Is there a Way to know if an audio element used alternate text?

In a simple audio element with alternate TTS text, can one determine 
whether the alternate text was spoken?  It seems the spec is silent on 
this, and I have found the question of whether the alternate content is 
evaluated to be interpreter-dependent.  For example:

<block>
<var name="ttsfail" expr="false"/>
<prompt>
<audio expr = "vox_file('greeting.wav')">
<value expr="ttsfail = true, ''"/>
Thank you for calling ... Your call is very important to us and may be 
monitored or recorded for training and quality control.
</audio>
</prompt>
<!--  execute following prompt only if the greeting.wav fetch failed and
       text inside the above audio element was spoken.  greeting.wav
       has a dual language recording, so this provides dual language TTS
       backup.  However, interpreter appears to evaluate inside the audio
       element even if the content is not used -->
<prompt cond="ttsfail" xml:lang="es-MX">
Gracias por llamar ... Su llamada es muy importante para nosotros y 
podrá ser monitoreada o grabada para capacitación y control de calidad.
</prompt>
</block>

The question is under what circumstances is the variable ttsfail set to 
true?

Dan