In this book, John Frampton sets out to reclaim the field of reduplication for derivational phonology. Using an impressive trove of data and detailed case studies, he presents a novel theory of reduplication (Distributed Reduplication or DR) and demonstrates how it can analyze a wide range of patterns of reduplication. He also exposes serious shortcomings in previous analyses, especially those in the framework of Optimality Theory (OT; Prince & Smolensky 1993/Reference Prince and Smolensky2004). Frampton's monograph makes the strongest and most persuasive case to date for a derivational analysis of reduplication, which is likely to be fully developed in the near future. Overall the author succeeds in pointing out serious inadequacies of many previous analyses, and convincingly demonstrates that DR can provide clever and elegant solutions for a wide range of problems.

However, Distributed Reduplication does not attain all of its goals. Important empirical problems remain, with some patterns escaping analysis, and others crucially relying on ad hoc stipulations. Frampton's critique of OT-based theories misses the versatility which allows OT to handle the difficult cases that he tackles with DR. Since DR fails to harness the explanatory power that OT at its best achieves, the die-hard OT phonologist will not be persuaded that the OT framework needs to be scrapped, although some cherished assumptions may need to be jettisoned.

Frampton's essential idea is that reduplication is ‘distributed’ in the sense that it is the surface result of three entirely separate operations: (i) juncture insertion (JI), (ii) transcription, and (iii) No Crossing Constraint repair (NCCR). JI is the only morphological step in the process; transcription and NCCR take place in the phonology.

Frampton first illustrates the workings of DR by analyzing reduplication in the Australian language Mangarayi, where the pattern jimgan ‘knowledgeable one’→jimgimgan ‘knowledgeable ones’ is found. Juncture insertion adds special markers called transcription junctures or t-junctures (typographically denoted with square brackets) before the first vowel and before the second vowel respectively. Transcription copies the timing slots which occur within those junctures, retaining the links to segments in the original and new timing slots; it then removes the t-junctures. NCCR duplicates the segments in order to allow delinking of impossible doubly-linked segments. (Although the terminology suggests that Frampton is referring to the well-known ‘No Crossing Constraint’, NCCR actually denotes Retraction Condition repair, a similar but distinct new operation. The term NCCR is used ‘to maintain the familiar terminology’ (32), which unfortunately is likely to mislead the casual reader.) An autosegmental derivation of the Mangarayi word jimgimgan is shown in (1).

1. (1)

   

By assumption, junctures must be paired by the time transcription occurs, but rules are not limited to inserting paired junctures. When a rule inserts only one juncture, its complementary juncture will be inserted by default closure at the opposite edge of the morphological domain. (Why default closure occurs precisely there is not explained.) This simplifies many rules. For example, the common case of left-edge C(onsonant)V(owel) reduplication will result from the rule in (2), which Frampton calls the C*V-rule:

1. (2) Ø→]/V

Along with the t-junctures, Frampton introduces another set of special markers, the angled brackets < and >, which mark phonological material which has already been copied or is exempt from being copied. Frampton's claim that these markers are logically necessary is not fully persuasive, but their existence is plausible enough given the rest of his framework. Although these markers are ‘bookkeeping symbols’ used to make sure that reduplication does not continue indefinitely, they also provide a crucial means to account for some patterns which would otherwise be difficult to explain. For example, reduplication in Chukchee involves forms like those in (3).

1. (3) Root.bsolutivesingular

   nutenute-nut‘earth, ground’

   inuinu-in(part of a reindeer leg)

   jilʔejilʔe-jil‘gopher’(54, ex. (74))

The generation of, for example, nute-nut follows, given right-transcription of a form like [nut<e>]. (Production of this pre-transcription form requires the use of more complicated goal-driven rules, which are discussed below.)

Frampton's basic model is simple; in fact by itself it is too simple. Since JI operates only over timing slots, something else is needed in the many cases where reduplicant shape appears to be prosodically determined. To solve this problem it is essential to realize that t-junctures (and the <>markers) are real phonological elements subject to manipulation by rules, especially by the prosodic adjustment rules, which do a great share of the work required to yield proper outputs. These rules are not given as a basic element in DR, but they assume a critical role in reduplicative analysis. Frampton illustrates this with the case of Ilocano, where left-edge heavy-syllable reduplication occurs. Here, the C*V rule applies in the morphology and is followed by a prosodic adjustment rule which extends the closing juncture rightwards until the material between the junctures constitutes a heavy syllable.

A further elaboration which often proves essential in Frampton's analyses is the goal-driven rule (GDR). Representing a Sound Pattern of English-style rule as , and g for some derivational constraint, Frampton defines a GDR as follows:

1. (4) if α does not satisfy g, , and β does satisfy g.

   (91, ex. (156))

For example, to produce the Chukchee examples above, we need to assume that JI involves the following two insertions:

1. (5)

   1. (a) Ø→[/__X

   2. (b) Ø→</ V__(111, ex. (209))

The first rule on its own would produce total reduplication, due to default closure which will occur at the right edge of the stem. The second rule inserts the marker < after the first vowel, with default closure also inserting the marker > at the right edge of the stem so that anything after the first vowel will be exempt from reduplication. To get the attested pattern, a further rule is needed. The generalization that emerges is that the copied portion is always bimoraic. This is incorporated into the following goal-oriented rule (91):

1. (6) second conjunct; bimoraic::<-right

The rule in (6) has the effect of moving the <to the right until the copied portion is bimoraic. Thus, the derivation of Chukchee jilʔe-jil ‘gopher’ will proceed as in (7).

1. (7) jilʔe

   JI[ji<lʔe]

   Default Closure[ji<lʔe>]

   Prosodic Adjustment [jil<ʔe>]

The structure will then undergo transcription and straightforwardly yield the attested output jilʔe- jil. (Multiple rules may also be associated with the same goal, in which case the rules will apply in a Paninian ordering and stop when a rule applies or when the end of the list is reached.)

The use of these GDRs narrows the distance between DR's derivational framework and one-step phonological theories like OT. Frampton is at pains to emphasize differences between the theories, such as the fact that goals are tied to particular rules, but he does not flesh out the empirical differences between a derivational theory with GDRs and a one-step theory like OT.

The issue is further complicated by the fact that Frampton is forced to allow extra constraints that limit the application of rules in order to account for cases such as that found in Yaqui, which has the following reduplicative ‘core paradigm’:

1. (8) Stem.eduplicated.loss

   1. (a) chikechi-chike‘comb one's hair’

   2. (b) chuktachuk-chukta‘cut with a knife’

   3. (c) kaateka-kaate‘walk’(108, ex. (202))

The forms in (8b) and (8c) suggest a heavy syllable reduplicant like the one in Ilocano, which Frampton derived by means of the C*V-rule and prosodic adjustment for a bimoraic copy. The problem is the form in (8a), in which a light syllable copy surfaces instead of the plausible but unattested form *chik-chike. The only way out is to constrain the prosodic adjustment rule. Frampton proposes a condition called Onset Permanence, which forbids prosodic adjustment to recruit onset segments into non-onset roles. Although this is a plausible condition, its use in DR opens a Pandora's box (as Frampton acknowledges): if output constraints like Onset Permanence are part of derivational phonology, then other output constraints may likewise be admitted, and the derivational nature of the system is endangered.

Notwithstanding these complications, DR remains attractively simple while accomplishing a great deal. This is demonstrated in Chapter 7, ‘Case studies’, which takes up nearly half of the book's pages. The case studies drawn from eleven languages illustrate the mechanics of DR in admirable fashion and show the range of patterns that the theory can analyze. The lengthy discussions of Washo and Sanskrit reduplication are particularly impressive and raise the bar high for future analyses of those patterns in any framework.

The problems that Frampton identifies with other analyses are genuine and should cause some serious rethinking on the part of the authors of these analyses and those who have relied on their work. For example, he shows that applied passive formation in Bantu presents a crucial problem for the Morphological Doubling Theory of Inkelas & Zoll (Reference Inkelas and Zoll2005). DR, on the other hand, adequately derives the complex attested patterns.

Frampton makes some strong points in his criticism of OT. For example, expanding on a point made in Raimy (Reference Raimy2000), he draws attention to the fact that the most common OT framework for reduplication makes various unwelcome predictions. Thus, a reranking of the constraints that McCarthy & Prince (Reference McCarthy and Prince1995) use in their analysis of Nuu-Chah-Nulth straightforwardly predicts the pattern in (9), which does not occur in natural language (207f.):

1. (9)

   1. (a) da ti kun→datiku-datikun

   2. (b) da tin kun→dati-datikun

   3. (c) dan ti ku→da-dantiku

All that would be necessary for this hypothetical pattern to emerge is a ranking of Faith-I(nput)O(output) >>No.oda >>Faith-B(ase)R(eduplicant). However, it can be argued that this problem arises due to particular assumptions made by McCarthy & Prince (such as the abstract red morpheme and base–reduplicant correspondence), and, as shown in Saba Kirchner (Reference Saba Kirchner2010), an OT theory of reduplication which eschews these assumptions will avoid the problem altogether.

Frampton rightly observes that ‘one of the more bizarre aspects of OT's conception of the language faculty is that a complex calculation is needed to move dog out of the lexicon and into a phonological calculation’ (206). But this is a dangerous line of critique, given that DR is also often driven to complex analyses that lack clear justification. Consider Yoruba reduplication, which copies an onset and inserts a fixed i as in dun→di-dun ‘to be tasty, sweet’. In DR, this pattern assumes that there is a morpheme whose underlying form is [<i>- as well as a JI rule of the form Ø→]/__V. This is a relatively complex account for a reduplicative pattern which is, as Frampton notes, widespread. By contrast, the analysis of Yoruba that is offered by Alderete et al. (Reference Alderete, Beckman, Benua, Gnanadesikan, McCarthy and Urbanczyk1999) rests on a typical ranking of well-motivated constraints and allows the fixed segment to emerge as an unmarked vowel. Moreover, Alderete et al.'s (Reference Alderete, Beckman, Benua, Gnanadesikan, McCarthy and Urbanczyk1999) analysis explains why the fixed segment in Yoruba reduplication is the default vowel i and why fixed segments are so often relatively unmarked. For Frampton and DR, this fact remains purely coincidental.

By showing how many different patterns DR can account for, Frampton reveals a problem with the permissiveness of the framework. DR-analyses of reduplication patterns in Sino-Korean (113) and Kinande (137) are welcome on the assumption that these patterns are productive (although this appears somewhat doubtful). But when Frampton shows that two simple DR rules can account for Arrernte Rabbit Talk (64), he opens himself up to criticism because Rabbit Talk is not a pattern which freely occurs in natural language, and no phonological theory should place it on a par with well-attested reduplication patterns. The fact that DR can account for Rabbit Talk – a language game which moves a word-initial string to the end of a word, similar to Pig Latin in English – suggests that it is too permissive a theory, which lacks a clear explanation for the very restricted occurrence of phenomenona like Pig Latin or Rabbit Talk in natural language.

Although DR does not yet seem to have all the answers for the analysis of reduplication, there is no denying that Frampton's monograph makes an important contribution. The author's attention to detail and use of a great volume of data make this a work that deserves to be studied as much by those who will disagree with his conclusions as by those who are persuaded by this theory of Distributed Reduplication.