Critical thinking (CT) is one of the components contributing to good thinking (Norris & Ennis, 1989), It is also necessary ability for the citizens in democratic society. Especially in the age of the Internet, the variety and complexity of information are greater than in the past. CT is saliently significant.

    The perfect performance of CT was dependent on domain-general and domain-specific knowledge, CT skills and CT dispositions. Undergraduates reported to have great CT dispositions but performed medium degree of CT skills (Din, 2020). Therefore, it is important to design programs to advance students’ CT skills under the considerations of the factors affecting CT skills performance. CT skills include evaluation of arguments, inference, interpretation, recognition of assumptions, and deduction. These cognitive operations are conducted in working memory (WM) (Halpern, 1998). Epistemological beliefs could affect CT performance (Schraw, 2001). We examined the prediction of WM and epistemological beliefs in CT skills.

    WM was involved in the processing and storing of information. A few researches showed that WM predicted the performance of reasoning tasks (Barrouillet & Lecas, 1999; Chuderski & Necka, 2012; Copeland & Radvansky, 2004). Baddeley and Hitch (1974) argued that WM had three components: A phonological loop, a visuospatial sketchpad, and the central executive. Individuals kept the speech or written information in the phonological loop. The central executive was responsible for attention control and thus helped the information related to the task keep accessible. We hypothesized that WM predicted the performance of CT skills. Short-term memory (STM) was different from WM, and was only involved in the processing of information. It was also hypothesized that STM predicted CT skills. Processing speed could affect reasoning performance and thus was used as a control variable in this study. We want to learn the net prediction of WM in CT skill performance after controlling the effect of processing speed.

    Epistemological beliefs referred to what individuals thought of the nature of knowledge and knowing. The number of the dimensions of epistemological beliefs were dependent on how and what researchers conceptualized them as. In this study, the epistemological beliefs scale consisted of five subscales: certainty of knowledge, simplicity of knowledge, source of knowledge (authority), quick learning, and innate ability. Individuals with naïve epistemological beliefs thought that knowledge was fixed, the aggregation of many facts, and conveyed from the authority, and that learning was a simple and linear process. Those with sophisticated epistemological beliefs considered the reliability and validity of information and arguments. It was hypothesized that the more sophisticated epistemological beliefs, the greater CT skills. We speculated that individuals with more sophisticated beliefs and less WM span might be willing to give careful consideration to the questions or information in order to make up for the less WM span. It was hypothesized that WM interacted with epistemological beliefs.

    We recruited participants from three universities in Tainan City, Taiwan. The three universities were representative of the undergraduates with different academic achievement respectively. Data were collected from 161 undergraduates (51 males and 110 females) but one participant was not analyzed because of not completing the epistemological beliefs scale. The average age was 20.87 (SD = 1.53).

    Participants were tested individually in a 60-minute session and fulfilled a demographic survey, the adapted Personal Epistemological Beliefs Scale, WM span scale, processing speed scale, Frontal Assessment Battery (Taiwan ed.), and Watson- Glaser Critical Thinking Appraisal (WGCTA) (Chinese ed.) in order. WM Index and Processing Speed Index of Wechsler Adult Intelligence Scale-Fourth Edition (WAIS) (Chinese ed.) were used. Frontal Assessment Battery was used to measure the executive function. The Chinese edition of WGCTA had five subtests: inference, recognition of assumptions, deduction, interpretation, and evaluation of arguments. The higher scores in the adapted Personal Epistemological Beliefs Scale, the more naïve. The high scores in other measures implied better cognitive performance. WGCTA and WAIS did not provide percentile ranks for every subtest. The original score was used to indicate participants’ performance of every subtask.

    CT skills were correlated with WM index, processing speed, and executive function (rs[158] = .40, .25, .28, p < .001), but not correlated with epistemological beliefs. The WM Index of WAIS was constructed by four measures, one involving in the storing of information (STM) and the others involving in the processing and storing. It was found that CT skills was correlated with STM (r = .37, p < .001) and also correlated with the other WM measures (rs = .36, .32, .28, p < .001). The correlation among the five CT subskills and other cognitive ability measures was computed. It was found that deduction was correlated with STM, WM index, processing speed, and executive function (rs = .31, .30, .23, .21, p < .01). Interpretation was correlated with STM, WM index, processing speed, and executive function (rs = .34, .39, .18, .20, p < .05). Deduction was correlated with the whole epistemological belief (r = -.22, p < .01) and also correlated with three components (certainty, simplicity, and authority) (rs= -.29, -.25, -.22, p < .01). Other CT subskills were not correlated with the whole epistemological belief but correlated with a few components of epistemological belief. To our surprise, the innate ability subscale was positively correlated with the whole CT skills performance (r = .18, p < .05) and also with the subtests of inference, recognition of assumptions, and interpretation (rs= .16, .23, .17, p < .05).

   Regression analysis showed that after processing speed controlled, WM and executive function predicted participants’ CT skill performance (b = .33, .58, p < .05), but the whole epistemological belief did not. Concerning every CT subskill, participants with greater WM performed better interpretation and deduction (b = 0.001, .003, p < .01).

    A regression analysis was conducted to examine the prediction of the four memory measures, processing speed, and executive function in CT skill performance. STM and executive function were predictive of CT skill performance (b = 73.83, .70, p < .01) but not all measures involving in the storing and processing of information were significant. While every CT subskill was used as dependent variable, the similar outcomes were only observed in the regression on deduction and interpretation.

    Another regression analysis was conducted to examine the prediction of the five components of the epistemological belief. It was found that simplicity belief predicted CT skills (b = -22.39, p < .01) as we hypothesized. It was surprised that innate ability belief had positive prediction (b = 17.24, p < .05). However, all of the participants performed sophisticated belief on the knowledge simplicity subscale (below 2.5 points), and thus what we found in the above description was that the strength of sophisticated belief in knowledge simplicity was predictive of CT skills. Not all participants showed sophisticated belief on the innate ability subscale (132 ones below 2.5 and 28 ones above 2.5) and there was no significant difference in CT skills between these two groups of the innate ability belief.

    We tested the interaction of WM subtests and sophisticated belief in knowledge simplicity. No significant interaction was found.

    The findings showed that the CT skill performance was predicted by STM as well as by WM defined by Baddeley and Hitch (1974). However not all measures of the processing and storing of information were significant. It might be due to the multicollinearity among the three measures. On the other hand, it was implied that STM might be more important than WM for conducting CT skills. Sophisticated beliefs in simplicity knowledge were significant but other factors of epistemological beliefs were not. This finding explained why the whole epistemological belief did not predict CT skills performance. The interaction between working memory and epistemological beliefs was not significant. Based on our findings, it was suggested that the programs for promoting CT skills performance had to take WM and the simplicity belief of knowledge into consideration. Further study were suggested.