1.The separation of primary and secondary stress

Assuming that the phonemes that make up words are organized into syllables, modern linguistic theory claims that rhythm is a manifestation of the fact that syllables are further grouped into constituents called feet, which are usually binary groupings of syllables. Standard Metrical Theory takes the grouping of syllables into feet as fundamental for the assignment of bounded primary stress. StressTyp is based on a separation of the treatment of primary and secondary stress. Such a separation has been proposed and defended on theoretical grounds by van der Hulst (1996). We offer several arguments here.

First, in our discussion of bounded systems (Chapter 15, §2.1) we have seen that weight-sensitive primary stress locations within the two-syllable window can take four different forms (at each edge of the word). Rhythm types, however, fall into just two categories, due to the fact that a rhythmic foot cannot contain two heavy syllables (because every heavy syllable constitutes a rhythmic beat). This difference, then, may make it necessary to describe primary and secondary stress patterns separately.

A second argument for the separation can be found in systems in which secondary stress starts at the edge that is opposite to the location of primary stress; we refer to such systems as polar. An example is given in (1).

(1) Garrwa (Garawan; Northern Territory, Australia): 

initial main stress, secondary stress on the penult and alternates before it.
(ˈnari)ŋin(ˌmukun)(ˌjina)(ˌmira)	‘at your own many’

In Table 1 we see exactly how many languages this argument is based on (data in Tables 1-4, and Figure 1, are based on the complete StressTyp sample of 510 languages).

In total, 39 languages (8% of all the languages in StressTyp) assign main and secondary stress from opposite directions. It would seem that the subset of languages for which we need separate direction parameters for main stress and rhythm is large enough to provide serious support for the separation of primary and secondary stress. 

A third argument for the separation of primary and secondary stress is formed by languages in which the rules for main stress and rhythm do not agree in their usage of weight, or more intricately, in the way in which they use it. In Kayardild, for instance, main stress is initial, regardless of weight, but long vowels receive a secondary stress, whereas short vowels do not (simplified for expositional purposes). 

(2) Kayardild (Tangkic; Queensland, Australia) 

ˈkunja

‘little’

ˈmaˌlaa

‘sea’

Table 2 shows the number of languages in which main stress is weight-sensitive and secondary stress is weight-insensitive, or vice versa. The last column in Table 2 shows the languages in which main and secondary stress disagree in their definition of what constitutes a heavy syllable. In such languages, closed syllables may, for instance, be heavy for stress but not for rhythm. 

In total, we need separate quantity statements for primary and secondary stress in 57 languages (11% of all the languages in StressTyp). All in all 17% of the languages in StressTyp support the separation hypothesis (Tables 1 and 2 combined, 10 overlapping languages).

A fourth argument for separating primary and secondary stress assignment lies in the fact that whereas lexical marking is quite normal for primary stress, even in systems that have dominant rule-governed locations, secondary stresses are never a matter of lexical marking. In this statement, we ignore so-called “cyclic stresses”, i.e. secondary stresses that correspond to primary stress locations in embedded morphemes in complex words.

2. Defining the values

For the languages for which we have information on rhythm, the following types are shown on the map. 

Values of Map 17A. Rhythm Types

Go to map
	Value	Representation
	Trochaic: left-hand syllable in the foot is strong	153
	Iambic: right-hand syllable in the foot is strong	31
	Dual: system has both trochaic and iambic feet	4
	Undetermined: no clear foot type	37
	Absent: no rhythmic stress	98
Total:		323

2.1. Trochaic rhythm

The prototypical rhythm pattern stresses every odd syllable from the left in languages with initial main stress, and every even syllable from the right in languages with penultimate main stress. This is achieved with trochaic feet, as in Ono (Trans-New Guinea; Papua New Guinea): (ˈmesi)(ˌkene) ‘you will sit’, (ˈari)(ˌmage)(ˌake) ‘he always goes’.

2.2. Iambic rhythm

Languages with iambic rhythm typically stress even syllables from the left or odd syllables from the right, as in Mapudungun (Araucanian; Chile): (eˈlu)(muˌyu) ‘give us’, (kiˈmu)(faˌlu)(wuˌlay) ‘he pretended not to know’.

2.3. Dual rhythm

This type is difficult to find. One needs to look at all the data of a language carefully to discover that one foot type is not sufficient to cover all cases. An example of such a language is Yuwaalaraay (Pama-Nyungan; New South Wales, Australia): (gaˌlu)(maˈya) ‘grandson’, (ˈgidul)(ˌgara) ‘snail’.

2.4. Undetermined rhythm

These languages do have secondary stresses, but the patterning is totally unclear. Languages like Malayalam (Dravidian; India) which place secondary stress on long vowels only fall under this type: ˈpukavaɳʈi ‘train’, ˈmutaˌlaaɭi ‘boss’.

2.5. No rhythm

Languages that have main stress but no rhythm abound. An example is Gorowa (Southern Cushitic; Tanzania): giramˈbooda ‘snuff’, oroˈmila ‘because’. It remains an open question whether the “no rhythm” languages do have a foot type grouping but lack a clear phonetic manifestation of this grouping; in some cases such "silent" grouping can be detected because it conditions other aspects of the phonology of a language.

3. Geographical distribution

We observe the following tendencies: 

1. Iambic rhythm occurs mostly in North and South America. 

2. South America and Australia always seem to have clear rhythmic patterns. 

3. Africa, on the other hand, shows little evidence for rhythmic patterns. 

As was noted in Chapter 14, we find an unequal division of languages over the two foot types that we use to describe the rhythmic properties of stress systems. In general, languages prefer the trochaic foot over the iambic foot. Note again that we divide languages into trochaic or iambic systems on the basis of their secondary stress properties. Unfortunately, that means that many languages in StressTyp are left out of the equation, simply because we do not have any secondary stress data for these languages. Statistics for the languages for which we do know something about secondary stress can be found in figure 1.

The hugely skewed distribution between trochaic and iambic systems in Figure 1 is hard to miss even in a cursory look at the typology of stress systems. Though it has never been quantified in this way before, the scarcity of iambic systems has already led some researchers on the path to total denial of the iamb’s existence as an abstract metrical entity. Hayes (1995) dispenses with all quantity-insensitive iambs, while among the most drastic of these studies, van de Vijver (1998) claims that all iambic systems can be reanalysed as trochaic ones (cf. van der Hulst 1999b for discussion). In StressTyp we choose no sides in this obviously highly theoretical debate. We will continue to offer the iambic foot as an option, because there are enough languages, in our view, whose stress patterns are clearly more straightforwardly described with the help of these iambic feet.

4. Theoretical issues

4.1. Correlations between rhythm type and edge

A correlation that has been noted many times in the past concerns the foot type used in secondary stress assignment and the edge at which this assignment starts. The claim is that languages with an iambic rhythmical pattern assign rhythm from left-to-right, as in Mapudungun in §2.2 (cf. Kager 1993). An example that makes this clear is given in (3). 

(3) Mapudungun 

(eˈlu)(aˌe)new

‘he will give me’

Rhythm assignment from right-to-left would have given us eluˈaeˌnew. (In this example we note that a left-over syllable, here on the right side, is not footed and hence has no rhythmic beat; this is a quite common phenomenon, and it has been argued that “degenerate”, i.e. monosyllabic, feet are uncommon or universally banned in quantity-insensitive languages.)

The reverse claim, that languages with trochaic rhythm start footing at the right edge, is not commonly deduced from the observation with respect to iambic languages. Linguists seem hesitant to claim anything in this case. So, we present the data for both types in Table 3.

We observe that the correlation between iambic rhythm and left-to-right footing is indeed quite high. It turns out to be even higher when we remove some "suspicious" cases from the list of right-to-left iambic languages. It appears that in many of these languages the choice between iambic and trochaic feet is not easily made, and where the descriptive source mentions that the language is iambic, this choice is often based on superficial evidence. For only 3 of these cases can we be quite certain that rhythm is iambic.

For trochaic patterns, we note that most trochaic languages start foot assignment at the left instead of the right edge. Yet, the difference is not large enough to speak of a clear correlation in this case. We may conclude, therefore, that the choice to state a correlation between foot type and direction for iambic systems is justified only in hindsight, by the StressTyp data. 

4.2. Correlation between rhythm type and relevance of weight

Hayes (1995) suggests another correlation, this time between rhythm type and weight-sensitivity, the so-called iambic-trochaic law. According to this “law” weight-insensitive languages (grouping units of equal weight, so to speak) are (preferred to be) trochaic. The reverse claim, that weight-sensitivity leads to iambic grouping, is less obvious, although one idea is that weight-differences based on quantity favor iambic patterns, while weight differences based on “intensity” favor trochaic patterns. However, it is not clear what intensity is (prominence?); moreover, many weight-sensitive trochaic patterns are based on quantity (cf. Rice 1992, Kager 1993 and van der Hulst 1999b for discussion). The following table indicates the relationship between the presence of weight-sensitivity and the rhythmic foot type:

4.3. Iterative and non-iterative secondary stress

So far we have assumed that the rhythmic structure of a word comprises the whole word. However, rhythm assignment can be ‘non-iterative’, which means only one secondary-stress foot is formed, as in the Ngalkbun (Australian) example in (4); here only one secondary stress occurs at the opposite edge from the main stress, though the word is long enough to allow more.

(4) Ngalkbun

ˌŋaʔyenjyenjtjuˈŋiyan

‘I will talk’

StressTyp reveals that the iterative option is clearly favored. 70% of the languages for which we have secondary stress information choose this option, while only 18% reveal non-iterative rhythm (no iterativity information for 12%). For the sake of completeness we mention here that 6 of the iterative secondary stress languages show ternary stress (the rhythm beat is followed, or preceded, by two unstressed syllables); cf. Rice 1992.

4.4. Foot typology

The precise grouping of syllables into feet is subject to theoretical discussion. In the classical theory, one heavy and one light syllable could be grouped together in a foot, but in later versions of the theory (the so-called Revised Theory) heavy syllables form feet by themselves. A choice on this issue does not, however, influence whether a language is classified as iambic or trochaic. We refer to Hayes (1995) and van der Hulst (1999b) for discussion of this issue.