Journal of Happiness Studies

, Volume 10, Issue 1, pp 1–17

Life Satisfaction is not a Balanced Estimator of the Good Life: Evidence from Reaction Time Measures and Self-Reported Emotions


    • Department of PsychologyUniversity of Tromsø
  • Hella I. Oelmann
    • Department of PsychologyUniversity of Tromsø
  • Anita L. Wang
    • Department of PsychologyUniversity of Tromsø
Research Paper

DOI: 10.1007/s10902-007-9058-1

Cite this article as:
Vittersø, J., Oelmann, H.I. & Wang, A.L. J Happiness Stud (2009) 10: 1. doi:10.1007/s10902-007-9058-1


A widespread belief in happiness research is that measures of life satisfaction capture the variety of an individual’s experiences along a single favorable—unfavorable dimension. The current article points to a possible violation of this assumption. With a combination of data and theory it argues that life satisfaction is not a balanced reflection of subjective experiences because the evaluation that drives judgments of life satisfaction is tilted towards pleasure and away from engagement. We postulate that feelings of pleasure are overrated in statements of life satisfaction, whereas feelings of engagement are underrepresented. Feelings of engagement and interest are further supposed to be indicators of personal growth. Measures of personal growth and life satisfaction should thus be unrelated. Empirical evidence for our model is provided by self-report and reaction time data from a Norwegian convenience sample (N = 30). Detection latencies for pleasant stimuli correlated significantly with life satisfaction but not with measures of personal growth. Life satisfaction correlated significantly with pleasant experiences, but not with engagement. Personal growth correlated significantly with feelings of engagement. Life satisfaction and personal growth were uncorrelated.


EmotionsLife satisfactionPleasureEngagementReaction timeBiasPersonal growth

1 Introduction

When Jeremy Bentham proposed that a good life simply amounts to the integral of the intensity and duration of experienced pleasure, he also insisted that all subjective feelings are reducible to a single dimension running from pain to pleasure (e.g., Sumner 1996). Some years later John Stuart Mill abandoned the idea of phenomenological homogeneity, but agreed that when the variety of subjective experiences are evaluated post hoc, they all submit to a single axis running from favorable to unfavorable. Today the relation between emotional qualia and the overall judgment of these experiences is captured by the concept of subjective well-being (SWB-Diener 1984). The premise of SWB is that the subjective part of a good life can be accounted for by a hedonic component (i.e., a feeling state) and an evaluative component. The latter is normally referred to as life satisfaction or global happiness, which is the idea of a subjective judgment as regards, not only feelings, but the appreciation of ones life as a whole.

The notion of SWB has been criticized for being either too hedonic (depending exclusively on the presence of pleasant feelings and absence of unpleasant feelings) and/or too attitudinal (depending on elusive subjective judgments about the favorability of a person’s own life). Nozick’s thought experiment of a “perfect feeling machine” (the brain in the jar metaphor-Nozick 1974) is an example of the hedonic criticism. Sen’s notion of capabilities is a response to the limitation inherited in the attitudinal part of subjective well-being (Sen 1993). Variations of the classical Greek concept of eudaimonia have been proposed as alternatives to the hedonic/attitude approach. For instance, inspired by Aristotle’s theory of motivation (Aristotle 1985), Ryff (1989) speculates that there must be six dimensions of well-being, and that one of them is personal growth. In the current article we too raise objections to the hedonic/attitudinal well-being model, and provide arguments in favor of what very roughly can be termed an eudaimonic approach to well-being.

There are literally thousands of interpretations of Aristotle’s idea of the human good. Most of these are very abstract, offering conceptualizations hardly precise enough for scientific purposes. For example, how should Aristotle’s statement about happiness as “activity of the soul in accordance with virtue” (McGill 1967, p. 17) be operationalized in testable terms? Rather than be swallowed by the hermeneutics of Aristotelian ethics, fragments of which frequently flutter around in the well-being literature, let it suffice for our purpose to follow Waterman (1984) and say that the eudaimonic approach has something to do with personal growth and the optimally functioning person. Waterman even takes the concept one step further and locates it in the hedonic domain by suggesting that eudaimonia should be understood as “the feeling accompanying behavior in the direction of, and consistent with, one’s true potential” (Waterman 1984, p. 16 - italics added).

Aristotle did not restrict his theory of optimal functioning to subjective feeling states (see for example McMahon 2006), but the idea of capturing personal growth by indicators of subjective experiences is worth pondering. At least two interesting questions follow from a subjective approach to eudaimonia. First, what are the phenomenological qualities of personal growth; what does it feel like to grow as a human being? And next, can such feelings be accounted for by the unidimensional pleasantness dimension? A first reaction to these questions should probably be that the term personal growth is an outdated concept, much too broad to be fruitfully studied by scientific methods. One reason why the notion of personal growth has not been replaced by more precise concepts is due to the influence from early scholars such as Maslow and Rogers. Another reason is the lack of agreement when it comes to the nature of the new concepts (although a series of suggestions are available, see for example Staudinger and Kunzmann (2005); Staudinger and Lindenberger (2003)). Although terms suggested by Staudinger and others (such as ‘optimal human functioning’) are promising candidates for future research, our article uses the imprecise and broad term personal growth in order to stick to current traditions in well-being research (e.g., Ryff 1989; Waterman 1993).

Turning to what it feels like to grow as a human being, we submit to the traditions from Piaget (Ginsburg and Opper 1969; Piaget 1952), Berlyne (Berlyne 1960; Berlyne 1970) and Tomkins (Tomkins and Izard 1965). This literature suggests that experiences like interest, engagement and curiosity are candidates for what personal growth may feel like. The Berlyne tradition in particular offers solid evidence that these experiences cannot be reduced to feelings of pleasure or pain (see Silvia (2006) for a recent review).

A number of studies, starting with Wundt (1896), have documented results that collectively amounts to a very strong argument in favor of the view that feeling engaged or interested is different from feeling pleased and satisfied. Nevertheless, hedonic psychologists such as Kahneman (e.g., Kahneman 1999) continue to argue that the pleasantness dimension satisfactorily summarizes all human feeling states. For Kahneman, the term pleasantness is not the important part of the argument. Rather he points out that all feeling states can be characterized as preferable (good), unpreferable (bad) or neutral. Accordingly, a life should be considered better if more time is spent in good states and less time in bad states, and “this formulation holds whether the good state is defined by positive affect or by intense engagement in a task or in a spiritual pursuit (Kahneman and Riis 2005, p. 294).

A different take on the relation between engagement and pleasure is offered by Seligman and his colleagues (e.g., Peterson et al. 2005; Seligman 2002; Seligman et al. 2005). Seligman argues that pleasure, meaning and engagement are the primary components of well-being, and that they all may be properly accounted for by the concept of life satisfaction. For instance, in his own interpretation of his Authentic happiness book, Seligman writes: “Well-being includes pleasure, engagement, and meaning, and the concept of life satisfaction may reflect all of these (Seligman 2002).” (Diener and Seligman 2004, p. 4). However, even if there are three constituents of well-being, Seligman does not argue that they are equally important in determining a person’s level of life satisfaction: “Importantly, the pursuit of engagement and the pursuit of meaning make much larger contribution to life satisfaction than does the pursuit of pleasure” (Seligman et al. 2005, p. 279). In a recent study, Seligman and his colleagues report evidence in support of the claim that an engagement orientation to happiness is a better predictor of life satisfaction compared with an orientation to pleasure (Peterson et al. 2005).

Due to the well documented demonstration that satisfaction and pleasure is highly correlated with each other, but only modestly correlated with engagement and interest, we think Seligman’s model is wrong. We believe the results reported in support of his model are due to weak measures of the concept of pleasure. For example, in two separate Norwegian studies, Seligman’s pleasure scale was found to be uncorrelated with self-reported feelings of pleasure in one of the studies (N = 22), and only slightly correlated with pleasure in the other (N = 422 and r = .14, p < .01) (Vittersø, Dyrbal and Røysamb 2006). Following the Wundt/Piaget/Berlyne traditions then, we claim that questionnaire items about satisfaction and pleasure are bad indicators for the feeling states of engagement and interest. Furthermore, we suggest that measures of self-reported life satisfaction are partly affected by a pleasantness bias. We argue that some individuals are more oriented toward what is good and pleasant in the world, but that feelings of engagement and interest are unchanged by such a pleasantness bias.

The Kahneman approach, the Seligman approach, and our approach produce three different predictions concerning the emotional correlates of life satisfaction. According to Seligman, engagement rather than pleasure should be the strongest predictor of life satisfaction. According to Kahneman, engagement and pleasure should be equally strong predictors, whereas our approach suggests that a pleasant feeling is a stronger predictor of life satisfaction than engagement. Two distinctions are important in our argument. The first points to the difference between feelings in the moment as opposed to feelings as memories. The second relates to the distinction between life satisfaction and personal growth.

1.1 Momentary and remembered experiences

In principle, an evaluation of subjective experiences could reflect the sum of good moments minus bad moments in a person’s life. In fact, this is the core idea in classic economics, although the tradition talked about utility rather than well-being (e.g., McFadden 2005; Mellers 2000). However, unknown to the founding fathers of economics, a mental calculation of a global utility (or life satisfaction) response would be far too demanding for the human brain to undertake. As pointed out by Kahneman and Riis (2005, p. 285), during a waking day a person would normally experience some 20,000 moments. The limitation of the sum-of-all-moments approach becomes particularly evident when, as we detail below, answers to questions about life satisfaction are given within a time frame of three or four seconds. Thus, survey participants must use a heuristic of some kind when they respond to the question of how satisfied they are with their life as a whole.

Even if a global evaluation of goodness is the result of a heuristic rather than a “hedonic calculus” of actual experiences. This evaluation could still reflect a rather unbiased ratio of good over bad things that have happened so far in a person’s life. However, more than a decade of extensive research on the relationship between how events are experienced as they occur and how they are retrospectively summarized has proven the unbiased ratio assumption wrong. In a nice review of this literature Ariely and Carmon (2003) conclude that rather than reflecting the affective shape of a subjective experience, a few central features (gestalt characteristics) seem to govern summary evaluations of those events. The slope, peak and end of the experience profile are examples of gestalt characteristics that determine the memory of an episode. It seems fair then, to consider self-reports of good and bad feelings as a biased heuristic in which some features play a more important role than others.

In addition to the gestalt characteristics dominating summaries of emotions, we suggest that another mechanism operates on overall evaluations of good and bad. Compared with feelings such as engagement and interest we believe that pleasant experiences are given higher weights on the good-bad scale. In other words, positive feelings come in different forms and the variety is not easily transformed into a common currency of goodness and badness. The heterogeneity of subjective feeling states is illustrated in the so-called affect circumplex in which different feeling states are organized around the circumference of a circle. The good–bad dimension (or interchangeably the pleasantness–unpleasantness dimension) is represented by the first axis. Activation (or interchangeably arousal) is measured along the second axis. Satisfaction is normally located within the area of approximately 0–20 degrees in such a configuration (with pleasure/goodness located at 0 degrees). Engagement on the other hand is located at approximately 60–80 degrees. Now, if one projects satisfaction and engagement town to the first axis it is obvious that the former gets a much higher value on the good dimension. The principle is illustrated in Fig. 1.
Fig. 1

A circumplex structure of affects, illustrating the theoretical location of satisfaction and engagement along the circumference of the circle, and their value on the good–bad (or pleasant–unpleasant) dimension

1.2 The pleasantness bias

From a well-being perspective, more rather than less good feelings should be preferable regardless of the level of arousal (or other elements) that could be added to the feeling state in question. Even if this statement holds logically, the question still remains as to whether a good-bad evaluation truly reflects the varieties of emotional feelings in an unbiased manner. If a bias towards pleasantness does exist among individuals with high scores on life satisfaction, episodes of pleasure will be overrepresented in surveys of subjective well-being. Other episodes of positive value, such as those characterized by engagement and interest, will be underreported even if, from a logical point of view, both states are preferable and both are considered to be attractive feelings.

There are several reasons to believe that measures of life satisfaction are insensitive to personal growth. One leans towards factor analytic efforts directed at demonstrating that measures of personal growth and life satisfaction fall into different dimensions. For example, after factor analyzing a set of 18 relevant scales, Compton and his colleagues found a large factor for measures of what they labeled subjective well-being as well as another factor for measures of personal growth. Their results provided support for the hypothesis that subjective well-being and personal growth are related, but not identical, constructs (Compton et al. 1996). Similarly, McGregor and Little (1998), Kafka and Kozma (2002) and Selnes et al. (2004) have reported that life satisfaction and personal growth load on different factors.

Another argument is informed by the attempts to differentiate a neurological pleasure system from an excitement system. Berridge (e.g., Berridge 2004) has for instance carried out a series of experiments showing the difference between a liking system (pleasure) and a wanting system (engagement). Panksepp (1998) argues that what he refers to as a seeking system has evolved in the human brain, characterized by intense interest in exploring the world around us, which lead us to become exited when we are about to get what we desire. Panksepp speculates that this may be one of the brain systems that generate and sustain curiosity in humans, even for intellectual pursuits. The feeling experienced during this emotional state is not pleasure as such, but the expectation that pleasure will be experienced. The important neurochemical ingredient in the seeking system is dopamine, and drugs that typically would stimulate the system are psychostimulants such as amphetamines and cocaine.

Panksepp suggests that pleasure is represented by another brain structure, which plays a quite different role in the government of behavior. It sustains the regulation of bodily processes including feeding, drinking and temperature regulation. Stimuli that promote a return to the homeostatic set-point are routinely experienced as pleasurable, while those that impair homeostasis are unpleasant. So, rather than feeling exited and curious, as is the case in the seeking system, the pleasure state is dominated by satisfaction. The important neurochemical ingredient in the pleasure system is endogenous opioids, and drugs that typically stimulate the system are narcotics such as morphine and heroine.

Taken together, these perspectives suggest that due to both methodological and theoretical reasons, satisfaction and engagement may be considered as separate constructs. Vittersø (2005) demonstrated this differences when, consistently over three studies, life satisfaction scores predicted positive emotions in situations of pleasantness and goal fulfillment, but not in situations of complexity and goal-obstruction. A measure of personal growth and engagement showed the opposite pattern. In the studies reported in Vittersø (2005) the personal growth indicator correlated with complex and challenging situations, but not with situations of pleasure and goal-fulfillment. Moreover, questionnaire data from another study revealed that engagement correlated with positive emotions reported during a problem solving task, but not with life satisfaction or with emotions after the task was accomplished. Life satisfaction did correlate with the positive emotions after the problem solving task was completed (Grape and Vittersø 2003). A similar pattern was replicated in a Day Reconstruction study among 82 Norwegian skiers. In this study, life satisfaction was highly correlated with positive emotions in general, but not with positive emotions during an effortful winter hike. Engagement on the other hand did not correlate with positive emotions in general, but did so with positive emotions during the effortful episodes. Engagement and life satisfaction were uncorrelated (Vittersø, Hetland and Søholt 2006). Finally, in a mood induction experiment, pleasure was successfully increased after showing pleasing pictures from the International Affective Picture System (IAPC). Engagement, on the other hand, was differently affected by the mood manipulation (Vittersø et al. 2007). In the current study we wanted to replicate and expand these results with a combination of self-reports and reaction time data.

1.3 Attention and the evaluation of goodness

A pleasantness bias could operate at the attentional level, orienting satisfied people toward pleasant information and away from unpleasant information. To test this possibility we used two different Reaction Time (RT) paradigms. The first was an attentional detection design (the pop-out test), and the other was an attention maintenance design (the dot-probe test).

The “pop-out effect” was demonstrated by Hansen and Hansen (1988). The idea is that emotional relevant stimuli guide attention according to the principles of the evolutionary-functional approach to emotions. The functional aspects of emotions imply, among other things, that we attend to different aspects of the environment when in different emotional states. For instance, when in a fearful mood, we concentrate attentional resources on possible threats (Öhman and Birbaumer 1993), whereas in a euphoric mood attention becomes biased towards cues signaling success and progress toward goal achievement (Isen 2003). Moreover, a quick detection of threatening cues is supposed to be more important than the detection of positive cues, so that we have an evolutionary developed bias towards very quick detection if possible threats (the “anger superiority effect”). Accordingly, Öhman and his colleagues (Öhman et al. 2000) have demonstrated that angry faces are detected quicker than happy faces, regardless of the background against which they are presented. Moreover, fearful individuals have been shown to react quicker to fear-relevant stimuli compared with control groups.

In the current article we take the knowledge about preattentive biases in fearful persons one step further. We suggest that a bias may exist for persons rating themselves as satisfied, favoring pleasant information that is processed at the attention level. High degrees of life satisfaction could partly be a result of a “Pollyanna tendency” that involves a bias towards quick shifts in attention toward pleasant stimuli in the environment. Scattered evidence for such a bias does exist. For instance, Robinson et al. (2003, Study 2) found that satisfied individuals (as measured by the Satisfaction With Life Scale—Pavot and Diener 1993) were faster to distinguish between positive and neutral words, relative to less satisfied individuals. A similar effect for measures of activated or engaged, emotional disposition (as measured by the PANAS—Watson et al. 1988) was not found.

Another attentional mechanism involves the maintenance of attention rather than detecting a stimulus (e.g., Posner and Petersen 1990). Whereas the pop-out test taps into a tendency of quickly shifting attention towards, for example, pleasant (or unpleasant) information, the dot-probe test measures a tendency to keep on processing, for example, pleasant (or unpleasant) information. The basic principle of the dot-probe test is to measure which out of several stimuli a participant is looking at. A bias in the tendency to keep on processing positive information was found by Derryberry and Reed (1994), although their study did not use the dot-probe design. They showed that a “holding power” rather than an “attracting” power of positive and negative stimuli operates in the motivational Behavioral Approach System. In Derryberry and Reed’s study, the focus of attention was on personality traits, but a “holding power bias” might also operate in the area of well-being, with measures such as the Satisfaction With Life Scale. We suggest that the dot-probe test is a suitable way of testing this assumption.

1.4 Hypotheses

If satisfaction is located in a pleasantness domain of well-being whereas personal growth is located in an engagement domain, it can be assumed that:
  1. 1)

    Life satisfaction is biased towards pleasantness rather than engagement.

  2. 2)

    Personal growth is biased towards engagement rather than pleasantness.

  3. 3)

    Life satisfaction and personal growth are relatively unrelated with each other.

These hypotheses are tested with correlations, path analysis and two different reaction time paradigms.

2 Method

2.1 Participants

Recruited from the University of Tromsø, Norway, a convenience sample of 30 subjects (music students, psychology students and technical staff) took part in the study, receiving a lottery ticket with a value of NOK 20,- (approximately 3 USD) for their participation. There are 18 females (60%) in the sample. All participants were selected within the age range of 18–50 years.

2.2 Procedure

The study was conducted in a quiet room, containing a chair and a table with a computer-screen and a keyboard. Participants were taken to the room one by one, and told that the study was about happiness and quality of life, and that they had to answer some questions and conduct some tasks, all of which would be done in front of the computer-screen. They where then taken through a short demonstration version of the computer-program, which demonstrated some of the questions and some of the tasks of the study. When necessary the demonstration was repeated until the participant was familiarized with and felt comfortable with the situation.

The program first asked about background information (age, gender and whether the participants were right handed or left handed). Next, the participants answered a series of commonly used subjective well-being questions, such as the Satisfaction With Life Scale and a one-item happiness question, and a series of positive emotion and negative emotion items. Participants gave their answers by pressing the number keys on the keyboard as each item was presented on the screen.

2.3 Measures

2.3.1 Self-reports

Life satisfaction was measured with the Satisfaction With Life Scale (SWLS—Pavot and Diener 1993), a five item inventory with items such as “I am satisfied with my life”. Responses were given on a numerical rating scale running from 1 (strongly disagree) to 7 (strongly agree). Cronbach’s alpha (α) for the SWLS was .87.

Happiness was measured by the following question: “All thing considered, would you say that you are: Very unhappy (1), Unhappy (2), Happy (3) or Very Happy (4)”.

Trait emotions were measured with the Basic Emotion Trait Test (BETT—Vittersø et al. 2005). The BETT contains a total of 15 items, three for each of the five basic emotions (Pleasure, Engagement1, Anger, Fear and Sadness). The Pleasure subscale (measured with the items happiness, joy, and contentment), and the Engagement subscale (measured with the items engagement, inspiration and interest) were collapsed into a positive trait emotion sumscore (α = .73), whereas Anger, Fear and Sadness were collapsed into a negative trait emotion sumscore (α = .77). Responses were given on a numerical rating scale running from 1 (not at all) to 7 (all the time).

State emotions were measured with the Basic Emotions State Test (BEST—Vittersø et al. 2005). Immediately after the Pop-out reaction time test (see below), participants were asked to report how (much) they felt while working with the task, with regard to 6 positive emotions (happiness, joy, contentment, engagement, inspiration and interest; Cronbach’s alpha = .80) and three negative emotions (anxiety, irritation and sadness; Cronbach’s alpha = .53). Answers were given on a 1 (No, not at all) to 7 (Yes, absolutely) scale. A total Emotionl State Score was calculated as positive emotions minus negative emotions (Cronbach’s alpha = .73). Personal Growth (PG)

This scale comprises four subscales: Subscale A, Curiosity (Amabile et al. 1994) with 5 items; subscale B, Flow (Kashdan 2004) with 3 items; subscale C, Complexity (from Cattell’s 16PF, available from IPIP 2002) with 5 items; and subscale D, Competence (from Cloninger’s TCI, available from IPIP 2002) with 5 items. Participants responded to statements on a Likert-like response format from 1 (Totally disagree) to 5 (Totally agree). Examples of the items are: “I enjoy trying to solve complex problems”, “When I am participating in an activity, I tend to get so involved that I lose track of time”, “I love to think up new ways of doing things”, and “I can perform a wide variety of tasks” for curiosity, flow, complexity and competence respectively. The Cronbach’s alpha of the Personal Growth composite scale was .80.

2.3.2 Reaction times Pop-out reaction time.
After the self-report part (except for the BEST), participants were exposed to a series of 3 × 3 matrices of schematically drawn faces of which half showed 8 identical and neutral faces and 1 happy face (Fig. 2a) and half showed 8 identical and neutral faces and 1sad face (Fig. 2b). We used schematically drawn faces because the use of real faces in pop-out studies is complicated (Öhman et al. 2000). The deviant face occurred randomly at all positions in the matrix over the 18 trials of this test. Participants were told that they would see nine faces on the screen, and that one of them differed from the others, by being either a happy face among neutral faces or being a sad face among neutral faces. A random half of the participants were told to press the “happy” button (the ‘m’ key marked with a red sticker) if the happy face appeared and a “sad” button (the ‘z’ key marked with a red sticker) if a sad face appeared. For the remaining half of the sample the instructions were reversed so that the “happy” button was the ‘z’ key and the sad button was the ‘m’ key. Participants were told to respond as quickly as possible. Each trial consisted of the following sequence: (a) a fixation point appeared at the center of the screen for 500 ms, (b) a 3 × 3 face matrix was then displayed until the participant pressed one of the two response keys on the keyboard. In the computation of pop-out RT means, we excluded error trials (i.e., pressing the “happy” button for a sad face and vice versa) and responses outside the range of ±3 SD from the individual’s own mean.
Fig. 2

Examples of 3 × 3 matrices of schematic faces used in the pop-out test. Deviant happy face (A) and deviant sad face (B) Dot-probe reaction time.
In the second series of trials, a total of 20 pairs of happy and sad faces were displayed on the computer screen. An example of such a pair of faces is given in Fig. 3. Each trial consisted of the following sequence: (a) a fixation point appeared at the center of the screen for 500 ms, (b) the happy and sad faces were displayed in the right and left positions on the screen for 3000 ms, (c) the two faces disappeared from the screen, and (d) an X appeared in the center of the screen location where one of the faces had been. The X remained on the screen until the participant pressed one of the two response keys on the keyboard. Participants were told that they would see two faces on the screen, and were instructed not to respond until an X appeared. They were told that when the X appeared on the screen they should respond as quickly as possible. If they saw the X appear on the right side of the screen, they should press the right button (the “m” key marked with a red sticker); if they saw the X appeared on the left side, they should press the left button (the “z” key marked with a red sticker). Happy and sad faces were presented randomly on the left and on the right side of the screen, and the X appeared randomly on either the happy or sad side of the screen.
Fig. 3

The happy and sad faces used in the dot-probe test. The happy face was randomly presented to the left and right part of the screen and the X appeared randomly behind either one of the faces

In the computation of dot-probe RT means, we excluded error trials (i.e., pressing the left side button when the X appeared on the right side and vice versa) and responses outside the range of ±3 SD from the individual’s own mean. For the sake of parsimony, we refer to the X when it appeared where the happy faces had been as the “Happy X”, and the X when it appeared where the sad faces had been as the “Sad X”.

3 Results

The mean SWLS sumscore was 24.9 (SD = 5.5) and the mean RT for responding to each of the items on the SWLS was 4724 ms (SD = 1634). The mean score of Happiness was 3.0 (SD = 0.67), with the mean RT being 9964 ms (SD = 4084). Mean scores for basic trait emotions were 5.2 (SD = 0.66), and 2.8 (SD = 0.53) for positive and negative emotions respectively. On average, participants spent 4204 ms (SD = 1456) on answering each of the basic trait emotion items. The mean PG score was 3.80 (SD = 0.44) and the mean RT for responding to each of the items was 4603 ms (SD = 1311). For the positive state emotions, the mean score was 3.75 (SD = 1.06) and the mean for negative state emotions was 1.37 (SD = 0.63). RT for responding to the state emotions: M = 3620, SD = 1190. Overall, sad faces were detected quicker than happy faces Ms = 935.16, SDs = 219.03; Mh = 1077.63, SDh = 302.56; t (29) = 4.37, p < .001). There were no significant correlations between scores on a variables and the RT for that variable, but the RT for SWLS, PG and BETT and BEST were strongly correlated (all p’s < .001), with a mean r = .74, and SDr = .08.

Following the Seligman approach, a higher correlation should be expected between engagement and life satisfaction, than between pleasantness and satisfaction. According to the Kahneman approach, the correlations between life satisfaction and pleasantness and engagement should be about equally high. Against these positions, our assumption was that pleasantness and life satisfaction would correlate highly whereas the correlation between engagement and life satisfaction would be small. Table 1 shows the relations between Life Satisfaction, Personal Growth and the Trait Emotions. The SWLS variable correlated significantly with Pleasantness (r = .39, p = .034), but not with the Engagement (r = .18, p = .349). The PG variable correlated significantly with engagement, and non-significantly with pleasantness (r = .39, p = .035; and r = .08, p = .673, respectively). The correlation between SWLS and PG was -.03 (p = .859). When aggregated into a positive trait emotion variable (pleasure + engagement), SWLS, but not PG correlated significantly with positive emotions in general.
Table 1

Pearson’s product-moment correlations and descriptives for the study variables (N = 30)














2. PG




3. PL_T





4. ENG_T






5. HAP







6. PO_Hap








7. PO_Sad









8. DP_Hap










9. DP_Sad






























Note. SWLS = Satisfaction with Life Scale; PG = Personal Growth; PL_T = Pleasure as Trait Emotion; ENG_T = Engagement as Trait Emotion; HAP = Happiness; PO_Hap = Pop-Out Happy face; PO_Sad = Pop-Out Sad face; DP_Hap = Dot-Probe X after a Happy face; DP_Sad = Dot-Probe X after a Sad face. SD = Standard Deviation; * = p < .05; ** = p < .01; *** = p < .001

Turning to momentary emotions, the pattern was reversed. PG but not SWLS was correlated with the aggregated positive state emotions. This pattern can also be displayed in a path analysis with latent independent variables (Fig. 4). A structural equation model confirmed that SWLS predicts emotions in general (β = .36, p < .05) but not in the moment (β = .09, ns). For PG the path to positive state emotion was significant (β = .59, p < .05) and the path to positive trait emotions was not (β = .29, ns). The data fit the model excellently (χ2 (41, N = 30) = 35.9, p = .696, CFI = 1.00, RMSEA = .00).
Fig. 4

Path model of the relations between Satisfaction with Life (SWLS), Personal Growth (PG) Positive Trait Emotion (PTE) and Positive State Emotions (PSE). Significant (p < .05) path coefficients are shown as standardized regression weights (betas). The association between PTE and PSE is a significant (p < .05) correlation coefficient

When it comes to the reaction time measures, some participants were quicker to react in the pop-out task on both sad and happy faces than others, with a Pearson product-moment correlation of r = .81 (p < .001), between detection latencies (reaction time) for happy and sad faces. To test the attentional detection bias hypothesis, we thus controlled for each participant’s reaction time by running a multiple regression analysis.

In the first regression model, RT for detecting happy faces was regressed on the SWLS, PG and the RT for sad faces. The three variable explained 71% of the variance in the happy Pop-out face variable (adjusted R2 = 71, F (3, 26) = 24.20, p > . 001). The SWLS variable contributed significantly to the prediction (β = −.27, p = .016) whereas PG did not (β = .09, p = .374). This indicates that as life satisfaction increases, detection time for happy faces decreased (Table 2). In a subsequent analysis, the SWLS variable was replaced with the single-item Happiness variable. The Happiness variable did not predict reaction time (β = .08, p = .527).
Table 2

Regression weights and t-values for reaction time for the happy faces in the pop-out and dot-probe tests as dependent variables and swls, pg and reaction time for the sad faces as independent variables (N = 30)


Pop-Out Test

Dot-Probe test




























Note. * = p < .05; *** = p < .001; B = Unstandardized regression weight; β = Standardized regression weight; SWLS = Satisfaction With Life Scale; PG = Personal Growth. SAD = Reaction Time for the Sad Faces

Table 2 also shows that when RT for the dot-probe (i.e., spotting the X beneath a happy face) was regressed on the SWLS, PG and the Sad dot-probe variable, the SWLS variable again predicted an attentional bias (β = −.22, p = .040). The three variables again explained 71% of the variance in the happy Dot-probe face variable (adjusted R2 = 71, F (3, 26) = 24.23, p > . 001). These results indicate that participants who reported high satisfaction with life, used shorter time to detect the X behind the happy face, after controlling for overall reaction time. Again, a subsequent analysis revealed that when SWLS was replaced with Happiness, no prediction occurred (β = −.07, p = .553).

4 Discussion

The study illustrated that in general, pleasantness but not engagement correlated with life satisfaction. Engagement, but not pleasantness, correlated with personal growth. Moreover, whereas positive emotions in general are predicted by life satisfaction (and not by personal growth), positive emotions during a particular task are predicted by personal growth (and not life satisfaction).

The results also indicated that individuals with high scores on life satisfaction are biased toward pleasant stimuli with regard to both attentional detection and attentional maintenance. Reaction times for recognizing a happy face among neutral faces correlated negatively with scores on the SWLS (the pop-out test). Similarly, satisfied people were quicker to detect an X on the screen when it appeared where a happy face had been than when it appeared where a sad face had been (the dot-probe test). The result from the dot-probe test indicates that satisfied people spent more time looking at the happy face compared with less satisfied people.

When it comes to the sensitivity of life satisfaction scales, our findings are inconsistent with both the notion that varieties of goodness are equally captured by such scales (Kahneman) and the idea that engagement is favored (Seligman). It rather seems that pleasantness holds a favored place. Taken together these results support the distinction between life satisfaction and personal growth and their two emotional counterparts pleasure and engagement.

Following the notion of optimal functioning, and the idea that feelings of satisfaction result from need fulfillment/goal achievement, it seems reasonable to suggest that the regulation of needs and goals are monitored by an evaluation along the dimension of good and bad. For instance, Carver and Scheier argue that people constantly evaluate whether their current experiences and actions are fulfilling their needs. If they are not, a “nagging sense that something isn’t right results” (Carver and Scheier 1992, p. 402). Although we agree with this reasoning, we believe that another mechanism is operating as a supplement in the process of self-regulation. In the bimodal well-being model (Vittersø 2005), engagement takes the role of committing persons to goals, even when the goals are difficult to accomplish. Without such commitments, humans would risk to give in for the “nagging sense” every time a goal is obstructed. Operating with a “pleasure” mode alone, few of our projects would be accomplished. Hence, when engaged in the execution of plans (in the interest of goal fulfillment), the feeling of engagement prevents the unpleasantness of temporal setbacks to stop behavior needed to accomplish complex and challenging goals. We interpret the presented results as supportive of the idea that both a pleasantness system and an engagement system have vital effects on the orchestration of human behavior, and thus the feelings involved in well-being.

The findings provide evidence in the debate on top-down versus bottom-up processes in subjective well-being. A series of studies have shown that part of the variance of subjective well-being measures is due to personal dispositions rather than situational influences (see Lyubomirsky et al. 2005 for a review). The tendencies reported in our study lend correlational evidence for a personological basis for life satisfaction, which is to say that satisfied people are more oriented toward the “sunny side of the street”. Or more precisely, what the findings indicate is that people differ with respect to how quickly they detect pleasant stimuli, and the shorter reaction time a person shows the higher scores he or she also shows on the life satisfaction scale. The direction concerning causality between the two variables is not tested in our design. Hence the top down mechanism might be a matter of attention rather than of evaluation as such. In any case the findings represent an argument in the debate on validity of self-reported well-being measures. Our study add to the literature in which measures of self-reported well-being have been found to correlate with objective physiological (e.g., Davidson 2005 and Urry et al. 2004) and medical (e.g., Cohen et al. 2003) criteria.

Another interesting question to consider is why the bias towards pleasant stimuli was found for the five-item satisfaction with life inventory, but not for the single-item happiness variable. The issue of reliability may explain this difference. In general single-item measures and the use of only four response categories reduces the reliability of a scale (e.g., Krosnick et al. 2005). For well-being research in particular Smith (1979) found that among people replying “very happy” to a question about general happiness with four response categories, there was a group that would choose an even more positive response such as “completely happy” if only they were offered such an option. Moreover, among those replying “pretty happy” there was a group that resists choosing “very happy” because they did not consider themselves in the top category. When this constraint was removed by the creation of a new top category (“completely”), about one-fourth of the “pretty happy” individuals switched into the “very happy” category.

An alternative explanation for our finding is that questions about overall happiness and general life satisfaction are tapping into different constructs and that only the latter function as a predictor for the pleasantness bias. This interpretation cannot be ruled out in our design. After all, the correlation between the single item happiness variable and the satisfaction with life scale was only .33 (cf. Table 1). Whether it is low reliability or conceptual discrepancies that drive the lack of a connection between the single item happiness scale and reaction time, the take home message from our comparison should be that care must be taken whenever one wants to draw conclusions about life satisfaction based upon 1–4 point rated questions about happiness. Although this caution should go without saying, the well-being literature is actually ripe with examples of the opposite, not at least in debates on income and well-being.

We note that our study is limited by a small sample size. For nullhypothesis significance testing (NHST) with a pre-specified H1, a small N work against the researcher and the issue of weak statistical power is less critical. However, the situation changes dramatically when the research hypothesis equals the null hypothesis. Some of our hypotheses are of this latter kind. In the cases where no relations were predicted, the effect sizes were generally very low (r’s < .10). One exception was the correlation between engagement and life satisfaction (r = .18, ns), but this was not surprising. In relevant literature, the correlations between these two constructs have repeatedly been reported to be in this area (e.g., Peterson et al. 2005; Vittersø 2005). Another effect size above .10 was the standardized regression weight from the dot-probe test upon personal growth (β = .19, ns). Currently we do not have any explanations as to why growth oriented individuals seem to look longer at the sad faces compared to less growth oriented individuals. Further research might look into whether this is a stable relation or just statistical noise. Moreover, future research would benefit by replicating the current results, to see whether our findings do generalize, not at least so in different cultural contexts.

In concluding we would like to emphasize that if life satisfaction is not an unbiased estimator of the good life, we seriously need to reconsider some of the conclusions drawn in the field of subjective well-being. If only pleasantness and not engagement is captured in the most commonly used measures in the field, supplementary indicators must be included in the conceptual and methodological toolbox for happiness researchers. After all, and as Linley et al. (2006) recently have reminded us, it is the optimal functioning person that constitutes the quintessential concern in positive psychology. Understanding the fully functioning person thus require knowledge about both the stability part (satisfaction) and the growth part (engagement) of human well-being.


In emotion research, the common parlour would be Interest rather than Engagement. We have chosen to use the term Engagement to be consistent with the jargon of current positive psychology.


Copyright information

© Springer Science+Business Media B.V. 2007