Common Sounds Audiograms: Quantitative Analyses and Recommendations

Abstract

A counseling tool routinely used by pediatric audiologists and early intervention-specialists is the often-named “common sounds audiogram” (CSA). Typically, a child's hearing detection thresholds are plotted on the CSA to indicate that child's audibility of speech and environmental sounds. Importantly, the CSA may be the first item that parents see when their child's hearing loss is explained. Thus, the accuracy of the CSA and its associated counseling information are integral to the parents' understanding of what their child can hear and to the parents' role in the child's future hearing care and interventions. Currently available CSAs were collected from professional societies, early intervention providers, device manufacturers, etc., and analyzed ( n  = 36). Analysis included quantification of sound elements, presence of counseling information, attribution of acoustic measurements, and errors. The analyses show that currently-available CSAs are wildly inconsistent as a group, not scientifically justified, and omit important information for counseling and interpretation. Variations found among currently available CSAs can lead to very different parental interpretations of the impact of a child's hearing loss on his/her access to sounds, especially spoken language. Such variations, presumably, could also lead to different recommendations regarding intervention and hearing devices. Recommendations are outlined for the development of a new, standard CSA.

A counseling tool routinely used by pediatric audiologists and early intervention-specialists for families and caregivers of children diagnosed with hearing loss is the often-named “common sounds audiogram” (CSA). Graphically, a CSA is similar to an audiogram, but with overlaid shapes and symbols corresponding to the sound levels and frequencies of speech and environmental sounds. While many varieties of the CSA are available, they generally include these three main elements: (1) the speech region, also known as the “speech banana” which is a shape that represents the dynamic range and spectral spread of speech; (2) letters that represent individual speech sounds, positioned at specific intensity levels and frequency locations; and (3) symbols that represent environmental sounds such as birds or an airplane, positioned at specific intensity levels and frequency locations.

A child's hearing detection thresholds are typically plotted on the CSA to indicate the audibility of speech and environmental sounds. Importantly, the CSA may be the first item that parents see when their child's hearing loss is explained. In an article published in Hearing Health , ¹ parent Joey Lynn Resciniti describes her initial encounter with the CSA and how the CSA helped her understand her child's hearing loss. “The first thing the audiologist showed us after the ABR (auditory brainstem response) testing was the “speech banana.” This was a confusing bit of information at first. Banana? Speech? The speech banana is a visual aid for a quick introduction to hearing loss and the varying levels of severity. […] All speech sounds (vowels as well as consonants and consonant pairs) are above where Julia can hear with her 55 dB loss. Theoretically, without hearing aids she can't hear any of those sounds. Things louder than 55 dB, like a dog barking or a piano, would be accessible for her without hearing aids. But the tricky part is that it isn't so cut and dry. Julia wasn't unable to hear all language, and she also wasn't always able to hear dogs barking.” This parent learned what her daughter could hear based on the CSA and her daughter's thresholds, but she also noted ambiguity in the interpretation.

Initial counseling sessions and parents' understanding of their child's hearing loss are integral parts of a child's future hearing care and interventions. Might CSAs, published by different reputable entities, lead to parents having very different understandings of what their child could hear? For example, imagine a child with detection thresholds of 40 dB HL (hearing level) from 250 to 8,000 Hz. These thresholds, when plotted on two readily available CSAs, display dramatically different speech-audibility levels. In Fig. 1A , ² approximately two-thirds of the “speech banana” and 16 of the 23 (70%) individual speech sounds are within this child's hearing range. By contrast, in Fig. 1B , ³ for these same hearing thresholds, only about one-fourth of the “speech banana” and only 4 of the 25 (16%) individual speech sounds would be within the patient's hearing range. These differences in hearing scenarios, due to differences in the CSAs, would presumably lead to significantly different sets of advice, especially regarding intervention and hearing devices. This observation, that different CSAs would lead to different interpretations of the impact of hearing loss on the audibility of speech sounds, is described in a well-regarded Audiology textbook ⁴ : “…this borders on malpractice, as a child suffering from middle-ear effusion (which often results in a hearing loss of 30 to 35 dB) might go untreated for months or years, simply because the parents were given the wrong counseling based on the wrong banana.” Thus, the CSA selected by the audiologist has the potential to impact the parents' perceived severity of their child's hearing loss and any subsequent actions taken by the family and caregivers.

Figure 1.

( A ) American Academy of Audiology CSA. ² (Repurposed and republished with the permission of the American Academy of Audiology. www.audiology.org .) ( B ) Central Institute for the Deaf CSA. ³ (Repurposed and republished with the permission of the Central Institute for the Deaf.)

Additionally, parents, such as Resciniti, note that their interpretation of their child's thresholds on the CSA does not match their daily experiences (i.e., how much their child seems to hear in typical situations). Several factors likely contribute to this. Even in quiet places, the audibility of speech depends not only on a listener's hearing thresholds but also depends strongly on the listener's distance from the person talking. Other factors are the talker's vocal effort, and the talker's gender and age (e.g., /s/ sounds produced by young children have much higher spectral peaks than do /s/ sounds produced by adult male voices, > 8 vs. 5.4 kHz; see Figure 5). ⁵ For a particular listener's hearing loss, speech energy might be mostly audible if the talker is an adult male standing 2 feet from the listener and exerting strong vocal effort, but mostly inaudible if the talker is a soft-spoken child distant (across a large room) from the listener. (Incidentally, similar factors affect the audibility of a dog bark, i.e., distance between the dog and the listener, vocal effort of the dog, and dog characteristics/species.) These important factors, however, are not specified on these two CSAs, and might not be discussed or mentioned when counseling parents of children with hearing loss.

With the aim of providing future audiologists' connections between speech acoustics coursework and clinical applications, a casual online search of CSAs was done by one of the authors. Substantial variations among CSAs were immediately apparent, and inspired this quantitative study. Since the CSA is a commonly used counseling tool, and its interpretation by parents may have significant effects on patient outcomes, it is important for CSAs to be consistent (i.e., standardized). Additionally, CSAs should be scientifically justified and should display clearly the most relevant speech information. This study, then, had three aims: (1) collect existing CSAs; (2) describe and analyze the CSAs, including the size and location of the “speech banana,” location of speech sound “letters,” counseling information, and cited references (if any); and (3) provide recommendations for a new CSA-like counseling tool.

Methods

Collection

Existing CSAs were collected using two primary methods, both using the internet. First, CSAs were sought directly by searching these types of sites: organizations and associations related to audiology and hearing loss, early intervention resources for hearing loss, audiology clinics, manufacturers of hearing devices, audiology textbooks, and journal articles. Second, CSAs were sought using the following keywords and phrases: “speech banana,” “common sounds audiogram,” “familiar sounds audiogram,” “understanding audiogram,” “what is an audiogram,” “pediatric audiogram,” “hearing loss early intervention,” and “my child has hearing loss.” For these keyword searches, two internet search engines, Google and DuckDuckGo, were employed.

Quantification

For each existing CSA, data were extracted using the software WebPlotDigitizer (version 4.1). When present in the CSA, data (i.e., dB HL and frequency values) were extracted corresponding to the speech region, speech-sound “letters,” and environmental sounds. For the area representing the speech region, minimum and maximum levels (dB HL) were extracted at half-octave frequencies from 125 to 8,000 Hz, and the lowest and highest frequencies (Hz) were also extracted (“frequency range”). These data were then used to characterize the speech region using two parameters: the mid-position (mid-point of the dynamic range) at 1,000 Hz and the percentage of the total audiogram area covered by the speech region. Speech-region area, as a percentage of the total audiogram area, was calculated by dividing the sum of the dynamic ranges of the speech region at each half-octave by the total area of a standard audiogram (12 half-octaves × 130 dB = 1,560 Hz × dB). Together, these two parameters, mid-position and area of the speech region, reflect the overall level and size of a particular CSA. For all speech-sound “letters” and environmental sounds, frequency (Hz) and level (dB HL) were measured, using the visual center of each letter (or letters), or of the environmental image.

Counseling Information

Each CSA and its surrounding context, such as web page or article, were examined for the presence of information regarding the interpretation of the audibility of speech sounds or of environmental sounds. Specifically, the presence (or absence) of information about these four items was documented: (1) level of speech (or of environmental sounds) in sound pressure level in decibels (dB SPL); any mention of dB SPL versus dB HL; (2) distance from the talker (or from the environmental sound source); mention of the relationship between distance and sound level; (3) talker's vocal effort; and (4) distinction between sound detection (or audibility) and sound recognition (or identification). The first three items were included to reflect whether the CSA's creators justify or explain the details of the speech region, especially its overall level (or mid-position ) and possibly its area. The last item was included because being aware of the difference between audibility and recognition is an important counseling item; parents or caregivers may not be aware of the distinction between sound detection and sound recognition, and may not know which one the CSA represents. For example, while a pair of speech sounds, such as /i/ and /u/, or /s/ and /ʃ/, may both be audible to a child, mere detection of these sounds does not necessarily ensure that the child will be able to discriminate among the detected sounds and, hence, learn to recognize each sound. CSAs were assigned one point for each specific item that was addressed, or present. Points were assigned generously; inclusion of any information related to the item, regardless of comprehensiveness or clarity, allowed full-point assignment (no fractional points were assigned). For each CSA, points were summed to create a total “counseling and interpretation” score, with a maximum value of four.

Citations, Errors, and Exclusion

For each CSA, any citation(s), related to source information, acoustic measurements, etc., were noted. Additionally, any errors in speech-sound labels, audiogram axes, etc., were documented. Exclusion criteria included poor-quality image (which would prevent accurate data extraction), duplication (i.e., the identical CSA was already included in the collection), and an inability to verify the source (e.g., an image on Pinterest with no webpage address).

Results

Collection

A total of 52 CSAs were collected. As a result of the exclusion criteria outlined previously, the number of CSAs for subsequent analysis was reduced from 52 to 36 (CSA IDs CSA01–CSA36). A primary aim of this study was to examine CSAs in the context of their use as counseling tools. Some CSAs, however, were identified as “non-counseling” CSAs and, hence, were excluded from the “counseling information” analysis. Such CSAs were found in academic journals and textbooks, making them less likely to be used as counseling tools. Of the 36 CSAs that were analyzed, 5 were deemed “non-counseling CSAs” (CSA32–CSA36) while the remaining 31 were identified as “counseling CSAs” (CSA01–CSA31; see Supplementary 1 for CSA source information).

Quantification

Speech Region

Thirty-four of the 36 (94%) CSAs demarked a speech region (indicated by color, outline, etc.). A summary of the speech region data is provided in Table 1 . Fig. 2 displays the mid-position at 1,000 Hz and area , as a percentage of the total audiogram area, for each of the 34 CSAs with a demarked speech region. The data are widespread in this figure indicating tremendous inconsistency across CSAs, inconsistencies in both the location of the speech region (mid-position), and in its size (area). Speech regions' mid-positions vary from 33 dB HL to 58 dB HL. That is, there is a 25-dB difference in level between the “weakest” and “most intense” speech regions, regions that represent presumably the same conversational speech acoustics. The areas of the speech regions vary from 14 to 32% of the total audiogram area, indicating that some speech regions are thin, or skinny, while others are thick. Again, although these speech regions' areas represent presumably the same conversational speech acoustics, their depictions on CSAs audiograms are not the same at all.

Table 1. Summary of the “speech region” data.

	Frequency range (Hz)	Number of half-octaves	Average dynamic range (dB)	Mid-position at 1,000 Hz (dB HL)	Area (% of total audiogram area)
Mean	7,484	11.4	30	45	22
Minimum	4,561	10.0	20	33	14
Maximum	13,026	13.0	41	58	32
SD	1,944	0.9	5.6	7.6	4.6

Figure 2.

Area (as a % of audiogram area) and mid-position (at 1,000 Hz, in dB HL) of CSA speech regions ( N  = 34).

Speech Sound “Letters”

Thirty of 36 (83%) CSAs included individual speech sound letters. Most often, single letters (e.g., “f,” “p”) were printed on the CSAs. Sometimes, though, letter sequences (“th,” “oo,” “ah”) were shown. Due to their special use in pediatric audiology, the “Ling” sounds, ^{6 7} /m/, /u/, /a/, /i/, /ʃ/, and /s/ were analyzed separately. It was often unclear which speech sounds (International Phonemic Alphabet [IPA]) were represented by the letters on the CSAs. Symbols that unambiguously represented a speech sound (e.g., “ee” for IPA /i/) were included in the letter-like symbol group that most closely matched that speech sound (“i”). Eight CSAs had at least one speech sound letter displayed in more than one location, or had a line that indicated a range of frequencies and levels. Speech sound letters that were represented in more than one location or had a line that indicated a range were included in an incidence analysis, but were not analyzed further. This was most common with the Ling sounds. Other, non-Ling-sound letters, however, were also found to be represented at more than one location on a single CSA, namely, “d,” “f,” “g,” and “th.”

The incidence of letters varied. The least common letter was “t,” which appeared in only five CSAs. The most common letter was “s,” which appeared in all of the 30 CSAs which had individual speech sound letters. More important are the locations, on the audiogram, of any “speech sound letters.” Fig. 3 shows the frequencies and levels of six of the most common letters representing non-Ling speech sounds: f, k, l, p, r, and v. Similarly, Fig. 4 shows the frequencies and levels of the six Ling sounds. Regardless of sound group (non-Ling or Ling), frequency and level values for the same speech sound vary widely across CSAs. For example, both the letters “f” and “p” are depicted in different CSAs at frequencies that differ by more than 1,000 Hz, while the letter “h” (not shown) is depicted at frequencies that differ by more than 3,500 Hz. Across the CSAs, the Ling sound, “sh,” is shown at frequencies that differ by more than 1,400 Hz. Levels of speech sound letters vary widely too. For example, across the CSAs, the letter “v” is shown at levels that differ by more than 20 dB (see Fig. 3 ).

Figure 3.

Level (dB HL) and frequency (Hz) values of the six most-common letters representing non-Ling speech sounds measured from CSAs.

Figure 4.

Level (dB HL) and frequency (Hz) values of the six Ling speech sounds measured from CSAs.

Environmental Sounds

Twenty-three of 36 (64%) CSAs included environmental sounds. Forty different environmental sounds were represented within this sample. Some sounds were unique to a single CSA, whereas others were common. Appearing in 20 CSAs, the three most commonly shown environmental sounds were birds, lawn mower, and airplane. Table 2 provides a summary of the location on the audiogram for these three most commonly represented environmental sounds. Fig. 5 shows the frequencies and levels on the audiogram of the 16 most common environmental sounds. Again, values are wildly inconsistent in both frequency (in Fig. 5 , see “dog”) and level (in Fig. 5 , see “vacuum”).

Table 2. Summary of frequency and level data for the three most commonly occurring environmental sounds in the analyzed CSAs.

	Birds ( n  = 20)		Lawn mower ( n  = 20)		Airplane ( n  = 20)
	Frequency (Hz)	Level (dB HL)	Frequency (Hz)	Level (dB HL)	Frequency (Hz)	Level (dB HL)
Mean	5,821	7.8	292	96	4,175	115
Minimum	4,095	−1.6	190	83	3,364	105
Maximum	11,367	36.6	504	120	6,682	126
SD	1,473	8.5	87	7.3	776	6.0

Abbreviations: CSA, common sounds audiogram; HL, hearing.

Figure 5.

Level (dB HL) and frequency (Hz) values of the 16 most common environmental sounds measured from CSAs.

Counseling and Interpretation

For the 31 counseling CSAs, information regarding counseling and interpretation are summarized in Tables 3 and 4 . The presence (or absence) of information related to these four items was documented: (1) level of speech or environmental sounds in dB SPL; any mention of dB SPL versus dB HL; (2) distance from the talker (or from the environmental sound source); mention of the relationship between distance and sound level; (3) talker's vocal effort; and (4) distinction between sound detection (or audibility) and sound recognition. None of the counseling CSAs mentioned the overall level of speech (e.g., 60 or 65 dB SPL) (see Table 3 ). Seven of the 31 CSAs (23%) mentioned distance, though a particular distance was not necessarily specified. Scoring was generous; one CSA stated “Variations in distances between listeners and speaker may affect the perceived intensity.” and was given 1 point for this counseling item. Ten of the 31 CSAs (32%) mentioned the vocal effort of the talker, while eight (26%) discussed, in some way, the difference between audibility (detection) and recognition. Fifteen of 31 (48%) counseling CSAs had a “counseling and interpretation” score of “0” and none of the counseling CSAs had a score of “4” (see Table 4 ).

Table 3. Number of occurrences and prevalence of the counseling information for the 31 counseling CSAs.

	dB SPL of sound	Distance of sound source	Vocal effort	Detection vs. recognition
Occurrences	0	7	10	8
Prevalence	0%	23%	32%	26%

Abbreviations: CSA, common sounds audiogram; SPL, sound pressure level.

Table 4. Distribution of the total “counseling and interpretation” scores; the maximum score is 4 points.

Total score	0	1	2	3	4
Number of CSAs	15	9	5	2	0

Abbreviation: CSA, common sounds audiogram.

Note: The mean total score for the 31 counseling CSAs was 0.8 (SD: 0.9).

Citation or Attribution

None of the 31 counseling CSAs mentioned original acoustic measurements. Eight of the 31 (26%) counseling CSAs included references or noted that the CSA was adapted from another source or sources. The two most commonly cited references, each with five citations, were Northern and Downs' Hearing in Children (various editions, and all except two CSAs did not include a specific figure or page) and American Academy of Audiology (AAA; unspecified sources or versions of their CSA). These two references were cited together three times, although one of these instances also included a citation of Aural Habilitation: The Foundations of Verbal Learning in Hearing-Impaired Children . ⁶ Together, they account for seven of the eight CSAs with any reference cited (i.e., seven of the eight CSAs that included any form of reference cited either Northern and Downs and/or AAA). The only other references cited were World Health Organization (unspecified source), Phonak (unspecified source), and brentbolthousephotography.com (unspecified source).

The speech regions in all five non-counseling CSAs either were based on acoustic measurements by the authors or included references to other acoustic studies.

Errors

Many errors were found. Examples of errors included inaccurate frequency scales, speech sound letters located outside of the speech region, inclusion of a letter that does not represent any speech sound, and likely confusions among printed letters. Five of 36 (14%) CSAs had errors in the frequency scale of the audiogram itself. This was typically in the form of inaccurate non-octave increment values. Of the 28 CSAs that included both a speech region and speech sound letters, 19 (68%) had speech sound letters that were not within the outlined speech region. This is considered an error, since the speech region is assumed to encompass the frequency and level ranges of long-term speech. One CSA included the letter “c” which is not uniquely associated with any speech sound; it is pronounced as /s/ or /k/ in English. Finally, several instances of errors, very likely due to visual confusion of sound letters, were observed. Though these errors cannot be confirmed, the similarity in location of the errant sound letter to a visually similar letter is suspicious. For example, in one CSA the letter “l” was located at a frequency and level consistent with the position of /i/ in most other CSAs. The creator of this CSA may have, due to a poor-quality image or poorly chosen fonts, confused “l” with “i.”

Discussion

The CSA is a medical counseling tool; it is used to describe the severity of a condition, and its use can affect patient outcomes. Such a counseling tool should be consistent or standardized; include important information for interpretation; be scientifically justified; and have no errors. In this study, 36 currently available CSAs were examined in the context of these criteria.

Quantitative analyses of the 36 CSAs revealed substantial inconsistencies in the speech region's vertical location ( mid-point ) and size ( area ), location of individual speech sound letters, and location of environmental sounds. In this sample, we measured mid-points of speech regions from 33 to 58 dB HL; this is an enormous 25-dB difference in vertical location of speech regions across these CSAs. The speech region's vertical location matters, as this reflects an assumed overall speech level (or equivalently, an assumed talker's vocal effort with an associated distance between the talker and listener). Of clinical importance when counseling, the use of CSAs with different mid-points can lead to very different parental interpretations of the impact of their child's hearing loss on the audibility of speech. The use of different CSAs could also lead to significantly different recommendations by audiologists regarding intervention and hearing devices. Finally, when aided thresholds are plotted on CSAs with different mid-points , different understandings of, or verification of, aided audibility would ensue.

Our analysis of the presence of counseling information accompanying the CSAs revealed that roughly half (15 of 31) of the counseling CSAs offered no information for interpretation and counseling (recall, only 31 of the 36 CSAs were analyzed for counseling information; 5 CSAs were deemed “non-counseling”). Consequently, the onus is on pediatric audiologists to provide clear and accurate interpretations of all the images and information on the CSA. Furthermore, when the CSA is absent counseling information (as was the case for roughly half of the CSAs examined here), parents have nothing to reference at times when the audiologist is not present. Yet, this type of counseling information is important for accurate interpretations of a child's hearing loss. It is paramount that parents and caregivers realize that audibility of speech is critical to spoken language development, and that they understand factors which affect the audibility of speech.

The lack of information about an assumed overall level of speech associated with the vertical location ( mid-point ) of the speech region on the CSA is a serious omission. Parents and caregivers should learn about the major factors that affect the overall speech level that reaches their child's ears (or device microphones), and hence that would affect the audibility of speech. The most noteworthy factor is distance between the talker and listener. When a CSA does not state that the speech region's vertical location is predicated on a talker being located 3 feet away (for example), it may not be clear that the patient would have poorer speech audibility when listening at distances greater than 3 feet. Parents should be counseled explicitly that as distance from a talker increases, speech levels (at the listener) decrease (by as much as 6 dB for each doubling of distance); that is, the speech region and the speech sound letters on the CSA shift upward. Parents could also be reminded that distances from a teacher to the rear of a classroom are generally much greater than 3 feet, reinforcing the importance for a child to wear his/her hearing devices and/or use a FM/DM system. Another factor that affects the overall speech level that reaches the child's ears is the talker's vocal effort. When shouting, compared to using a casual vocal effort, talkers can increase their average speech levels by about 30 dB, depending on the individual and type of talker (child, adult female, or adult male). ⁸ When a CSA does not specify that speech audibility is based on conversational vocal effort, then parents may not understand that audibility would increase when vocal effort is changed from conversational (casual) to shouting, or that audibility would decrease when a talker speaks softly. Lastly, when information is absent regarding the distinction between speech sound detection (or audibility) and speech sound recognition, then a parent may erroneously assume that detection is synonymous with recognition. Parents' understanding of this distinction is crucial. CSAs should include explicit counseling instructions so parents can learn: detection of a speech sound is a necessary but not sufficient condition for a child to learn and recognize that speech sound. This fact is aptly described by Pascoe, ⁹ “…although it is true that mere detection of a sound does not ensure its recognition, it is even more true that without detection the probabilities of correct identification are greatly diminished.”

The CSAs analyzed in this study provided scant scientific justification for their depictions of speech regions and/or speech sound letters. Most CSAs (23 of the 31 counseling CSAs) did not state, at all, any source of information (scientific article, textbook, etc.) to justify the mid-point and/or area of the drawn speech region or speech sound letters. (All five of the non-counseling CSAs either made their own acoustic measurements or included references to other studies, citing scientific journals or textbooks.) Of the eight counseling CSAs that cited sources, seven cited sources that upon further research have little or no clear scientific basis. Specifically, these seven CSAs cited Northern and Downs' Hearing in Children (various editions; most CSAs did not include a specific figure or page reference) or the AAA (unspecified sources or versions of their CSA).

A brief “history of the CSA,” as found in the textbooks written by Northern and Downs, follows, as this may partially explain the variations found among currently available CSAs. In the 1st edition of Hearing in Children (1974), ¹⁰ there is no CSA. In the 2nd edition of their textbook (1978), ¹¹ Northern and Downs introduce a figure (Figure 1.11: “Frequency spectrum of familiar sounds”; p. 12) that shows speech sound letters and environmental sounds on an audiogram. There is not, however, any speech region drawn on the audiogram. Some people consider this figure to be the very first “Common sounds-” or “Familiar sounds audiogram.” In this 2nd edition, there are no references or mention of any data or acoustic measurements that would justify the locations of the speech sound letters and/or images in the figure. The narrative, in which Figure 1.11 is referenced, is about why a 15-dB loss can result in language delays; the authors state, “The reasons lie in the nature of speech sounds, with the major amount of speech energy residing in the voiced vowels and consonants. The unvoiced consonants ( s, p, t, k, th, f, sh ) contain so little speech energy that they often fall below even normal hearing thresholds in average rapid conversation (Figure 1.11).” The 3rd (1984) ¹² and 4th (1991) ¹³ editions of Hearing in Children contain this same CSA, published as Figure 1.5 (p. 7) and Figure 1.9 (p. 17), respectively. In the 5th (2002) ¹⁴ and 6th (2014) ¹⁵ editions of Hearing in Children , a new version of the CSA appears. ^a Compared to the CSAs published previously, images were added, deleted, and modernized; fonts were changed; and a speech region was added (Figure 1-9, p. 18 and Figure 1-4, p. 13). In essence, then, there are two versions of a CSA published by Northern and Downs, the “old N&D” CSA (2nd, 3rd, and 4th editions) and the “new N&D” CSA (5th and 6th editions) which includes a speech region. Interestingly, the “new N&D” has the same speech sound letters displayed in the “old N&D” CSA, but many sound letters are in different locations; some are shifted up (e.g., the letters “n,” “ng,” “e,” and “u,” are depicted at levels about 10 dB lower than in previous figures) and some are shifted right (e.g., the letters “ch” and “sh” are shifted from about 1,500 Hz to about 2,000–3,000 Hz). Also in the “new N&D” CSA, many of the environmental sounds are represented at frequency and level positions that differ substantially from their positions in the “old N&D” CSA. For example, “dog” is shown at 500 Hz and 80 dB HL in the “old N&D” CSA, but at 250 Hz and 70 dB HL in the “new N&D” CSA, while “lawnmower” shifts to a higher frequency but a lower level (earlier version: [250 Hz, 100 dB HL]; new version: [500 Hz, 90 dB HL]).

In the 3rd edition of their book (1984), the authors, Northern and Downs, introduced a CSA-like figure (see Figure 1.6: “Average range of speech energy in dB HL is shown in audiogram (A).”; p. 9). ^b Panel A in this figure shows an audiogram with a shaded region; solid lines along the top and bottom edges represent the “softest speech energy” and “loudest speech energy,” while a dashed line in-between represents “average speech energy” in dB HL. In the caption, Northern and Downs state, “Audiograms adapted from Skinner (1978) ¹⁶ and Dudich et al (1975).” ¹⁷ The speech sound letters displayed in the “old N&D” CSA are also displayed in this CSA-like figure at exactly the same (Hz, dB HL) locations, but no environmental sounds are displayed. This CSA-like figure, a shaded speech region with speech sound letters but no environmental sounds, is maintained “as is” in subsequent editions (4th–6th) of Hearing in Children with minor changes in the figure captions and font style. Since some speech sound letters' positions in the “new N&D” CSA differ from those in the “old N&D” CSA, but the speech sound letters remain the same throughout all versions of the CSA-like figures, there is a consequent discrepancy in some speech sound letters' positions of the “new N&D” CSA and the CSA-like figures in the later editions of Hearing in Children (5th and 6th). For example, in the 5th edition, the positions [Hz, dB HL] of the sound letters in the “new N&D” CSA (Figure 1-9, p. 18) are inconsistent with their positions in the CSA-like figure (Figure 1-8, p. 16). The letters “m,” “d,” and “b” are located at about 28 dB HL in the “new N&D” CSA, yet are located at about 40 dB HL in the CSA-like figure; these three letters are located at the same frequencies in both figures. And, the letter, “f,” is located at about (3,500 Hz, 18 dB HL) in the “new N&D” CSA, but is positioned at about (5,000 Hz, 20 dB HL) in the CSA-like figure. Interestingly, the shaded regions (“speech banana”) in the “new N&D” CSA and in the CSA-like figures do not cover the same areas in the audiogram. In particular, the top edges (“loudest” speech sounds) are near 52 dB HL in the “new N&D” CSAs, but are near 62 dB HL in the CSA-like figures.

^a The CSAs in the 5th and 6th editions, aka “new N&D” CSA, appear identical with one exception—a “telephone” image is present in the 5th edition but not in the 6th edition.

^b This CSA-like figure is one of the five “non-counseling” CSAs included in our analyses. See later parts of the “Discussion” section.

For completeness, Northern and Downs also published, in some editions, figures that display the dynamic range (DR) of conversational speech on an audiogram. A “DR” figure appeared first in the 4th edition of Hearing in Children (1991), and also appears, with slight modifications, in the 5th and 6th editions. For these DR figures, the authors cite either Olsen ¹⁸ in the 4th edition or Chial ¹⁹ in the 5th and 6th editions.

The other multiple-cited source is the AAA. The current version of the AAA CSA has a copyright date of 2014. According to the AAA website, where the CSA may be purchased, it was “Developed and approved by Academy audiologists […].” ² There is no additional information about how this CSA (CSA01) was developed, such as original acoustic measurements or references to those of others. The AAA also has an older, slightly different version of a CSA with no copyright date; this version was also included and analyzed in this sample of 31 counseling CSAs (CSA02). The older version says it was adapted from the CSA in the 4th edition of Hearing in Children . ¹³ Notably, these two AAA versions of the CSA differ substantially in overall level; compared to the older version, there is a 15-dB downward shift in the speech region (downward, on the audiogram, or “louder”) and a roughly 10-dB downward shift in the positions of the speech sound letters. The basis for this change in level is unknown (i.e., CSA01 → CSA02). When the AAA is cited by others for their own CSAs, it is also unknown which of these two AAA CSA versions (undated or 2014) is referenced. It is interesting to note that while the AAA and Hearing in Children are the most commonly cited references for other CSAs, at one point, one references the other.

The five non-counseling CSAs provide references and/or report their own acoustic measurements (Chial, ¹⁹ Fant, ²⁰ Ling and Ling, ⁶ Northern and Downs, ^{12 13 14 15} Olsen ¹⁸ ). Visually, these “non-counseling” CSAs are speech regions drawn on an audiogram scale (i.e., dB HL). None of these five display environmental sounds; two display only a speech region (Chial, ¹⁹ Olsen ¹⁸ ); one assigns sections of the speech region to specific acoustic characteristics (e.g., fundamental frequency, first formants of vowels; Fant ²⁰ ); one displays five speech sounds at either multiple points or ranges (Ling and Ling ⁶ ); and one displays speech sounds (Northern and Downs ^{12 13 14 15} ). Fant ²⁰ and Ling and Ling ⁶ report their own acoustic measurements, while the others cite data from others. Fant's region is based on speech measurements, from a 1-m distance, of adult male voices speaking Swedish. Ling and Ling ⁶ say their data show “the relative intensity levels (HL) of the main components of the five sounds [u, a, i, ʃ, s] spoken at two yards” from an adult male. Both Chial ¹⁹ and Olsen ¹⁸ cite Pascoe (1980), ⁹ which in turn cites Pascoe (1975), ²¹ Dunn and White, ²² and Pearsons et al ⁸ ; the speech regions represent speech at an overall level of 66 dB SPL (or, equivalently, the level of a “raised” voice according to Pearsons et al ⁸ ), and the dynamic range reflects the 10th- and 90th-percentiles of the 1/3-octave-band level distributions of speech reported by Dunn and White. ²² Northern and Downs cite Skinner, ¹⁶ which in turn cites the same articles of Pascoe (1975) ²¹ and Dunn and White, ²² and also Pascoe (1978). ²³ Similar to Pascoe, ⁹ Skinner ¹⁶ displays data corresponding to an overall speech level of 65 dB SPL. Thus, in general, overall level and dynamic range of speech regions are documented fairly well, while positions of individual speech sounds are documented less well.

Finally, a substantial number of errors were observed within this sample of 36 CSAs. Errors were observed, however, only in the counseling CSAs; no errors were observed in the non-counseling CSAs. While this list is not exhaustive, errors included incorrect labeling along the octave-frequency scale, speech sound letters located outside of the speech region, speech regions that extended beyond the edges/boundaries of the audiogram, and inclusion of the letter “c” (which does not correspond to any speech sound in English). Errors, such as these, may impede parents' understanding of child's hearing loss. It is possible that some errors (e.g., the letter “l” occurring in a handful of CSAs where most CSAs had depicted the letter “i”) reflect carelessness or sloppiness, or that CSAs were created by persons who understand graphic-design but not the scientific nature of the items displayed.

Recommendations

The CSA is an important part of pediatric audiology counseling. Many versions of the CSA are available from professional societies, early intervention providers, device manufacturers, and elsewhere. As a group, however, these CSAs are inconsistent, not scientifically justified, and do not include important information for interpretation. A comprehensive CSA should be developed based on the recommendations outlined later.

For these recommendations, adult female speech should be the basis of all the speech acoustic properties. Female speech is recommended because its acoustic properties represent a middle-ground between the acoustic properties of children's and adult male's speech. Young children with hearing loss must hear their own speech, as well as that of their similarly aged (young) peers. Additionally, pediatric patients need to hear well the speech of pre-school teachers, early interventionists, and caregivers—all of who are predominantly adult females. The developers of the International Speech Test Signal (ISTS) hearing-aid test signal chose female speech for some of the same reasons. ²⁴

Speech region (or speech banana) : The counseling tool should include a speech region that represents normal, conversational vocal effort at a specified distance of 1 meter, which corresponds to an average level of approximately 57 dB SPL. ^{8 25} The speech region should also reflect the dynamic range of adult female speech, for American English; these data are fairly consistent across several studies. ^{24 26 27} The chosen dynamic range, however, should be specified explicitly as 10th versus 90th percentiles, or 99th versus 30th, etc. Using 99th versus 30th percentiles, the dynamic range of adult female spoken English is approximately 30 dB for the 1/3-octave bands from 125 to 8,000 Hz. ²⁴

Speech sound “letters” : For simplicity, we recommend the display of very few sound letters. Rather than displaying every sound in English (∼42) and cluttering the speech region, we recommend there be a focus on a handful sounds, such as the Ling sounds. This would emphasize to the patients' families that sound patterns, rather than single locations, may be needed for sound recognition. Two [Hz, dB HL] locations could be used to represent the first two formant frequencies of each of the three vowels. Spectral ranges could be shown for /ʃ/ and /s/. Again, speech acoustics, based on adult female speech, of the six sounds (the Ling sounds: /m/, /u/, /a/, /i/, /ʃ/, and /s/) is recommended. Adult female acoustic data are available for vowels from Hillenbrand et al, ²⁸ for fricatives from Jongman et al ²⁹ and Pittman et al, ⁵ and for all six Ling sounds from Scollie et al. ³⁰

Environmental sounds : This counseling tool should not include environmental sounds since the primary focus of this tool is speech audibility for spoken language acquisition.

Counseling and interpretation : This counseling tool should have information in text stating explicitly: (1) the speech area and speech sounds represent speech energy for conversational vocal effort for a distance of 1 meter (3 feet) between the talker and listener; when distances are greater than 1 meter, the speech area and speech sounds would shift up on the audiogram (i.e., all speech sounds would be less intense and hence “softer,” or possibly inaudible, to the listener); (2) speech detection alone is not sufficient for acquiring spoken language; sounds must be heard well-enough to recognize individual speech sound patterns across spectral regions.

Attribution/citation/justification : This CSA should be scientifically based and should provide references to acoustic measurements of speech, as well as specifics such as the overall speech level and that acoustic properties are shown for adult female speech. Any assumptions should be stated explicitly.

Conclusions

The CSA is a commonly used counseling tool, and its interpretation may have a large effect on patient outcomes. There is, however, an abundance of currently available versions that are inconsistent, are not scientifically justified, and do not include important information for parental interpretation. These findings justify the need to develop an accurate and consistent CSA-like counseling tool.