Understanding Self-Criticism: A Systematic Review of Qualitative Approaches

Abstract

Objective

Self-criticism (SC) is a central transdiagnostic factor in several psychopathological conditions, influencing the development and maintenance of symptomatology. The importance of this construct has stimulated quantitative and qualitative research about it. The main purpose of this systematic review is to highlight which qualitative methods have been used most frequently and which are most suitable for studying SC.

Method

We conducted a systematic search by searching the following databases to identify publications: PsycINFO, PsycARTICLES, MEDLINE, Scopus, Web of Science, PubMed, and ERIC (all years up to and including January 2024). We aggregated search terms into two concepts for all databases: “self-criticism” and “qualitative analysis”.

Results

After removing duplicates, we screened a total of 852 records, resulting in the identification of 28 full-text articles that we assessed for eligibility. Upon closer examination, there was consensus that 16 of those studies met the inclusion criteria. Data extracted from the included studies revealed the lack of a shared approach regarding qualitative analysis of SC. Some studies employed a top-down coding approach, others used a bottom-up coding approach, and a few combined both methods. Consensual qualitative research and thematic analysis have been used most widely in the field. Furthermore, the data revealed no consensus among researchers on the conceptualization of SC. SC categories identified through qualitative analysis frequently do not align with existing theoretical models, and these categories are rarely re-examined in subsequent studies.

Conclusions

There is a need to test existing theoretical models of SC through qualitative analysis and to develop new models that should be examined with qualitative and quantitative methods in different clinical populations to fully capture the complexity and multidimensionality of SC.

Introduction

Relevance of self-criticism

In recent years, the construct of self-criticism (SC) has received increasing interest in the field of scientific research. SC is considered a transdiagnostic factor and a central phenomenon in various psychopathological disorders, influencing their development and maintenance, acting as a vulnerability factor, accentuating symptom presentation, or hindering and inhibiting psychological changes (Papa et al., 2024; Kim et al., 2020; Löw et al., 2020; Rose & Rimes, 2018; Werner et al., 2019; Warren et al., 2016; Gilbert & Irons, 2005).

Several authors have explored its implications in psychopathology (Löw et al., 2020; Werner et al., 2019). Indeed, a large body of research has highlighted how SC is present in different psychopathological conditions, including mood disorders, eating disorders, social anxiety, obsessive-compulsive disorder, psychotic disorders, personality disorders, Tourette syndrome, body-focused repetitive behaviors like trichotillomania, chronic fatigue syndrome, and stress (Löw et al., 2020; Werner et al., 2019).

SC is also considered to be a negative outcome factor in psychotherapy (Kannan & Levitt, 2013; Löw et al., 2020). Individuals who exhibit high levels of SC often derive limited benefits from psychotherapy, display resistance to treatment (Marshall et al., 2008; Rector et al., 2000), manifest difficulties in maintaining and establishing a therapeutic alliance with their therapist (Whelton et al., 2007, and exhibit greater resistance in restructuring their dysfunctional beliefs (Gilbert, 2014). Patients who are particularly self-critical do not improve with standard cognitive therapy. Despite understanding the illogical nature of their dysfunctional negative thoughts, actions, or processes, they continue to feel uneasy, guilty, and self-accusatory, and they maintain a judgmental attitude toward their inner experiences, displaying higher levels of psychological distress (Petrocchi et al., 2021; Saliani et al., 2024).

The relevance of the construct has led several authors in recent years to systematically study its role both in psychopathology (Löw et al., 2020; Werner et al., 2019; McIntyre et al., 2018) and in psychotherapeutic interventions (Wakelin et al., 2022; Vidal & Soldevilla, 2023), analyzing the effective interventions in reducing SC. The currently available results document that interventions based on self-compassion (Wakelin et al., 2022; Vidal & Soldevilla, 2023; Kannan & Levitt, 2013) are effective, although research in this area is growing and there is cross-sectional interest in other therapeutic approaches (Young et al., 2003; Arntz, 2018; Pugh, 2019).

Definition of self-criticism

The clinical attention to SC gained momentum with the advent of psychoanalysis. Since then, various theorists, including Freud (1917) as well as cognitive-behavioral (Beck et al., 1979) and humanistic-existential (Deci & Ryan, 2012; Rogers, 1961) theorists, have presented different conceptualizations of SC (Chang, 2008). Despite the diverse theoretical perspectives, these authors have consistently emphasized that SC-related feelings and cognitions are integral components of psychopathology.

SC involves negatively judging one’s actions, thoughts, and self, accompanied by feelings of worthlessness, inability, and inadequacy (Blatt & Zuroff, 1992; Gilbert et al., 2004; Shahar, 2015). The process also encompasses self-scrutiny, negative self-evaluation, self-judgment, and negative self-talk (Blatt & Zuroff, 1992; Gilbert et al., 2004; Shahar, 2015), giving rise to negative emotional reactions such as shame, anger, guilt, and self-loathing (Whelton & Greenberg, 2005), and contributing to experiences of disapproval and criticism (Blatt, 1974; Kannan & Levitt, 2013; Shahar, 2015). SC is an internal process in which individuals judge and attack themselves using hostile self-dialogue: people can activate an integrated system of beliefs, emotions, and attitudes that often take the form of negative and coercive inner dialogues and hostile inner voices (Longe et al., 2010). Various theoretical models have been developed to conceptualize SC (for a comprehensive review of the literature, see Zaccari et al., 2024). Among the most well-known models is the two-polarity model of personality and psychopathology development by Blatt and colleagues (Blatt, 2007, 2008; Blatt et al., 1976; Blatt & Luyten, 2009). They conceptualized SC as an intense preoccupation with self-definition and characterized it as a personality flaw focused on concerns with success and self-esteem. High levels of SC lead to emotions of failure, guilt, inferiority, and shame (Blatt & Luyten, 2009; Whelton & Greenberg, 2005). Based on this conceptualization, Blatt (2004) distinguished two types of SC: comparative SC, understood as the tendency to compare our characteristics with those of others (perceptions of hostility and criticism from others), and internalized SC, deriving from the comparison with one’s own ideal (i.e., feeling below one’s ideals).

Gilbert (2014) made a further theoretical contribution by explaining SC through a cognitive-evolutionary approach, closely linked to a specific intervention for addressing SC: compassion-focused therapy (CFT; Gilbert, 2010, 2014). According to this conceptualization, SC is a phenomenon connected to the motivational system for competition and social rank. Gilbert et al. (2004) highlighted two forms and functions of SC: a sense of inadequacy and inferiority and a sense of disgust and hatred toward oneself called inadequate self and hated-self. Inadequate self indicates that SC “focuses [s] on a sense of personal inadequacy” and the need “to improve and do better” (linked to competitive concerns). Hated-self is “linked to self-hating and wanting not to improve aspects of the self but to get rid of them” (Gilbert, 2022, p. 170).

Shahar (2015) formulated the Axis of Criticism Model (ACRIM). The model posits that the origins of SC lie in parental criticism and in the child’s (and later the adolescent and adult’s) failed attempt to develop their true self through authenticity and self-knowledge (SK). According to this perspective, high levels of SC stem from a lack of authenticity and self-knowledge as well as from experiences of criticism expressed by significant others. Shahar (2015) characterized self-knowledge as a distorted manifestation of SK that creates a gap between an individual’s evolving self-knowledge and accurate self-knowledge.

Despite a general agreement among scholars that SC entails a negative internal dialogue and a critical and hostile self-judgment (Gilbert et al., 2004; Shahar, 2015), the literature has provided diverse descriptions and definitions regarding the multidimensional nature of SC (Rose & Rimes, 2018). This diversity in SC definitions may be linked to the development of different theoretical models or to the overlap of SC with other constructs (Zaccari et al., 2024).

Self-criticism research approaches

Because SC has been implicated in psychological suffering, there has been intense research interest regarding the detection of SC. In fact, the importance of this construct in psychopathology and, consequently, the development of the aforementioned interventions, has stimulated researchers to explore the construct from empirical as well as qualitative perspectives.

From the perspective of empirical research, several instruments aimed at detecting SC have been developed based on the theoretical models of experts in the field of SC, including the Forms of Self-Criticizing/Attacking Reassuring Scale (FSCRS; Gilbert et al., 2004) and the Levels of Self-Criticism Scale: Comparative Self-criticism and Internalized Self-Criticism (LOSC; Thompson & Zuroff, 2004). Other instruments such as the Self-Critical Rumination Scale (SCRS; Smart et al., 2016) and the Depressive Experiences Questionnaire (DEQ; Blatt et al., 1976) contain a few items or factors related to SC, but they do not specifically detect SC. This fact highlights that even concerning measurement tools, there is an overlap between SC and other constructs or psychological processes. The content of the questionnaires varies depending on the authors’ theoretical orientation. Therefore, some authors (e.g., Rose & Rimes, 2018) have recommended future research to study valid and reliable measures of SC to ensure adequate detection and to capture greater subjective variability that could account for the complexity of the construct even in different clinical conditions (Zaccari et al., 2024). Considering these limitations, several authors have focused on studying SC from a qualitative point of view.

With the advent of psychotherapeutic techniques and measurement tools, clinicians have recently turned their attention to exploring SC through qualitative analysis. Addressing the critical voices and the accompanying emotional responses triggered by SC has underscored the need for in-depth investigations, utilizing qualitative research methods, into the content, targets, and emotional reactions associated with this phenomenon (Austin et al., 2021; Bailey et al., 2020, 2022; Bauer & Bonanno, 2001; Gilbert & Irons, 2004; Granek, 2006; Halamová et al., 2019, 2020, 2021; Koróniová et al., 2020; Redden et al., 2021; Ripoll-Núñez & Naismith, 2022; Thew et al., 2017; Toolan et al., 2019).

In psychology, qualitative research aims to deeply understand and interpret psychological phenomena using non-numerical data sources like texts, interviews, and observations (Creswell & Poth, 2017; Denzin & Lincoln, 2018; Miles et al., 2019; Patton, 2014). These methods delve into personal experiences, perspectives, and meanings, striving to capture the intricate layers and depth of psychological constructs. They are commonly employed to explore complex and nuanced aspects of psychology, offering detailed insights that enrich both theoretical understanding and practical applications in the field. The most well-known and commonly used approaches include thematic analysis (TA; Braun & Clarke, 2006, 2014), consensual qualitative research (CQR; Hill, 2012; Hill et al., 1997), and a revised version of grounded theory (Glaser & Strauss, 1967), which focuses on generating theories directly from collected data rather than starting with predefined hypotheses.

Qualitative analyses facilitate the generation of hypotheses and theories, making them useful in the initial phase of research by providing a foundation for more in-depth quantitative investigations. Additionally, they capture individual variability through interviews, observations, and discourse analysis, allowing for the detection and appreciation of personal differences in how people approach and experience the construct. Finally, they enable the exploration of new research areas. In summary, integrating qualitative analysis with quantitative approaches provides a more comprehensive and profound understanding of a psychological construct, enriching our understanding and guiding further research and interventions.

To our knowledge, there are currently no published systematic reviews that have highlighted which methodologies are used most frequently in qualitative research, and which are most suitable to study SC. There have been no conclusions drawn from the qualitative research regarding the various forms of SC and, therefore, what has been found beyond the psychometric data.

It appears important to have an overall view of how SC has been studied so far through qualitative methods. A systematic review can serve as a starting point for formulating a comprehensive view of the phenomenon.

Rationale

SC is a construct of broad scientific interest that is very relevant in clinical manifestations as a factor that exacerbates psychological suffering and hinders treatment, leading to negative psychotherapeutic outcomes. Indeed, several systematic reviews and meta-analyses have highlighted the increased interest in SC concerning its implications in psychopathology and psychotherapy. Furthermore, there are several theoretical models on SC that offer different conceptualizations (Zaccari et al., 2024), and researchers have developed various assessment tools, often based on these theoretical models, to identify SC (Rose & Rimes, 2018). While there is an extensive body of empirical studies that have used standardized questionnaires to detect SC, according to some authors, the available instruments do not always seem to capture the complexity of the construct, and there is a need to place a greater emphasis on the qualitative aspects of SC. In response, researchers have begun to investigate SC from a qualitative perspective—employing methods that go beyond a quantitative empirical approach that uses standardized tools—to identify categories of SC. To date, there are no systematic reviews on the topic, unlike the authoritative contributions on the role of SC in psychopathology and psychotherapy; therefore, this is the first systematic review of the literature that explores SC from a qualitative perspective, stimulating new lines of research on this topic.

Systematizing the available knowledge in this research field would be useful for clinicians and the scientific community to recognize how SC has been studied from a qualitative perspective. This could lead to important epistemological reflections and have significant consequences on the therapeutic process, including the development of intervention protocols.

Research question

The main purpose of this systematic review is to identify the current trends in qualitative research of SC. Specifically, we aim (a) to explore which empirical contributions have conducted qualitative analyses on SC; (b) to clarify which qualitative research methodologies have been developed to study SC; (c) to determine whether there is consensus within the scientific community regarding the employed qualitative analyses; (d) to analyze points of convergence and divergence, as well as potential limitations among various contributions that have addressed qualitative analyses of SC; and (e) to analyze future recommendations that have been provided by researchers to better understand whether they go in the same direction. Instead, we seek to provide a systematic overview of qualitative research methods used in the field of SC, to draw well-founded conclusions about the use of qualitative methodologies for SC analysis, and to stimulate reflection on the results and research trajectories, focusing on what is missing and what should be produced in empirical research.

Method

A systematic literature review was conducted following the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) guidelines (Page et al., 2021). The entire process of identification and selection of studies, based on the inclusion criteria used, is detailed in the flowchart in figure 1.

Figure 1.

Flow diagram of the search and screening process of the contributions included in this systematic review

Abbreviations: NR = not retrieved; E = excluded.

Eligibility criteria

The inclusion criteria were: (a) the study examines SC; (b) the study proposes a qualitative analysis methodology for SC; and (c) any qualitative analysis to examine the SC construct. The exclusion criteria were: (a) not original contributions (reviews, book reviews, gray literature, and comments); (b) not published in an international peer-reviewed journal. The sample type (gender, age, ethnicity, nationality, and clinical or nonclinical status) and language of the article were not exclusion criteria.

Search strategy

Before operationalizing the search syntax, we searched for terms related to SC in the thesaurus of various databases and Medical Subject Heading (MeSH) terms. We also analyzed keywords and search terms in the filters of authoritative systematic reviews and meta-analyses (Löw et al., 2020; McIntyre et al., 2018; Wakelin et al., 2022; Werner et al., 2019; Zaccari et al., 2024). We followed the same procedure for terms related to qualitative analyses (Carrera-Fernández et al., 2014). In addition, we examined studies and authoritative contributions on other psychological constructs analyzed from a qualitative perspective (Katzenmajer-Pump & Balázs, 2021). Following this endeavor, we aggregated the search terms into two concepts for all databases: “self-criticism” and “qualitative analysis” (the search string is detailed in Appendix A). The literature search was conducted on 7 scientific databases: PsycINFO, PsycARTICLES, MEDLINE, Scopus, Web of Science, PubMed, and ERIC (all years up to and including January 2024).

Selection of studies

Study selection comprised several steps: Eight reviewers (MF, FS, VC, LT, MGF, VP, SJC, and MGG) independently conducted the electronic searches of the 7 databases and screened every title and abstract to determine the eligibility of the study for inclusion. The first author (VZ) double-checked the search results. Any critical issues or ambiguities noted by the reviewers in the title and abstract were addressed through a careful reading of the full text of the article by several reviewers to determine its suitability.

The initial searches of the electronic databases carried out with a specific filter, identified a total of 2306 articles, which were then examined by each author to verify their suitability. Duplicates were removed using the Zotero software and a total of 852 records were screened resulting in the identification of 28 full-text articles for further analysis. Any disagreements were resolved through discussions among the authors. Full texts that were not available in the databases were requested from the authors via email. Finally, 16 articles were selected for inclusion in this systematic review.

The identification and screening operations are summarized in the flow diagram (figure 1).

Data extraction and synthesis

A module was developed to extract information (see table 1). For each included article, the following metadata was extracted: study (author names and publication year), country, sample, general aim, design, type of qualitative analysis, the detection procedure (i.e., how SC was detected), the coding method (i.e., how the themes and variables related to SC to be coded were detected, the rationale for extracting the variables, the variables that were selected, how many judges or raters were included in the coding method, and how inter-rater reliability was calculated), the coding process, and the identified categories of SC. The extracted data were collected by 8 authors (MF, FS, VC, LT, MGF, VP, SJC, and MGG) and rechecked by the first author (VZ). The gathered information was discussed among all authors until a consensus was reached. Furthermore, for each article, the conceptualization of SC and the theoretical models considered in the coding process were extracted.

Table 1.

Main information extracted from the 16 included studies

Study	Country	Sample				General aim	Design	Type of qualitative analysis	Detection procedure	Coding method	Coding process	Identified SC categories
		N	Type	Gender	Age (years)
				N males	Mean
Austin et al. (2021)	Netherlands	26	CO	10	48	To explore experiences of SC and self-compassion in people with cancer	Cross-sectional	Thematic analysis (Braun & Clarke, 2014)	Audio-recorded and transcribed individual and group semi-structured interviews	1. Two researchers read all transcripts, detected interesting themes, and selected text fragments. 2. The two researchers independently and inductively generalized initial codes, and subsequently sorted them into broader themes. 3. The results were discussed and reviewed to generate a final version of codes and themes. IRR:NS	- Bottom up	Four themes for SC: 1. Being harsh or strict with yourself 2. Feeling guilty or angry 3. Feeling useless or like a burden 4. Feeling ashamed and not wanting to show weakness
Bailey et al. (2020)	Slovakia	1	CP (D)		NS	To examine how SC, self-protection, and self- compassion are expressed by a client with depressive symptoms	Single-case	Consensual qualitative research (Hill, 2012)	Transcribed two-chair dialogue during an EFT video session	1. Self-critical sentences were extracted by the first author. 2. Two researchers individually developed domains, subdomains, and categorizations. 3. The results were discussed, and a consensus was reached. 4. The auditor gave feedback 5. In a final group discussion, the changes were integrated. IRR:NS	- Bottom up	Three domains for SC: 1. Behavioral aspect (2 subdomains,2 categories, 4 characteristics). 2. Cognitive aspect (2 subdomains,3 categories, 5 characteristics). 3. Emotional aspect (1 subdomain, 3 categories).
Bailey et al. (2022)	Slovakia	12	CP (D, ANX)		NS	To examine how SC, self-protection, and self-compassion are articulated by clients	Cross-sectional	CQR(Hill, 2012)	Transcribed two-chair dialogue during EFT video sessions	1. The team agreed on three domains from a previous pilot study (Bailey et al., 2020). 2. Each team member developed separate subdomains, categories, and characteristics. 3. The team came together and discussed the core ideas. 4. An auditor gave feedback. 5. The team discussed the results until they reached a consensus. 6. The auditor gave their second feedback. 7. In a final group discussion, the changes were integrated. IRR:NS	- Top down - Bottom up	Three domains for SC: 1. Behavioral aspect (2 subdomains,5 categories, 14 characteristics) 2. Cognitive aspect (4 subdomains,7 categories, 16 characteristics) 3. Emotional aspect (2 subdomains,6 categories, 6 characteristics)
Bauer & Bonanno (2001)	United States	69	NC	33%	47.4	To examine negative and positive self-evaluations of people who had experienced the death of a spouse	Longitudinal	Detailed coding system for narrative self-evaluation (Bauer, 1997)	1. Interviews about the relationship with the deceased spouse were audio-recorded and transcribed. 2. Judges segmented each transcriptinto narrative units based on their intuitive understanding of the natural boundaries of a complete thought or idea. 3. The final markers for narrative unit boundaries were determined by using the majority ratings of the judges.	1. Transcripts were coded by two raters. 2. Presence-absence codings were made for positive and negative self-evaluations in each NU. 3. Each self-evaluation was coded as either doing-based or being-based. IRR:NS	-Top down	NS
Gilbert & Irons (2004)	United Kingdom	8	CP (D)	2	NS	To explore triggers and forms of SC in patients with depression.	Longitudinal	NS	Participants keptdiaries to provideresponses to thefollowing: Q1. what situationsor events triggeredtheir self-criticalthinking. Q2. how thesesituations madethem think aboutthem-selves. Q3. how thesethoughts madethem feel.	NS IRR:NS	- Top down	Themes for each question: Q1. Trigger situations/events (e.g., family, work) Q2. Thoughts about themselves (e.g., inadequate, incompetent) Q3. Feelings associated with the thoughts (e.g., unhappy, anxious)
Granek (2006)	Canada	6	CP (D)	1	NS(range:25-30)	To explore experiences of depression using a hermeneutical approach	Cross-sectional	Methodical hermeneutical approach (Rennie, 2000), a revised version of grounded theory (Glaser & Strauss, 1967)	1. Open-ended interviews about experiences of depression 2. The author transcribed all the interviews herself.	1. Transcriptions were broken into initial meaning units 2. The interpreted meaning of each unit was represented by one or more categories. 3. The meaning unit was assigned to the category or categories. 4. The main categories were interpreted to include subcategories. 5. Once all the data had been assigned to categories, the major themes stood out clearly. 6. Themes were tested for validity and usefulness as representing the phenomena under inquiry. IRR:NS	-Bottom up	NS
Halamova etal. (2019)	Slovakia	10	NC	2	21.7	To categorize the participants' descriptions of their self-critical imagery and to identify differences between high and low self-critical participants	Cross-sectional	CQR (Hill etal., 1997)	Open-ended questions about the content of their self-critical parts during the imagery	1. Two assessors each received the raw data. 2. Each member of the team categorized the data (separately for the high and low SC). 4. A consensus on the domains, subdomains, categories, subcategories, and characteristics was reached. 5. The auditor checked the first draft of the data and provided feedback. 6. After a group discussion, the auditor's comments were implemented in the final version of categorization. IRR:NS	-Bottom up	Six major domains for SC: 1. Emotions 2. Appearance 3. Voice 4. Cognitions 5. Needs 6. Behavior Other: -23 domains (NS) -26 subdomains (NS)
Halamova etal. (2020)	Slovakia	151	NC	37	22.2	To examine free associations for criticism and SC	Cross-sectional	CQR (Hill, 2012)	Free association task	1. Four researchers and one auditor 2. Researchers noted their expectations. 3. Each member obtained the data and individually created their own categories and domains. 4. A consensus was sought through discussions as a team. 5. The auditor gave feedback. 6. A consensus was reached by the core team and approved by the auditor. IRR:NS	-Bottom up	Four domains for SC: 1. Behavioral aspect (3 subdomains, 6 categories). 2. Preconditions (2 subdomains, 4 categories). 3. Cognitive aspect (1 subdomain, 2 categories). 4. Emotional aspect (1 subdomain, 2 categories).
Halamova etal. (2021)	Slovakia	20	NC	5	27.7	To analyze and categorize participants' subjective statements when self-criticizing	Cross-sectional	CQR (Hill etal., 1997)	Video-recorded short self-critical dialogues using the two-chair technique	1. Three coauthors selected from the whole sample the most expressive and self-critical participants. 2. After transcription, they created domains, subdomains, and categories. 3. The auditor provided feedback. 4. The proposed changes were implemented to produce the final version of the categorization. IRR:NS	-Bottom up	Three domains for SC: 1. Emotional aspects of self-critic (1 subdomain, 2 categories). 2. Cognitive aspects of self-critic (3 subdomains, 6 categories) 3. Behavioral aspects of self-critic (2 subdomains, 5 categories)
Koroniova etal. (2020)	Slovakia	27	NC	2	22	To identify and categorize participants' description of inner critical, protective, and compassionate parts of the self	Longitudinal	CQR (Hill et al., 1997)	1. Guided imagery of the inner critical self 2. Open-ended questions on the imagery	1. Two assessors 2. Core team members wrote their expectations about what the participants would answer prior to working with the data. 2. Each assessor independently created an ensemble of the expected categories. 3. The final categories were agreed upon after a group discussion, and the raw data were allocated to the designated categories (separate categories for high self-critics and low self-critics). 4. The auditor checked the categorization, and her feedback was acted on following team discussion and unanimity. IRR:NS	-Top-down	Six major domains for SC: 1. Emotions 2. Appearance 3. Voice 4. Cognitions 5. Needs 6. Behaviour 30 domains (NS) 14 subdomains (NS)
Redden et al. (2021)	Canada	103	NC	59	24	To identify thoughts, compassionate and uncompassionate, toward the self	Cross-sectional	Completecodingapproach(thematicanalysis; Braun & Clarke, 2013)	Twenty-four illustrated scenarios created to evoke compassionate or uncompassionate responses	1. Five assessors met weekly to discuss the responses and to look for patterns. 2. All responses were sorted into categories, based on their similar functions. 3. The categories were re-named to capture the coding team's interpretations according to the self-compassion theory. IRR: Specified	-Bottom up -Top-down	Two categories for self-judgment: 1. Unequivocal, broadly self-critical statements 2. Annoyance or critical feelings toward oneself
Ripoll-Nunez & Naismith (2022)	Colombia	18	CP (D, ANX, PTSD)		37.5	To explore self-perceptions of women who had experienced intimate partner violence or gender-based violence	Cross-sectional	Thematic analysis (Braun & Clarke, 2006)	Audio recorded and transcribed CFT sessions (questions about SC)	1. Session transcripts were analyzed using the NVivo 12.1 software. 2. Two assessors provisionally defined the themes based on the existing literature. 3. Three students each coded one third of the data, and then coded two sessions of each other coders' data. 4. The whole team revised the themes, and the two professors produced the final interpretation. 5. To increase validity of results, the participants were asked to check the results. IRR:NS	-Bottom-up -Top-down	The SC theme included: - Perfectionistic standards - Perceived incompetence at coping with difficulties associated with the violence - Perceived failures in parenting
Thew et al. (2017)	United Kingdom	78	NC CP (D, ED)	NS	NS	To explore the phenomenology of SC, and its relationship with rumination and perfectionism	Cross-sectional	Framework analysis (Gale etal., 2013)	Audio-recorded and transcribed semi-structured interviews	1. Transcripts were reviewed by one judge to create categories. 2. A code set was developed for each category through the raw data and was used to construct a summary matrix. 3. The records were coded by two independent judges. IRR: specified	-Top-down -Bottom-up	NS
Toolan et al. (2019)	Ireland	14	CP (GAD)	2	NS(range:23-56)	To explore the relationship between self-worry messages and self-critical messages within worry dialogues in patients with GAD	Multiple case	Descriptive-interpretative qualitative research (Elliott & Timulak, 2005; Timulak & Elliott, 2019)	Audio-/video-recorded and transcribed worry dialogues (Timulak etal., 2017) during EFT sessions	1. Two independent judges highlighted on the transcripts the SC messages relatedto self-worry statements, by comparing them to the dialogues of five additional participants (not included in this study). 2. Next, they developed a table with the self-worry and SC messages for each participant (90 pairs of messages). 3. One judge added a summary statement for each linked self-worry and SC message. 4. The summary statements were grouped into six clusters, first separately for each participant and then througha cross-case analysis of all 14 participants IRR:NS	-Bottom-up	Six clusters of paired worry-critic messages: 1. (W) 1 need to be prepared for future events because(C) I'm weak and a failure. 2. (W) 1 need to stop worrying and (C) I'm flawed for being a worrier. 3. (W) People will negatively judge me if 1 engage with them because (C) I'm not good enough. 4. (W) If 1 don't worry, there will be negative consequences and (C) 1 will be responsible. 5. (W) 1 worry/ ruminate that 1 cause/d some damage because (C) I'm incompetent. 6. (W) 1 must always be prepared, or else others will take advantage of me because (C) I'm a walkover
Vanheule etal. (2008)	Belgium	2	CP (D)		NS	To assess whether the self-critical patient would be concerned mainly with self-critical autonomous themes, rather than with dependent ones	Cross-sectional	Clinical content analysis (NS)	Unstructured clinical interview	Two judges were asked to verify whether nine themes (five self-critical or autonomous and four dependent or sociotropic) were present in the transcript. IRR:NS	Top-down	SC themes: 1. Establishing, protecting, and maintaining a viable self-concept in relationship 2. Concern about autonomy and control in a relationship 3. Other-directed aggression 4. Self-directed aggression: self-worth and failure 5. Self-directed aggression: self-mutilation and suicide attempts
Whelton & Henkel-man (2002)	Canada	22	NC	NS (22.7%)	23.8	To examine the types of SC in individuals with or without vulnerability to depression	Cross-sectional	Phenome-nological analysis (Colaizzi, 1978)	Video-recorded self-critical dialogues	1. Video-recordings were transcribed. 2. Twenty-eight first-order themes were extracted. 3. The extracted themes were grouped into eight more abstract categories. IRR:NS	Bottom-up	Eight categories for SC: 1. Demands and orders 2. Exhorting and preaching 3. Explanations and excuses 4. Inducing fear and anxiety 5. Concern, protection, and support 6. Description 7. Explore/puzzle/ existential 8. Self-attack and condemnation

Abbreviations: ANX = anxiety; C = clinical sample; CFT = compassion-focused therapy; CP = Clinical sample with psychopathological disorders; CO = clinical sample with organic disorders; D = depression; ED = eating disorders; EFT = emotion focused therapy; GAD = generalized anxiety disorder; IRR = inter-rater reliability; NC = non-clinical sample; NS = not specified; PTSD = post-traumatic stress disorder; SC = self-criticism.

Critical appraisal: assessment of methodological quality of qualitative studies

The Critical Appraisal of the qualitative studies that met the eligibility criteria was carried out with the Critical Appraisal Skills Program (CASP; 2024). This tool is aimed at the critical analysis of qualitative studies and aims to evaluate the rigor, credibility, and relevance of the results presented (Noyes et al., 2018; Hannes & Bennett, 2017; Hannes & Macaitis, 2012). It is endorsed by Cochrane and the World Health Organization for use in qualitative evidence synthesis (Hannes, 2011). It helps in considering, respectively, the appropriateness and scientific punctuality of the research methodology, the clarity of the conclusions, and the significance of the study with respect to the initial aims of the bibliographic investigation. CASP Checklist consists of ten questions, each of which focuses on a different methodological aspect of a qualitative study, divided into 3 sections (see table 2): Section A (Are the results valid?), B (What are the results?), and C (Will the results help locally?). In this process, each evaluated question could be assigned the following qualitative scores: “Yes” (1 point), “Can’t tell” (0.5 points), or “No” (0 points). When a paper received a “Yes” in two-thirds of the sections of the CASP, it was rated as “High”; “Moderate” quality was considered when the score was between 4 and 6 “Yes” responses; and finally, if more than two-thirds of the replies were “No,” the paper was recorded as “Low” quality, as used in previous qualitative review studies (Ricoy-Cano et al., 2020). One of the authors (VZ) performed the assessment of the methodological quality of the qualitative studies included, and the assessment procedure was verified by two reviewers (MF and FS). The discrepancies that emerged in the evaluation and scoring of the studies were resolved by consensus among the reviewers.

Table 2.

CASP (Critical Appraisal Skills Programme) qualitative checklist scores for the methodological quality assessment of the qualitative studies included

Study	Q1	Q2	Q3	Q4	Q5	Q6	Q7	Q8	Q9	Q10	Score	Quality rating
Austin et al. (2021)	Y	Y	Y	Y	Y	Y	Y	Y	Y	Y	10	High
Bailey et al. (2020)	Y	Y	Y	CT	Y	N	Y	Y	Y	Y	8.5	High
Bailey et al. (2022)	Y	Y	Y	Y	Y	N	Y	Y	Y	Y	9	High
Bauer & Bonanno (2001)	Y	Y	N	Y	Y	N	CT	Y	Y	CT	7	High
Gilbert & Irons (2004)	Y	Y	Y	Y	Y	N	CT	N	CT	Y	7	High
Granek (2006)	Y	Y	Y	Y	Y	N	Y	N	N	N	6	Moderate
Halamová et al. (2019)	Y	Y	Y	CT	Y	N	Y	Y	Y	Y	8.5	High
Halamová et al. (2020)	Y	Y	Y	CT	Y	N	CT	Y	Y	Y	8	High
Halamová et al. (2021)	Y	Y	Y	Y	Y	N	Y	Y	Y	Y	9	High
Koróniová et al. (2020)	Y	Y	Y	Y	Y	N	Y	Y	Y	Y	9	High
Redden et al. (2021)	Y	Y	Y	Y	Y	N	Y	Y	Y	Y	9	High
Ripoll-Núñez & Naismith (2022)	Y	Y	Y	Y	Y	N	Y	Y	Y	Y	9	High
Thew et al. (2017)	Y	Y	Y	Y	Y	N	Y	CT	N	N	6.5	Moderate
Toolan et al. (2019)	Y	Y	CT	Y	Y	N	Y	Y	Y	Y	8.5	High
Vanheule et al. (2008)	N	Y	N	Y	CT	N	CT	N	Y	N	4	Moderate
Whelton & Henkelman (2002)	Y	Y	Y	Y	Y	N	N	N	Y	CT	6.5	Moderate

Abbreviations: Y = Yes; N = No; CT = Can't tell; Question 1 = Clear statement of aim; Item 2 = Appropriate qualitative methodology; Item 3 = Appropriate research design; Item 4 = Sampling; Item 5 = Data collection; Item 6 = Researcher reflexivity; Item 7 = Ethical consideration; Item 8 = Appropriate data analysis; Item 9 = Clear statement of findings; Item 10 = Research value. Section A (validity of the results); Section B (results); Section C (value of the research).

Results

Study characteristics

A total of 16 studies were included in the final review. The articles were published between 2001 and 2022. The sample sizes ranged from 1 participant (Bailey et al., 2020; Vanheule et al., 2008) to 151 participants (Halamová et al., 2020). The studies exhibited heterogeneity in terms of the populations under investigation. Out of these 16 studies, 9 recruited a clinical sample, while 7 recruited a non-clinical sample. The primary diagnoses considered were depression (Bailey et al., 2020, 2022; Gilbert & Irons, 2004; Granek, 2006; Ripoll-Núñez & Naismith, 2022; Thew et al., 2017; Vanheule et al., 2008), anxiety disorders (Bailey et al., 2022; Ripoll-Núñez & Naismith, 2022; Toolan et al., 2019), eating disorders (Thew et al., 2017), and post-traumatic stress disorder (Ripoll-Núñez & Naismith, 2022). One study considered patients with cancer (Austin et al., 2021). All studies had a predominantly female sample. Four studies had an entirely female sample (Bailey et al., 2020, 2022; Ripoll-Núñez & Naismith, 2022; Vanheule et al., 2008). Gilbert and Irons (2004) considered a sample including 9 participants, of whom two were male. However, the authors only performed the analyses on 8 participants; they did not indicate the gender of the excluded participant. The average age of the participants ranged from 18 to 56 years (Austin et al., 2021; Bauer & Bonanno, 2001; Granek, 2006; Halamová et al., 2019, 2020, 2021; Koróniová et al., 2020; Redden et al., 2021; Ripoll-Núñez & Naismith, 2022; Thew et al., 2017; Toolan et al., 2019; Whelton & Henkelman, 2002). Four studies did not include the average age of the participants (Bailey et al., 2020, 2022; Gilbert & Irons, 2004; Vanheule et al., 2008). See table 1 for an overview of all sample characteristics.

Regarding the research design, 11 studies used a cross-sectional design (Austin et al., 2021; Bailey et al., 2022; Granek, 2006; Halamová et al., 2019, 2020, 2021; Redden et al., 2021; Ripoll-Núñez & Naismith, 2022; Thew et al., 2017; Vanheule et al., 2008; Whelton & Henkelman, 2002), 3 studies employed a longitudinal design (Bauer & Bonanno, 2001; Gilbert & Irons, 2004; Koróniová et al., 2020), 1 study used a single-case design (Bailey et al., 2020), and 1 study used a multiple-case design (Toolan et al., 2019). The studies were conducted in the Netherlands (Austin et al., 2021), Slovakia (Bailey et al., 2020, 2022; Halamová et al., 2019, 2020, 2021; Koróniová et al., 2020), the United States (Bauer & Bonanno, 2001), the United Kingdom (Gilbert & Irons, 2004; Thew et al., 2017), Canada (Granek, 2006; Redden et al., 2021; Whelton & Henkelman, 2002), Colombia (Ripoll-Núñez & Naismith, 2022), Ireland (Toolan et al., 2019), and Belgium (Vanheule et al., 2008). This geographical distribution highlights the global interest in exploring SC within different cultural contexts.

Methodological quality assessment of studies included

The methodological quality of the articles included in the systematic review emerged as “High” (mean quality = 7.91). Twelve articles exhibited “High” quality (Austin et al., 2021; Bailey et al., 2020, 2022; Bauer & Bonanno, 2001; Gilbert & Irons, 2004; Halamová et al., 2019, 2020, 2021; Koróniová et al., 2020; Redden et al., 2021; Ripoll-Núñez & Naismith, 2022; Toolan et al., 2019), which represents 75% of the total studies. Four studies presented “Moderate” quality (Granek, 2006; Thew et al., 2017; Vanheule et al., 2008; Whelton & Henkelman, 2002), which represents 25% of the total; no study showed “Low” quality.

Regarding Section A (validity of the results), discrepant scores were shown in Questions 3, 4, 5 and 6. In Question 3, which refers to the adequacy of the research design to its objectives, 3 studies (Bauer & Bonanno, 2001; Toolan et al., 2019; Vanheule et al., 2008) were inadequate concerning this criterion. In Question 4, which refers to the recruitment strategy in line with the aims of the research, 3 studies (Bailey et al., 2020; Halamová et al., 2019, 2020) did not meet this criterion. For Question 5, about how appropriate the sample recruitment strategies are, only 1 study (Vanheule et al., 2008) doesn't give enough information to assess this aspect, while all the other studies meet the criterion.

Only 1 study (Austin et al., 2021) adequately met the requirements of Question 6, which deals with considerations in the relationship between researchers and participants.

Regarding Section B, other discrepant results emerge in Question 7, 5 studies (Bauer & Bonanno, 2001; Gilbert & Irons, 2004; Halamová et al., 2020; Vanheule et al., 2008; Whelton & Henkelman, 2002) did not take ethical issues into account. In Question 8, 5 studies (Gilbert & Irons, 2004; Granek, 2006; Thew et al., 2017; Vanheule et al., 2008; Whelton & Henkelman, 2002) did not conduct appropriate data analysis. In Question 9, 3 studies (Gilbert & Irons, 2004; Granek, 2006; Thew et al., 2017) did not report a clear statement of findings. Finally, in Section C (value of the research), in Question 10, 5 studies (Bauer & Bonanno, 2001; Granek, 2006; Thew et al., 2017; Vanheule et al., 2008; Whelton & Henkelman, 2002) did not meet this criterion. A quality assessment of the studies included is presented in table 2.

Study aims

All included studies provided a clear statement of their aims and objectives. Regardless of the variations in the specific aims, all articles explored the role and phenomenology of SC across different populations. Ten studies specifically intended to explore forms and expressions of SC (Austin et al., 2021; Bailey et al., 2020, 2022; Gilbert & Irons, 2004; Halamová et al., 2019, 2020, 2021; Koróniová et al., 2020; Thew et al., 2017; Whelton & Henkelman, 2002). In contrast, the remaining 6 studies pursued fewer specific objectives. Bauer and Bonanno (2001) tested the hypothesis that a limited number of negative self-evaluations predicts lower levels of grief over time, specifically referring to emotional pain associated with the loss of a loved one. Granek (2006) inquired into the nature of depression among young adults. Redden et al. (2021) aimed to identify uncompassionate thoughts toward the self. Ripoll-Núñez & Naismith (2022) evaluated the self-perceptions of women who had experienced partner or gender-based violence. Toolan et al. (2019) intended to explore the relationship between self-worry messages and self-critical messages within worry dialogues. Finally, Vanheule et al. (2008) aimed to investigate whether patients with subtypes of depression exhibiting a pronounced inclination toward SC tend to focus more on critical aspects of themselves rather than seeking emotional support from others.

Qualitative analysis and methods used

Data Collection

The included studies employed different methods to gather qualitative data on SC. In 3 studies, data were collected using semi-structured interviews (Austin et al., 2021; Bauer & Bonanno, 2001; Thew et al., 2017); in addition, Granek (2006) used open-ended interviews, and Vanheule et al. (2008) used unstructured clinical interviews. Two articles employed open-ended questions about an imagery task, during the task itself (Halamová et al., 2019) or afterwards (Koróniová et al., 2020). Three studies employed two-chair dialogues (Bailey et al., 2020, 2022; Halamová et al., 2021), while Toolan et al. (2019) used worry dialogues. The remaining studies collected data in different ways, through diaries (Gilbert & Irons, 2004), written free association tasks (Halamová et al., 2020), questions about SC during CFT sessions (Ripoll-Núñez & Naismith, 2022), automatic thoughts evoked by illustrated scenarios (Redden et al., 2021), and self-critic dialogues (Whelton & Henkelman, 2002).

Coding procedures

The included studies employed different approaches to analyze qualitative data. In top-down coding, also known as deductive coding, researchers interpret data based on predetermined categories derived from the existing literature or theoretical frameworks. This process is often used to test specific hypotheses or to validate existing theories. On the other hand, in bottom-up coding, also known as inductive coding, themes and categories emerge from the raw data, without any preexisting concept. This kind of process is appropriate when researchers want to explore a new phenomenon or construct. Among the included studies, 4 employed a top-down coding approach (Bauer & Bonanno, 2001; Gilbert & Irons, 2004; Koróniová et al., 2020; Vanheule et al., 2008), 8 used a bottom-up coding approach (Austin et al., 2021; Bailey et al., 2020; Granek, 2006; Halamová et al., 2019, 2020, 2021; Toolan et al., 2019; Whelton & Henkelman, 2002), and 4 used a combination of both top-down and bottom-up coding approaches (Bailey et al., 2022; Redden et al., 2021; Ripoll-Núñez & Naismith, 2022; Thew et al., 2017).

Qualitative analysis

In terms of the type of qualitative analysis conducted, the most widely used methods were CQR (Hill, 2012; Hill et al., 1997) and TA (Braun & Clarke, 2006, 2014; Clarke & Braun, 2013). CQR and TA are primarily bottom-up coding methods, but they can also incorporate elements of a top-down approach. Four studies (Bailey et al., 2020; Halamová et al., 2019, 2020, 2021) used CQR as a pure bottom-up process. In these studies, multiple assessors independently generated codes based on the raw data. Then, an auditor checked the categorization and there was discussion among the team members, until the final categories were determined through unanimous agreement. Bailey et al. (2022) employed CQR as a bottom-up approach but involving top-down elements, as the researchers initially agreed on 3 domains from a previous study. Koróniová et al. (2020) used CQR as a mostly top-down coding process: the team members defined and agreed on the categories before working with the data. Austin et al. (2021) used TA as a purely bottom-up process, while Redden et al. (2021) and Ripoll-Núñez & Naismith (2022) adapted TA to include both top-down and bottom-up elements. In the former study, after familiarizing themselves with the data, two assessors provisionally defined the themes by drawing on existing literature. Then, the entire team revised the themes until reaching a final categorization (Ripoll-Núñez & Naismith, 2022). In the latter study, the team identified, and gradually adjusted, content codes based on the raw data. However, they introduced a top-down approach in a subsequent stage, as the team renamed the response categories to align them with the theoretical framework. In the second part of their study, the coders received training on theory and content codes before coding responses individually, suggesting a top-down coding process (Redden et al., 2021). Granek (2006) used a methodical hermeneutical approach (Rennie, 2000), a revised version of grounded theory (Glaser & Strauss, 1967). This is a bottom-up approach, as data are initially divided into smaller units, which are then grouped to identify categories and themes. Toolan et al. (2019) analyzed their data by following the guidelines for generic descriptive‐interpretative qualitative research, as described by Elliott and Timulak (2005). They appeared to follow a bottom-up coding process, as the final clusters of linked self-worry and SC messages emerged from the data (transcribed worry dialogues). Whelton and Henkelman (2002) stated that they used a phenomenological method of qualitative analysis to ground the development of the category system empirically (Colaizzi, 1978). They followed a bottom-up approach, extracting sentences from transcripts to formulate first-order themes, then grouped them into second-order clusters. Among the studies that used a top-down coding process, Bauer and Bonanno (2001) coded narrative transcripts for the presence or absence of both doing-based and being-based self-evaluations, thus imposing predetermined categories on the data. Gilbert and Irons (2004) also used a top-down coding process; their SC categories were represented by the 3 questions they asked the participants. Vanheule et al. (2008) carried out a clinical content analysis, where transcripts of unstructured clinical interviews were categorized into predetermined themes. Finally, Thew et al. (2017) followed a procedure based on framework analysis (Gale et al., 2013), combining elements of both top-down and bottom-up approaches. They applied predetermined categories to the data (a top-down element) while developing codes for each category iteratively based on the raw data (a bottom-up element).

Of note, only 4 studies calculated inter-rater reliability (Bauer & Bonanno, 2001; Redden et al., 2021; Ripoll-Núñez & Naismith, 2022; Thew et al., 2017). Bauer and Bonanno (2001) measured reliability by using the kappa statistic, with an overall inter-rater reliability of .76. Redden et al. (2021) evaluated reliability using two-way mixed effects intra-class correlations, finding that most correlations were excellent (> .75), with a few being good (between .60 and .74), fair (between .40 and .59), or poor (< .39). Categories with poor and fair reliability were infrequent. Thew et al. (2017) ensured reliability by having two independent raters review and code audio-recordings, resulting in good consistency of coding choices between the rater pairs (85%–96%). Ripoll-Núñez & Naismith (2022) had 3 students each code one third of the data, and then they coded two sessions of each other coder’s data. However, they did not report any numeric value for inter-rater reliability.

Main extracted or coded categories of self-criticism

The included studies varied in their categorizationof the phenomenological aspects of SC. Several studies identified clusters or categories within the participants’ transcripts, either through analysis or by predefining them, and then assigned the participants’ responses to these predetermined categories. Five articles (Bailey et al., 2020, 2022; Halamová et al., 2019, 2020, 2021) examined the various ways in which SC is evident, identifying 3 specific domains and their subdomains: cognitive, emotional, and behavioral. Halamová et al. (2020) also identified the preconditions domain, which encompasses factors predisposing individuals to engage in SC and includes two subdomains: internal (e.g., self-worth) and external (e.g., rival, enemy) sources. In their previous study, Halamová et al. (2019) identified 6 major domains of SC, including the categories of appearance, voice, and needs.

Koróniová et al. (2020) utilized the same primary areas of focus as Halamová et al. (2019). However, the authors established predetermined categories to classify the responses of participants with high and low SC. These responses were collected both before and after an EFT intervention. Other authors similarly employed a top-down approach. Bauer and Bonanno (2001) utilized interviews to categorize self-evaluation valence (positive/negative) and self-evaluation context levels (doing-based/being-based). Gilbert and Irons (2004) conducted interviews to examine the factors that led to self-critical thinking (situations or events that triggered such thinking), the thoughts that had self-critical content (how these situations influenced their thoughts about themselves), and the associated emotions (how these thoughts affected their feelings). Vanheule et al. (2008) opted to contrast the occurrence of 5 self-critical themes with other dependent themes in an interview transcript. The main SC themes explored in the study were: the establishment, protection, and maintenance of a valid self-concept within the relationship; the concern for autonomy and control within the relationship; aggression directed toward others; aggression directed toward oneself, specifically related to self-esteem and failure; and self-directed aggression in the form of self-mutilation and suicide attempts.

Three articles focused on the cognitive content of SC through bottom-up analyses (Austin et al., 2021; Toolan et al., 2019; Whelton & Henkelman, 2002). Austin et al. (2021) found 4 main themes of SC: being harsh or strict with yourself, including setting high and pressing standards, repressing felt emotions, minimizing one’s needs, not accepting choices and mistakes, and mistreating the body; feeling guilty or angry, including not being there for others, clashing with others, and enacting unhealthy behaviors; feeling useless or like a burden for not being able to fulfill social roles; and feeling ashamed of the visible signs of illness and not wanting to show weakness to avoid pity. Whelton and Henkelman (2002) grouped the different types of SC into 8 clusters: requests and orders; exhortation and preaching; explanation and apology; fear and anxiety induction; worry, protection, and support; description; exploration/puzzle/existential; and self-attack and condemnation. Toolan et al. (2019) identified 6 clusters of SC in relation to worry. Cluster 1 summarizes future worries about negative events as a result of inability to cope with such events. Cluster 2 groups worry about feeling defective as a result of worry. Cluster 3 includes concerns about being judged negatively for not feeling good enough. Cluster 4 comprises the fear of liability for not caring in case a negative event occurs. Cluster 5 includes concerns about causing harm through incompetence. Cluster 6 contains the concern of always having to be ready to prevent others from taking advantage and making a judgment of weakness.

Other studies used both bottom-up and top-down approaches, tracking thoughts and emotions (Redden et al., 2021) and themes of thoughts and emotions related to SC (Ripoll-Núñez & Naismith, 2022; Thew et al., 2017). Similarly to Austin et al. (2021), Ripoll-Núñez & Naismith (2022) analyzed the most frequent thought content of SC; they found that SC is related to the claim of pursuing perfectionistic standards, the ability to make decisions appropriate to violence, and the perception of having failed as a parent. Thew et al. (2017) produced an overview of the presence and impact of SC, exploring several aspects (presence, frequency, duration, external and internal triggers, and effects of SC). They found that SC is present several times during the day and has a considerable duration, which varies contextually with several factors. The most frequently reported external triggers are those involving a mistake or failure at something or being in social situations or with friends, while internal triggers mainly involve a low mood. SC, in most cases, is an automatic process over which there is little control, and the main negative effect of SC concerns a low mood. The main ongoing change concerns the desire to reduce self-critical thoughts. Two main positive effects emerged for SC: a beneficial effect on motivation and momentum and the potential for self-improvement. Redden et al. (2021) identified 2 categories of self-judgment: unequivocal and largely self-critical statements and feelings of annoyance or criticism toward oneself. Finally, Granek (2006) did not elaborate on categories of SC but found that SC toward oneself often manifests itself in response to an external gaze.

Discussion

The main purpose of this study was to identify the current trends in qualitative research within the field of SC. To achieve this goal, we conducted a systematic review of the studies in which the authors performed a qualitative analysis of the construct. After identifying the included studies, we extracted the relevant information and analyzed it to draw conclusions about the current state of the art in the field. We were interested in identifying which coding methods have been employed to analyze SC from a qualitative point of view and the obtained results.

Among the 16 studies included in our review, 12 were assessed as having high methodological quality, while the remaining 4 were of moderate quality. It is a positive outcome that no low-quality studies were included in the review.

Although many studies demonstrated high methodological quality, their research objectives were often not exclusively focused on SC but also addressed related phenomena or constructs.

In particular, regarding the aims, only 11 studies (Austin et al., 2021; Bailey et al., 2020, 2022; Gilbert & Irons, 2004; Halamová et al., 2019, 2020, 2021; Koróniová et al., 2020; Redden et al., 2021; Thew et al., 2017; Whelton & Henkelman, 2002) specifically aimed to explore forms and expressions of SC in various populations. This highlights how the construct is not always studied in a "pure" form, but rather in relation to various phenomena.

The variation in the objectives among the included studies poses challenges in comparing findings and drawing comprehensive conclusions about the phenomenology of SC. Studies that did not prioritize SC as their primary focus may have overlooked relevant aspects not directly related to their specific objectives.

Regarding the studies primarily focused on exploring SC, the results obtained from the qualitative analysis reflect a lack of dialogue and interaction among different perspectives on SC, as noted by Zaccari et al. (2024). Indeed, our findings highlight the lack of agreement among the included studies regarding the identified categories, which implies a fragmentation in the conceptualization of SC, probably due to the heterogeneous aims of the studies.

The included studies varied in the number and nature of the categories or dimensions of SC the authors identified. Some studies (e.g., Bailey et al., 2020 2022; Halamová et al., 2019, 2020, 2021) identified and categorized subdomains within the emotional, cognitive, and behavioral domain of SC, while other studies (Austin et al., 2021; Whelton & Henkelman, 2002) identified other themes or clusters of SC. However, even among studies that identified the same components, we noted a lack of coherence. For example, although the cognitive component of SC always refers to self-evaluations, the included studies identified different subdomains. The differences in the identified subdomains across studies suggest a need for further research to clarify the cognitive domain of SC.

Regarding the emotional domain, different authors have identified various emotions associated with SC, while others have entirely overlooked this aspect. This lack of consistency in defining and categorizing emotional subdomains of SC within the literature highlights the need for further research to establish a more unified framework. The gap in the investigation of the negative emotional reaction to SC represents a significant limitation. Gilbert (2022) also underlined the relevance of this component in his theoretical conceptualization of SC. He argued that SC becomes psychopathological when it involves a hostile attack on the self, eliciting shame. Furthermore, therapeutic interventions specifically target the emotional reaction, rather than the self-critical process itself. For example, in chairwork, the individual is asked to recall childhood situations in which they experienced the same emotional state as that experienced in response to SC, to trace back to their historical vulnerability. This implies that exploring in more detail the emotional reactions derived from SC is essential to gain a more comprehensive understanding of SC and its role in different psychopathological contexts, thus allowing for well-tailored interventions.

We also noted incoherence among the subdomains within the behavioral dimension. Different authors focus either on the behavior that is the source of criticism (Bailey et al., 2020; Bailey et al., 2022) or on the different strategies for managing criticism (Halamová et al., 2021; Halamová et al., 2020).

The results in this dimension show that the emotional reactions and behavioral implications that follow criticism are not well-defined, remaining vague and unsystematized. The emotional and behavioral dimensions appear very important from a clinical perspective, as it has been widely reported in the literature (Blatt et al., 1976; Blatt & Luyten, 2009; Shahar, 2015; Gilbert et al., 2004; Gilbert & Simos, 2022) that different emotional reactions to criticism lead to different coping behaviors. Several theoretical models on SC (Blatt et al., 1976; Blatt & Luyten, 2009; Shahar, 2015; Gilbert et al., 2004), as well as the therapeutic techniques developed to alleviate levels of SC (i.e., Compassionate Mind Training, Compassion-Focused Therapy, Schema Therapy, Chairwork, Emotion-Focused Therapy; Young et al., 2003; Gilbert & Procter, 2006; Greenberg, 2011; Gilbert, 2014; Arntz, 2018; Pugh, 2019; Gilbert & Simos, 2022), have focused on emotional reactions related to SC (e.g., guilt, shame, sadness, self-contempt, etc.) or implications for action (coping strategies such as perfectionism or hyper compensatory behaviors aimed at success or avoiding mistakes, avoidance strategies, submission, and self-punitive or self-harming behaviors), which are targets for intervention and can be found in various clinical manifestations. An interesting line of research highlights how, for example, the reactions of guilt and shame related to SC imply different behaviors aimed at restoring both the moral self and the ideal self (Miceli & Castelfranchi, 2018). Miceli and Castelfranchi (2018) argue that guilt and shame, both defined as self-critical emotions, imply different motivational consequences: criticism that elicits guilt would motivate reparative or self-punishing behaviors to restore the moral self, while shame would motivate withdrawal or greater efforts to build the desired identity and thus the ideal self. No study specifically highlights how the analysis of the overall critical judgment of the self can result in different behaviors or emotional reactions.

Furthermore, a clear framework for analyzing the phenomenology of SC cannot be established due to the existence of numerous subcategories within the same domains that have been identified in the various studies. From a clinical perspective, the lack of consensus regarding the components of SC does not allow one to investigate the specific involvement of SC in different mental disorders. The lack of clarity in defining the construct might lead to unfocused treatment strategies and, consequently, negative outcomes (Zaccari et al., 2024). It is important to initiate an epistemological reflection on how self-criticism may imply different action tendencies depending on the emotional state experienced (Zaccari et al., 2024). The plurality of approaches and study results underlines the complexity of SC and the need for comprehensive research to fully understand the dimensions and implications of this construct in psychopathology.

Of note, among the studies included in the present review, none conducted a qualitative analysis of SC based on existing theoretical models, thus explaining the heterogeneity of the categories and subcategories the authors identified.

Another essential aspect to consider when interpreting the results of this review is the potential overlap between SC and other psychological constructs, such as worry and perfectionism. As noted by Toolan et al. (2019), worry may co-occur with SC, as constant worrying can lead to negative thoughts about oneself (e.g., seeing oneself as a weak person), and self-critical people may tend to worry frequently about their actions or potential negative outcomes. Concerning the relationship between perfectionism and SC, Bailey et al. (2020, 2022) identified perfectionism as a category within the cognitive domain of SC, aligning with Shahar’s and Blatt’s perspective that SC involves a demand for perfection and leads to hostile reactions when these standards are not met. However, within the framework of perfectionism, SC has been studied as a component of this construct (Shafran et al., 2002), highlighting the potential overlap between SC and perfectionism. To sum up, it is crucial to deepen our understanding of the association between SC and other constructs, such as worry and perfectionism, to acknowledge their potential intersection and to differentiate them accurately.

Implications for Qualitative Research

From a methodological point of view, the included studies employed different methods. The most widely qualitative methods employed to investigate SC are CQR (Bailey et al., 2020, 2022; Halamová et al., 2019, 2020, 2021; Koróniová et al., 2020) and TA (Austin et al., 2021; Redden et al., 2021; Ripoll-Núñez & Naismith, 2022). These findings are consistent with the analysis of the literature carried out by Carrera-Fernández et al. (2014), who provided a general overview of the state of qualitative research in psychology. They identified both CQR and TA as the “most current methods,” that is, the most widely used methods in articles that have been published since 2000. Furthermore, they found that CQR was the second most used method in psychology publications. This result suggests the suitability of CQR for qualitative investigations in psychology. The suitability of these methods is also supported by several studies where CQR and TA have been employed to investigate various psychological constructs qualitatively, such as perfectionism (Binder et al., 2023; Merrell et al., 2011; Nealis & Mackinnon, 2018).

Although CQR and TA are widely used qualitative methodologies in psychological and social research, they have different strengths and weaknesses that deserve attention. CQR promotes a collaborative approach involving the researcher and a team of raters in data analysis, fostering discussion and a consensus on interpretations. It appears particularly suited for exploring and understanding individual experiences in depth, allowing researchers to capture the complexity of the studied phenomena. The consensus process among raters reduces the risk of bias and contributes to ensuring a balanced view of the data. It can be adapted and customized to fit specific research needs and study contexts. On the other hand, TA represents a flexible methodology that can be applied to a wide range of qualitative data, enabling researchers to explore a variety of themes and concepts. Additionally, TA follows a systematic process involving data coding, theme identification, and organization into a coherent interpretive model, allowing researchers to perform a detailed examination of the data, and to identify patterns, relationships, and implicit meanings within the collected material.

While CQR and TA provide robust and systematic approaches for qualitative data analysis, both methodologies have limitations. In CQR, if the group of analysts is not sufficiently diversified in terms of experience, background, and perspectives, there may be a loss of diversity of opinions, limiting the understanding of the phenomenon under study. Despite efforts to reduce bias through consensus among analysts, personal influences may still emerge and affect data interpretations. Due to the intensive participant-centered approach, the generalizability of CQR results to other populations or contexts may be limited. Regarding TA, if data analysis is not conducted thoroughly, there may be superficiality in understanding the emerging themes, leading to the loss of significant details and nuance. Because TA involves a degree of subjectivity in coding and theme identification, there may be a lack of objectivity in the results. Furthermore, in attempting to identify cross-cutting themes, the richness of the specific context in which the data emerged may be lost, limiting the overall understanding of the phenomenon. In summary, both CQR and TA present potential challenges that require attention during research planning, execution, and interpretation. It is important for researchers to be aware of these limitations and to adopt strategies to manage them as effectively as possible.

Although these qualitative methods have emerged as the most used by researchers, at the same time, there is a lack of consensus within the scientific community regarding the methods that should be employed for qualitative analysis of SC. The included studies certainly revealed areas of convergence in exploring the phenomenology of SC, but there were also many areas of divergence. We found these discrepancies in the methods as well as the SC categories the researchers analyzed or identified, highlighting that there is no consensus regarding the conceptualization of the construct (Zaccari et al., 2024). Regarding data collection, two-chair dialogues have been identified as the most used psychotherapeutic technique for detecting and observing SC from a qualitative perspective (Bailey et al., 2020, 2022; Halamová et al., 2021). Compared to interviews, this technique allows participants to experience and express self-critical thoughts and feelings in real-time. Hence, this approach provides a deeper understanding of their self-critical tendencies than recalling and reporting them during interviews.

Regarding the employed coding procedures, the included studies used bottom-up approaches more often than top-down approaches. In other words, researchers usually explore SC without any predetermined category in mind. Because bottom-up coding strongly relies on the raw data itself, it ensures adherence to participants’ experiences and allows researchers to discover unexpected insights within the dataset. However, the analysis is not guided by theoretical frameworks, so the coding process might lack direction or focus. Moreover, generalizing results derived from pure bottom-up coding processes might be challenging, as the categories heavily rely on raw data from a specific population. On the other hand, top-down coding allows researchers to test specific hypotheses or theories, thus providing a clear direction for the analysis. Although top-down coding is based on predetermined categories, generalization of the results might be easier, but this approach may involve confirmation biases: researchers may be influenced by preexisting theories while interpreting the data. At the same time, researchers may overlook findings that do not fit their predetermined categories. Because both coding methods have advantages and disadvantages, researchers might employ mixed-method approaches combining both top-down and bottom-up coding, thus maximizing the strengths of both methods and minimizing their respective weaknesses. Among the 16 studies included in the present review, only 4 (Bailey et al., 2022; Redden et al., 2021; Ripoll-Núñez & Naismith, 2022; Thew et al., 2017) employed mixed methods. To conclude, the combination of top-down and bottom-up approaches allows for flexibility. While the identification of domains from the literature provides structure and guidance, the subsequent codification allows for adaptation to unexpected findings or emerging themes. Therefore, the integration of qualitative and quantitative approaches can further enrich the understanding of SC, thus leading to more robust and meaningful results.

The results obtained in most of the included studies should be interpreted with caution, as only 4 studies reported calculating inter-rater reliability. Given the lack of data regarding inter-rater reliability, there is a risk that the interpretations of the data might be subjective or inconsistent. We highly recommend that future qualitative research on SC to assess inter-rater reliability to enhance the trustworthiness of research findings.

Implications for psychopathology

Among the clinical samples, the authors enrolled individuals with various psychiatric or organic conditions. Among those studies enrolling clinical samples, the most frequent diagnosis was depression (Bailey et al., 2020, 2022; Gilbert & Irons, 2004; Granek, 2006; Ripoll-Núñez & Naismith, 2022; Thew et al., 2017; Vanheule et al., 2008), followed by anxiety disorders (Bailey et al., 2022; Ripoll-Núñez & Naismith, 2022; Toolan et al., 2019), eating disorders (Thew et al., 2017), and post-traumatic stress disorder (Ripoll-Núñez & Naismith, 2022). Depression might be the most represented psychopathological condition because SC was first conceptualized within a model of depression (Blatt et al., 1982; Blatt & Zuroff, 1992), which identified SC and the dependent personalities as personality configurations that predispose individuals to the development of the disorder. Besides being an important risk factor in the etiopathogenesis of depression, SC is a common symptom of depressive disorder, and thus this construct has been extensively studied in relation to depression. On the other hand, the inclusion of participants with different clinical conditions might be attributed to the conceptualization of SC as a transdiagnostic construct in subsequent models (Gilbert et al., 2004; Shahar, 2015). Variations in sample characteristics across studies (e.g., nationality, cultural background, and health condition) may limit the generalizability of the findings to broader populations. In particular, the expression of SC may vary significantly across cultures, depending on the specific norms, values, and beliefs of each culture (Heine, 2003). Furthermore, health conditions, referring to both psychological and organic disorders, can significantly influence an individual’s experiences of SC. For example, individuals with different psychopathological conditions may exhibit different patterns of SC compared to those without such conditions (Werner et al., 2019). Therefore, the results derived from studies conducted on populations with specific cultural characteristics and clinical conditions should be interpreted with caution. It follows that the inclusion of individuals with various health conditions in studies on SC is essential to capture the multifaceted nature of SC and its manifestations across diverse populations. Authoritative systematic reviews and meta-analyses (Löw et al., 2020; McIntyre et al., 2018; Wakelin et al., 2022; Werner et al., 2019) have highlighted how SC is present in various clinical disorders and represents a negative factor for therapeutic outcomes, constituting an obstacle to change. However, no contribution has yet captured the different manifestations of the phenomenon within various clinical conditions (Zaccari et al., 2024). Therefore, deepening knowledge in this field is crucial for understanding the complexity of the construct and for designing targeted and effective interventions for the specific types of SC across different psychopathological profiles.

Strengths and limitations

This review is subject to several limitations. First, we did not analyze the quantitative methods used in the included studies. The combined approach of quantitative and qualitative methods could provide a more comprehensive and accurate understanding of the phenomenology of SC, as well as a more precise empirical validation of the theoretical models present in the literature. Second, we were unable to analyze some of the studies that we selected in the initial phase of the review because the full text was unavailable. These studies might have provided additional valuable information to enrich the research. Additionally, in the preliminary phase of the review, we did not consider the keyword “grounded theory.” It should be noted that there may be additional relevant studies that we did not include in this review.

Despite these limitations, we believe that our results, obtained through a systematic review based on the PRISMA guidelines, can provide a substantial contribution to the understanding and deepening of SC. Our review emphasizes the importance of establishing a shared understanding of the phenomenology of SC, particularly through studies employing qualitative methods. Therefore, this review can help to guide future research with the aim of achieving a deeper understanding of SC, especially in relation to its implications for psychopathology and psychotherapy. We believe that research in this area should be intensified to capture the various qualitative manifestations of the SC construct. There is clearly a need to develop more precise and consistent categories that fully reflect the complexity and multidimensionality of SC. It is necessary to test the current theoretical models through qualitative analysis or to develop new models that consider the complexity of SC, which could then be examined using qualitative and quantitative methods in different clinical and non-clinical populations to capture the variability and different facets or declinations of SC. Shedding light on the qualitative components of this transdiagnostic construct would enhance our understanding of how specific components of SC are associated with different psychological variables and, consequently, with different psychopathological conditions. An advance in the research field would allow for more targeted and effective treatments.

Conclusion

We aimed to provide a comprehensive overview of the qualitative methods frequently employed in the literature to study SC. This exploration highlighted that there is no single, shared trend of qualitative analysis for this construct, and revealed a critical gap in the literature regarding a lack of consensus among researchers about the conceptualization of SC (Zaccari et al., 2024). The identified categories of SC, through qualitative analysis, often do not reflect the existing theoretical models, especially for studies that employed a bottom-up coding approach, and often the extracted categories were not retested or reanalyzed by other studies. In fact, the authors of all the included studies have recommended exploring the construct rigorously and systematically in future studies. Overall, the multifaceted nature of SC in the literature highlights the need for a comprehensive understanding that incorporates various theoretical perspectives. Future research could benefit from integrating these different viewpoints to gain a more nuanced understanding of SC through qualitative methods.

Author contributions

VZ and all co-authors assumed the overall responsibility for conceptualizing the systematic review. MF, FS, VC, LT, MGF, VP, SJC, and MGG searched the scientific databases for articles and systematically evaluated each study for relevance. VZ, MF, FS, VC, LT, MGF, VP, SJC, MGG, FDO and FM were involved in the analysis and interpretation of the data, as well as the drafting and revision of the final article. All authors contributed to the article and approved the final version.