Journal o f Psycholinguistic Research, Vol. 1, No. 2, 1972
Cross-Linguistic Pause and Rate Phenomena in Adults and Adolescents Daniel C. O'Connell 1 and Sabine Kowal 1
Received September 29, 1971
Three groups o f 40 Ss /German adolescents and American adults and adolescents) read two passages and retold them. In confirmation o f O'Connell, Kowal, and H(~rmann (1969) for German adults, a number o f pause and rate measures were significantly different for semantically ordinary or unusual passages. Comparisons among the four experiments manifested different patterns o f pauses and rate across the two languages and age brackets.
T h e research of O'Connell, Kowal, and H6rmann (1969) presented evidence that variations in semantic context can produce changes in the number and length of unfilled pauses in a situation in which syntactic and other variations are minimized. Their Ss were adult native speakers of German. Each S read two paragraphs aloud and after each paragraph retold the story without further instructions. Each paragraph consisted of five sentences in German, each containing 23 syllables. The third sentence was either in accord with the story or an unusual occurrence (depending on a simple exchange of subject and object). Both the number and length of unfdled pauses were greater in the unusual stories as compared with the usual ones. In the readings the effect was limited to the critical sentence and the pauses immediately thereafter. Support for the research reported here has come from two grants-in-aid of research from the Society of the Sigma Xi. 1St. Louis University, St. Louis, Missouri. 155 ~) 1972 Plenum Publishing Corporation, 227 West 17th Street, New York, N.Y. 10011.
O'Connell and Kowal
In order to investigate the generality of the influence of semantic context on pause phenomena, the present research was planned as three replications. The first made use of adolescents, all native German speakers, of ages 12 to 14. The second and the third made use of adult and adolescent native American Ss respectively and translated stimulus materials which retained the same number o f syllables in each sentence as in the original German. In view of the slow development of semantic differentiation relative to the growth of grammar in children at an earlier stage (McNeill, 1970), it is plausible that, even in adolescents, some semantic insensitivity, relative to adult performance, still persists. Hence, it was hypothesized in addition that differences due to the semantically unusual condition would be diminished in the adolescent groups.
METHOD A total of 40 men and women volunteers served as Ss in each experiment. In Exp. I, they were adolescents of ages 11 to 13 at two Gymr~sien in West Berlin. In Exp. ]I, they were adults at St. Louis University. In b o t h experiments, the number o f men and women in each group was roughly the same, but there was no effort to investigate sex differences, since there appeared no evidence in the data of O'Connell, Kowal, and H6rmann (1969) to warrant the hypothesis of such differences. In Exp. III, Ss were adolescents at St. Louis University High School. In this last case, only boys of ages 13 to 15 were Ss, a circumstance necessitated by the unavailability of a mixed group of younger Ss. In all instances, a native speaker (German or English) was the E, and Ss were run individually. Each was seated at a table before a microphone, with E seated somewhat to the side and out of the S's line of vision. The following instructions were read aloud in German (Exp. I) or English (Exp. II and Exp. III): In the following experiment we wish to investigate the influence on behavior of reading aloud. You will receive two cards, on each of which a little story appears. Please read the story aloud. Then put the card face down on the table and repeat aloud the story which you have just read. Since we will be recording what you say, we ask you to refrain from asking questions during the experiment. Then S was asked to begin. The experimental sequence was always the following: first reading (R 1), first story (S 1), second reading ( I ~ ) , and second story (82). The two readings are given here in their normal (N) versions with the abnormal (A) critical sentences in parentheses:
Cross-Linguistic Pause and Rate Phenomena in Adults and Adolescents
The Student Story (S). He had already been disturbed for the whole class period by the tiny white paperwads. A student by the window seemed to be producing them and tossing them into the front rows. The young teacher contemplated for a moment, whether or not he should punish the student. (The student contemplated for a moment whether or not he should punish the young teacher.) In his dismay, he breathed a sigh of relief as the period f-many came to an end. Perhaps he ought to discuss the matter seriously with the student during the recess. The Child Story (C). In the late evening, a fight pick-up truck drove at high speed down the highway into the valley. Suddenly the driver realized to his utter horror that the brakes were not functioning. The pick-up truck ran over a youngster, as he was crossing the highway at the same moment. (A youngster ran over the pick-up truck, as he was crossing the highway at the same moment.) Only with great effort did the driver succeed in keeping the skidding truck on the highway. With brakes screeching, he was finally able to bring it to a halt at the foot of the hill. In terms of the stimulus materials (R 1 and R2), therefore, there were eight sequences: N c A s, N c N s, A c A s, A c Ns, N s A c , N s N c , A s A c , and A s N c . In each experiment five Ss received each experimental sequence. The readings and the corresponding stories o f each S were tape recorded by means of an Uher Royal tape recorder. Literal transcriptions for readings and stories were typed for all Ss. Tabulation o f results was as in O'Connell, Kowal, and H6rmann (1969) with the exception that the data o f Exp. II and Exp. III were transferred to a time record o f the acoustic energy by means o f a Briiel and Kjaer Audio Frequency Spectrometer and Level Recorder. For Exp. I, the time record was made by means o f a Schwarzer Physioscript (Type PST 6 S 50/1) recording machine. A cutoff point o f 250 msec was adopted as minimum length for unfilled pauses, in keeping with Goldman-Eisler (1968). These intervals were measured from the time records and were not dependent on estimates o f judges. Both f'filed and unfilled pauses (FPs and UPs) were defined as in Maclay and Osgood (1959). Departure o f stories from the original semantic content was also noted. RESULTS Analyses of variance were run for response measures within each o f the experiments: mean length of UPs after the c~tical (third) sentence in R I and R 2, mean time between the completion o f R~ and the beginning of S~, and mean syl./sec in S 2. In general, the differences due to the main factor of N and A versions were not as striking as in the O'Connell, Kowal, and H6rmann (1969) study. Nonetheless, in each experiment, one or another response measure indicated a significant effect o f the AN variable, and only these analyses are reported below.
O'Connell and Kowal
Comparisons Within Experiments In Exp. I, mean length of UPs after the'critical (third) sentence in R 2 was significantly shorter in N (369 reset) than in A (741 reset), F (1, 32) = 11.99, P < 0.005. No other F tests were significant. The analysis is summarized in Table I and the means and standard deviations are presented in Table II. In Exp. II, an analysis of the mean time between completion of R~ and the beginning of S 1 yielded a significant AN • SC interaction, F ( 1 , 3 6 ) = Table I. Analyses of Variance of Length o f UPs After Sentence 3 in R2 in Exp. I and Syl./sec in $2 in Exp. II II
Exp. I, length of UPs after sentence 3 in R 2 Source of variation AN SC A_N_ AN X SC AN X A N SC X A N _ A N X SC X A _ N
1 1 1 1 1 1
142.51 12.88 7.83 0.13 8.74 3.54
142.51 12.88 7.83 0.13 8.74 3.54
Exp. II, syl./sec in S2 Sum of
11.987 b 1.0184 0.659 0.011 0.735 0.298
0.32 0.73 0.26 0.23 4.10 0.44
0.32 0.73 0.26 0.23 4.10 0.44
0.474 1.066 0.374 0.329 5.988 a 0.645
ap < 0.025.
be < 0.005. Table II. Means (X) and Standard Deviations (S.D.) for Length o f UPs After Sentences 3 in R2 in Exp. I and Syl./sec in $2 in Exp. II III
Condition R2 and S2
Exp. I, length of UPs after sentence 3 in R 2
Exp. II, syl./sec in S2
Cross-Linguistic Pause and Rate Phenomena in Adul~ts and Adolescents
4.46, P < 0.05. T h e analysis is summarized in Table III and the means and standard deviations are presented in Table IV'. The analysis o f mean syl./sec in S2 yielded a significant AN • A_N_ (A N indicates: following upon an R 1 and S I in the A or N version) interaction, F (I, 32) = 5,99, P < 0.025. The analysis is summarized in Table I and the means and standard deviations are presented in Table II. In Exp. III, mean length o f UPs after the critical sentence in R~ was significantly shorter in N (622 msec) than in A (841 msec), F ( 1 , 3 6 ) = 10.89, P < 0.005. Finally, an analysis of the mean time between completion of R~
Table III. Analyses of Variance of Time Between R 1 and S 1 in Exp. II, Length of UPs After Sentence 3 in R 1 and Time Between RI and $1 in Exp. Ill I
of Sum of variation df squares AN SC ANXSC Error
Exp. II, time between R 1 and S1 Variante
Exp. III, length of UPs after sentence 3 in R 1 Sum of squares
Exp. III, time between R 1 and S1 Sum of Varisquares ance
47.961 47.961 10.888 b 60.762 60.762 8.119 b 1 3.192 3.192 0.071 1 1.122 1.122 0.025 6.400 6.400 1.453 9.120 9.120 1.219 1 202.050 202.050 4.465 a 10.816 10.816 2.455 1.560 1.560 0.208 36 1625.075 45.252 158~574 4.405 269.427 7.484 I
IIr FII[~ i
ap 13.2) UPs for adolescents, t (158) = 3.93,P < 0.002, than for adults. A similar t test between mean number of UPs in readings for Germans and Americans (15.1 > 14.6) was nonsignificant, t (158) = 0.57. An overall two-tailed t test comparing the mean length of UPs in readings (R 1 and R 2 combined) by adolescents and adults (559 msec < 727 msec) was short of significance, t (158) = 1.76, 0.10 > P >0.05. A one-tailed option would have indicated significance at the 0.05 level. The corresponding t test between mean length of UPs in readings for Germans and Americans (610 msec < 677 msec) was nonsignificant, t (158) = 0.73. 2. Pauses in Stories. In the stories,, the number of UPs varied less systematically from condition A to N than in the readings. In S~, however, the number of FPs was greater in condition A than N for Germans (90 > 68), but greater in N than A for Americans (87 > 70). A conservative test comparing the number of Germans and Americans with and without FPs in S, for version A alone is just short of significance: 9(2 = 3.66, 0.10 > P > 0.05. The mean length of UPs in the stories was a more reliable measure; in all the experiments, UPs of condition A were longer than those of condition N. The effects over all stories (S 1 and S2 combined) are included in Fig. 2. They reflect the reliability across experiments of the comparisons within experiments already reported. However, the overall two-tailed t tests comparing mean length of UPs in stories (S 1 and S2 combined) for adolescents and adults (851 < 1143) and Germans and Americans (935 < 1059), respectively, were nonsignificant: t (158) = 1.35 and t (158) = 1.54, 0.15 > P > 0.10. 3. Semantic Content of Stories. Finally, the sensitivity of Ss to the A condition, as reflected in their reproduction of the same semantic content in the story, differed from experiment to experiment. In each of the adult experiments 23 out of 40 A readings were reproduced as A in the corresponding story; in the German and American adolescents, 7 and 18 were reproduced as A respectively. Comparison of the German adolescents and adults for number of correct reproductions of version A shows significance: 9(~ = 8.53, P < 0.01. A similar comparison of German and American adolescents also shows significance: X 2 = 4.84, P < 0.05.
The comparisons between experiments as well as those within experiments indidate that an unusual turn of events noted in the reading of a story literally gives people pause. The consistent finding that the unusual version yields more frequent and longer UPs confirms the findings of O'ConneU, Kowal, and H6rmann (1969) and lends support to their hypotheses regarding
O'Connell and Kowal
the cognitive function of pauses. The generality of these effects across two languages and two age brackets as well as their extraordinary reliability, as indicated in the figures in the present research, all point to the usefulness of temporal response measures in the analysis of oral language behavior. To date such measures have been largely neglected or relegated to the domain of clinical investigations. Their sensitivity to very slight semantic variations shows promise for further use to uncover cognitive aspects of language behavior. In a more theoretical context, however, we are very far from a general theory of pauses as indicants of cognitive processes. Attempts that have been made thus far to interpret consistent findings for this response measure are closely bound to the specific aspects of each experiment. One fact seems evident: Pauses are functional for speaker and hearer and they serve a variety of functions. Concretely this means that the specifics of each experiment must be considered: Subtle changes in design, artifacts and distortions in measurement and data analyses, and imprecise or inconsistent terminology regarding hesitation, pause, and rate phenomena from one E to another have obfuscated the issue greatly. Goldman-Eisler's (1968) experiments are prime examples of some of these complexities, as Boomer (1970) has recently pointed out. For that reason, despite the acuteness of her insights regarding the cognitive functions of pauses, Goldman-Eisler's formulations must be considered at this time prototheories. In general, any theory about the function of pauses (unfilled and filled) should be devised so as to account for and integrate the following considerations: Unfilled pauses differ in terms of location, number, and length in different types of speech varying from rote reading to completely spontaneous speech (e.g., Levin, Silverman, and Ford, 1967; Martin and Strange, 1968; O'Connell, Kowal, and H6rmann, 1969). UnFilled and f'filed pauses seem to be sensitive to cognitive processes preceding them, as reflected in length and number of pauses after the abnormal sentence in the present study; they also seem to reflect anticipatory cognitive processes, as shown in longer pauses occurring before words with low transition probability (e.g., Maclay and Osgood, I959; Goldman-Eisler, 1968). Unfilled pauses may serve a syntactic function to the extent that they help to structure strings of words perceptually in accord with the underlying grammatical structure. They are used for this purpose in reading by adults with little interindividual variability (O'Connell, Kowal, and H6rmann, 1969). That this is a learned aspect of the function of pauses is indicated by the larger number of "ungrammatical" pauses in the reading of adolescents in the present study. Unfilled pauses may also serve a semantic function in that they indicate the active search for single words, recently studied by means of the cloze procedure (e.g., Cook, 1969; Goldman-Eisler, 1968). A possible inter-
Cross-Lingui~c Pause and Rate Phenomena in Adults and Adolescents
action of both syntactic and semantic aspects of pauses could be studied in an experiment dealing with ambiguous sentences. As has been pointed out in the present study, filled and unfilled pauses are sensitive to age difference. So far there is no other experimental evidence supporting this result. On the contrary, Levin, Silverman, and Ford (1967), in a study on hesitations in children's speech of four different age levels found that there were no variations in their data "attributable to the different ages of the Ss." [p. 563] At the same time they acknowledge, however, that these results may be due to the small number of Ss used in their study. One of our hypotheses related to age differences concerned a quite different response measure: semantic sensitivity as shown in the correct reproduction of version A. The intermediate level of American adolescents-superior to the German adolescents, but inferior to American and German adults-may reflect the fact that the American adolescents were two years older on the average than the Germans, A selection of Ss which began as a reluctant necessity turned out to differentiate the groups in complete accord with the hypothesis of increasing semantic sensitivity with age. The present study also yielded similarities-and some suggestions of differences-in filled and unfilled pauses due to language. Comparable crosscultural data have not yet been provided. Ramsay (1968) has indeed stated on the basis of rather limited evidence that "in general speech patterns of Dutch subjects are similar to those of English-speaking subjects." [p. 54] In a study on foreign accent, Black et al. (1966) had native Hindi, Spanish, and Japanese speakers read a passage in English and in their respective mother tongue and found that "the pauses used in native languages and English did not differ significantly." [p. 240] And finally, a cross-cultural study of speech rate in Japanese and Americans (Osser and Peng, 1964) "fated to reveal significant differences in speech-rate between the two groups." [p. 124] The authors suggest t h e possibility that further cross-linguistic comparisons may prove speech rate within a certain range to be another language universal. Nevertheless, the present research suggests that there may also be differences across languages as well as commonalities, at least with respect to pause phenomena. In fact, Osser and Peng (1964) acknowledge that "judgement of speech rate, when a foreign language is being spoken, may depend on significant differences between languages in the way in which they allow speakers to distribute their pauses." [p. 124] To return finally to the reading process, developmental psycholinguistics has had little to say from a theoretical point of view. We must agree with Athey (1971) that, for the time being, we need partial models to tell us "how language functions in reading for children of different ages, from different backgrounds, in different social and educational contexts." [p. 9]
O'Connell and Kowal
ACKNOWLEDGMENTS The authors wish to acknowledge the assistance of Tracy Connell and Jill Szawara in the collection of data.
REFERENCES Athey, I. (1971). Language models and reading. Paper presented at the meeting of the American Psychological Association, Washington, D.C., September 1971. Black, J. W., Tosi, O., Singh, S., and Takefuta, Y. (1966). A study of pauses in oral reading of one's native language and English. Lang. Speech 9, 237-241. Boomer, D. S. (1970). Review of F. Goldman-Eisler, Psycholinffuisties: Experiments in Spontaneous Speech. Lingua 25, 152-164. Cook, M. (1969). Transition probabilities and the incidence of idled pauses. Psychonomic ScL I, 191-192. Goldman-Eisler, F. (1968). Psycholinguistics: Experiments in Spontaneous Speech. Academic Press, London. Levin, H., Silverman, I., and Ford, B. I. (1967). Hesitations in children's speech during explanation and description. J. Verb. Learn. Verb. Behav. 6,560-564. Maclay, H., and Osgood, C. E. (1959). Hesitation phenomena in spontaneous English speech. Word 15, 19-44. Martin, J., and Strange, W. (1968). Determinants of hesitations in spontaneous speech. J. Exper. Psych. 76,474-479. Mdqeill, D. (1970). The Acquisition of Language: The Study of Developmental Psycholinguistics. Harper & Row, New York. O'ConneU, D. C., Kowal, S., and H6"rmann, H. (1969). Semantic determinants of pauses. Psychologische Forschung 33, 50-67. Osser, H., and Peng, F. (1964). A cross-cultural study of speech rate. Lang. Speech 7, 120-125. Ramsay, R. W. (1968). Speech patterns and personality. Lang. Speech 11 54-63.