BREAKDOWN OF PRECEDENCE WITH COCHLEAR IMPLANTS - SIMULATIONS SHOW
IMPORTANCE OF SPECTRAL CUES 

Bernhard U. Seeber(1) and Ervin Hafter

3210 Tolman Hall, Dept. of Psychology, Univ. of California Berkeley, USA
(1) now at: MRC Institute of Hearing Research, Nottingham, UK

The precedence effect shows the suppression of a lagging copy of
a sound, the echo, in the presence of the leading sound, the
direct sound. It is considered a binaural phenomenon, but the
results of the present studies suggest that it has to be seen in
the larger context of auditory scene analysis, especially with
cochlear implants (CIs).  

Using current devices, CI-subjects have very limited access to
interaural time differences (ITDs) and spectral information,
while interaural level differences (ILDs) as well as
representations of the temporal envelope are relatively
available. This limited information is often sufficient to
provide localization in quiet, but the presence of other sounds
in the background might disturb it. In the first part we present
results of a study on the precedence effect with bilateral
CI-subjects. A 10ms noise-burst or a CVC-word were played from
+30° or -30° accompanied by a delayed, identical sound copy from
the opposite side loudspeaker. Subjects had to indicate with a
light pointer the perceived sound direction or, in the event of
two perceived sounds, to indicate in separate runs either the
left- or rightmost sound location. Results show evidence for 3
different outcomes:  
(1) Some subjects localized a single sound between lead and lag
    in the middle between both speakers, even for long delays. 
(2) Two subjects showed precedence, with a single sound localized
    towards the lead at short delays. 
(3) Two subjects showed a breakdown of precedence, with lead and
    lag sounds localized separately, beginning with relatively
	 short delays.  
Result (3) is interesting as it shows that CI-patients were able
to separate and locate two sounds simultaneously in a situation
where normal hearing subjects would hear only one sound. This is
contrary to the belief that less information should produce less
segregation ability. This breakdown of precedence was replicated
with normal hearing subjects in a similar paradigm using a
noise-band vocoder CI-simulation.  

In hopes of explaining this breakdown of precedence, we varied
the interaural match of the carrier frequencies of a sinusoidal
vocoder and found that precedence broke down if carriers were not
matched in frequency. However, for zero or small frequency
offsets some subjects showed precedence and fusion into a single
image. Interestingly, the interaural phase of the carrier played
only a limited role which shows that precedence of ongoing sounds
can be solely based on ILDs and envelope-ITDs. Together, these
results suggest that proper place matching of CI-electrodes is
important for the analysis of concurrent sounds. 

Support provided by NIH RO1 DCD 00087, NOHR 018750, and the MRC.