Information Distribution, Interdependence, and Activity Levels Donald P. Hayes; Leo Meltzer; Signe Lundberg Sociometry, Vol. 31, No. 2. (Jun., 1968), pp. 162-179. Stable URL: http://links.jstor.org/sici?sici=0038-0431%28196806%2931%3A2%3C162%3AIDIAAL%3E2.0.CO%3B2-Y Sociometry is currently published by American Sociological Association.

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/about/terms.html. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/journals/asa.html. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.

The JSTOR Archive is a trusted digital repository providing for long-term preservation and access to leading academic journals and scholarly literature from around the world. The Archive is supported by libraries, scholarly societies, publishers, and foundations. It is an initiative of JSTOR, a not-for-profit organization with a mission to help the scholarly community take advantage of advances in technology. For more information regarding JSTOR, please contact [email protected].

http://www.jstor.org Mon Jul 16 17:28:20 2007

Information Distribution, Interdependence, and Activity Levels* DONALD P. HAYES, LEO MELTZER,

AND

SIGNE LUNDBERG

Cornell University

Pairs of subjects instructed a silent third person for 40 minutes i n the building o f a tinkertoy structure. One of three fornzs of the structure's diagram (complete, partial, or ambiguous information) was presented, i n a factorial design, t o each member of 60 dyads. Effects on ego's vocal activity of ( a ) the information given t o ego, ( b ) the information given t o his partner, and ( c ) the interaction of ego's and partner's information were assessed. I n virtually all conditions partner's information significantly affected the amount ego talked; i n only some conditions did ego's information influence his own activity; and i n no condition was there a significant effect on ego of the interaction between ego's and partner's information. These effects varied b y t y p e of dependency (induced b y the distribution of information), b y type of subject (conscriptee or volunteer), and over time. The sheer amount of activity, particularly vocal activity, has proven to be a useful indicator in many investigations (Bales, 1953; Bales, 1958; Bass, 1949; Bass, 1960; Bavelas et al., 1965; Caudill, 1958; Chapple and Vaughn, 1944; Lott and Lott, 1961; Riecken, 1958). However, its determinants cannot be said to be well understood. For example, some studies suggest that the more active persons are the better informed, while other studies suggest that activity increases as the ambiguity of the information available increases. I n one group of studies, stemming primarily from the work of Bales and his associates, a direct association between one's total verbal activity and the perceived quality of his contribution (as rated by others in the situation) has been found (Bales, 1958; Bass, 1949; Norfleet, 1948). One possible interpretation is that the amount of talk is directly related to the quality and quantity of the speaker's information. This interpretation has yet to be checked experimentally. I t is quite possible that talkative persons talk for reasons other than the possession of special knowledge (Riecken, 1958). Other research indicates that ambiguous rather than good information may

* Some of the results described here were presented in a paper read a t the annual meeting of the American Psychological Association, September, 1967. The research was supported by Public Health Service Grant MH 11829-01 and by National Science Foundation Grant NSF GS-920. The authors wish to acknowledge the suggestions received in discussions with R. Darlington, S. Jones, W. W. Lambert, J. M. Roberts, S. Searle, and R. M. Williams, Jr. 162

INFORMATION DISTRIBUTION

163

increase the amount of communication. According to Allport and Postman (1947, chapters 1-2), the amount of rumor transmission is directly related to both the importance and the ambiguity of (uncertainty about) the issue. Many authors theorize that needs for consensual validation (Sullivan, 1953) or for cognitive consistency (well reviewed in Feldman, 1966) often lead to ambiguity-reducing communications. The direct relationship between ambiguity and activity may also be due to the linguistic difficulties encountered when either party attempts to describe an ill-defined subject to another (Glanzer and Clark, 1964; Krauss and Weinheimer, 1966; Moscovici, 1967). The implication of these diverse views is that poor information may, a t times, generate more communication than good information does. I t is not clear whether the inmplications stemming from these two sets of studies are contradictory, since the conditions under which their results hold have been insufficiently specified. In particular it would seem necessary to specify the level of information possessed by relevant others. For example, a man with ambiguous information might talk a great deal if paired with another person similarly informed, but he might become quite inactive if paired with a very well informed person. T o clarify these speculations, we conducted an experiment in which the dependent variable was vocal activity in dyads and the two main independent variables were: (a) amount of information possessed by a person (ego); and (b) amount of information possessed by ego's partner (alter). PROCEDURE

SUBJECTS.TWOseries of experimental sessions were conducted, separated by a period of two months. In each, 60 subjects were paired a t random. The first series made use of students required by their introductory sociology class to participate in a study (hereafter: conscriptees) while the second series made use of volunteers recruited from sociology and social psychology classes (hereafter: volunteers) who were each paid $2.00. TASKAND INSTRUCTIONS. Each pair of subjects was given the task of jointly instructing a third person (hereafter: builder)-a trained, paid participant-in how to assemble a complex structure. This type of task was utilized to minimize the effects of special knowledge or recent experience. Subjects were unable to examine one another's diagram because they sat approximately 8 feet apart separated by low screens. The builder had all the pieces before him, in full view of the subjects, but he had no diagram. Subjects read the following instructions: You have the task of building a tinkertoy model in the shortest possible time. You will be given only 40 minutes for completing this task so you must work quickly and efficiently. The aim is to get the builder to complete as much of the structure as possible in the time available. On the table in front of you is your diagram. Using it, you should tell the builder exactly how to assemble the model. You may not do any of the actual

164

SOCIOMETRY

building yourself. The builder does not have any plans and must not ask any questions. He is not allowed to communicate with the instructors in any way. Please keep your diagram on the table and stay seated until the task is completed and Mrs. Lundberg signals the end of the experiment. The model you are to build is labeled Carnival Loop. Remember that you have only 40 minutes to complete the model so you must work efficiently.

EXPERIMENTAL TRFATMENTS. Variations in the experimental treatments were produced by providing each subject with one of the following forms of the diagram: ( a ) The Total-Clear (TC) diagram was an 8 x 10 inch color reproduction (Fig. 1, upper left). (b) The Partial-Clear (PC) diagram, was one-half of the same large color reproduction cut in such a way that the most difficult portions of the structure were divided equally. When both partners received a PC treatment, one received the diagram shown a t the lower left of Fig. 1 while the other received the one at the lower right. In all other treatments involving PC, the diagram used was the one a t the lower right. (c) The Total-Ambiguous (TA) diagram (Fig. 1, upper right) was a 2 x 4 inch, black and white reproduction of the original with a heavy overlay of blue crayon. I t was still possible to read it but only with difficulty. I n each of the two experimental series, 10 subjects (5 pairs) were assigned to each of the following treatments: (a) both subjects had TC; (b) both subjects had TA; (c) both subjects had PC; (d) one subject had TC, one had TA; (e) one subject had TC, one had PC; (f) one subject had TA, one had PC. For most purposes, a subject was classified both ANALYTICTREATMENTS. by his own information and by his partner's information. Thus the 120 subjects could be classified as follows: Information Treatment1 Subject's Partner's TC-TC TC-PC TC-TA PC-TC PC-PC PC-TA TA-TC TA-PC TA-TA

Number of Subjects Conscriptees 10 5 5 5 10 5 5 5 10

Volunteers 10 5 5 5 10 5 5 5 10

I The abbreviations presented here will be used throughout the paper. The letters to the left of the hyphen refer to ego's information treatment; the letters to the right of the hyphen refer to ego's partner's information treatment. When not hyphenated, the symbol (TC, TA, or PC) will refer to either ego or his partner, depending upon context. These

INFORMATION DISTRIBUTION FIGURE 1

Diagrams Used i n the Experimental Treatments

Total Ambiguous Total Clear (In color- 114 scale)

Partial Clear

(In color-1/4 sca.le)

(In ~ e r o x1/3-scale blue crayon overlay)

Partial Clear

(In color-114 scale)

symbols refer to the independent variables, which are applied to both ego and his partner. The dependent variable (described elsewhere) is always a measure of ego's vocal activity.

166

SOCIOMETRY

REDUCTION OF N FOR ANALYTIC PURPOSES. AS noted immediately above, the N's in the 9 analytic treatments are unequal. The investigation required a number of analyses of variance (hereafter: ANOVAs), but computer programs for ANOVAs with unequal cells were not locally available. Therefore, in any computation involving the cells with I0 cases, the sums of scores, the sums of squares, and the N's were simply h a l ~ e d .All ~ tables report the analytically reduced N's rather than the N's actually run in the experimental sessions. DEPENDENT VARIABLE.Vocal activity, the dependent variable, was measured automatically by means of bone-conductance microphones (Hayes and Meltzer, 1967) connected to a seven-channel analog-to-digital converter and a computer tape recorder (Hayes and Meltzer, 1966). The score used for each subject was his own percent-of-total-time-in-talk (hereafter: activity). If both persons talked simultaneously, each person's own talking time was credited as if'he alone were speaking. Thus the theoretical range of percentof-talk was 0 to 100 for each subject. Pauses within speech of 0.5 second or less were counted as speech. Brief utterances of 0.5 second or less were counted as silence. ANALYSIS 1. VALIDATION

PROBLEM.HOW adequate was the manipulation of information in the several treatments? Since a direct measure of the amount of information in the diagrams was not readily available, an indirect check was made by asking the question: Did the subjects perceive that the information in their diagrams produced varying levels of dependence upon their partners? Our expectations were based on three simple assumptions: one, the more information ego has, the less his dependency on partner; two, the more information his partner has, the more dependent ego is upon partner; and three, the effects described in the first rule are stronger than those described in the second. From these assumptions, the predictions shown in the third column of Table 1 follow. METHOD.At the end of each session, a brief questionnaire was distributed including the item: "How valuable were the contributions which your partner made?" The response alternatives, and the number of persons giving each answer, were: They were essential to completing the model (58) ; they were a Since we were not aware of a mathematical rationale for this technique, we performed a n empirical check suggested by a mathematical statistician, Professor Shayle Searle. Five cases were randomly deleted from each cell of 10, the 3-way ANOVA was then computed, and the deleted cases were then returned. This random deletion method was repeated 19 times. The F's for each main effect varied considerably from repetition to repetition, but the average F's of the 20 ANOVAs were in every case virtually identical to those based upon the technique reported above in the text.

INFORMATION DISTRIBUTION

accurate but I could have also given the same instructions (29) ; they were irrelevant to building the model ( 2 ) ; they were inaccurate or misleading ( 1) . The answers were dichotomized, with the first answer ("essential") considered as indicating "high-dependence-upon-partner" and the remaining three considered as indicating "low depesldence upon partner." RESULTS.The results are shown a t the right of Table 1. TABLE 1

Ego's Predicted and Actual Perceived Dependence Upon

Partner, b y Information Treatment

Information Treatments (Ego's first ; Partner's second)

TA-TC and PC-TC TA-TA, TA-PC, PC-TA, and PC-PC TC-TC TC-TA and TC-PC

Predicted Rank Order in Dependence

Actqfl Percent I-Iigh Dependence"

(1) Most (2)

(3)

85 78 35

40 10

(4) Least

30

20

Total

N

20

D r s c u s s ~ o ~Perceived . dependency is directly related to the amount of partner's information and even more strongly, inversely related to one's own amount of information. Since the subjects responded as the experimenters expected, we believe that the experimental manipulation had the intended effects on the subjects. ANALYSIS 2. THE EFFECTS OF INFORMATION

PROBLEM. HOWmuch of ego's participation is a function of (a) the information he himself possesses, (b ) the information his partner possesses, and (c) the combination of one's own and one's partner's information? METHOD.The means and standard deviations in percent-of-time-in-talk (activity) were computed for each of the nine analytic treatments. A threeway ANOVA was also computed, whose main-effect variables were: (a) Ego's Information; (b) Partner's Information; and (c) Conscriptees versus VoIunteers. Tests for homogeneity of variances among the several treatments proved negative, so no transformations on the data were carried out. RESULTS.All nine of the Volunteer treatment means were higher than the Conscriptee means (see Table 2). Table 3 shows that the difference between these two sets of means is significant. The only other significant condition in Table 3 is Partner's Information level. As seen in Table 2, subjects whose partners possessed the TC diagram were generally less talkative than subjects whose partner had either a TA diagram or a PC diagram. Considering the latter two both as "low i n f o m -

SOCIOMETRY

168

TABLE 2

Means and Standard Deviations of Ego's Activity b y Treatment and Tyge of Subject Type of Subject

Information Treatments

Ego's

Conscriptees

N

Partner's

-X

All Subjects

Volunteers

N

SD

X

N

SD

-X

SD

TC TC TC PC TC TA PC TC PC PC PC TA TA TC TA PC TA TA Mean of the 9 treatment values SD of the 9 treatment values

tion," it can be seen that one's own activity is dependent upon and inversely related to one's partner's information level. The most notable treatment difference in Table 2 is that between TC-PC and PC-TC. This relation appears to say that if ego has good information, but his partner has poor information, ego is very active; whereas if ego has poor information but is paired with an informed partner, ego will be inactive. But note that these large differences do not occur if poor information is operationalized by the ambiguous, obscured diagram (i.e., TA). Drscussro~.Perhaps the reader is as surprised a t the results in this section as we, a t first, were. I t seemed only common sense that a person's taskoriented activity would be influenced by his own information; and it seemed only a bit more sophisticated to expect that the effects of one's own informaTABLE 3 Analysis of Vaviance in Ego's Activity Source of Variation

Conscriptees vs. Volunteers (C) Ego's Information (E) Partner's Information (P) C x E C x P E x P C x E x P Within Total

df

MS

F

P

INFORMATION DISTRIBUTION

169

tion might be contingent upon the information partner has. We were quite unprepared to find that not only were neither of these effects significant, but that partner's information, alone, influenced ego. Post hoc, then, we offer some speculations to clarify both the positive and negative results, beginning first with the significant relation of alter's information to ego's parti~ipation.~ Alter's information may determine whether he responds positively or negatively to ego, and that this in turn would-as the operant conditioning literature suggests-shape the responses of ego (see, for example: Krasner, 1958; Levin and Shapiro, 1962; RiIatarazzo et al., 1965; Oakes et al., 1960; Salzinger, 1959). Since persons with excellent information (here, TC) probably find others' remarks a t best redundant, they would not be as likely to reward their partner's contributions, and they may very well react negatively. Those with incomplete or inadequate information (here, PC or TA) wouId, we suspect, be information-hungry and thus interested in stimulating or maintaining contributions by their partners. In either case, alter's information determines the reaction of ego. Let us turn now to the non-significant comparisons in Table 3. I t would be incorrect to conclude from these that neither a person's own information nor the interaction of his own and his co-workers' information affect his own activity level. The failure to demonstrate such significant effects might, paradoxically, have occurred either because the range of information selected for this experiment was too restricted or because it was too extensive. If there had been an additional treatment called "No Information" (no diagram a t all), it seems safe to assume that such egos would have behaved quite differently from others with partial or full information. On the other hand, an overall ANOVA such as the one presented in Table 3 may mask relationships which are more readily manifest in a subset of the total group of treatments. (To illustrate, suppose a "true effect" existed in a set of treatments. If additional "treatments" were then added whose values were drawn from a table of random numbers, the original effects would be lost in the mass of random events.) At this point in this paper, then, the reader should simply conclude that partner's information has been shown to affect ego's activity level, but that no ego-effects or interaction-effects were strong enough to differentiate among the nine treatments compared. The remainder of this paper may be considered as an attempt to pursue the question of the generality of the results presented in Analysis 2. Does the ego-effect or the interaction-effect appear a t all, and does the partnerThe significant differences between volunteers and conscriptees will be deferred and discussed in the section entitled Analysis 4, The Effects of Volunteering.

170

SOCIOMETRY

effect continue to hold up, when one examines the results under the following conditions: variations in type of interdependency (Analysis 3 ) ; conscriptee versus volunteer subjects (Analysis 4) ; and temporal phase in the interaction process (Analysis S ) ? These questions will be investigated through a set of increasingly molecular sub-analyses of the data so far presented. ANALYSIS 3. INFORMATION'S EFFECTS ON PARTICIPATION

BY TYPE OF INTERDEPENDENCY

PROBLEM. I t seems likely that ambiguous information produces a qualitatively different type of interdependence than does partial, but clear-cut information. The former is assumed to generate needs for consensual validation; the latter is assumed to produce alternations of total independence and total dependence. Are there differences in activity produced by these two different types of interdependency? Inspection of Table 2 shows that the two "low information" conditions do not have identical effects. For example, the TC-PC treatment is greatly different from the PC-TC treatment, but the TC-TA treatment is virtually identical to the TA-TC treatment. This section pursues these results. METHOD.Three 3-way ANOVAs were computed based on the data presented in Table 2, but each ANOVA omitted a different one of the three information condition^.^ In each case the question asked was the same as the one posed in Analysis 2, namely, how much of ego's participation is a function of (a) the information he himself possesses, (b) the information his partner possesses, and (c) the combination of one's o m and one's partner's information? In addition, Conscriptees vs. Volunteers was included as one main effect. RESULTS.The results are shown in Table 4. No results approach conventional levels of significance except in the comparisons involving the TC and PC diagrams, where all three main effects were significant. Note that Ego's Information is significant in this particular comparison, even though it did not show up as significant in Table 3 (Analysis 2). DISCUSSION.Why should the three main effects influence the subjects' activity in the set of treatments permuting the TC and PC diagrams but not in the set permuting TC and TA or the set permuting PC and TA? One plausible answer stems from the assumption that two strangers, of equal status, will probably adhere to elementary norms of courtesy and fairness, and will tend, insofar as this is possible, to share the floor approximately 4Thus the cells included in the three ANOVAs were: (1) TC-TC, TC-PC, PC-TC, PC-PC; ( 2 ) TC-TC, TC-TA, TA-TC, TA-TA; and ( 3 ) PC-PC, PC-TA, TA-PC, and TA-TA.

INFORMATION DISTRIBUTION

171

TABLE 4 Analyses of Variance in Ego's Activity, b y Pairs of Information Treatments TC and P C Treatments

Source of Variation

Conscriptees vs. Volunteers (C) Ego'sInformation(E) Partner's Information (p) CxE CxP E xP CxExP Within Total

d.f.

MS

1 1

740 922

F

PC and TA Treatments

TC and T A Treatments

P

5.1 <.04 6.4 <.02

1 1877 13.0 <.01 1 11 0.1 >.50 1 144 1.0 <.32 1 9 0.1 >.50 1 170 1.2 <.28 32 145 39 - A

MS

F

MS

P

F

P

548 2.2 <.I5 348 1.4 <.24

320 1.5 < . 2 3 51 0.2 >.SO

231 78 280 336 45 250

190 6 61 112 10 218

-

0.9 0.3 1.1 1.3 0.2

<.34 >.50 <.30 <.26 >.50

-

-

A

A



-

0.9 0.0 0.3 0.5 0.1

-

<.34 >.50 >.50 >.50 >.50

-

equally. When one partner literally has no information, this becomes impossible. I n the PC-PC treatment each person has no information for about half of the structure, so we can expect an alternation, with ego talking about as much as his partner. I n the TC-TC treatment, the partners have no strong dependence upon one another, but each may be expected to follow the courtesy rule by alternating the floor. There is no a priori reason, then, for the ego given the PC-PC treatment to differ in his total participation from the ego given the TC-TC treatment. The actual data (Table 2 ) show that the TC-TC egos to be active 31 per cent of the time, and the PC-PC egos 35 percent of the time. Both of these figures are very close to the overall mean for all 9 treatments, i.e., 33 per cent. When one subject has TC and the other PC, the holder of TC has full information during an entire session, whereas the holder of PC has good information half the time and no information the rest of the time. If sharing occurs during the time when both have information, and if only TC talks when PC's diagram is blank, then simple arithmetic indicates that TC would talk, over the entire session, about three times as much as PC. If, during the 50 per cent of the total time that PC has information he "attempts to make up for lost time," by doing all of the talking, then the two persons would, of course, speak an equal total amount. If we make the more realistic compromise assumption that both courtesy-sharing and attempts by PC to catch up are operative, we then expect the ratio of participation to be somewhere between 3: 1 and 1: 1. The data of Table 2 show that when TC and PC were paired, TC spoke 43 per cent of the time and PC spoke 20 per cent of the time, a bit more than a 2: 1 ratio. I t is hardly surprising that when a person has literally no information he

172

SOCIOMETRY

makes little or no contribution. But one might expect the holders of the T A diagram ( a badly obscured, small, and colorless diagram) also to defer to their better endowed partners. Yet the means and the ANOVAs show that these persons talked as much as did holders of the T C diagram. No information inhibits participation but ambiguous information apparently does not. The holders of the ambiguous diagram differed in several ways from the holders of the clear ones. We have evidence that they were considerably less accurate in the structures they built. They also perceived themselves as being more dependent upon their partners than did the holders of clear diagrams (see Table 1). There may have been important differences in the quality and content of their verbal contributions but unfortunately we do not have data bearing on this latter point. I n spite of these differences, there were no differences in sheer quantity of talk. The practical implication is that ambiguous information may be highly dysfunctional for task-oriented groups. Persons with such information, whether seeking consensual validation or attempting to master a challenge by overcoming their poor information, may waste a great deal of a group's time with their inferior contributions. Since their information is ambiguous, neither they nor their partners are likely to recognize readily that their information is deficient16 and so they may neither be self-inhibited nor extinguished by the negative reactions of others. ANALYSIS 4. THE EFFECTS OF VOLUNTEERING

PROBLEM.Analysis 2 showed that the paid volunteers were more active than the conscriptees. This section elaborates the differences and similarities between these two types of subjects. METHODAND RESULTS.TOcompare the volunteers and conscriptees, five results need to be considered: (1) I n every treatment the volunteers talk more than do the conscriptees (Table 2). ( 2 ) The product-moment correlation between these two sets of means is +.81. (3) The average variance of individual subjects about the mean of their treatments is nearly the same for conscriptees (141.5) and volunteers (136.1). (4) The variability among the nine treatment means is greater in the conscriptee sample (55.9) than in the volunteer sample (33.9). (5) The correlation (rho) between the conscriptee means and the increments of the volunteers over the conscriptees is -.56. In summary, the treatment means were similarly ordered for both volunteers and conscriptees, the volunteers showed an increment in activity 5 I n accord with this interpretation, our post-experimental questionnaire shows that subjects did not perceive a clear difference in the quality of contributions of TC and T A partners.

INFORMATION DISTRIBUTION

173

over the conscriptees in every treatment, but these increments were largest in those treatments where the conscriptees were particularly inactive. DISCUSSION. Rosenthal's (1966) research on volunteers showed them to be more sociable and motivated than non-volunteers. From his evidence, we expected the volunteers to be more active vocally, as in fact they were in every treatment. There are two lines of explanation for the fact that the activity levels among the volunteers did not rise uniformly over the level of conscriptees. The first and most likely explanation is simply the statistical unlikelihood that nine independent means in the second series would rise in proportion to the values in the first series. The second explanation is more complex. Suppose two talkative people find themselves paired off for a 40-minute period. At high levels of activity, the parties are in near zero-sum condition where, barring a rude shouting episode, one person's talk is a t the other's expense. Informal rules of courtesy, in effect, put a ceiling on the total activity levels in dyads. To raise already high levels of vocal activity would require that pauses be reduced; it would also require better synchronization of changes in who has the floor, otherwise there will be greater frequency of interruptions and simultaneous speech. Low levels of activity have ample space within the conversational period for one or both parties to increase activity without violating rules of courtesy. In short, the suspected ceiling for this activity dimension may make it far easier for participants to increase the volume of their vocal activity from 30 to 40 per cent of the total-time-in-talk-per-person than from 40 to 50 per cent. I t follows that the amount of increased talkativeness among the volunteers should be inversely related to the level of activity among the conscriptees; i.e., the treatments whose means were lowest in the first experiment should experience the greatest gains, while the treatments with the largest initial activity level should increase the least. The results cited above support this prediction. I t would be reasonable to infer from Rosenthal's work that volunteers habitually talk closer to the hypothetic ceiling of activity than non-volunteers. If this reasoning is valid, in effect the experimental treatments "worked" on the volunteers as they did upon conscriptees (cf. similar ordering of treatment means) but the treatments were not as potent in their effects on the volunteers (cf. smaller variance of the volunteer's treatment means) due to an effective ceiling on participation among the habitually more active subjects. ANALYSIS 5. THE TREATMENTS EXAMNED OVER TIME

PROBLEM. The problem is to determine when the treatments utilized in previous analyses took effect and how stable these effects were over time.

SOCIOMETRY

METHODAND RESULTS.Separate three-way ANOVAs were carried out on ego's activity for each of the eight five-minute periods in the 40-minute experimental session. The main effects were the same as those in Analysis 2, namely, Conscriptees vs. Volunteers, Ego's Information, and Partner's Information. I t is not necessary to present these results in tabular form since they can, for the most part, be succinctly summarized. The influence of Partner's Information is manifest as early as the first five minutes. The significant differences between conscriptees and volunteers first appears during the second five-minute period. The large F-values for those two effects (Partner's Information and Conscriptees vs. Volunteers) remains relatively stable throughout the remainder of the experiment. The other main effect-Ego's Information-has consistently small F values throughout the experiment. Concerning the last point, it should be recalled that in Analysis 3 a significant Ego's Information effect was found when the analysis was restricted to the combinations of TC and PC treatments. Moreover, the major source of the other experimental effects (Partner's Information and Conscriptees vs. Volunteers) also appeared to be traceable to combinations of the TC and PC diagrams. Therefore a detailed examination of these particular information conditions was conducted. Separate two-way analyses of variance were carried out on ego's activity for each of the eight five-minute periods within the conscriptee group (8 ANOVAs), within the volunteer group (8 ANOVAs), and for both groups combined (8 ANOVAs). Since, in these analyses the results for the conscriptees turn out to be similar to those for the volunteers, results are shown here only for the combined data. These results are summarized in Figure 2. Care must be taken in interpreting this figure, since the calculated values of F for each period are not independent. DISCUSSION. This temporal analysis indicates that the conclusions relating information and activity might very well be different in experiments of varying lengths. Among these treatment groups, both Ego's Information and Partner's Information were significant-surprisingly-as early as the first five minutes. Had the experiment been halted after fifteen minutes, our conclusion would have been revised to read: as experience in the dyad grew, the impact of the information treatments weakened. The conclusions from this experiment would be still different if the experiment had been halted after 30 minutes. A thirty-minute experiment shows that after an initial decline in the relative importance of the ego and partner's information treatments, there is a substantial rise in the effects of the information treatments on vocal activity. During the latter part of this hypothetical 30-minute experiment, the information conditions appear to be producing very large differences among the four experimental groups. This period is followed in the actual experiment by evidence of a sharp decline in the impact of information upon activity in the last ten, and particularly the last five minutes.

INFORMATION DISTRIBUTION

175

FIGURE 2 F ratios (1, 36 d.f.) from ANOVAs in Ego's activity, computed separately for each of the 5-minute periods of the 40-minute session, within combinations of the TC and PC treatments only. F's for interactions are not shown since none reached even the 40 per cent level of significance.

.01 level 1, 36 d.f, .05 level 1, 36 d.f.

I

0-4

I

5-9

I

I

I

I

I

I

10-14 15-19 20-24 25-29 30-34 35-39

Periods in Minutes Separate analyses of the conscriptees and volunteers for these same four treatments show essentially the pattern just described. I t is only this similarity of the pattern of F ratios over time in these two independent studies which makes any interpretation, other than that the results are a product of random variation, plausible. Only one important feature distinguishes the conscriptees from the volunteers: the absence among the conscriptees of a significant ego's information effect during the 20th to 35th minutes. If the pattern of F ratios for ego's and partner's information over time shown in Figure 2 is not coincidental, what might be responsible for the pattern? The interpretation we favor is that the level of interdependence between the members of each dyad was not a constant throughout the experiment. The most difficult portioz~ of the structure (the upper half) was assembled a t a time coincident with the period of the experiment in which the information effects were strongest (i.e., during the middle and last half of

176

SOCIOMETRY

the 40-minute period). The period in which the information effects were relatively weak (5-15th minutes) coincided with the period in which the lower half of the structure was being assembled, a time-consuming but relatively straight-forward instructional task, whose needs for interdependent activities were less. In short, not only may the experimental treatment groups differ in their relative dependencies upon one another, but in pursuing the task itself, additional variation in dependency may have been induced by the task itself which affected levels of vocal activity. The broader implication is that levels and varieties of interdependence among interacting individuals may be a much more transitory and shifting phenomenon than has previously been suspected. GENERAL DISCUSSION

The reader will recall that we began with a seeming contradiction in the literature concerning the relationship between vocal activity and information. One generalization asserts a direct relationship between activity and the quality of information. Another generalization asserts that activity increases as ambiguity and uncertainty increase. Of the three kinds of information treatments, there can be no doubt at all that the TC diagram objectively exceeds, in both amount and certainty, the information that can be found in the other two diagrams. For this reason, we could predict from the first generalization that the TC subjects will be more active than either the T A or PC subjects, in those dyads in which they were paired. The evidence is inconsistent. Contrary to prediction, in the TC-TA dyads, those with the ambiguous diagrams talked as much as the TC subjects. But as predicted, when paired with PC subjects, the TC subjects talked much more. The second generalization predicts that the TA and PC subjects will be more active than the T C subjects since the latter have less ambiguity and uncertainty in their diagrams. The results of our experiment are again inconsistent with these predictions, as the TC subjects talked as much as or more than either the TA or PC subjects. Both generalizations are far too simple and unqualified to predict the complexity of the results found in this experiment. I n seeking to explain the complexity of the findings, it becomes clear that there are a number of difficulties of both a substantive and methodological nature. Our initial objectives for the research required only that three qualitatively different diagrams be used, but now that the experiment is over, we have become dissatisfied with these qualitative differences and want to know more precisely how the diagrams differed from one another. Clearly, what we called the total-but-clear (TC) diagram contains uncertainty in the more difficult parts of the structure, and, even more, the PC structure contains

INFORMATION DISTRIBUTION

much uncertainty a t the places where the two halves of the diagrams come together. The T A diagram also differed from the others in terms of its smaller size, the absence of color coding of the pieces and the blue overlay, all contributing in unknown amounts to uncertainty. A single measurement of the amount of uncertainty in each diagram would have been enormously valuable in enabling us to make more specific predictions and strengthen our generalizations about information, interdependency and activity. The experimental results might have been clearer, with a smaller error variance, if better control had been exercised over the builder to whom the subjects gave instructions. Builders had been instructed to be passive automatons, but our observations made evident the difficulty of the role and the contribution they made to the total variation of subjects7 activity found in the experiment. Further consideration should be given to controlling the source of variation produced by the type of subject used since the volunteers behaved very differently from the conscriptees. Finally, if our speculations are valid about the way in which information distributions produced different levels of interdependency (Table I ) , there may have been important differences between the groups in how much talk was addressed to one's partner and to the builder. The principal target of communication in the low interdependence conditions must have been the builder, there being relatively little need to communicate between the subjects given the low uncertainty in their diagram. High interdependency dyads may have addressed a disproportionately large share of their communications to each other and less to the builder, which would be consistent with the fact that these dyads did not make as much progress with the structure as did those groups in which one subject held the T C diagram.6 Attention to the target of communication would facilitate determination of how differences in information influence the patterns of seeking and giving help or the tendencies of some group members to "go it alone." CONCLUSIONS

Given the limitations and the complexity of the findings, we offer the following generalizations as bases for further research rather than as firm conclusions: (1) Verbal activity in task-oriented groups is sometimes a function of the amount of one's own information, but under certain conditions persons with vastly different information may behave alike. OBecent pilot work by us, noting who communicates to whom in this task, provides tentative support for these speculations.

178

SOCIOMETRY

( 2 ) Verbal activity is more dependent upon the information available to one's partner than upon one's own information.

(3) There is no evidence indicating that the interaction between one's own and one's partner's information influences vocal activity. (4) Different patterns of perceived interdependence can be produced by the way information is distributed.

(5) The impact of information upon activity takes effect very rapidly but may vary within short periods of time as the shifting nature of the task changes the nature of the dependency between subjects. ( 6 ) Two types of inadequate information, partial-but-clear and full-butambiguous, produce strikingly different responses in those who possess them or who are paired with persons having them. Partial information restricts activity of the holder and enhances the activity of the holder's partner; ambiguous information does not have these effects.

( 7 ) Volunteer subjects are consistently more active in all experimental treatments when compared with conscripted subjects. REFERENCES Allport, Gordon W., and Leo Postman 1947 Chapters 1-2 in The Psychology of Rumor. New York: Holt. Bales, R. F. 1953 "The equilibrium problem in small groups." In Talcott Parsons, Robert F. Bales, and Edward A. Shils (eds.), Working Papers in the Theory of Action. Glencoe: Free Press. 1958 "Task roles and social roles in problem solving groups." Pp. 437-447 in Eleanor E. Maccoby et al. (eds.), Readings in Social Psychology. New York: Holt. Bass, Bernard M . "An analysis of leaderless group discussions." Journal of Applied Psychology 1949 33 (December) :527-533. 1960 Chapters 17-18 in Leadership, Psychology, and Organizational Behavior. New York: Harper. Bavelas, A., A. H. Hastorf, A. E. Gross, and W. R. Kite 1965 "Experiments on the alteration of group structure." Journal of Experimental Social Psychology 1 (January) :55-70. Caudill, William "Status and role in group interaction." Pp. 231-265 in The Psychiatric Hos1958 pital as a Small Society. Cambridge: Harvard University Press. Chapple, E. D., and W. T. Vaughn "A clinical method for studying the factor of human relations in disease." 1944 Journal of Laboratory and Clinical Medicine 29 (January) :I-18. Feldman, She1 (ed.) Cognitive Consistency. New York: Academic Press. 1966

INFORMATION DISTRIBUTION

179

Glanzer, M., and W. H. Clark 1964 "The verbal-loop hypothesis: conventional figures." American Journal of Psychology 77 (December) :621-626. Hayes, D. P., and L. Meltzer ''A general purpose recording system for social and sociologists." 1966 Paper read a t the annual meeting of the Upstate New York Sociological Association, Rochester, N. Y. "Bone conducting microphones." American Journal of Psychology 80 (De1967 cember) :619-624. Homms, George C. The Human Group. New York: Harcourt Brace. 1950 Krasner, L. "Studies of the conditioning of verbal behavior." Psychological Bulletin 55 1958 (May) :148-170. Krauss, R. M., and S. Weinheimer 1966 "Concurrent feedback, confirmation, and the encoding of referents in verbal communication." Journal of Personality and Social Psychology 4 (September) :343-346. Levin, G., and D. Shapiro "The operant conditioning of conversation." Journal of the Experimental 1962 Analysis of Behavior 5 (July) :309-316. Lott, A. J., and B. E. Lott 1961 "Group cohesiveness, communication level and conformity." Journal of Abnormal and Social Psychology 62 (March) :40&412. Matarazzo, J. D., A. N. Wiens, and G. Saslow 1965 "Studies in interview speech behavior." Pp. 179-210 in Leonard Krasner and Leonard P. Ullmann (eds.), Research in Behavior Modification. New York: Holt, Rinehart and Winston. Moscovici, S. 1967 "Communication processes and language properties." Pp. 226-271 in Leonard Berkowitz (ed.), Advances in Experimental Social Psychology. Vol. 3. New York: Academic Press. Norfleet, B. 1948 "Interpersonal relations and group productivity." Journal of Social Issues 4 (Spring) :66-69. Oakes, W. P., A. E. Droge, and B. August 1960 "Reinforcement effects on participation in group discussion." Psychological Reports 7 (December) :503-514. Riecken, H. W. 1958 "The effect of talkativeness on ability to influence group solutions of problems." Sociometry 21 (December) :309-321. Rosenthal, Robert Experimenter Effects in Behavioral Research. New York: Appleton-Century1966 Crofts. Salzinger, K. "Experimental manipulation of verbal behavior." Journal of General Psy1959 chology 61 (July) :65-94. Sullivan, Harry S. The Interpersonal Theory of Psychiatry. New York: Norton. 1953