Mismatch of envelope and temporal fine structure information destroys the precedence effect

Bernhard U. Seeber

MRC Institute of Hearing Research, Nottingham, UK

The precedence effect (PE) describes our ability to suppress detrimental effects of 
reflections on localization of sound sources in rooms. For short delays between a direct 
sound (lead) and its reflection (lag) only one sound is heard and localized at the location 
of the lead. 

The information used by the auditory system to obtain the PE is still under debate. Models 
based on cross-correlation analysis rely on binaural cues extracted from the temporal fine 
structure (TFS) at low frequencies while envelope information is used to start a lateral 
inhibition process (e.g. Lindemann, 1986). Another modeling approach bases the PE entirely 
on envelope or repeated onset information (Zurek, 1987). The fact that some of our patients 
with bilateral cochlear implants (CIs) showed the PE for ongoing sounds suggests that 
envelope information can be sufficient (c.f. CIAP 2007); however, since most of our patients 
localized one sound in-between lead and lag, rather than at the lead, envelope information 
seems  not to be sufficient in most cases. It can be expected that a perfect representation 
of TFS information will not be possible with CIs in the near future, making the study of the
mechanisms and the limits of the PE - and with it the ability to function in multi-source 
environments - a high priority. 

The contribution of TFS and envelope information to the PE was studied for noises of 300 ms 
duration. The envelopes of lead and lag were sinusoidally amplitude modulated (AM) and the 
modulation phase (delay) between lead and lag was varied across conditions. Subjects 
adjusted the delay of the TFS in the lag to find the echo threshold while the envelope delay
was kept constant. To simulate imperfect TFS representation, the correlation between the TFS
in the lead and the lag was varied. 

Both AM and TFS strongly influenced the precedence effect: 
1)No echo thresholds could be obtained for inconsistent delays between TFS and envelope, 
  suggesting that both cues contribute to the PE of ongoing sounds and need to be 
  consistently represented in future CIs. 
2)For consistent delays the presence of AM elevated echo thresholds above the unmodulated 
  baseline. This suggests that AM aids the PE potentially through the evaluation of multiple
  onsets or by serving as a grouping cue.
3)Despite partial TFS decorrelation between lead and lag the presence of AM made it possible
  to obtain echo thresholds, a potential route to overcome effects of imperfect TFS 
  representation in electric stimulation. 

Support provided by the Intramural Programme of the MRC.