Consequential unscripted interactions: A conceptual and empirical description (Hample, Dai, Zhan, &...

60
Consequential Unscripted Interactions: A Conceptual and Empirical Description A consequential unscripted interaction (CUI) is a conversation that a person understands to be important but does not know how to do. Readers can probably remember episodes such as being uncertain how to ask for a date, how to require use of a condom, how to express romantic love for the first time, how to ask for a job or promotion, or even just how to explain that you would rather not play shortstop. The two key features, consequentiality and unscriptedness, are both subjective, and take their definitional status solely from the viewpoint of a particular interactant. Perhaps only a small group of people regards some episode as a CUI (e.g., recent immigrants having trouble asking police officers to help them). Sometimes registering something as a CUI merely indicates one’s life stage (e.g., asking for a date for the first time). Or perhaps a CUI is relatively widespread (e.g., learning to insist on safe sex practices in the

Transcript of Consequential unscripted interactions: A conceptual and empirical description (Hample, Dai, Zhan, &...

Consequential Unscripted Interactions: A Conceptual and

Empirical Description

A consequential unscripted interaction (CUI) is a

conversation that a person understands to be important but

does not know how to do. Readers can probably remember

episodes such as being uncertain how to ask for a date, how

to require use of a condom, how to express romantic love for

the first time, how to ask for a job or promotion, or even

just how to explain that you would rather not play

shortstop. The two key features, consequentiality and

unscriptedness, are both subjective, and take their

definitional status solely from the viewpoint of a

particular interactant. Perhaps only a small group of

people regards some episode as a CUI (e.g., recent

immigrants having trouble asking police officers to help

them). Sometimes registering something as a CUI merely

indicates one’s life stage (e.g., asking for a date for the

first time). Or perhaps a CUI is relatively widespread

(e.g., learning to insist on safe sex practices in the

1980s). In other words, CUIs can be localized or

widespread. By definition, they are always consequential,

on a person-by-person basis.

Both scholarly and popular literature discusses a

related idea, the difficult conversation. CUIs and difficult

conversations probably overlap considerably in ordinary

life, but they are distinguishable conceptually. Difficult

conversations are simply hard to do. They might be

challenging because the person does not know how to do them,

making them also CUIs, but the difficulty may arise from

other sources as well. In fact, the literature commonly

suggests that what makes these conversations difficult is

their emotionality. For instance, nurses, even after

training, find it spiritually draining to give bad news to

patients or their families. A supervisor who knows exactly

what script to follow in giving a negative performance

appraisal may still stew over the experience and feel badly

about doing it. Difficult conversations are consequential,

just as CUIs are, but their oppressiveness may not derive

from being unscripted. Learning a script may not make these

episodes less difficult. CUIs by definition are unscripted

and consequential, and whatever emotional load they impose

derives from both of these features, not merely the second.

Consequentiality implies that CUIs will probably also have

emotional weight, but this is an empirical matter for CUIs

rather than a definitional attribute, as it is for difficult

conversations. Nonetheless, the likelihood is that most of

the relevant conversations will be both difficult and CUIs.

We will soon report on the emotionality of CUIs.

A great deal of the “difficult conversations” scholarly

work focuses on healthcare contexts, commonly giving bad

news to patients (e.g., Browning, Meyer, Truog, & Solomon,

2007; Lamiani, Meyer, Browning, Brodsky, & Todres, 2009;

Lamiani, Meyer, Leone, Vegni, Browning, Rider, Truog, &

Moja, 2011; Makoul, Zick, Aakhus, Neely, & Roemer, 2010).

Discussing racial matters in therapy has also been a matter

of investigation (Cardemil, & Battle, 2003). Another

substantial body of work deals with family and personal

issues (e.g., Keating, Russell, Cornacchione, & Smith, 2013;

Smith, Cornacchione, Keating, & Russell, 2014; Stone, Patton

& Heen, 1999). Some of this close relationships research

includes focused analysis of explaining racial matters to

children (Copenhaver-Johnson, 2006) or declaring one’s

sexual orientation (Manning, in press). Recent applied work

deals with teaching children and other family members to

deal with post-deployment problems among U.S. military

(e.g., Knobloch, & Theiss, 2012; Wilson, Wilkum, Chernichky,

MacDermid Wadsworth, & Broniarczyk, 2011). The commonness

of the term “difficult conversations” is almost certainly

due to a well-received advice book that often has workplace

interactions in mind (Stone, Patton, & Heen, 1999). All

these research topics range from the obviously heart-rending

(e.g., receiving a terminal diagnosis) to the mildly amusing

(e.g., explaining twerking to one’s parents; Wayne, 2013).

Many of the studies just cited involve the development

of interventions designed to guide people through difficult

conversations. From our point of view, it is interesting

that all those programs can be summarized as teaching

scripts to people. Not having a suitable script for an

episode is a defining feature of a CUI. But knowing a

script will not necessarily make it more pleasant to tell a

family their father is dying or telling an employee that

s/he is losing a job. The conversation can remain difficult

even when the participant is competently following good

instructions.

Even though difficult conversations are essentially

identified due to their emotionality, we notice that the

intervention scripts seem mainly aimed at improving

communicator competence rather than relieving communicator

distress (e.g., Browning et al., 2007; Lamiani, et al.,

2011; Villagran, et al., 2010). The motivation for these

interventions seems to be workplace-practical: how to get

the conversations done efficiently and clearly. Less

attention is being given to how it feels to say these

things, one after the other. The emotionality of these

conversations is sometimes felt most pointedly before the

episode, in some cases with a sense of relief or other

positive feeling appearing once the conversation is

completed (e.g., Keating, et al., 2013; Manning, in press).

Emotions before, during, and after the conversations are

separate phenomena, not to be elided.

By “scripts,” we have in mind a known set of directions

for behaving, normally less precise about prescribing actual

utterances than the script an actor in a play would have.

Honeycutt and Bryan (2011) have analyzed a great many

interpersonal scripts, approving of some and critiquing

others. Besides the Honeycutt and Bryan book, we have in

mind Schank (1982) as a theoretical frame and Kellermann,

Broetzmann, Lim, and Kitao (1989) and Honeycutt, Cantrill,

and Greene (1989) as research exemplars. A more unusual

example is Copes, Hochstetler, and Cherbonneau (2012), who

studied carjackers’ scripts (or lack of them) when

confronted with unexpected victim resistance. The present

investigation does not require an elaborate explanation of

the idea of scripts, however, and so we will move forward

with the general idea of script/schema/schemata that has

appeared in many lines of research over the decades.

While we do not move to the point of offering any

scripts in the current paper, we observe an oddity about the

interventions when CUIs and difficult conversations are

compared. The interventions offer scripts (and so are

apparently aimed at CUIs) but do so in the context of

difficult conversations (whose defining attribute, punishing

communicator emotionality, is not always addressed by the

scripts). The distinction between CUIs, notable for

participants’ uncertainty as to how to proceed, and

difficult conversations, defined by their emotionality,

would seem to be informative to several scholarly and

practical communities.

The ultimate aim of this research project is, in fact,

to generate scripts so that people can be instructed and/or

practiced in the important conversations for which they have

great uncertainty as to their competence. This paper begins

the project by obtaining useful descriptions of CUIs as they

actually appear in people’s lives. We will code open-ended

descriptions to find out what topics are involved in CUIs,

what consequences people are concerned about, with whom

people try to participate in a CUI, and various other

descriptive matters. These results should help researchers

identify and select particular CUIs for further work. In

addition, based on the material just reviewed, we

anticipated several elements of the CUI experience, and

formed closed-ended scales to measure people’s uncertainty,

their emotional reactions, how forceful the memory of the

CUI was, and various subjective outcomes from the

interactions. This is obviously an exploratory study, and

so we specify our immediate aims as research questions:

RQ1: What are the characteristics of CUIs, as recalled

by participants?

RQ2: How are uncertainties, feelings, rumination, and

outcomes related to one another, in the view of respondents?

RQ3: Are the topics and consequences of the recalled

CUIs distinguishable by reference to uncertainties,

feelings, rumination, and outcomes?

Method

Participants

We collected data online from two different samples.

The first was 257 undergraduates enrolled in communication

classes in a mid-Atlantic public university. They received

minor extra credit for their participation. Three

undergraduates gave such a limited description of their CUI

that the descriptions were not codable, and inspection

revealed that their closed-ended responses formed simple and

unlikely patterns, so they were dropped from the sample.

Undergraduate participants’ average age was 19.5 years (SD =

1.8). Women were 68% of the sample. Their most common

relational status was single (60%), followed by single in a

serious relationship now (29%) and single serious

relationship before (9%). Only 2% were married. Their most

common self-identified ethnicity was Euro-American (54%),

followed by Asian-American (9%), African-American (8%),

Hispanic-American (6%), and Asian (4%). Some (23%) declined

to estimate their family’s approximate annual income, but

45% thought it was more than $100,000.

The second sample was 230 Mechanical Turk workers,

recruited through Amazon’s mTurk system. Each was paid

$0.50 for completing the survey, using Amazon’s normal mTurk

procedures. Their average age was 36 years (SD = 12.9).

Males comprised 44% of the older adult sample. Their modal

marital status was married (40%), followed by single in

serious relationship now (21%), and single (21%). Most

(63%) self-described themselves as Euro-American, with 10%

Asian-American, and 9% African-American. Their modal family

income was $20,000 to $40,000 (29%), with 20% less than

that, 20% in the range $40,000 to $60,000, and 15% between

$60,000 and $80,000. Most (61%) were employed full time,

22% part time, 4% retired, and 18% unemployed (possibly

including full time students). The modal level of education

was university graduate (37%), followed by some university

(23%), at least some graduate education (15%), and community

college graduate (13%). Only 13% had a high school degree

or less.

Procedures

After completing demographic items, respondents were

instructed as follows:

This study is about important difficult

conversations. By “important” we mean that the

conversation had (or seemed to have) genuine

consequences for your life. Those consequences might

have been emotional, financial, relationship-oriented,

educational, or anything else you considered to be

important at the time. By “difficult” we mean

conversations that were challenging to participate in

because you weren’t sure how to conduct yourself, or

perhaps the other person didn’t know how to

participate. The conversation may have occurred when

you were adult or earlier – either is fine for the

purposes of the study.

The survey essentially asks you to describe one

conversation that was both important and difficult, a

conversation that you have actually had. You will be

asked to describe the conversation in your own words,

and then we will have some specific items for you to

fill out about it.

Their open-ended responses (“please describe the important

difficult conversation”) supplied the materials that were

coded. Remaining elements of the surveys were closed-ended

Likert items, detailed below.

Coding

Two of the authors coded the CUI descriptions. We

first read through all the descriptions from both samples to

see what seemed to be present, and then developed a draft

codebook. We anticipated some of the coded variables before

we examined the descriptions (e.g., topic) and others

appeared to us as we read (e.g., personal or institutional

identities). The two coders applied the codebook to 20

descriptions (combining both samples), calculated their

reliability, resolved their disagreements, revised the

codebook, and then did another round of coding. Some

variables (e.g., topic) were immediately easy to code but

others (e.g., emotion, consequences) took several rounds of

codebook revision. Eventually all variables were coded with

acceptable reliability. Tables 1 and 2 shows the coding

results for both samples, as well as the Krippendorf’s

alphas for each variable. All reliabilities exceeded

minimum acceptable levels for Krippendorf’s alpha.

The main codes were as follows. Topic was of course an

obvious concern, and we found a number of discernible

categories relating to sexual relations, illegal activity,

medical problems, financial issues, romantic matters,

work/job concerns, academic issues, and some conversations

dealing with death. The more specific categories are in

Table 1. We noticed that some of the descriptions were

ambiguous about whether or not the CUIs appeared to have

actually taken place, so we coded whether there was clear

evidence that the conversation happened. We coded settings

twice, once as to whether the conversation took place in a

private or public location (defined as whether others were

present who could hear or see the conversation), and once

into more specific settings, such as academic, workplace,

residence, and so forth. The channel within which the CUI

was conducted was coded, as was whether the participants in

the CUI were enacting personal or institutional identities. We

wondered whether the CUIs were consequential because they

were relevant to a decision. We noted the general tenor of the

respondent’s reaction, according to whether it was cooperative or

not (here, and for the next code, “respondent” refers to the

person responding to the CUI’s initiation, not necessarily

to the person who provided the data). Similarly, we coded

the respondent’s emotional status in a gross way, simply noting

whether it was expressed as being positive or negative.

Finally, we coded the nature of the CUI’s consequentiality,

finding subcategories dealing with self-image, close

relationships, resources (e.g., money), health, and the

future. These are detailed in Table 2, along with all the

other coding results except those bearing on topic, which

are in Table 1.

Closed-ended Instruments

Remaining measures were Likert items ranging from

strongly disagree to strongly agree on a 1-10 point scale.

Descriptive statistics, including Cronbach’s alphas, are in

Table 3.

We asked respondents to report self certainty before the

conversation began as well as to estimate the other’s certainty,

both with 3 items (e.g., I was expecting it to happen; the

other person knew how to conduct himself/herself). We also

solicited estimates of the same things during the

conversation (e.g., I knew what I was going to say next; the

other person seemed to know how it was going to end).

Respondents also recalled the degree and frequency with

which they reflected on the conversation after it was over,

using 7 items (e.g., I thought of things I could have done

better; I became more settled in my mind about what

happened).

Remaining measures were straightforward single-item

instruments, used instead of multi-item instruments out of

concern for respondent fatigue or annoyance. All of these

were developed in Hample (in press). Politeness (we were both

polite to one another) and rudeness (at least one of us was

rude to the other) were both measured because a conversation

can have both qualities or neither. Participants also rated

CUIs both for their argumentative (at least one person gave

his/her reasons for what s/he said or wanted) and conflictive

(the conversation contained clear disagreement) qualities.

The distinction is that argumentative exchanges require

reasoning and conflicts require explicit notice of

incompatible goals. A series of discrete emotions were

found to be most common in everyday conversations in Hample

(in press). These were assessed both during and after the

CUI. These were anger (I was angry), uncertainty (I was

uncertain), disappointed (I was disappointed in myself), happy

(I was happy with the conversation), sad (the conversation

made me feel sad), guilty (I felt guilty during the

interaction), surprised (I felt some surprise during the

conversation), worried (I was worried during the

interaction), thankful (I was thankful that I had this

conversation), bored (I felt the conversation to be boring),

and concerned for self image (I was concerned about giving a

negative impression of myself). In addition, during the CUI

we also assessed carefully edited messages (I was carefully

editing what I was saying).

A few other descriptive items (e.g., how many people

were involved in the conversation?) were also gathered and

will be reported momentarily.

Results

Research Question 1: Characteristics of CUIs

Here are the leading characteristics of the CUI

interactions in our two samples. Information on zero or

very low-frequency results is available from the authors.

First we present some contextualizing descriptions, and then

proceed to Tables 1 and 2.

Undergraduates reported that they had the CUI about two

years before the survey, when they were 17.7 years old (SD =

2.8), and estimated that the other main person in the CUI

was noticeably older, 35.2 years (SD = 15.1). More than

half (58%) of the CUIs were dyadic, but 27% had a third

person present, and 15% had four or more people

participating. The most common other main person was parent

(42%), followed by romantic partner (22%), and friend (13%).

More had CUIs with a subordinate (7%) than with a supervisor

(1%). The other main conversant was somewhat more likely to

be male (51%). The respondent (49%) and other main person

(48%) were about equally likely to have initiated the CUI,

with only 3% being instigated by a third party.

The mTurk respondents reported that on average, they

had the CUI when they were 31 years old (SD = 11.9; i.e.,

about 5 years prior to the survey), and they estimated that

the other main person was 40 years old (SD = 16.6). Nearly

three-quarters (74%) of the CUIs were dyadic, with 18%

involving a third person, and 9% involving four or more

people. The main other participant was most commonly a

romantic partner (34%) or a parent (27%). Friends were less

frequent partners (9%), having about the same frequency as

one’s supervisor (10%). Acquaintances (3%) and strangers

(4%) were about as common as one’s child (4%), but somewhat

more rare than one’s sibling (6%). The other main

participant was most often female (54%). The CUI was most

often initiated by the respondent (59%), followed by the

other main person (37%) and a third party (4%).

The mTurk respondents were older and their longer life

experience may have given rise to several of the obvious

differences between the two samples. Their CUIs were more

distant in time from the data collection, perhaps suggesting

that they do not have as many recent CUIs to recall. Their

average age at the time of the CUIs was well beyond ordinary

university graduation age, implying that some of the things

registered as CUIs by the undergraduate respondents may have

been replaced or lessened in importance as respondents moved

into their next decade. The older adults were more likely to

have experienced the CUI in a dyadic interaction. Romantic

partners were more likely interactants for them than

parents, compared to the undergraduates, but these were

still the most common CUI partners for both groups. The

older adults were more likely to have initiated the CUI than

the undergraduates were. This point about initiation needs

some reflection: why would a person (49% of the

undergraduates and 59% of the mTurk respondents) initiate a

consequential unscripted conversation? Perhaps the topic’s

consequentiality impelled the interaction, or perhaps the

conversations moved in unanticipated directions; our data do

not distinguish these possibilities.

Table 1 reports our classifications of the CUI topics

for both groups. The largest general topic categories for

both groups were romantic relationships (25% of

undergraduate topics and 24% for mTurk respondents), work or

job issues (10% and 11%), medical matters (9% and 16%),

finances (9% and 13%), and academic matters (14% and 5%).

Some differences between the samples are evident but the

same general kinds of issues seemed to resonate with both

groups. Within these larger categories, some differences

between the samples appeared, but most of the observed

frequencies were too low to make statistical significance

testing worthwhile. In fact, if our aim had been to do such

analyses, we would have condensed or omitted many of the

codes in the table. Instead, we want to communicate as many

details as possible, even if they are rare, so that future

work can pick up one or another of these topics and explore

it in view of its nature as a CUI.

Table 2 contains the results of the rest of our coding

for the open-ended CUI descriptions. In Table 2, sometimes

the “other” or “uncodable” categories are the most

interesting, precisely because these matters were not

important enough for respondents to mention spontaneously.

For example, few thought that whether the CUI’s setting was

private or public was very salient (is this truly

unimportant, or is it just assumed that the CUIs will be

done in private?). The physical setting of the

conversations was most often unmentioned, whether we were

coding privacy or location. The possibility that the

respondent was cooperative or avoidant during the

conversation was not described about half the time, and the

valence of the CUI respondents’ emotional reactions was not

apparent to us two thirds of the time. We need to think

about whether these unmentioned things are genuinely

unimportant, under what circumstances they become salient,

and whether encouraging awareness of these matters might be

a first step in an intervention.

The Table 2 results are more affirmatively informative

in other respects. Some of the descriptions did not give

clear evidence that the CUIs actually took place (2% for

undergraduates and 7% for mTurk informants). Perhaps their

nature was foreseen and a threat avoided, or perhaps the

descriptions were just not well written. Respondent

reaction is relevant to this point, because there we found

that about a quarter of the time people in both samples were

uncooperative or somehow avoidant during the conversation.

More than a quarter displayed a negative emotional tone to

us, compared to only a handful who had positive affect. The

most common consequentiality for both samples concerned the

relationship between the two conversants: this was our code

for 37% of the undergraduate reports and 39% of the mTurk

descriptions. No other consequentiality approaches this one

in frequency. However, utilitarian matters (in contrast to

relational ones) also appeared: career and academic matters

were important, along with health issues in both samples.

Settings were more often private than public (but mostly not

inferable from the descriptions), and CUIs were most

commonly conducted face to face (with both samples often not

finding this important to mention, either because it never

mattered or perhaps because they assumed that such

conversations would be private and face to face; however,

some respondents indicated that they chose texting or social

media because face to face contact was not feasible).

People participated in CUIs out of their personal

identities, rather than as supervisors, police officers, or

other institutional role occupants. Scripts may be more

commonly available for institutional identities than

personal ones.

So we have a general sense that CUIs are private

experiences, emotionally negative to recall or participate

in, and involve key matters of personal and relational

identity. However, these are only simple generalizations,

and some readers may find that the most interesting CUIs are

public events, tap institutional identities, and result in

positive feelings such as satisfaction.

Research Question 2: Relationships Among Uncertainties,

Feelings, Rumination, and Outcomes

Besides asking for free-form descriptions of CUIs, we

also requested that respondents provide us with some

specific information that we felt pointed directly to our

theorized nature for CUIs. These included people’s

(un)certainty about participating in the conversations,

their perceptions of the CUI’s characters, and their

feelings while participating. Table 3 provides descriptive

information on these measures (along with comparisons of the

two samples), and Table 4 shows their correlations.

Correlations involving the self-reports of emotions are in

Table 5.

In reading Table 3, readers should remember that the

theoretical midpoint of these scales is 5.5, because this

will help contextualize which matters stood out for

respondents (e.g., politeness, argumentativeness,

uncertainty, thankfulness) and which seemed irrelevant or

rare to the CUI experience (e.g., rudeness, boredom). The

two samples had quite comparable scores on nearly all the

measures. The only real pattern of note is that the

undergraduates had higher estimates of whether the other

participant knew what was going on. While undergraduates

and the older adults had somewhat different patterns of

topic (Table 1) and consequentiality (Table 2), the general

nature of CUIs seems not to have been very different for

them.

Table 4 shows how these subjective ratings of the CUI

experience (excluding emotions, which are in Table 5)

related to one another. The two samples have very

comparable correlation matrices, with only a few

coefficients being as much as .10 different between the two

samples. Let us mention some of the most interesting

results.

An initial interest is the four certainty measures,

because these implement our idea that CUIs are unscripted.

Uncertainty about how to proceed was a shared experience,

with self and other uncertainty before and during the

conversation mostly having modest positive associations (the

only exception being before the CUI for the mTurk

respondents). Shared variance was only about 4%, however,

so these are not strong associations. Own uncertainty

dissipated in both samples (see the means in Table 3), but

still had very large positive correlations between the

before and during measures for both samples. Estimates of

the other person’s certainty followed this same

correlational pattern. Own certainty predicted that the

conversation would be argumentative (that is, reason-filled)

but not particularly conflictive (i.e., characterized by

disagreement).

The distinction between recalling that a CUI was

argumentative and that it was conflictive has immediate

implications for how to design interventions. Use of

reasons implies that people are honestly try to think their

way through the problem, but simple disagreement suggests

that either they have little idea how to start or have lost

hope as the episode progresses. Participants’ uncertainty

had negligible associations with conflictual episodic

character, but the more confident people were about how to

proceed, the more they participated in exchanges of reasons

and evidence. Both argumentative and conflictive episodes

were recalled as stimulating reflection after the episode,

implying that people might be open to instruction.

Another element that points toward whether

interventions will be more or less challenging is the

emotionality of the episodes. Own certainty (i.e.,

knowledge of scripts) tended to produce slightly more polite

exchanges, but rudeness was associated with the other person

seeming not to know how to act. Rudeness was strongly

associated with being conflictive, but not with being

argumentative. The Table 3 results about emotional levels

indicate that most of the self-reported feelings were around

the theoretical midpoint of the scales or lower. Besides

uncertainty, only sadness and worry seemed to stand out as

potential barriers to help, and these were somewhat balanced

by surprise and thankfulness.

Table 5 deals explicitly with the interrelationships

among the discrete emotions. We chose to analyze the

correlations between the “during” and “after” recalled

feelings for both samples. Since these were all single-item

measures, it is worthwhile to begin by pointing out that the

diagonals in the matrices indicate very high correlations

between the measures at the two time points. Since these

are conceptually separated by the experience of the CUI,

these are not quite test-retest correlations, so they

indicate at best a basement measure of reliability, and we

judge those measures to be satisfactory. Those diagonal

coefficients indicate a considerable stability of feelings

during and after the CUIs.

The emotions differed in their levels of mutual

connectedness. Boredom, for example, was not associated

with any of the other feelings (and Table 3 indicates that

this feeling was pretty much never associated with CUIs).

Concern for self-image was also largely independent of the

other feelings, except for guilt and disappointment. All of

the other self-reported discrete emotions had substantial

connections with other elements of our battery. We will

take notice here of the connections that seem most

interesting to us.

Self-reported uncertainty, one of the more salient

feelings in Table 3, was directly connected to several other

emotions. In both samples, higher feelings of uncertainty

predicted more anger, disappointment, sadness, guilt,

surprise, and worry. More certainty produced more happiness

and thankfulness.

Although Table 3 does not indicate that anger was very

commonly felt, Table 5 shows that it was emotionally

dangerous. High anger predicted more uncertainty,

disappointment, sadness, surprise, and worry. Lower anger

was associated with more happiness and thankfulness.

Happiness and thankfulness regarding the conversation

might well be two epiphenomena of successful interventions.

They have fairly similar emotional profiles here. Greater

happiness and thankfulness were associated with less anger,

uncertainty, disappointment, sadness, guilt, and worry.

This is a coherent emotional profile for helpers to target,

and possibly an entry at any point in this dynamic pattern

will ripple out to the other feelings.

Table 5 reports many more associations but we will

leave these to interested readers. Our overall answer to

the second research question is as follows. CUIs tended to

be experienced as having fairly routine (or absent) levels

of emotions, except for uncertainty, argumentativeness,

politeness, and thankfulness, all of which seem encouraging

for the project of intervention. In fact, these tend to be

positively correlated and so might constitute a single

experiential density. Uncertainty about the conversation, a

defining feature of CUIs, was one of the most central

feelings in our emotional matrices (Table 5). It had

implications or associations for all the other discrete

emotions except boredom. Our theoretical orientation

implies that interventions should aim at the absence of

scripts for CUIs, but here is empirical evidence that this

is also a central feature of people’s experience of them.

Research Question 3: Associations between CUI

Characteristics and Uncertainties, Feelings, Rumination, and

Outcomes

Our final research question seeks to connect the

leading CUI characteristics (those that we coded) with the

self-reports we also gathered. To simplify the analyses and

increase cell counts (and because we have not been impressed

with many of the differences between our samples), we

combined our two samples for these analyses, which are

summarized in Tables 6, 7, and 8. We found it most useful

to combine many of our codes and to focus on those that

seemed most informative and frequent. Within each focused

set of comparisons, we sought to determine whether there

were mean differences on the closed-ended measures.

Table 6 shows how the various self-reports varied

according to the general topic category of the CUI.

Respondents’ certainty was highest (and higher than the

theoretical midpoint, 5.5) for CUIs dealing with academic,

workplace, and financial matters. The greatest uncertainty

was reported when the conversations concerned sex, medical,

or death issues. This was roughly the same pattern

regardless of whether certainty was being reported as felt

before or during the CUI. The other person’s levels of

certainty followed almost the identical pattern, indicating

that topics had the same level of estimated scriptedness for

both participants. Politeness and rudeness had opposite

patterns, of course, with the death, work, financial, and

medical conversations being conducted the most pleasantly.

The CUIs that were least characterized by reason-giving were

those dealing with death. All the other topics had scores

well above the theoretical midpoint. Death was also least

conflictive, and romantic relationships stood out as having

the most disagreement. Topic was not significantly

associated with the degree to which people reviewed or

mulled over the CUIs afterwards. Table 6 also reports all

the emotional effects. Nearly all of these were

statistically significant. Details are left to the readers,

especially those particularly interested in one of the

topical categories.

Table 7 is parallel to Table 6, except that the topical

categories have been replaced with other codings of interest

and reasonably balanced frequencies. Because so many of the

comparisons were weak and essentially null, only the means

for the significant comparisons are displayed.

A glance shows that the codes that distinguished the

various self-reported reactions most clearly were those

involved with consequences, which were reduced to

relationship between the parties, health concerns, and

issues about someone’s future. CUIs that bore on

interpersonal relationships were associated with the

greatest uncertainty about how to proceed, and people

reported more security in their thinking about their

futures. Relational consequences also led to the least

politeness and most conflict. The emotional patterns are

not simple to summarize, and so are again left to the

reader.

We were able to code whether the respondent conversed

from a personal identity or an institutional one (e.g.,

teacher, subordinate). Participating in a CUI from a

personal footing meant that the other person was more

uncertain about the conversation, the interaction was less

polite, and was more conflictive. Prior to the engagement,

people acting out of their personal identities felt more

uncertainty, disappointment, sadness, and guilt. After the

CUI was finished, people in personal identities still had

more anger, sadness, and guilt.

Finally we simplified our respondent’s reaction codes

to indicate merely whether the person conversed

cooperatively or not (e.g., avoided the topic). Cooperation

was associated with the other person being more certain how

the conversation would go, more politeness, and less

conflict. Cooperative respondents reported having less

anger and more happiness and thankfulness prior to the CUI.

After the CUI, cooperative people self-reported less anger,

uncertainty, and worry, and more happiness and thankfulness.

The least important of these codes was channel.

Whether the CUI occurred face-to-face or not seemed to make

little difference. Whether the CUI’s setting was public or

private was not important across the board, but private

exchanges were nonetheless associated with more reflection

afterwards, more politeness during the conversation, less

conflict, and less anger, uncertainty, and disappointment

prior to the exchange.

Finally, Table 8 reports results of several ANOVAs

using relational effects as a predictor variable. This was

respondents’ self-report of the CUI’s effect on the

relationship between the people involved. The categories

were the relationship ended, changed importantly for the

worse, got a little worse, continued unchanged, improved a

little, and changed in an important way for the better.

Table 8 shows that relational outcome was generally

predictive of the other self-reports. These tended to be

ordered by the positivity of the relational outcome, but it

should be noticed that at many points respondents seemed to

favor stability or small changes in the relationships over

larger improvements. Certainty was an exception, because

for several of the measures, ending the relationship was

associated with great clarity. Argumentativeness and

conflictiveness were also high when the relationship ended

as a result of the CUI. Thankfulness had an interesting

curvilinear pattern, being highest when the relationship

ended or improved the most. Many of the other emotional

reports, however, showed improvement in valence as the

relational outcome improved.

In sum, Tables 6, 7, and 8 point us to this answer to

our final research question. The CUI topics affected nearly

every self-reported variable we measured, suggesting that

this would be a productive way to begin dividing up the

community CUI research project. Topics dealing with sex,

medical issues, and death generated the greatest uncertainty

in our data sets. Our consequence codes were almost as

important as topic in distinguishing the various feelings

and impressions that respondents communicated to us. Here,

the relationship issues were associated with the least

certainty about how to proceed. We also noticed that

cooperative patterns of engagement by respondents seemed to

be the most constructive. Ending the relationship was

sometimes met with certainty, relief, and thankfulness.

Improving the relationship also had some positive emotional

associations.

Discussion

This has been an exploratory investigation of what we

propose as an interesting and potentially productive idea,

the consequential unscripted interaction (CUI). More or

less by definition, these identify opportunities to improve

people’s lives if we can convey useful interactional scripts

or training to them. Although this paper has taken a

theoretical and abstract tone, our real ambition is for

researchers to identify CUIs, perhaps from the descriptions

we have given, and move on to teach people more confident

and useful ways of interacting.

We have listed the topics we found in two different

samples, as well as the consequences that made the

conversations important to our informants. Our

instrumentation might be useful in developing and testing

interventions.

Almost all respondents in both samples (only one

respondent explicitly said he/she had never encountered a

CUI) reported that they had been involved in a CUI. They

seemed to have little trouble reporting an experience in

which they felt they had an important conversation for which

they did not have a script. Thus, it seems that CUIs are

fairly common phenomena in everyday life, and so developing

people’s skills to handle those situations probably will be

very helpful.

In both samples, the CUI experiences tended to be

negative, with topics dealing with death, illness,

relational threats, and so forth. Some people do not have

scripts available for these important but negative topics,

perhaps because people are generally reluctant to discuss

negative information due to the discomfort associated with

it (Conlee & Tesser, 1973), and thus have few experiences,

either direct or virtual. For example, romantic

relationship dissolution were found to be the most mentioned

topic (more than 10% for both samples), followed by romantic

relationship maintenance (also more than 10% for both

samples, mostly about complaints in romantic relationships).

Negative medical-related information and conversation about

death were also mentioned repeatedly in both samples.

Relatedly, data indicated that almost one third of

respondents had negative emotions, with only 2.3% of the

undergraduates and 0.4% of the mTurk respondents reporting

positive emotions. This result may indicate that CUI issues

naturally induce negative emotions or that CUIs are so

difficult to handle that some people cannot do them, thus

making training for CUIs useful.

It is also interesting that CUIs were reported to

happen more often in private settings, such as residences,

about private matters, and to affect interpersonal

relationships. People tend to have more developed schemas

and scripts for familiar topics (Honeycutt & Bryan, 2011;

Schank, 1982), so that the schema-activation process should

start most effortlessly when encountering a situation that

fits previously developed schemas. It can be expected that

people normally have much experience dealing with personal

issues (e.g., regarding mTurk sample’s average age, they

should have at least some experience of romantic

relationship maintenance), which should foster more

developed corresponding schemas and scripts. Interventions

might therefore attempt to scaffold new scripts from

familiar ones.

References

Browning, D. M., Meyer, E. C., Truog, R. D., & Solomon, M.

Z. (2007). Difficult conversations in health care:

Cultivating relational learning to address the hidden

curriculum. Academic Medicine, 82, 905-913.

Cardemil, E. V., & Battle, C. L. (2003). Guess who’s coming

to therapy? Getting comfortable with conversations

about race and ethnicity in psychotherapy. Professional

Psychology: Research and Practice, 34, 278-286.

Conlee, M. C., & Tesser, A. (1973). The effects of

recipient desire to hear on news transmission.

Sociometry, 36, 588-599.

Copenhaver-Johnson, J. (2006). Talking to children about

race: The importance of inviting difficult

conversations. Childhood Education, 83, 12-22. doi:

10.1080/00094056.2006.10522869

Copes, H., Hochstetler, A., & Cherbonneau, M. (2012).

Getting the upper hand: Scripts for managing victim

resistance in carjackings. Journal of Research in Crime and

Delinquency, 49, 249-268. doi: 10.1177/0022427810397949

Hample, D. (in press; 2015). A theory of interpersonal

goals and situations. Communication Research. doi:

10.1177/0093650214565899.

Honeycutt, J. M., & Bryan, S. P. (2011). Scripts and

communication for relationships. New York: Peter Lang.

Honeycutt, J. M., Cantrill, J. G., & Greene, R. W. (1989).

Memory Structures for Relational Escalation A Cognitive

Test of the Sequencing of Relational Actions and

Stages. Human Communication Research, 16, 62-90.

Keating, D. M., Russell, J. C., Cornacchione, J., & Smith,

S. (2013). Family communication patterns and difficult

family conversations. Journal of Applied Communication

Research, 41, 160-180. doi: 10.1080/00909882.2013.781659

Kellermann, K., Broetzmann, S., Lim, T. S., & Kitao, K.

(1989). The conversation MOP: Scenes in the stream of

discourse. Discourse Processes, 12, 27-61.

Knobloch, L. K., & Theiss, J. A. (2012). Experiences of US

military couples during the post-deployment transition

Applying the relational turbulence model. Journal of Social

and Personal Relationships, 29, 423-450.

Lamiani, G., Meyer, E. C., Browning, D. M., Brodsky, D., &

Todres, I. D. (2009). Analysis of enacted difficult

conversations in neonatal intensive care. Journal of

Perinatology, 29, 310-316.

Lamiani, G., Meyer, E. C., Leone, D., Vegni, E., Browning,

D. M., Rider, E. A., Truog, R. D., & Moja, E. A.

(2011). Cross-cultural adaptation of an innovative

approach to learning about difficult conversations in

healthcare. Medical Teacher, 33, e57-e64.

Makoul, G., Zick, A. B., Aakhus, M., Neely, K. J., & Roemer,

P. E. (2010). Using an online forum to encourage

reflection about difficult conversations in medicine.

Patient Education and Counseling, 79, 83-86.

Manning, J. (in press). Coming out conversations and

gay/bisexual men’s sexual health: A constitutive model

study. In V. L. Harvey, & T. H. Housel (Eds.), Health

care disparities and the LGBT population. Lanham, MD: Peter

Lang.

Schank, R. C. (1982). Dynamic memory. Cambridge: Cambridge

University Press.

Smith, S. W., Cornacchione, J., Keating, D. M., & Russell,

J. (2014, May). The impact of multiple goals and discrete

emotions on attitudes toward, and engagement in, difficult family

conversations. Paper presented to the annual meeting of

the International Communication Association, Seattle,

WA.

Stone, D., Patton, B., & Heen, S. (1999). Difficult conversations:

How to discuss what matters most. New York: Penguin.

Villagran, M., Goldsmith, J., Wittenberg-Lyles, E., &

Baldwin, P. (2010). Creating COMFORT: A communication-

based model for breaking bad news. Communication

Education, 59, 220-234. doi: 10.1080/03634521003624031

Wayne, T. (Sept. 1, 2013). Explaining twerking to your

parents. New York Times, p. SR2.

Wilson, S. R., Wilkum, K., Chernichky, S. M., MacDermid

Wadswroth, S. M., & Broniarczyk, K. M. (2011).

Passport toward success: Description and evaluation of

a program designed to help children and families

reconnect after a military deployment. Journal of Applied

Communication Research, 39, 223-249. doi:

10.1080/00909882.2011.585399.

Table 1: Coding of Open-Ended CUI Topic Descriptions

Krippendorf’s Undergrads mTurkalpha f % f %

Topic .73Sex (16) (6.4) (15) (6.6) Sexual orientation 3 1.2 10

4.3 Sex education 8 3.1 4 1.7 Sex talk 2 0.8 0 - Sex other 3 1.2 1 0.4Illegal Activity (9) (3.6) (8)

(3.5) Of respondent 3 1.2 3 1.3 Of other 0 - 1 0.4 Of third party 2 0.8 2

0.9Religion 4 1.6 2 0.9Medical (22) (8.8) (36)

(15.9) Addiction 2 0.8 8 3.5 Mental illness 10 3.9 5 2.2 Physical illness/injury 10 3.9 12

5.2 Old age 0 - 4 1.7 Other 0 - 7 3.0Finances (22) (8.8) (30)

(13.2) Employment 0 - 5 2.2 Debt 3 1.2 9 3.9 College 15 5.8 4 1.7 Purchase or bargaining 0 - 1 0.4 Other 4 1.6 11 4.8Romantic Relationship (63) (25.1) (54)

(23.8) Initiation 4 1.6 2 0.9 Escalation 1 0.4 0 - Dissolution 31 12.1 26 11.3

Negotiate or maintain 27 10.5 26 11.3Work/Job (25) (10.0) (25)

(11.0) Salary 0 - 3 1.3 Resign 1 0.4 5 2.2 Complaint 2 0.8 4 1.7 Transfer/change job 2 0.8 3

1.3 Other 20 7.8 10 4.3Academic (34) (13.5) (12) (5.3) College choice 13 5.1 2 0.9 Grades 3 1.2 0 - Failing school 1 0.4 1 0.4 Complaint 1 0.4 0 - Other 16 6.2 9 3.9Death (18) (7.2) (9)

(4.0) Own reaction 13 5.1 2 0.9 Another’s reaction 1 0.4 3

1.3 Explain to another 4 1.6 4

1.7Other 42 16.3 33 14.3

Table 2: Coding of Open-Ended CUI Descriptions Except for Topic

Krippendorf’s Undergrads mTurkalpha f % f %

Did Conversation Happen? .78 Yes 248 96.5 213 92.6 No 6 2.3 16 7.0Public Private Setting .79 Public 19 7.4 9 3.9 Private 51 19.8 34 14.8 Unclear 181 70.4 183 79.6Respondent’s Setting .94 Academic 15 5.8 2 0.9 Residence 35 13.6 15 6.5 Commercial 5 1.9 3 1.3 Own workplace 11 4.3 22 9.6 Mobile (e.g., car) 3 1.2 3

1.3 Healthcare facility 3 1.2 8

3.5 Other/unclear 179 69.6 174

75.7Channel .91 Face to face 119 46.3 77 33.5 Telephone 12 4.7 11 4.8 Text message 1 0.4 1 0.4 Social network private msg 2 0.8 0

- Other 117 45.5 138 60.0Personal/institutional ID .88 Both personal identity 207 80.5 185 80.4 R institutional, O personal 4 1.6 1

0.4 R personal, O institutional 23 8.9 13

5.7 Both institutional identity 17 6.6 26

11.3

Relevant to a Decision? .71 Yes 81 31.5 60 26.1 No 168 65.4 167 72.6Respondent Reaction .70 Cooperative 68 26.5 48 20.9 Uncoop, cut off conversation 1 0.4 0

- Uncoop, avoided conversation 2 0.8 3

1.3 Uncoop, made no response 7 2.7 6

2.6 Uncoop, other 49 19.1 53 23.0 Unclear 124 48.2 117 50.9Respondent Emotion .72 Positive 6 2.3 1 0.4 Negative 71 27.6 85 37.0 Neutral or unclear 174 67.7 141

61.3Key Consequentiality .71 Own self image 10 3.9 7 3.0 Other’s self image 1 0.4 1

0.4 Relationship betw the two 94 36.6 89

38.7 Relationship w 3d party 13 5.1 10

4.3 Own resources 9 3.5 8 3.5 Other’s resources 2 0.8 2 0.9 Joint resources 3 1.2 3 1.3 Third party resources 1 0.4 2 0.9 Own health 17 6.6 6 2.6 Other’s health 4 1.6 14 6.1 Third party’s health 3 1.2 5

2.2 Future: career 20 7.8 22 9.6 Future: academic 37 14.4 13 5.7 Future: family status 3 1.2 11

4.8 Future: finances 3 1.2 17 7.4

Other 31 12.1 14 6.1

Table 3: Descriptive Statistics for Closed-Ended Measures

No. Undergrads mTurkItems alpha Mean SD alpha

Mean SD tSelf Certain Before 3 .83 5.62 2.65 .79 5.97 2.62

1.47Other Certain Before 3 .83 5.66 2.67 .86 4.96

2.77 2.86**Self Certain During 3 .83 4.35 2.39 .82 4.70 2.50

1.57Other Certain During 3 .90 5.51 2.53 .93 5.03

2.76 1.99*Reflected After 7 .78 6.41 1.86 .74 6.12 1.86

1.68Politeness 1 - 7.09 3.00 - 6.77 3.18

1.14Rudeness 1 - 3.48 3.07 - 3.73 3.40 0.86Argumentative 1 - 7.91 2.26 - 7.85 2.51

0.29Conflictive 1 - 4.71 3.21 - 5.59 3.47

2.88**During CUI Angry 1 - 4.57 3.39 - 4.63 3.56 0.21 Uncertain 1 - 6.91 2.78 - 7.04 2.92

0.50 Disappointed in Self 1 - 4.44 3.18 - 4.77

3.40 1.13 Happy w Conversation 1 - 4.14 2.91 - 3.66 2.93

1.83 Sad b/c Conversation 1 - 6.69 3.21 - 7.12 3.08

1.50 Guilty 1 - 4.80 3.33 - 5.34 3.31 1.80 Surprised 1 - 6.21 2.85 - 5.75 3.11

1.67 Worried 1 - 7.30 2.47 - 7.60 2.47 1.34 Thankful 1 - 6.33 3.11 - 6.11 3.32

0.76

Bored 1 - 2.13 1.74 - 2.15 2.060.14

Concerned for Self Image 1 - 5.11 3.29 - 4.433.26 2.26*

Carefully Edited Messages 1 - 5.86 3.06 - 5.803.12 0.21

After CUI Angry 1 - 4.31 3.41 - 4.62 3.60 0.96 Uncertain 1 - 6.02 3.24 - 6.19 3.26

0.56 Disappointed in Self 1 - 4.19 3.20 - 4.53

3.43 1.13 Happy w Conversation 1 - 4.72 3.10 - 4.06 3.18

2.32* Sad b/c Conversation 1 - 6.16 3.36 - 6.78 3.25

2.06* Guilty 1 - 4.56 3.28 - 4.96 3.44 1.30 Surprised 1 - 6.05 2.91 - 5.79 3.16

0.94 Worried 1 - 5.86 3.24 - 6.27 3.27 1.39 Thankful 1 - 6.28 3.12 - 5.96 3.43

1.09 Bored 1 - 1.95 1.49 - 1.91 1.89

0.21 Concerned for Self Image 1 - 3.89 3.07 - 3.96

3.17 0.23Note. N for the undergraduate sample was 257, and for the mTurk sample was 230. t tests are between the two samples, and df = 485, except when correction for unequal variances had to be made.

* p < .05 ** p < .01*** p < .001

Table 4: Correlations Among Closed-Ended Self-Report Measures, Except Emotions

1 2 3 4 5 6 7 8

Undergraduates1 Self Certain Before2 Other Certain Before .153 Self Certain During .63 .104 Other Certain During .14 .71 .185 Reflective After .05 .08 -.03 .176 Politeness .12 .28 .08 .11 -.137 Rudeness -.01 -.17 -.03 -.06 .15 -.808 Argumentative .21 .16 .14 .16 .18 .09 .049 Conflictive .06 -.12 -.01 -.05 .25 -.63 .64

.21

mTurk Workers1 Self Certain Before2 Other Certain Before .013 Self Certain During .71 .134 Other Certain During .07 .78 .265 Reflective After -.04 -.02 -.15 .036 Politeness .18 .33 .11 .26 .037 Rudeness -.02 -.24 .06 -.11 .08 -.798 Argumentative .24 .07 .12 .07 .24 .17 .019 Conflictive .10 -.14 .09 -.06 .20 -.42 .57

.25Note. For the undergraduate sample, correlations of |.13| or higher are significant at p < .05, two-tailed. For the mTurk sample, the parallel figure is |.14|.

Table 5: Correlations Among Closed-Ended Self-Report Measures of Emotions

During the CUI 1 2 3 4 5 6 7 8 9 10 11

UndergraduatesAfter CUI1 Anger .80 .32 .21 -.44 .38 .09 .22 .28 -.44 .01 -.062 Uncert .33 .56 .34 -.50 .47 .18 .32 .42 -.36 -.07 .003 Disapp .20 .35 .76 -.42 .35 .46 .13 .27 -.27 .01 .244 Happy -.38 -.35 -.20 .78 -.50 -.13 -.17 -.27 .63 .02 .085 Sad .37 .41 .27 -.54 .87 .31 .25 .43 -.32 -.09 -.066 Guilty .13 .22 .48 -.38 .42 .77.10 .24 -.18 .08 .247 Surpr .17 .32 .10 -.25 .25 .10.67 .34 -.17 -.09 .098 Worry .37 .44 .41 -.53 .53 .23 .35 .55 -.38 .01 .129 Thank -.35 -.15 -.19 .52 -.20 -.02 -.14 -.12 .85 -.07 -.0210 Bored .07 -.07 .11 -.00 .11 .03 -.20 -.18 -.10 .64 .1511 Self .04 .21 .41 -.21 .16 .34 .10 .15 -.15 .20 .65

mTurk WorkersAfter CUI1 Anger .81 .21 .08 -.31 .39 .14 .34 .12 -.30 .06 -.08

2 Uncert .30 .57 .30 -.37 .37 .26 .26 .37 -.24 -.06 .163 Disapp .18 .34 .77 -.23 .42 .67 .21 .32 -.18 .14 .364 Happy -.28 -.10 -.17 .75 -.41 -.17 -.11 -.09 .58 .17 .085 Sad .26 .24 .27 -.49 .83 .39.21 .44 -.23 -.18 .026 Guilty .11 .26 .58 -.19 .50 .81 .17 .38 -.06 .02 .307 Surpr .23 .34 .16 -.08 .14 .12 .72 .32 -.06 -.07 .028 Worry .26 .41 .28 -.42 .44 .26 .34 .48 -.27 -.12 .089 Thank -.24 -.05 -.09 .55 -.25 -.09 -.09 .01 .83 .01 .0610 Bored .06 -.15 .10 .23 -.11 .03 -.02 -.24 -.07 .86 .1911 Self .00 .19 .42 -.04 .12 .35.16 .21 -.07 .19 .73Note. For the undergraduate sample, correlations of |.13| or higher are significant at p < .05, two-tailed. For the mTurk sample, the parallel figure is |.14|.

Table 6: Mean Differences by Topic

Sex Med- Fin- Rom- Work Aca- Deathical ances antic demic

p <Self Certainty Before CUI 4.7a 5.1a 6.6bc 5.6ab

6.9c 7.0c 4.5a .001Other Certainty Before CUI 4.2a 4.6a 6.3c 4.8a 6.5c 6.0bc

5.0ab .001Self Certainty During CUI 3.4a 4.1abc 5.2c 4.6bc

5.1bc 5.2c 4.0ab .01Other Certainty During CUI 4.6a 4.3a 5.9bc 4.9ab

6.6c 6.1c 4.7a .001Reflected Afterwards 6.1ab 6.3ab 6.0ab

6.6b 6.1ab 6.3ab 5.6a =.11Polite During 6.8ab 7.6bc 7.6bc 6.0a

8.2c 7.5bc 8.5c .001Rude During 3.4bc 2.7ab 3.3bc 4.4c

2.7ab 3.0ab 1.9a .001Argumentative During 7.4b 7.7b 8.3b 8.4b 8.1b 8.0b 6.2a

.001Conflictive During 4.6b 4.2b 4.7b 6.5c 4.0b 5.0b 2.6a

.001During CUIAnger 4.6ab 4.2ab 3.9ab 5.5b

3.7a 3.4a 4.4ab .01Uncertain 6.8ab 7.8b 6.9ab 7.2b 5.8a 6.9ab

6.8ab .05Disappointed 3.6ab 4.6bc 5.6c 5.0bc

3.8ab 5.3c 2.5a .001Happy 4.5cd 3.2ab 4.5cd 3.5bc

5.1d 5.2d 2.1a .001Sad 5.2a 8.4c 6.6b 80c 4.6a 5.1a 8.5c .001Guilty 4.3ab 5.4b 5.3b 5.5b 4.0ab

4.7ab 3.6a .05Surprised 7.0b 5.7a 6.0ab 6.2ab 5.3a 5.5a

6.5ab =.09Worried 7.7b 8.1b 7.6b 7.6b 7.0ab 6.4a 7.8b

.01

Thankful 6.3ab 6.0ab 6.8b 6.8b 6.1ab6.7b 5.2a =.21

Bored 1.9ab 1.9ab 2.2ab 1.9ab2.7b 2.7b 1.5a .05

Concerned Neg Impression 4.6bc 3.7b 4.8bc 4.9bc6.3d 5.5cd 2.3a .001

Carefully Edited 6.1bc 5.1ab 5.6b 6.1bc7.0c 5.9bc 4.2a .01

After CUIAnger 4.2ab 4.0ab 4.3ab 5.1b

3.4a 3.2a 4.3ab .05Uncertain 4.9a 6.7b 5.8ab 6.3ab 5.1a 5.3ab

6.4ab .05Disappointed 3.0ab 4.3bc 5.1c 4.5c

3.7abc 5.1c 2.6a .01Happy 5.7d 4.0ab 4.8bcd 4.1abc

5.3cd 5.8d 2.8a .001Sad 4.9ab 7.8de 6.1bc 7.3cd

4.3a 4.7a 9.0e .001Guilty 3.5a 5.1bc 5.2c 5.1bc 3.5a

4.3abc 3.7ab .01Surprised 67.9b 5.9ab 5.8ab 6.1ab

5.0a 4.9a 6.6b .05Worried 5.5ab 7.1c 5.5ab 6.3bc 5.2ab

4.8a 6.5bc .01Thankful 6.1a 6.1a 6.6a 6.6a 6.0a 6.8a 5.9a =.70Bored 1.9abc 1.6a 2.0abc 1.7ab

2.4bc 2.6c 1.3a .01Concerned Neg Impression 3.7b 3.2b 3.9b 4.4b 4.5b 4.4b 1.5a

.001Note. For all analyses, df = 6/374. Means with the same letter within a row did not differ significantly according to Duncan post hoc tests at p < .05. The p values refer to the overall test of the one-way ANOVA. Entrees in that column beginning with “=” were not statistically significant.

Table 7: Mean Differences by Non-Topic Codes

Public/Private Channel R’s Identity Reaction ConsequencesPub Priv ftf Nftf Pers Instit Coop NCoopRelatn Health Fut

SCertB 5.7b 4.9a6.7c

OCertB 5.6 4.1 5.2 6.4 4.6a 4.6a 6.4bSCertDOCertD 5.1 6.7 5.7 5.1 4.8a 4.4a 6.4bReflec 7.3 6.3Polite 6.0 8.2 6.8 8.2 8.1 6.6 5.7a 7.7b

8.0bRude 4.5 2.4 2.2 4.0 4.7b 2.7a 2.6aArgtvConfl 5.4 3.9 5.3 4.1 4.0 5.5 6.4b 4.4a

4.4aDuringAngry 5.8 3.9 3.5 5.0 5.5b 4.3a

3.4aUncert 8.3 6.8 7.1 5.7Disapp 5.6 4.0 4.4 5.8 4.7 3.7Happy 4.5 3.7 3.7a 3.5a

4.7bSad 7.1 4.9 7.7b 7.8b 5.4aGuilty 5.1 4.1 5.8b 5.6ab

4.8aSurprWorry 7.6ab

8.3b 7.1aThank 6.9 6.0BoredNeg Impr 4.6 6.1 4.8ab 3.9a

5.3bEdited 5.7 6.7AfterAnger 4.6 3.5 3.2 4.9 5.3b 3.9a

3.4a

Uncert 5.5 6.3Disapp 4.1 5.8Happy 5.3 4.1 4.3a 4.2a

5.3bSad 6.7 4.6 7.1b 7.2b 5.0aGuilty 4.9 3.5SurprWorry 5.3 6.3 6.3ab

7.0b 5.5aThank 6.9 5.9Bored 1.9 2.4 1.6 2.0 1.7a 1.9ab

2.2bNeg ImprNote. Only when the ANOVA returned at statistically significant result are the means for a coded variable’s categories reported. Blanks indicate that the test was insignificant. The row variables are in the same order as in Table 6. Public/Private refers to whether the setting was public or private. Channel was recoded into face-to-face or not. Identities were recoded to indicate whether the respondent acted in his/her personal identity or an institutional one. Respondents’ reactions were recoded intocooperative or not. Consequences were recoded into relationship, health, and future issues, and the other categories omitted. Duncan post hoc tests were done for consequences.Table 8: Self-Report Means by Relational Outcomes

End Imp Little No Imp ImpWorse Worse Dif Little Bettr

p <Self Cert Before 5.7abc 4.8a 5.2ab 5.8bc 6.4c

5.9bc =.07Other Cert Before 4.9a 4.9a 5.0ab 6.0b 5.5ab 5.1ab

.05Self Cert During 4.7b 3.8a 3.6a 4.8b 4.9b 4.3ab .05Other Cert During 5.7b 5.7b 4.4a 5.7b 5.1ab 4.9ab

.05Reflected After 6.6a 6.0a 6.5a 6.0a 6.3a 6.3a =.19

Polite During 5.0ab 4.4a 5.6b 7.6c 7.6c 7.9b .001Rude During 5.5c 5.7c 4.3b 2.8a 3.0a 2.9a .001Argument During 8.2a 8.1a 7.3a 7.4a 8.2a 8.0a .05Conflict During 7.3c 7.5c 5.9b 4.3a 4.7a 4.2a .001During CUIAngry 6.6c 6.6c 5.2b 3.8a 4.1ab 3.8a .001Uncertain 7.3a 7.0a 7.6a 7.0a 6.8a 6.8a =.58Disappointed 4.9a 4.5a 5.5a 4.6a 4.7a 4.2a =.42Happy 2.9a 2.1a 2.7a 3.9b 5.0b 4.6b .001Sad 8.2b 8.8b 8.3b 6.0a 6.0a 6.6a .001Guilty 5.3ab 5.5ab 6.1b 4.5a 5.0ab

5.0ab =.19Surprised 6.1a 6.6a 6.3a 5.4a 6.1a 6.2a =.22Worried 7.9bc 7.9bc 8.4c 6.8a 7.1ab

7.6abc .01Thankful 6.0c 3.3a 4.5b 5.4bc 7.3d 7.5d .001Bored 1.9a 2.2ab 1.8a 2.9b 2.2ab 1.8a

.001Concern Neg Image 5.2a 4.7a 4.9a 4.8a 5.0a 4.5a =.69Carefully Edited 6.0a 5.1a 6.4a 5.7a 6.0a 5.8a =.60After CUIAngry 6.7bc 7.3c 6.0b 3.8a 3.9a 3.1a .001Uncertain 7.1b 7.2b 7.7b 5.8a 5.4a 5.6a .001Disappointed 4.9ab 4.9ab 6.0b 4.4a 4.1a 3.6a

.001Happy 3.2b 2.0a 3.5b 4.2b 5.4c 5.4c .001Sad 7.7b 8.8b 7.9b 5.9a 5.4a 6.0a .001Guilty 5.2ab 5.1ab 5.8b 4.4ab 4.3a

4.7ab =.21Surprised 6.4a 6.6a 5.9a 5.5a 6.2a 5.8a =.25Worried 6.9bc 7.7cd 8.1d 5.9ab 5.0a 5.5a

.001Thankful 5.7b 3.6a 4.6ab 5.4b 7.1c 7.3c .001Bored 1.6a 2.1a 1.8a 2.7b 1.9a 1.5a .001Concern Neg Image 4.6b 4.5b 4.7b 4.2ab 3.6ab 3.2a

.01Note. Means with the same letter within a row did not differsignificantly according to Duncan post hoc tests at p < .05.The p values refer to the overall test of the one-way ANOVA.

Entrees in that column beginning with “=” were not statistically significant.