The key grouping structures relevant to metrical stress theory are the categories of the prosodic hierarchy. Prosodic categories can be divided into two types: interface categories and rhythmic categories. The interface categories are the utterance, the intonational phrase, the phonological phrase, and the prosodic word. The rhythmic categories are the foot, the syllable, and the mora. The key principles governing prosodic grouping are Constituency, Strict Succession, and Headedness. Constituency insists that prosodic groupings occur in the dominance relationship specified by the prosodic hierarchy. Strict Succession insists that phonological representations not skip prosodic categories moving lower to higher in hierarchy. Headedness insists that every instance of a prosodic category designate one of its immediate constituents as its head (its most prominent constituent). The combination of Headedness and Strict Succession insists that phonological representations not skip prosodic categories moving in the either direction, either lower to higher or higher to lower. Two special configurations play key roles in the theory: recursion and overlap. Interface categories may exhibit recursion, but recursion of rhythmic categories is prohibited by the Simple Layering condition. Instances of the same prosodic category may overlap so that they share a constituent.

As in music, stress and accent in natural language are phenomenal prominences. A phenomenal prominence is always the most salient aspect of an acoustic contrast. A stress or accent might consist of a higher pitch, a greater amplitude, or a longer duration. It might also arise from differences in aspiration, vowel quality, or voicing. The primary purpose of stress and accent is to indicate a form’s temporal structure. It does this by indicating the positions of metrical prominences on the metrical grid. When phenomenal prominences correspond to metrical prominences, as they do in both music and language, they indicate the locations of metrical prominences and overall temporal organization. The key difference between metrical patterns in music and metrical patterns in language is that the former are typically more cyclic – or repetitive – than the latter with a more even distribution of prominences. Metrical organization is always rich and constructed automatically. Even when presented with a series of identical isochronous pulses, a hearer will automatically construct an analysis with multiple metrical levels. Stress and accent indicate which metrical analysis a listener should construct. This typically requires minimal information. A single accent per form can distinguish between the four perfect grid patterns, the simplest binary metrical patterns.

There are two general well-formedness principles that shape the metrical grid: clash avoidance and lapse avoidance. Clash is a configuration on the metrical grid where two entries on one level do not have an intervening entry on the next level higher. Clash avoidance plays a key role in restricting the range of binary default systems predicted under a Weak Bracketing approach. It prefers patterns where overlapping feet share grid entries rather than mapping to separate entries. Clash avoidance is also plays a key role in producing rhythm rule effects, where prosodic word-level and phonological phrase-level prominences shift to create a more regular, even distribution. A lapse is a configuration where two adjacent entries on one level both fail to support an entry on the next level higher. Lapse avoidance plays a key role in creating the fine-grained prominence distinctions on the metrical grid that cannot be accounted for with MAP constraints. Finally, extended lapse is a configuration where three adjacent entries on one level all fail to support an entry on the next level higher. Extended lapse avoidance is the key factor in producing the few ternary patterns found among the world’s languages.

Two asymmetric constraint families help to shape the metrical grid at prosodic boundaries. The NONFINALITY family of constraints prohibits prominence at the end of a domain. NONFINALITY constraints produce a wide range of effects and provide a uniform account of phenomena that otherwise appear to be unrelated. One NONFINALITY constraint helps to produce the falling stress contour of compounds in English. Other NONFINALITY constraints reproduce traditional foot extrametricality effects. Still other NONFINALITY constraints make metrical prominence sensitive to syllable weight. They ensure that stress avoids light syllables, sometimes shifting stress to a heavy syllable and sometimes producing lengthening effects, both iambic and trochaic. Finally, a NONFINALITY constraint helps to introduce clash or lapse near the right edge of a form. The INITIAL PROMINENCE family requires prominence at the beginning of a domain. The main role that INITIAL PROMINENCE constraints have played in previous analyses is introducing clash or lapse at the left edge of a form. Together, the NONFINALITY and INITIAL PROMINENCE constraints are responsible for the asymmetries found in the typology of word-level prominence patterns. They introduce clash and lapse configurations near the edges of prosodic words, allowing the grammar to produce patterns beyond the perfect alternation patterns.

A variety of mechanisms for producing directional orientations have been proposed in the literature. Two serial approaches are the Perfect Grid Rules of Prince (1983) and the Parametric Foot Construction approach of Hayes (1995). The Perfect Grid Rules are implemented in a grid-only framework. They construct the metrical grid entry by entry working either from left to right or from right to left. The Parametric Foot Construction Rules perform a parallel function in a foot-based framework. They build feet from syllables one at a time working either from left to right or from right to left. In a parallel approach like Optimality Theory, the grammar cannot create structures step by step. It cannot start at one edge and work towards the other. Alignment constraints allow parallel approaches to reproduce the directional effects of the earlier serial approaches. Alignment constraints accomplish this by discouraging misalignment between domain edges. Distance-sensitive Alignment takes degree of misalignment into account. It assesses a greater number of violations as the degree of misalignment increases. Distance-insensitive Alignment does not take degree of misalignment into account. It assesses the same single violation for each instance of misalignment, regardless of degree.

The foot is the key prosodic category involved in constructing stress and accent patterns. There is compelling evidence for binary feet – feet that are either disyllabic or bimoraic – but no compelling evidence for feet larger than two syllables. Feet always have syllables as their constituents. While some proposals involve mora-based footing, building feet on moras unnecessarily flouts fundamental restrictions on the prosodic hierarchy. Weak Layering approaches, including the Layered Foot approach, are susceptible to the Odd-Parity Input Problem. The Odd-Parity Input Problem is a set of pathological predictions that arise from the need to achieve exhaustive binary parsing in odd-parity forms. It has two sub-problems. The first is the Odd Heavy Problem, a quirky type of quantity sensitivity where exhaustive binary parsing is achieved in odd-parity forms by parsing a single odd-numbered heavy syllable as a monosyllabic foot. The second is the Even Output Problem, where an odd-parity input is converted to an even-parity output to achieve exhaustive binary parsing. Unlike Weak Layering approaches, Weak Bracketing is not susceptible to the Odd-Parity Input Problem.

Chapter 9 summarizes the main points addressed in previous chapters. The main issues addressed in Chapter 1 are phenomenal prominence, metrical prominence, and the relationship between them. Chapter 2 addresses the Prosodic Hierarchy and structural prominence. Chapter 3 examines the typology of word stress. Chapter 4 examines two correspondence relationships: the relationship between prosodic categories and grid entries and the relationship between syntactic categories and prosodic categories. Directionality effects are addressed in Chapter 5, and grid well-formedness are addressed in Chapter 6. Chapter 7 examines boundary effects, and Chapter 8 focuses on feet.

Default prominence patterns divide into two types: fixed and proportional. In fixed patterns, the number of prominences is constant regardless of the length of the form. Fixed patterns can be single (one prominence per form) or dual (two prominences per form). In proportional patterns, the number of prominences increases as the length of the form increases. Proportional prominence patterns can be either binary (prominence on every second syllable) or ternary (prominence on every third syllable). Phenomenal prominence systems may be either fixed or proportional. Metrical prominence patterns are always proportional. Metrical patterns are typically binary, very rarely ternary. All single and dual phenomenal patterns indicate binary metrical patterns. The simplest binary metrical patterns are the four patterns that exhibit perfect alternation. While each of the four perfect alternation patterns is attested, only a few of the possible departures from perfect alternation are attested. The pattern of attestation exhibits asymmetries that become clear when considering iambic-trochaic mirror image pairs. At most one member of such pairs is attested. With only four reasonably clear examples, ternary stress patterns are extremely rare, and is it impossible to make any significant generalizations about them with any degree of confidence.

The relationship between the metrical grid and the prosodic hierarchy and the relation between prosodic structure and syntactic structure are both relationships and relations of Correspondence. Correspondence is a representational link between two representational objects. Entries on the metrical grid and instances of prosodic categories may correspond, and instances of prosodic categories and instances of syntactic categories may correspond. Mapping is the correspondence relation between instances of prosodic categories and entries on the metrical grid. The mapping relation is one of the key factors influencing the grid’s construction. Mapping is governed by a handful of key principles, including Hierarchy Coordination. The prosodic hierarchy and the metrical grid are both hierarchies and they map to each other as hierarchies. Mapping is required by the violable MAP family of constraints, constraints that require prosodic categories to map to grid entries. The MATCH family of constraints requires faithful correspondence between prosodic categories and syntactic or morphological categories. It requires both that the correspondence relation exist and that that correspondents share key elements. Simple MATCH constraints require correspondents to have exactly the same set of terminal elements. LexMatch constraints require correspondents to have the same set of lexical terminal elements. LexMatch constraints ignore functional terminal elements.