SIMPLE REACTION-TIMES TO SPEECH AND NON-SPEECH STIMULI! D. B. Fry (University College London)
In experiments previously reported (Fry, 1970, 1971) reaction-time was measured for very simple speech tasks involving, on the perceptual side, a choice between minimal pairs of English words and, on the motor side, the utterance of a vowel sound. It was found that choice reaction-time is affected by the complexity of the recognition task which subjects are asked to perform. Thus a decision between pit and bit heard in isolation, depending on the cue of voice onset time, takes less time than a decision between lay and ray, which depends on second and third formant configuration, while a choice between splay and spray, where the same distinction figures in a consonant cluster, takes very much longer still. These times were obtained with a key-pressing response so that no motor speech action was involved. The results indicate that the more complex the processing operation, the longer the time required by the central processing unit for speech. A second series of experiments aimed at estimating the minimum time required for speech reception and speech production by setting subjects the simplest possible speech tasks. Two types of stimulus were combined with two response modes. The.s.ttmuli were an isolated vowel [a:], and a short pulse of a pure tone, lasting 20 msec., which was heard as a click. The response modes were uttering the vowel [a:] or pressing a key. Each stimulus was paired with each response mode and simple reaction-times were measured in all four conditions. The results showed that subjects gave relatively long reaction-times when asked to make a speech response to the click and they reported that this stimulus seemed to have an inhibiting effect on the speech mechanism. The long reaction-times may therefore have been related not so much to the fact that a non-speech stimulus was being given as to some physical characteristic of the particular stimulus. Two experiments are reported in the present paper, the first of them I
Supported by a grant from the Science Research Council.
Cortex (1975) 11, 355-360.
356
D. B. Fry
designed to find a suitable non-speech stimulus for use in contrast with a speech stimulus, and the second constituting a repetition of the previous experiment with speech and non-speech stimuli and responses.
EXPERIMENT
I
Simple reaction-time was measured when subjects responded by uttering [a:] to each of four stimuli, one speech and three non-speech. The purpose of the experiment was to determine whether a click has a specific effect in delaying a subject's response or whether comparable reaction-times would be obtained when other kinds of non-speech stimulus were used. The duration and intensity of an auditory stimulus are well known to influence reaction-time. A click is necessarily of short duration and it is difficult to ensure that its intensity shall be comparable with that of other stimuli used. In order that the intensity of all stimuli should be as closely controlled as possible, a rapidly reading peak voltmeter was designed, using a light emitting diode display. The response of this meter is such that one cycle of a 1000 Hz tone applied to the device gives the same reading as a continuous tone of equal frequency and leveP The stimuli provided for subjects were all derived from tape recordings and the peak-reading meter was used both in recording and in replaying the stimuli, which were relayed to subjects binaurally through calibrated headphones. The speech stimulus used in the experiment was a single utterance of the English vowel [a:] spoken on a monotone by a single speaker. A tape recording of this utterance was cut to give a stimulus lasting 450 msec. and this was re-recorded many times to provide the test tape. When the stimulus was relayed to subjects, controls were set so as to give a sound pressure level of approximately 60 dB at the subject's ears and all other stimuli were adjusted to give the same peak reading. . The first non-speech stimulus was a complex tone of the same fundamental frequency as the vowel, obtained by filtering a square-wave of this frequency. The sound included a long range of harmonics without any marked peaks in the spectrum so that no vowel-like formants were audible. This periodic sound had the same peak intensity as the vowel and was cut to' the same duration of 450 msec., but the perceptual impression was of a sound not resembling speech. The second non-speech stimulus was the recorded sound of a telephone bell, again adjusted to have the same peak intensity and duration as the vowel sound. The third was a click produced by a short burst of pure tone of 2210Hz, lasting 25 msec. and registering the same peak intensity as the other stimuli. For each of the four stimuli a test tape was made up bearing 30 identical sounds. The inter-stimulus interval varied throughout each test over a range from 2400 msec. to 3700 msec. and intervals of different length were distributed randomly through the test. The first five items in each test were treated as practice items and the times for these' were disregarded in the treatment of results. The subject was told that he would hear in the headphones a certain sound repeated 30 times and that each time he heard it he was to begin to say [a:] as soon as he possibly could after the sound began, being careful to avoid anticipating the arrival of the sound. It was impressed on subjects that they could utter 2 The peak voltmeter was designed and built by Stephen Bywaters, whose technical help is gratefully acknowledged.
Simple reaction times
357
the vowel on any comfortable level of pitch and loudness and that it did not matter what length they made their own vowel. Voice responses were recorded through a microphone held at a comfortable distance from the subject's lips by means of a light boom attached to the telephone headset through which the stimuli were delivered. When each stimulus condition was set up, the subject was given a short rehearsal in which he made his voice response to a few samples of the given stimulus; these trials were in addition to the five practice items in each test run. Throughout all tests, the stimuli and the subject's responses were recorded on a two-channel tape recorder so that response times could later be measured. In previous experiments and in preliminary runs no significant effect was detected of the order in which stimulus tapes were presented, but nonetheless effects of order were balanced out as far as possible in this experiment by using all orders of presentation of the four types of stimulus once each. Reaction-times were obtained for all four stimuli from 24 subjects, all females between the ages of 18 and 25 years.
RESULTS
Each subject made 25 responses to each stimulus, making 100 responses per subject and 2400 reaction-times in all, which were measured by means of an electronic counter. Since the range of reaction-times is limited at the lower end by a physiological minimum, the distribution tends to be positively skewed and statistical processing of the data was therefore based on the logarithm of reaction-time. The mean reaction-time for each stimulus is shown in Table I. Analysis of variance indicated a significant effect of change TABLE I
Stimulus
Mean reaction· time (msec.)
Speech (vowel)
223
Tone
217 274
Bell
Click
~26
of stimulus and t-tests showed that reaction-time to the click stimulus was significantly longer than that to the vowel, the bell and the tone (in all cases p < 0.001). The mean reaction-time for the tone was significantly shorter than that for the bell (p. < 0.01) but the difference between the means for the vowel and the tone only approached significance (p < 0.1
>
0.05).
On the basis of these results there is good reason to suppose that the long latency of the speech mechanism in response to a click stimulus is
358
D. B.Fry
largely due to its short duration since the other two non-speech stimuli, which were equal in duration to the vowel sound, gave means very close to that for the vowel. Since there was no significant difference between the m:~ans for the telephone bell and the vowel stimulus, these two stimuli were adhpted for a second experiment which repeated reaction-time measurements for all combinations of speech and non-speech stimulus with speech and nonspeech response modes. EXPERIMENT
II
This experiment employed a 2 X 2 factorial design with 2 stimuli and 2 response modes. The two stimuli were those already described: a natural utterance of the vowel [a:] with a duration of 450 msec. delivered to subjects bin aurally at a level of approximately 60 dB SPL, and a single ring of a telephone bell with the same duration and giving the same peak reading as the vowel stumulus. The stimuli were delivered from the same test tapes as in the previous experiment. The two response modes were key-pressing and a voice response. The key presented enough resistance to allow the subject to rest his finger lightly on it throughout a test run. Voice responses were recorded in the same conditions as before. During experimental runs both the succession of stimuli and the subject's responses were recorded on a two-channel tape recorder and the time interval between the onset of the stimulus and the beginning of the subject's response was subsequently measured by means of an electronic counter. The subject was told that he would hear in the telephones a certain sound repeated 30 times; at each occurrence he was to make his response as soon as he possibly could after the stimulus began. It was impressed on all subjects that they should not wait for a stimulus to end before responding, though they should take care not to anticipate the arrival of the sound. Subjects used one finger of the dominant hand in making the key-pressing response (in only one of the subjects was this the left hand). When making a voice response, subjects were told, as in the previous experiment, that they should say the vowel [a:] on any comfortable level of pitch and loudness and make it any length they liked. For each experimental condition a subject was given a short rehearsal in which he heard several stimuli and responded in the appropriate mode, and in addition each test run began with five practice items. Effects of order of presentation of stimuli and of response mode were again balanced out as far as possible by employing 24 subjects and using all possible orders of stimulus and response once each. The subjects, a different group from those used in Experiment I, were all female and between the ages of 18 and 25 years. RESULTS
Each subject experienced four experimental conditions: the bell stimulus with a key response, the speech stimulus with a key response, the bell stimulus with a voice response and the speech stimulus with a voice response. As before, the statistical treatment of the data was based on the logarithm of reaction-time. The mean for each condition is given in Table II.
Simple reaction times
359
TABLE II
Mean reaction-time (msec.)
Condition
167 180 201
Bell- key (BK) Speech - key (SK) Bell- voice (BV) Speech - voice (SV)
213
The mean reaction-times at once suggest a patterning of the effects since the key response provides the two shorter times and the voice response the two longer, while in each pair the bell stimulus produces shorter times than the speech stimulus. The shorter stimulus paired with the shorter response mode (condition BK) gives the shortest time and the combination of longer stimulus and longer response (condition SV) gives the longest reactiontime. In the crossed conditions, the longer response mode (voice) is responsible for a considerably greater increase in reaction-time than the longer stimulus (speech). A simple subtraction of the means shows that the difference between the bell and the speech stimulus is 13 msec. in one response mode and 12 msec. in the other; the difference between the key and the voice response is 34 msec. with one stimulus and 33 msec. with the other. This is a strong indication that there is no interaction between stimulus and response in this experiment and this is borne out by more detailed treatment of the distributions of reaction-time. Analysis of variance showed that the main effects of stimulus and of response mode were highly significant; in the case of the stimulus, F (1, 23) = 16.17, p < 0.001, and for the response mode, F (1,23) = 42.6, p < 0.001. There was no significant interaction of stimulus and response (F 0.3).
=
DISCUSSION
The speech tasks presented to the subjects in the second experiment, the perception and production of a vowel sound, represent minimal involvement of the speech processing and programming circuits and the results therefore give an indication of the order of the shortest reaction-times to be expected in experiments with speech. A response requiring the reception of speech is not likely to be made in less than about 180 msec. and where the response involves the action of the motor speech mechanism, an increase in reaction-time of not less than 30 msec. is probable. The absence of interaction between stimulus and response in this experiment is almost certainly associated with the very simple nature of the task. Previous work
360
D. B, Pry
has shown not only that reaction-time is affected by the complexity of a speech stimulus, but that the degree of right ear advantage in processing speech is also greater for a more complex stimulus (Fry, 1974). This suggests that the more complex the speech stimulus, the greater the involvement of the central speech processor, located in the left cerebral hemisphere in right-handed people. It is likely that the brain makes use of this central unit in both speech reception and speech production and it would be surprising therefore if no evidence of interaction were forthcoming in experiments employing more complex verbal tasks. Work is at present in progress to examine this question. SUMMARY
Two experiments were carried out involving the measurement of simple reaction-time when subjects responded to speech and to non-speech stimuli. In the first, subjects were required to make a speech response (uttering the vowel [a:]) to one speech stimulus (the vowel [a:]) and three non-speech stimuli (a complex tone, a telephone bell and a click). The click stimulus gave significantly longer reaction-times than the other three stimuli; since all stimuli were equated for peak intensity delivered to the subjects' ears, this was due to the short duration of the click (25 msec.). There was no evidence that compatibility between the speech stimulus and the speech response had any influence on reaction-time. The second experiment employed a 2 X 2 design with 2 stimuli and 2 response modes. The stimuli were the vowel [a:] and the telephone bell; the response modes were key-pressing and uttering the vowel [a:]. The speech stimulus and the speech response gave significantly longer reaction-times than the non-speech stimulus and response. The minimum time for a reaction requiring speech reception is of the order of 180 msec. and the use of the motor speech mechanism adds about 30 msec. to reaction-time. Again no interaction was found between stimulus and response and this is probably due to the extremely simple nature of the speech tasks imposed. REFERENCES
FRY, D. B. (1970) Reaction-time experiments in the study of speech processing, Proc. 6th Int. Congo Phon. Sci., Prague, 337-346. (1971) Time constants in speech, in Form and Substance, ed. by L. L. Hammerich, R. Jakobson and E. Zwirner, Copenhagen. (1974) Right ear advantage for speech presented monaurally, "Language and Speech," 17, 142-151.
Prof. D. B. Fry, Department of Phonetics and Linguistics, University College London, Gower Street, London, WClE 6BT.
In experiments previously reported (Fry, 1970, 1971) reaction-time was measured for very simple speech tasks involving, on the perceptual side, a choice between minimal pairs of English words and, on the motor side, the utterance of a vowel sound. It was found that choice reaction-time is affected by the complexity of the recognition task which subjects are asked to perform. Thus a decision between pit and bit heard in isolation, depending on the cue of voice onset time, takes less time than a decision between lay and ray, which depends on second and third formant configuration, while a choice between splay and spray, where the same distinction figures in a consonant cluster, takes very much longer still. These times were obtained with a key-pressing response so that no motor speech action was involved. The results indicate that the more complex the processing operation, the longer the time required by the central processing unit for speech. A second series of experiments aimed at estimating the minimum time required for speech reception and speech production by setting subjects the simplest possible speech tasks. Two types of stimulus were combined with two response modes. The.s.ttmuli were an isolated vowel [a:], and a short pulse of a pure tone, lasting 20 msec., which was heard as a click. The response modes were uttering the vowel [a:] or pressing a key. Each stimulus was paired with each response mode and simple reaction-times were measured in all four conditions. The results showed that subjects gave relatively long reaction-times when asked to make a speech response to the click and they reported that this stimulus seemed to have an inhibiting effect on the speech mechanism. The long reaction-times may therefore have been related not so much to the fact that a non-speech stimulus was being given as to some physical characteristic of the particular stimulus. Two experiments are reported in the present paper, the first of them I
Supported by a grant from the Science Research Council.
Cortex (1975) 11, 355-360.
356
D. B. Fry
designed to find a suitable non-speech stimulus for use in contrast with a speech stimulus, and the second constituting a repetition of the previous experiment with speech and non-speech stimuli and responses.
EXPERIMENT
I
Simple reaction-time was measured when subjects responded by uttering [a:] to each of four stimuli, one speech and three non-speech. The purpose of the experiment was to determine whether a click has a specific effect in delaying a subject's response or whether comparable reaction-times would be obtained when other kinds of non-speech stimulus were used. The duration and intensity of an auditory stimulus are well known to influence reaction-time. A click is necessarily of short duration and it is difficult to ensure that its intensity shall be comparable with that of other stimuli used. In order that the intensity of all stimuli should be as closely controlled as possible, a rapidly reading peak voltmeter was designed, using a light emitting diode display. The response of this meter is such that one cycle of a 1000 Hz tone applied to the device gives the same reading as a continuous tone of equal frequency and leveP The stimuli provided for subjects were all derived from tape recordings and the peak-reading meter was used both in recording and in replaying the stimuli, which were relayed to subjects binaurally through calibrated headphones. The speech stimulus used in the experiment was a single utterance of the English vowel [a:] spoken on a monotone by a single speaker. A tape recording of this utterance was cut to give a stimulus lasting 450 msec. and this was re-recorded many times to provide the test tape. When the stimulus was relayed to subjects, controls were set so as to give a sound pressure level of approximately 60 dB at the subject's ears and all other stimuli were adjusted to give the same peak reading. . The first non-speech stimulus was a complex tone of the same fundamental frequency as the vowel, obtained by filtering a square-wave of this frequency. The sound included a long range of harmonics without any marked peaks in the spectrum so that no vowel-like formants were audible. This periodic sound had the same peak intensity as the vowel and was cut to' the same duration of 450 msec., but the perceptual impression was of a sound not resembling speech. The second non-speech stimulus was the recorded sound of a telephone bell, again adjusted to have the same peak intensity and duration as the vowel sound. The third was a click produced by a short burst of pure tone of 2210Hz, lasting 25 msec. and registering the same peak intensity as the other stimuli. For each of the four stimuli a test tape was made up bearing 30 identical sounds. The inter-stimulus interval varied throughout each test over a range from 2400 msec. to 3700 msec. and intervals of different length were distributed randomly through the test. The first five items in each test were treated as practice items and the times for these' were disregarded in the treatment of results. The subject was told that he would hear in the headphones a certain sound repeated 30 times and that each time he heard it he was to begin to say [a:] as soon as he possibly could after the sound began, being careful to avoid anticipating the arrival of the sound. It was impressed on subjects that they could utter 2 The peak voltmeter was designed and built by Stephen Bywaters, whose technical help is gratefully acknowledged.
Simple reaction times
357
the vowel on any comfortable level of pitch and loudness and that it did not matter what length they made their own vowel. Voice responses were recorded through a microphone held at a comfortable distance from the subject's lips by means of a light boom attached to the telephone headset through which the stimuli were delivered. When each stimulus condition was set up, the subject was given a short rehearsal in which he made his voice response to a few samples of the given stimulus; these trials were in addition to the five practice items in each test run. Throughout all tests, the stimuli and the subject's responses were recorded on a two-channel tape recorder so that response times could later be measured. In previous experiments and in preliminary runs no significant effect was detected of the order in which stimulus tapes were presented, but nonetheless effects of order were balanced out as far as possible in this experiment by using all orders of presentation of the four types of stimulus once each. Reaction-times were obtained for all four stimuli from 24 subjects, all females between the ages of 18 and 25 years.
RESULTS
Each subject made 25 responses to each stimulus, making 100 responses per subject and 2400 reaction-times in all, which were measured by means of an electronic counter. Since the range of reaction-times is limited at the lower end by a physiological minimum, the distribution tends to be positively skewed and statistical processing of the data was therefore based on the logarithm of reaction-time. The mean reaction-time for each stimulus is shown in Table I. Analysis of variance indicated a significant effect of change TABLE I
Stimulus
Mean reaction· time (msec.)
Speech (vowel)
223
Tone
217 274
Bell
Click
~26
of stimulus and t-tests showed that reaction-time to the click stimulus was significantly longer than that to the vowel, the bell and the tone (in all cases p < 0.001). The mean reaction-time for the tone was significantly shorter than that for the bell (p. < 0.01) but the difference between the means for the vowel and the tone only approached significance (p < 0.1
>
0.05).
On the basis of these results there is good reason to suppose that the long latency of the speech mechanism in response to a click stimulus is
358
D. B.Fry
largely due to its short duration since the other two non-speech stimuli, which were equal in duration to the vowel sound, gave means very close to that for the vowel. Since there was no significant difference between the m:~ans for the telephone bell and the vowel stimulus, these two stimuli were adhpted for a second experiment which repeated reaction-time measurements for all combinations of speech and non-speech stimulus with speech and nonspeech response modes. EXPERIMENT
II
This experiment employed a 2 X 2 factorial design with 2 stimuli and 2 response modes. The two stimuli were those already described: a natural utterance of the vowel [a:] with a duration of 450 msec. delivered to subjects bin aurally at a level of approximately 60 dB SPL, and a single ring of a telephone bell with the same duration and giving the same peak reading as the vowel stumulus. The stimuli were delivered from the same test tapes as in the previous experiment. The two response modes were key-pressing and a voice response. The key presented enough resistance to allow the subject to rest his finger lightly on it throughout a test run. Voice responses were recorded in the same conditions as before. During experimental runs both the succession of stimuli and the subject's responses were recorded on a two-channel tape recorder and the time interval between the onset of the stimulus and the beginning of the subject's response was subsequently measured by means of an electronic counter. The subject was told that he would hear in the telephones a certain sound repeated 30 times; at each occurrence he was to make his response as soon as he possibly could after the stimulus began. It was impressed on all subjects that they should not wait for a stimulus to end before responding, though they should take care not to anticipate the arrival of the sound. Subjects used one finger of the dominant hand in making the key-pressing response (in only one of the subjects was this the left hand). When making a voice response, subjects were told, as in the previous experiment, that they should say the vowel [a:] on any comfortable level of pitch and loudness and make it any length they liked. For each experimental condition a subject was given a short rehearsal in which he heard several stimuli and responded in the appropriate mode, and in addition each test run began with five practice items. Effects of order of presentation of stimuli and of response mode were again balanced out as far as possible by employing 24 subjects and using all possible orders of stimulus and response once each. The subjects, a different group from those used in Experiment I, were all female and between the ages of 18 and 25 years. RESULTS
Each subject experienced four experimental conditions: the bell stimulus with a key response, the speech stimulus with a key response, the bell stimulus with a voice response and the speech stimulus with a voice response. As before, the statistical treatment of the data was based on the logarithm of reaction-time. The mean for each condition is given in Table II.
Simple reaction times
359
TABLE II
Mean reaction-time (msec.)
Condition
167 180 201
Bell- key (BK) Speech - key (SK) Bell- voice (BV) Speech - voice (SV)
213
The mean reaction-times at once suggest a patterning of the effects since the key response provides the two shorter times and the voice response the two longer, while in each pair the bell stimulus produces shorter times than the speech stimulus. The shorter stimulus paired with the shorter response mode (condition BK) gives the shortest time and the combination of longer stimulus and longer response (condition SV) gives the longest reactiontime. In the crossed conditions, the longer response mode (voice) is responsible for a considerably greater increase in reaction-time than the longer stimulus (speech). A simple subtraction of the means shows that the difference between the bell and the speech stimulus is 13 msec. in one response mode and 12 msec. in the other; the difference between the key and the voice response is 34 msec. with one stimulus and 33 msec. with the other. This is a strong indication that there is no interaction between stimulus and response in this experiment and this is borne out by more detailed treatment of the distributions of reaction-time. Analysis of variance showed that the main effects of stimulus and of response mode were highly significant; in the case of the stimulus, F (1, 23) = 16.17, p < 0.001, and for the response mode, F (1,23) = 42.6, p < 0.001. There was no significant interaction of stimulus and response (F 0.3).
=
DISCUSSION
The speech tasks presented to the subjects in the second experiment, the perception and production of a vowel sound, represent minimal involvement of the speech processing and programming circuits and the results therefore give an indication of the order of the shortest reaction-times to be expected in experiments with speech. A response requiring the reception of speech is not likely to be made in less than about 180 msec. and where the response involves the action of the motor speech mechanism, an increase in reaction-time of not less than 30 msec. is probable. The absence of interaction between stimulus and response in this experiment is almost certainly associated with the very simple nature of the task. Previous work
360
D. B, Pry
has shown not only that reaction-time is affected by the complexity of a speech stimulus, but that the degree of right ear advantage in processing speech is also greater for a more complex stimulus (Fry, 1974). This suggests that the more complex the speech stimulus, the greater the involvement of the central speech processor, located in the left cerebral hemisphere in right-handed people. It is likely that the brain makes use of this central unit in both speech reception and speech production and it would be surprising therefore if no evidence of interaction were forthcoming in experiments employing more complex verbal tasks. Work is at present in progress to examine this question. SUMMARY
Two experiments were carried out involving the measurement of simple reaction-time when subjects responded to speech and to non-speech stimuli. In the first, subjects were required to make a speech response (uttering the vowel [a:]) to one speech stimulus (the vowel [a:]) and three non-speech stimuli (a complex tone, a telephone bell and a click). The click stimulus gave significantly longer reaction-times than the other three stimuli; since all stimuli were equated for peak intensity delivered to the subjects' ears, this was due to the short duration of the click (25 msec.). There was no evidence that compatibility between the speech stimulus and the speech response had any influence on reaction-time. The second experiment employed a 2 X 2 design with 2 stimuli and 2 response modes. The stimuli were the vowel [a:] and the telephone bell; the response modes were key-pressing and uttering the vowel [a:]. The speech stimulus and the speech response gave significantly longer reaction-times than the non-speech stimulus and response. The minimum time for a reaction requiring speech reception is of the order of 180 msec. and the use of the motor speech mechanism adds about 30 msec. to reaction-time. Again no interaction was found between stimulus and response and this is probably due to the extremely simple nature of the speech tasks imposed. REFERENCES
FRY, D. B. (1970) Reaction-time experiments in the study of speech processing, Proc. 6th Int. Congo Phon. Sci., Prague, 337-346. (1971) Time constants in speech, in Form and Substance, ed. by L. L. Hammerich, R. Jakobson and E. Zwirner, Copenhagen. (1974) Right ear advantage for speech presented monaurally, "Language and Speech," 17, 142-151.
Prof. D. B. Fry, Department of Phonetics and Linguistics, University College London, Gower Street, London, WClE 6BT.
