse) a central form both in the oral tradition and in the written tradition of folk inspiration, 
as evidenced by the work of the 19th-century Galician writer Rosalia de Castro. In the hands 
of Galician authors, even of those writing in Spanish, verso de moinheira has never been con-
strained by the learned connotations that arte mayor had acquired in mediaeval Spanish, and 
thus remained free from some of the stylistic limitations that Mena imposed on his Laberinto. 
Despite their different minimum and maximum line lengths, the identity between the verso de 
moinheira and arte mayor has been affirmed by Henríquez Ureña (2003: 205–207). Proof that 
the absence of certain line types does not necessarily point to a different metre comes from 
Santillana’s “Comedieta de Ponça” = “Little Comedy of Ponça” (Santillana 2003: 295–356), whose 
verse is clearly the same as Mena’s – 5#6 lines predominate, appearing side by side with a few 
4#5, 4#6 and 6#6 lines – despite lacking 5#5 lines without contiguous vowels.
30 To be sure, a Laberinto line like “O virtuo#sa, magnífica gue{rra}” (l. 2019) can be scanned 
as 3#6, but also as 4#6 (the latter is the option preferred by editors, who indicate this reading 
by means of diaeresis: virtüoso = virtuous).
31 Despite its title, Valle-Inclán’s composition was not for singing but for recitation.



74 Jesús M. Saavedra Carballido

considered to distinguish among verse types (which are generally allowed 
to mix oxytonic, paroxytonic and proparoxytonic endings), so we cannot be 
certain that this is a necessary requirement for all arte mayor. After all, the 
insertion of two extrametrical syllables at line end is a permissible variation 
in authors other than Mena. Witness this example by Villasandino (Dutton, 
González Cuenca eds. 1993: 162):

(35) Filósofo firme e grant metafísico
  […]
  non entendades que só tan çentífico

Like other lines in the same composition (analysed as arte mayor in Saavedra 
Molina 1945: 116, and Piera 1981: 132), these two have the ending X {x x}, 
absent from Laberinto but included among the accepted elements of arte 
mayor by such scholars as Foulché-Delbosc (1902: 102–103), Navarro Tomás 
(1974: 115–116) and Saavedra Molina (1945: 27, 30). Despite its exclusion 
from Mena’s composition, double extrametricality can be provisionally permit-
ted as an option in arte mayor (cf. Piera 1981: 100, 131–134).32

After this comparison between Laberinto and works by other authors, it 
appears that the necessary characteristics of arte mayor may be fewer than 
those that are constant in Mena’s composition. In this light, the latter comes 
over as a relatively strict version of the metre.33

32 Further doubts about the demand to avoid double extrametricality in arte mayor are cast by 
an isosyllabic 10-position line like “que tienen fama en este mundo va{rio}”, by the late-16th-
century writer Bartolomé Carrasco de Figueroa. In the composition to which it belongs, it 
appears surrounded by proparoxytonic lines alone, and is therefore supposed to have two extra-
metrical syllables rather than one (cf. Alatorre 2007: 194). Theoretically, the same interpretation 
could be given to similar endings with contiguous vowels in Laberinto, such as “privados de 
to#da visiva poten{çia}” (l. 148, of the 5#6 type).
33 If we compare the constant characteristics of the Laberinto verse with Tarlinskaja’s descrip-
tion of the English dolnik, a couple of interesting differences emerge. Tarlinskaja speaks of void 
weak positions, most often the first one, but also internally (cf. Tarlinskaja 1993: 12–13); but in 
Laberinto all internal weak positions are realized by at least one syllable. The other difference 
is that the only constant feature that she contemplates in the realization of the metre is that 
the internal stresses of polysyllables are avoided in weak positions (the distribution of other 
stresses and nonstresses is fairly predictable, but always with exceptions), while in Laberinto 
the constant feature is that colon-final strong positions are always realized by maxima.



75Juan de Mena’s arte mayor verse in Laberinto de fortuna ...

Possible characteristics of other arte mayor

The use of final 4-syllable cola might not be Imperial’s only departure from the 
arte mayor standard subsequently set by Mena. Imperial also has initial 3-syl-
lable cola and final 7-syllable cola, and in some of his lines the initial colon is 
the longer. These phenomena indicate that his metrical rules might be looser 
than those followed by Mena and by other authors (Dutton, González Cuenca 
eds. 1993: 312): Imperial might have admitted both headless lines with 3-syl-
lable cola (e.g. “son donze#llas de grant exçelen{çia}, a line of the 3#6 type) 
and the interspersion of lines divided by epic caesura (e.g. “la quinta lla{man} 
+ Conjurado Sermón”, of the 4+6 type); also, he might have combined his 
cola quite freely (e.g. “e vence voluntat # desenfrena{da}”, of the 6#4 type).34 
By these rules – which may still result in arte mayor, but which certainly do 
not coincide with those of its mainstream variety, i.e. of arte mayor qua long 
unified dolnik with four strong positions, let alone with those of the stricter 
mainstream represented by Laberinto – all of the lines interpreted as incorrect 
in the preceding sections would become automatically correct.

Conclusion

The anisosyllabic metre of Mena’s Laberinto has often been considered the most 
perfect example of Spanish arte mayor. I have analysed the verse employed by 
Mena – and by other authors, from the Middle Ages to the 20th-century – as 
a kind of syllable-stress dolnik with a predominant ternary rhythm.

My claim that arte mayor is a syllable-stress form is based on the following 
considerations. If a metre that only regulates syllables (or metrical elements 
based on syllables) is syllabic, and a metre that only regulates stresses (or met-
rical elements based on stress) is accentual, then it is reasonable to interpret 
as syllabic-accentual any metre that regulates (metrified) syllables and (met-
rified) stresses in some way. Crucially, Spanish metres usually regulate both 
the syllable count and the minimal number and placement of stresses along 
the line: all colon- and line-final strong positions must be realized by stresses. 
The latter rule applies to the immense majority of Spanish metres; so much 
so that traditional Spanish metrists impose it even on places where it does not 

34 Some of Imperial’s lines are ambiguous; e.g. “La del semblan{te} + nin alegre nin tris{te}” 
(of the 4+6 type), which can also be interpreted as “La del semblante nin # alegre nin tris{te}” 
(of the 6#5 type).



76 Jesús M. Saavedra Carballido

apply (hence their prescription of stress wrenching in the middle of cola and 
lines). In relation to Spanish, then, Duffell is eminently correct when he says 
that, in general, “The patterns of Romance [… metrical] verse are based on 
two linguistic features, syllable count and accentuation. No Romance verse 
is purely syllabic”, as the “metres contain mandatory positions for accented 
syllables” (1996: 211).

From the perspective of metrical traditions where stresses are regulated in 
a different way, it might seem that Spanish metres are too irregular, as certain 
stretches within the lines seem to follow no discernible organization; but the 
fact remains that they are as strictly (even if differently) regulated as many oth-
ers – e.g. English ones, whose general syllable-stress character no one seriously 
denies. In this respect, one could say that Spanish metres simply obey the rule 
that correspondence to metrical template tends to be lax at the beginnings 
of units and strict at the ends (cf. Hayes 1983: 373). In Spanish metres, these 
(colon- and line-) ends can be placed quite far apart (e.g. in strict a minore 
10-position lines in which only the 4th and 10th positions are always filled 
by a stress, while other stresses, despite following certain statistical tenden-
cies, are optional), or as close together as every four syllables (e.g. in strict 
7-position lines in which the 3th and 7th positions must be filled by a stress, 
while other stresses are optional). At most we could say that in the stretches 
preceding those ends, Spanish metres may display syllabic tendencies (though 
in many metres stresses are statistically regulated even at the beginning of cola 
and lines). But, in general, the issue is qualitative rather than quantitative: 
in a language like Spanish in which stress is not fixed or always predictable, 
it is not really possible to tell the length of a metrical line just by counting 
the (metrified) syllables. At least one (metrified) stress must be taken into 
account. Take a simple case in point: a 7-syllable line can realize a 7-position 
template without final extrametricality (if the line ends …X), a 6-position 
template (if the line ends …X {x}), a 5-position template (if the line ends …X  
{x x}), a 4-position template (e.g. the construct “adelantán{dosenos}”: x x x 
X {x x x}), and, theoretically, even a 3-position template (e.g. the construct 
“acabán{dosetelos}”: x x X {x x x x}). If verse like this cannot be identified 
without paying attention to syllables and stresses, then it surely belongs to 
the syllable-stress kind.

As regards the dolnik hypothesis, it consists in assuming strong positions 
to be realized by one syllable and separated by another two or one from the 
syllable realizing the nearest strong position. Arguing that the realization rules 
of Mena’s arte mayor do not count prosodic syllables proper but metrically 
relevant segments (metrified syllables) and that they do not regulate directly 
the distribution of stresses and nonstresses but of (usually stressed) maxima 



77Juan de Mena’s arte mayor verse in Laberinto de fortuna ...

and (unstressed) nonmaxima (metrified stresses and nonstresses), this article 
begun by analysing the constant and necessary characteristics of the verse of 
Laberinto: two rising cola, each with two strong positions; extrametricality 
aside, each colon is realized by four to six syllables; as in other dolnik, every 
strong position is realized by one syllable and every internal weak position by 
one or two; colon-final strong positions must be realized by maxima, whether 
stressed or not; both cola are kept distinct by a midline boundary rule, such 
that the syllables realizing the last strong position of the initial colon and 
the first strong position of the final colon cannot belong to the same syntac-
tic word, in spite of which the lines can be said to be unified; unusually for 
Spanish metres, headlessness is permitted (but only as long as the initial colon 
is realized by at least four syllables); the final colon cannot be shorter than the 
initial one, so it must be realized by at least five syllables; extrametricality only 
appears at line end, and it is confined to one nonmaximum, so proparoxytonic 
line ends and extrametrical atonic words are not tolerated.

I have tried to explain as many arte mayor features as possible – some 
admittedly rare, though shared by other Spanish metres – as correct and as 
necessary. Hence, to begin with, the attempt to explain the arte mayor of 
Laberinto as a dolnik rather than as a ternary metre allowing headlessness, a 
decision intended to accommodate lines whose final colon is realized by five 
syllables instead of six, and, in addition, the substitution of stress maxima 
for stresses, a move designed to account for line-end atonic words. It is clear, 
however, that the metre of Laberinto, like other Spanish metres, regulates the 
line length – whose variation it keeps within precise limits – and the placement 
of maxima at colon and line end more strictly than the placement of other 
maxima elsewhere in the line. As a result, between maxima – as against the 
syllables realizing strong positions – there can be less than one or more than 
two syllables. If it can be stated that – to make room for all the vocabulary in 
the language and to avoid monotony – syllable-stress dolnik in English can dis-
play accentual tendencies, when weak positions are left unrealized (Tarlinskaja 
1993: 13, 199; Duffell 2008: 34; Attridge 2013: 107–108, 155 n. 11), so, con-
versely, it is tempting to suggest that Spanish dolnik tends towards syllabism, 
when the distribution of maxima and nonmaxima does not reveal but obscures 
(outside colon and line ends) the distribution of strong and weak positions; 
but, given its strong preference for a ternary rhythm, it can be concluded that 
Laberinto is not characterized by such syllabic tendencies.

In relation to rhythm, it is clear that in Spanish 1- and 2-syllable intervals 
cannot be the only defining characteristic of dolnik proper: one can write verse 
with a rhythm that mixes iambs and anapaests while sticking to isosyllab-
ism (e.g. romance, a strict sequence of 7-position lines in which each position 



78 Jesús M. Saavedra Carballido

amounts to one syllable, and in which it does not usually matter if the rhythm 
of some lines is binary, of some ternary and of others mixed). Some authors 
have spoken of dolnik when dealing with Spanish compositions made up of 
isosyllabic lines displaying mixed rhythms only (e.g. Gasparov 1996: 136). To be 
sure, this can be done when it can be demonstrated that such lines evolve from 
anisosyllabic dolnik, but this latter, original variety should never be forgotten.

Not all the constant features of Laberinto are necessary in other kinds of 
arte mayor. As it turns out, Mena’s verse is a singularly strict variety of main-
stream arte mayor (defined as long unified dolnik with four strong positions). 
In other mainstream arte mayor, the final colon can be realized by four syl-
lables (though still not shorter than the initial one), and the line end can 
be proparoxytonic. Lines with these characteristics do not appear in Mena’s 
work, or else they are so scarce as to be better viewed as incorrect; but if they 
appeared, or if they were more frequent, it would not be reasonable to say that 
they result in a substantially different metre.

I have provisionally argued for the existence of yet other kinds of arte 
mayor. These seem to have only two necessary characteristics: four strong posi-
tions (the second and fourth colon-final and corresponding to maxima) and a 
midline boundary between syntactic words. The possibilities accepted here but 
not in mainstream arte mayor would be: headlessness combined with initial 
cola of fewer than four syllables, and the existence of a looser variety and of 
two stricter varieties. In the loosest version arte mayor becomes a taktovik and/
or a divided 2-hemistich line (with internal extrametricality, i.e. epic caesura), 
and it allows a final colon that is shorter than the initial one.35 As regards the 
versions that are stricter than the mainstream, one amounts to a ternary metre 
whose final colon is always realized by six syllables, the permitted combina-
tions of lines types being thus limited to 4#6, 5#6 and 6#6; the strictest version 
of arte mayor is not only rigorously ternary but isosyllabic: all lines are of the 
4#6 or 5#6 or 6#6 type.36

In order to achieve a full understanding of the rhythmic range of arte mayor  
(cf. Duffell 1999: 73–74), it is necessary to perform a complete statistical 
analysis of its variable characteristics, beginning with the ternary or binary 
movement that predominates (the latter is the norm, for example, in Valle-
Inclán’s “Los pobres de Dios”). While studying the characteristics of all the 

35 It would be logical to expect initial cola falling short of four syllables and the interchange-
ability of cola in certain varieties, other than Mena’s, of mainstream arte mayor; however, this 
does not seem be the case (the reasons need studying).
36 For example, Santillana’s composition beginning “Robadas havían el Austro e Borea” appar-
ently consists of 5#6 lines alone (Santillana 2003: 285–294).



79Juan de Mena’s arte mayor verse in Laberinto de fortuna ...

forms of Spanish arte mayor that might exist, it is nevertheless important to 
remember Clarke’s claim that there have been as many varieties of this verse 
as there are practitioners of it, and that “the rule” was “that each poet was a 
law unto himself in the matter of rhythm interpretation” (1964: 210). It might 
well be that arte mayor is less a single metre than a cluster of different forms 
with some family resemblance (cf. Clarke 1964: 92). Thus, it must be borne in 
mind that “arte mayor from the beginning enjoyed a loose form along with 
the restricted, which later gradually took precedence over its rival”, leading 
to “the monotony of rigid pattern” (60, 50). While some arte mayor texts 
seem to be genuine dolnik, others might be best described as strict ternary 
verse and even – e.g. when the placement of the colon boundary is rigorously 
fixed – as effectively amphibrachic; still others might eventually be identified 
as taktovik. This would mean applying Piera’s second account of arte mayor, 
if not to Laberinto, as he did, then to other texts; more generally, this would 
lead to accept the existence in Spanish, and therefore in Romance languages, 
of the taktovik form hitherto proved to exist only in languages like Russian, 
English and Latvian (cf. Gasparov 1996; Fabb, Halle 2008).37 As a result, the 
necessary characteristics of arte mayor might be reduced in number or even 
declared nonexistent, at which point the very label arte mayor, whatever its 
convenience as an umbrella term, would lose much of its analytic usefulness.

Meanwhile, the results obtained here warrant the provisional conclu-
sion that the dolnik form has been cultivated not only in Slavic, Baltic and 
Germanic languages, from Russian to English, but also in a Romance language 
like Spanish. In this respect, it cannot be forgotten that, before the emergence 
of arte mayor, a number of similar forms had already been cultivated in other 
Romance languages from Galician-Portuguese to Italian (cf. Burger 1957: 

37 Some authors – of whom Geers was perhaps the first (Geers 1930: 179–18; cf. Morley 1933: 
968), but certainly not the last (e.g. Pellen 1985: 10; Orduna 1987: 10–11, 23; García Calvo 2006: 
1617–1623) – have characterized the rhymed verse of the Spanish epic Cantar de Mio Cid = 
Song of My Cid (composed c.1200) as a long, two-part metre with four strong positions; in the 
nomenclature used here, it would correspond to either taktovik or loose anapaest. (Succinct 
descriptions of the form can be found in Gasparov 1996: 134–135; Domínguez Caparrós 2014: 
159–161.) The problem is that the text of the only extant manuscript does not seem to bear out 
these characterizations (even their proponents sometimes recognize the existence of many lines 
with the “wrong” number of stresses and/or syllables). This is perhaps because, as some argue 
(e.g. Fernández, Brío 2004), the work was meant to be performed to the accompaniment of (now 
lost) music. To solve these difficulties, other authors have suggested more strong positions: as 
many as six (though he speaks of “stresses”, this seems to be the view favoured by Smith 1979: 
52) or even eight (speaking of “beats”, Leonard 1931: 293). From today’s perspective, the Cid 
reads much like rhymed free verse.



80 Jesús M. Saavedra Carballido

40–43). And, just like arte mayor has coexisted with other Spanish metres 
that look much like its short counterparts,38 future research may uncover other 
Romance traditions in which dolnik metres are similarly varied. A complete 
comparative analysis of these forms, on the linguistic basis argued for here, 
is still pending.

References

Alatorre, Antonio 2007. Cuatro ensayos sobre arte poética. México D.F.: El Colegio 
de México.

Attridge, Derek 1982. The Rhythms of English Poetry. London: Longman.

Attridge, Derek 2012. The case for the English dolnik; or, how not to introduce 
prosody. In: Poetics Today 33(1), 1–26. https://doi.org/10.1215/03335372-1505522

Attridge, Derek 2013. Moving Words: Forms of English Poetry. Oxford: Oxford 
University Press.

Balaguer, Joaquín 1995. Apuntes para una historia prosódica de la métrica castellana. 
Santo Domingo: Editora Corripio.

Bello, Andrés 1884. Principios de la ortolojía i métrica de la lengua castellana. In: 
Consejo de Instrucción Pública (ed.), Obras completas de don Andrés Bello 5: 
Opúsculos gramaticales. Santiago de Chile: Pedro G. Ramírez, 3–229.

Burger, Michel 1957. Recherches sur la structure et l’origine des vers romans. Genève: 
Droz.

Chisholm, David 1995. Prosodic aspects of German hexameter verse. In: Poetics Today 
16(3), 523–545. https://doi.org/10.2307/1773177

Clarke, Dorothy Clotelle 1964. Morphology of Fifteenth-Century Castilian Verse. 
Pittsburgh: Duquesne University Press.

Correas, Gonzalo 1954. Arte de la lengua española castellana. Madrid: CSIC.

Díaz Rengifo, Juan 2012. Arte poética española. Kassel: Edition Reichenberger.

38 Let us just recall such anisosyllabic forms as the verso de seguidilla, with between four and 
six syllables up to the rightmost (metrified) stress, and the verso semilibre menor, with between 
three and six syllables (for the former, cf. Navarro Tomás 1974: 177–182, 240–241, 292–293, 
340; for the latter, cf. Navarro Tomás 1974: 524). Both can be argued to be underlain by the 
template w s w s {x}, without and with optional headlessness, respectively.



81Juan de Mena’s arte mayor verse in Laberinto de fortuna ...

D’Introno, Francesco; Teso, Enrique del; Weston, Rosemary 1995. Fonética y fonología 
actual del español. Madrid: Cátedra.

Domínguez Caparrós, José 2014. Métrica española. 3rd rev. ed. Madrid. UNED.

Duffell, Martin J. 1996. Chaucer, Gower, and the History of the Hendecasyllable. 
In: McCully, Christopher B.; Anderson, John J. (eds), English Historical Metrics. 
Cambridge: Cambridge University Press, 210–218.

Duffell, Martin J. 1999. Modern Metrical Theory and the Verso de Arte Mayor. London: 
Department of Hispanic Studies, Queen Mary and Westfield College, University 
of London.

Duffell, Martin J. 2008. Triple time in English verse. In: Rhythmica: Revista Española 
de Métrica Comparada 5–6, 23–48.

Dutton, Brian; González Cuenca, Joaquín (eds.) 1993. Cancionero de Juan Alfonso de 
Baena. Madrid: Visor.

Fabb, Nigel; Halle, Morris 2008. Meter in Poetry: A New Theory. Cambridge: Cambridge 
University Press.

Fernández, Guillermo; Brío, Clara del 2004. Sobre la métrica del Cantar de Mio Cid. 
Música y épica: La cantilación de las gestas. In: Lemir 8, 1–37.

Foulché-Delbosc, Raymond 1902. Étude sur le Laberinto de Juan de Mena. In: Revue 
hispanique: Recueil consacré à l’étude des langues, des littératures et de l’histoire des 
pays castillans, catalans, et portugais 9(29), 75–138.

García Calvo, Agustín 2006. Tratado de rítmica y prosodia y de métrica y versificación. 
Zamora: Lucina.

Gasparov, Mikhail L. 1996. A History of European Versification. Oxford: Oxford 
University Press.

Geers, Gerardus Johannes 1930. Algo sobre versificación española. In: Neophilologus 
15, 178–183. https://doi.org/10.1007/BF01510199

Góngora, Luis de 1985. Sonetos completos. Madrid: Castalia.

Halle, Morris; Keyser, Samuel Jay 1966. Chaucer and the study of prosody. In: College 
English 28(3), 187–219. https://doi.org/10.2307/374038

Halle, Morris; Keyser, Samuel Jay 1971. Illustration and defense of a theory of the iambic 
pentameter. In: College English 33(2), 154–176. https://doi.org/10.2307/374744

Hanson, Kristin 2008. Review of Fabb, Nigel; Halle, Morris 2008. In: Language 84(2), 
370–386. https://doi.org/10.1353/lan.0.0026



82 Jesús M. Saavedra Carballido

Harris, James W. 1983. Syllable Structure and Stress in Spanish: A Non-Linear Analysis. 
Cambridge and London: MIT Press.

Hayes, Bruce 1983. A grid-based theory of English meter. In: Linguistic Inquiry 14(3), 
357–393.

Henríquez Ureña, Pedro 2003. Estudios de versificación española. In: Obras completas 3: 
Estudios métricos. Santo Domingo: Secretaría de Estado de Cultura, Editora 
nacional, 163–440.

Hualde, José Ignacio 2009. Unstressed words in Spanish. In: Language Sciences 31(2/3), 
199–212. https://doi.org/10.1016/j.langsci.2008.12.003

Hualde, José Ignacio 2010. Secondary stress and stress clash in Spanish. In: Ortega-
Llebaria, Marta (ed.), Selected Proceedings of the 4th Conference on Laboratory 
Approaches to Spanish Phonology. Somerville: Cascadilla Proceedings Project, 11–19.

Hualde, José Ignacio 2014. Los sonidos del español. Cambridge: Cambridge University 
Press.

Hualde, José Ignacio; Nadeu, Marianna 2014. Rhetorical stress in Spanish. In: Hulst, 
Harry van der (ed.), Word Stress: Theoretical and Typological Issues. Cambridge 
and New York: Cambridge University Press, 228–252.

Jakobson, Roman 1966. Linguistics and poetics. In: Sebeok, Thomas A. (ed.), Style in 
Language. Cambridge: MIT Press, 350–377.

Ker, William Paton 1899. Analogies between English and Spanish verse (arte mayor). 
In: Transactions of the Philological Society 24(1), 113–128. 

 https://doi.org/10.1111/j.1467-968X.1899.tb00138.x

Kiparsky, Paul 1977. The rhythmic structure of English verse. In: Linguistic Inquiry 8, 
189–247.

Lázaro Carreter, Fernando 1972. La poética del arte mayor castellano. In: Studia 
Hispanica in Honorem R. Lapesa 1. Madrid: Gredos, 343–378.

Le Gentil, Pierre 1953. La poésie lyrique espagnole et portugaise à la fin du moyen âge 2. 
Rennes: Plihon.

León, Luis de 2011. Poesía. 13th ed. Madrid: Cátedra.

Leonard, William Ellery 1931. The Recovery of the Metre of the Cid. In: PMLA 46(2), 
289–306. https://doi.org/10.2307/458034

Mena, Juan de 1979. Laberinto de fortuna. Madrid: Cátedra.

Mena, Juan de 1995. Laberinto de fortuna. Madrid: Castalia.



83Juan de Mena’s arte mayor verse in Laberinto de fortuna ...

Morales-Front, Alfonso 2014. El acento. In: Núñez Cedeño, Rafael; Colina, Sonia; 
Bradley, Travis G. (eds.), Fonología generativa contemporánea de la lengua española, 
2nd rev. ed. Washington, D.C.: Georgetown University Press, 235–265.

Morel-Fatio, Alfred 1894. L’arte mayor et l’hendécasyllabe dans la poésie castillane 
du XVe siècle et du commencement du XVIe siècle. In: Romania 23(90), 209–231.

 https://doi.org/10.3406/roma.1894.5820

Morley, S. Griswold. 1933. Recent theories about the meter of the Cid. In: PMLA 48(4), 
965–980. https://doi.org/10.2307/458191

Navarro Tomás, Tomás 1967. Manual de pronunciación española. 6th ed. New York 
and London: Hafner.

Navarro Tomás, Tomás 1974. Métrica española: Reseña histórica y descriptiva. 4th ed. 
Barcelona: Labor.

Nebrija, Antonio de 2011. Gramática sobre la lengua castellana. Barcelona: RAE; 
Galaxia Gutenberg.

Nespor, Marina; Vogel, Irene 1989. On clashes and lapses. In: Phonology 6(1), 69–116. 
https://doi.org/10.1017/S0952675700000956

Orduna, Germán 1987. Función expresiva de la tirada y de la estructura fónico-rítmica 
del verso en la creación del Poema de Mio Cid. In: Incipit 7, 7–34.

Pellen, René 1985. Le modèle du vers épique espagnol, à partir de la formule cidienne 
“El que en buen hora…” (Exploitation des concordances pour l’analyse des 
structures textuelles). In: Cahiers d’Études Hispaniques Médiévales 10(1), 5–37.

 https://doi.org/10.3406/cehm.1985.948

Piera, Carlos 1981. Spanish Verse and the Theory of Meter [unpublished PhD thesis]. 
Los Angeles: UCLA.

Piera, Carlos 2003a. Del verso español y los universales métricos. In: Mairal, Ricardo; 
Gil; Juana (eds.), En torno a los universales lingüísticos. Madrid: Cambridge 
University Press; Akal, 265–303.

Piera, Carlos 2003b. Intonational factors in metrics. In: Michaux, Christine; Dominicy, 
Marc (eds.), Linguistic Approaches to Poetry. Amsterdam: John Benjamins, 206–228.

Piera, Carlos 2008. Southern Romance. In: Fabb, Nigel; Halle, Morris Meter in Poetry: 
A New Theory. Cambridge: Cambridge University Press, 94–132.

Quilis, Antonio 1993. Tratado de fonología y fonética española. Madrid: Gredos.

RAE [Real Academia Española] 2011. Nueva gramática de la lengua española 3: 
Fonética y fonología. Madrid: Espasa.



84 Jesús M. Saavedra Carballido

Roca, Iggy M. 1986. Secondary stress and metrical rhythm. In: Phonology Yearbook 3, 
341–370. https://doi.org/10.1017/S0952675700000683

Rodríguez Vázquez, Rosalía 2010. The Rhythm of Speech, Verse and Vocal Music:  
A New Theory. Bern: Peter Lang.

Roubaud, Jacques 1971. Mètre et vers: deux applications de la métrique générative de 
Halle-Keyser. In: Poétique 7, 366–387.

Saavedra Molina, Julio 1945. El verso de arte mayor. In: Anales de La Universidad de 
Chile 57–58, 5–127.

Salinas, Francisco de 1983. Siete libros sobre la música. Madrid: Alpuerto.

Santillana, Marquis of 2003. Poesías completas. Madrid: Castalia.

Smith, Colin 1979. La métrica del Poema de Mio Cid: Nuevas posibilidades. In: Nueva 
Revista de Filología Hispánica 28(1), 30–56. 

 https://doi.org/10.24201/nrfh.v28i1.1729

Tarlinskaja, Marina 1987. Shakespeare’s Verse: Iambic Pentameter and the Poet’s 
Idiosyncracies. New York: Peter Lang.

Tarlinskaja, Marina 1992. Metrical typology: English, German, and Russian dolnik 
verse. In: Comparative Literature 44(1), 1–21. https://doi.org/10.2307/1771165

Tarlinskaja, Marina 1993. Strict Stress Meter in English Poetry Compared with German 
and Russian. Calgary: University of Calgary Press.

Tarlinskaja, Marina 1995. Beyond “loose iamb”: the form and themes of the English 
“dolnik”. In: Poetics Today 16(3), 493–522. https://doi.org/10.2307/1773176

Tarlinskaja, Marina 2002. Verse text: its meter and its oral rendition. In: Küper, 
Christoph (ed.), Meter, Rhythm and Performance. Metrum, Rhythmus, Performanz. 
Frankfurt am Main: Peter Lang, 39–55.

Toledo, Guillermo Andrés 2002. Acentos en español: un problema para la fonología 
métrica. In: Verba 29, 119–138.

Tsur, Reuven 1998. Poetic Rhythm: Structure and Performance. Berne: Peter Lang.

Valle-Inclán, Ramón María del 1995. Claves líricas. 4th ed. Madrid: Espasa Calpe.

Vega, Garcilaso de la 2001. Obra poética y textos en prosa. Barcelona: Crítica.

Wright, George T. 1988. Shakespeare’s Metrical Art. Berkeley: University of California 
Press.



85Juan de Mena’s arte mayor verse in Laberinto de fortuna ...

Addenda

1. Underlying metrical presuppositions; the maximum

1a. Metrical presuppositions
Within the linguistically-inflected approach adopted in this article (inspired 
by Tarlinskaja 2002; cf. Jakobson 1966: 364–366), I assume a basic distinc-
tion between performance or delivery (Jakobson’s delivery design and delivery 
instance), rhythm (verse instance) and metre (verse design).

The performance includes such disparate phenomena as silent reading and 
a reciter’s oral rendition.

The rhythm is that of the authorial realization of verse, established by the 
analyst in view of phonological knowledge and, when necessary, of the work 
done by textual editors.

As regards the metre, I take it to comprise three elements (cf. Kiparsky 1977: 
190): a metrical template made up of strong and weak positions; metrical filters 
selecting what counts as a syllable and what syllables count as salient, that is, by 
means of which the prosodic material is metrified; realization rules connecting 
the template and the metrified materials, thus governing the author’s choice 
of linguistic elements and the audience’s expectations. My analysis focuses 
on the rhythm of the authorial realization and on the metre, not on a kind 
of delivery or on any individual performance. I assume that the metrified 
materials include neither pauses nor vocalic lengthenings (in Spanish, these 
two phonetic staples of temporal and musical theories of verse performance 
play no phonological role). I reject any tinkering with the linguistic features 
of lines, including the stresses, so that they will conform to the metre, a move 
which would “violate the linguistic givens of [the] language” as described by 
phonologists independently of verse analysis (Halle, Keyser 1966: 188), and 
which would render “almost any arrangement” of linguistic traits “metrically 
acceptable” (Halle, Keyser 1971: 162). In Spanish, this would push metrical 
verse in the direction of song. The treatment of stress in Spanish song involves 
the erasure of all preexisting stresses and the addition of new ones as called for 
by the tune (cf. Rodríguez Vázquez 2010; Domínguez Caparrós 2014: 75–76 
n. 52): at best, this amounts to rendering all Spanish verse purely syllabic; at 
worst, to doing away with any linguistic regulation. Respecting linguistic stric-
tures is a guarantee of methodological rigour and of the reproducibility of the 
analysis. Moreover, I distinguish those realizations that are correct (metrical, 
in an evaluative sense) from those that are incorrect (unmetrical). Finally, I 
focus on constant features; calculations of variable tendencies are confined to 
these Addenda. Without denying the usefulness of statistical methods when 



86 Jesús M. Saavedra Carballido

it comes to appreciating important stylistic nuances in the realization, to ana-
lysing the local effects of deviations from a given statistical norm, and even to 
fully characterizing a metre, I assume that, on their own, statistics cannot tell 
us what the metre is and what constitutes a correct realization thereof.

1b. The maximum
According to Halle and Keyser’s original definition, the maximum is “a syllable 
bearing linguistically determined stress that is greater than that of the two 
syllables adjacent to it in the same verse” (Halle, Keyser 1966: 197). Different 
versions of the maximum appear in other joint works by Halle and Keyser, 
where they suggest, for example, that it is “a stressed syllable […] located 
between two unstressed syllables in the same syntactic constituent within a 
line of verse” (Halle, Keyser 1971: 156), and also in Fabb, Halle (2008) and 
in Piera (2008), where different metres are said to demand different kinds of 
maxima based on different kinds of prosodic features (not only stress, but also 
tone, for example). The idea has a precedent in the theories of the Russian for-
malists; Jakobson, for example, states that in Russian “a stressed syllable cannot 
fall on the upbeat [weak position] if the downbeat [strong position] is fulfilled 
by an unstressed syllable of the same word” (1966: 361). Though by another 
name, the idea of the maximum also appears in Kiparsky (1977), according to 
whom English metres generally ban from weak positions a stress flanked by 
nonstresses in lexical polysyllables, such that the middle syllable in the word 
importance can only realize a strong position (but there might be exceptions 
to this rule too; cf. Shakespeare’s and Milton’s placement of words like Tiresias 
and universal in Kiparsky 1977: 201–203; and like temptation and dividable in 
Tsur 1998: 360). Hayes offers his own version of the maximum, calling it peak: 
a syllable stronger than “at least one of its neighbors” (1983: 376). The notion 
of the maximum is also implicit in Tarlinskaja’s works; she asserts, much like 
Kiparsky, that in English verse an internal stress in polysyllables like continues 
cannot realize weak positions (Tarlinskaja 1993: 10).

2. Spanish metrics and phonology

2a. Spanish metrics
In Spanish metrical verse, there is a distinction between verso simple and verso 
compuesto, that is, between what Piera calls a “unified” line (1981: 149), and 
what Clarke calls a “divided” or “compound” line (1964: 51 n. 1, 169). The short 
unified line (verso de arte menor) is usually distinguished from the long unified 



87Juan de Mena’s arte mayor verse in Laberinto de fortuna ...

line (called, somewhat confusingly when dealing with Mena’s metre, verso de 
arte mayor) in terms of their lengths. Where the frontier falls between both 
types is a bone of contention (cf. Domínguez Caparrós 2014: 134). Arguably, 
however (cf. Domínguez Caparrós 2014: 49), the difference is that the short 
unified line consists of one single metrical part or colon, while its long coun-
terpart consists of at least two cola of equal or similar length, without internal 
extrametricality (this means that a word can straddle two cola). Metrically, the 
divided line can be explained as a combination of two or more independent 
hemistichs, i.e. of different unified lines placed side by side, all of which can 
have some extrametrical material at the end (so a word cannot straddle two 
hemistichs). The phenomenon of the divided line is very frequent in Spanish, 
where its use has long been systematically regulated; its sporadic counterpart 
is the phenomenon known as epic caesura (cf. Domínguez Caparrós 2014: 94 
n. 67; for English, cf. Attridge 1982: 245; Wright 1988: 165–169). The defining 
characteristic of the divided line is that neither the length nor the structure of 
one hemistich affect the length or the structure of the next. Thus, one hemistich 
may end in one or more extrametrical syllables, or display no extrametricality 
at all; in the latter case, if the next hemistich begins with a strong position, two 
strong positions can appear back-to-back within the line (this does not mean 
that in between there may be unrealized or void weak positions: pace some 
analysts of arte mayor, in Spanish verse this phenomenon does not occur).

2b. Spanish syllables
In Spanish, only vocalic phonemes can be heads of syllables; there are two 
high vowels (/i/, written i and y, and /u/, written u and ü), two mid vowels 
(/e/, written e, and /o/, written o) and one low vowel (/a/, written a). If, as 
Quilis claims (1993: 182–183), Spanish semivowels (glides) are only allophonic 
variants ([j], [w]) of vocalic phonemes (/i/, /u/), then we can treat them just 
like any other vowel (cf. RAE 2011: 342). In principle, then, utterances have 
as many syllables as vowels, but almost any sequence of contiguous vowels, 
with or without an intervening h (which in present-day Spanish is silent), 
can function as one single syllable in speech, as long as it does not follow the 
order lower–higher–lower (in which case we pronounce between two and as 
many syllables as vocalic phonemes there are in the sequence). Similarly, in 
metrical verse almost any vocalic sequence, with or without an intervening 
h, amounts to one or as many (metrified) syllables as vowels there are in it; 
if the sequence follows the order lower–higher–lower we count between two 
and as many syllables as vowels we find. Thus, the 5-vowel sequence in odio a 



88 Jesús M. Saavedra Carballido

Eurípides = I hate Euripides can be treated as any number of syllables between 
one and five (cf. Navarro Tomás 1967: 72, 150–151). Witness how one strong 
metrical position is realized by the ending – a stressed vowel plus an unstressed
vowel (plus a consonant) – of venían = they came in this 10-position line by 
Garcilaso de la Vega: “de dos pastores que venían cantan{do}” (Vega 2001: 
244). Obviously, when verse lines are said aloud, all the vowels can be uttered, 
and pronounced as belonging to the same prosodic syllable or not: “it is just 
that some of them do not count for metrical purposes” (Piera 2008: 97).

2c. Spanish stress
In Spanish, syntactic words are prosodically divided into tónicas and átonas, 
that is, tonic (always stressed) and atonic (usually unstressed). The list of tonic 
forms includes all content words: nouns, adjectives, verbs and adverbs (Quilis 
1993: 390–395; RAE 2011: 370–376), and most interjections. Simple words 
have one stress alone; compounds can have one or two stresses. As regards 
simple words, I take it that all trisyllabic and longer words and locutions are 
tonic. Some function monosyllables and disyllables – including the preposition 
según(d) = according to/as (Quilis 1993: 392; RAE 2011: 417) – are tonic too. 
As regards compounds, some behave like simple words in that only the final 
element is tonic; but each part of the adverbs in -mente and of many other 
compounds counts as a tonic word; thus, for example, the adverb felizmente = 
happily has two stresses (Piera 1981: 214; Quilis 1993: 391; RAE 2011: 392–397; 
Hualde 2009: 202–204; 2010: 12; 2014: 226, 231–233). All other words – always 
function words in proclitic positions, including the poetic interjection o(h) 
and the conjunction sino(n) = but (which I take to be oxytonic) – are atonic. 
(Enclitics are integrated in the preceding content word – necessarily a verb, 
whose preexisting stress does not move – such that the noun dique = dam and 
the verb plus enclitic dime = tell me have the same prosodic structure.)

Secondary stresses in Spanish remain a controversial issue. Their place-
ment is usually explained either as conditioning that of the primary stress, 
or as being conditioned by it (cf. Morales-Front 2014: 259). In the former 
interpretation, secondary stresses are considered to obey the lexical rules that 
determine the structure of words; in the latter interpretation, they are regarded 
as postlexical (added by the utterer); in both interpretations, the assignment 
of secondary stresses is governed by the belief that Spanish words cannot 
have more than two contiguous nonstresses, at least to the left of the primary 
stress (Navarro Tomás 1967: 195–196; Harris 1983: 85–86, 96–97; Roca 1986; 
D’Introno et al. 1995: 156–173, 411–436). Although their existence is not 



89Juan de Mena’s arte mayor verse in Laberinto de fortuna ...

explicitly denied by any author, I interpret the absence of references to lexical 
secondary stresses, in works such as Quilis (1993), as an implicit rejection. In 
general, the phonologists that admit the existence of lexical secondary stresses 
tend to accept Navarro Tomás’s view that one or more of them can be placed to 
the left of the primary stress of the same word (on the first syllable and/or on 
a syllable separated from that primary stress by an odd number of unstressed 
syllables), and/or to the right of the primary stress (on a syllable separated 
from that primary stress by an odd number of syllables). Though postlexical 
secondary stresses are rejected by some authors (e.g. Toledo 2002: 135), their 
existence is widely recognized. According to Roca (1986), one or more of them 
can be inserted to the left of the primary stress of a word, but only on syllables 
separated from the primary stress by an odd number of syllables. According 
to Hualde (2009: 205–207; 2010; 2014: 252–253) and Hualde, Nadeu (2014: 
244–247), one or more postlexical secondary stresses can optionally be placed 
to the left of the primary stress of a word, on the first syllable or on a syllable 
preceding the primary stress by an odd number of syllables, and/or on the 
final syllable that there may be to the right of that primary stress, provided 
that this syllable is an enclitic separated from the primary stress by at least one 
intervening syllable (for the latter phenomenon, cf. Navarro Tomás 1967: 196; 
RAE 2011: 412–413); however, postlexical secondary stresses placed to the 
left of a primary stress only appear in certain registers, where they serve, for 
example, to draw attention to the importance of the utterance (Hualde 2009: 
206; 2010: 13–14; 2014: 252–253; Hualde, Nadeu 2014: 231), and postlexical 
secondary stresses to the right of the primary stress are even rarer. Postlexical 
strategies like stress deletion and (leftward) displacement have been suggested 
for a Romance language like Italian (cf. Nespor, Vogel 1989), but the latest 
linguistic research does not mention the possibility of these rhythmic proce-
dures in Spanish.

3. Arte mayor: Theories and problems 

3a. Previous theories of arte mayor
At the close of the 15th century, Nebrija described arte mayor as a divided 
line amounting to an “adónico doblado” = “double adonic”: a repetition of 
the 5-syllable sequence dactyl-plus-“spondee” – i.e. trochee – totalling (due 
to the possible presence of one extrametrical syllable at the beginning of each 
sequence and the possible absence of a syllable at their end) between eight 
and eleven syllables up to the rightmost stress (2011: 71–73; cf. Balaguer 1995: 
29–45). Navarro Tomás, from the camp of temporal metrics, speaks of two 



90 Jesús M. Saavedra Carballido

dactylic hemistichs of four syllables up to the rightmost stress (the nonstresses 
following this stress are optional, the dactylic ending is rare); either hemistich 
can optionally present 1- or 2-syllable initial extrametricality, thus reaching up 
to six syllables (normally five); in theory the total number of syllables of this 
divided line goes from eight to fourteen, but in practice it goes from nine to 
thirteen, as the optional nonstresses between the hemistichs are one or two, 
never zero or four; the rhythm is usually ternary within each hemistich, yet 
it can be binary (1974: 115–117). These are the only major accounts positing 
a falling metre.

In the late 16th century, Salinas described arte mayor as made up of amphi-
brachs (1983: 573–574). Somewhat later, Correas added that it contained from 
nine to eleven syllables, that some feet could be headless and that others could 
be truncated (Correas 1954: 472-475). More recently, other authors have fol-
lowed suit (Bello 1884: 141–145; Piera 1981; Duffell 1999).

For Díaz Rengifo, also writing in the late 16th century, arte mayor was a 
divided form combining two rising hemistichs of five syllables, the second and 
fifth stressed, and with or without one extrametrical nonstress at the end of the 
hemistichs, which together add up to ten or eleven syllables (2012: 178–179). 
For Burger arte mayor is a divided form combining hemistichs of four and 
five syllables up to the rightmost stress; all other possible stresses depend on 
the versifier’s personal style (1957: 34–35). Morel-Fatio analysed arte mayor 
as a divided form akin to the French decassyllable, with a break after the fifth 
syllable, “nécessairement accentué sur la cinquième syllable de chaque hémis-
tiche” = “necessarily stressed on the fifth syllable of each hemistich”, and whose 
syllabic irregularities must be ironed out either by the editor or in performance 
(1894: 211, 219–221). By suggesting the latter kind of correction, Morel-Fatio 
shows what some current metrical theories regard a misunderstanding of the 
relation between the metre, the rhythm of the authorial realization and the 
performance by a silent reader or by a reciter. The same charge can be lev-
elled, as diagnosed by Duffell (1999: 11, 83–84), at the description provided 
by Lázaro Carreter (1972), which is still popular among textual editors (cf. 
John G. Cummins’s remarks in Mena 1979: 44–46; Maxim P. A. Kerkhof ’s 
comments in Mena 1995: 105 n. 149). Contrary to those scholars who, as Piera 
puts it, believe that in Mena’s days “reading habits did not impose the pattern 
stresses at the expense of the natural stress configuration” (1981: 144), Lázaro 
Carreter advocates what Clarke calls a “mechanical” chanting of the lines 
(1964: 56). He understands the arte mayor metre as consisting in divided lines, 
each hemistich with two strong positions realized by their respective stresses 
and separated by two nonstresses, and he believes in the “omnipotencia” = 
“omnipotence” of this metre (1972: 350), to which the authorial realization and 



91Juan de Mena’s arte mayor verse in Laberinto de fortuna ...

what he regards as the correct performance are subordinated: thus, stresses 
are added, deleted or displaced by the author so that they coincide with strong 
positions; and those stresses that do not undergo this authorial treatment must 
be wrenched, when necessary, by the performer.

Alternatively, arte mayor has been analysed not as a divided line but as a 
long unified line. Foulché-Delbosc posits a line consisting of two parts (equiva-
lent to two cola). Each part has between four and seven syllabes = syllables, 
including two syllabes accentuées = accented syllables (strong positions), each 
realized by a stress separated by two syllabes non accentuées = unaccented syl-
lables (weak positions) corresponding to nonstresses; other nonstresses can be 
added at the beginning and at the end of each part. Together, both parts total 
between nine and twelve (usually eleven) syllables up to the last accented sylla-
ble. The structure of one part affects the other, to the effect that the last accented 
syllable of the initial part and the first of the final part must be separated by at 
least one unaccented syllable. Foulché-Delbosc contends that, when the first 
stress of a given part appears earlier than expected, this part has only one 
accented syllable, the last (1902: 94–103). In total he describes thirty-six line 
types in Laberinto and thirty-two further arte mayor possibilities (including 
lines with proparoxytonic endings). From other perspectives, the arte mayor 
line has been analysed as a unified line in which the initial colon contains 
between four and six syllables, and the final one contains between five and six 
(Roubaud 1971: 380–381); as a line containing from seven to sixteen syllables 
(Saavedra Molina 1945: 65, 70), and as a metre whose weak positions are 
realized, at the beginning of the line, by zero to two syllables, and, elsewhere, 
by one to three – a form whose lines could therefore have, were it nor for the 
introduction of some ad hoc restrictions, between seven and fifteen syllables 
up to the rightmost stress (Piera 2008: 122).

3b. Secondary stresses in arte mayor
Spanish metrists tend to assume that secondary stresses coincide with (what 
they think to be) strong positions not realized by stressed heads (e.g. Balaguer 
1995: 193 ff.; Domínguez Caparrós 2014: 75–76). In the absence of conclu-
sive evidence, I assume that secondary stresses are inexistent or irrelevant for 
Laberinto. After all, irrespective of what many Spanish metrists and editors 
may think (cf. Cummins in Mena 1979: 177), all the linguistic descriptions 
of Spanish stress agree that, unless we adopt a chanting style, in a line like 
“reconoscerán, # maguer que fero{çe}” (l. 2186 of Laberinto), the first word 
would never present a secondary stress on the second syllable, but only, if at 



92 Jesús M. Saavedra Carballido

all, on the first and/or third (the resulting scansions being x x x x X or X x x 
x X or X x X x X, or even x x X x X, but never x X x x X); in brief, this word 
does not conform to the ternary rhythm that many metrists demand from all 
lines of arte mayor. To this it must be added that in no variety of arte mayor 
has the last syllable of proparoxytones ever been said to realize the rightmost 
strong position of the line, so the alleged stresses at the end of rápido = fast 
and of pensándolo = thinking it would not even behave as such.

4. Mena’s Laberinto

4a. Analysed lines according to line type and to colon and line ends
There follows a list of the lines from Laberinto without consecutive or elided 
vowels, and without h between vowels: there are 442 of them, i.e. 18.6 % of 
the 2376 lines that make up Mena’s work. There are two that I consider metri-
cally incorrect: *224 (example (22)) and *1563 (example (21)). The ones that 
I consider metrically correct (440 lines, 18.5 %), and on which all subsequent 
statistics are based, are grouped according to line types, with those whose cola 
have been disambiguated (41 lines, 9.3 %) marked with a ? sign. Oxytonic 
(“masculine”) endings are marked m (43 lines, 9.8 %), paroxytonic (“femi-
nine”) endings are marked f, proparoxytonic (“dactylic”) endings are marked 
d (Laberinto consisting of unified lines, the paroxytonic and proparoxytonic 
endings of initial cola actually belong to the corresponding final cola; par-
oxytonic endings of final cola are extrametrical; proparoxytonic final-colon 
endings are systematically avoided):

•	 5#6 type (288 lines, 65.5 %):

−	f#f (“amphibrachic” type; 243 lines, 55.2 %): 3, 9, 10, 21, 40, 55, 88, 114, 115, 
131, 155, 156, 163, 168, 174, 179, 201, 208, 211, 232, 259, 260, 284, 288, 311, 
365, 371, 374, 390, 391, 401, 404, 424, 428, 439, 443, 445, 465, 467, 474, 481, 491, 
498, 499, 509, 526, 541, 550, 556, 565, 588, 635, 651, 656, 657, 661, 664, 674, 682, 
693, 694, 707, 715, 740, 752, 754, 758, 759, 765, 803, 807, 812, 813, 815, 828, 830, 
843, 850, 856, 876, 881, 883, 888, 896, 899, 905, 909, 918, 921, 926, 936, 944, 951, 
983, 991, 996, 1040, 1049, 1067, 1100, 1107, 1109, 1111, 1119, 1120, 1121, 1126, 
1150, 1153, 1160, 1161, 1166, 1167, 1178, 1181, 1184, 1186, 1198, 1204, 1207, 
1218, 1232, 1235, 1239, 1253, 1257, 1262, 1265, 1283, 1306, 1307, 1309, 1318, 
1352, 1360, 1378, 1407, 1416, 1440, 1472, 1477, 1480, 1484, 1494, 1496, 1497, 
1500, 1502, 1504, 1506, 1509, 1514, 1520, 1538, 1554, 1569, 1573, 1575, 1576, 
1588, 1600, 1604, 1609, 1611, 1629, 1640, 1642, 1643, 1645, 1646, 1647, 1658, 



93Juan de Mena’s arte mayor verse in Laberinto de fortuna ...

1665, 1688, 1703, 1710, 1731, 1733, 1738, 1773, 1789, 1792, 1796, 1801, 1802, 
1807, 1809, 1812, 1816, 1819, 1831, 1837, 1853, 1865, 1878, 1898, 1902, 1906, 
1910, 1911, 1914, 1916, 1939, 1960, 1976, 1980, 2003, 2014, 2017, 2019, 2037, 
2041, 2053, 2111, 2115, 2138, 2141, 2142, 2146, 2153, 2168, 2173, 2181, 2200, 
2221, 2224, 2241, 2244, 2263, 2268, 2277, 2285, 2287, 2288, 2289, 2295, 2310, 
2313, 2323, 2333, 2346, 2351, 2357.

−	f#m (24 lines, 5.5 %): 319, 353, 409, 459, 504, 641, 681, 688, 705, 724, 728, 854, 
861, 959, 1270, 1281, 1285, 1540, 1727, 1881, 1888, 1926, 2001, 2060.

−	d#f (14 lines, 3.2 %): 569, 730, 963, 965, 973, 1031, 1073, 1374, 1605, 1698, 1849, 
2129, 2201, 2283.

−	m#f (4 lines, 0.9 %): 561?, 848, 1383?, 1448.

−	d#m (3 lines, 0.7 %): 643, 957, 1342.

•	 4#6 type (140 lines, 31.8 %):

−	f#f (116 lines, 26.4 %): 15, 19?, 22, 54?, 56, 71, 82, 125?, 158?, 254, 454, 485?, 
495, 496, 507, 519, 528?, 572, 576, 580, 627?, 639, 649, 679, 704, 720, 733, 735, 
738, 789, 832, 836, 842, 845, 882?, 884, 887, 889, 911, 962, 971, 986, 1011, 1016, 
1027?, 1043, 1063, 1069?, 1125, 1127, 1144, 1154, 1159?, 1185?, 1225?, 1259, 
1297, 1298?, 1312, 1328, 1358, 1364, 1369?, 1375, 1393, 1408, 1411, 1415, 1457, 
1501, 1517?, 1528, 1547, 1656, 1672, 1699, 1708, 1711, 1729?, 1739?, 1755?, 
1776, 1790?, 1810, 1811?, 1818, 1843, 1869, 1873?, 1899, 1915, 1936, 1977, 1990, 
1992?, 1995?, 1996?, 2026?, 2035, 2039, 2066, 2086?, 2107, 2119, 2127, 2216, 
2235?, 2264, 2272, 2292?, 2294, 2321, 2343, 2344, 2353?, 2363? 

−	d#f (13 lines, 3 %): 506, 546, 702, 941, 1022, 1029, 1140, 1704, 1856, 1972, 2000, 
2081, 2094.

−	f#m (11 lines, 2.5 %): 27?, 559, 855, 1455, 1591, 1844, 1848, 1931, 2005, 2057, 
2275?

•	 5#5 type (8 lines, 1.8 %):

−	m#f (5 lines, 1.1 %): 79?, 1056, 2034?, 2186, 2286?

−	m#m (2 lines, 0.5 %): 894?, 1590.

−	f#f (1 line, 0.2 %): 1037.

•	 4#5 type (3 lines, 0.7 %)

−	m#f (2 lines, 0.5 %): 167?, 307?

−	f#m (1 line, 0.2 %): 1965.



94 Jesús M. Saavedra Carballido

•	 6#6 type (1 line, 0.2 %):

−	f#m: 939.

All lines without contiguous vowels in which straddling proparoxytones end in 
6-syllable cola are correct. The only line without contiguous vowels in which a 
straddling proparoxytone does not end in a 6-syllable colon is l. *1563 (exam-
ple (21)), which has been considered incorrect.

4b. General rhythmic tendencies in Mena’s Laberinto
The relation of stresses and nonstresses in each line type is the following (only 
syllables realizing metrical positions are taken into account, as all extrametri-
cal syllables are nonstresses and nonmaxima; colon-final strong positions are 
underlined):

•	 5#6:

Syllable no.: 1 2 3 4 5 6 7 8 9 10 11
% Stresses: 1.4 86.5 7.6 5.2 99.7 0 10.1 74.7 8 2.4 100
% Maxima: 13.9 97.6 28.5 5.2 100 0.7 25.7 87.2 26.7 2.4 100

•	 4#6:

Syllable no.: 1 2 3 4 5 6 7 8 9 10
% Stresses: 56.4 5.7 3.6 100 0 12.1 74.3 12.9 3.6 100
% Maxima: 96.4 45.7 3.6 100 0 27.1 85 29.3 3.6 100

•	 5#5:

Syllable no.: 1 2 3 4 5 6 7 8 9 10
% Stresses: 12.5 50 0 0 100 25 50 25 0 100
% Maxima: 50 62.5 12.5 0 100 50 50 50 0 100

•	 4#5:

Syllable no.: 1 2 3 4 5 6 7 8 9
% Stresses: 33.3 0 0 100 0 0 66.7 0 100
% Maxima: 100 0 0 100 66.7 33.3 66.7 0 100

•	 6#6:

Syllable no.: 1 2 3 4 5 6 7 8 9 10 11 12
% Stresses: 0 0 100 0 0 100 0 0 100 0 0 100
% Maxima: 0 0 100 0 0 100 0 0 100 0 0 100



95Juan de Mena’s arte mayor verse in Laberinto de fortuna ...

4c. Ternary rhythmic tendencies in Mena’s Laberinto
In lines in which the final colon is realized by six syllables, ternary tendencies 
can hold throughout the line (i.e. both within and across cola):

•	 Lines with four noncontiguous maxima marking a ternary rhythm: (i) all max-
ima are stressed (112 lines, 25.5 %): 3, 9, 10, 21, 55, 114, 115, 131, 155, 174, 201, 
254, 374, 391, 424, 428, 439, 467, 499, 541, 556, 576, 635, 639, 679, 694, 704, 715, 
740, 754, 758, 765, 789, 845, 876, 905, 936, 939, 951, 962, 971, 973, 986, 1016, 
1029, 1049, 1107, 1109, 1120, 1144, 1160, 1166, 1178, 1181, 1198, 1262, 1283, 
1285, 1408, 1477, 1480, 1497, 1501, 1502, 1506, 1573, 1588, 1600, 1604, 1640, 
1642, 1647, 1658, 1665, 1703, 1708, 1731, 1802, 1809, 1818, 1831, 1843, 1849, 
1856, 1881, 1902, 1906, 1914, 1936, 1939, 1960, 1977, 2014, 2017, 2019, 2053, 
2066, 2081, 2127, 2129, 2141, 2173, 2200, 2221, 2244, 2263, 2264, 2277, 2288, 
2289, 2310, 2313; (ii) not all maxima are stressed (15 lines, 3.4 %): 507, 519, 569, 
702, 883, 957, 1411, 1457, 1554, 1591, 1699, 2000, 2216, 2283, 2285.

•	 Lines with more than four maxima but only four noncontiguous stresses marking 
a ternary rhythm (72 lines, 16.4 %): 71, 82, 88, 156, 168, 211, 259, 260, 409, 445, 
454, 496, 546, 561, 565, 688, 724, 730, 733, 738, 752, 812, 828, 830, 848, 854, 855, 
888, 926, 944, 959, 996, 1011, 1073, 1111, 1161, 1167, 1184, 1186, 1207, 1232, 
1257, 1259, 1306, 1307, 1309, 1342, 1360, 1383, 1440, 1455, 1484, 1528, 1540, 
1569, 1646, 1672, 1727, 1733, 1801, 1816, 1848, 1916, 2039, 2060, 2094, 2153, 
2181, 2268, 2272, 2287, 2351.

•	 Lines with four noncontiguous maxima not marking a ternary rhythm: (i) all max-
ima are stressed (8 lines, 1.8 %): 284, 651, 803, 1185, 1235, 1472, 2115, 2119; (ii) not 
all maxima are stressed (7 lines, 1.6 %): 856, 1496, 1688, 1796, 1926, 2168, 2275.

•	 Lines with more than four maxima but only four noncontiguous stresses not 
marking a ternary rhythm (20 lines, 4.5 %): 465, 474, 498, 526, 707, 815, 896, 
1407, 1520, 1547, 1575, 1609, 1853, 1992, 2019, 2026, 2037, 2111, 2292, 2346.

In lines in which the final colon is realized by five syllables, ternary tendencies 
hold either within cola or across them:

•	 Lines with four noncontiguous maxima marking a ternary rhythm in which not 
all maxima are stressed (1 line, 0.2 %): 1965 (ternary tendencies within cola).

•	 Lines with more than four maxima but only four noncontiguous stresses marking 
a ternary rhythm (1 line, 0.2 %): 1590 (ternary tendencies within cola).



96 Jesús M. Saavedra Carballido

•	 Lines with four noncontiguous maxima not marking a ternary rhythm in which 
not all maxima are stressed (1 line, 0.2 %): 2286.

•	 Lines with more than four maxima but only four noncontiguous stresses not 
marking a ternary rhythm (1 line, 0.2 %): 894 (ternary tendencies across cola).

4d. Realization of strong positions by nonmaxima placed next to maxima
Lines without vowels in which a nonmaximum realizing a strong position is 
next to a maximum, at least one of which belongs to a polysyllable (the pro-
visional acceptance of headlessness in 4-syllable initial cola means that the 
displacement of the maximum always occurs in the final colon):

•	 in which the nonmaximum and the maximum do not belong to the same word: 
(i) within one foot, with leftward displacement of the maximum (14 lines, 3.2 %): 
528, 705, 883, 909, 1185, 1472, 1688, 1796, 1807, 1865, 1926, 2168, 2275, 2346;  
(ii) across feet, with rightward displacement of the maximum (6 lines, 1.4 %, 
none of them with 4-syllable initial cola): 815, 1031, 1575, 1995, 1996, 2111.

•	 in which both the nonmaximum and the maximum belong to the same word, 
which therefore thwarts metrical expectations on more than one occasion:  
(i) within one foot, with leftward displacement of the maximum (20 lines, 4.5 %): 
27, 54, 232, 284, 353, 651, 803, 832, 889, 896, 911, 1239, 1407, 1415, 1496, 1547, 
1873, 2037, 2115, 2295; (ii) across feet, with rightward displacement of the maxi-
mum (9 lines, 2 %): 319, 1609, 1611, 1755, 1837, 1972, 2026, 2119, 2292.