Proceedings of Meetings on Acoustics

Transcription

1 Proceedings of Meetings on Acoustics Volume 19, ICA 2013 Montreal Montreal, Canada 2-7 June 2013 Psychological and Physiological Acoustics Session 1pPPb: Psychoacoustics and Perception (Poster Session) 1pPPb19. Boundary effects on the illusory continuity of an interrupted glide through a notched noise. Valter Ciocca* and Nicholas R. Haywood *Corresponding author's address: School of Audiology and Speech Sciences, University of British Columbia, 2177 Wesbrook Mall, Vancouver, V6T 1Z3, British Columbia, Canada, vciocca@mail.ubc.ca This study investigated the illusory continuity of an interrupted frequency glide through a notched-noise burst. A 2I-2AFC procedure was used to measure detection of the (target) portion of the frequency glide that overlapped in time with the noise. The portions of the glide preceding and following the noise (flankers) could be present or absent. The center frequency of the notch coincided with either the frequency end-point of the flanker that preceded the noise, or the onset frequency of the flanker that followed the noise. A control condition with a wideband noise burst (absent notch) was also included. Performance was poorest in the absent notch condition, and was significantly poorer with present than with absent flankers. This suggests that listeners perceptually restored the missing target when flankers were present. Performance was also less accurate (indicating stronger illusory continuity) when the notch was centered on the end-point of the flanker that preceded the noise. These results suggest that the masking of the onset of the flanker following the noise provides a stronger cue to the perception of continuity than the masking of the offset of the flanker that precedes the noise. Published by the Acoustical Society of America through the American Institute of Physics 2013 Acoustical Society of America [DOI: / ] Received 21 Jan 2013; published 2 Jun 2013 Proceedings of Meetings on Acoustics, Vol. 19, (2013) Page 1

2 INTRODUCTION When the middle portion of a signal is replaced by a louder masker, listeners typically perceive the signal as continuous during the masker (Miller and Licklider, 1950, Thurlow, 1957; Vicario, 1960). Warren et al. (1972) proposed that the illusory percept of a continuous signal resulting from a physically interrupted signal ( perceived continuity ) occurs when two requirements are met. Firstly, the masking sound should contain sufficient energy to mask a truly continuous signal. The second requirement is that there should be no contextual evidence that the signal has stopped during the time that the masker is presented (i.e., no evidence of signal offset and/or re-onset; see also Bregman and Dannenbring, 1977). These requirements has been called the energetic masking and the edge masking requirements, respectively (Haywood et al., 2011). Haywood et al. (2011) investigated perceived continuity when a temporal (silent) gap is introduced around the temporal centre of the masker. They found that listeners perceived a strong continuity with a 20-ms temporal gap, in apparent violation of the energetic masking requirement. They concluded that energetic masking is not a necessary condition for perceiving the illusory continuity of an interrupted signal, and proposed that edge masking is a necessary requirement. They also noted that edge masking alone cannot account for continuity, as perceived continuity was weakened by a sufficiently long (60-ms) temporal gap within the masker (even though the signal edges immediately preceding and following the masker were still masked with such stimuli). The present study employed a forced-choice procedure in which listeners detected the presence of a target tone during a noise-burst masker. When present, two surrounding flanker tones were expected to impair performance at this task. This is because the flanker tones were expected to be heard as continuous through the masker, and so the perceptually continuous tone heard when the target was absent was expected to be indistinguishable from the truly continuous tone heard when the target was present (Haywood et al., 2011; Ciocca and Haywood, 2011). The current study further explored the conditions under which perceived continuity can be perceived in the absence of energetic masking. The target was a 150-ms frequency glide; the flankers were two frequency glides that were separated by a 150-ms silent gap. The target and the flankers were always aligned along a log-frequency trajectory. The centre frequency of a spectral notch was varied within a broadband noise masker that was simultaneous with the target. In three notch conditions, the noise was expected to mask either the offset of the first flanker only ( offset edge ), the onset of the second flanker only ( re-onset edge ), or was expected to mask both signal edges. Critically, performance at the task was expected to be accurate in the notched masker conditions when the flankers were absent from the sequence. Evidence of this would suggest that impaired performance in the flankers-present conditions could not be attributed simply to an energetic masking effect of the noise-burst on the target. By using these spectral notch conditions, the study had the following two goals: 1. to investigate the effect of independently masking each signal edge, and 2. to replicate of previous findings that perceived continuity can occur in the absence of energetic masking with notched maskers (Ciocca and Haywood, 2011). METHOD Participants Fourteen listeners with normal hearing (20 db HL or better at the octave frequencies between Hz) took part in this experiment. Before participating in the experiment, all listeners reported no known hearing problems and an age of between years. The mean age of the listeners was 25.7 years, and the group reported a broad range of musical experience. All listeners had previously completed a related experiment (Ciocca and Haywood, 2011); both experiments were run during a single session.. Stimuli and Apparatus The target tone was a 150-ms ascending frequency glide that spanned half an octave octave, between an onset frequency of 1189 Hz and an offset frequency of 1681 Hz (spanning ½ octave, centred at 1414 Hz). All frequency trajectories were calculated in logarithmic space. The target was presented at 43 dba, and had 10-ms raised-cosine ramps at onset and offset. A noise-burst masker was presented simultaneously with the target tone. The spectral properties of the masker were varied in order to obtain four masker conditions. The broadband masker comprised a white noise burst that was band-pass filtered between 500 Hz and 4000 Hz (roll-off = 56 db per octave); this Proceedings of Meetings on Acoustics, Vol. 19, (2013) Page 2

3 masker noise was presented at a level of 57 dba. This masker noise was selected because it was found to have a strong masking effect on the target tone in a related experiment (Ciocca and Haywood, 2011). The three other maskers were obtained by applying a band-stop filter (roll-off equal to 56 db/octave) to the broadband masker ( notched maskers). The band-stop filter spanned 4 semitones (ST). For the offset-notch masker, the notch was centred at the offset frequency of the first flanker (1189 Hz; cut-off frequencies of 1059 Hz and 1335 Hz). For the centre-notch masker, the band-stop filter was centred at 1414 Hz (cut-off frequencies of 1259 Hz and 1587 Hz). For the onset-notch masker, the notch was centred at the onset frequency of the second flanker (1681 Hz; cut-off frequencies of 1498 Hz and 1887 Hz). Note that the lower cut-off of the centre-notch masker was 1 ST above the onset frequency of the target tone (1189 Hz), which corresponds to a separation of 0.45 ERB N (Glasberg and Moore, 1990). The higher cut-off of the centre-notch masker was 1 ST below the offset frequency of the target (1681 Hz), which corresponds to a separation of 0.43 ERB N (Glasberg and Moore, 1990). Therefore, the centre-notch masker was expected to have a masking effect on both the offset of the first flanker and the onset of the second flanker. Each masker had 10-ms raised-cosine ramps at onset and offset. The notched maskers were expected to be less capable of masking the target tone than the broadband masker. For each type of masker, the experiment included conditions in which the target-plus-masker pair was immediately preceded and followed by frequency glides ( flanker tones). Each flanker lasted for 375 ms, and had the same level as the target. Hence, the flanker conditions had an overall sequence duration of 900 ms. The flankers had 10-ms, raised-cosine ramps at onset and offset. The offset ramp of the first flanker overlapped with the onset ramp of the target-plus-noise stimuli; similarly, the onset ramp of the second flanker overlapped with the offset ramp of the target-plus-noise stimuli. The frequency trajectory of the flankers was matched with that of the target (rate of change equal to 1 octave per 300-ms). This meant that the leading flanker rose from 500 Hz to 1189 Hz, and the following flanker rose from 1681 Hz to 4000 Hz. For the flankers-absent conditions, the two flankers were each replaced with a silence of equivalent duration so that the overall duration of a trial was preserved. Figure 1 shows a schematic representation of the spectro-temporal properties of the stimuli for the three notched masker conditions with flankers; the red dotted line represents the target. FIGURE 1. Flanker conditions with notched maskers. The spectral notch in the masker is centred around either the offset of the first flanker ( offset-notch masker ; left panel), the onset of the second flanker ( onset-notch masker; right panel), or the centre frequency of the target ( centre-notch masker ; centre panel). Target represented by red dotted line. All stimuli were digitally synthesized on a Capybara 320 sound computation processor that was controlled via the Kyma X software package ( , Symbolic Sounds Corporation) running on an imac desktop computer ( Apple Inc.). The presentation of the stimuli and the data collection were controlled by custom software written with the Revolution 3.0 software package (Runtime Revolution Ltd., ) running on an imac desktop computer. The stimuli were output through a MOTU Ultralite external audio interface that was also used to set the Proceedings of Meetings on Acoustics, Vol. 19, (2013) Page 3

4 overall output level of the stimuli. Participants listened to the stimuli through Sennheiser HD 280 Professional earphones in a double-walled sound-attenuating chamber. Procedure The experiment used a 2I-2AFC procedure. In each trial, listeners were presented with two intervals that differed only in whether the target tone was present or absent. Listeners reported which of the two intervals contained the target tone by pressing either 1 (for the first interval) or 9 (for the second interval) on a computer keyboard. The two intervals were presented sequentially, and were separated by a 500-ms silence. Listeners were allowed to respond only after the sound presented in the second interval had ceased. After each response there was a 500-ms pause before the next trial began automatically. During this pause, on-screen visual feedback indicated whether the previous response was correct or incorrect. In total, there were 8 unique conditions: (4 masker types 2 flanker conditions), and these conditions were organized into trial blocks. Every condition was presented twice within each block in one presentation the target tone was present in the first interval, and in the other presentation it was present in the second interval. The order of the conditions within each block was randomized. The main experiment comprised a total of 10 blocks, so that each condition was presented 20 times in total. The main experiment was completed in a single session that lasted for approximately 15 minutes. Prior to the main experiment, listeners completed a brief training session that comprised two blocks of the stimuli used in the main experiment. This session was used to familiarize the listener with the stimuli, and so the data from this session were not analyzed. RESULTS AND DISCUSSION The responses were converted to d scores using the following formula: d = 1/ 2 [z(h) z(f)], where z(h) is the standardized score of the proportion of hits, and z(f) is the standardized score of the proportion of false alarms (Macmillan and Creelman, 2005). The average d scores for each condition are displayed in Figure 2. FIGURE 2. The average sensitivity (d ) scores for each masker type are displayed for flanker-absent (open circles) and flankerpresent (filled squares) conditions. The error bars represent ± 1 standard error. Proceedings of Meetings on Acoustics, Vol. 19, (2013) Page 4

5 A two-way repeated measures ANOVA was carried out to analyze these data; flanker (present vs. absent) and masker type (broadband, offset notch, centre-notch and onset-notch) were the two within-subjects factors. As expected, accuracy was significantly higher when flankers were absent (main effect of flanker, F(1, 13) = 24.9, p < 0.001). Sensitivity was also affected by the type of masker (main effect of masker type, F(3,39) = 30.2, p < 0.001). The two-way interaction was statistically significant, F(3,39) = 3.56, p < 0.05, indicating that the effect of the flanker factor varied depending on masker type. Specifically, listeners were significantly more accurate when flankers were absent for all masker types (Tukey HSD tests, p < 0.01), except for the broadband masker (Tukey HSD test, p > 0.05). When the flankers were absent, sensitivity was not significantly different among the notched masker conditions (Tukey HSD tests, p > 0.05). When flankers were present, sensitivity scores were significantly higher for the onset-notch masker than for the offset-notch masker (Tukey HSD, p < 0.05). No difference was observed between the centre-notch and the other notched maskers. Previous results provided evidence that poor performance in the 2I-2AFC task for flanker-present conditions is likely due to listeners perception of illusory continuity (Haywood et al., 2011). In a manner consistent with the previous finding, the presence of the flankers in the current experiment caused a decrease in sensitivity in the detection of the target. Hence, the results support the notion that perceived continuity can occur when the masker does not energetically mask the target (see also, Riecke et al., 2008, for similar findings). When the flankers were present, sensitivity was significantly lower (indicating stronger perceived continuity) in the offset-notch condition than in the onset-notch condition. Since the target was equally detectable for these notched maskers when flankers were absent, these findings suggest that there is an asymmetry in the perceptual weight given to the onset and offset signal edges. If replicated in future studies, this result should be taken into account by perceptual and physiological models of perceived continuity (see, for example, Riecke et al., 2008; Petkov et al., 2007; Vinnik et al., 2010). CONCLUSION For the three notched-noise conditions, performance at the task was good when the flanker tones were absent, and performance decreased significantly when the flanker tones were present in the sequence. This impairment in performance is attributed to the perceived continuity of the flanker tones through the masker (Haywood et al., 2011; Ciocca and Haywood, 2011), and suggests that perceived continuity can occur when a portion of a frequencyvarying signal is replaced by a notched masker that does not produce energetic masking of the missing portion of the signal. This finding is consistent with the results of Ciocca and Haywood (2011). The findings also suggest that the masking of the signal re-onset edge after the masker might be more important than the masking of the offset edge for the occurrence of perceived continuity. ACKNOWLEDGMENTS This work was supported by a startup fund awarded to the first author by the Faculty of Medicine, University of British Columbia. REFERENCES Bregman, A.S., and Dannenbring, G.L. (1977). Auditory continuity and amplitude edges, Canad. J. Psychol, 31, Ciocca, V. and Haywood, N. R. (2011). Illusory continuity and masking: Evidence for an illusory tone percept through a notched noise, J. Acoust. Soc. Am. 129, Haywood, N., Chang I.-C. J., and Ciocca V. (2011). Perceived tonal continuity through two noise bursts separated by silence, J. Acoust. Soc. Am. 130, Miller, G.A., and Licklider, J.C.R. (1950). The intelligibility of interrupted speech, J. Acoust. Soc. Am. 22, Petkov, C.I., O Connor, K.N., and Sutter, M.L. (2007). Encoding of illusory continuity in primary auditory cortex, Neuron 54, Riecke, L., Van Opstal, A.J., and Formisano, E. (2008). The auditory continuity illusion: a parametric investigation and filter model, Percept. Psychophys. 70, Thurlow, W.R. (1957). An auditory figure ground effect, Am. J. Psychol. 70, Vicario, G. (1960). L effetto tunnel acustico, Rivista di Psicologia. 54, Proceedings of Meetings on Acoustics, Vol. 19, (2013) Page 5

6 Vinnik, E., Itsov, P., and Balaban, E. (2010). A proposed neural mechanism underlying auditory continuity illusions, J. Acoust. Soc. Am. 128, 1-6. Warren, R.M., Obusek, C.J., and Ackroff, J.M. (1972). Auditory induction: Perceptual synthesis of absent sounds, Science 176, Proceedings of Meetings on Acoustics, Vol. 19, (2013) Page 6