Neuropsychological impairment has been suggested as a central factor underlining poor insight. Poor insight is associated with a secondary deficit in neurocognition due to structural [10-12] and/or functional brain deficits [4,13], especially frontal or parietal dysfunction [14,15]. Aleman et al. (2006) [9] have demonstrated that poor insight is associated with poor functioning in a range of cognitive domains, including intelligence quotient (IQ), memory, and the set-shifting and error-monitoring aspects of executive function. Several studies have also noted that metacognition has the potential to influence insight in individuals with schizophrenia [16-18]. The term "metacognition" was first coined by Semerari et al [19], and is defined as the "general capacity to think about thinking" [19,20]. Of note, metacognition is considered to concern a wide range of internally and socially driven cognitive acts [19,21]. Through metacognition, individuals not only process information that they encounter, but they can also react to and think about their own mental states and those of others [19,21]. Specifically, there is growing evidence [17,18] that metacognition is not only one of many clinical or psychological variables linked to insight in schizophrenia, but metacognition is also a factor that moderate the effects of other factors such as self-reflectivity (the ability to comprehend one's own mental state) [17-19,21,22] and mastery (the ability to form knowledge about one's own mental states and those of others and to use that knowledge to response to psychological challenges) [17-19,23] that underline an individual's awareness of illness. Lysaker et al. (2011) [18], for instance, suggested that metacognitive abilities, rated by the Metacognition Assessment Scale [19], may be linked to insight in individuals with schizophrenia independent of concurrent impairments in neurocognition. Thus, poor insight in schizophrenia may result in part from deficits in metacognitive capacities, namely self-reflectivity and mastery.

According to Beck et al. (2004), an important extension of the insight concept was introduced with the description of "cognitive insight," which was defined as a patient's current capacity to evaluate his or her anomalous experiences and atypical interpretations of events [24,25]. Those authors provided a conceptual dissociation between clinical and cognitive insight, suggesting that cognitive insight is a form of cognitive flexibility that involves first an ability to distance oneself from distorted beliefs and misinterpretations and then to reappraise these beliefs and recognize erroneous conclusions [24,25]. In a more specific way, at least four aspects of cognitive insight can be influenced by psychosis, according to Beck and Warman's research [25]: (a) impairment of the ability to be objective concerning delusional experiences and cognitive distortions, (b) reduced capacity to put these experiences into perspective, (c) unresponsiveness to corrective information from others, and (d) overconfidence in delusional judgments. Recent findings have also highlighted the potential importance of cognitive insight as a mediator of response to cognitive behavioral therapy of psychosis [26]. In fact, Beck's theory offered a stronger theoretical basis for cognitive insight and supported the contention that it is both identifiable and quantifiable [24,25]. Thus, cognitive insight was first operationalized with the publication of the 15-item Beck Cognitive Insight Scale (BCIS) [24,25].

The initial study by Beck et al. showed that the BCIS could measure individuals' capacity for distancing themselves from and re-evaluating anomalous beliefs and misinterpretations [24,25]. The BCIS, which is a 15-item self-report measure, is composed of two subscales: self-reflectiveness (SR) and self-certainty (SC) [24,25]. The former includes items measuring objectivity, reflectiveness, and openness to feedback, and the latter measures the certainty of one's beliefs and judgments [24,25]. Beck and colleagues proposed that high levels of certainty might diminish the capacity for self-reflection; thus, a composite index providing an estimate of overall cognitive insight was calculated by subtracting the score for the SC subscale from the score for the SR subscale [24,25]. The two subscale scores were only weakly intercorrelated, which indicated that they represent two different dimensions of cognitive insight [24]. The BCIS in this regard proved to be an indirect tool for evaluating the impairment of the 'higher level' functions in schizophrenia, and, more specifically, impairment in the process of distancing oneself from highly salient (delusional) beliefs and viewing them in terms of executive functions [24,25,27].

The majority of studies that have investigated the relationship between the overall cognitive insight of schizophrenia patients (measured by the composite index scale of the BCIS) and clinical insight (measured by the Scale to Assess Unawareness of Mental Disorder [24,28-30], the Positive and Negative Syndrome Scale [29-33], and the Birchwood Insight Scale [34]) have found that these two variables are significantly related and demonstrate convergent and criterion validity, respectively. The reliability and validity of the BCIS have been demonstrated in a mixed group of inpatients with psychosis and depression [24,35], groups of inpatients or outpatients with schizophrenia spectrum disorders [28,31,34], and a group of patients with bipolar disorder [32,35]. The BCIS has also been applied to nonclinical populations [32,33,36-38] and the internal consistency of the BCIS is similar between clinical and nonclinical samples [32,37].

A review of the literature indicates that clinical insight is associated with depression in patients with psychosis [33,39-41]; however, the findings of studies examining the relationship between cognitive insight and depression have been mixed. Two studies [33,37] found a correlation between depression as measured by the Beck Depression Inventory-II (BDI-II) and cognitive insight in patients with schizophrenia or schizoaffective disorder; however, another study [24] did not find such a correlation in individuals with psychotic disorders. In addition, Pedrelli et al. did not find an association between cognitive insight and depression as measured by the Hamilton Rating Scale for Depression in middle-aged and older patients with schizophrenia and schizoaffective disorder [34]. However, in an investigation of psychotherapy for individuals with schizophrenia or schizoaffective disorder, Granholm et al. (2005) [26] discovered a relationship between increased cognitive insight and increased depression midway through the treatment. To date, researchers have become increasingly interested in the relationship between cognitive insight and depression in individuals with psychotic disorders. However, there has thus far been relatively little research into this area using a non-psychiatric population sample. Two studies that investigated the BCIS with a normal population did not include a measure of depression [36,38], which makes it difficult to draw general conclusions about the relationship between cognitive insight and depression.

The BCIS has been used for comparisons between individuals with a psychotic diagnosis and healthy controls. One earlier study reported that patients scored significantly higher than controls on SC, but differences in SR were not observed between the two groups [37]. In another study, Engh et al. (2007) [32] found no difference in SR or SC subscales between individuals with schizophrenia, those with bipolar disorder, and normal controls. However, Martin et al. (2010) [38] found that healthy controls exhibited higher SR, lower SC, and a higher composite index than patients with schizophrenia. The failure of some previous studies to differentiate the SR between patients and controls could be due to a high percentage of SR items being omitted by the controls [32,38], cultural differences in the way individuals understand questions on the scale [37,38], and insufficient sample size [32,37,38]. At present, high scores on the SR subscale and low scores on the SC subscale are regarded as being normal [24,25,38,42]. However, this theoretical view has not been sufficiently supported by direct research to clarify possible impairments in psychosis and to answer the question as to whether increased SR and decreased SC are evidence of improvement [38]. In addition, cutoff scores that would allow a categorical determination of the presence or absence of impaired cognitive insight, as measured by the composite index, have not been clearly determined. In light of these concerns, the present study investigated the psychometric properties and factor structure of the BCIS with a large nonpsychiatric population and compared the results to those collected from individuals with schizophrenia.

Researchers and clinicians have expanded upon Beck's original work with numerous studies focusing on the extent to which the construct of the BCIS contributes to our understanding of a wider range of cognitive insight. To date, the BCIS has already been translated into several languages, including Turkish [28], French [31], Norwegian [32], Japanese [33], Spanish [43], Korean [44], Chinese [45], and Taiwanese [46], and its validity and psychometric properties have been reported in each of these languages. Research focusing on the discriminative properties of the BCIS (i.e., thresholds that draw on the combination of sensitivity and specificity) is scarce, however. To date, only one study has reported that the composite index reliably discriminates between outpatients and nonpsychiatric individuals (i.e., AUC, 0.641; SE, 0.033; 95% Confidence interval (95% CI), 0.575-0.707; nonparametric P < 0.001) [38]. Visual inspection of the ROC curve suggested that no single part of the curve maximizes specificity and sensitivity [38].

Two factors, SR and SC, emerged in exploratory factor analyses using principal axis factoring and varimax rotation; the two factors accounted for 39.5% of the variance. Interestingly, the two factors were the same as the factors determined by Beck et al. in the original BCIS [24] in a sample of inpatients with schizophrenia, schizoaffective disorder, or mood disorder; however, the factors differed from a study by Kao et al. (2010) [46], conducted in samples of patients with schizophrenia, schizoaffective disorder, major depressive disorder without psychotic features, and healthy controls. The present findings were also consistent with a previous study by Martin et al. (2010) [38], which confirmed that the basic factor structure and internal consistency of the BCIS were similar for the normal population. Recently, Uchida et al. (2009) [33] also observed acceptable internal consistency of the BCIS in a sample of nonpsychiatric Japanese individuals. These findings support the generalizability of the two-factor model of cognitive insight to both nonpsychiatric and psychotic populations. The test-retest reliability intraclass coefficients of the Taiwanese BCIS confirmed the stability of cognitive insight in nonpsychiatric and psychotic populations, thus indicating the reliability of the Taiwanese BCIS.

The alpha coefficient values for the SR and SC subscale scores in the nonpsychiatric group were 0.75 and 0.78, respectively, and these values indicated that the internal consistencies of the Taiwanese BCIS subscales were adequate for research purposes. Moreover, the present sample's alpha coefficients for outpatients who had been stabilized in an outpatient setting were higher than the alpha coefficients (0.68 for SR and 0.60 for SC) that Beck et al. [24] found for their acute sample. These latter alpha coefficients, however, are similar in magnitude to those reported by Pedrelli et al. (2004) [34] (i.e., 0.7 for SR and 0.55 for SC) for 164 middle-aged and older outpatients diagnosed with either schizophrenia or a schizoaffective disorder. A partial explanation for the inconsistent results may lie in the fact that these low coefficients alpha are partially attributed to the severity of the patients' current symptoms, especially for those patients with thought disturbances and concentration difficulties [24]. In the psychotic group, the alpha coefficient for the SC scores was < 0.7, but this value was considered acceptable for the present research purpose because these subscales were composed of fewer that ten items [24,62,63].

To our knowledge, this is the largest population-based examination of the factor structure of the BCIS. Testing for multiple a priori models and indices has established that the BCIS has acceptable validity and reliability in both nonpsychiatric and psychotic samples. Consistent with the multidimensional view of cognitive insight, the CFAs supported a two-factor structure underlying the BCIS in non-psychiatric and psychotic populations. Most of GFI statistics performed better with GFI, AGFI, TLI, and RMSEA. However, two of them (CFI, NNFI) are slightly lower than the cut off previously recommended [55,56] and could be considered as acceptable. In CFA, previous studies have used the ratio χ² /df as an index to assess the consistency of different models' factor structure. Having χ² /df < 2 means that the model fits well, but it is important to note that the larger the sample size, the bigger the χ² [64]. Because of the large sample size (n = 507) in the present study, it might not be appropriate to use χ² /df as the index to assess the fit of the model.

The present findings add to the previous studies [38] that have suggested that participants with psychosis have impaired SR when compared with healthy controls. A partial explanation for this may lie in the fact that deficits in metacognition are a stable feature of schizophrenia [22] and that such deficits may be key features of the relationship between schizophrenia and self-reflectivity [17-19,21,22]. Although the same spectrum of cognitive dysfunctions should also be expressed in nonpsychiatric individuals, the severity of cognitive dysfunction should increase towards the schizophrenia end of the continuum. In other words, the core clinical and subclinical features of schizophrenia, such as limitations in the individuals' metacognitive capacity to reflect upon their difficulty in thinking as well as to recognize and correct their errors, will be found across the dimension of cognitive insight [17,18,21,22].

Previous studies investigating the relationship between cognitive insight and depression have reported conflicting results. Several studies have replicated the finding of Beck et al. [24] that depression was not related to cognitive insight in patients with a psychotic diagnosis [30-32,34]. Three studies [35,37,65], however, found that better cognitive insight (composite index and SR) is related to worse depression in patients with schizophrenia or schizoaffective disorder when looking at a cross-section. Our results replicate earlier findings [35,37,65] concerning the association between cognitive insight and depression in patients with schizophrenia. The results suggest that schizophrenia patients with depression have better cognitive insight than those without comorbid depression. This finding could be due to the fact that as individuals begin to believe they are mentally ill, they are confronted with and demoralized by self-stigma [17,45,66-70], which is, in turn, related to poor self-esteem [17,45,67-70], depressed mood [17,68], and hopelessness [17,68-70]. In empirical studies of this hypothesis, previous research has noted that the association of insight with depression [68], low quality of life [68], negative self-esteem [68,69], hope [69], and social functioning [69] is moderated by self-stigma. A previous study also reported that having a greater level of cognitive insight was significantly associated with experiencing a high level of self-stigma [45]. The increased SR or cognitive insight in such individuals can be used to indicate a negative attitude to self or "bad me" distortions of thinking [71,72], given that individuals with depression are characterized by a variety of negative biases about the self or negative self appraisal [72,73]. It may be of interest for future research to examine whether certain kinds of psychosocial intervention might be beneficial in improving SR or cognitive insight, while decreasing rather than increasing depressive symptoms. Furthermore, the relationships between cognitive insight and depression could be studied longitudinally. Unexpectedly, the present study found that depression was not significantly correlated with the BCIS index or its subscales in nonpsychiatric individuals. It is unlikely that this was due to a floor effect or a limited range of scores on the BDI (see Table ​Table1).1). Thus, it can be reasoned that the relationship of the SR dimension of cognitive insight to depression is not driven by a specific cognitive process such as self-reflectivity but rather is related to the experience of adopting the stigmatizing label and internalizing the stigma. As with unexpected and negative findings, future studies are required before any firm conclusions can be drawn.

The area under the ROC curve in the present study demonstrated that the Taiwanese BCIS satisfactorily differentiated patients with schizophrenia from the nonpsychiatric controls. In addition to being useful as an outcome measure for assessing responses to cognitive insight, the BCIS might be employed as a screening instrument to identify patients who are willing to examine their erroneous beliefs and accept the types of corrective feedback afforded by cognitive therapy [26]. Colis et al. (2006) also suggested that the BCIS might be used to screen for patients who exhibit low levels of cognitive insight and require more stabilization with psychotropic medications before engaging in psychotherapy [35]. Therefore, in the present study, our goal was to validate use of the BCIS to screen for cognitive insight in clinical settings where one may want to maximize the sensitivity of a test. It is important to note that the risks associated with the profound damage of psychotic relapses and the difficulties in treating psychotic disorders are high. In addition, the costs associated with further assessment of those patients with psychosis who appear to be unaware of their illness, their need for treatment, and the social consequences of psychotic disorder are relatively low. Thus, a screening measure of cognitive insight that would identify this "high-risk" population would be valuable, providing good sensitivity at the cost of some specificity. For this reason, we suggested a BCIS index cutoff value of 3 (sensitivity, 87%; specificity, 51%) for the detection of moderate or severe impairments in cognitive insight in a Taiwanese population. However, cutoff scores provided in this and similar studies largely depend on sample characteristics (in particular, sample sizes and types of psychiatric disorders or conditions considered in the studies) as well as on the procedures used during assessment, diagnosis, and data analysis. Given the number of differences across studies in these and other variables, it is difficult to compare the current results with previous evidence; cutoff scores provided in this study should therefore be treated with caution [74].