Prosodic licensing and the development of phonological and morphological representations
Dinnsen, Amsterdam: John Benjamins. This material is under copyright, and John Benjamins benjamins. Superaddivity in Bambara words 1 Superadditivity and limitations on syllable complexity in Bambara words Christopher R. The premise of superadditivity is that although marked structures are accommodated in a system, particular structures cannot co-occur in a given domain.
This arises because the simultaneous, additive violation of constraints within a domain arguably incurs an additional penalty. Thus, languages may limit the number of phonologically complex structures in a domain. We consider superadditivity in CB, which places strict limitations on the type and distribution of complex syllables within a word. We also discuss these data as they relate to models of phonological acquisition which maintain that outright bans on multiple complex structures do not occur in adult language yet are frequently encountered in developing languages.
Introduction Languages may act to limit the type and number of phonological complexities permitted in a word. While a language may not disallow all instances of phonological complexity, complex structures may occur less frequently than otherwise expected. Such tendencies and restrictions have been attributed to detrimental superadditivity effects e. Albright , that would arise when certain combinations of markedness constraints are violated within a single domain e.
The general premise of superadditivity is that although marked structures may be individually accommodated in a system, there exist some marked structures that cannot co-occur in a given domain. Scholars have argued that superadditivity arises because the simultaneous, additive violation of constraints on certain marked structures within a single system or domain incurs an additional penalty.
- The Mask of Anarchy: The Destruction of Liberia and the Religious Roots of an African Civil War.
- A Practical Guide to Security Assessments!
- The role of onsets in primary and secondary stress patterns.
- Recent work | Phonological Typologies Laboratory?
- The Brazen Gambit.
- Jae Yung Song.
This proposition diverges from standard conceptions of Harmonic Grammar HG: e. Strict domination is such that only a single constraint is ultimately responsible for evaluating a candidate, effectively masking the effect of constraints ranked lower in the hierarchy. Because of strict domination, OT does not easily address gang effects. The formal mechanism of LCC still in the spirit of strict domination requires that the locally- 1 There is growing interest in methods of additive and superadditive evaluation in formal phonology.
Some scholars have argued that instantiations of LCC are not grammatically intuitive and may in some instances predict unattested phonological grammars e. While our analysis of superadditivity effects employs a type of local conjunction of constraints, we aim to show that these constraints do not wield quite the same power that analogous constraints have in typical LCC analyses. This paper has several aims. First, we shed light on the types of superadditive effects found in natural language by considering three interrelated phenomena in Colloquial Bambara CB.
That is, CB avoids words that contain multiple instances of some, but not all, complex syllable types. To demonstrate this, we employ an HG framework supplemented by superadditively conjoined constraints to demonstrate that the additive effect of violating certain combinations of constraints is more costly, phonologically-speaking, than the sum of their individual parts.
Lastly, we discuss how these CB data relate to models of phonological acquisition e. This language, described in Green , has emerged or diverged from the standard, more phonologically-conservative urban variety of the language. The data in this paper are drawn from Green and our subsequent research on Bambara. Bambara has a typical seven-vowel system in which high vowels are preferentially, but not exclusively, deleted.
VCD operates alongside VS, producing CVV syllables with derived long vowels by removing intervocalic velar stops flanked by identical vowels. Derived CVV syllables differ from syllables with phonemic long vowels which are limited in their distribution to morpheme- initial positions and may be absent in the synchronic phonologies of some speakers Dumestre Whether by the action of VS or VCD, there is an overall drive towards minimization in 3 VS can remove vowels of any height from a particular syllable or word position, but phonotactics and metrical restrictions govern the process Green There is independent evidence that high vowels may delete in certain Standard Bambara words to create phonetic CCV syllables e.
Dumestre A non- syncopated form of a given word will surface in certain constructions when syncope cannot occur for reasons of metrical structure Green Tonal melodies on syncopated forms are also clearly derived from those in SB. We illustrate below that in a restricted set of instances, a second deletion is permitted.
Typical instances of single deletion via VS are in 1. Examples 1j-k show variation between CCV and CVC outcomes where identical vowels can be removed to create syllables with licit margin phonotactics. Examples 1l-n , however, show that VS will not occur when appropriate phonotactic conditions are not met. These reductions occur analogously in shorter words, as in 2. Words in 3 show VCD removing an intervocalic velar stop flanked by identical vowels. In these words, phonotactic restrictions preclude VS.
Thus, in 3 , VCD acts on an intervocalic velar stop within this domain when it is flanked by identical vowels, e. In some instances, this competition ends in a stalemate in which either process but not both may apply. This is shown in 4. In 5c , when deletion targets are separated by a foot boundary, only a single outcome is possible. They also offer insight into the distribution of complex syllables in CB. By the action of VS, CB admits 5 Others offer independent perspectives on certain characteristics of foot structure in Mande languages Vydrine and in Bambara tonology e.
These and the current approach differ most apparently in their assumptions of foot headedness.
CV, and CCV. CV words.
Distribution of complex syllables in Colloquial Bambara VS and VCD are constrained by Bambara foot structure and reference the prosodic foot domain when selecting and acting upon their respective targets. The examples above consider the outcomes of VS and VCD when acting alone or in competition with one another for a deletion target within a single disyllabic foot. Compounds and certain morphological derivatives offer a look beyond a single domain where VS and VCD have deletion targets located in adjacent feet.
By exploring such CB words, we can observe the potential for and permissibility of speakers to introduce multiple deletions and therefore multiple complex syllables into Bambara. Data below illustrate that the introduction of multiple complex syllables to a single word is highly regulated due to interactions between certain phonological constraints. Data in 6 illustrate this choice and show that both CVV.
CV and CV. CVV words are possible in CB. In the former, VCD acts within the leftmost of two disyllabic feet, while in the latter, VCD acts on a target in the rightmost foot. CVV words are conspicuously absent from these data. CCV are not found. This may be surprising considering that in some instances e. We show this in 8. CVV words are disallowed in CB. The preferred reduction yields a complex syllable in the leftmost foot, closer to the left edge of the word.
One possibility to account for this is that higher level metrical restrictions preclude the co-occurrence of adjacent heavy syllables. We turn next to words that have adjacent feet with potential VS deletion targets. What we encounter is more complex, yet similar, to that observed in 8 for VCD. For words with multiple VS targets, the patterns of complex syllable formation rely on the types of vowels available and selected for reduction. The simplest cases concern words with adjacent domains containing non- high vowel VS targets. In these words, like the representative case of 9 , only a single deletion generating a single complex syllable is permitted.
The complex syllable generated is again in the leftmost foot, creating complexity at the left edge of the word. Like the individual constituents in 8a-b, d-e , the constituent words in 9a-b are free to reduce by VS in isolation.
However, when the constituents are compounded in 9c , only a single reduction is possible. Here, CB permits multiple deletions within a single word. The outcome is CCV. Note that this outcome is possible in both monomorphs 10a and compounds 10d. CCV word; however the process is tightly constrained such that a second deletion is allowed only when at least one of the two VS targets is a high vowel. This result raises several important questions. Secondly, how does Bambara regulate the highly selective process of minimization such that only words containing certain combinations of vowels are potentially subject to deletion?
Lastly, what formal mechanism is best employed to capture these complex relationships? To address these questions, we entertain potential analyses of these minimizations in OT vs. These constraints capture the parallel preferences for consonants of specific sonorities to be found in certain syllable margin positions. M1 and M2 constraints may also be conjoined in a local domain i.
An adaptation of the split margin syllable is in Figure 1. These limits are problematic for strict domination, which effectively overpredicts segmental deletion and the creation of complex syllables in CB. A Standard OT, strict domination analysis predicts that optimal outcomes in CB have fewer syllable peaks in all instances. Note that HG constraint weights are arbitrary; however the ratio between them is the key factor in the analysis Farris-Trimble This better captures what occurs in CB, where the cumulative violation of margin markedness constraints is a less favorable or less harmonic outcome than what is attested, i.
The candidates in 11 , now evaluated in HG, are shown in We begin with words for which multiple reductions arise via the removal of high vowels, as in By employing HG, we view the effect of constraints banning different syllable peaks high vs. These relationships are key in correctly predicting outcomes like 13d , where by removing two high vowels, CB creates a word with two CCV syllables. This involves removing one high and one non-high vowel to create a CCV.
CCV word, where the benefit of removing a non-high peak outweighs retaining it, if only slightly. While it is beyond the scope of this paper to explore this, the restriction is likely due to a requirement for dissimilar M2 consonants within a word. CCV words. Compared to Standard OT, HG also arrives at the attested outcome in 12 where only a single reduction is possible in words containing two non-high vowel deletion targets.
For example, there are words like 7f , in which, all else being equal, one might predict a double deletion, yet only a single deletion is permitted.
Serie: Language Acquisition and Language Disorders » Bokklubben
We propose that what is at issue in this and related instances is the particular combination of constraint violations that would be necessary to accommodate this double reduction in this case, violations of MAX[-hi] and MAX-K. We argue below that this combination of constraints and two others in CB cannot co-occur within particular domains, and as a result, their violations combine superadditively to create a more harmonically complete phonological grammar.
Superadditive ordering Smolensky and LSS entertain the hypothesis that some grammars are non- harmonically complete such that neither strict domination nor a weighted harmonic analysis accurately predict all attested outcomes. As such, neither ranking nor simple summation of constraint violations arrives at correct, grammatical predictions. These authors suggest that non- harmonically complete phonologies necessitate the introduction of local constraint conjunction via superadditive ordering relationships between some constraints.
Although we return to discussion of such constraint conjunctions below, it is important to note that the constraints proposed here introduce subtle, rather than sweeping, effects of cumulativity in CB. To begin, consider first the predicted yet unattested HG outcome in 15 utilizing the constraints employed above. It appears, therefore, that HG encounters the same issue of overpredictability observed in standard OT.
The outcome in 15 , and in particular the difference between the total harmony scores for the predicted vs. This same difference but importantly in the opposite direction separates the incorrectly predicted winner 16b from the attested winner 16a. Thus, it is impossible via simple summation of violation scores to arrive at both attested outcomes in 16 , even though the weight ratios between these constraints have successfully predicted attested outcomes straightforwardly for other word types, e.
CB, therefore, appears to be non-harmonically complete. For non- harmonically complete systems, Smolensky proposes that local conjunction of select constraints in HG introduces harmonicity into the grammar. We follow Smolensky in exploring this possibility in CB. We know that CB prefers when phonotactically possible to delete high vowels; however it will delete a non-high vowel only when the alternative is not to reduce at all. This disharmonic relationship suggests a superadditive constraint relationship. That is, while violating one or the other of these constraints in a word is normally unproblematic, violating them both presents a situation where their combined violation is equal to their sum plus an additional penalty.
The addition of this constraint introduces harmonicity into the grammar, such that the grammar now accurately predicts certain CB outcomes wherein some doubly unfaithful forms are permitted 17d , while others are disallowed 17a. CCV word formed by these constraints may be possible. We expect the reverse to be precluded for reasons of illicit iambic structure, i. The authors suggest that combinations of faithfulness constraints, on the other hand, could be ordered in a sublinear relationship such that faithfulness violations yield dissimilarity, and thus the penalty for being unfaithful to an already dissimilar structure might be less costly.
They point out, however, that this proposition may be problematic. Indeed, the instances discussed by LSS concern featural rather than structural faithfulness. For our purposes in CB, proposing a sublinear relationship between faithfulness constraints fails to capture the data. In 17 and below, a superlinear relationship between relevant faithfulness constraints better captures the interaction between illicit constraint combinations.
When MAX is violated to remove a peak, the other remaining structures that MAX still protects against removing are no less representative of faithful structure than they were previously. We propose, therefore, that faithfulness constraints, too, can be superlinearly conjoined. Recall from 12 that outcomes reduced by two non-high vowel deletions are also unacceptable in CB. That this constraint is, in fact, too powerful becomes clear from a very limited number of words that accommodate multiple deletions to create CVC. Reduction by non-high vowel loss is the least preferred means of minimization in CB, and MAX[-hi] is the highest weight faithfulness constraint of those considered in the hierarchy.
CB disallows multiple deletions of non-high vowels within a word, suggesting another case of superadditivity. That is, while removing one non-high vowel is unfavorable, removing more than one is far worse; so much so that it is prevented. This is shown in This non-high vowel, however, is never removed by VS. The result is CB words that are fully faithful to their SB input. In these words, removing the first syllable high vowel would create impermissible obstruent-obstruent complex onsets, while removing the word-final vowel would create a disallowed coda. Creating a word-internal obstruent-sonorant complex onset is acceptable in other instances, but here, the option is not possible.
Tableau 20 shows that the grammar overpredicts non-high vowel deletion. The recurrence of this value and the ratio of constraint violations that it represents suggests that here, too, disharmony is at play. The disharmony once again appears related to the dispreference in CB to violate MAX[-hi] except as a last resort.
In 20 , the incorrectly predicted candidate avoids creating an impermissible complex onset by instead removing a non-high vowel. Resisting the drive toward minimization by total preservation of peak faithfulness is preferred in this situation. Although this seems problematic formally, it is intuitive in terms of cross-linguistic generalizations on markedness. The outcome in 20 supports this assertion, i. CB behaves as predicted by barring against this marked state of affairs. An important detail, however, concerns the domain of conjunction for these particular combined constraints.
Next, unlike constraints in OT LCC, the power of the proposed superadditively conjoined constraints is comparatively minimal; their weight is comparable to syllable margin constraints. Finally, the larger question as to the types of constraints that should be permitted to conjoin is still an open area of debate e.
The next section discusses some further theoretical and analytical concerns related to our findings. Summary and conclusion CB exhibits a drive towards minimization by removing segments, and therefore syllables, from its SB input. Generally, this involves violating faithfulness constraints i. MAX to avoid violating more costly markedness constraints i. CB data show that high vowels are marked and easily removed, and while non-high vowels are not exempt from removal, this is a less preferred option.
Complex syllables are permitted in CB within the bounds of phonotactics and metrical structure, yet the number and co-occurrence of these syllables in a given word is restricted. We have shown that some words accommodate more than one complex syllable; however their co-occurrence is lexically gradient. Escudero, M. Warren eds. Combiths, P.
Serie: Language Acquisition and Language Disorders
Quantifying phonological knowledge in children with phonological disorder. Treatment targets for co-occurring speech-language impairment: A case study. Barlow, J. The syllable. Damico eds. Keffala, B. International Journal of Bilingualism, 22 1 , Influences of phonological context on tense marking in Spanish-English dual language learners.
Journal of Speech, Language, and Hearing Research, 60, — Sonority in acquisition: A review. Ball ed. London: Equinox. Dinnsen, D. Phonological disorders: Theoretical and experimental findings. Lidz, W. Pater eds. Age of acquisition and allophony in Spanish-English bilinguals. Frontiers in Psychology, 5, PMID: Barlow eds.
Perspectives on phonological theory and acquisition: In honor of Daniel A. Dinnsen pp. Philadelphia: John Benjamins. Fabiano-Smith, L. Dialect density in bilingual Puerto Rican Spanish-English speaking children. Linguistic Approaches to Bilingualism, 4,