From: LAURA SLOCUM This response paper to Thomas Berg's "Sound change in child language: A study of inter-word variation" (Language and Speech, 1995) served to challenge Berg's original assumption that covert contrast was not apparent in young children's production of alveolar and velar sounds (t, d, k, g). The authors provide acoustic and articulatory evidence of covert contrast and suggest that additional research is needed in order to understand the perceptual features which impact listener judgments. Comments/Questions: Each of this week's readings deals (either explicitly or implicitly) with covert contrast and whether children at young ages are able to produce subtle phonological distinctions imperceptible to most listening ears. For example, when a child produces "th" as a substitution error for /s/, are the acoustic and articulatory parameters of that "th" distinguishable from a target "th"? If so, then can we assume that the child's phonological system includes representations of both /s/ and "th" (even though "th" may consistently be produced for /s/)? It seems rather obvious to me that this would be the case as children at young ages are able to recognize production differences in others despite being able to produce such distinctions themselves (as in the case of the child who insists he said "fis," not "fis"), and it comes as no surprise that the acoustic and articulatory parameters of a target "th" would be different from the child's "misarticulation" of /s/. I do question, however, the accuracy of Baum & McNutt's (1990) statement that: "Results of the current study indicate that frontal misarticulation of /s/ cannot be characterized as a "th" for /s/ substitution" (p. 62). Coming to this from a speech-language pathology perspective with an emphasis on overall intelligibility of speech, the listener's perception is deserving of consideration. The child may very well be marking differences in such a way that indicate awareness within his or her phonological system; however, (audible) phonetic constraints persist. Once a child's ability for "discrimination" (so-to-speak) between target sounds has been established, how would this information affect therapy? There are children for whom "correct" articulatory gestures are not able to be produced (e.g., due to underlying organic causes) despite their intact phonological systems. As speech-language pathologists, our goal is ultimately communicative efficacy and I question the authors' assumption that subtle measurable differences in acoustic and articulatory phenomena are more important (or valid) than perceived sound qualities. Additionally, since Edwards, Gibbon & Fourakis (1997) show that phonologically disordered children are generally more variable in production of all sounds (t, k, and t for k substitutions) than their typically-developing peers, how can we accurately determine acoustic and articulatory parameters for "target" and "misarticulated" sounds in this population? Furthermore, at least one of these studies (and I think two of them) state that the children being studied had participated in a course of therapy prior to data collection. This is not trivial, as the children have presumably been working on "proper" articulatory gestures and their approximations. Their productions, therefore, do not represent a "natural" state of disordered speech. From: Kirk Baker This paper looks at temporal, intensity and spectral measures of fronted /s/ misproductions in children. Address two main points: 1) Is fronted /s/ realized as /T/; no. 2) Are there age related differences in misproductions; no. p.54 Why did they use an 8.5kHz low-pass filter but report Daniloff Wilcox Stephens 1980's /s/-fronters with energy peaks in 13-14kHz range (p.52)? Why didn't they use a sample rate high enough to look at this range of spectrum? p.58 Why are their fricative - vowel dB differences so much greater than DWS80? p.59 Is high centroid correlated with small front cavity? So centroid for /s/ should be higher than for /S/? If I don't have it backwards, I'm confused about why /s/ centroid 6369Hz is higher than /T/ 5730Hz. Is it because there's no cavity anterior to tongue-teeth constriction for /T/, and centroid reflects that anterior resonance? From: ASIMINA SYRIKA -I find it hard to understand the authors' argument about the abstract nature of phonemes. Why do they see phonemes as abstract units? What are the implications of this argument for the phonemes' language specificity and acquisition by children? From: Junko Davis I agree with the view that "acoustic and/or articulatory analysis are necessary to supplement auditory-perceptual judgments... (p.209)." However, I have been wondering for a long time how we should balance these physical and psychological perspectives. For instance, when I examine the accuracy of voicing, I look at the spectrogram if I am not sure. The lack of a voice bar makes me feel secure about making the judgment of the target as "voiceless," but does this mean that I am incorporating a possible covert contrast into my judgment? From: ejkong@ling.ohio-state.edu It comments on Berg's case study on one kid's continuous development of velar consonants, suggesting that possibly there might have existed the covert contrast in Berg's data which reveals the gradual acqusition process of abstracting feature/segments categories. Questions: - Can we go over what 'skewness' and 'kurtosis' indicate about the place of articuation? (I got the idea of 'skewness' thanks to Fangfang but not 'kurtosis'.) And what does 'moment' mean in the spectra? ('center of gravity' would be the same thing to 'centroid'?) - From the phonological acquisition point of view, the covert contrast provides the strong evidence that the contrasts are gradually shaped: i.e., there are stage where the contrasts are in kid's phonological representation but not phonetically realized well enough to be recognized by others. Then, what would this imply from the speech pathology point of view when it's not correctly perceived by people anyway? - Are we assuming that the covert contrast between segments is one of the stage that kids undergo in the developmenal process? Have the previous researches on the covert contrast investigated only on the kids with phonological disorder but not on normal kids? (such as a longitudinal study finding out the covert contrast from the normal baby who is just before getting the contrast in the developmental stages) From: Kathleen Currie Hall Notes on: Edwards, Jan, Fiona Gibbon, and Marios Fourakis. 1997. On discrete changes in the acquisition of the alveolar/velar stop consonant contrast. Language and Speech, 40(2): 203-210. Summary: In this article, the authors respond to a previous article by Thomas Berg, which traced the acquisition of velar stops in word-initial position of a single child. While they praise his work and his large corpus, and especially his conclusion that acquisition was gradual and not abrupt, they disagree with his conclusion about _why_ this acquisition is gradual, claiming that it is a matter of gradual motor control rather than gradual psychological realization. Details: 1. Berg claims that features and phoneme segments are "connected" and that the connection between these two is built up gradually over time -- hence gradual acquisition of velar stops, for example. 2. Edwards, Gibbon, and Fourakis claim that "features" and "phonemes" are only abstract representations and that the "acquisition" of an abstract concept like "velar" involves the acquisition of _many_ motor skills (since "velar" may differ depending on e.g. following context). 3. Therefore, they claim that the source of gradual learning is simply the gradual nature of learning motor control. 4. They argue that motor control can be shown to be gradual through the existence of covert contrasts: i.e., that there is clearly not a categorical jump between a single, undifferentiated motor gesture and a pair of fully adult-like motor gestures. 5. The rest of the paper largely covers the basics of covert contrast and shows examples of it. 6. A final point is that linguists need to be careful in terms of using transcriptions when the whole point of covert contrast is that it is covert . . . . Questions: 1. I'm not sure I know why Berg's explanantion is completely incompatible with that of Edwards, Gibbon, and Fourakis (perhaps because I haven't read his paper?). Is it the case, in cases of covert contrast, that there is generally a tendency to differentiate, or that there is usually a reliable way to differentiate a child's productions, but one that is not detectable to adult ears (e.g. in the EPG data, it looks like I could definitively categorize all the productions of [d] and [g])? If the latter, then there is a sense in which each stage along the gradient path from one undifferentiated gesture to two adult gestures is categorical. It doesn't seem impossible to me that a child could have good control over these categorical gestures, perhaps being able to differentiate his own productions as clearly as if they were adult-like, but that in another context, he doesn't know that he should be producing the _same_ contrast (even though he could in theory be producing a different covert contrast). Hence it could be only gradual that the child realizes that the sounds in these two contexts are in fact both "velar" (precisely because it is an abstraction) and begins producing them the same way. (Note that even if the child doesn't have good motor control over the covert contrast, this idea could still work -- i.e., it could be BOTH a case of immature motor control AND a lack of strong psychological connection between an abstract feature and a set of sounds.) 2. Could we go over the definitions of the various "spectral moments" they use here? From: HELENA RIHA p. 206 Figure 1 EPG data for two subjects This figure is particularly interesting because it shows the children's tongue contact patterns for /d/ and /g/. In particular, they have simultaneous alveolar and velar closures when producing /da:/. It seems that when they eventually learn to make the /d/-/g/ contrast, they will subtract a velar closure from their articulation of /d/ to produce a 'correct' /d/. What does this subtraction of gestures say for theories of phonological development? How would Rice & Avery account for this kind of progression in the acquisition of phonological contrasts? p. 203 "Melanie's acquisition of word-initial velars was gradual rather than abrupt ..." Did her acquisition of word-initial velars depend on acquisition of labials and dentals in different position? Did acquisition of labials have to precede that of velars? Also, gradualness of acquisition seems logical if features and phonemes are an abstraction. Children need language experience and a certain level cognitive maturity to move away from specific phonetic contexts to an abstract representation of features and phonemes. p. 205 "... children do not learn to produce phonological contrasts in a strictly categorical or discrete way." Does this counter Rice & Avery's (1995) learning path in which inventories are build up in a monotonic fashion (p. 35)? p. 207 "... the distinctions are not categorical." How does this square with Rice & Avery's notion of the development of opposing consonant contrasts in place of articulation? (p. 39) From: Fangfang Li this paper mainly questions Berg's claim of absolute alveolar/velar contrast, by invoking a whole body of literature on covert contrast, in which an instrumental-tested but perceptually un-noticeable difference in many studies on substitutions produced by children with phonological disorder as well as normal-developing children. questions: 1. what exactly does "moment" mean in "moment analysis"? not sure what it means by first moment, second moment, etc. 2. what is kurtosis? how can it used to describe the general shape of the spectrum? 3. is covert contrast a "necessary" stage in language acquisition? can a discrete contrast happen during a process at all? From: Kirk Baker I agree with the spirit of Edwards et al's conclusion that acoustic analysis might can reduce speculation about the articulatory configuration that produced a speech event, but I've been impressed with how differently we're interpreting ins trumental output in our transcriptions. At least for my transcriptions, the 'phonetic' symbol that I stick on an acoustic event is really a perceptual label and not an articulatory one. And even their Fig 1 palatogram data (p.206) isn't that easy to interpret without knowing the time-course of contact/release.