Despite a long history of conceiving emotional differences as temperamental (Allport, 1937; Campos et al., 1983; Goldsmith, 1993, 1994), the distinction can be confusing. Here, we describe definitional differences between emotion and temperament from the perspective of Lab-TAB. Let us suppose that a single infant is tested in the Stranger Approach episode of the Lab-TAB and shows a fearful facial expression when the stranger first begins to approach. The infant's fearful facial expression should not necessarily be conceived as temperamental, but instead is an emotional reaction. What if the infant continues to show facial fear reactions throughout the Stranger Approach, is this pattern of facial expressions temperamental? Again, no. We have simply observed a series of facial expressions. Now, if 100 infants are tested using a Stranger Approach and we quantify their individual differences in patterns of facial expression, can we call these differences temperamental? Yes; temperament requires an individual differences perspective. However, our designation of these differences as temperament is simply definitional; it is open to dispute. If we now quantify individual differences in the sample to other responses that prove to be correlated with facial fear (such as withdrawal, freezing, looking away, and whimpering or crying) and infer a trait that accounts for correlated responses, we are on firmer ground for calling the pattern of reactions temperamental. Still, competing explanations for the correlated responses exist (e.g., individual differences in the effects of infant state, such as tiredness or hunger). Measured emotional reactions to other novel or threatening stimuli besides the stranger that converge across situation and extend over time (days, weeks, and longer) and demonstrate cross-situational consistency as well as longitudinal stability adds to our confidence that we are observing the manifestation of a temperamental trait, using Allport's (1937) definition. Individual instances of emotional behaviors are never properly viewed as temperamental; patterns of consistent and stable individual differences in emotion manifestation are temperamental, or at least they are consistent with temperament theory.

At the most recent counting, we have received over 600 requests for the Pre-locomotor, Locomotor and Preschool Lab-TAB manuals from researchers in 36 different countries. We do not intend for Lab-TAB administration, coding or scoring to be prescriptive; therefore, researchers may use different coding schemes and modify administration of the episodes depending on their needs. Developmental scientists who have used the manuals over the last 25 years have addressed important questions about the early development of temperament and emotion from a multi-method perspective. For example, one group of researchers has used infant Lab-TAB data to examine the contextual basis of maternal perceptions of temperament (Hane et al., 2006), and another has examined interactions between negative temperament and parenting behaviors as risk factors for child obesity and obesity interventions (Anzman-Frasca et al., 2013, 2014). Others have examined early social and nonsocial fear and other temperament dimensions assessed in Lab-TAB episodes to predict the development of anxiety in early school years (Buss, 2011; Kiel and Buss, 2011; Brooker et al., 2016). Another group has noted remarkable stability in five temperament dimensions (Dyson et al., 2015) as well as an interaction between elevated cortisol levels and a high quality relationship with the primary parent predicting lower negative emotionality and higher positive emotionality in children from age three to six (Kopala-Sibley et al., 2017), using Lab-TAB assessments. However, psychometric properties of the infant Lab-TAB as a whole have not been examined in great depth or detail although we have examined properties of a home-administered version of Lab-TAB for older children, at age 4.5 years (Gagne et al., 2011). Although many investigators use one or a few Lab-TAB episodes as suitable for their research (e.g., Braungart-Rieker et al., 2010; Degnan et al., 2011; He et al., 2013), a need remains for a comprehensive examination of multiple dimensions of infant temperament to garner a more holistic, person-centered understanding of infant emotional development.

Following an intellectual tradition dating back to Darwin (1872), discrete emotion theorists such as Tomkins (1982), Ekman (1982), Izard (1972), Izard and Malatesta (1987), and Plutchik (1980) identify fear, anger, sadness, joy, interest, disgust, and surprise as fundamental emotions, along with other feeling states (Ekman, 1982, 1989). Temperament theorists traditionally include activity level as aspects of temperament (Rothbart, 1981; Goldsmith and Campos, 1986) and sometimes point out affective features of this overtly non-emotional temperament construct (Goldsmith and Campos, 1986; Goldsmith et al., 1987). For the purposes of this paper, we examine individual differences in fear, anger, sadness, positive affect, interest/persistence, and activity level.

Although researchers often use a more general measure of negative affect in early infancy, we believe that a broad dimension of proneness to distress is less useful than more focal dimensions of fearfulness, anger proneness, and sadness, which show differences in function, expression, and development (Campos et al., 1983; Clifford et al., 2015). Previous research (Braungart-Rieker et al., 2010; Clifford et al., 2015) has indicated etiological as well as behavioral differences in fear, anger and sadness. Though all conceptually negative, anger is related to approach behaviors whereas fear and sadness are related to withdrawal behaviors. Fear is typically expressed during novel situations whereas sadness is expressed upon loss of an object or need. Thus, evidence from outside Lab-TAB dictates that these three dimensions represent different aspects of hedonic negativity and thus should be examined separately.

The idea that temperament dimensions manifest in a diverse set of behaviors is well-established. More discrete and quantitative measurement of these dimensions in infancy becomes difficult because we must depend entirely on behavioral observations without the benefit of knowing with certainty whether the infant interprets the context of assessment, including the affective incentives, in the way that investigators intend. Temperament measures can take many forms: parent reports of infant's responses to various scenarios or laboratory assessment of infant reactions to specific stimuli are often utilized to assess infant behaviors (Rothbart and Goldsmith, 1985). Despite some convergence across parent reported and laboratory assessed infant behaviors, the differences between these two modes of assessment have been explicated extensively (Goldsmith and Rieser-Danner, 1990). Among the key differences are that parents generalize across contexts and time and bring their own biases to bear in interpreting behavior whereas laboratory measures reflect objective, minimally interpreted responses to a specific context over period as short as a minute. Thus, the lack of strong convergence is unsurprising.

The magnitude of gender differences in toddler and child temperament appears to be modest and variable depending on which dimension of temperament is under investigation (Chaplin and Aldao, 2013). A recent meta-analysis of gender differences in temperament indicated that boys tend to be higher in approach based tendencies such as anger, exuberance, and activity level whereas girls are higher in shyness, some fear behaviors, and inhibitory control (Else-Quest et al., 2006). Our results in this paper further examines this issue in the pursuit of describing and validating a laboratory based assessment of infant temperament.

Emotional behaviors vary throughout development; early in infancy, we generally conceive of emotional behaviors as solely reactive in nature. The variety of these reactive behavior increases during infancy, and their solely reactive nature also changes. From 6 to 12 months of age (the age period examined empirically in this paper), infants grow rapidly. During this period, infants develop more muscle tone and the ability to stabilize themselves and sit more easily; they begin to crawl and then to pull themselves to standing, and finally begin to walk. These advances in motor control align with cognitive and motivational changes and support an expanding repertoire of emotional expressions (Izard and Malatesta, 1987). Moreover, increased motor and cognitive control allows infants to better regulate emotions as they approach 12 months of age (Thompson, 1991). In other words, infants develop more ways to express emotion concurrently with developing ways to downregulate emotions—especially hedonically negative emotions—more effectively. Temperament theorists generally expect that individual differences in the tendency to respond to emotionally laden stimuli remain relatively stable across infancy (Goldsmith et al., 1987; Rothbart and Bates, 1998).

In a sample of 6 and 12 month infants, we discuss scale development and scoring of tasks that elicit the primary emotions and describe gender differences and stability in infant temperament. Further, we use multiple methods (parent report and laboratory-based, elicited behaviors) to examine relations across reporter. We address the following questions: (1) How do we construct scores for laboratory assessments of infant temperament? (2) What degree of convergence exists between laboratory and parent report infant temperament? (3) To what extent does gender relate to infant temperament, and are temperament dimensions stable from 6 to 12 months?

Parents and their infant twins (N = 1,076 infants) were assessed in both the laboratory and via questionnaire methods. Some 92.8% of infants were Caucasian (3.2% Hispanic), 3.5% of infants were African-American, and 1.7% were Asian-American. Approximately half the infants (n = 550, 51.5%) were female. Mothers were on average 32 years old (M = 31.86, SD = 4.79) and fathers were 34 years old (M = 33.67, SD = 5.68), with median family income above $50,000. Seventy-five percent (75.4%) of mothers and 77.1% of fathers had completed college and 14.9 and 20.6% of mothers and fathers, respectively, had a high school degree.

Infants participated in videotaped laboratory assessments of temperament at 6 and 12 months of age. Mothers brought their infants into the lab and completed the Lab-TAB with each infant tested in a tandem fashion (i.e., separate testing of each twin, conducted in a multi-room laboratory with a large team, such that co-twins could be tested at the same time). Following the laboratory visit, mothers and fathers completed questionnaires measuring infant temperament. This study was carried out in accordance with the recommendations of the University of Wisconsin-Madison's Institutional Review Board with written informed consent from all subjects.

Recruitment was ongoing throughout the course of the study; thus, more families participated in the 12-month than 6-month assessment (see Table 2 for Ns, which reflect both the continuing recruitment and missing data). Analyses used full information maximum likelihood estimation (FIML; Enders, 2010), a robust method for approximating unbiased parameter estimates by utilizing all available data.

The Lab-TAB yields several scores (e.g., facial affect ratings) for each epoch (intervals of a few seconds) across each trial within each episode of the protocol. The resulting amount of raw data from as few as 10 episodes can be overwhelming. Here, we describe a standardized procedure for scoring the Lab-TAB and reducing the data to more useable episode-level variables for each desired construct. Figure 1 depicts our data reduction strategy to derive temperament composites from Lab-TAB raw data.

The first step in data reduction is to create response parameters for behaviors across the episode. For example, in the Masks episode the infant receives a raw score for facial fear on each epoch (n = 2) for each mask (n = 4), resulting in 8 scores. Thus, we average facial fear across the 8 epochs and use this as a mean facial fear score for Masks. By examining the mean level of each behavior that the infant exhibits, we quantify the infant's average reactivity for each episode. Consequently, shorter latencies to show facial fear, longer durations of facial fear, and higher levels of facial fear will all be reflected in higher means. Next, we determine the peak level of facial fear for the episode, whenever this peak might have occurred during the 8 epochs. Peaks and means tend to be correlated, but peaks better capture an infant's tendency to respond to stimuli using extreme behavioral responses. Two infants may have the same mean facial fear scores, but very different peak scores.

Finally, we calculate the latency, or the time before, the occurrence of a behavior. We do so indirectly by counting the number of epochs before the first occurrence of facial fear, for example. If facial fear is apparent right away, then the latency it takes to observe it is 0 because it occurs in the first epoch. If facial fear is not present until the third mask, the latency is 4 because it occurs in the fifth epoch. We then reverse score these latencies to reflect the speed with which the infant exhibited the behavior. This procedure results in mean, peak, and speed scores that are all in the same direction (i.e., higher scores imply more intense or frequent occurrence of a behavior) and tend to be correlated within an episode. We suggest that researchers follow this approach initially then use the mean, peak and speed scores as best fits their research question.

For this paper, we combined mean and peak scores across trials within each episode to create an overall episode score. Factor analyses to determine which scoring components to include in each episodic score and then reliability statistics (Cronbach's alphas) were calculated for each episode at each age. The speed scores did not add significant variation to our episode-level scoring, and these speeds are therefore not included in further analyses. Researchers who are interested in examining how quickly an infant may react to a stimulus, or more generally in examining emotion expression using a chronometric approach (Davidson, 1998) may opt to include the speed scores as well. The resulting episode-level scores include each of the scoring components listed in Table 1. For the fear episodes, escape behaviors and bodily fear (freezing) are mutually exclusive behaviors, so only bodily fear is included in the Stranger Approach and Masks episode-level scoring. We also combined scores across episodes to create an overall score for each temperament dimension of interest. For example, Masks and Stranger Approach both assess infant fear so we combined these for an overall Fear composite (see Figure 1).

Table 2 provides the reliability statistics for the episode-level and composite scores derived from each Lab-TAB episode. For the IBQ scales, alphas were satisfactory and generally consistent with the literature. The IBQ Soothability scale tended to somewhat less internally consistency than the other scales, possibly because the nature of the items on this scale ask about the effectiveness of different soothing techniques. High internal consistency would require most techniques to be equally effective, which may not be the case. Mother-father correlational agreement on all IBQ scales was significant at both ages in this large sample. The magnitude of the agreement might be characterized as modest-to-moderate (correlations in the 0.30–0.50 s), except for Soothability.

Turning to the Lab-TAB (lower portion of Table 2), internal consistency estimates averaged 0.794 at 6 months and 0.786 at 12 months. Variation in internal consistency across composites was limited, with Anger measures tending to be highly consistent and Sadness measures somewhat less so. Although internal consistency was enhanced by the composite construction process and would be lower in a replication sample, we judged these levels of internal consistency as satisfactory.

We created an SES variable measuring mother and father education as well as family income. SES was inversely related to parent reported IBQ Fear at 6 and 12 months of age as well as IBQ Anger and at 12 months of age. It is possible that lower SES parents encounter more external stressors which affect their reporting of infant behaviors. Only one significant relation emerged between SES and observed Lab-TAB temperament (during Basket of Toys at 6 months of age, r = 0.13, p < 0.01); thus, we did not include SES in further analyses.

Inter-correlations of the IBQ scales are presented in the upper left portions of Table 3 (6 months) and 4 (12 months). These correlational patterns align with previous literature examining the IBQ (Putnam et al., 2001). Scales assessing infant negativity, namely Fear and Anger, were significantly correlated with each other at 6 and 12 months. Scales assessing infant regulatory behaviors (Soothability and Duration of Orienting) were also significantly related at each age. Smiling and Laughter was positively correlated with Soothability and Duration of Orienting at both ages. Only at age 12 months was the most hedonically positive scale (Smiling and Laughter) negatively correlated with the hedonically negative Fear and Anger scales. Finally, infants who were higher in Activity Level also expressed more emotionally driven behaviors, as reflected in the Fear, Anger and Smiling and Laughter scales at each age.

Correlations within the Lab-TAB episodes and dimensions are shown in the lower right portions of Table 3 (6 months) and 4 (12 months). The first result is that our strategy for creating cross-episode composites was supported although the magnitude of cross-episode convergence was usually modest. Episodes were significantly correlated (all ps < 0.01) with the other episode within each dimension at each age: Stranger Approach and Masks (fear) scores were related at 6 (r = 0.15) and 12 months (r = 0.18); Arm Restraint and Car Seat Anger were related at 6 (r = 0.28) and 12 months (r = 0.17); Arm Restraint and Car Seat Sadness were related at 6 (r = 0.29) and 12 months (r = 0.27); and finally, Puppets and Peekaboo were related at 6 (r = 0.28) and 12 months (r = 0.28). Thus, we created Lab-TAB dimensions for Fear, Anger, Sadness, and Positive Affect at each age.

In general, episodes within the negative temperament dimension were significantly intercorrelated (Stranger Approach, Masks, Arm Restraint, Car Seat). This pattern was also reflected in the relations among the composites, such that Fear, Anger, and Sadness were consistently related to each other at 6 and 12 months. Activity level was related to Anger, Sadness and Positive Affect episodes at each age as well. Further, Positive Affect was negatively correlated with Anger at 6 and 12 months.

Associations between conceptually related questionnaire and observational measures speak to criterion-oriented validity, and those associations were not strong, generally speaking. Parent reported infant fear was related to Car Seat anger and sadness at 6 months (but not the expected fear measures from Lab-TAB), but by 12 months of age parent rated fear was correlated with laboratory measured fear, anger, and sadness. Parent reported distress/anger was modestly correlated, in most but not all cases, with increased anger and decreased sadness in the laboratory at both 6 and 12 months. Parent reported smiling/positivity was unrelated to any laboratory assessed temperament at 6 months, but by 12 months it was related to increased anger and decreased sadness during the Arm Restraint episode. Parent rated infant regulation, namely the Soothability and Orienting scales, were inversely related to laboratory positive affect at each age, but increased fear and interest and decreased anger and sadness at 12 months.

Our third aim was to examine stability of temperament across time, and determine whether this was similar for boys and girls. Table 5 lists the mean levels of each parent reported and laboratory observed temperament dimension by age and gender. Parents reported girls to be very modestly higher in Fear at both 6 (Cohen's d = 0.21) and 12 months (Cohen's d = 0.15), which was a salient finding, given no other gender differences for parent reported temperament. A similar trend occurred for observed fear for the Stranger and in the Masks episodes of the Lab-TAB although the differences between boys and girls did not reach statistical significance (p = 0.06). However, when Lab-TAB episodes were combined to create an overall observed fear composite, girls exhibited more fear than boys at 6 months but not 12 months.

In the positive affect dimension, no differences in Puppets or overall positive affect at either age occurred, but a significant gender difference did occur for the Peekaboo episode. Boys were rated higher in positive affect coded from the Peekaboo episode of the Lab-TAB at both 6 and 12 months. Boys were scored higher in interest/persistence as measured by the Slides episode and in activity level across time (see Table 5).

Because we standardize the Lab-TAB raw data prior to creating episode scores, and are unable to examine mean level differences in observed temperament. However, we can look at rank order stability across time. Bivariate correlations indicated longitudinal (rank-order) stability in both the parent reported and observed measures of temperament; see correlations across time in Table 6 (left side of the table). The stabilities of IBQ scales (mother + father averages) ranged from 0.41 to 0.59. Although the stabilities of observed behavior reflected in the Lab-TAB composites was substantially lower than the questionnaire scales, only one Lab-TAB episode, Car Seat (Anger), did not show significant longitudinal stability between 6 and 12 months of age. As expected, the cross-episode dimension composites (last 4 rows of Table 6) were more stable than the single episode measures.

Regarding the stability of temperament across gender, correlations between 6 and 12 months for are presented on the right side of Table 6. We first correlated infant temperament measures across time for girls and boys, and then transformed the Pearson r values to z-scores to compare the degree to which stability existed across gender. For parent reported temperament, significant stability was observed for each scale, and this stability did not differ by infant gender. For Lab-TAB, boys exhibited significantly greater stability than girls in sadness and activity level, indicated by significant z-scores for Car Seat sadness, activity level, and the overall sadness dimension.

Although fine grained measurement of temperament is informative, researchers often examine infant behaviors in terms of much broader factors such as positivity or surgency, negative affect, and regulation (Putnam et al., 2001). In our study, within measurement method, temperament scales indicative of broader positivity, negativity, and regulatory dimensions were related. That is, infant smiling and laughter, or positive affect, was inversely related to negative dimensions such as fear; fear, and anger were positively related at a given age; and soothability and orienting were positively related. In addition, similar to Bridges et al. (1993), anger and orienting were inversely related.

When incorporating Lab-TAB assessment of infant behaviors, some research questions may pertain to an infant's discrete response to a specific situation, for example, a tendency to respond by struggling to distressing goal blockage such as in the Arm Restraint episode. The use of a single episode from the Lab-TAB may be sufficient to incorporate into such studies. However, to examine inter-relationships among temperament constructs, multiple episodes are obviously needed, and multiple episodes within a domain, such as anger, are needed to claim cross-situational generality. In our study, we first determined scores for individual laboratory episodes by calculating the mean and peak of each infant behavior. In this way, we consider an infant's characteristic reaction throughout the relevant stimulus exposure (mean of response) as well as the most extreme behavior exhibited (peak).

Additionally, temperament theory intimates that temperament reflects an individual's propensity toward emotional reactions; that is, individuals' tendency to respond similarly across multiple situations that share affective incentive characteristics. Accordingly, we also created composite scores to reflect broad (cross-response modality, cross-episode) temperament propensities. Scores from each of the two episodes chosen within each temperament dimension (i.e., fear, anger) were correlated at each age, indicating that the episodes designed to represent a temperament composite elicited similar behaviors from infants. These correlations, however, were only modest, indicating potent situational effects.

In behavioral research, numerous factors typically affect observed behaviors. The adult personality literature offers support for moderate cross-situational consistency of behavior, with many caveats and qualifications (Kenrick and Funder, 1988). Modest cross-episode convergence in the infant version of Lab-TAB may be due to multiple sources. First, infant temperament traits may be a weak organizer of behavior, relative to the situational differences between episodes with apparently similar affective incentives. Second, relative to adults, infants possess a less differentiated sense of self, less developed patterns of affective and self-regulation, and no developed social attitudes. Each of these features of personality may support temperament differences. Third, contextual or incentive differences between episodes may alter the patterning of individual differences in temperament. Or, stated another way, temperament traits may be narrower than researchers tend to believe; for example, “anger proneness” might be too broadly conceptualized. Fourth, state variables in infancy, such as hunger or tiredness, may be powerful enough to obscure temperamental differences. Fifth, the novelty of laboratory context is constant, in a sense, across all Lab-TAB episodes, and this novelty may also obscure temperamental differences that would be apparent in more familiar contexts. Thus, the finding that behaviors across episodes are only modestly correlated should not be surprising. Despite these competing influences on behavior, some evidence of cross-situational consistency in Lab-TAB responses is apparent.

Although most of what we know about infant temperament comes from research using parent reports, laboratory based assessment can elucidate patterns of emotional reactivity during specific situations. Parent reports and observed behavior are inconsistently correlated, and several investigators have elaborated potential reasons for this inconsistency (Rothbart and Goldsmith, 1985; Goldsmith and Hewitt, 2003; Saudino, 2003; Stifter et al., 2008). In a previous report on scale development and validation of the Preschool Lab-TAB (Gagne et al., 2011), laboratory-based assessments of fear, shyness, activity level and inhibitory control showed low to moderate correlations with corresponding mother- reported temperament dimensions. We and others (e.g., Gartstein and Marmion, 2008; Braungart-Rieker et al., 2010) found moderate convergence between parent reported and laboratory observed infant fear and anger, but not positive affect (Gartstein and Marmion, 2008) or interest/persistence (Matheny and Wilson, 1981; Carranza Carnicero et al., 2000).

The reason for discrepant findings for different aspects of temperament might be that infants and young children actually do exhibit distress across the home and laboratory environments more consistently than they express positive affect and interest/persistence in the two types of environment. Alternatively, parents may be more attuned to interpersonally disruptive infant distress or anger (Huebner and Izard, 1988) than to other less obtrusive temperamental traits. Thus, plausible reasons exist for lack of convergence between seemingly corresponding IBQ and Lab-TAB assessments in some cases. Nonetheless, by considering both observed and parent measures of temperament, we gain a clearer picture of the development of multiple temperament dimensions across infancy. Moreover, the pattern of significant correlations of IBQ by Lab-TAB findings in Table 4 is fairly systematic and sensible. For instance, fear is perhaps the most salient aspect of temperament to parents. Parents' IBQ Fear reports are related to a variety of the more objective Lab-TAB measures, a finding that suggests parents may over-assimilate infant behaviors to their own version of a fear/inhibition/shyness concept, which is broader than our concept of fear. An alternate interpretation of this finding is that the novelty of the laboratory context affects the assessment of multiple dimensions of temperament. Also, we observe cross-method correlations between activity level and positive affect, suggesting a role for approach motivation in structuring these correlations.

Consistent with previous literature (Else-Quest et al., 2006) we found gender differences in parent reported and observed temperament, such that girls were higher in the fear dimensions of temperament, and boys were higher in approach related behaviors such as positive affect and activity level.

Further, few studies use laboratory measures of infant temperament when assessing continuity. In ours and others' research, infants' expressions of fear, anger, and positive affect as well as activity level increase from 6 to 12 months (Carranza Carnicero et al., 2000; Rothbart and Bates, 2006; Planalp et al., 2016); however, less is known about infant sadness and interest/persistence. We found rank order stability of temperament across time, reflected by longitudinal correlations within scale, episode, or dimension.

Although our results indicate significant stability within temperament dimension, temperament can be less stable across developmental stages, such as the widely implicative transition at about age 8–9 months (Emde et al., 1975). For example, infants who express more distress at 6 months are not necessarily higher in anger or fear by school age (Lemerise and Dodge, 2008). Cognitive and motor development may impact the ways in which individuals can express emotion. Thus, although temperament is relatively stable within developmental periods such as mid-infancy we are not able to predict stability across longer periods of time. We discuss findings within each temperament dimension below.

Although we conclude that the Lab-TAB offers numerous advantages over subjective parental report, several limitations are also noteworthy. First, administering the full Lab-TAB battery, which we did not do in this study, is time-consuming and expensive, especially when the episodes are repeated to assess stability. Even with the 20 episodes described in the locomotor version of the Lab-TAB manual, we still only sample limited components of the infant's behavioral repertoire and thus only a limited range of affective individual differences. For instance, we do not tap the affective dimension of disgust. Also, for ethical reasons, we do not use more intense elicitors of negative affect that occasionally occur in daily life.

Another set of limitations is characteristic of most laboratory assessments in the social, affective, and even cognitive domains. Examiner differences, scorer sensitivity effects, infant state effects (hunger, tiredness), effects of the novel laboratory environment, and other subtle effects may impact results. We also examined convergent validity of the Lab-TAB episodes and dimensions with parent reported temperament, but did not address discriminant validity between temperament dimensions. Although we considered discriminant validity in the construction of Lab-TAB, we did not consider it in this paper. The issue of discriminant validity, particularly as it applies to Lab-TAB measurement of anger and sadness issue is complex. As explained above, we score anger and sadness from the same episodes, and anger and sadness are correlated due to the common element of measurement, much as questionnaire items on the IBQ that follow from the same stem are correlated.

More generally, formulations about discriminant validity often occur in the absence of theory that specifies how latent constructs should be related. We posit that temperament is affective individuality, so we must use affect theories to offer expectations about discriminant validity. For instance, popular circumplex theories of affect (e.g., Plutchik, 1980) specify that affects at opposite poles of the circumplex would be uncorrelated, but those nearby on the circumplex would be highly correlated. Other theories of affect generate different expectations for the association of components of affect and thus for expectation of what would constitute evidence of discriminant validity. Given these complications, we did not address discriminant validity comprehensively in this manuscript although we do so in more limited contexts in other papers (Goldsmith and Gagne, 2012; Clifford et al., 2015).

Despite these limitations, we provide more support and validation for the use of an infant laboratory based temperament assessment. The Lab-TAB indicated moderate relations with parent reports, consistency across time, and replicated gender differences from prior research. In addition, we provide information on the use of both episode-level vs. dimensional scoring, which allows researchers flexibility depending on their question of interest.

We conclude that employing the pre-locomotor and locomotor versions of the Lab-TAB as a behavioral assessment of temperament in infancy is feasible and fruitful for examining temperament using an individual differences approach. Researchers will vary on administration (single episodes for each temperament dimension vs. multiple episodes), coding and scoring (e.g., some will prefer to use more global scoring than we did, cf. Dyson et al., 2012). Lab-TAB can be used to examine temperament within age, longitudinally across infancy (as we did) and into childhood, and within and across temperament dimensions. In sum, the Lab-TAB is a useful tool for examining fine grained individual differences in early emerging infant temperament.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.