Grandfather Effects: A Longitudinal Case Study of the Phonological Acquisition of Intervocalic Consonants in English

Abstract

In this article, I present a longitudinal study of a child's (male, aged 3;0–3;4) acquisition of intervocalic consonants characterized within the framework of Optimality Theory (OT). At Stage I, the child presents with unusual error patterns, weakening and labial substitution, and shows evidence of phonologically opaque surface forms. These patterns persist at Stage II despite elaboration of phonemic contrasts intervocalically. In addition, a new error pattern emerges, nasal harmony, along with additional opaque forms. Although the errors exhibited by the child are somewhat atypical of those generally observed in acquisition, they are, nevertheless, entirely consistent with cross-linguistic facts. Further, although phonological opacity is a linguistic phenomenon that poses a challenge to nonderivational theoretical frameworks such as OT, I offer McCarthy's (2002; 2003a) comparative markedness approach as a solution to this problem that is theoretically superior to alternatives.

ACKNOWLEDGMENTS

This research was supported in part by a grant from the National Institutes of Health to San Diego State University (DC05754). Special thanks to Timothy Tipton and Ashley Jung for assistance with data collection, transcription, and analyses. Thanks also to Dan Dinnsen, Heather Goad, and three anonymous reviewers for very helpful comments on earlier drafts of this article.

Notes

The verb grandfathered in is drawn from American law and refers to grandfather clauses that allowed certain rules (or laws) to continue to apply in certain existing situations, whereas new rules were to apply to all future situations. To be grandfathered in refers to this type of exemption from application of the new rules (see also McCarthy (2002), Note 2).

The abbreviation “dim.” refers to the diminutive form of a word as in doggie.

As a reviewer noted, target skating is included with data set (4a), whereas targets ladder and riding are included with data set (4b) despite the fact that all three are produced with a flap in intervocalic context in the adult system. The data are organized as such based on assumptions about how these forms are represented underlyingly in the adult form. The word skating is assumed to have an underlying /t/, and ladder and riding are represented with /d/. The same appears to hold true for Z's grammar given that he systematically produces these targets forms differently ([f] for the former, [w] for the latter) despite their neutralized status in adult surface forms.

The choice of word-initial /v/ as the treatment target was made independently of the child's errors on intervocalic consonants in particular and instead was determined based on other aspects of the child's sound system as well as criteria of the larger study from which he was drawn.

Harmony at this stage affected progressive forms primarily. Z did show some evidence of both right-to-left and left-to-right nasal harmony at Stage I, although this was nonsystematic and unpredictable, and did not apply in progressive forms: [wæmin] ‘wagon’, [m∧mi] mud (dim.)', [mæmi] ‘magic’, but [diwin] ‘digging’, [feiwin] ‘shaving’. These forms suggest that the harmony pattern had already begun to emerge in his sound system prior to the initiation of treatment, although by Stage II, the directionality of harmony was right-to-left only. Although I do not present any formal account of this pretreatment pattern, I take up the issue further in section 4 in light of variability in children's sound systems (Boersma (1998)).

At both stages of development, Z's productions showed evidence that voiced obstruents and sonorants patterned as a natural class in his grammar, suggesting that Z's grammar may be best characterized as a Sonorant Voice language (as characterized by Avery (1996), Goad and Brannen (2000), Piggott (1992), Rice (1993)). Such languages are reported to show allophonic variation between voiced obstruents, nasals, and/or approximants, revealing of their function as a natural class. The patterning of voiced and voiceless consonants was not the primary focus of this study, and the account that I present is shown to be adequate for differentiating the two sound classes and in accounting for the observed patterns in Z's productions (see also Flemming (2004)).

One might be tempted to assume that Z's errors are attributable to motoric immaturity or even oral-motor deficiency. Indeed, weakening patterns are noted to occur in the speech of (for example, individuals with cleft palate) (Harding and Grunwell (1998)), due to insufficient oral closure. Recall, however, that an oral-motor evaluation showed that Z was functioning normally and showing no structural abnormalities. Further, Z consistently produced stop closure for target labial stops intervocalically (as well as both labial and coronal stops word initially) prior to treatment.

In the clinical literature (e.g., Grunwell (1985), Ingram (1981)), this labial substitution pattern has often been referred to as labialization, a term I avoid here given its more widespread usage in reference to the pattern of secondary articulation as in [k ^w ].

Rose and Walker (2004) distinguished “nasal harmony” from “nasal agreement.” In the former, intervening vowels and other consonants are nasalized, whereas in the latter, they are not. This latter definition is most similar to the types of long-distance harmony observed in child language. However, in this case, intervening vowels are likely nasalized due to the more general process of vowel nasalization observed in English.

I address the relative (un)markedness of labials in the sound systems of Z, T, as well as languages such as Sri Lankan Portuguese Creole in section 4 (see also Hume (2003), Hume and Tserdanelis (2002)).

Although the exact characterization of this constraint is not clear, it is certain that the constraint or constraints that achieve this effect must function like phonotactic constraints, prohibiting all nonlabials from occurring intervocalically.

All place, manner, and voicing features are assumed to be privative, following Rose and Walker (2004) and references therein.

This constraint, as it stands, would rule out initial bilabial stops. As I show very shortly, this constraint is problematic for other reasons as well, and will be modified.

As *pbm is a co-occurrence constraint, even the mapping of /VpV/ → [VmV] would violate _O*pbm because the labial stop from the input also occurs in the output. Faithfulness constraints (i.e., Ident(nasal) and Ident(voice)) prevent these forms from surfacing.

The relative ranking of the constraints _N*pbm and _O*pbm is not a problem for word-initial stops given that only fully faithful (old) bilabial stops are allowed to occur in that context.

On the other hand, because underlying approximants always surface as approximants (as [w]), Ident-io(approximant), which prevents deletion of [approximant] from inputs, must be ranked relatively high in Z's grammar at both points in time.

I will revise this ranking in section 4 in light of additional constraints that are introduced for Stage II relevant to the nasal harmony pattern.

I reconsider the ranking of these two constraints at Stage I in section 4.

At Stage I, of course, *{\nhookfnt} (both new and old versions) was ranked high (undominated) given that all dorsals were prohibited. Coronal nasals were allowed to surface faithfully (except intervocalically), yet they never occurred as substitutes for other sounds. This warrants a high (undominated) ranking for _N*n but a relatively low ranking (below Ident-io(nasal)) for _O*n at that stage.

Frameworks such as output–output correspondence (Benua (1995)) and sympathy theory (McCarthy (1999)) do not account for opacity effects of this nature, nor were they intended to do so.