Emotions are dynamic processes that take place over time—often rapidly, but sometimes gradually— involving a cascade of different response systems, such as those mentioned earlier. Each response system has its own dynamic and duration. For example, the central nervous system is very fast (with negative stimuli producing changes at around milliseconds; Smith, Cacioppo, Larsen, & Chartrand, 2003), the cardiac system somewhat slower (with changes taking place on a beat-by-beat basis), the skin conductance system somewhat slower still (at least 2 seconds poststimulus before a response can occur), and the hormone system is even slower. One critical measurement issue involves isolating the targeted response system within this temporal cascade. When does the response begin and when does it end? Structuring observation of these response systems within the right temporal window will greatly increase the chances of observing emotion-related changes. Imprecision at this stage can dilute the targeted emotional episode within a wash of emotionirrelevant moments (see Levenson, 1988).
Another issue concerns making sure that the measure has sufficient temporal resolution to properly assess the chosen response system. Some indicators of emotion—for example, an increase in cardiac output—might last less than 6 seconds (e.g., Witvliet & Vrana, 2000). The cardiovascular system is exquisitely controlled, and so this response system typically returns to preemotion levels relatively quickly. The experience of emotion, however, could last much longer, or even be assessed later, using recall measures. Cardiac measures cannot be recalled, nor can output be averaged over long time periods because a transient emotion signal can be lost in the average. Self-report measures aggregated over that same time span, however, should accurately reflect the experiential component of emotion.
The discussion so far assumes that the researcher is interested in assessing emotions as states. To conceptualize emotions as states is to emphasize the short-lived (typically), intense (relatively), and situation-caused (mostly) changes in response systems responsible for emotional experience and for supporting resultant behaviors. Because the state conception identifies emotions as quickly changing, measures should be temporally fine-grained, with a temporal resolution that is smaller (ideally much smaller to provide reliable aggregate measures) than the expected duration of the transient emotion-related change.
A third issue about assessing emotion states concerns the temporal proximity of emotion measures to the emotion state itself. Researchers should consider obtaining measures during an emotion experience. This is certainly feasible for measures of the expressive system, such as those obtained from video records or through physiological recording devices, but perhaps this is less feasible for measures obtained via self-report or through a cognitive assessment or through emotion-sensitive tasks, which typically interrupt the ongoing emotion state. Also, emotion measures can be obtained on more than one occasion so as to assess change. When pre- and postmeasures are not feasible or practical, measures that minimize the delay between emotion experience and emotion measurement should be sought.
Although emotions are typically thought of as states, these states nevertheless fluctuate around some mean or average level for each person. Persons differ reliably from each other in their average level of various emotions (Larsen, 2000). As such, emotions can also be conceptualized as having enduring traitlike components, that these emotion traits relate to causes "inside" the person (e.g., personality), thereby exhibiting some degree of consistency and stability (Diener & Larsen, 1984). The concept of an emotion trait refers to the set point or expected value for a person on a particular emotion, other things being equal (George, 1996).
Emotions are thus hybrid phenomena, consisting of both trait and state components, allowing the researcher to focus on one or the other component in addressing various questions (Watson & Telle-gen, 2002).
Why is the distinction between state and trait emotion important to the researcher? First, researchers should be aware that people bring emotional dispositions to the assessment setting; not everyone shows up in the same emotional state. To the extent that emotion dispositions refer to the "expected value" of emotion for an individual, people are likely to have predictable emotional levels. This level may work according to the law of initial values to influence subsequent reactivity. Understanding how emotion traits work, and the causes and consequences of specific emotional dispositions, will help psychologists predict and explain specific emotional reactions. Lastly, states and traits can easily be confused, the variance components that are due to each can become blended, and so researchers need to be aware of this distinction. It is not difficult to find papers in the emotion literature where a correlation is computed between some measured emotion (say positive affect) and some other variable (say helping), and the authors interpret this as a state relationship, as in "people are more likely to help when in a positive mood." However, when measuring emotions in people "off the street," researchers are as likely to be tapping emotion traits as states. Consequently, it may really be that the kind of person who is most likely to be in a positive state (high trait PA) is also the most likely kind of person to be helpful. To infer state effects when emotion traits have been measured is to confound the two sources of variability in emotion measures. To infer state effects researchers should use experimental designs where a manipulation of emotion serves as the independent variable, an emotion measure is included as a manipulation check, and some other theoretical construct serves as the dependent variable.
Emotion traits are receiving a good deal of attention from personality researchers, as well as from psychologists interested in motivation and the biological bases of behavior. When it comes to trait emotion, the dimensional perspective maybe the most useful, whereas for the state approach the categorical view of emotion measurement may be most useful. Zelenski and Larsen (2002) presented data showing that structural conclusions about the emotion domain are related to whether the researcher is analyzing between- or within-subject correlations. That is, because emotions are states, they vary within subjects over time. Such variability is inherently different from between-subject variability. Within-subject analyses (of each subject across 60 measurement occasions in the Zelenski & Larsen, 2002 data) yielded structural support for multiple categories of emotion, whereas between-subject analyses (of each subjects' average emotion scores aggregated over the 60 observations) yielded support for the dimensional structure, with the factors being positive and negative emotionality. Positive affect and negative affect exhibit different structural relationships depending on whether they are assessed as states or traits (Schmukle, Egloff, & Burns, 2002).
Was this article helpful?