Shape arises from structure: Dispersedness through con-

decide between the two phonemes even if the input she receives may be mapped to either one. She does so in a ratio analogous to the distribution ratio in her input.

What is important about the notion of inventory espoused in Boersma and Hamann (2008), is that it strikes a balance between the functional goals of

’Ease of Articulation’ and ’Perceptual Clarity’, without using teleological or imprecisely defined constraints. Following Boersma and Hamann (2008) in their application of Occam’s Razor, I propose that Parallel Bidirectional Phonology and Phonetics is preferable to Dispersion Theory.

In the above mentioned study, Boersma and Hamann make a number of simplifying assumptions. Most of these are relatively uncontroversial, but in paragraph 5.3, one is made that reveals one of the most serious challenges for PDBB as a theory about inventories. A telling quotation is given below:

We describe here the situation when a child already has correct lex-ical representations, but not yet an adult-like prelexlex-ical perception.

That is, she already knows which lexical items have an underlying /S/ and which have an underlying /s/...

In other words, PBPP can predict the phonetic contours of phonemes, but it cannot independently decide which phonemes are part of a language’s in-ventory, and which are not. In order to reach a stable state, the number and labeling of phonemes must be known to the learner. The phonetic identities of members of the inventory are determined by constraints interacting on the /surface form/, [auditory form] and [articulatory form]. Thus, although the shape of the inventory is emergent and non-teleological, PBPP leaves open the question of phonological representations: inventories of abstract entities seg-ments that are composed of features. The structure of the inventory is not the subject matter of Boersma and Hamann (2008). Because PBPP operates on an inventory of phonemes whose members are determined elsewhere, it is still a holistic theory: given a phonemic inventory, it can predict its phonetic shape. It cannot independently predict the phonological identity of the inven-tory members. Note, however, that it is very well possible to integrate the theory of Feature Co-occurrence Constraints developed in the current thesis into the Parallel Bidirectional Phonetics and Phonology model. This will be pursued further in chapter 5, after we have properly investigated the Feature Co-occurrence Constraint theory. It should be mentioned, however, that the GLA has been successfully applied to modelling real world language acquisi-tion (Boersma & Levelt, 1999, for example).

2.2.3 Shape arises from structure: Dispersedness through

Theory does not concern itself with the structure of the inventory (in fact, it is questionable whether the structure of the inventory can be described in DT, with its emphasis on phonetic representations and antagonistic attitude towards input-output mappings), it aims to describe dispersion effects, rather than derive or explain these. Parallel Bidirectional Phonetics and Phonology, on the other hand, distances itself from the overly teleological perspective of DT, and demonstrates that dispersion effects can be seen to emerge once a rich, integrated model of phonetics and phonology is assumed. Still, however, it does not say much about the structure of the inventory, and in fact, must pre-suppose knowledge of phonemic categories in the learner.

Taking a different route, Hall (2011) shows that dispersion effects arise naturally and predictably when we mix two pre-existing theoretical concepts:

Contrast and Enhancement. The theory of phonetic enhancement (as expressed in, for example, Keyser & Stevens, 2006) holds that, while underlying phono-logical representations are specified only for a limited, language-specific set of features (that is, languages select a sub-set from the set of universally avail-able features), in the course of speech production these are supplemented with features that enhance the features of segments that are otherwise in danger of losing perceptibility. As an example, let us assume a three-member vowel in-ventory /a, i, u/. Here, /u/ need only be specified for [+back], and rounding is supplied later to enhance the effect of backness. Similarly, speakers of English add rounding to /S/ to enhance the effect of [-anterior].

The problem with enhancement theory is that it does not, by itself, supply a principled means to decide which features are primary, phonological, and stored lexically, and which features are enhancements. At the same time, the Modified Contrastive Hierarchy has no means of deciding by itself how redundant (non-contrastive) features are filled in at the surface. Hence, Hall (2011) proposes to combine the two, according to the following schema (Hall, 2011, example 10):

(21) Elements of a theory of contrast and enhancement

a. Phonological feature specifications are assigned by the Successive Division Algorithm

b. Only these contrastive feature specifications are phonologically ac-tive

c. In phonetic implementation, redundant properties of segments tend to be filled in in ways that enhance the auditory impression of their contrastive features

d. Phonetic enhancement is variable across languages, speakers and contexts,and the distinctness of phonemes is sometimes reduced by other factors, such as auditory overlap (Stevens & Keyser, 2010,§4) Hall (2011) then proceeds to list some ways in which (21c) can take place, ranging from the amplification of the phonetic dimensions inherent in a feature specified as per (21a), to enhancing a contrastive (and thus lexically specified)

feature’s phonetic correlate by a separate phonetic event that increases the salience of the first.

To illustrate how the contrast and enhancement approach derives dispersed inventories, a very simple demonstration suffices (but see Hall, 2011 for many more). Consider again the inventory /i a u/, which we discussed in section 2.1.3 above. To characterise this inventory, two binary features are minimally neces-sary (and, since we are dealing with contrastive specifications, also maximally).

Two features means two ways of ordering divisions, which means two possible hierarchies. Example (18) is repeated here in (22) below, for clarity.

(22) Possible specifications for /i a u/

a. [high] >> [round]

i a u

[high] + - +

[round] - +

b. [round] >> [high]

i a u

[round] - - + [high] +

-Now consider a sub-optimally dispersed inventory /1 9 0/. The question is, why such an inventory is unattested. If we apply the Successive Division Algorithm to this inventory, we again need two features. /1/ and /0/ are [+high], to the exclusion of /9/. /0/ is [+round], to the exclusion of /9/ and /1/. It follows that we need [±high] and [±round] minimally (and maximally). Again, there are two possible sequences of applying the binary divisions, listed in (23):

(23) Possible specifications for /i 1 9 0/

a. [high] >> [round]

1 9 0

[high] + - +

[round] - +

b. [round] >> [high]

1 9 0

[round] - - + [high] +

-As becomes clear immediately, there is no difference in the feature specification of /i a u / and /1 9 0/, and, as enhancement can only apply to the specified features, it is unlikely for any of the two possible schemata in (23a) or (23b) to surface as /1 9 0/: /a/ is lower then /9/; for example, and the rounded-ness of /u/ is more enhanced (by backrounded-ness) then that of /0/. Furthermore, the possible feature specifications for /i a u/ are a subset of the possible feature specifications of /1 9 0/ (for example, a system with [±low] and [±back] could describe the former, but not the latter). Thus, /1 9 0/ cannot be distinguished

from /i a u/, but the reverse does hold. Hall (2011) shows that the contrast and enhancement theory excludes other unattested vowel inventories (such as horizontal or diagonal ones) through a similar logic, and that the same holds for consonant inventories. The key to the success of the contrast and enhance-ment approach appears to be that, rather than mix and confound phonetic and phonological effects, it separates phonetics from phonology in a principled way, and therewith separates phonetic and phonological tendencies. It has long been observed that inventories tend to be dispersed through phonetic space (shape), but also that they tend to be symmetrical in their phonological spec-ification (structure). By allowing dispersion effects to act at the level of the feature, and allowing only contrastive features to be specified, the contrast and enhancement approach appears to have found a equilibrium between phonetic and phonological forces, which makes correct predictions, too.

There, are, however, some issues with the contrast and enhancement theory.

First of all, there is the matter of underspecification. Phonetic enhancement applies at the feature level, and only contrastive features are specified. In a standard feed-forward model of the phonology-phonetics interface, however, the phonetic implementation module has no access to underlying structures. This means that underspecification must be permanent, in the sense that it holds at the phonological output level. Redundant features are thus not filled in by the phonology, and cannot be supplied by the phonetics unless they enhance one or more contrastively specified features. Whether this is a real problem must be tested empirically.

A second problem is that enhancement does not really hold at the feature level only: only those correlates of features are enhanced that result in more contrast. Hence, not only the phonology must have access to the entire inventory in online processing, the same holds for the phonetic implementation module.

It appears, thus, that the contrast and enhancement theory is, in some sense, holistic.

Finally, the Successive Division Algorithm is successful in deciding the un-derlying representation of entire inventories; it remains to be seen, however, whether it is a suitable model for a learner whose inventory grows over time.

In other words, whereas it is an excellent solution to the logical problem of language acquisition, it remains to be seen how well it fares in the light of the developmental problem of language acquisition. This is an issue we will address in chapter 5.