What underlies between-frequency gap detection? Shuji Mori Kyushu - - PowerPoint PPT Presentation

2014 Symposium on Across-Channel Processing in Human Audition Niagara-on-the-lake, Canada, 2014/08/06

Shuji Mori

Kyushu University

What underlies between-frequency gap detection?

1

‘What’ and ‘Where’

What makes BF gap detection so difficult? Across-channel processing Relative timing/attention shift Channel bandwidth Onset cue Where in auditory pathway does it take place? Peripheral Central Primary auditory cortex

Psychophysics Lab. Department of Informatics Faculty of ISEE

2

Relative timing

Monitoring offset of leading and onset of trailing marker (Phillips, 1999)

Discontinuity detection

Reflects central processes Performed peripherally

Psychophysics Lab. Department of Informatics Faculty of ISEE

3

Attention shift

Auditory attention

Directed to frequency channel (attention band; Scharf et al., 1987) Enhances auditory processing

Two hypotheses

• 1. Attention dwell time (Fitzgibbons et al., 1974)

Minimum time spent at one channel before shifting to another channel

• 2. Attention disruption (Phillips et al., 1997)

Imprecise time-stamping when shifting to unattended channel

Can be differentiated in terms of psychometric function (Kikuchi et al., 2014)

Psychophysics Lab. Department of Informatics Faculty of ISEE

4

Psychometric function

Within-frequency (Florentine et al., 1990)

Proportion of ‘Yes’ responses 0.5 1.0 1 10 100 Gap duration (msec)

Between-frequency Attention disrupt. Dwell time Internal noise

Psychophysics Lab. Department of Informatics Faculty of ISEE

0.0 0.20 0.40 0.60 0.80 1.0 0.1 1 10 100

Py Gap duration (msec)

0.0 0.20 0.40 0.60 0.80 1.0 0.1 1 10 100

Py Gap duration (msec)

0.0 0.20 0.40 0.60 0.80 1.0 0.1 1 10 100

Py Gap duration (msec)

5

Kikuchi et al. (2014)

P1 P2 P3

800-800 800-1600 1600-800 800-3200 3200-800

Psychophysics Lab. Department of Informatics Faculty of ISEE

6

Problems on attention shift

• Attention shifts instantaneously (Scharf et al., 2007)
• Attention can be directed to multiple frequencies

(Schlauch & Hafter, 1991) No study yet to manipulate attention in BF gap detection

Psychophysics Lab. Department of Informatics Faculty of ISEE

7

Channel bandwidth

Formby & Forrest (1991) Estimate channel bandwidth from BF gap detection

(Adapted from Formby & Forrest, 1991, p.836, Fig.4) (Formby & Forrest, 1991, p.834, Fig.3)

About half of bandwidth of typical auditory filters (e.g. Patterson & Moore, 1986)

Psychophysics Lab. Department of Informatics Faculty of ISEE

8

Channel bandwidth

Single- and multiple-channel models (Forrest & Formby, 1996; Heinz et al., 1996)

(Forrest & Formby, 1996, p.24, FIGURE 1)

Psychophysics Lab. Department of Informatics Faculty of ISEE

9

Channel bandwidth

Single- and multiple-channel models (Forrest & Formby, 1996; Heinz et al., 1996) Gap thresholds reflect narrowed channel bandwidth

(Forrest & Formby, 1996, p.29, FIGURE 5)

Psychophysics Lab. Department of Informatics Faculty of ISEE

10

Problems on bandwidth account

No explanation of why channel is narrowed under BF gap detection Empirical evidence lacking for effect of bandwidth on gap detection

Psychophysics Lab. Department of Informatics Faculty of ISEE

11

Trailing marker onset

WF

Amplitude Time

• nset

BF

Amplitude Time

• nset
• nset

Can be accomplished by onset detection (≈discontinuity detection) Onset cue unreliable

Psychophysics Lab. Department of Informatics Faculty of ISEE

12

Onset account

Availability of TM onset cue distinguishes between WF and BF gap detection

 Neuronal onset responses match WF gap detection Werner et al. (2001)

(Werner et al., 2001, p.741, Figures 2 and 3)

Psychophysics Lab. Department of Informatics Faculty of ISEE

13

Onset account

 Reducing onset-cue availability impairs gap detection Oxenham (2000) Inducing amplitude difference to two WF markers Amplitude Time Worsens gap detection to BF level Grose et al. (2007) Presenting secondary tone with TM worsens BF gap detection TM onset obscured by the tone

Psychophysics Lab. Department of Informatics Faculty of ISEE

14

Onset account

Eggermont (2000) Single-cell recording at cat auditory cortex Manipulating LM duration

(Eggermont, 2010, pp.1458, Fig.6; pp.1459, Fig.7)

TM onset responses appear 40-55 ms after LM onset

Corresponds to behavioral data (Phillips et al., 1997)

Psychophysics Lab. Department of Informatics Faculty of ISEE

15

Problems on onset account

Only explains qualitative categorical difference between WF and BF

Unable to deal with frequency separation effects on BF gap detection

(Phillips et al., 1997, JASA, pp.3697, Fig.2)

Psychophysics Lab. Department of Informatics Faculty of ISEE

16

‘Where’ in auditory pathway

Peripheral

Auditory filter (Formby & Forrest, 1991; Forrest & Formby, 1996)

Central

Channel monitoring (Phillips et al., 1997) Attentional operation (Fitzgibbons et al., 1974)

Primary auditory cortex

• Broadly-tuned onset-sensitive neurons (Eggermont, 2000)
• Comparable MMN for WF and BF generated near PAC (Heinrich

et al., 2004)

• Frequency-separate regions of onset responses (Mitsudo,

Hironaga)

Psychophysics Lab. Department of Informatics Faculty of ISEE

17

Our approach

ABR MEG

(Coren et al., 1994, Sensation & perception, pp.204, Fig.6-17)

0.0 0.20 0.40 0.60 0.80 1.0 0.1 1 10 100

Py Gap duration (msec)

Psychophysics

Psychophysics Lab. Department of Informatics Faculty of ISEE

18

Auditory brainstem response

Reflects onset responses of auditory nerves and brainstem neurons Mostly measured to WF TM onset

Werner et al., 2001; Poth et al., 2001

Grose et al. (2007)

Measured ABR to BF TM onset consisting of two-tone complex

Psychophysics Lab. Department of Informatics Faculty of ISEE

19

Method

Participants

10 healthy male students (mean 22.9 yrs)

Stimuli

LM & TM – 0.5-oct. bandnoise of 50 ms (rise/fall 3 ms) 45 dB SPL monaurally presented to left ear LM/TM center frequency (Hz): 800/800, 800/1600, 1600/800, 800/3200, 3200/800 No gap Gap duration

Psychophysics Lab. Department of Informatics Faculty of ISEE

20

Method

Procedure

threshold measurement 2IFC 1-up 6-down procedure to obtain 89.1% accuracy gap thresholds

5 10 15 20 25 30 800/800 800/1600 1600/800 800/3200 3200/800

Gap Threshold (ms) LM/TM Frequency (Hz)

Psychophysics Lab. Department of Informatics Faculty of ISEE

21

Method

ABR measurement Gap durations set to 0 (no gap), and 0.5, 1.0, and 1.5 times of individual gap thresholds for each LM/TM frequency 2000 presentations for each gap/frequency at 3-Hz rate Recorded at Cz with a reference at A2 and a ground at Fpz Band-pass filtering between 100 and 3000 Hz 100 kHz sampling rate

Psychophysics Lab. Department of Informatics Faculty of ISEE

22

Method

Amplitudes and latencies extracted from individual ground averages

5msec 0.1μV

TM onset Wave V Latency Amplitude

Psychophysics Lab. Department of Informatics Faculty of ISEE

23

Results

Mean amplitudes and latencies of 10 ps

0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.5T T 1.5T 0 0.5T T 1.5T 0 0.5T T 1.5T

800/800 800/1600 1600/800 800/3200 3200/800

Amplitude (microV) Gap Condition

11 12 13 14 15 0.5T T 1.5T 0 0.5T T 1.5T 0 0.5T T 1.5T

800/800 800/1600 1600/800 800/3200 3200/800

Latency (msec) Gap Condition

Psychophysics Lab. Department of Informatics Faculty of ISEE

24

Results

Mean amplitudes as a function of gap duration in ms

0.06 0.08 0.1 0.12 0.14 0.16 0.18 5 10 15 20 25 30 35

800/800 800/1600 1600/800 800/3200 3200/800

Amplitude (microV) Gap Duration (ms)

6~7 ms + 50-ms LM duration ≈ 55-ms from LM onset

Psychophysics Lab. Department of Informatics Faculty of ISEE

25

Discussion

Increased ABR amplitude (TM onset response?) at ~55 ms after LM onset

 TM onset responses appear 40-55 ms after LM onset (Eggermont)

(Eggermont, 2010, pp.1459, Fig.7)

Psychophysics Lab. Department of Informatics Faculty of ISEE

26

Discussion

 ABR to TM onset reflects physical duration rather than psychophysical threshold

 TM onset response is not a sole determinant of gap threshold  Very high accuracy (89.1%) of gap detection criterion may contribute to the discrepancy  LM duration needs to be manipulated

 ABR reflects broadly tuned mechanism

 Low (suppressed?) ABR observed for BF below 55 ms

Psychophysics Lab. Department of Informatics Faculty of ISEE

27

Conclusion

What makes BF gap detection so difficult?  Unavailability of TM onset cue  Other processes to be identified for frequency separation effect Where in auditory pathway does it take place?  As early as brainstem for onset cue  Primary auditory cortex  Peripheral?

Psychophysics Lab. Department of Informatics Faculty of ISEE

28

Comprehensive approach

ABR MEG

(Coren et al., 1994, Sensation & perception, pp.204, Fig.6-17)

0.0 0.20 0.40 0.60 0.80 1.0 0.1 1 10 100

Py Gap duration (msec)

Psychophysics DSAM simulations

Psychophysics Lab. Department of Informatics Faculty of ISEE

29

Currently ongoing…

BF gap detection with close frequency separation

0.0 0.20 0.40 0.60 0.80 1.0 10 20 30 40

800/800 800/830 800/880 800/1000

Py Gap duration (msec)

Followed by MEG, ABR, and DSAM

Psychophysics Lab. Department of Informatics Faculty of ISEE

30

Acknowledgement

Kyushu University Psychophysics lab, ISEE: Faculty of Medicine

Nobuyuki Hiirose Shozo Tobimatsu Takako Mitsudo Naruhito Hironaga Yousuke Kikuchi 0ther students and staffs

University of Toronto

Willy Wong

Psychophysics Lab. Department of Informatics Faculty of ISEE

31

Thank you for your attention

Psychophysics Lab. Department of Informatics Faculty of ISEE