2. Default patterns

Among the possible default patterns that a language might exhibit, there are only four with a perfect binary alternation. In other words, there are only four default patterns where prominence is located on every other position so that a form contains neither clashes nor lapses. These four perfect alternation patterns can be divided into two pairs, each pair consisting of one trochaic pattern and one iambic pattern. The first pair, the minimal alternation patterns, have the fewest prominences possible without a lapse. The minimal alternation patterns are illustrated in (4) with the structures they would be assigned in the Weak Bracketing framework. In (4) and throughout this article, entries on the metrical grid are positioned above the syllable string, and foot structure is indicated by association lines below it. The head syllables of feet are indicated by vertical association lines.

The second pair, the maximal alternation patterns, have the most prominences possible without a clash. The maximal alternation patterns are illustrated in (5).

Under the Weak Bracketing approach, default patterns with clash or lapse emerge as variations on the perfect alternation patterns, and are due either to a requirement that a prominence occur in initial position or to a requirement that a prominence avoid final position.

The portion of the typology that concerns us here is the group of iambic default patterns that emerge as variations on the iambic minimal alternation pattern in (4b). This group emerges under a requirement that prominence avoid final syllables. Two members of the group have been discussed previously in the context of the Weak Bracketing analysis. The first is the variation where a lapse occurs at the right edge in even-parity forms. The pattern in (6a) is just like the iambic minimal alternation pattern in (4b), except that the final prominence of even-parity forms is absent, resulting in a lapse. We will refer to this variation as the final lapse variation. The second variation is one where a clash configuration occurs near the right edge in even-parity forms. The pattern in (6b) is just like the iambic minimal alternation pattern in (4b) except that the final prominence in even-parity forms shifts one position to the left, resulting in a clash. This variation is often referred to as iambic reversal, since the final foot in even-parity forms switches from iambic to trochaic.

The final lapse variation can be seen in Choctaw (Nicklas Reference Nicklas1972, Reference Nicklas and Crawford1975), Hixkaryana (Derbyshire Reference Derbyshire1985) and Central Alaskan Yupik (Miyaoka Reference Miyaoka1985; Woodbury Reference Woodbury1987).Footnote ³ Examples from Choctaw are given in (7). In longer Choctaw forms, syllables with long vowels are stressed. Every even-numbered syllable from the left, except the final syllable, has a long vowel.

The iambic reversal variation can be seen in De’kwana (Hall Reference Hall1988), Aguaruna (Payne Reference Payne and Payne1990; Hung Reference Hung1994), Southern Paiute (Sapir Reference Sapir1930) and Axininca Campa (Payne Reference Payne1981). Examples from De’kwana are given in (8).

The Sentani default pattern constitutes a third variation on iambic minimal alternation. As (9) illustrates, the Sentani variation combines aspects of the previous two. As in the two previous variations, odd-parity forms are identical to the odd-parity forms of iambic minimal alternation from (4b). In even-parity forms, the final prominence shifts one position to the left as in the iambic reversal variation in (6b), but it does not always result in iambic reversal’s clash. It results in a clash in forms where the initial and final foot are adjacent (i.e., forms with four syllables), as in (9a), but not in forms where they are not (i.e., forms with more than four syllables), as in (9c). In forms where the initial and final foot are not adjacent, the prominence expected on the antepenult is removed, resulting in a lapse. Even-parity forms with more than four syllables, then, have a lapse like the final lapse variation, though the lapse arises in a different location.

Below, we demonstrate how the Sentani variation emerges under circumstances similar to those that produce the final lapse and iambic reversal variations. Like the other variations, the Sentani variation emerges under a requirement that prominence avoid final position. Unlike the other variations, however, the Sentani variation requires a constraint restricting the position of primary stress. It can only emerge when primary stress is restricted to the final foot. The Sentani variation also requires a constraint insisting that the initial foot, in particular, be stressed (in addition to a constraint insisting that feet generally be stressed). Both of the additional requirements are available from constraint families – alignment (McCarthy & Prince Reference McCarthy, Prince, Booij and Marle1993; McCarthy Reference McCarthy2003; Hyde Reference Hyde2012a) and initial prominence (Hyde Reference Hyde2002, Reference Hyde2016a) – that have previously been established as fundamental to the Weak Bracketing analysis.

3. The Weak Bracketing analysis

Before presenting the Weak Bracketing analysis of Sentani, we briefly examine the Weak Bracketing analysis of iambic minimal alternation, iambic final lapse and iambic reversal. (For a more thorough discussion, see Hyde Reference Hyde2016a.) The five constraints most immediately responsible for iambic minimal alternation and its variations are discussed below, beginning with two alignment constraints.

We employ the Relation-Specific Alignment formulation of Hyde (Reference Hyde2012a, Reference Hyde2016a; see also McCarthy Reference McCarthy2003 and Hyde & Paramore Reference Hyde, Paramore, Hansson, Farris-Trimble, McMullin and Pulleyblank2016). For those unfamiliar with the formulation, the general schemas for Relation-Specific Alignment constraints are given in (10). Each schema has two components separated by a slash. The component to the right of the slash is the prohibited configuration of misalignment, and the component to the left of the slash defines a locus of violation. The prohibited configuration is constructed from three categories. ACat1 and ACat2 are the two categories whose edges are being aligned, and SCat is the ‘separator category’, whose intervention between the relevant edges of ACat1 and ACat2 constitutes misalignment. The locus of violation can consist of either two categories or three. It can consist of the two aligned categories, ACat1 and ACat2, or it can consist of the two aligned categories plus the separator category, SCat.

Note that the separator category is only optionally included in the locus of violation. When the separator category is included, the constraint is distance-sensitive: the number of violations it assesses increases as the degree of misalignment increases. When the separator category is omitted, the constraint is distance-insensitive: it assesses a single violation for each pair of misaligned categories regardless of the degree of misalignment.

The alignment constraints most relevant to producing iambic minimal alternation and its variants are AllRight(σ_Hd) and AllLeft(σ_Hd), given in (11). AllRight(σ_Hd) draws the head syllables of feet towards the right edge of the prosodic word, and AllLeft(σ_Hd) draws the head syllables of feet towards the left edge.

Since the separator category ‘σ’ is included in the loci of violation, the constraints are distance-sensitive. They register degrees of misalignment rather then the simple fact of misalignment. AllRight(σ_Hd) assesses a number of violations that is equal the number of syllables intervening between each foot and the right edge of the prosodic word in which it occurs. Similarly, AllLeft(σ_Hd), assesses a number of violations that is equal the number of syllables intervening between each foot and the left edge of the prosodic word in which it occurs.

Constraints that map prosodic categories to entries on the metrical grid play a key role in the Weak Bracketing framework. The general formulation for Map constraints (Hyde Reference Hyde2002) is given in (12), where κ is the prosodic category to be mapped to the grid and x_κ is a κ-level grid entry.

The Map constraint that is key at this point in the discussion is Map(f), given in (13). Map(f) requires that each foot have a foot-level grid entry within its domain.

Though foot-level grid entries must coincide with head syllables (indicated with vertical association lines), it is not case that all head syllables will coincide with grid entries. We will encounter an additional Map constraint, Map(ω), in §4. Map(ω) requires prosodic words to map to prosodic-word level grid entries.

The next constraint key to producing iambic minimal alternation and its variants is *Clash (Liberman & Prince Reference Liberman and Prince1977; Prince Reference Prince1983). *Clash requires that any two adjacent entries on one level of the metrical grid have an intervening entry on the next level lower. This prevents grid entries from occurring too close together.

The final key constraint is a Nonfinality constraint. NonFin(x_f), given in (15), prohibits foot-level grid entries from occurring on prosodic word-final syllables.

Iambic minimal alternation and its variations all share the same pattern for their odd-parity forms. This lack of distinction arises because all four of the active constraints required to differentiate the three patterns – AllRight(σ_Hd), Map(f), *Clash and NonFin(x_f) – can be maximally satisfied in words with an odd number of syllables, which means that their ranking in relation to one another does not alter the optimal candidate for such forms. As long as AllRight(σ_Hd) dominates AllLeft(σ_Hd), as the tableau in (16) illustrates, a right-aligned iambic pattern emerges regardless of the rankings between AllRight(σ_Hd), Map(f), *Clash and NonFin(x_f). In the optimal candidate, a pair of overlapping feet at the right edge of the prosodic word share a grid entry, allowing both to satisfy Map(f) while avoiding NonFin(x_f) and *Clash violations.

Iambic minimal alternation and its variations, then, differ only in their even-parity forms. Which of the even-parity patterns emerges depends on the ranking of AllRight(σ_Hd), Map(f), *Clash and NonFin(x_f). We examine the rankings that produce the different even-parity patterns below.

4. The Sentani variation: six-syllable forms

Where the iambic reversal variation employs clash in even-parity forms to avoid final prominence, and the iambic final lapse variation employs lapse, the Sentani variation employs both. In forms with four-syllables – forms where the initial and final feet are adjacent – Sentani uses clash to avoid prominence on the final syllable. In six-syllable and longer even-parity forms – forms where the initial and final feet are not adjacent – Sentani uses lapse to avoid prominence on the final syllable. We consider six-syllable and longer even-parity forms in this section and four-syllable forms in §5.

The Sentani variation cannot emerge under the set of constraints discussed thus far because the pattern found in six-syllable and longer even-parity forms is harmonically bounded by the iambic final lapse pattern with respect to these constraints.Footnote ⁴ As the tableau in (23) demonstrates, the two patterns perform equally well on NonFin(x_f), Map(f) and *Clash, but iambic final lapse performs better on AllRight(σ_Hd). The result is that the Sentani pattern cannot be optimal under any ranking of this set of constraints.

The reason that the six-syllable Sentani pattern is harmonically bounded by iambic final lapse with this set of constraints is that leaving the penultimate foot unmapped does not offer all of the same advantages as leaving the final foot unmapped. While the six-syllable Sentani pattern avoids final prominence and clash, it does not avoid the final trochaic foot that results in an additional AllRight(σ_Hd) violation.

Despite not offering the advantage of avoiding a final trochee, lapse arises in the Sentani variation for essentially the same reasons as in the iambic final lapse pattern: it arises to avoid final stress and clash. The lapse arises in a different position in Sentani because the position of lapse in the final lapse pattern is simply unavailable in Sentani. As discussed above, primary stress always occurs within the final foot of a Sentani form. This has two consequences. The first is that the foot hosting primary stress – the final foot – has to be trochaic in order to avoid final stress. The second is that the final two syllables – the syllables that constitute the final foot – are unavailable for hosting a lapse. To avoid the clash that would otherwise result from a final trochaic foot, the penultimate foot is left unmapped, resulting in a lapse preceding the primary stress.

The first consideration in producing the Sentani pattern, then, is to locate primary stress within the final foot. The two constraints most directly responsible for locating primary stress in Sentani are Map(ω), given in (24), and AllRight(f_Hd), given with its counterpart, AllLeft(f_Hd), in (25). Map(ω), requires that each prosodic word have a prosodic word-level grid entry within its domain. The prosodic word-level grid entry must fall within a head foot.

AllRight(f_Hd) draws the head foot of the prosodic word toward the prosodic word’s right edge; AllLeft(f_Hd) draws the head foot of the prosodic word toward the prosodic word’s left edge. Because the primary prominence must always fall within the head foot, restricting the position of the head foot also restricts the position of the primary prominence.

As the tableau in (26) illustrates, the combined effect of AllRight(f_Hd) and Map(ω) is to enforce mapping of the final foot in particular. AllRight(f_Hd) and Map(ω) can be satisfied simultaneously only when the final foot is mapped. When the prosodic word maps to a prosodic word-level grid entry that is over a head foot in final position, as in candidate (26w), both constraints are satisfied. The mapping status of the penultimate foot is of no concern. AllRight(f_Hd) is only violated if the head foot is not the final foot, as in candidate (26b), and Map(ω) is only violated if the prosodic word does not map to a prosodic word-level entry, as in candidate (26a). In contrast, the effect of Map(f) is more general. Map(f) requires every foot to map to the metrical grid. If either the final foot is left unmapped, as in candidates (26a) and (26b), or the penultimate foot is left unmapped, as in candidate (26w), Map(f) is violated.

To streamline the discussion, we will assume from this point on that Map(ω) is undominated, and we will not consider candidates without a primary prominence. Given this situation, a high-ranking AllRight(f_Hd) can be used to ensure that the obligatory primary prominence occupies the final foot and that the final foot cannot be unmapped.

As the tableau in (27) demonstrates, the six-syllable pattern of the Sentani variation emerges under two key ranking conditions. The first condition is that NonFin(x_f) and AllRight(f_Hd) both dominate AllRight(σ_Hd). Ranking NonFin(x_f) above AllRight(σ_Hd) excludes the option of a stressed final iamb, as in candidate (27d), and ranking AllRight(f_Hd) above AllRight(σ_Hd) excludes the option of shifting the primary stress off of a final iamb so that the final iamb can remain stressless, as in candidate (27c). The second key ranking condition is that AllRight(σ_Hd) and *Clash both dominate Map(f). Ranking AllRight(σ_Hd) above Map(f) excludes the option of a thoroughly trochaic pattern, as in candidate (27b), and ranking *Clash above Map(f) excludes the option of tolerating a clash configuration, as in candidate (27a). In the optimal candidate (27w) a stressless iamb precedes a stressed final trochee. This allows (27w) to maintain mostly iambic footing and primary stress on the final foot while avoiding clash and stress on the final syllable.

While the ranking conditions demonstrated in (27) produce the correct result for six-syllable and longer forms, they are unable to produce the correct result for four-syllable forms. The ranking conditions would leave the penultimate foot stressless in four-syllable forms as they do in longer even-parity ones. Since the penultimate foot is initial in four-syllable forms and non-initial in six-syllable and longer forms, however, we can obtain the correct result with a constraint that requires initial feet to be stressed.

5. The Sentani variation: four-syllable forms

The constraint family that can require stress on initial elements is the Initial Prominence family. InProm(x_f, f), given in (28), is the Initial Prominence constraint requiring stress on prosodic word-initial feet. InProm(x_f, f) insists that some foot-level grid entry coincide with the initial foot of every prosodic word.

Adding InProm(x_f, f) to the constraint ranking in (26), ranked above *Clash, is sufficient to produce the correct results for four-syllable forms.

The tableau in (29) demonstrates the result of ranking both AllRight(σ_Hd) and InProm(x_f, f) above *Clash for a four-syllable form. Note that the tableau does not consider output candidates that fail to maintain primary stress on a final trochaic foot, as required by the ranking NonFin(x_f), AllRight(f_Hd) ≫AllRight(σ_Hd). Ranking AllRight(σ_Hd) above *Clash excludes the option of a thoroughly trochaic pattern, as in candidate (29b). Ranking InProm(x_f, f) above *Clash excludes the option of a stressless initial/penultimate foot, as in candidate (29a). The optimal candidate (29w) tolerates a clash to ensure that the initial foot is iambic and stressed while also ensuring that the final foot contains a primary stress that avoids the final syllable. (The low-ranked Map(f) is included in the tableau to highlight the fact that it plays no role in this context.)

A Hasse diagram of the constraint ranking responsible for the Sentani variation is given in (30):

6. Binary vs. ternary

A reviewer states that the default pattern of Sentani is consistent with a ternary pattern (with nonfinality), and that it provides support for internally layered ternary (ILT) feet (Martínez-Paricio & Kager Reference Martínez-Paricio and Kager2015). While we agree that the available data for the default pattern is consistent with a ternary pattern with nonfinality, the ILT approach is not actually capable of producing this pattern or similar key patterns.

As the reviewer states, the available data for the Sentani default pattern could be analysed in terms of amphibrachs – ILT feet with left-aligned internal iambs. Forms that do not divide evenly into ternary feet would use binary feet to exhaustively parse leftover syllables. One iamb would be employed at the left edge in $3n+2$ forms, and two in $3n+1$ forms. In four-syllable forms a final iamb would be replaced by a final trochee to avoid final stress. The structures in (31) illustrate the ternary foot analysis for forms of up to seven syllables – the longest in available data. The proposed binary Weak Bracketing structures are repeated alongside them for comparison.

The predictions of the binary and ternary analyses would start to diverge beginning with nine-syllable forms. The binary pattern described by Elenbaas is given in (32) with the proposed Weak Bracketing structures.

The ternary pattern described by the reviewer is given in (33) with ILT structures. Note that the predicted stress patterns differ in forms with nine syllables (compare (32b) with (33b)), and in all forms with eleven or more syllables (compare (32d)–(32f) with (33d)–(33f)).

Martínez-Paricio & Kager (Reference Martínez-Paricio and Kager2015: Supplementary Material, p. 10) identify a version of the ternary pattern in (31) and (33) without nonfinality (a version where four-syllable forms have two iambs) as one of the predictions of their ILT account, and they indicate that it is unattested. The pattern is entirely plausible. It is similar to the ternary default pattern of Chugach (Leer Reference Leer1985a,Reference Leerb,Reference Leerc), where the iambs would occur to the right of the amphibrachs rather than to the left (Martínez-Paricio & Kager Reference Martínez-Paricio and Kager2015: 483 and Supplementary Material, p. 10; Martínez-Paricio & Kager Reference Martínez-Paricio and Kager2016).

This may seem like good news for the ILT framework. It is not. In considering forms long enough to see the difference between the binary pattern described by Elenbaas and the potential ternary alternative, it becomes clear that ILT cannot actually produce the ternary alternative. The distance-insensitive alignment constraints that ILT uses to produce directionality effects simply break down at longer lengths.

One of the advantages claimed for layered feet is that they can replace unparsed medial syllables in patterns that would require them in standard Weak Layering accounts. In standard Weak Layering accounts (Itô & Mester [Reference Itô and Mester1992] Reference Itô and Mester2003), it is necessary to directly influence the position of medial feet relative to medial unparsed syllables in order to distinguish between the trochaic patterns in (34a) and (34b), for example. Influencing the position of medial feet directly requires distance-sensitive alignment constraints (see Alber Reference Alber2005 and Hyde Reference Hyde2012a for discussion).

Internally layered feet are claimed to eliminate the need for medial unparsed syllables. The pattern in (34a) is recreated in (35a) with a dactyl at the left edge of the form, and the pattern in (34b) is recreated in (35b) with an amphibrach at the right edge. There is no need to position medial feet directly. All that is needed is to position a peripheral ternary foot of the right type.

Since ternary feet are thought to eliminate the need for unparsed medial syllables, then, ternary feet are also thought to eliminate the need to influence the position of medial feet directly and, therefore, the need for distance-sensitive alignment constraints.Footnote ⁵ The possibility of eliminating distance-sensitive alignment is supposed by some to be quite desirable, because the violations for distance-sensitive constraints can increase quadratically as the length of the form increases (though it has yet to be demonstrated that this actually has any practical empirical consequences).

The two constraints that are most directly responsible for locating feet in the ILT framework are given in (36) in the Relation-Specific Alignment formulation (see §3). In (36), f_T is a ternary foot, f_B a binary foot and f a foot of any size.Footnote ⁶

Since they omit the separator category in the definition of the locus of violation, both constraints are distance-insensitive: they assess violations for the simple fact of misalignment rather than measuring degrees of misalignment. Ternary-R assesses a single violation for each ternary foot that has a foot (of any size) to its right. Binary-L assesses a single violation for each binary foot that has a foot (of any size) to its left.

As indicated in (37) and (38), Ternary-R and Binary-L correctly locate the three feet needed to parse seven-syllable and eight-syllable forms. The two constraints are best satisfied when a ternary foot occurs at the right edge and a binary foot at the left edge. It is not necessary to influence the position of the medial foot directly. In the seven-syllable form in (37), there is one ternary foot and two binary feet. In the desired winner, the two binary feet precede the ternary foot. Ternary-R is the key constraint in this case. It locates the single ternary foot at the right edge, forcing the two binary feet to the left. Binary-L is also satisfied as well as possible given this particular collection of feet.

In the eight-syllable form in (38), there are two ternary feet and one binary foot. In the desired winner, the binary foot precedes the two ternary feet. Binary-L is the key constraint in this case. It locates the single binary foot at the left edge, forcing the two ternary feet to the right. Ternary-R is also satisfied as well as possible given this particular collection of feet.

In general, the approach works as long as there is no more than one type of medial foot. In other words, it works as long as the form contains no more than one ternary foot or no more than one binary foot. Whenever there is more than one of both types, however, both types must occur medially, and Ternary-R and Binary-L are unable to fix the position of the medial feet.

The effect is apparent in $3n+1$ forms that have at least ten syllables. To produce the correct pattern in a ten-syllable form, for example, two binary feet would need to precede two ternary feet. Since they cannot influence the position of medial feet, however, Ternary-R and Binary-L cannot actually locate them in the correct positions. As the tableau in (39) demonstrates, the candidate where the medial trochaic foot precedes the medial binary foot ties with the desired winner, in which the medial trochaic foot follows the medial binary foot. Note that we would not be able to distinguish between the candidates by considering additional constraints. The two candidates perform identically on every constraint in the ILT constraint set.

The problem arises in every $3n+1$ form with ten syllables or more. ILT will be unable to choose between two options for ten-syllable forms, among three options for 13-syllable forms, among four options for 16-syllable forms, and so on.

Some might be tempted to say that this circumstance has little practical consequence, since the longest Sentani forms available have just seven syllables. It should be kept in mind, however, that this circumstance also prevents ILT from producing the Chugach pattern, and Chugach does have forms of the relevant lengths.

ILT’s distance-insensitive constraints, then, are inadequate for implementing the potential ternary analysis of Sentani and similar patterns. While ILT could simply adopt distance-sensitive constraints, doing so would mean abandoning one of the most significant motivations for positing internally layered feet in the first place. ILT’s proponents have pointed to other virtues, of course, but these virtues are typically not unique. Much of the supporting evidence presented for internally layered feet is based on segmental processes that have been acknowledged to be captured by overlapping binary feet as well (Martínez-Paricio Reference Martínez-Paricio2013; Martínez-Paricio & Kager Reference Martínez-Paricio and Kager2016). In many cases, the relevant processes were first discussed in the context of overlapping binary feet (Hyde Reference Hyde2002). It is also possible to point to additional deficiencies of ILT. ILT is generally even less restrictive than standard Weak Layering, and it suffers from the same types of pathological predications (Hyde Reference Hyde2016b, Reference Hyde2025).

While ILT is unable to execute the potential ternary analysis of Sentani, there is ample evidence that Elenbaas’s description of Sentani as a binary pattern is correct in any case. The Sentani default pattern can be perturbed by the avoidance of stress on schwa in open syllables and by a limited weight sensitivity. Neither phenomenon is straightforward, and the interaction between them is even less so. Trying to provide an analysis would take us far beyond the scope of this squib, but it is worth noting that the perturbations nearly always result in forms that unambiguously exhibit a binary pattern. Occasionally, the results are ambiguous, but they are never unambiguously ternary. What follows provides several examples in which perturbations reveal the binary pattern of Sentani stress placement.

In Sentani, secondary stress tends to avoid occurring on a schwa in an open syllable. In forms with four or more syllables, for example, a secondary stress occurs on the peninitial syllable in the default pattern. To avoid stress on schwa in an open syllable, however, the default penintial stress can shift to an initial syllable without a schwa, as in (41).

If both the first and second syllables are open with schwa, then the shift fails to occur, as in (42). Crucially, the stress does not shift to the third syllable even when doing so would avoid schwa without producing a clash, as in the six-syllable (42b). To parse a six-syllable form, the binary analysis would employ three binary feet, but the ternary analysis would employ two ternary feet. The explanation for the failure to shift is straightforward if the stress is always confined to a binary foot at the left edge, as the binary analysis of six-syllable forms assumes. The explanation must be more subtle, however, if the stress occurs in a ternary foot at the left edge, as the ternary analysis of six-syllable forms assumes.

When stress is able to shift from second to first syllable in six-syllable forms, the results further support binary footing. As (43) illustrates, a secondary stress that would occur on the second syllable in the default pattern shifts to the initial syllable to avoid schwa. In addition to the stress that is shifted, however, an entirely new stress missing from the default pattern appears on the third syllable. The result is straightforwardly consistent with binary footing and straightforwardly inconsistent with ternary footing.

Unlike secondary stress, primary stress seems never to shift to avoid schwa. In the forms in (44), the primary stress remains on the penult despite the possibility of avoiding schwa by shifting to either the right or the left. On first thought, this is unexpected. If secondary stress is incompatible with schwa, then primary stress should be even more so. Under both the binary analysis and the ternary analysis, the effect of Nonfinality, discussed above, straightforwardly explains the inability of primary stress to shift from the penult to the ultima. Under the binary analysis, confining primary stress to a final binary foot also straightforwardly explains its inability to shift from the penult to the antepenult. Under the ternary analysis, however, if the primary stress is confined to a final ternary foot, as it would be in the examples in (44), there would have to be a more subtle reason for its inability to shift from the penult to the antepenult.

Finally, we turn briefly to Sentani’s limited weight sensitivity. Final heavy syllables always receive the primary stress, and initial heavy syllables always receive at least a secondary stress. While providing an analysis of Sentani’s weight sensitivity would take us well beyond the scope of this squib, it is worth noting that the perturbations due to syllable weight always result in patterns that are consistent with binary feet. In three-syllable forms, for example, whether the final syllable or the initial syllable is heavy, the result always has two stresses, indicating two binary feet, rather than the single stress we would expect with a ternary foot.

While the data we have for the Sentani default pattern are consistent with either a binary or a ternary analysis, then, forms where the default pattern has been perturbed suggest that a binary analysis is correct. Of course, it may be possible for an especially clever ternary approach to supply binary feet for the necessary cases in Sentani. (ILT is ruled out as a candidate analysis by its inability to produce the default pattern.) However, the prima facie case for the binary analysis is a strong one.