B.Ev-D A Coding Scheme for
Descriptions of Behavioral Events

Bertram F. Malle, University of Oregon

Version 1.3, April 1997


Back to Current Research at the Social Interaction Lab

Back to Social Interaction Lab home page


The following coding system is used to classify the types of behaviors people describe in everyday language (Malle & Pearce, in press). The coding is limited to concrete behavioral/psychological events--that is, actions, thoughts, and feelings. We thus exclude dispositions ("she is intelligent," "he is handsome") and demographic characteristics ("she is only 19 years old") and focus instead on events that are typically described by verbs. For example, if we asked a person (let's call her A) to describe (from the observer perspective) her partner's experiences during a particular short getting-to-know interaction, A might say:

    B seemed fairly comfortable talking to me, considering the circumstances. She didn't make a lot of eye contact, but the conversation flowed smoothly. She asked a lot of questions about school...
To code the behavioral events that A describes, we have to focus on the verbs that she uses (most behavioral events are described using verbs). Among these verbs we have to rule out (a) those that do not refer to behavioral events during the interaction of interest (i.e., those that happened before or after the situation of interest) and (b) verbs that don't have the other person or the self as their agent (e.g., `we,' the impersonal `you' and `it,' things and processes). (c) As mentioned before, we need to exclude those expressions that refer to general dispositions (often formulated in present tense, in contrast to events that happened during the interaction ,which are formulated in past tense).

Despite the dominance of verbs, some events are expressed in nouns (e.g., "my questions," "her answers," "his eye-contact," "my posture"), some in adjectives (e.g., "she was anxious," "I was comfortable").

Once we have identified the behavioral event of interest, we need to code them for perspective (self vs. other, or sometimes called actor vs. observer) and for event type (observable vs. unobservable as well as intentional vs. unintentional). These codes are described in turn (see B.Ev for greater detail on these general codes).

1. Who is the Agent--Self or Other?

We must first determine whether a given behavioral event is ascribed to self (actor perspective) or to other (observer perspective). Clear contrast cases are "I was frustrated" (self/actor) and "she tapped on the table" (other/observer). The grammatical subject of the verb typically indicates the agent: First-person propositions indicate self, third-person propositions indicate other. Exceptions include statements in passive voice (which are ruled uncodeable; see below) or propositions such as "my heart rate increased," which is a self-ascribed event, even though the grammatical agent is not a person. Statements starting with it, such as "it was rather uncomfortable," are codeable only if the context makes clear that the speaker refers to him- or herself rather than to the whole situation. "It was fun to talk to somebody about my home town" clearly refers to joy that the speaker experienced herself.

2. Was the Event Observable?

Psychology has traditionally distinguished between publicly observable events (e.g., behaviors, actions) and publicly unobservable events (e.g., bodily sensations, feelings and thoughts). An event is publicly unobservable when only the agent himself can, in principle, perceive it (e.g., "I was very frustrated").

Note that many emotions, though primarily unobservable, have observable "symptoms." The person who was frustrated in the previous example might have subsequently displayed her frustration through (observable) facial expressions; the original event, however--the frustration itself--is still unobservable. Even though the person may have shown observable symptoms of frustration, such symptoms were not the event that was described. Similarly, "She seemed very insecure when talking" refers to an unobservable event. The observer may have noted observable signs of insecurity in the other person, but what he describes is not these signs but the insecurity itself (an unobservable event). Likewise, "being upset" is usually coded as an internal event both from the actor and the observer viewpoint, even though an observer uses observable clues to infer this event, whereas the actor often simply feels it.

Further examples of unobservable events include: "I wanted to change topics." "I didn't like her sitting so close." "She seemed pleased with my comment." "I did detect a slight uneasiness."

Examples of observable events include: "I asked many questions." "I began to relax." "He constantly switched topics." "He was fidgeting in his chair."

3. Was the Event Intentional?

An action is intentional if the agent chose/decided to do it. For example, "He told me about a ski accident" refers to an intentional event; apparently, the person chose to talk about that accident. By contrast, "She was blushing" or "She understood my point" are not intentional because one hardly chooses to blush or to understand.

Examples of unintentional behaviors include: "I noticed that he was stuttering"; "...topics that interested me." "I ran out of things to say." "Her energy level was quite high." "She was also 8 minutes late." "His voice was very soft."

Examples of intentional behaviors include: "She started talking about people she knew here," "I tried moving away from that topic." "He basically just answered the questions I asked."

4. Putting it Together: Four Types of Events

The observability of a behavioral event and its intentionality are orthogonal and thus combine to produce four (2 x 2) types of behaviors (see Table 1).

Table 1.
Classification of behavioral events in folk explanations.

                            Intentional                  Unintentional

Observable                  1 (actions)                  2 (mere behaviors)    
Unobservable                3 (intentional thoughts)     4 (experiences)      

Behavior type 1 ("action") is observable and intentional. Examples include "I asked general questions." "It was easy to talk to him." "I had to lead the conversation." "He didn't talk about himself the entire time." "She made herself comfortable." "He was smiling more."

Behavior type 2 ("mere behavior") is observable and unintentional. Examples include "I was fidgety." "I started to relax little by little." "He seemed very relaxed." "Her torso moved forward (when she became surprised)."

Behavior type 3 ("intentional thought") is intentional and unobservable. Examples include "I had a hard time thinking about the conversation." "I tried following the names and who did what." "She seemed to enjoy visualizing Jakarta." "He was probably searching for answers and things to say."

Behavior type 4 ("experience") is unintentional and unobservable. These events include bodily states ("I was feeling tired." "I guess he felt a little uncomfortable"), sensations ("I noticed his scar."), emotions/feelings ("I really enjoyed talking to him." "He must have become frustrated with the situation"), and mere desires ("I wished he would stop smiling." "She may have desired to have a more in-depth conversation.")

5. Uncodeable Statements

To code the behavioral events that research particiapnts describe, we focus on the verbs that they use because almost all behavioral events are described using verbs. However, we have to rule out those verbs that do not refer to relevant behavioral events.

  • We need to exclude verbs that don't have the other person or the self as their agent (e.g., `we,' the impersonal `you' and `it,' as well as things and processes).

  • We need to exclude dispositions such as "I ramble" and general descriptions such as "her voice is quiet" because they don't refer to a concrete behavioral event. Dispositions are typically formulated in present tense, so present tense statements are generally uncodeable.

  • We also need to exclude descriptions of events that happened before or after the situation of interest (e.g., the conversation in the various Events studies) are uncodeable as well (e.g., "I feel like she...").

6. Specific Coding Problems

The coding of naturally occurring behavioral event descriptions is occasionally quite difficult. Some of the recurring difficulties are described below, along with coding suggestions.

"He seemed to ...," "She appeared to..."

Whenever a person uses these phrases that, it is an inference. Therefore, what is inferred cannot be observable because one wouldn't have to mark an observation with the phrase "he seemed." For example, one wouldn't say, "He seemed to wear a shirt" after having seen the person..

Consequently, "He seemed pretty relaxed" is event type 4 because the writer infers the internal state of relaxation in the other person. If the writer had said "He was pretty relaxed," event type 2 would be more appropriate.

The verbs "He seemed to" or "She appeared to" are markers for "I infer that he/she..." and are therefore not coded in these phrases. "He appeared to be nervous" is coded as event type 4. Similarly, "He appeared to be very sure of himself" is not coded at all because it describes a trait inference, and traits are not coded.

"To try to...," "To attempt to..."

These verbs are ambiguous in that they can refer either

(a) to a mental act of trying (i.e., an intention), as in "I tried to come up with new topics of conversation," which is coded as 3; or

(b) to a continued, effortful, intentional observable action, as in "I mainly tried to talk about things we had in common," which is coded as 1.

The coder must decide whether the writer is referring to the mental act alone or to the observable action, which was done with effort and often repeatedly. Foe example, "I tried to make sure we were both talking for an equal amount of time" refers more to the intention than to whatever action she took to fulfill the intention, thus 3.

The verbs "to appear to" or "to seem to" can help resolve this ambiguity because they indicate an inference, which then must be an unobservable event. For example, "He appeared to try to contribute..." is coded as 3 because the writer seems to refer to the other person's internal tryings. Similarly, "He seemed to be trying to be friendly" is coded as 3 because the writer infers the other person's intention. By contrast, "Throughout the conversation he tried to be friendly," is a repeated, effortful set of actions that should be coded as 1.

A similar reasoning applies to the verb "to be willing to." For example, "The fact that [...] made her seem more willing to do the experiment." Here the writer infers both to an internal state of willingness (probably a 4) and to the doing of the experiment (which is not coded because the event goes beyond the recorded conversation).

"It was difficult..." "it was easy..."

If writers modify a behavior by saying it was difficult, it introduces a problem for coders. Difficulty itself is not a behavioral event, just a modification of a given behavioral event. As a result, "it was difficult" should not be coded separately. For example, "It was a little difficult to keep creating conversation." Apparently, the person was creating conversation, which was difficult, but she was still working hard to do it. Thus, the event code is 1.

The difficult event can also be an unobservable one: "It was difficult not thinking of her brother." The behavioral event that was difficult was "not thinking," which should be coded as a 3 (trying to not think...).

The same considerations apply to "easy." For example, "it was easy to talk to the student" is a 1 (that it was easy was a feature of the talking).

Finally, a common descriptor, "he was easy to talk to" counts as a trait inference (code 0).

Unintentional thoughts

Event type 3 refers to thoughts or other cognitive states that the agent brought about intentionally (e.g., controlled imagery, intentions, plans, memory search processes). But many thoughts are not intentional. For example, "He may have thought I talked too much" refers to a thought that was not intentional (thus, 4) . Similarly, "I thought he was a little reserved" refers to an unintentional thought (4). Other verbs that are unintentional and must be coded as 4 are "to wonder" (e.g., "I wondered what her childhood was like," 4), "to forget," "to know," "to realize," "to recognize," "to find out," "to understand," "to believe," "to be conscious" (e.g., "I was very conscious of my actions")

Coinciding Verbs

Some phrases contain two coinciding verbs both of which must be coded. "I felt comfortable talking with her" (there is the comfort, 4, and there is also the talking, 1). Similarly, "I was nervous while talking with her" refers to two events (4, 1), and so does "It felt good [4]relating my story to her [1]" or "I was glad [4] to meet someone who....[2]."

The verb "to find oneself doing" also requires two codes, as in "I found myself [4] really opening up [2] to her." The particular meaning of this verb strongly suggests that the behavior following it (what one found oneself doing) is not intentional, for if I intend to do something I am unlikely to "find myself" doing it.

Two coinciding verbs should be coded separately only if both verbs refer to behavioral events that actually occurred. This is not the case when the first verb is a mental state and the second describes the mere "content" of the first. For example, "I wanted to laugh" refers to desire (4), but the laughing apparently did not occur. Cf. the section on "trying," "attempting," etc.

Phrases with the verb "to wonder" sometimes require two codes, sometimes one. If what the person wonders about actually occurred, two codes apply, as in "I wondered [4] what I was sounding [2] like." If what the person wonders about is an abstract content, only one code applies, as in "Sometimes I wondered [4] what he was going to talk about next."

Could, couldn't

These auxiliaries are not themselves behavioral events, but they help code the intentionality of the behavior they refer to . "Could" is often associated with an intentional behavior, and "couldn't," with an unintentional one. For example, "I could share [1] these things with him" refers to sharing that the person apparently chose to do (but perhaps she was surprised how easy it was). By contrast, "I couldn't get in a word [2] because he kept talking" refers to unintentional nontalking, caused by the partner's continuous talking.

Nonverbal behavior, body language, posture

As a rule, nonverbal behavior, body language, posture, etc. are unintentional (2). Most of these behaviors can in principle be intentionally controlled but usually aren't. Unless there are indications that they were in fact intentionally controlled, such as to achieve a certain goal, they should be coded as type 2, as in the following examples:

  • "He was leaning back [2] in his chair most of the time,"
  • "I basically during the whole time leaned back [2] in my chair with my legs crossed [2] and used my hands [2] when I talked [1],"
  • "I did notice [4] that I was sitting with my legs and arms crossed [2] and leaning on the side of the chair farthest from her [2],"
  • "She talked very softly [2]" "At the end she uncrossed her arms [2]"
  • "His eyes opened wide [2] when he talked about [1]..."
Eye behavior

Eye behavior can be intentional or unintentional, depending on the verb used and on the context. "To see" is unintentional because it corresponds to a perceptual discovery. "To look," however, can be intentional, as in "She was looking at me [1] while I was talking. [1]." By contrast, "He was not looking at me the whole time" or "She did not look around the room much" are more likely to be unintentional (2). Similarly, "He was having a few problems looking me straight in the eye" refers to the partner's unintentional evasive eye behavior (he probably tried to look her straight in the eye but couldn't).

The word "eye contact" cannot be coded when it involves both people (e.g., "The lack of eye contact between us..."). But it can be coded when it corresponds to "looking," as in "He would hold eye contact when I talked."

Relaxed, friendly, honest, open..

"Relaxed" can refer to a trait ("John is very relaxed"), but it can also refer to an observable behavior ("he was pretty relaxed when he spoke" [2]) or an unobservable one ("he seemed relaxed" [4]). The latter distinction can be difficult to make. Inference markers such a "to seem" or the context of other descriptors may help (e.g., "He was relaxed, not fidgety" [2] versus "She was relaxed, not nervous" [4]).

"Open" can also refer to a trait ("Sarah is an open person") or to a behavior ("She was very open during our conversation"), which is typically intentional, 1. The same holds for "friendly," which can be a trait or a behavior (e.g., "He was friendly throughout").