Temporal induction

introduction | demonstration | investigate | reading | credits | downloading | home

Introduction

Under certain conditions, signals may be perceived as continuous even when temporal gaps are deliberated inserted and filled with noise (Miller & Licklider, 1950). This phenomenon applies to a wide range of signals from tones to speech, and has been termed temporal induction, apparent continuity, or the phoneme restoration effect when applied to speech signals (Warren, 1970). Temporal induction has since been shown to be one of a class of effects which includes contralateral (Warren & Bashford, 1976) and spectral induction (Warren et al, 1996). The current demonstration concerne temporal induction and allows the main stimulus properties of noise level and duration to be direectly manipulated for a variety of signal types.

Note: this demonstration uses the Matlab signal processing toolbox in a minor way (to perform bandpass filtering).

The demonstration

Start the demo by typing 'ti' at the MATLAB prompt. Then load a sound file (1). A spectrogram of the sound appears in the lower panel (2). Click on the spectrogram to hear it.

To produce speech with noise-filled gaps, simply click in the upper grid (3). This produces a stimulus with noise bursts of level and duration determined by the location of the click. For instance, the cross indicates that a level of around -3 dB and a duration of just under 200 ms has been chosen. A spectrogram of the resulting stimulus is produced, and the sound is produced. Your task is to determine the range of (duration,level) pairs for which the underlying signal appears continuous.

By default, the noise burst are wideband, but the centre frequency and bandwidth of the noise can be changed using relevant menu options (4,5).

Noise bursts can be regular or irregular, with a noise/signal duration of 1:1, or singular. This option is chosen via the multiple? menu (6).

Things to investigate

1. Start with a tone signal and adjust the level for a constant duration of, say, 200 ms, until the tone appears to continue behind the noise. Now, keeping the level constant, change the duration. Repeat the test with irregular noise bursts.
2. Repeat the procedure above for the other signal types. Are the (level,duration) pairs similar for each signal type?
3. Use a non-wideband noise masker, and explore the necessary conditions for induction to occur.
4. What other factors are involved at extreme durations and at high noise levels?
5. What factors govern the intelligibility of interrupted speech (that is, speech with no noise in the gaps) as a function of gap duration?

References

1. Miller & Licklider (1950). JASA, 22, 167-173.
2. Warren (1970). Science, 167, 392-393.
3. Warren & Bashford (1976). Percep. & Psychophys., 20, 380-386.
4. Warren et al (1997). Percep. & Psychophys., 59(2), 275-283.

Further reading

• A recent summary of these and related phenomena can be found in Warren (1996). Proc. ESCA Workshop on The Auditory Basis of Speech Perception, Keele.

Credits etc

Produced by: Martin Cooke

Release date: June 22 1998

Permissions: This demonstration may be used and modified freely by anyone. It may be distributed in unmodified form.