How do you construct a SPA test?

 

1.  Why use minimal pairs?

Minimal pairs are pairs of words which differ in one sound only (e.g. date-gate). By using minimal pairs, it is possible to investigate the perception of a specific contrast between two sounds which are acoustically similar (e.g. /d/ versus /g/).

The minimal pairs included in the test battery are chosen on the basis of knowledge of speech perception development as some speech contrasts are acquired earlier than others.  By testing listeners on word pairs graded in terms of speech pattern complexity, it is possible to assess the stage of speech perceptual development achieved, which is particularly useful when testing children.
 

2. Why use synthetic speech?

By using high-quality computer-generated versions of words, constructed using a speech synthesizer, it is possible to control the speech patterns presented to the listener, and to keep constant all speech patterns except for those being evaluated as acoustic cues. By presenting sounds in which certain acoustic cues have been removed or "neutralised", it is possible to ascertain how important they are for the listener.
 

3. How to make up a 'date-gate' test?

First, make recordings of a speaker saying the words and then choose one clear example of each.

Click on the word to listen to the words date and gate spoken by a female speaker.
 

Make a fully synthetic copy of these words using the Klatt speech synthesiser.

Click on the word to listen to the words date and gate produced by the Klatt synthesiser.
 
Decide which speech patterns you need to vary to change the perception of the initial consonant from /d/ to /g/. Research has shown that the initial burst frequency and second formant transitions into the vowel are important acoustic cues to the difference between these two consonants.

Produce a 'stimulus continuum' by changing these patterns in equal steps between the values appropriate for /d/ and /g/. You can have as many steps as you want, but we tend to use a six-step continuum.

Click on the step number to listen to individual steps along the /d/ to /g/ continuum. Both the burst frequency and formant transition cues are being varied:

    step 1        step 2        step 3        step 4        step 5        step 6
 

4. How to test which acoustic cue the listener is using to differentiate /d/ from /g/?

Produce a stimulus continuum in which all acoustic cues are varying (in this case, both the burst and formant transition, as shown above).

Produce a continuum in which only one cue is varying (e.g. burst). You can do this by keeping the other cue (e.g. formant transition) constant at an intermediate value. If a listener does the test well when both burst and formant transition are varying but cannot hear the constrast when only the formant transition is varying, this will suggest that the burst is the main acoustic cue for that listener in this test. 
 

Click on the step number to listen to individual steps along the /d/ to /g/ continuum: Only the burst is being varied in this set:

                                    step 1        step 2        step 3        step 4        step 5        step 6

Click on the step number to listen to individual steps along the /d/ to /g/ continuum: Only the formant transition is being varied in this set:

                                    step 1        step 2        step 3        step 4        step 5        step 6

 
This page created with Netscape Navigator Gold