Cross-linguistic perspective on “atypical” error patterns

As a clinical phonology instructor working in a linguistically and culturally diverse environment I am conscious of the need to prepare my students for clinical practice with children who may speak languages that are not English. I am forced to narrow my focus in the classroom to English normative data because there is only so much time and because most of the available clinical data pertains to English. At the same time I know that those students who plan to practice in English-dominant environments will have some bilingual children on their caseloads. That part of my class that plans to stay in Montreal will be working in a bilingual environment, switching hourly between English dominant and French dominant patients who often know a second or third language. Finally a good proportion of the class are international students who plan to return to their home countries to practice in a language that is neither English nor French. Therefore the question of how best to ensure that they have analytical tools and background knowledge that will allow them to apply what they have learned in an English context to other languages is paramount. Sharynne McLeod’s work and the Multilingual Children’s Speech website  is enormously important in this regard.

I really enjoyed the systematic review by Hambly, Wren, McLeod and Roulstone on “the influence of bilingualism on speech production”. I especially liked the introductory summary of Stackhouse and Wells’ psycholinguistic model whereby phonological development is “simplified into three phases: input → storage → output”. This model figures prominently in our book and was one influence on the organization of our discussion of treatment procedures in Part III. In their review, Hambly et al. describe many papers that recount instances of apparently atypical error patterns in bilingual children with typical or atypical speech development. Many of these patterns reflect transfer from one language to another such as unexpected voicing patterns, trills or spirants crossing over from first language Punjabi, Russian or Spanish to second-language English. The papers reviewed also described patterns that were rare across languages however: for example, the rare occurrence of backing in a variety of first and second language contexts lead some authors to the conclusion that backing is universally rare and therefore always atypical. Francoise and I are aware of a context in which an error pattern that might be mistaken for backing is not rare however. Our papers describing typical and atypical French phonology take a multilinear and multirepresentational approach and highlight the disconnect between the persistence of a phonological pattern approach to the description of children’s speech despite the important insights offered by the psycholinguistic approach to phonological development.

Although the theoretical underpinnings of phonological processes are not necessarily implied when clinicians and researchers speak of phonological patterns I think that it is helpful to recall the theoretical roots of this approach. The original idea was that the child’s underlying representations were adultlike even though the child’s motor abilities were not up to the task of producing speech that matched adult expectations. Therefore one assumed that the child’s underlying representation for a word such as “cone” would be /kon/ but an innate simplification process would front the /k/, leading to [ton]* in the output. The error pattern in “cone” → [ton]*, “key” → [ti]*, “cake” → [tet]* is called velar fronting, implying that an underlying present velar is fronted during the transition to the output form. The opposite pattern, in which a front sound, e.g., /t, d, s, z/ is produced as a back sound, e.g., /k,ɡ,ʃ,ʒ /, is referred to as backing but is atypical by virtue of infrequence. I am not going to recount all of the evidence against this position here but suffice it to say if it doesn’t work for phonological development in the case of a child learning one language according to a typical trajectory it is even less tenable as an explanation for mismatches in bilingual children’s speech and it is clearly not tenable as an explanation for mismatches in the speech of children with a phonological disorder.

Continuing with the example of backing I am going to show that this is a poor description of apparent backing errors in English or French speaking children with a phonological disorder. In English speaking children, backing alveolar consonants is indeed rare and thus termed “atypical”. The atypical nature of the error is further enhanced by the phonological problem that it represents: when analyzed from a multilinear perspective, fronting means that the child is using the default Coronal feature or is delinking the marked Dorsal feature, resulting in a simplification; on the other hand, backing is in most contexts phonologically impossible since it involves adding a marked feature unless the child has a very unusual phonological system in which Dorsal is the default place feature. How then does the error arise? Answering this question requires that one examine the articulatory roots of the child’s problem as was done by Fiona Gibbon in her description of “Undifferentiated Lingual Gestures in Children with Articulatory/Phonological Disorders.” Electropalatography shows that this error pattern is rooted in an abnormal lingual gesture during speech that is pervasive, affecting all the lingual phonemes and often resulting in other atypical errors such as lateral distortions of sibilants. One thing that is important and interesting about the child’s error pattern is that it is revealed to be not backing of front sounds at all since all the lingual phones are produced with a similarly undifferentiated gesture (neither front nor back but whole tongue dorsum in contact with the palate); the listener’s percept of a front or back sound is determined by the timing of the release phase of the gesture. In this case, in English, the error is a highly atypical error reflecting a motor speech problem that is not accurately described as backing.

Now, turning to French, we have published raw data from francophone children with a phonological disorder in the Canadian Journal of Speech-Language Pathology and Audiology: we have observed that in both typical and atypical development it is very common for French speaking children to substitute a back sound for a front sound in a very particular context, specifically the /ʁ/-clusters. We show first of all that the output form [kʁ] is easiest for the children to acquire and thus the place features of the two consonants appear to facilitate production of each other, e.g., as in ‘crayon’ /kʁɛjɔ̃/ → [kʁɛjɔ̃]. Second, in the case of /tʁ/-clusters, spreading of place from the /ʁ/ to the unspecified place node on the /t/ is very common resulting in an apparent backing error, e.g., as in ‘train’ /tʁɛ̃/ → [kʁɛ̃]*. Interestingly, we see the spreading of the Dorsal feature from the second segment in the cluster back to the first segment of the cluster even when the /ʁ/ target is produced as [w] and in the case of /w/ clusters (recall that this phone has both labial and dorsal place features), e.g., “doigt” /dwa/ → [ɡwa]*. Finally, we observe even more unusual productions such as, ‘framboise’ /fʁɑ̃bwaz/ → [kwɑ̃bjaz]*, suggesting that the features of /f/ are difficult for the child to capture perceptually in these words, perhaps resulting in substitution of a preferred form in the complex onset. We have seen similar error patterns in the speech of francophone children with typically developing speech. The conclusion to be drawn in the case of French is that this error is not a backing error at all; it is a spreading error that makes perfect sense in its typical context. In this case it is not atypical, neither from the perspective of its frequency nor from the perspective of its phonological origins.

I have shown that an error pattern that may look similar on the surface can have a very different origin in two languages and thus be atypical in one and typical in the other. My overall conclusion from this discussion only partly echoes that made by Hambly et al: it is true that SLPs need more information about how phonology develops in children who are learning different languages and multiple languages. However it is even more important for SLPs to have the analytical tools to describe children’s phonology at multiple levels of representation (perceptual, articulatory, phonological) and to identify the origin of children’s error patterns after taking all these domains of phonological learning into account.

Further Reading

If you are not familiar with multilinear phonology, here is a source  intended for a clinical audience:
Bernhardt, B., & Stoel-Gammon, C. (1994). Nonlinear phonology: Introduction and clinical application. Journal of Speech and Hearing Research, 37, 123-143.

For another example of “backing” that is typical in another language, see:
Li, F., Edwards, J., & Beckman, M. E. (2009). Contrast and covert contrast: The phonetic development of voiceless sibilant fricatives in English and Japanese toddlers. Journal of Phonetics, 37(1), 111-124.

For our complete multirepresentational explanation of francophone children’s cluster productions, ask us for a copy of this manuscript:
Rvachew, S. & Brosseau-Lapre, F. (accepted with minor revisions). Pre- and post-treatment production of syllable initial /ʁ/-clusters by French-speaking children. In M. Yavas (Ed.), Unusual productions in phonology: universals and language-specific considerations. Psychology Press/Taylor Francis.

Conversations with SLPs (2)

Some of you know that SAILS, my speech perception intervention software, is available for free to any speech therapist working with children who speak a North American dialect of English. The license is available from McGill University and I receive many requests every week along with some feedback and questions from users.  Gissella wrote to me with the following interesting question and I would like to share it and the answer with my readers:

“I am currently using the SAILS program with some children and have found it very helpful. I was wondering if the program could be used by parents with the help (training and treatment) of the treating S-LP. If that is possible, how would the parent go around getting the software.”

I answered that I am not comfortable with releasing the licence directly to parents because I would prefer that the parent be working under the guidance of an SLP so the best thing is for the SLP to apply for the license on behalf of the parent via flintbox .  (By the way I have worked out a way to make this old software run on 64-bit computers – you can download instructions in the revised BACKGROUNDER from the flintbox site. The tablet app is still under construction but I am hopeful that it will be available in 2014).

Further to the topic of parent application of SAILS, I also sent Gissella a copy of my paper – Rvachew, Nowak, & Cloutier (2004) – describing a randomized control trial in which we taught parents to administer SAILS to their children in the clinic, after their child’s regularly scheduled speech therapy session. Children in the control group played with Living Books after their speech therapy session and their parent asked them questions about the story according to sImagecripts that we provided. Both groups showed similar gains in phonological awareness but the SAILS (Sp Percn) group made dramatically better improvements in articulation accuracy as measured by number of errors on the Goldman-Fristoe Test of Articulation (see Figure inserted into this post) and by Percent Consonants Correct as reported in the paper. In that study we didn’t send SAILS home with the parents – as Gissella commented in a subsequent e-mail: “I now realize that the frequency of once per week seems to be sufficient, and therefore it can be done within the clinic most of the time, no need for the parent to do this at home.” This is absolutely correct, in all of my studies we have found that the intensity of the speech perception intervention does not have to be that great – a 20 minute session once per week during the first three sessions on a new phoneme works just fine.

In a follow-up e-mail, Gissella had another really interesting question however; “often times parents ask if they can have this program at home as they see it is easy to use and enjoyable for the child. There has been one case where we had treated the phonological delay, but there was residual articulation (frontal lisp) that will be treated at a later time. The parent had seen how SAILS worked for the phonology part and was interested in continuing exposing the child to the samples for /s/.”  I think that this would be a really interesting application for SAILS but unfortunately there is no empirical evidence to prove that this would be effective. I have always wanted to do a randomized control trial in which children with residual errors in kindergarten were treated with SAILS and then measures of speech accuracy, functional communication and psychosocial outcomes were collected at 6-month intervals for the next two years. Would we see more cases of “spontaneous” resolution of the residual errors in the SAILS group than in the control group? Unfortunately I can’t get decent funding for randomized control trials that involve even children with severe speech sound disorders so I doubt that I could get enough funding for something like this (to have enough statistical power it would have to be a big multisite study with a lot of funding even though 18 percent of 8 year olds can be expected to have speech errors; see Roulstone et al., 2009). You cannot believe how often other researchers, even those working in the field of speech sound disorders, tell me that this most common of neurodevelopmental disorders is just “not important” (see Bishop, 2010)!

Online Gaming and Speech Therapy

I have just read this marvelous paper tweeted out by @vaughanbell: Stafford, T., & Dewar, M. (2013). Tracing the Trajectory of Skill Learning With a Very Large Sample of Online Game Players. Psychological Science. He was impressed by the very large sample size (N = 854,064) but I am impressed by the relevance of this paper for speech therapy. The researchers used “detailed records of practice activity from an on-line game” and used it to test hypotheses about learning in the game which requires “rapid perceptual decision making and motor responses”. Gratifyingly for us as speech-language pathologists, the results confirm the principles of motor learning that are currently promoted for successful treatment of childhood apraxia of speech (CAS), specifically practice intensity, distributed practice and variable practice conditions (for application of these principles to the treatment of apraxia of speech see for example Gildersleeve-Neuman in the ASHA Leader or Tricia McCabe’s ReST program).

There was one concept raised in the paper that was a little bit novel with respect to the CAS literature however: specifically, the authors talk about the “exploration/exploitation” dilemma. In the context of this simple but bizarrely fun computer game (found here at The Welcome Collection)  you can explore the axon growing environment when first learning to play or you can settle into a strategy of simply clicking on the closest protein in your circle of influence. The latter strategy will work to grow your axon which is the object of the game but you will miss out on learning how to maneuver your circle of influence so as to actively find the “power proteins” that advance the growth of your axon. Exploration has a cost in that it leads to more variable performance early on but the benefit is potentially better performance with longer experience. In fact, Stafford et al. observed a close relationship between higher early variance in performance and better performance during later attempts. This trade-off between exploration and exploitation reminded me of the importance of the expansion stage in early speech development and the implications for intervention with young children with CAS.

In Table 10-1 of Developmental Phonological Disorders: Foundations of Clinical Practice we suggest learning outcomes and therapeutic strategies to correspond to four stages of speech development as follows: 1. Expansion stage (explore possibilities of the vocal system); 2. Babbling and integrative stage (controlled variability); 3. Early speech development (expanding repertoire of phones and word shapes to achieve intelligible speech); and 4. Late speech development (ongoing refinements to achieve adultlike speech accuracy and precision). These stages are described in greater detail in Chapter 3 which covers the literature on the development of speech motor control. The expansion stage typically occurs during months 3 through 6 and is characterized by a variety of vocalizations that are not very speech-like (squeals, growls, raspberries and so on) as well as the appearance of fully resonant vowels and marginal babble. It is my experience that SLPs do not appreciate the importance of the expansion stage to normal speech development or understand its significance when planning an intervention program for children with limited if any speech capacity. Therefore I highlight this point in Chapter 10, as follows:

“The importance of the expansion stage in the laying of building blocks for later speech development is easy to forget when choosing goals for speech therapy, a topic to which we return shortly. Another important achievement during the infant period is the acquisition of canonical syllables when the child learns to control the variable parameters explored during the expansion stage, coordinating them to produce well-formed syllables in the context of babble, jargon, and early words. …Typical descriptions of speech acquisition focus on reductions in variability with age. … Therefore, it is not surprising that traditional speech therapy procedures are designed to enhance consistency and reduce variability in the production of phonemes with practice. However, variability is not always an impediment to speech learning and children with DPD often suffer from insufficient variability in their repertoire of speech behaviors. Performance variability can be viewed as facilitating, detrimental, or irrelevant to a successful outcome depending on the motor learning context (Vereijken, 2010). For example, the highly variable vocalizations of the expansion stage provide a complex foundation for the emergence of speechlike vocalizations at later stages. Infants who are described as being “quiet” during the first year of life lack sufficient variability for normal motor speech development. The normally developing infant harnesses rather than reduces this variability to coordinate the separate respiratory, phonatory, resonance, and articulatory components to produce babble in the next stage. Throughout the next 16 or so years there will be a continual interplay between adaptive variability to meet new challenges and increased stability to enhance precision. (p. 758)”

 I often talk to SLPs who are frustrated by failed efforts to teach new phones via imitation to children with severe speech sound disorders. However children with limited vocal repertoires must first be encouraged to freely explore their vocal systems. I describe procedures to encourage vocal play in detail in the book, following Dethorne, Johnson, Walder, and Mahurin-Smith (2009) and supplementing with examples of implementation from my own clinical experience. I hope that Stafford et al.’s interesting research and this amusing little game leads to more reflection about the role of exploration and variability in speech motor learning.