Humans, Bias, and Communication Phenomena

Even if the specifics of the processes of coding meaning and information in intentional (or unintentional) communication are not fully understood, some of the results may be visible on a social level. If individuals have exchanged information, there might be a convergence in the meanings they associate with certain signs or actions. There my be an increased tendency to see things from each other’s perspective. Through conversation they establish a common context, a dictionary of shared words, and a collection of common connotations and references which grows with each interaction. In Eco’s terms, the “semantic trees” exchange elements; the possible meanings attached to the symbolic referents they employ become more similar. In a sense, a micro-language is developed, a subculture between the interactants, a set of expectations and conventions. If it were possible to attach some metric to the degree of meaning convergence, this might be a usable measure of information transfer.

Fig. 7 Diagram shows a summary of differences in mean semantic loadings of terms for English (star) and Japanese (circle) speakers. In this spatial representation, emotion terms that are judged as more similar (by a method similar to Osgood’s) are closer to each other than terms that are judged less similar. The dimensions of the diagram are from a Principal Component Analysis. Dimension 1 appears to correspond to what Osgood referred to as the Evaluative Factor (good-bad, pleasant-unpleasant, positive-negative) and Dimension 2 appears related to his Activity Factor (fast-slow, active-passive, excitable-calm). (copied without permission form Romney et. al. 1999)

Although not directed specifically towards information transfer, Charles Osgood’s “semantic differential” (1952) has some useful and conceptually similar elements. Osgood developed a methodology for using sets of 7-point scales with polar terms to measure the meaning of particular concepts. (ex. PACIFIST: Kind < - - - - - - -> Cruel ) He reports that the evaluations were very consistent across subjects in several studies. When factor analysis was performed, there was a strong tendency for ratings to occur along a few underlying dimensions. In other words, a great deal of the variance could be explained by an “evaluative factor” along with a “strength factor,” an “activity factor,” and “several others not clearly defined in this rough approximation.” (Osgood, 1952, p. 228). This suggests the possibility of representing the semantic associations of words by positioning the words as points in a multi-dimensional space so that the coordinates of the point give the factor loadings for the word. (Figure 7) As the connotations for the word and its usage change over time, repeated surveys might show gradual change in the factor loadings, resulting in the point “drifting” over time as its general meaning changes. Osgood mentions a study by Stagner and Osgood in which:

… a set of scales was used to measure the ‘meaning’ of particular concepts, such as PACIFIST, RUSSIAN, DICTATOR, and NEUTRALITY. Successive samples of subjects were tested between April, 1940, and March, 1942 (including a sample obtained just prior to the Pearl Harbor incident). … The feasibility and efficiency of using this method to record the changing structures of social stereotypes (e.g., the changing meanings of a set of social signs) were demonstrated. That a total shift from an essentially pacifistic to an essentially militaristic frame of reference had been accomplished, even before the Pearl Harbor incident provided the spark to overt expression, was clearly evident in the data. (Osgood, 1952, p. 226)

It seems likely that similar effects would be present at the level of individual interaction as well. Perhaps if two individuals were to take a semantic differential test on a specific subject, asked to discuss the subject with each other, and then re-tested, there might be some measurable convergence of the points representing the relevant concepts in their respective “meaning-spaces”. The idea of “information transfer” in this kind of methodological and conceptual framework would be based more on changes in conceptual linkages and semantic loadings than on reduction of uncertainty.

Related to this is a research tradition resulting from a series of “serial transmission” experiments done by F.C. Bartlett in the 30’s. The basic idea is that successive individuals in a transmission chain read a statement or story and then write what they remember as the input for the next person:

In both Bartlett’s original study and our serial reproduction phase, subjects assimilated the folk story with its unfamiliar ideas and obscure connections to their own culturally determined cognitive categories, or as Bartlett called them, Schemata. Bartlett’s subjects and ours condensed, highlighted, and rationalized the story to enhance its apparent coherence and consistency. (Kurke, et. al. 1989, p. 15)

There are two lines of thought here which I think it is important to develop: Systematic transmission bias, and cognitive “schema” or “frames” as a tool for organizing informational input. As Kurke (and presumably Bartlett) implies, bias and framing are closely linked. It is interesting to consider schema not just as a filtering system for highlighting and assembling pertinent perceptual or meaning elements from an information stream, but also as a cultural unit or information chunk itself. In other words, a different way of imagining the convergence and creation of shared meaning between interactants is to think of transmission as resulting in their possessing increasingly similar interpretive “frames”. The process of creating a shared dictionary of meanings is in a sense a process of generating a common point-of-view from which to interpret the codes of communication. Individuals who possess the same schema will agree on what portions of raw perceptual data are relevant and which should be ignored or discarded.

The schema or frame is a good concept for describing the “meaning” sense of information transmission because the literature surrounding the term seems to be well aware of the potential for people to adopt new interpretive frames, to alternate between conflicting interpretations, and to refine or increase the detail and precision of the schemas they employ to make sense of their surroundings.

… the key modeling principle here is that a person’s knowledge determines his ability to distinguish states of the world, and if a person cannot distinguish between several states of the world, he must take the same action in all of them. (Chwe, 1999, p. 2)

Transmission of a schema (or refinement of the detail in an existing one) need not be a permanent alteration to the perceiver’s perceptual framework, and may not always result in convergence of the interactants’ meanings. As individuals tend to discard or dis-attend information which is inconsistent with their current interpretation of what is going on, the schema they hold will likely affect the information they transmit. This is important for the filtering aspects. Losses and errors will not be random, instead they are likely to be strongly biased towards occurring in the parts of a message which are not understood or conflict with the teller’s original perspective.

[True white noise in a message] …would appear to a later human unit as bizarre, imperfect, randomly distorted, and, beyond a certain degree of information loss, unintelligible But the output of the human transmission and memory unit, no matter what degree of information loss, is apt to appear to a latter human unit as intelligible and usable as a base of action. This appearance of plausibility and comprehensibility in the output can accompany a total loss of the input message. Human beings as transmission units have this characteristic of rationalizing, of filling gaps, of providing outputs that lead to action rather than paralysis. (Campell, 1958, p. 341)

… through anticipatory monitoring of his own intended output, he makes an active effort to produce a coherent output by suppressing remembered detail that does not now seem to fit and by confabulating detail when gaps are conspicuous. Transmission thus typically involves reconstruction. The less meaningful the material, the fewer notions the transmitter has about the appropriate characteristics of the output, the more machinelike and random will the error output appear. (Campell, 1958, p. 342)

Campell (1958) gives a good general overview and classification of the kinds of bias which occur frequently when humans engage in serial transmission tasks. Except for the last category, these biases occur at the level of the individuals in the chain and may or may not describe the results of the overall process.

1) Abbreviation, simplification, condensation and loss of detail.

2) Middle message loss.

3) Closure (tendency to complete according to a schema, make consistent.)

4) Tendency towards symmetry or “good figure”. He points out that this is interesting from an information theory perspective, as regular or symmetrical forms have less information (can be described with fewer terms) than irregular forms of the same class.

5) Enhancement of contrast, figure/ground effects, categorization (things must be white OR black, not both)

6) Bias towards central tendency (ignoring unusual cases).

7) Assimilation to prior input (distortion to agree with previous information.)

8) Assimilation to expected message (hearing what one expects to hear.)

9) Assimilation to own attitudes.

10) Assimilation to reward and punishment: salience.

11) Distortion to please receiver.

12) Assimilation to prior output (desire to appear consistent to self and others).

13) Adaptation level, contrast error, or coding relativism (relativity of judgments to background and current context)

14) The tendency to make errors in output by duplicating input (failure to translate.)

15) Assimilation to prior coding assignments (holdover from other tasks.)

16) Coding contamination from associated cues (stereotyping, using one trait to incorrectly infer the presence of others.)

17) Assimilation to evaluative coding (things must be good or bad rather than large or small.)

18) Misc. biases (population stereotypes, response bias, survey layout problems, etc.)

19) Over dependence upon single input sources (ignoring the larger picture or context.)

20) Assimilation to other channels.

21) Conformity, pseudo confirmation (effects of communication and consensus in groups – people agree with each other rather than with what they saw)

Like anything dealing with information and communication, the list is complex, somewhat arbitrary, and sometimes contradictory. Some of the categories seem most relevant to the specific laboratory experiments from which they were derived, where others are almost broad enough to be “laws” of communication – laws which will almost by definition have loopholes and outlying cases. How does one distinguish, for example, between messages which will be subject to “enhancement of contrast” (exaggeration of distinctions) and those which will be “biased towards a central tendency”? But Campell in no way claims to have done much more than scratch the surface of the categorizations and inferences which could be drawn at that date.10 It is also important to consider the type of real world transmission tasks for which these categorizations are relevant. Campbell’s review was done in part under the auspices of the Office of Naval Research. The idealized hierarchical serial communications chains of the military as corporate worlds which the survey pertained to may not be typical of the processes experienced by individuals in everyday life.

In fact, I would argue that most social information and communication phenomena we experience are not well described by a serial transmission metaphor. I do not by any sense mean that they will therefore be exempt from any of the biases. If anything, the process will be more complex and there may be more effects to account for. Campell was aware of this:

In other settings, however, humans beings perform as groups, or operate as parallel units with interaction tangential to the stream of communication. In such situations it is possible that error tendencies arise in addition to those characteristics of single individuals. (Campell, 1958, p. 360)

As Campell suggests with his last category of conformity and pseudo-confirmation, one of the consistent findings in social psychology is that, not surprisingly, the decision and opinion formation processes of groups is usually not simply some kind of “sum” of the individuals’ processes. Asch (1952), Sheriff (1936) and many others have repeatedly demonstrated that interactions between group members strongly affect individuals’ statements about even seemingly basic perceptual judgments. It is not obvious how individual-level biases mentioned by Campell will affect system- or group-level dynamics when interactions are not delineated by the specific interaction geometries of a particular study.

Consider an experiment in the serial transmission tradition by F.L Brissey (1961). One group of subjects watched a short film (Hit and Run Driver) wrote a description, and took a relevancy weighted T/F test about its contents. The members of the second group each read a description from the first group, wrote their own description, and took the test. I’m simplifying the explanation, but basically data were collected on a set of 5-stage transmission chains. The basic results were not all that surprising:

In all groups the information retained is of greater relevance than is either the misinformation or the lack of information.” … “In other words, in successive stages of serial reproduction, information of ever increasing importance is lost.” [But] “A study of individual chains … reveals pronounced variations from one to another. It appears that the accuracy and the completeness of a given chain is closely related to the effectiveness of given communicants. A relatively ineffective message midway though an otherwise excellent chain markedly and irreversibly curtails the content of the series. (Brissey, 1961, p. 215)

So most of the individual chains did not in fact follow the orderly loss curve of the group average shown in figure 8, and often had strong step-functions in them. This may not be that significant in a serial transmission setup, where each individual is allowed only one input – the average “quantitative” loss will still be the same. But if individuals are allowed the more natural role of integrative units assembling an interpretation from multiple sources, the results could conceivably be very different. Imagine, for example, that all the individuals are connected together in a network so that each hears multiple versions of the story before repeating it. Even if several individuals have very bad communications skills and convey very little information, a person farther on “down” the chain may still get the full story from someone else. It is also important to consider that different people may lose different aspects of the story in their retelling. This means that there is the possibility that a person who receives several chains which contain very little information would still be able to re-assemble the story because each chain carried dissimilar items of the narrative. Of course, there is also the possibility of people assembling the incorrect or “bad” components of the story into a version which is equally self-consistent but very inaccurate – and it could be transmitted on down the chain instead.

Fig. 8 Graph shows the relative amount of material (mean number of words) in descriptions produced by the participants at successive positions on the transmission chain in Brissey’s serial reproduction study. (data from Brissey, 1961)

Questions about the transmission of inaccurate information are often discussed in the literature on rumor transmission. Some early theories held that transmission would always lead to loss of details and “factual” information, but later work shows that the opposite is often the case. Several studies emphasize the effects of rumor “importance” and situational “ambiguity,” but often had mixed results. Ralph Rosnow (1991), in a short review of work on rumors, states that contemporary theory describes four main effects. More rumors are transmitted in experimental conditions in which subjects are very anxious, and there seem to be smaller effects due to credulity (believability of the rumor) and the “uncertainty” of the subjects. What is not clear is the relationship between rumor and information transmission in general. Do ambiguous or uncertain situations simply induce more communication attempts by the subjects, hence more transmission of accurate as well as inaccurate items? Are some situations particularly likely to cause errors and distortions? How good are people at sifting out the actual facts from garbled accounts? It seems likely that effects will be partially dependent on context, message content, connectivity, and numerous other variables.

To date, I have found little discussion or analytical work on the idea of people uninvolved in the original transmission re-assembling messages from partial information. Certainly it is a very real effect which should be taken into account in human communication processes. Although the problem is related to that of noise and error correction in a communication channel, it is also tied up in the cognitive abilities of humans as sense-makers. As I mentioned before, people have the ability to independently assemble similar interpretations, schemas, or concepts from the same data without any communication occurring between them. Or they may be able to pull together apparently disconnected information and events to intuit facts or relationships which have not been directly communicated. Even the absence of communication may provide information. This presents an additional challenge for any attempts to analyze short term studies in purely information theoretical terms – the base of common-sense knowledge employed by individuals to disambiguate events and messages is vast. Who’s to say when a situation from a childhood detective novel will suggest an alternative interpretation of an experimental question? The order in which information is received may strongly affect how it is understood as well. People often seem to demonstrate a faithfulness to hard-won concepts and interpretations: “This tendency to distort messages in the direction of identity with previous inputs is probably the most pervasive of the systematic biases” (Campbell 1958, p. 346)

There are also many different kinds of group level properties than can act to filter or bias the noise in transmission. Kurke, et. al. (1989), mention the concept of “uncertainty absorption” in organizations.

When people absorb uncertainty, they often draw inferences from evidence ‘using the categories of the organization’s conventionally accepted conceptual scheme’.” “… the more complex the data that are perceived and the less adequate the organization’s language, the closer to the source of information will the uncertainty absorption take place. (March & Simon, 1958) [quoted in Kurke et. al. 1989 p. 6]

… serial reproduction removes information about beginnings, contingencies, and options and replaces it with information that suggests inevitability, givens, and certainty. (Kurke, Weick, Ravlin 1989, p.19)

In other words, there may be certain kinds of messages which a group simply does not want to hear. Individuals may feel implicit (or in political situations, explicit) pressure from the group to “soften their language”.

As the statement is passed from person to person it may be softened out of existence or modified into a more desirable form.11 This, again, brings up the issues of instrumentality and intentionality in transmission. Humans are often aware of some of the potential impacts of the information they transmit. There are certainly many specific situations which are not well described by a general bias as they are highly dependent on the content of the message. But even these instrumental situations (lying explicitly or by omission, controlling who is told what) may show some general trends when aggregated. In their study and literature review O’Reilly and Roberts (1974) mention fairly pervasive effects of status differences in hierarchies, and strong effects of trust on information communicated. Kurke et. al. (1989) mention in passing that females generally out-performed males in a serial transmission study.

Inherent in all these kinds of information transmission chain experiments are all sorts of interesting possibilities for methodological artifacts. This discussion began with the assertion that there are as yet no established empirical techniques for quantitatively measuring information or its transmission. The problem certainly hasn’t gone away. Word number counts and judged coding schemes (Kurke, et. al. ) or relevancy weighted T/F items (Brissey) are certainly a first step, but for all the reasons mentioned above, they are certainly not complete or appropriate measure for all situations. This is especially true when the media across which information transmission occurs is not spoken language or written text. The transfer of schemas or frames seems like a level of phenomena for which it is particularly difficult to define an appropriate metric. Yet this seems to be a conceptualization of information which is fairly in line with the way we commonly speak of it: as some “thing” which can be passed from person to person. At this level it might be more accurate to speak of what is being transmitted as an “idea” rather than “information” – although an idea (if it is a good one), provides an explanatory framework and so may fit well with the concept of “reduction of uncertainty” I’m beginning to stray somewhat out onto a philosophical morass here, but I think it is still worthwhile to consider the communicative potential of other modes – especially modes in which people may be less aware of their own instrumentality.

Humans, like most animals, communicate a great deal with their bodies, not only with position and movement (proxemics), but also through active ornamentation and display through apparel and action. Body language in humans is often a relatively unconscious behavior, but we also use it instrumentally in some situations. When they are being threatening, charming, or seductive, people tend to actively deploy cues of position, gesture, and expression as part of a behavioral collection aimed at producing a certain effect. Some of these cues seem to be relatively “hard wired”, while others are highly culture, locale, and context specific. Even very young infants show a smiling response when shown a smiling face. All humans tend to maintain a “personal space” between conversational interactants, and the distances often change as the relationship between the people changes. But the distances also very systematically depending on the ethnic background, class, and power relationship of the interactants. (Hall, 1966)

Proxemics and facial expression can communicate a great deal of information about the emotional status and intentions of an individual, but they are also used as signals from in-groups to out-groups. Couples, for example, often exhibit “With” behavior (Goffman, 1971) in which they make displays which indicate the connection between them – that they are with each other – both to each other and to a larger audience. Goffman presents the interesting perspective that most of public social behavior is in fact a sort of role playing or semi-intentional display of behavior in a near theatrical sense. This role-playing, and its interpretation by observers, applies to everyday life, not just in situations in which there is an active intent to misrepresent.

Here note that the expressive idiom of the individual’s society and group ensure that evidence of his assumptions about himself will be made available not only through his performing of his main substantive obligations, but also through expressive means, comprising the way he handles himself while in the presence of others or while having dealings with them … the individual exudes assumptions about himself. These provide others with a running portent, a stream of expression which tells them what place he expects to have… (Goffman, 1971, p. 344)

The idea that “behavioral communication” may be nearly ubiquitous should emphasize that information transmission does not require explicit linguistic behavior and may be almost unceasing and unavoidable when individuals enter into perceptual contact with each other.

As I have already mentioned, even a lack of communicative behavior can send a strong signal. This is true from both the social and information theoretic perspectives. There are often situations in which individuals convey disapproval through a marked lack of comment, or individuals may attempt to avoid interaction as a way of communicating anger or disdain. Socially, there is a great difference between not speaking to someone, and not speaking to someone of whose presence you are aware. In the information theory sense, absence of communication is a distinguishable state (perhaps the “space-between-words” character in Shannon’s examples) which can have its own expected probability. If the observed frequency does not match what is predicted, information is transmitted.

Even without its negative connotations, non-communication or off-hand communication (brief, stereotyped greetings, etc.) can carry information because of the social context in which they are embedded. Inaction or a failure to disagree is often taken to be a sign of social ratification of an existing situation or framing interpretation of what is happening. This is true not just in the trivial example of the Western wedding ceremony, but also as a consistent finding in social psychology. If you say “Good morning” to me and walk away, I can take that not only as ratification of a presumed relationship in which it is appropriate for us to great each other (and no more), but also as partial evidence that you are not aware of any information of sudden and dramatic importance, (imminent asteroid collisions with earth, rabid dogs in the immediate vicinity) and that my behavior conforms to your expectations as well. It is possible to interpret this as a part of a “social ratification” or conformity effect in which an individual’s behavior is endorsed or opposed by a social group: “When changes occur, it is usually only when the individual perceives that his group approves, or that support comes from a dissident sub group, or from an outside group toward which the individual sees himself moving or whose presumed standards he accepts” (Menzel & Katz, 1955 ). As Stanley Milgram (1969) demonstrated in his classic studies of obedience to authority, the power of such conformity effects is remarkable, and often under-estimated.

Unlike most animals, humans have gained considerable ability to alter their appearance on an almost daily basis. Although clothing in its current elaborate form may be a relatively recent invention on the evolutionary time scale,12 it is a powerful communicative element and shapes a great deal of interaction. Clothing and personal appearance provide powerful (and manipulable) markers of wealth, social and instrumental roles, group alignment, ethnic background, class, and event ratification. There seems to be some consensus that fashion choices are partially determined by imitation (Schrank & Gilmore, 1973). If so, fashion displays can be considered cultural behavioral messages. Certain “clothing messages” might become associated with certain groups or social subcultures (or be adopted by the groups because of their symbolic properties). Once established, the association between the clothing and the group’s attributes may become a stereotyped property of a larger cultural repertoire which can be drawn upon or “borrowed” by others. However, an effective reference requires that the displayer and receiver both have enough experience with the stereotype to catch what is being cited.

It is interesting to think of the role of fashion in identity display. Conformity and individuality can take on information theoretical properties in this context. If, in a group of people, everyone conforms very closely to the norm, no information is available to an outsider about any individual except the identity of the group category they have adopted. The perspective from within the group, however, would be slightly different. Because the properties common to all members are ever-present, they are not surprising; they carry no information. To an insider, only the subtle divergence from the norm will be noticeable, and members will be distinguished using characteristics which would be imperceptible to an outsider. Only the “gross” attributes of a group will be accessible to an outsider, until they have enough experience and familiarity with the norm to be able to make more finely grained categorizations. This may partially account for why, in segregated societies, people may say that members of the minority race “all look the same.” If this is in fact the case, there might be a fairly consistent ratio of trait contact (a threshold) at which members of the majority are able to distinguish among members of a minority, and this value might be derivable from information theoretic considerations. For groups, greater divergence from a norm facilitates the perception of individual identity, but too drastic a divergence among group members may weaken or eliminate outsider’s perception of the identity of the group as a cohesive whole. As much as fashion is an active and strategic, rather than a passive and receptive choice, individuals must consciously or unconsciously strategize about what kind of displays will allow them to be classed with their desired group while still presenting an identity as an individual to various audiences, both inside and outside the group

Fig. 9 Graphs show metrical measurements of skirt length of women’s formal evening dresses taken from high fashion magazines and averaged to create yearly means. The first graph shows the rate of change (derivative) and the second shows the within-year variance of skirt length. (copied without permission from Lowe & Lowe 1990)

These kinds of “strategic” adoption phenomena might drive interesting population-level trait patterns. To use dynamical terminology, the perceived group norm may act simultaneously as a repeller and attractor. The fashion transmission literature often assumes a phenomenon of leadership: there are certain individuals with desirable characteristics which others will try to emulate. (Schrank & Gilmore, 1973) Although the traits displayed by the “leaders” might initially act as attractors, they could also become repellers as they are assimilated by a group norm. If all individuals must present a display which is norm-divergent in order to proclaim an identity, but at the same time must stay norm-associated for acceptance and attribution to their relevant social groups, then there might be strong tendencies for directional or divergent drift in group or population averages. Because the normal value for a trait moves with the group distribution, it acts as sort of a reverse “carrot-on-a-stick.” In a sense this is the “runaway trait” phenomena described by Boyd and Richerson (1988). In the case of clothing, the magnitude of the drift may be somewhat bounded, as it is subject to a few functional and aesthetic or “moral” constraints. There are, in fact, observable national trends in fashion measurements (such as dress dimensions of formal wear, see Figure 9) which occur within very short time frames and frequently change direction. (Lowe & Lowe 1990)

It seems likely that there are many traits which, when compared to fashion, are more passively or unintentionally displayed and adopted – either because they occur early in development, are so normative to a culture that they cannot really be perceived until they are contrasted, or because the process occurs at such a “deep” level evolutionarily that individuals have no real control over display and adoption. Perhaps language would be an example of the first, and accent, dialect and cultural institutions are examples of the second. Although this is really just speculation, culturally determined sex roles and courtship behavior could be an example of the third. (There might be strong evolutionary pressure to conform to local courtship patterns.) For a trait to be “invisibly” transmitted in the second sense, it would have to occur with a high frequency in the population or group with which an individual is in contact. As I discussed earlier, what appears normal to one sub-group need not be normal in another, and, if groups are segregated, the group norm doesn’t have to agree with the average of the en

It is important to consider what other structural or contextual features might permit the formation of groups or spatialized patchiness of attributes like dialect or accent. Historically, this could be due partially to the effects of physical distance and the effective isolation of human social groups. But so called “mass media” which allow the display and reception of information across vast distances have drastically redefined the structural framework withen which social information transmission occurs. Innovations like writing, printed material, and the book add complex temporal effects as well. When information can be symbolically encoded in a material form it can be preserved, lying “dormant” until it is stumbled across at some later date and decoded. The concept commonly employed to generalize the structural effects of mass media is one of diffusion, as opposed to impermeable isolation. The concept of diffusion in social systems is by no means new. The basic ideas, at least as applied to culture, have been present in anthropology in one form or another for at least 150 years. But do diffusion (or “epidemic” in the Blue Tits example) principles accurately describe cultural phenomena? How about on the fine time- and space-scales of social groups?

Deutschmann and Danielson (1960) used phone interviews to track the “diffusion of knowledge of the major news story” in several cities. (Figure 10) Their work was based in part on Elihu Katz’s (1957) hypothesis of the “two-step flow of communication” in which ideas often flow from radio and print to opinion leaders and from these to the less active sections of the population. It would be interesting to know what role conversations and contacts between individuals play, as well as how effective the news media are at conveying a particular message. Little is actually understood about what the dynamics of news items are once the newspapers reach the streets.

…Then follows a second stage in the history of the story, about which little is known. We call this the period of news diffusion because in it the facts of the story filter through community life – diffuse through it so to speak – color it, change its complexion, reach and effect in some way nearly every person in it … It appears that for this class of events and general adult populations, one to two days is required for completion of the diffusion process even with full ‘bulletin’ and ‘banner headline’ treatment by the mass media. We found no rapid ‘word-of-mouth’ diffusion similar to that reported by Miller for his college sample. It should be noted that his respondents were homogenous, to a considerable degree spatially contiguous, and with a pattern of regular movement which naturally brought many persons in contact with many others. (Deutschmann & Danielson, 1960, p. 345)

Fig. 10 Results of phone survey of knowledge of major news stories: President Eisenhower’s stroke, the launch of the Explorer I satellite, and Alaskan statehood. (copied without permission from Deutschmann & Danielson, 1960 p. 348)

In their study, and in a related follow-up by Greenberg (1964), an emphasis is placed on the “importance” of a story as determining whether its transmission occurs more by direct contact with mass media or through personal communication. Both studies suggest that more important stories are more likely to be heard from personal sources. But to estimate the relative importance of social transmission it is necessary to consider how the news-worthiness of the item will couple with the normal patterns of social and media contact of a group: “Personal contacts play their largest role in diffusion when news is of greatest and least importance.” (Greenberg 1964, p. 489) When news items are carried by the mass media, most people tend to learn of them from that source. But as usual, the specific content of the message may have strong effects as well: “Events which go unnoticed by the majority may be deliberately chosen [for transmission]… by the few because those events have some functional importance.” And “gossip” news items which are only pertinent for members of a certain social group are unlikely to receive mass media treatment. There may also be strong self-monitored informational “boundaries” between social groups because members realize that their gossip will be of little interest to outsiders and so neglect to discuss it with them – making the groups effectively isolated with respect to certain dimensions.

The “two-step flow” literature based on Katz, and certainly the “diffusion of innovations” (Rogers, 1962; Valente, 1996) and the fashion-diffusion literature, often emphasize the idea that there are certain individuals who play the roles of brokers, opinion leaders, “early-adopters” or innovators. This may be more true of slower processes like innovation adoption in agriculture, but apply less to rapidly percolating phenomena like news or gossip. After a survey of relevant studies Katz points out that,

…It became clear that opinion leadership could not be viewed as a ‘trait’ which some people possess and others do not, although the voting study sometimes implied this view. Instead, it seems quite apparent that the opinion leader is influential at certain times and with respect to certain substantive areas by virtue of the fact that he is ’empowered’ to be so by other members of his group. (Katz, 1957, p. 68)

This suggests that it might be important and worthwhile to consider the effects of social structure in more detail than the diffusion perspective permits.

10 It is interesting to compare these biases and their implications with the Boyd and Richerson’s (1988) classes of cultural transmission bias. Part of the argument I’m trying to make is that cultural transmission may be communication “writ large.” 11 Kurk, et. al. (1989, p. 4) give a fascinating and terrifying example of the behavior of US troops during the Vietnam War. “At My Lai…the order sent down from headquarters to a brigade said, ‘On no occasion must hamlets be burned down.’ The brigade radioed the battalion, “Do not burn down any hamlets unless you are absolutely convinced that the Viet Cong are in them.’ The battalion radioed the infantry company, ‘If you think there are any Viet Cong in the hamlet, burn it down.’ The company commander told his troops, ‘Burn down the hamlet.'” 12 There are many possible sociobiologically styled arguments for why it might be adaptive to use personal appearance for signaling. High status items, like Armani suits for Western men, could function as the “peacock’s tail” in sexual selection. They make you look good within a particular cultural schema, but they, and the accompanying appearance and lifestyle, are expensive to obtain and maintain, thereby providing a potentially useful indicator of wealth, power, and status to would-be mates.

10 It is interesting to compare these biases and their implications with the Boyd and Richerson’s (1988) classes of cultural transmission bias. Part of the argument I’m trying to make is that cultural transmission may be communication “writ large.”
11 Kurk, et. al. (1989, p. 4) give a fascinating and terrifying example of the behavior of US troops during the Vietnam War. “At My Lai…the order sent down from headquarters to a brigade said, ‘On no occasion must hamlets be burned down.’ The brigade radioed the battalion, “Do not burn down any hamlets unless you are absolutely convinced that the Viet Cong are in them.’ The battalion radioed the infantry company, ‘If you think there are any Viet Cong in the hamlet, burn it down.’ The company commander told his troops, ‘Burn down the hamlet.'”
12 There are many possible sociobiologically styled arguments for why it might be adaptive to use personal appearance for signaling. High status items, like Armani suits for Western men, could function as the “peacock’s tail” in sexual selection. They make you look good within a particular cultural schema, but they, and the accompanying appearance and lifestyle, are expensive to obtain and maintain, thereby providing a potentially useful indicator of wealth, power, and status to would-be mates.

Leave a Reply

Your email address will not be published. Required fields are marked *