Majority Rule in Harmonic Serialism

Andrew Lamont

doi:10.3765/amp.v7i0.4546

Abstract

Majority Rule is an unattested process where agreement is controlled by the largest class in the input. As a function from inputs to outputs, Majority Rule requires more computational expressivity than do attested phonological transformations. This paper examines how Majority Rule arises in parallel Optimality Theory and Harmonic Serialism. It is shown that in HS, Majority Rule relies on globally evaluated output constraints, which are known to produce computationally complex pathologies. However, without them, HS is unable to produce iterative harmony at all. We propose adopting directional constraint evaluation in HS as a way of modeling harmony while maintaining local representations.

Highlights

Majority Rule is an unattested process where agreement is controlled by the largest class in the input (Lombardi, 1999; Bakovic, 2000). (1) illustrates Majority Rule with a running example of sibilant harmony; here the classes are [+anterior] sibilants, such as /s/, and [−anterior] sibilants, such as /S/
This paper examines how Majority Rule arises in parallel Optimality Theory (Prince & Smolensky, 1993/2004) and Harmonic Serialism (HS) (Prince & Smolensky, 1993/2004; McCarthy, 2000)
If locus replacement improved on the motivating output constraint by swapping loci for less offensive loci, harmonic bounding would no longer be an obstacle to harmony

Summary

Introduction

Majority Rule is an unattested process where agreement is controlled by the largest class in the input (Lombardi, 1999; Bakovic, 2000). (1) illustrates Majority Rule with a running example of sibilant harmony; here the classes are [+anterior] sibilants, such as /s/, and [−anterior] sibilants, such as /S/. A disharmonic input cannot surface faithfully, and must either be mapped onto an output with only [+anterior] sibilants or one with only [−anterior] sibilants. The choice between these mappings depends only on how many members each class has in the input (represented in (1) by m and n): whichever class is larger will surface. As a function from inputs to outputs, Majority Rule requires more computational expressivity than do attested phonological transformations (Riggle, 2004; Gerdemann & Hulden, 2012; Heinz & Lai, 2013). Any model that is only expressive enough to compute attested phonological transformations cannot produce Majority Rule.

Majority Rule in parallel Optimality Theory

The scope of the problem

Directional constraint evaluation in Harmonic Serialism

Conclusion and future directions