Results and Discussion

The data set was first trimmed in order to reduce the effect of outlier latencies.

First, all response times above 3000 ms and below 100 ms were cut. Second, an upper bound mean was calculated for each individual participant by adding 2.5 standard deviations to each individual's mean latency score. All response times that fell above a particular individual's upper bound score were cut. This trimming procedure resulted in the elimination of 0.03% of overall responses. Third, because the speed of participants’

incorrect responses would have been difficult to interpret in terms of accessibility (see Bargh, Chaiken, Govender, & Pratto, 1992), we excluded from the analysis incorrect

9 Each block contained the same amount of generated words and non-words.

10 As the critical words differed for each participant in the experimental condition (and therefore for each yoked control participant), the computer program automatically matched each critical word with a non-word out of a beforehand created non-non-word pool.

responses (2.5 % of the responses). Finally, the response times were log-transformed (natural logarithm function) in order to adjust for the skewness of the original latency distributions (Winer, 1971); for the sake of clarity, however, non-transformed means are presented.

Response times were collapsed across the eight situation-words and the eight behavior-words¹¹. A 2 (condition: implementation intention vs. yoked control) x 2 (word type: situation-word vs. behavior-word) analysis of variance (ANOVA) on the mean of the transformed response latencies was conducted, with condition as a between-participant variable and word type as a within-participant variable. The analysis yielded a significant main effect for condition, F(1,60) = 6.33, p < .05. There was no significant main effect for word type, F < 1 and no significant word type "

condition interaction, F < 1. Simple effect analysis showed that response latencies significantly differed between conditions for situation-words, F(1,60) = 6.24, p < .05, and for behavior-words, F(1,60) = 5.9, p < .05. Namely, participants in the experimental condition responded faster (in ms) to situation-words (M = 634) and behavior-words (M

= 632) than participants in the yoked control condition (M = 717, M = 708, respectively). The results are presented in Figure 5. In addition, mean response times and their standard deviations are indicated in Appendix B.

The results of Experiment 1 support the predictions. First, applying the standard assumption that faster latencies reflect more activation, the reported faster response latencies for situation-words in the experimental condition compared to the yoked control condition replicate previous findings of implementation intention formation leading to heightened accessibility of the specified situation (Gollwitzer, 1993, 1996).

However, this finding provides more straightforward evidence of heightened activation of the mental representation of the specified situation (upon having formed an implementation intention), as the measure of activation consisted of an indirect measure, namely lexical decision (whereas previous findings are based on direct measures such as attention, recall, and detection rates.)

11 Separate analysis for critical words associated to different goal-domains (health and interpersonal relation) did not reveal any significant differences on participants’ response latencies. Therefore, data corresponding to both goals were collapsed and will not be discussed further.

Figure 5. Response latencies on lexical decision for situation- and behavior-words as a function of condition Experiment 1 (implementation intention condition [n = 31], yoked control condition [n = 31]).

Second, and of special interest, faster response latencies in the experimental condition compared to the yoked control condition were also found for behavior-words, indicating that the formation of an implementation intention not only enhances the accessibility of the specified situation, but also of the goal-directed behavior. Hence, this finding represents initial evidence that heightened activation of the then-component occurs upon having formed an if-then plan. In sum, the present findings indicate that linking a specified situation to a goal-directed behavior in an if-then format (i.e., forming an implementation intention) leads to enhanced activation of the mental representation of both components of an implementation intention (i.e., the specified situation and the goal-directed behavior). In addition, given that the shorter latencies for the situation- and behavior-words in the experimental condition compared to the control condition were found after a time-delay between the encoding of the words and the measurement of activation, these results are interpreted as evidence for persisting activation of the two components of an implementation intention related to their superior status, and not due to their semantic relation (as activation that results from a semantic relation of two constructs has a faster rate of decay than activation due to

constructs’ functional relation; e.g., Higgins, Bargh, & Lombardi, 1985; Srull & Wyer, 1979).

However, one could argue that this conclusion is premature. Although the basic design of Experiment 1 was plausible as starting point of primary investigation of a construct’s mental representation (namely of the mental representation of implementation intention’s components), it allows for alternative explanations of the results. First, one could argue that the higher activation of the situation- and behavior-words in the experimental condition compared to the yoked control condition is not a result of a superior status of the specified situation and the goal-directed behavior due to having formed an implementation intention, but rather a generation effect of the components. In other words, shorter latencies for the generated situation- and behavior-words in the experimental condition compared to the longer latencies for the provided situation- and behavior-words in the control condition might be a mere consequence of better encoding of these words due to their generation in the experimental condition.

Second, an alternative explanation for the heightened activation of the specified situation and the goal-directed behavior in the experimental condition could be a sole consequence of goal activation. Kruglanski and colleagues (e.g., Kruglanski et al., 2002; Shah & Kruglanksi, 2002, 2003) have demonstrated the capability of goals to prime (i.e., to activate) their attainment means and attributed this finding to the cognitive association of goals and means (see section 1.3.3 Goal Systems Theory herein). Therefore, one could argue that having asked participants in the experimental condition of Experiment 1 to generate personally relevant goals resulted in activation of possible attainment means (i.e., viable situations and goal-directed behaviors) independent of the formation of plans. In other words, participants that held a certain goal might have been more sensitive to situations and behaviors that could fulfill the goal, and therefore, the same findings (shorter latencies for the situation- and behavior-words compared to participants in the control condition) might have been attained without having participants in the experimental condition explicitly generate viable situations and goal-directed behaviors in order to form implementation intentions.

Third, as the LDT only included critical words and non-words, but did not include any word controls to the critical words (i.e., other words that did not appear in the

implementation intentions)¹², one might even argue that faster response times in the implementation intention condition for both critical word-types, namely for the situation- and behavior-words, do not indicate a heightened accessibility of the mental representation of both components, but rather a motivational effect that leads participants to an overall superior performance. Experiment 2 was designed to test the robustness of present findings by addressing their possible alternative explanations.

2.2 EXPERIMENT 2: PLAN ACTIVATION AS FUNCTIONAL RELATION