Articulatory Phonetics for Residual Speech Sound Disorders: A Focus on /r/

Abstract

Effective treatment for children with Residual Speech Sound Disorders (RSSD) requires in-depth knowledge of articulatory phonetics, but this level of detail may not be provided as part of typical clinical coursework. . Incorporating contemporary work in the basic articulatory sciences into clinical training becomes especially important at a time when new imaging technologies such as and ultrasound continue to inform our clinical understanding of speech disorders. This is particularly the case for the speech sound most likely to persist among children with RSSD--the North American English rhotic sound, /r/. The goal of this paper is to review important information about articulatory phonetics as it affects children with RSSD who present with /r/ production difficulties. The data presented are largely drawn from ultrasound and Magnetic Resonance Imaging (MRI) studies. This information will be placed into a clinical context by comparing productions of typical adult speakers to successful vs misarticulated productions of two children with persistent /r/ difficulties.

Introduction

Most speech production difficulties in children resolve before the age of seven. Accordingly, the vast bulk of therapy for speech sound disorders is aimed at children in this age group. As other articles in this issue make clear, however, there is a small group of children whose speech production difficulties persist beyond this age. For purposes of this paper, the diagnostic category for these children will be termed Residual Speech Sound Disorders (RSSD) (although see Flipsen [this volume] for discussion of within-category heterogeneity). It is a relatively small group—between 1 and 2% of children [1]. This group, however, is responsible for a significant amount of frustration on the part of clinicians [2,3] and a significant secondary market in remediation advice directed to frustrated clinicians. The issue is particularly important because poor phonetic production skills in children with RSSD are associated with reduced phonological processing and reading skills [4].

Naturally enough, phonetic training in the typical speech language pathology program tends to concentrate on those aspects of phonetic knowledge most applicable to younger children. As it happens, however, treating children with RSSD requires deeper knowledge of articulation than is typically provided as part of clinical education. This may be because, unlike children who respond to therapy at earlier ages, children with RSSD have effectively become enmeshed in ineffective articulatory habits that must be “unwound” and rebuilt. Because these habits are embedded in speech, they must be carefully analyzed from three points of view: (1) phonological, (2) phonetic, and (3) physiological. The issue of phonetic knowledge is particularly important for clinicians treating RSSD, because the increasing availability of imaging data, and in particular ultrasound imaging data, is creating an opportunity for clinical strategies to be refined and refocused [3,5-8].

Although children with RSSD may present with misarticulations of a number of different sounds, the most numerous, and most frustrating, are children with misarticulations involving the English rhotic phoneme /r/ [2]. Accordingly, this article will concentrate on the phonetics of /r/. In particular, it will emphasize the acoustic and articulatory role of tongue configurations in the pharyngeal portion of the vocal tract.

In the initial portion of this article we will review linguistic and phonetic background information for this sound, including the use of phonetic symbols. The second portion of the article is devoted to a short review of what we know about the articulation of /r/ in typical adult speakers, along with illustrations from Magnetic Resonance and x-ray images from a range of speakers. In the final portion, we review discuss specific differences between successfully generated and misarticulated productions by two children with a history of RSSD for /r/, as shown by Magnetic Resonance Imaging (MRI) data, in the context of ongoing treatment..

Linguistic Background

Linguists agree that the rhotic liquid of English is a single phoneme [9] and that certain articulatory movements must occur in order for a typical acoustic profile and an acceptable percept to occur. In IPA notation, this sound is represented by [ɹ], which specifies that the sound in question is an approximant with a primary constriction at a point along the palate that may range from alveolar to post-alveolar to palatovelar. The American phonetic tradition, which is followed by most clinicians, is to use the Roman alphabet symbol [r]. American phoneticians have a strong tradition of distinguishing between cases where /r/ is used as a syllabic onset or coda (i.e. a consonant), vs. when it is used as a syllabic nucleus (i.e. more like a vowel). One method is to use a single syllable with a syllabic diacritic, [ṛ]. An alternate tradition is to use the [ɚ] and [ɝ] symbols for unstressed and stressed syllabic nuclei. This usage is historically linked to attempts among phoneticians to document similarities and differences across the different rhotic and non-rhotic, or “r-less” dialects of English but has lingered in descriptions of rhotic dialects as a means of indicating situations where the rhotic liquid acts phonologically as a syllable nucleus.

In general, clinical practice has followed the more common American tradition of multiple symbols for /r/. This preference has gained traction from the convenience of marking a very real clinical phenomenon. Children with residual speech sound disorders frequently show an asymmetrical pattern of ability to produce positional variants of /r/. In other words, many children with RSSD have difficulty with /r/ in postvocalic position but find syllable-initial position easier. Others have more difficulty with syllable-initial position than postvocalic position[10]. Clinicians have typically responded to this asymmetrical pattern by organizing therapy protocols and strategies according to a “consonantal”, or “vocalic”, vs “syllable-initial” or “postvocalic” schema[11]. Notably, however, these clinical schemas differ largely in the determination of which phonetic contexts, words, and larger sequences are most important for stimulation and practice, and in methods for encouraging children with an acceptable variant in one position to generalize to the other position. Because the phoneme is the same, the acoustic profile is internally consistent, and the same articulatory movements must occur for the same acoustic profile to be generated, the schemas for different positional variants involve the use of similar clinical instructions for shaping vocal tract postures [11].

Because this paper is focused on the basic phonetics, and in particular the articulatory phonetics, of late-acquired sounds, we will emphasize the articulatory sameness of the rhotic phoneme across different syllable positions by using the symbol /r/ (with a syllabic diacritic when required) rather than the two symbols /r/ and /ɚ/. The symbol is shown with slashes rather than brackets to indicate the status of /r/ as a single phoneme. A full treatment of the phonetics of positional variants of /r/ is beyond the scope of this paper. One fact, however, is worth mentioning. For rhotic dialects, there is little difference between the vocal tract shapes that distinguish consonantal (i.e. onset and coda) productions of /r/ and vocalic (i.e. syllabic) productions of /r/. Consonantal productions appear to be stronger and more extreme versions of syllabic productions. Thus, what is true of syllabic productions can be reasonably extended to onset and coda productions.

Phonetic Reasons for RSSD Children’s “Resistance” to Therapy

Most clinicians have an arsenal of techniques for teaching sounds to children. These typically focus on the tongue shapes illustrated in textbooks, or on introspective analysis of their own productions. A common approach for teaching /r/, for instance, is to instruct the child to position the tongue dorsum so as to feel the molar teeth[11-13]. Other clinicians may instruct the child to feel the alveolar ridge with the tip of the tongue and then curl the tip of the tongue backwards[11,14]. In many cases, however, RSSD children have not responded to standard techniques. While there are many possibilities for why these children do not respond—e.g. difficulty following instructions, reduced auditory perception, etc. [15]—one major possibility is that the tongue configuration inherent in the instructions does not work for that child’s vocal tract. It is therefore essential that clinicians possess articulatory knowledge of /r/ (and other sounds) at a level of detail required to provide articulatory alternatives when particular clinical techniques are not effective.

Variability of Tongue Shapes

Phonetics textbooks generally describe sounds in terms of their primary “place of articulation”. For acoustical purposes, this is better described as “primary place of constriction”—that is, the primary place where the vocal tract airspace is “constricted”, or narrowed so as to produce the desired acoustic result. However, for many sounds the vocal tract airspace is also narrowed in more than one location. These sounds are referred to as “doubly” or even “triply” articulated. The order of places of articulation as “primary”, “secondary”, or “tertiary” follows the degree of constriction; the primary place of articulation is the location where the constriction is narrowest. The secondary or tertiary locations are necessary for production of an acceptable acoustic version of the phoneme, but their degree of constriction is either less, or more variable in different prosodic or phonetic contexts[16,17]. One example is the vowel-semivowel pair /u,w/. Although /u/ is typically listed as a back vowel, and the semivowel /w/ is listed as a consonant, both involve a constriction of the lips and the tongue dorsum in the vicinity of the velopalate [18].

With regard to /r/, clinicians are taught to focus on a single “place of articulation” (in the alveolar/postalveolar region). Most textbooks mention two alternative configurations, (1) a “bunched” configuration with the tongue dorsum raised toward the middle portion of the palate and the tongue tip lower than the dorsum, and (2) a “retroflex” configuration with the tongue tip raised toward the front portion of the palate and the dorsum lower than the tongue tip. These configurations are illustrated in the stylized x-ray tracings on the left side of Figure 1 [9,18]. However, tongue configurations for /r/ are significantly more variable, spanning a continuum from “bunched” to “retroflex” and including several types with both tongue tip and dorsum raised. This variability is illustrated in Figure 2, which shows midsagittal Magnetic Resonance (MR) images of 21 different typical speakers producing sustained syllabic /ṛ/ as in the word “pour”. The figure shows 22 images because one of the speakers (Speaker 5, shown in the second and fifth column, going from left to right) was able to produce a perceptually “correct” /r/ using two different tongue configurations. Each speaker’s production sounded perceptually “correct” as determined by trained listeners. Because the MR images of Figure 2 were obtained from syllabic /ṛ/ productions, one question that might arise is whether the tongue configurations also occur in syllable onset position. Note, however, that a parallel set of ultrasound images was collected from each speaker producing words with syllable onset /r/ in the word “role”. Each of the tongue shapes shown in Figure 2 also occurred in syllable onset position. A similar range of variability across speakers has been noted by other investigators [9,19-21].

Figure 1.

Types of American English tongue configurations for /r/ as identified by Delattre and Freeman (1968) from x-ray tracings of 44 subjects. Only types seen in “rhotic” dialects are shown. All types were found in both prevocalic (syllable onset), syllabic (nucleus) and postvocalic (syllable rime) positions. Tongue configuration types have been renumbered for purposes of this paper. Figure adapted with permission from Hagiwara (1995). Type #1 and Type #4 are examples of configurations typically called bunched (#1) and retroflex (#4). They correspond roughly to images shown in Figure 2 for Speakers 22 and 5.

Figure 2.

Midsagittal Magnetic Resonance images of 21 different typical adult native speakers of rhotic dialects of American English speakers producing sustained /ṛ/ as in the word “pour”. All images are shown in the midsagittal plane facing left. As noted in the text, similar vocal tract shapes have been recorded dynamically from the same set of speakers in prevocalic positions using ultrasound. There are 22 images because Speaker 5 was trained to produce a retroflex shape similar to that of #4 in Figure 1, as well as his natural bunched shape, similar to that of #1 in Figure 1. Note that MR images reflect the density of hydrogen atoms in tissue vs. air. Bone, air and teeth all appear as dark areas on the image, meaning that it is not possible to separate teeth from air when they are contiguous. Codes identifying each speaker are numerical and appear at the bottom right edge of each image.

Looking at Figure 2, several points are apparent. First, the variability across typical speakers suggests that there are a number of different ways to shape the vocal tract to produce the acoustics of /r/. The different tongue shapes in the figure are arranged roughly according to their similarity to each other and to illustrate the continuum from “bunched” configurations with the tongue dorsum raised and the tongue tip down, to “retroflex” configurations with the tongue dorsum lowered and the tongue tip raised, through configurations with both tip and dorsum raised. While logically it is possible that a single shape would work for all vocal tracts, the variability we see suggests that different speakers use different paths to find a shape that works for them, and it is likely that some shapes work better with some vocal tracts than others. For instance, although Speaker 5 was able to produce a perceptually correct /r/ using both a raised and lowered tongue tip, his natural tongue configuration was that of the classic bunched shape shown in the second column. Significant coaching was required to achieve the tip-up shape illustrated in the column on the far right and to sustain it for the 20 seconds required for the MRI scan. From an intervention point of view, this means that teaching a single tongue configuration for /r/ may not work for all vocal tracts.

Second, the different tongue configurations in Figure 2 show a range of locations for “place of articulation”—that is, the point of greatest vocal tract narrowing along the palate. For some speakers, this point is closer to the alveolar ridge. For some speakers, it is in the palatovelar position very close to the location of narrowing for the /u/ vowel, the homorganic semivowel /w/, and velar stops /k/ and /g/. Thus, although many textbooks refer to /r/ as having an alveolar place of articulation, it is more accurate to say that it has a relatively undefined “palatal” or “post-alveolar” primary place of articulation. As noted above, this is in fact the current stance of the international phonetic association for the IPA symbol /ɹ/.[22]

Third, and perhaps most importantly, each of the tongue configurations shown in Figure 2 exhibits an aspect of /r/ production that has been relatively unknown among clinicians until recently. This is the existence of a secondary place of constriction in the pharynx as an obligatory feature of the /r/ phoneme. The existence of a pharyngeal constriction was noted in Delattre & Freeman’s (1968) study, but has not typically been emphasized in textbooks or educational materials directed to clinicians. The mechanism for this pharyngeal narrowing is a movement of the tongue root toward the back pharyngeal wall, independent of the movement of the tongue dorsum or tongue blade/tip toward the palate. (Note that when the tongue root moves backwards it can sit flush with the epiglottis, and the narrowest point of the vocal tract in the pharynx may in fact include the projection of the epiglottis into the pharyngeal space.) Separate acoustic investigations [23,24] [25,26] have also shown that for attested vocal tract configurations this pharyngeal constriction is a major contributor to the acoustical profile of /r/. Theoretically, it is possible to produce the characteristic acoustic profile of /r/ using lingual movements alone [26,27], but corresponding vocal tract configurations have not been discovered in actual speakers[23,24,28]. The degree of pharyngeal constriction is not the same for all speakers. As Delattre & Freeman[9] noted from x-ray observation, and Zhou et al. [24] showed using MRI data and acoustical modeling, the pharyngeal constriction in American English speakers tends to be narrower for tongue configurations with a raised tongue tip, and wider for tongue configurations with a “bunched” configuration and a lowered tongue tip.

Role of Tongue Groove and Sides

It is well known that the coronal (i.e. cross-sectional) profile of the front part of the tongue for /r/ is either relatively flat or shows a mild groove [11]. This profile is illustrated in Figure 3 below, which shows coronal slice images of the tongue blade during /r/ from two typical adult speakers, along with midsagittal images showing the location of these slices along the vocal tract.

Figure 3.

Corresponding Midsagittal (left) and coronal (middle and right) plane images of Speakers 22 and 5 from Figure 2. Red lines labelled 1 and 2 on the midsagittal image indicate the location of the coronal slice. The lateral edges of the tongue at this point along the vocal tract are shown by red arrows on the coronal slice images. As with Figure 2, both air and teeth are shown as dark, while tissue is shown as gray or white depending on hydrogen atom density. The palate is shown as a bony ridge above the tongue.

What is not well known among clinicians is that for /r/ the midsagittal grove continues along the length of the tongue through the tongue dorsum and root. Further, if the tongue dorsum or the tongue blade or both are lifted toward the palate while the tongue root is retracted toward the back pharyngeal wall, the midsagittal shape of the tongue along the groove will often show a dip, or dimple at the point of separation between the tongue dorsum and tongue root. This dip between the tongue dorsum and root is particularly noticeable for the speaker productions in the second column from the left of Figure 2 (Speakers 3, 7,11,5 and 17). This aspect of tongue configuration is extremely important for understanding images of the tongue because different impressions of tongue configuration can be gained depending on whether the image is derived from a slice along the midline groove, along one or the other side of the tongue, or some combination. Thus, imaging modalities such as ultrasound and MR, that effectively slice the vocal tract into sections, will show a different view of sagittal tongue shape depending on the thickness of the slices, and for narrow slices, whether they are imaging the tongue down the middle or along the sides.

This issue is illustrated best by modern x-ray imaging technology using a radiographically opaque medium such as barium to outline the tongue groove. An example is shown in Figure 4, which shows the tongue outline during a production of “year” in the sentence “where were you a year ago?” The speaker was an elderly man with normal speech who was imaged during a routine clinical swallowing evaluation[29]. His production of /r/ in “year” was typical for speakers of a rhotic dialect of North American English.

Figure 4.

Modified Barium Swallow image of elderly male speaker of American English producing coda /r/ as in “year” with barium medium showing contrast between tongue sides and midline groove. The leftmost panel shows the untouched image. The rightmost panel shows outlines drawn along to show the contrast between the sides of the tongue and the midline groove.

In this figure, the left hand panel shows the untouched image. The shape of the tongue sides are visible against the airspace of the vocal tract, while the shape of the groove through the tongue midline is shown by the darker line where barium material pooled and clung. The right-hand panel shows a hand-drawn outline of the different paths of the tongue sides and groove, along with anatomical landmarks. Note that the shape of the groove includes a dip between the tongue dorsum and the tongue root, similar to that shown by Speakers 3, 7, 11, 5 and 17. The shape of the tongue configuration along the sides shows the tongue tip lower than the tongue dorsum. The point of greatest constriction is in the vicinity of the mid-palate. The shape of the palate is cut off in the image at its highest point, but the shape of the tongue itself is clear. Note also that the vallecula is shown as a clear space between the tongue root and the epiglottis where the barium material has pooled. (This pooling of prandial material was the main clinical focus of the barium swallow study.)

It is important to note that because all of the 22 images of Figure 2 were made along a midsagittal plane, they represent the midline portion of the tongue and thus are a reasonable guide to the shape of the groove from tongue front to tongue root. At the same time, an observer looking only at a midsagittal image may underestimate the degree to which the sides of the tongue project into the pharyngeal airspace, and may thus underestimate the degree of pharyngeal narrowing. This point is illustrated in Figures 5, which shows two horizontal (axial) slices through the vocal tract (right-hand panels) along with the relevant locations of the slices in the midsagittal plane (left-hand panels), for Speaker 5. The two axial slices show the midline tongue groove at the level of the tongue tip and at the narrowest portion of the pharynx. Tongue configurations for other speakers shown in Figure 2 likewise show that the tongue sides are higher and show less of a dip.

Figure 5.

Corresponding Magnetic Resonance Images in Midsagittal and Axial (Horizontal) planes from Typical Adult Speaker 5 of Figure 2 at level of tongue tip (left) and level of narrowest constriction between tongue root and back wall of pharynx (right). Note that a section of the velum hanging in the airspace behind the tongue dorsum can be seen as a bar of flesh bracketed by dark areas before and behind.

Clinical Relevance of Pharyngeal Constriction and Tongue Groove

Clinicians can use accurate knowledge of these aspects of typical production to evaluate their favorite strategies for eliciting /r/, or to devise new strategies. To take an example, a well-known strategy for eliciting /r/ is to instruct a child to “make a boat shape with the tongue”[30]. This instruction may encourage the child to form a groove along the midline of the tongue dorsum and root, while maintaining a constriction along the palate. Because forming a groove requires depression of the tongue along the midline, the effect may be to move the tongue root toward the back pharyngeal wall, producing a pharyngeal constriction.

Many clinicians are taught that /r/ has an alveolar constriction alone, and that children with /r/ misarticulations should be discouraged from making a constriction at the lips. In our clinic, it is common to find that RSSD children have been instructed to spread or pull back their lips when attempting to produce /r/. This admonition probably stems from the fact that many /r/ misarticulations sound like /w/ or /u/, which have relatively extreme and visible lip constrictions involving protrusion as well as narrowing. However, typical adult /r/ also involves a narrowing of the vocal tract at the lips. The lip movement involved is mild compared to constrictions common to sounds such as /u/ and /w/, and it is typically more extreme in syllable-initial position [9] [31] [32] [33]. Acoustically and perceptually, however, lip constriction is the least relevant contributor to a correct /r/. In other words, if the tongue configuration is appropriate for /r/, adding a lip constriction may affect its naturalness but not its phonemic identity. Conversely, if the tongue configuration is not appropriate, changing the lip constriction cannot substantially improve it [23,24].

Although many textbooks suggest that the retroflex tongue configuration with tongue tip raised and dorsum lowered (shown for Speakers 5, 1 and 22 in Figure 2) is the most prevalent version, larger studies have found this to be the rarest type [9,20,21]. Speakers who use this extreme retroflex configuration tend to use it only in syllable-initial position and that many speakers switch to a more bunched configuration depending on phonetic context [9]. Similarly, in a study of 27 adults using ultrasound imaging Mielke and colleagues found that a significant fraction of adult typical speakers switch configurations across contexts. These facts are very relevant to intervention. It is likely that retroflex tongue positions are more rarely used as a speaker’s natural /r/, and/or are more likely to be switched out, because the anatomy that favors them is rarer. Possibly, coarticulation with other sounds and/or prosodic conditions is more difficult. As indicated above, it is likely that RSSD children need explicit instruction and target structuring in order to generalize newly learned articulatory strategies across different syllables, words and sentences. Thus, even if a child is successful at producing /r/ with a retroflex tongue configuration in some contexts, they may not be able to use it in all contexts. If progress in therapy has ground to a halt, it may be helpful to teach a different tongue configuration and to explore its use across contexts.

One question that arises is whether the most appropriate /r/ tongue shape for a particular child can be predicted from anatomy. If so, clinicians might be able to predict the most appropriate tongue configuration for /r/ from observations gained in an oral mechanism examination. There is strong evidence that vowel articulation is influenced by palate shape [34,35] and by pharyngeal cavity space [36]. Although these apply to vowels rather than /r/ specifically, they do suggest that, for any one speaker, anatomical influences such as palate shape, pharyngeal space, and oral cavity length limit the range of tongue configurations that will produce a “correct” /r/. On the other hand, Westbury and colleagues did not find a correlation between palate length and /r/ configuration types[19,37]. It is clear that many vocal tracts are broadly compatible with both shapes, since many typical speakers use different tongue configurations across different phonetic and prosodic contexts[20,38,39], and some vary tongue shape in response to a palatal prosthesis[40]. Further, the typical oral mechanism evaluation does not include an analysis of oral cavity length or pharyngeal area. More research is needed in order to provide clinical guidance of this nature. Given the degree of variability in typical adult speakers, clinicians should not assume that they know, a priori, which tongue configuration is the one that will work best for a particular child. Instead, a child’s most workable tongue configuration should be determined by trial and error[10] .

Phonetics of Misarticulations in RSSD Children

The preceding sections of this paper deal with typical adult patterns of producing /r/. In the following sections, we will discuss three types of commonly found patterns of misarticulation. These error productions will be contrasted with typical child and adult productions of related sounds. The data will be drawn from a Magnetic Resonance Imaging study of RSSD children between 8:2 to 11:10 years of age. Each child produced /r/ in a target word, and sustained that production for 8 seconds. Children with a RSSD diagnosis were imaged producing both a typical misarticulated version of /r/ and their “best” /r/. This was accomplished by asking children to identify words they had trouble with in terms of /r/ articulation, and words they were most successful with. They were instructed to start to say the word in question and then to sustain the /r/. Children were familiarized with the task in a training session using ultrasound before they undertook the MRI session. During the MR session itself, if movement artifacts in the image were identified, the production trial for that sound was repeated. Images with significant movement artifacts were discarded.

To ensure that the children’s misarticulated and correct productions in MRI sessions were typical of their utterances under more normal circumstances, the midsagittal MR images were compared to midsagittal ultrasound images collected during the children’s productions of the same target words during therapy sessions. A trained clinician with experience of the children’s performance during therapy was also present for the MR sessions and determined that the children’s productions were impressionistically similar to their characteristic misarticulations and corrected productions in therapy sessions. Note that this could be done for the MR sessions only at the very beginning of the 8 second sustained interval, because the onset of noise from the magnet masked the subsequent sound from the child. Figure 6 shows examples from two children who each produced a clearly misarticulated and an acceptably “correct” version of /r/. The misarticulated version is labelled “error” and the more acceptable version “good”.

Figure 6.

Magnetic Resonance (MR) Images of two children in therapy for Residual Speech Sound Disorders (RSSD). Child Speaker 1 (age 10.5) is shown on the left and Child Speaker 2 (age 9.0) is shown on the right. Sustained misarticulated versions of /ṛ/ are shown above versions judged to be “good”.

Child Speaker 1 originally came to the clinic for ultrasound therapy after many years of conventional therapy in his school setting. He was unable to produce an acceptable /r/ in any context. In word-initial and cluster contexts, his productions were transcribed as /w/ but were sometimes heard as /l/. In all other contexts, his productions were typically transcribed as sounding like /ʊ/. He had been instructed to use an exaggerated retroflex tongue configuration but more frequently used a tongue configuration with a high, humped dorsum and a lowered tongue tip with no sign of a separate tongue root movement toward the back pharyngeal wall. Therapy focused on suppressing the exaggerated retroflex tongue configuration and on finding an alternative configuration for the tongue blade/tip or dorsum that he could combine with tongue root movement in a backwards direction. After 4 sessions of therapy with ultrasound, he could produce a good /r/ in most words if prompted to remember his best tongue shape. During good /r/’s he showed consistent tongue root retraction. His error /r/’s showed inconsistent tongue root retraction along with inconsistent control of the sides of the tongue. The MR session was held during this phase in his treatment.

The tongue shapes shown in Figure 6 for Child Speaker 1 (left side) are representative of the tongue configurations he showed with ultrasound in therapy. The error vs. good /r/ tongue configurations are clearly different in both overall shape and in the location of vocal tract constrictions. For both, the primary constriction is made by the tongue blade in the vicinity of the alveolar ridge. For both, there is an apparent secondary constriction in the region of the pharynx. For the error /r/, however, the tongue tip is lower and the entire front of the tongue including the tongue blade and tip is more retracted. In contrast, for the good /r/, the tongue tip is raised to be closer to the alveolar ridge itself and the entire blade/tip complex is stretched to be more forward in the mouth. This causes the primary constriction itself to extend for a longer distance, to encompass a more forward location, and to be notably narrower along its length. Although it is hard to see from this midsagittal view, the retraction of the tongue root is more pronounced in the good /r/. Perhaps the most notable aspect of the good /r/ tongue configuration is the suggestion of a dip in the midline groove between the tongue front and tongue root. This dip is not present in the error /r/ tongue configuration.

It is also worth pointing out that although this speaker’s error tongue shape resembles that of many adult typical speakers shown in Figure 2 (e.g. Speaker 5), the similarity breaks down when the total shape of the vocal tract is taken into account. Relative to the adult vocal tract shape, the child error production shows a palatal constriction that is slightly further back. Further, his error production a reduced pharyngeal constriction. The child speaker’s good /r/ more closely resembles the typical adult vocal tract configurations shown in the two left-most columns of Figure 2, both in terms of constriction location and degree.

Although the good /r/ configuration was elicited for this child using ultrasound imaging, similar results may be obtained with clinical strategies that encourage separation of these two parts of the tongue, or enhancement of the midline groove. An example of the former is the instruction “say “ah” and raise the tip of your tongue without moving the rest of your tongue”. This works because the vowel /α/ involves pharyngeal constriction. An example of the latter is the previously mentioned “make a boat with your tongue” instruction [11,30].

Child Speaker 2 also came to the clinic for ultrasound therapy after many years of conventional therapy and he was originally unable to produce an acceptable /r/ in any context. In word-initial and cluster contexts his productions were transcribed as sounding like /w/ but in other contexts they were transcribed as sounding like /ɔ/. The evaluating clinician also noticed that his /r/’s were accompanied by what she called a “gurgling” sound. This clinical observation is commonly noted when native English speaking clinicians hear something like a uvular fricative or trill. Sounds with a uvular place of articulation are common in languages such as French and German. The place of articulation for such sounds is the upper pharynx [41].

When he came to the clinic, this child used a tongue configuration with the tongue tip raised in an exaggeratedly retroflex posture with the tongue curled back. Although encouragement of a tongue curl posture is a popular remediation strategy, the posture itself has not been observed in typical speakers of American English[9,11,38]. There was no sign of a separate tongue root movement toward the back pharyngeal wall. Since the child had not succeeded in generalizing to correct /r/ in a variety of words, therapy focused on experimentation with different tongue shapes that allowed him to maintain a constriction along the palate along with simultaneous tongue root retraction. This was done using ultrasound and the MR images of Figure 2 as a guide for reference. This child’s error /r/ and good /r/ tongue configurations are shown on the right side of Figure 6. The MR session was held during a phase in his treatment where he managed to separate the tongue front from the tongue root movement but still inconsistently produced a “gurgling” sound when his palatal constriction was too far back. They are representative of his ultrasound tongue configurations during therapy.

Unlike Child Speaker 1, the general tongue shapes for Child Speaker 2 are similar for both error and good /r/’s. Both show a clear dip in the midline groove and resemble tongue shapes shown in Figure 3 for adult typical speakers. Both show pharyngeal constriction by the tongue root as well as palatal constriction by the tongue front at the same level of the spinal column. Both show a raised tongue blade and a lowered tongue tip. For his error /r/’s, however, this child has positioned his tongue front and tongue root constrictions in the wrong part of the vocal tract. In particular, the location of the palatal constriction is too far back along the palate and there is too much air in the front cavity under the alveolar ridge. Further, the pharyngeal constriction appears to be too narrow. This vocal tract configuration used by Child Speaker 2 for /r/ is similar to that described for uvular trills in languages that use them.[42] It is likely that the impression of “gurgling” derives from the child’s attempt to move the entire tongue backward rather than to separate movements of the tongue front and tongue root, causing the aerodynamic conditions for trill or fricative vibration in the upper pharynx.

The good /r/ production corrects this problem by moving the front of the tongue (i.e. the blade/tip) forward along the alveolar ridge so that the narrowest part is directly below the highest portion of the palate. Note that in the good /r/, the tongue root remains very close to the back wall of the pharynx, meaning that the two parts of the tongue are stretched away from one another. Because the good /r/ does not include any gurgling or frication noise, we can conclude that the constriction is wide enough to obviate aerodynamic conditions for a fricative or trill.

In our clinic, access to the MR image for this child was extremely helpful in explaining why his progress in therapy had halted. From an ultrasound image, it is not possible to determine exactly where the parts of the tongue are with respect to other structures of the vocal tract that are not imaged, such as the palate or the pharyngeal walls. In the case of Child Speaker 2, both his error and his good /r/ productions showed what looked like the same tongue shape on the ultrasound screen. The MR images, however, showed that he needed to increase the distance between his tongue root and tongue front, stretch the front part of the tongue further forward along the palate, and pull the tongue root slightly away from the pharyngeal wall. This observation suggests that any child who presents with a “gurgling” or uvular-sound error productions may be showing the maladaptive behavior of moving the tongue backwards as a single unit.

Summary and Conclusions

Remediation for these RSSD children requires the development of new articulatory habits and the development of new interactions between the motor planning components and the phonological components involved in speech. Careful consideration of imaging and other phonetic data from typical speakers can elucidate the maladaptive behaviors behind misarticulations, and suggest alternative intervention strategies. For instance, data from multiple sources indicate that accurate production of /r/ requires a tongue root movement toward the back pharyngeal wall along with a tongue front (i.e. either blade/tip or dorsum/blade) movement toward the palate. Although they showed different types of misarticulations, for both Child Speakers, improvement in /r/ production was accompanied by clear separation of the tongue front and tongue root. Typical speakers use a number of different tongue configurations to produce /r/, and many speakers switch back and forth between tongue configurations according to different prosodic and/or phonetic contexts [3,7,43,44]. Incorporation of this knowledge, and additional knowledge derived from articulatory techniques, could prove highly beneficial to clinicians treating children with RSSD, and to the practice of speech-language pathology more broadly.

Abbreviations

CCHMC

Cincinnati Children’s Hospital Medical Center

MRI

Magnetic Resonance Imaging

UC

University of Cincinnati

RSSD

Residual Speech Sound Disorder