This was the first study to examine irony comprehension among individuals with BPD. Participants were presented with both ironic and literal text messages varying in praising and critical intention. Within a signal detection framework, we assessed negativity biases and the ability to discriminate literal from ironic remarks. Biases were distinguished on two levels: implicit tendencies measured in the choice of the literalness of the statement (ironic vs. literal) and explicit ratings of perceived intention (critical to praising).
Participants with BPD exhibited more difficulty differentiating literal from ironic remarks than HC. Yet, group differences did not vary with critical or praising content, showing that it was the literality of the stimulus, not the intention, affecting performance. For both groups, ironic praise was harder to detect than ironic criticism, replicating that ironic criticism is easier to process (70, 71), mostly because it is much more common (71–73). A reduced sensibility was explained by borderline pathology beyond groups, confirming findings among healthy adults with borderline traits in a clinical sample (61) and corroborating the idea of dimensional approaches to personality disorders (74). The current results are commensurate with those in social cognition paradigms using sarcasm as a stimulus (15, 22, 44). For the first time, these impairments have been confirmed with respect to verbal irony.
Other forms of nonliteral language, such as metaphors, have recently been shown to be preserved in BPD (75). This is of particular importance, as metaphor comprehension is commonly impaired in schizophrenia (76) with whom BPD patients share symptoms (32, 33). In addition to emphasizing that linguistic phenomena are subject to different cognitive structures, this shows that psychopathologies may be subject to them as well. Both metaphor and irony require an abstraction from the literal words, but irony further demands to integrate multiple mental states (8, 77, 78). Ironists do not intend to deceive, but seek duplicitous understanding. As such, irony proves particularly challenging for epistemic trust, as the listener must decide which of the conflicting pieces of information to trust and which to question. In our study, errors in the detection of irony implied that in some instances patients decided to stick to the literal meaning, even when an incongruence between context and target sentence suggested otherwise. A lack of epistemic trust may make it more difficult for individuals with BPD to decide which information is relevant in an ambiguous context (10). As a result, they may adhere to one perspective (16, 79) and choose a context-inappropriate interpretation. Indeed, shifts in the representation of the self and of others have long been deemed problematic in BPD (31). Empirically, patients experience difficulties alternating between egocentric and altercentric perspectives with face-morphing tasks (52) and show overlapping self-other boundaries on a bodily and cognitive level (51, 52). Accordingly, in our study, patient’s personal distress in response to others’ emotions was higher and cognitive perspective-taking lower than those of HC, replicating previous findings on self-reported empathy scales in BPD (80–82). However, both were unrelated to outcomes in our study of irony. Future studies should include more complex social cognition paradigms that may be more commensurate with metacognitive processes than with self-ratings (46), and should compare speech varying in self-other representation (e.g., deceit and faux pas).
Contrary to our expectations, HC (and not BPD) tended to interpret stimuli critically when looking at implicit response biases. These were measured in participants’ interpretation of a remark as literal or ironic. Interpreting literal praise ironically HC ascribed negative intent to literal praise (“I have an A in my test” “You are clearly not smart”), while ironic criticism was interpreted as literal criticism. The same negativity bias was evident in interpretations of ironic praise and literal criticism: HC tended to interpret these statements literally, considering ironic praise as literal critique and literal critique as such. Although individuals with BPD’s difficulty to determine the literality of a statement may be explained by overattribution (83), our results did not indicate that these attributions are hostile when measured implicitly.
Response biases were further analyzed with explicit ratings of perceived intention. Both groups tended to rate ironic utterances as less praising and critical than literal ones, confirming the well-known perception of irony as “tinged” with the literal meaning (84, 85). In contrast to implicit biases, patients with BPD showed a negative interpretation bias in explicit ratings. They perceived praising remarks as less praising than HC. So far, only a limited number of studies have investigated the effect of positive social stimuli in BPD (47). Our findings are in line with BPD participants’ fear of positive appraisal (86), negative ratings of appreciating video-clips (87) or self-referential information (88) and approach-avoidance behavior (89, 90). Muting the positive experience of praise has major implications, since positive feedback is a crucial part of the therapeutic process (91) and of positive interactions with others. Yet, contrary to other studies (87) and patients with BPD’s heightened rejection sensitivity (92) clinical participants did not differ to HC in perception of critical remarks.
Several factors may contribute to the discrepancy between explicit and implicit biases in BPD. First, the current stimuli were written text messages. Negativity biases in BPD have mostly been found in facial emotion recognition (3, 47, 93, 94), especially when combined with other modalities (95, 96). Further, most biases accompany anger and disgust (3, 93) or neutral stimuli (3, 59, 97). Critique may be less arousing than expressions of anger. And irony is almost impossible to be neutral, as its principal function is to tacitly convey an opinion of the ironized content (50). Yet, unlike most emotion recognition paradigms (47), the current one allowed for an implicit assessment of response tendencies. In this context, our findings correspond with Kobeleva et al.’s (98) comparing approach-avoidance reactions with photographs of positive and negative expressions. In their study, patients tended to rate positive faces as less approachable, but there were no differences between BPD and HC in avoidance and approach behaviors. Similarly, Franzen et al. (99) measured participant investments in a virtual trust game without asking for evaluations of the trustees. HC considered the trustees’ facial expressions and adjusted their investments accordingly, while BPD relied exclusively on the fairness of the trustees’ decisions. Our results align with these findings and indicate that negative appraisals in BPD tend to emerge in contexts in which emotional ratings are salient.
We are aware of several limitations. First, our patient sample had a high verbal IQ and educational background, which may be less prevalent in BPD among the general population (100, 101). Further, patients were recruited in a specialized ward for dialectal behavioral therapy (102), which trains the differentiation between self and other and emotion regulation. This might have minimized group differences. Second, our stimuli did not contain prosody or facial expressions. However, the study focused on text messages - a conversational context that is a major part of current communication. Third, the concept of irony transcends verbal irony, such as situational irony, hyperbole or understatement (103). Fourth, we did not account for experienced abuse or neglect, which is considered to be the origin of epistemic distrust (10). Lastly, this study did not include clinical controls, leaving the question of clinical specificity to be explored. As both personality dimensions often overlap (33), it is important to analyze distinct and/or shared mechanisms of disorders in the schizophrenic spectrum and BPD. Contrary to other studies (28, 29, 61, 104, 105), schizotypal symptoms did not explain irony detection beyond borderline symptoms in our study, although patients scored high on both. This further stresses the need to control for borderline symptoms when assessing irony comprehension in different pathologies.