Determining Threshold Level for Speech

General Considerations for Clinical Determination of the Speech Threshold

Instrumentation and Calibration. Speech threshold audiometry shall be accomplished with a speech audiometer or diagnostic audiometer capable of transducing speech as defined and calibrated according to the American National Standard Specifications for Audiometers (ANSI S3.6-1969, or subsequent superseding standards). The reader is referred to Asha ( 1987) for a review of recommended procedures for calibration of speech signals delivered via earphones.

Test Environment. The test environment shall meet the criteria for background noise in audiometric rooms as specified by the American National Standard Criteria for Permissible Ambient Noise During Audiometric Testing (ANSI 53.1-1977 or subsequent superseding standards).

Test Material. Spondaic words are the standard test materials for determination of the speech recognition threshold. An alphabetical list of 36 spondaic words, appropriate for adults, is presented in Appendix A. Appendix B presents 15 homogeneous words from Auditec recordings of CID W-1 ( Young, Dudley, & Gunter, 1982). These words are homogeneous in terms of audibility in those specific recordings. Recommended procedures for determining homogeneity of other tape-recorded lists are described in Wilson and Margolis ( 1983) and Young et al. ( 1982). Numerous variables associated with monitored live voice testing renders exact specification of homogeneity impossible with this procedure.

Appendix C presents a list of spondaic words appropriate for most children aged 5 to 12 years. These words can be represented pictorially, and if so, would be appropriate for children under age 6, depending on their capabilities.

There may be circumstances or individuals that require a modification of the standard word list, in which case alternative test materials would be used. The selection of such materials should be made with consideration of the person under test. Some factors to consider are age, language facility, and the physical condition of the client. Different lists of spondaic words may be required for different clinical populations and to accommodate certain physical or developmental limitations. It should be recognized, however, that the use of speech stimuli with less homogeneity than spondaic words may compromise the reliability of this measure.

When circumstances or individual capabilities prevent determination of a speech recognition threshold, the speech detection threshold may be determined instead. The speech detection threshold (in dB) should be consistent with the best pure tone threshold (in dB) between 250 and 4000 Hz ( Olsen & Matkin, 1979) and should also be obtained at levels 8–9 dB weaker than the speech recognition threshold ( Chaiklin, 1959). The type of speech material is not as critical for this measure because it reflects detection and not recognition. Some common materials are speech babble, running speech, or familiar words. Nevertheless, specification of the type of material helps to ensure test-retest reliability and may be useful information for future hearing evaluations.

Response Mode. The usual response mode for obtaining the speech recognition threshold is repetition of the stimulus item. For many patients it is not possible to obtain verbal responses, necessitating the use of alterative response modes. Many alternatives are acceptable but must convey recognition of test items from a closed set of choices. For example, response modes can take the form of picture pointing, signing, or visual scanning. In addition, if a picture-pointing task is used for obtaining the speech recognition threshold in children, then the clinician should be cautious in choosing the number of response items. Too few items increases the probability of chance performance and too many items may be distracting and increase response time (e.g., between 8 and 12 words usually is appropriate).

For assessing the speech detection threshold, a number of response modes can be used to convey signal detection. Usually, these response modes are non-verbal. Whenever a response mode other than repetition of a spondaic word is used, it should be specified in recording and reporting the results.

Recorded vs. Live Voice Presentation of the Test Material. Either a recorded or a monitored live voice technique can be used to obtain the speech threshold. Recorded presentation of the test material is the preferred procedure. The use of recorded material standardizes the composition and presentation of the test list. It allows for better control of the intensity of the test items and ensures that the speech pattern of the recorded talker will be consistent to each client. In other words, recorded presentation is preferred because the stimuli are consistent to each patient tested with a given set of recorded test materials. It must be remembered that phonographic recordings and tape recordings become worn after extensive use, introducing distortion and noise into the test system. The audiologist must be alert to such problems and periodically replace phonograph records or tape recordings. Further, cartridge needles must be replaced periodically, and tape playback units must be cleaned and demagnetized regularly. Many of these problems may be obviated by use of digitized speech recordings.

The use of recorded test materials may limit flexibility of the test procedure in terms of selection of test words and rate of presentation. Certain clinical situations arise which favor use of monitored live voice presentation. The disadvantages of live voice presentation are the problems and difficulty in monitoring the test words to a consistent hearing level. In addition, it is impossible to present each spondaic word in the same manner to every client, even if each syllable of each spondaic word peaks at 0 dB on the VU-meter. When monitored live voice is used, it should be noted with the test results.

Recording of Results. The speech recognition threshold or speech detection threshold shall be recorded in dB HL. The results should be recorded for each ear on the same form that contains the client's results for pure tone audiometry. Additional space should be available to report other pertinent information that describes the test situation, such as alternative materials or response modes.

Masking of Nontest Ear. When the obtained speech recognition threshold (or speech defection threshold) in one ear exceeds the apparent speech recognition threshold (or speech detection threshold) or a pure tone bone conduction threshold at either 500, 1000, 2000 or 4000 Hz in the contralateral ear by 40 dB or more, masking should be applied to the nontest ear. Of course, the criteria for the use of masking should consider the signal's spectrum and transducer. The appropriate masker for a speech stimulus must have a wideband spectrum (e.g., white noise or speech-spectrum noise). The level of effective masking used should be sufficient to eliminate reception by the nontest ear without causing overmasking and should be recorded on the same form as that used to record audiometric results.

Return to Top

Determination of Speech Recognition Threshold

The basic procedure consists of instructions, familiarization, a single series of descending threshold determination, and calculation of threshold hearing level.

Instructions. The specific wording of the instructions must be phrased in language appropriate to the client and should

1. Orient the client to the nature of the task,

2. Specify the client's mode of response.

3. Indicate that the test material is speech and specify that the client should respond with only words from the test list, and

4. Stress the need for the client to respond at faint listening levels and encourage the client to guess.

Familiarization. The first step in the testing procedure is to familiarize the client with the exact spondaic words in the word list. This necessary step ensures that

1. The test vocabulary is familiar to the client,

2. The client can auditorily recognize each test word, and

3. The client's responses can be accurately interpreted by the clinician.

The audiologist may read the test list to the client in a face-to-face situation or present the test list through a speech audiometer. In either case, visual cues should be eliminated during familiarization with the test words. The client repeats or in some other way demonstrates recognition of each word on the list.

The audiologist should emphasize that the client is to respond only with words from the test list. If the client has any difficulty understanding or responding to any spondaic word, then that word should be eliminated from the test list. Any spondaic word that the audiologist has any difficulty understanding should be eliminated from the test list.

Familiarization with the test list is essential to control for the effects of prior knowledge of test vocabulary on the speech recognition threshold ( Tillman & Jerger, 1959). It should not be eliminated from the procedure.

Descending Threshold Determination. The method described in this document closely follows the procedures originally used in PAL Auditory Test Number 9 ( Hudgins et al., 1947) and CID W-2 recordings ( Hirsh et al., 1952). The basic procedure, which was described by Wilson et al. ( 1973), involves a preliminary phase and a test phase. During the test phase, different spondaic word pairs are presented at 2 dB decrements until at least five of the last six test words are missed.

Threshold for the hearing level of speech is calculated to estimate the 50% point on the psychometric function. When variations in the standard technique are used, the nature of the variation should be recorded with the results. The current procedure is a 2 dB descending approach but can be modified to a 5 dB descending approach. (Footnotes detail modifications for using a 5 dB rather than a 2 dB descending approach.)

Return to Top

Descending Technique

1. Preliminary Phase to Obtain Starting Level

   1. Set the hearing level to 30–40 dB above the estimated speech recognition threshold and present one spondaic word to the client. ^[1] If the response is correct, then descend in 10 dB decrements, presenting one spondaic word at each level until the client responds incorrectly. If the client does not respond correctly to the first spondaic word at the first level, then increase the level in 20 dB steps until a correct response is obtained ( Martin & Stauffer. 1975). Then initiate the 10 dB decrements ( Martin & Stauffer, 1975). In cases where a speech recognition threshold cannot be estimated prior to testing, the procedure for determining a starting level reported by Martin and Stauffer ( 1975) may be used. ^[2]

   2. When one word is missed, present a second spondaic word at the same level. Continue this process of descending in 10 dB steps until a level is reached at which two consecutive words are missed at the same hearing level.

   3. Increase the level by 10 dB (above the level at which two spondaic words were missed). This defines the starting level.

2. Test Phase

   1. Present two spondaic words at the starting level and at each successive 2 dB decrement. ^[3]

   2. Continue this process if five out of the first six words are repeated correctly. If this criterion is not met, then increase the starting level by 4–10 dB.

   3. The descending series is terminated when the client responds incorrectly to five of the last six words presented. ^[4]

Calculation of Threshold. Threshold is calculated by subtracting the total number of correct responses from the starting level and adding a correct factor of 1. ^[5] This calculation is based on a statistical precedent ( Spearman, 1908) for estimating threshold at the 50% point of the psychometric function. The calculation procedure is the Spearman-Karber method ( Finney, 1952) and is given by the formula: where T = threshold, i = initial test intensity, d = dB decrement between steps, r = correct responses, and n = number of words per dB step. While the equation may appear cumbersome, it may be reduced to the simple form:

Figure 1 presents a worksheet for tallying responses and calculating the speech recognition threshold for both 2 dB and 5 dB step sizes, using this formula.

Speech recognition thresholds determined with this procedure and calculation method provide standardization and therefore reduce variability in threshold estimates. An ascending procedure, which may be warranted in certain instances, is described in Footnote ^[6] .

Determination of the speech detection threshold involves a detection task that is similar to the one used in pure tone threshold audiometry. Stimulus familiarization is unnecessary. Test technique should follow the one used for pure tone threshold estimation, including appropriate modification for the client's capabilities (see ASHA Guidelines for Manual Pure Tone Threshold Audiometry, 1978).