Practical Applications: Some Examples

Spectrographic analysis is a very useful tool in applied disciplines. The increasingly affordable software on the market has made its use viable in a clinical context and in foreign language teaching. The utility of spectrographic analysis comes from the fact that it pro vides quantitative and objective data on a wide range of speech parameters (e.g. nasalization, vowel quality, segmental duration, place and manner of articulation, voice onset time), and greatly enhances the scope of auditory-based perceptual judgments of speech. It is particularly helpful in monitoring changes during remediation (clinical and/or classroom context).

Real-time spectrographic displays have been used in speech training with individuals who have severe and profound hearing impairments. Although some have expressed concerns about the usefulness of spectrograms in speech training, citing their complex and abstract nature, several investigators have pointed out their successful use with hearing-impaired adults (Maki 1980, 1983; Maki et al. 1981) and children (Ertmer and Stark 1995; Ertmer et al. 1996) with respect to contrasts such as voiced–voiceless, durational differences, tongue position for vowels, and differences in manner of consonant production. Recently, Ertmer (2004) examined how well children with normal hearing and children with impaired hearing can recognize spectrographic cues for vowels and consonants, and the ages at which these visual cues are distinguished. Subjects’ training activities involved instruction, highlighting of target spectrographic cues, matching of spectrograms by the children, and feedback on correctness. Results showed that a variety of spectrographic cues were recognized with greater-than-chance accuracy at each age level and across both hearing statuses. On average, formant cues were recognized with greater accuracy than consonant manner features, making vowels, diphthongs, and vowel-like approximants easier to recognize. This saliency, Ertmer suggests, may be the result of a combination of greater duration, darker energy traces, and distinctive visual patterns.

For another example, we can consider problems related to VOT, which may be of concern to a variety of populations. Individuals who are suffering from aphasia (language deficit due to damage to certain regions of the brain, in this case Broca’s area) seem to be unable to control the timing between the release of a stop and the onset of voicing. Individuals with apraxia (neurological disorder of motor programming for speech) tend to have problems in areas such as timing and coordination that might lead to troubles in VOT production. In several studies, dysarthria (a group of speech disorders resulting from a disease or damage to neural mechanisms that regulate speech movements) patients have been reported to show increased variability in their VOT productions. VOT deficiencies have also been reported in relation to hearing impairment.

The significance of VOT deficiencies is not restricted to the clinical context and is very important in foreign language teaching and accent reduction con texts. The difference between English /p, t, k/ and /b, d, g/ is not voiced/voiceless in initial position, as /b, d, g/ may also have no voicing before the release of the stop closure. Thus, the aspiration of /p, t, k/ is very important for a segment to be perceived as /p/ rather than /b/ (and there is a similar situation for /t/ vs. /d/ and for /k/ vs. /g/). Learners of English coming from Romance languages with unaspirated /p, t, k/ (e.g. Spanish, Portuguese) face a big challenge in learning the English patterns, because their productions may be (and indeed, in many cases, are) perceived as /b, d, g/ by native speakers of English. While aspiration of /p, t, k/ is perceived by the hearer in a binary fashion (i.e. all-or-none occurrence), studies on first and second language acquisition production data show that delay of the onset of voicing is typically gradual in the case of voice lag. Thus, the value of acoustic data becomes ever more important during the remediation process. In considering the productions of the client (learner/patient), their progress needs to be constantly evaluated via instrumentation. Monitoring the changes and making the learners/patients aware of this progress, however incomplete, would encourage them, and thus accelerate the remediation process.

It is frequently pointed out that transcribing vowels accurately is more difficult than doing the same for consonants. This task is made harder when we deal with different vowel sounds that are not part of our native inventory, or different vowel sounds we might encounter in disordered speech. The key here is to identify the differences between the system to be remedied and the target system. Although the cardinal vowel system may be helpful, this requires a very strong phonetic training if one is to rely on it. Acoustic analysis provides a viable alternative here. As mentioned earlier, vowels can be accurately described by the frequencies of their first three (in most cases, two) formants. Thus, with the help of spectrographic analysis, the practitioner will be in a position to identify the nature and the extent of the mismatches between the target system and the system of the patient/student, and plan the remediation accordingly.

Apart from its utility in identifying different vowel qualities, acoustic analysis is also a powerful tool for vowel durations, an issue that may be crucial in both a clinical context and foreign language teaching. For example, as we saw earlier, voicing is frequently absent during closure of final /b, d, g/. Because vowels preceding voiced obstruents are lengthened, the final consonant will be perceived as voiced even though no specific evidence of voicing is present. When the learner/patient fails to implement this expected lengthening, the practitioner can identify this unambiguously via spectrographic analysis. Also, during remediation she or he can carefully monitor changes in vowel durations via spectrographic data. This can help measure progress by accurately describing productions that may not be evident perceptually.

Failure to implement voicing contrasts among obstruents is not uncommon in some aphasic patients. For example, a 63-year-old female Broca’s patient (Code and Ball 1982) did not have the voicing contrast involving fricatives (e.g. proofing vs. proving, pence vs. pens). Analyzing her data spectrographically, however, the investigators were able to find out that her productions for such pairs were not homophonous, and that although she did not have the voiced/voiceless distinctions in target fricatives, she did maintain the contrasts via the length of the vowel before target fricatives; that is, vowels were (just as in normal speech) longer before voiced (lenis) targets than before voiceless (fortis) targets.

The same subject’s data were also instructive with respect to the duration of target fricatives. As pointed out earlier, the duration of frication of voiceless targets is longer than that of voiced ones. The subject in the above-mentioned study also made a difference (somewhat smaller than in normal speech) in the duration between voiceless and voiced targets. Thus, out of three parameters that are available to contrast /f/ with /v/, /s/ with /z/, and so on (voicing of the target, vowel length before the target, duration of frication), the patient was able to utilize the last two. Without the use of spectrographic data, the investigators would not have been able to find out that the patient was making the phonemic distinctions and that the disability was phonetic.

All of the above point to the conclusion that spectrographic analysis allows a quantification of mismatches between the normative data and the client’s productions. In addition to the fact that such quantification is indispensable for the diagnosis of trouble spots, it greatly enhances, by monitoring changes, the ability of the professional who deals with the remediation process.

Finally, we can also mention that spectrographic data can help professionals dealing with voice disorders, training of the singing voice, speaker identification, identification of the correlates of speech in stressful conditions, and intoxicated speech.