本研究檢測認知腳本失驗與對等推論對消費者反應之影響。研究－以腳本分析法瞭解受試者對熟悉之消費情境是否存有認知腳本，研究二透過實驗設計操弄認知腳本失驗與對等推論，藉以分析受試者情感與行為意圖反應程度。研究結果顯示，對於熟悉之消費情境確實可引出消費者之核心腳本，而操弄不同腳本對受試者情感與行為意圖反應發現，正向失驗帶來較高評價，負向失驗反應出較負面評價，結論符合傳統失驗理論。對等推論則在歸因過程中形成干擾效果，並對正負向失驗腳本產生擴大作用，形成「高者愈高、低者愈低」效果，但對標準腳本情境則不會產生干擾作用。

This article applies the underlying reason behind disconfirmation model and correspondent inference theory. The reason was supported by Binter, Booms, and Tetreault' (1990) result. Disconfirmation leads to attribution, which leads to judgment of satisfaction. The authors propose those consumers' feelings and judgments in case of disconfirmed expectations will be different, and conjecture there is an interactive effect between the cognitive script of disconfirmation and correspondent inference. The attribution effect of correspondent inference may impose a significant moderating effect on consumer responses. After reviewing related literature, the authors advocate that the script-theoretic approach provides an opportunity to clarify the expectation components of consumer behavior models. Two studies examine whether the consumer responses depend in part on the cognitive script of disconfirmation and correspondent inference. Study 1 elicits consumers' general script of a familiar professional service. Study 2 shows that the cognitive script of disconfirmation and correspondent inference has an interactive effect on participant's responses. Script theory is more applicable to industries with high interpersonal contact (Anderson, 1983), so the authors chose hairstyling as the context in this study because of its high familiarity and consumption frequencies. In Study 1, one hundred twenty-six samples from several hair salons participated in this study. Their average age was 31.23 years (SD=7.6). Participants who were asked about a social interaction phenomenon they had experienced at a hair salon. Participants wrote down all the events involved in the service encounter in sequential order. According to script analysis (Bower et al., 1979), it takes three steps to obtain a standardized script (i.e., neutral script). The result of elicitation of a hairstyling script shows that totally fifty-three actions were mentioned by participants. Second, the authors retained sixteen actions, those with a mention frequency of 40% or above, to comprise the core script. Third, by using paired-comparisons, the authors constructed a general hairstyling script (i.e., neutral script). The purpose of Study 2 was to investigate the strength of the moderating effect of correspondent inference on the relationship between cognitive script of disconfirmation and consumer’s responses. In order to obtain a better understanding about the disconfirmation script occurring during the consumption process and their outcomes, the authors executed a pilot study with the critical incident technique (CIT). The authors collected one hundred thirty-one satisfactory and one hundred nine dissatisfactory incidents. The authors used these incidents to explore the most realistic positive/negative disconfirmation situations for further experiment. The positive script was named ”warm treatment script” and the negative script was labeled as ”blame-sidestepping script”. They will be used as the treatments for the experiment of Study 2. Study 2 manipulated the disconfirmation script and correspondent inference to display variety and demonstrates the interaction effect on the consumer's judgment. A total of one hundred sixty-eight undergraduate students participated in the experiment for course credit and were assigned randomly to one of the three between-subjects cells. Each respondent filled out one kind of situational script questionnaire. A 3 X 2 (script(S): positive/neutral/negative X correspond inference score (CI): high /low) mixed design was employed in this experiment. The first factor was manipulated as a between-subject factor. The neutral script derived from Study 1. The positive and negative scripts differing on expectation were chosen on the result of the pilot study (described earlier). Correspondent inference was manipulated by within-subjects. The median split (M(subscript median)= 35) of CI divided the respondents into two levels for low and high. Two manipulation checks were handled. The first was for script manipulation. Its purpose was to find out whether the participants all activated from the same controlled cognitive script (i.e. core script) prior to the experiment. Manipulation was implemented by adding several irrelevant actions (Smith and Houston 1985) to the core scripts from Study 1. No significant difference was shown in script scores among the participants. In other words, the three groups (positive/neutral/negative) of participants all activated from the same core script. The second manipulation was to check the level of the perceived upper or low disconfirmation expectations. The participants were asked to anchor the matching level using a five-point Likert scale. As expected, negative script was perceived to be more unexpected than neutral script and positive script. The two studies expounded in this article converge on several conclusions. First, the results clearly indicate the consumer expectation can be elicited through cognitive script because we found that the general public has a cognitive script for the specific service(e.g. hairstyling) s/he is familiar with. Second, disconfirmation is a robust phenomenon, as it can be conducted by cognitive script analysis. Higher positive responses come from positive disconfirmation script; in contrast, negative script leads to more negative responses. Finally, attributing effect of correspondent inference may moderate disconfirmation cognitive script on the participant's perception and behavioral intention. The augmenting effect is strongly proved. Given a positive disconfirmation script, respondents with higher correspondent inference feel even more satisfied than in neutral circumstances; in the same vein, given a negative unexpected script, their dissatisfaction exceeds that felt by those with low correspondent inference. Moreover, correspondent inference only has the moderating effect in case of disconfirmation cognitive script. It has no bearing in neutral circumstances.