Spectral overlap and interaural time difference sensitivity: Possible role of binaural interference

Abstract

A follow-up experiment to those conducted by Brown and Yost [(2011). J. Acoust. Soc. Am. 130, 358–364; (2013). Basic Aspects of Hearing: Physiology and Perception (Springer, London, UK)] examined interaural time difference (ITD) discrimination for a low-frequency target noise band flanked by monotic noise bands that were either lower-frequency than the target band, higher-frequency, or both. The flanking bands were either spectrally contiguous with the target band or spectrally separated. Significant interference in ITD processing occurred in the presence of the high-frequency flanking band. Results are discussed by way of a comparison of the conditions in the present study to those in studies of binaural interference. The possible role of attention is also discussed.

1. Introduction

Cochlear implant patients who retain some residual low-frequency hearing (either in the implanted or the unimplanted ear) often show improved intelligibility in a competing background because of the ability to combine their electric stimulation with low-frequency acoustic stimulation (EAS, see Gifford et al., 2009). When the patient has residual low-frequency hearing bilaterally, one might consider stimulating both ears acoustically. The current study is motivated by the eventual possibility of improving speech intelligibility by delivering interaural time differences (ITD) via bilateral low-frequency acoustic stimulation. Indeed Brown and Bacon (2007) showed that when a monaural 4-channel vocoder was combined with dichotic 500-Hz lowpass speech (to simulate EAS conditions in which residual hearing is present in both ears), an ITD of 600 μs improved speech understanding in different backgrounds compared to conditions in which the ITD was zero.

Given the nature of hearing loss and the risks to hearing preservation during cochlear implant surgery, it seems reasonable to expect that the frequency range of residual hearing might differ across the two ears. This was addressed in a study by Brown and Yost (2011; hereinafter referred to as experiment 1), in which the ITD sensitivity of listeners with normal hearing was measured when the bandwidths of filtered noise were varied interaurally. The low-frequency noise band delivered to the left ear was either 50–125 Hz, or 50–250 Hz. To the right ear, the high-pass cutoff frequency was always 50 Hz, and the low-pass cutoff was either equal in frequency to that in the left ear, or higher; see the left panel in Fig. 1 for a schematic of these stimulus conditions. Results showed that as the interaural spectral asymmetry increased, ITD sensitivity decreased. This was an interesting result, because, even with a fixed region of interaural spectral overlap, the presence of higher-frequency energy in the right ear proved deleterious to performance (Brown and Yost, 2011). We argued that the decrease in ITD sensitivity was not because of an increasing overall interaural level difference at the right ear (which occurred as the right-ear noise bandwidth increased), and findings from a subsequent study definitively ruled this out (Brown and Yost, 2013).

Fig. 1.

Representation of the spectrum of noises at each ear used in experiments 1, 2, and the present study: (top) left; (bottom) right. (Left) In experiment 1 (Brown and Yost, 2011), the upper cutoff of the bandpass filter was increased at the right ear. (Middle) In experiment 2 (Brown and Yost, 2013), the center frequency of the band of noise at the right ear was shifted up relative to that at the left ear (the bandwidths remained equal). (Right) In the present study, there was a band of spectrally overlapping noise at both ears (interaural spectral overlap, ISO), and monotic flanking bands above and below the ISO band were moved away from the ISO band (difference in monotic flankers, ΔMF).

In the second study (Brown and Yost, 2013, hereinafter referred to as experiment 2), the noise bandwidth and, thus, the overall interaural level difference at each ear was kept the same. Specifically, the band of noise presented to both ears was either $\frac{1}{3}$, $\frac{2}{3}$, or 1 octave wide. In the left ear, the center frequency of the noise band was always 250 Hz. In the right ear, it was systematically increased to frequencies above 250 Hz, which effectively slid the noise band in the right ear upward in frequency relative to that in the left ear (see Fig. 1, middle panel, for a schematic representation). Thresholds for ITD discrimination increased as a result of increasing the center frequency of the noise band presented to the right ear (i.e., ITD sensitivity decreased as the amount of binaural spectral overlap decreased).

In discussing these findings (Brown and Yost, 2013), we pointed out that a “three-band” condition resulted when the bands of noise presented to the two ears had different center frequencies but still had regions of spectral overlap: (1) a band of low-frequency noise at only the left ear, (2) a band of interaurally spectrally overlapping (ISO) noise presented to both ears, and (3) a band of high-frequency noise at only the right ear (see Fig. 1, middle panel). Along similar lines, the conditions in experiment 1 (Brown and Yost, 2011) could also be described as representing a multi-band configuration: a region of interaural spectral overlap flanked by a high-frequency monotic band (Fig. 1, left panel). In both of these studies, the various bands were spectrally contiguous (Fig. 1, left and middle panels). We conjectured that the decline in ITD discrimination performance we measured in these experiments might represent a form of binaural interference, i.e., a reduction in the processing of interaural time differences present for a given target sound when simultaneous “interfering” sounds diverge from the target in both frequency and interaural differences (see Heller and Richards, 2010, for a review of the binaural interference literature). Binaural interference has typically been reported under conditions in which the interferers are spectrally remote and binaural, while in the Brown and Yost (2011, 2013) studies the “flanking” bands could be considered spectrally contiguous monotic interfering bands. The goal of the current paper was to explore ITD discrimination for stimulus configurations that were similar to those used in experiments 1 and 2 (Brown and Yost, 2011, 2013), but where the noise bands were not spectrally contiguous (i.e., Fig. 1, right panel). That is, we wanted to explore in more detail how binaural interference might relate to ITD discrimination involving stimuli that are spectrally different across the two ears.

2. Methods

2.1. Subjects

Six subjects who reported normal hearing participated. All subjects wore Sennheiser HD250 headphones (Old Lyme, CT) while seated in a sound proof room. The procedures were approved by the Arizona State University Institutional Review Board for the Protection of Human Subjects.

2.2. Stimuli

The basic stimulus was a 200-ms band of noise shaped with a 10-ms raised cosine rise-decay time and presented with a random (from presentation to presentation) overall monotic band is higher in frequency between 86 and 90 dB sound pressure level (SPL). One spectral band was identical in frequency across the ears (the ISO band). There were also two monotic flanking bands, one lower in frequency than the ISO band and delivered to the left ear only, and one higher in frequency and delivered to the right ear only. The distance between the ISO band and the monotic flankers (ΔMF) was manipulated. In some cases the spectral distance was zero (ΔMF = 0); that is, the monotic flankers were spectrally contiguous with the ISO band, a configuration that was equivalent to that used in experiment 2 (Brown and Yost, 2013). In other cases the flanking bands were spectrally separated from the ISO band (ΔMF > 0). Conditions included either both flanking bands, the high-frequency flanker only, the low-frequency flanker only, or neither flanker.

The high- and low-frequency cutoffs of the ISO, lower-frequency, and higher-frequency flanking bands are shown in Table 1, along with their corresponding bandwidths, in $\frac{1}{6}$ octave units (the low- and high-frequency flanking bands always had the same bandwidth). Table 1 also displays the values of ΔMF. The cutoffs were chosen so that when ΔMF = 0, the conditions are identical to several conditions in Brown and Yost (2013).

Table 1.

Cutoff frequencies (Hz), bandwidths ($\frac{1}{6}$ octaves), and differences in monotic flanker (ΔMF, $\frac{1}{6}$ octaves).^a

Cutoff frequencies (Hz)						Bandwidths ($\frac{1}{6}$ octaves)		ΔMF ($\frac{1}{6}$ octaves)
ISO region		Low flanker		High flanker
HiPass	LoPass	HiPass	LoPass	HiPass	LoPass	ISO	Flankers
223	315	198	223	315	354	3	1	0
223	315	157	177	397	446	3	1	2
223	315	125	140	500	562	3	1	4
223	315	99	112	630	708	3	1	6
250	315	198	250	315	397	2	2	0
250	315	157	198	397	500	2	2	2
250	315	125	157	500	630	2	2	4
250	315	99	125	630	794	2	2	6
281	315	198	281	315	446	1	3	0
281	315	157	223	397	562	1	3	2
281	315	125	177	500	708	1	3	4
281	315	99	140	630	892	1	3	6
315	315	198	315	315	500	0	4	0
315	315	157	250	397	630	0	4	2
315	315	125	198	500	794	0	4	4
315	315	99	157	630	1000	0	4	6

^aLeftmost 6 columns: High (HiPass) and low (LoPass) pass cutoff frequencies (Hz) for the Interaural Spectral Overlap (ISO) band and the Low and High Flanker bands. Toward the right: Bandwidths for the ISO and flanker bands (the Low and High Flankers have the same bandwidth). Rightmost column: Distance of Monotic Flanker from ISO (ΔMF) in units of 1/6th Octaves (zero means the flankers were contiguous with the ISO band).

2.3. Procedure

A two-interval, two-alternative forced choice task was employed in which the noise in one interval contained an ITD of one-half the nominal ITD called for by the track favoring the left ear, and the other interval contained the same ITD favoring the right ear. The interval containing the left leading ITD was randomly determined, and the listeners indicated in which interval the image was perceived more to the left. A two-down, one-up adaptive tracking procedure (tracking the 70.7% correct point on the psychometric function) with feedback was used to estimate ITD thresholds. The initial ITD was 500 μs, and step sizes were 50 μs for the first two reversals and 20 μs for the last six reversals (runs were eight-reversals long). Thresholds were the average of the values at the last six reversals with an average of at least three such thresholds used to estimate the final threshold. This procedure was identical to that used in experiments 1 and 2 (Brown and Yost, 2011, 2013).

3. Results

Figure 2 shows mean interaural time difference (ITD) thresholds (μs) as a function of ΔMF for six subjects. The different panels represent performance in the presence of both monotic flankers (left panel), the low-frequency flanker only (middle panel), or the high-frequency flanker only (right panel). Performance when no flankers were present (ISO band only) is represented in each panel as “None” (the “None” data are the same for each panel and are replotted for reference). The parameter on each figure is the width of the ISO band in $\frac{1}{6}$ octave units. The bandwidths of the high- and low-frequency flankers were always equal for a given condition and were inversely proportional to the ISO bandwidth (see Table 1).

Fig. 2.

(Color online) Mean ITD thresholds (μs) for detecting an ITD in the ISO band as a function of ΔMF (distance of monotic flanking bands from the ISO bands in $\frac{1}{3}$ octave units). Error bars are 1 ± 1 standard error. “None” means there were no flanking bands. The parameter in each plot is ISO bandwidth in $\frac{1}{6}$ octave units. (Left) The data when both monotic flankers were present; (middle) when only the low-frequency flanker was present; and (right) when only the high-frequency flanker was present.

As shown in Fig. 2, the influence of the monotic flanking bands on ITD sensitivity in the ISO band depends to a large extent on which flanker is present. Although there is a noticeable increase in ITD thresholds when both flankers are present (compare “None” with the other data in the left panel), this increase appears to be in large part because of the presence of the high-frequency flanker. The low-frequency flanker contributed relatively little to the decline in performance observed. These observations are corroborated by inferential statistics. A repeated-measures analysis of variance (ANOVA) was first conducted, with flankers, ISO width, and ΔMF the main factors. This analysis revealed significant main effects for each factor, p < 0.05. We then conducted three additional ANOVAs on each flanker condition (Both, Low, and High) separately, using ISO width and ΔMF as the two factors (because of the multiple analyses, a Bonferroni correction was applied to all of the p values). The No-flanker data were included in each analysis. For the low-flanker data, neither factor revealed a significant effect, p > 0.05. For the Both-flanker and the High-flanker data, both factors were statistically significant, p < 0.05. Thus, neither the proximity in frequency of the low-frequency flanker to the ISO region, nor its relative width had a statistically significant effect on ITD sensitivity, whereas both of these factors were implicated for the high-frequency flanker.

Separate Bonferroni-adjusted post hoc Tukey tests were performed on the Both-flanker and High-flanker data targeting the levels of the ΔMF variable. For the High-flanker data, performance when ΔMF was $\frac{1}{3}$, $\frac{2}{3}$, or $\frac{3}{3}$ ($\frac{2}{6}$, $\frac{4}{6}$, or $\frac{6}{6}$) of an octave was significantly different than when no flankers were present, and performance in all other groups was statistically equivalent. For the Both-flanker data, performance when ΔMF was $\frac{1}{3}$, $\frac{2}{3}$, or $\frac{3}{3}$ of an octave was significantly different than when either no flankers were present, or ΔMF was zero. All other comparisons did not reveal statistically significant results.

4. Discussion

This study is the third in a series of related experiments examining the relationship between interaural spectral overlap and ITD sensitivity. The results confirm and expand upon those in the first two studies of the series (Brown and Yost, 2011, 2013), both of which showed that ITD thresholds increased as the area of monotic high-frequency energy became broader. In the current experiment, results indicate that the presence of low-frequency monotic energy elicits only a small (and statistically non-significant) change in ITD thresholds as compared to conditions in which no flanker is present. On the other hand, the presence of a monotic flanking band on the high-frequency side of the ISO band is shown to significantly “interfere” with ITD processing. Interestingly, this interference is only observed (at least statistically) when the high-frequency monotic flanker is not contiguous with the ISO region. The data for ΔMF = 0 when the bands were $\frac{2}{3}$ octave wide are very similar to those from the identical conditions in experiment 2 (Brown and Yost, 2013).

As noted earlier, there are several similarities between the conditions used in our recent series of studies (Brown and Yost, 2011, 2013; this paper) and those examining binaural interference (see Heller and Richards, 2010). Like our experiments, studies of binaural interference involving ITDs often employ a binaural target stimulus containing the ITD to be discriminated, and flanking diotic or dichotic interference stimuli that are spectrally separated from the target. Bands of noise similar to those employed in our studies are often composed of both the target and interference stimuli. In the present paper, the ISO band is analogous to the target band and the flanking bands analogous to the interfering bands.

There are also several differences between the stimulus conditions used in most studies of binaural interference and those used in our studies. Specifically, (1) the flanking (interfering) bands in the current series of studies are monotic, while they are binaural in studies of binaural interference; (2) the flanking bands in the current series of studies are often spectrally contiguous with the target (ISO) band, while they are spectrally separated in studies of binaural interference; (3) the frequency range used in the current series of studies is in the low frequencies (<1000 Hz), while it typically spans a large range that reaches into the high frequencies (>1000 Hz) in studies of binaural interference; and (4) the spectral separations between flankers (interferers) and ISO (target) bands in our studies are relatively small (<2 octaves), while they span two or more octaves in some studies of binaural interference.

It is generally the case that binaural interference stimuli produce an increase in the ITD thresholds measured for the target stimulus, and that interference of ITD processing occurs even when the interference stimulus is more than a critical bandwidth removed from the spectral region of the target. Most studies of binaural interference (Heller and Richards, 2010) show more interference of ITD processing when the interferer is lower in frequency than the target stimulus (but see Dye et al., 1996, in which most subjects showed greater binaural interference for higher-frequency interferers). This is opposite the effect found in the present study (see Fig. 2). This may be because of one or more of the differences cited above; in other words, the stimuli used in the current series of experiments may be so different from those used to examine binaural interference that the present data do not represent binaural interference conditions. However, we tend to see enough similarities across studies to believe that the current set of experiments do implicate, at least in part, some type of binaural interference not unlike that which has been observed previously (see Heller and Richards, 2010). For example, the low-frequency flanking stimuli used in the present study are very low (<315 Hz) compared to those used in studies of binaural interference. Since the stimuli were not presented at equal sensation levels across the different spectral conditions, the loudness of these low-frequency flanking bands is low, which might have reduced their effect on ITD discrimination. In any event, the stimulus conditions from the current set of studies differ enough from those of most binaural interference studies that we are not willing to conclude that our data represent binaural interference as it has been studied previously. While there is little doubt of interference to ITD processing in our studies, the determination of whether the interference is “binaural interference” will probably require additional research.

In all three of our experiments, the monotic bands interfere with the detection of subtle changes in the ITD present in the binaural target band (i.e., the monotic bands never lead to lower ITD thresholds as compared to conditions in which they are not present). As Brown and Yost (2013) showed, such increases in ITD thresholds when monotic flanking bands are present are not due to any overall increase in the interaural level difference that might exist between the stimuli presented to the ears. The most parsimonious explanation that we have been able to offer so far for the results of experiments 1 and 2 is based on an attention argument. That is, the presence of the monotic bands diverts the listener's attention to their presence at one ear or the other, making it more difficult for the listener to attend to the relatively subtle change in the lateralized image associated with an ITD change occurring near the middle of the head. The ITD occurs in the ISO band and since it is an ITD change about the midline, the lateralized image would be toward the middle of the head (particularly near threshold), while the monotic flankers would appear entirely at one ear or the other. The data from this paper are consistent with this observation, but with an important caveat: The degree to which ITD threshold is elevated depends on the spectral content of the monotic flanking bands. If a flanking band contains energy at frequencies higher than those of the ISO band, there is a much greater impact on ITD thresholds than if the monotic flanking band contains only lower-frequency energy than the ISO band. This caveat assumes that the low-frequency interferers have sufficient loudness to divert a subject's attention from the ISO band. If there were not sufficient loudness, the caveat might be stated in terms of the loudness of the interferer rather than just its spectral content.¹

What is the effect of spectral distance of flankers? The lateralized monotic images at each ear would likely become more salient as the flanking bands were moved away in frequency from the ISO band. This might make attending to the lateralized images associated with the ISO band even more difficult as the flanking bands were spectrally moved away from the ISO band. This is consistent with the data when either both bands are present (this paper and experiment 2, Brown and Yost, 2013) or only the high-frequency band is present (this paper and experiment 1, Brown and Yost, 2011).

Francart and Wouters (2007) conducted a study similar to our experiment 1 (Brown and Yost, 2011) but they measured interaural level difference (ILD) discrimination for $\frac{1}{3}$ octave bands of noise, where the band in one ear was shifted upward in frequency relative to that in the other ear. Results indicated an increase in ILD thresholds with increasing spectral separation of the noise bands between the ears for the center frequencies they tested (250 to 4000 Hz). Thus, it appears as if both ILD and ITD thresholds increase with increasing separation of spectral information in the two ears. With minimal spectral overlap of the two noise bands, ILD thresholds are in the range of 4–5 dB and ITD thresholds are between 200 and 500 μs, depending on the listener. While these values are much larger than interaural thresholds obtained for stimuli that completely spectrally overlap between the ears, they still represent an ability to do the task, even when there is minimal spectral overlap between the ears. While Francart and Wouters (2007) did not consider the possible role of attention or binaural interference in their study, the similarity of their stimulus conditions to those in our recent set of experiments (Brown and Yost, 2011, 2013; this paper) suggests that attention and/or binaural interference may be implicated for ILD processing as well.

This study was partially motivated by the idea that EAS patients with residual hearing in both ears might be able to process ITDs even when the spectral regions of residual hearing are not the same at the two ears (a similar motivation was discussed in Francart and Wouters, 2007). These studies suggest that as long as there is a region of interaural spectral overlap, EAS patients who have residual acoustic hearing bilaterally might be able to process interaural differences as long as the spectral regions of residual hearing at each ear do not vary by much. The inability of such EAS patients to use interaural differences may not be due solely to processing ILD and ITD cues in the region of spectral overlap. Processing ITD and ILD cues may also have something to do with one's ability to attend to binaural cues in regions of spectral overlap (where there is still residual hearing) when there is also stimulation occurring at only one ear or the other, especially when the energy in such a monotic band is higher in frequency than that in the binaural spectral band, at least at the low frequencies used in the current study. In such cases it appears as if a patient might have difficulty attending to interaural changes (ITD or ILD) in the region of spectral overlap because their attention is drawn to one ear or the other based on the presence of monotic energy at that ear.