A Selective Effect of Partial Sleep Deprivation on Metaphor Generation Among Healthy Young Adults

ABSTRACT

Partial sleep deprivation (SD), a common phenomenon in modern life, is known to impair cognitive and linguistic processes. This study investigates its selective effect on metaphor generation, differentiating between conventional and novel metaphors. Twenty‐five healthy young adults underwent two sessions involving either regular sleep or partial SD conditions. Participants completed tasks evaluating metaphor generation, cognitive flexibility (Stroop Test), and executive function (Wisconsin Card Sorting Test). Results revealed that partial SD increased the production of conventional metaphors, associated with bottom‐up processes and default mode network (DMN) activation. However, it did not significantly affect novel metaphor generation, which relies on top‐down cognitive control and frontoparietal network (FPN) activity. Furthermore, Stroop Test findings showed slower reaction times under incongruent conditions after partial SD, indicating reduced cognitive control. These results suggest that partial SD selectively impairs top‐down processes while enhancing bottom‐up retrieval, leading to a shift in metaphor generation patterns. This research highlights the distinct neural mechanisms underlying different types of metaphor generation and their differential sensitivity to partial SD, providing theoretical and psycholinguistic insights into the cognitive effects of SD.

1. Introduction

Sleep deprivation (SD) is becoming increasingly prevalent in the modern world. In recent years, substantial evidence has accumulated regarding the negative impact of sleep restriction and deprivation on physiological and behavioural functions, including memory consolidation (Walker 2008). SD has been shown to impair various language and cognitive processes such as speech perception, non‐word reading, and auditory temporal order judgement (Fostick et al. 2014; Liu et al. 2024). Evidence also suggests that SD affects cognition, reducing attention, learning, and working memory, while being associated with heightened neural reactivity (Dai et al. 2012). Affecting these cognitive abilities may in turn impact figurative language processing. Specifically, the effect of SD on metaphor generation has not been tested before.

Metaphor is a well‐known form of figurative language. Generating metaphors for a given concept requires the activation of a concept from a seemingly unrelated domain of knowledge and establishing a cross‐domain mapping between the two (Gentner and Wolff 1997). This process goes beyond the process of metaphor comprehension. Metaphor comprehension relies on the activation of a base concept within the semantic memory network and assessing how its attributes align with those of the target concept. This process demands a flexible search to allow access to relevant features while inhibiting irrelevant ones for both the target and the base term. The ability to generate creative metaphors is linked to fluid intelligence (Beaty and Silvia 2013), executive control (Silvia and Beaty 2012), selective attention (Menashe et al. 2020), verbal fluency (Kasirer and Mashal 2016), idea fluency (Kasirer and Mashal 2018), and a flexible semantic network (Li et al. 2021). On the other hand, the ability to generate conventional, familiar metaphors primarily relies on vocabulary knowledge and meaning retrieval from the mental lexicon. Thus, generating metaphors is regarded as a facet of creative cognition that capitalises on cognitive control and mental flexibility.

A well‐known effect, termed the Serial Order effect, posits that generated ideas become more creative over time. According to Mednick's (1962) classic associative model, common ideas are more readily accessible in semantic memory. As a result, familiar ideas are activated and retrieved before less common, more distantly related concepts. Thus, activation in the semantic network spreads gradually from close and highly frequent associations to remote and less frequent ones. However, this effect is moderated by the involvement of executive functions, such that as fluid intelligence increases, the serial order effect decreases (Beaty and Silvia 2012). A recent study demonstrated that when participants were asked to generate as many original and creative explanations of a given concept, 70% of the responses were novel metaphors, but about 20% were conventional ones and about 10% were literal explanations (Lifshitz‐Ben‐Basat and Mashal 2021). Thus, when people are tasked with generating creative metaphoric explanations for given concepts, they still generate conventional familiar metaphors.

From a brain research perspective, generating metaphors involves an interplay between brain networks that are interwoven during creative cognition processes. Generating conventional, familiar metaphors depends largely on a bottom‐up, sense‐retrieval process, associated with the retrieval of information from long‐term memory. Generating novel, creative metaphors, on the other hand, involves both bottom‐up and top‐down processes. These processes encompass idea generation, which is associated with the default mode network (DMN), a network linked to internally directed attention and spontaneous cognition (Andrews‐Hanna 2012; Beaty et al. 2014; Mayseless et al. 2014). In contrast, idea evaluation is associated with the executive control network (ECN) (Beaty et al. 2016; Benedek 2018) or the frontoparietal network (FPN) (Jung et al. 2013; Abraham 2014; Dietrich and Haider 2017). Recent evidence from studies on divergent thinking suggests that core ideas emerge in posterior brain regions located within the DMN/semantic system. Additionally, the prefrontal cortex (PFC), as part of the ECN, plays a role in evaluating and regulating creative outputs, constraining these ideas to meet task requirements (Beaty et al. 2016; Benedek 2018; Vartanian et al. 2018). An additional network, the salience network (SN), is involved in modulating the transitions between these two networks, helping to balance idea generation and top‐down control. This interaction is essential for achieving the flow and flexibility required in verbal creative thinking (Beaty et al. 2016; Vartanian et al. 2020). The top‐down idea evaluation process is also linked to the dorsal attention network (DAN), which is consistently involved in verbal and visual working memory tasks as well as tasks requiring attentional control (Majerus et al. 2018).

DMN and FPN collaborate closely during creative and divergent thinking processes. However, research suggests that their relationship during cognitive tasks is often competitive and antagonistic. Uddin et al. (2009) investigated the functional connectivity between components of the DMN and FPN, proposing that the activation of one network suppresses the other, depending on task demands. Similarly, Xin and Lei (2015) found an increased anti‐correlation between the DMN and FPN during working memory tasks compared to the resting state. More recently, Koshino, Osaka, et al. (2023) examined the relationship between DMN and FPN in the context of age‐related cognitive decline during a working memory task. They found that older adults exhibited weaker negative correlations (i.e., reduced antagonism) between the DMN and FPN compared to young adults, suggesting that diminished antagonism may contribute to cognitive decline. Thus, generating metaphors—a creative thinking process—may engage both a bottom‐up process driven by the DMN and a top‐down control process mediated by the FPN, with the SN functioning as a dynamic “switch,” regulating these networks in response to cognitive demands (Schimmelpfennig et al. 2023).

Cognitive control and cognitive flexibility are two executive functions that have been examined in the context of sleep deprivation (SD). Recent studies highlight a selective effect of SD on executive functions. While many convergent and rule‐based reasoning, decision‐making, and planning tasks appear relatively unaffected by SD, evidence suggests that tasks requiring sustained attention and higher‐level processing, such as reading comprehension, were negatively affected by SD (Pilcher et al. 2007). Furthermore, Gevers et al. (2015) observed that some top‐down cognitive control processes were impaired by SD, while bottom‐up processes essential for word retrieval remained relatively intact. Moreover, Kusztor et al. (2019) explored the selective effects of SD on subcomponents of cognitive processes using ERP recordings during Stroop test performances. They found that 24 h of SD led to a decline in sustained attention, reflecting a gradual breakdown of top‐down control. Conversely, automatic processing and the associated ERP waveforms demonstrated greater resilience to SD. The paucity of studies that tested the effects of SD on more creative and innovative aspects of cognition, such as divergent thinking, shows inconclusive results. One study indicated that 32 h of sleep deprivation resulted in participants producing fewer creative responses compared to normally rested individuals. This was demonstrated by a decrease in the number of unique or original ideas generated and a reduced capacity to adapt strategies (Horne 1988; Harrison and Horne 1999). However, as reviewed by Jones and Harrison (2001), several clinical and experimental studies on SD individuals have reported impairments in the Wisconsin Card Sorting Test (WCST), particularly in the tendency to persist with ineffective strategies. Nevertheless, some studies have found no significant effects. It should be noted that in addition to these inconsistent findings, it remains unknown how partial SD affects creative and flexible thinking performance.

As discussed above, there is growing interest in the effects of sleep deprivation on cognitive skills, particularly in relation to language and its underlying neural processing (Dai et al. 2020; Killgore 2010; Sana et al. 2021). However, the impact of partial sleep deprivation—an increasingly common phenomenon—on language, and specifically on metaphor generation, remains unexplored. In the present study, we aimed to test for the first time whether a reduction of 3 h from the duration of regular sleep, hence partial SD, differentially affects the generation of conventional versus novel metaphors. Cognitive control and mental flexibility, abilities that underlie metaphor generation, were also assessed, using The Stroop test and WCST. Given the distinct cognitive abilities required for novel vs. conventional metaphor and the distinct neural mechanisms underlying each type of metaphor, we hypothesise that partial sleep deprivation may affect them differently. Based on the known effect of SD on cognition, we hypothesised that conventional metaphor generation may not be affected by SD, while novel metaphor production, which relies more heavily on cognitive control, may decrease.

2. Method

2.1. Participants

Participants were 25 healthy adults (11 men, 14 women) aged 20–33 years (M = 25.16 years, SD = 3.16). This sample size is based on previous studies on the effect of SD on cognitive skills that employed similar sample sizes (Drummond and Brown 2001; Casement et al. 2006; Pekçetin et al. 2021). All participants were native Hebrew speakers. All participants provided signed informed consent. The study was approved by the Ariel University Ethics Committee.

2.2. Sleep Monitoring

Participants' sleep was monitored using the participants' smart watches (Garmin vivoactive 4S) or Phillips actiwatch2, a non‐invasive tools for monitoring human rest/activity cycles. The output file of each participant, which includes information about the duration and quality of sleep, as well as the participants' reports regarding their sleep on the measured nights (exact sleeping hours and sleep duration), was used to confirm their sleep or sleep deprivation. Participants were asked to record their regular, full‐night sleep duration for three nights before their participation in the study. Then the mean duration of regular full‐night sleep was calculated for each participant. Partial sleep deprivation was defined as 3 h less than the average participant's regular full‐night sleep duration.

2.3. Behavioural Measurements

2.3.1. Tel Aviv Creativity Test (TACT) (Milgram and Livne 2006)

This test is based on the Torrance tests of creative thinking (TTCT) (Torrance 1966), designed to assess a person's creative thinking. In the current study, the test was used to establish a baseline measure of participants' creativity. Two subtests from the verbal version of the TACT were administered: Uses (Subtest A) and Similarities (Subtest B). In the Uses subset, participants were asked to think of all possible different uses of a given object (four different objects). For example: ‘Please mention all the possible ways to use a chair’. In the Similarities subtest, participants were asked to point out as many similarities they could find between two objects. For example: ‘In what ways are a radio and telephone similar?’. Reliability analyses indicated that Cronbach's alpha was 0.79 for the Uses subset and 0.85 for the Similarities subtest (Kasirer and Mashal 2018). For each participant, a total creativity score was calculated as the sum of the scores from the two subtests.

2.3.2. Metaphor Generation Task (Kasirer and Mashal 2014)

This test evaluates the ability to generate metaphors by presenting participants with concepts related to common emotions (e.g., ‘Feeling sad is like…’). Participants are instructed to complete the phrases in as many unique and original ways as possible, avoiding repetition, synonyms, or rephrasing of the given phrase. There is no time limit for responses. To ensure the participant's thought process and verbal fluency are not disrupted, responses are recorded and later transcribed. Two independent judges evaluated all responses, categorising them as literal or figurative. Discrepancies between judges were resolved by a third judge—a certified speech and language pathologist with expertise in this field. The reliability between judges was high (average of 97%) as was found in our previous study (Lifshitz‐Ben‐Basat and Mashal 2021). Metaphors are classified as creative if they are unfamiliar, unique, and novel non‐literal expressions (e.g., ‘Feeling worthless is like … a totally smashed lemon’), conventional if they involve familiar expressions or idioms (e.g., ‘Feeling embarrassed is like … having a red face’), and literal if they lack figurative meaning (e.g., ‘Feeling successful is like … a victory’). Several outcomes were coded: the total number of novel metaphors, conventional metaphors, and literal responses, as well as the overall number of responses generated. Additionally, the mean number of novel metaphors and the mean number of conventional metaphors across all participants were calculated and analysed. To avoid learning effect between consecutive sessions, two versions of the test were created, each consisting of five sentences. Participants were randomly assigned one version per study session.

2.3.3. The Stroop Test (Stroop 1935)

This test assesses the ability to inhibit cognitive interference elicited by incongruent stimuli. The test measures selective attention, cognitive flexibility, and processing speed abilities that might be influenced by sleep deprivation. During the test, 33 written colour names were shown to the participants. The test was delivered via personal PC screen using DirectRT software (version 2014). The colour names were written in various font colours which did not always fit the written colour name (e.g., the word ‘Blue’ written in red font). Participants were instructed to mention the colour of the font in which the word is written while ignoring the content of the word. Participants' performances were coded in terms of accuracy and reaction times and were analysed separately for congruent stimuli (where the ink colour of the word matched the content of the written word) and incongruent (when the ink colour of the word did not match the content of the written word).

2.3.4. The WCST (Heaton 1981)

This test was used as a measure of executive functions, set‐shifting, problem‐solving, and behavioural flexibility that leans on working memory to integrate feedback given on performance. In this study, the WCST abbreviated version (WCST‐64: Computer Version) (Kongs et al. 2000) was used. This version of the WCST consists of four key cards and 64 response cards with geometric figures that vary according to three perceptual dimensions (colour, form, and number). The task requires participants to find the correct classification principle by trial and error and examiner feedback. Participants were asked to match the cards without explicit instruction about the regularity according to which it is sorted. They were only told ‘right’ or ‘wrong’ after each choice. The correct sorting rule is either by colour, form, or number on any given trial. The task ended after the completion of six rules (categories) or 64 trials, whichever comes first, and it required 10 correct responses before the sorting rule was changed. Total accuracy and reaction time were measured.

3. Procedure

Each participant participated in two study sessions spanning 1 week with a break of 2 to 3 days between the sessions. Participants were instructed to arrive at one study session after a regular night's sleep and at the other study session after a night in which they slept 3 h less than their regular sleep duration. The order of the sessions was randomly controlled so that half of the participants came to the first session after partial SD and to the second session after regular night sleep, and half of the participants in the opposite order. The first session included gathering personal information and filling in informed consent. Then participants filled out the metaphor generation task, the Stroop test, and the WISC test. At the second session, the metaphor generation task was administered (the complementary version to what they filled in the first session) and the Stroop and WISC tests. Study design can be seen in Figure 1.

FIGURE 1.

Study procedure. TACT, Tel‐Aviv Creativity Test; WISC, the Wisconsin Card Sorting Test.

4. Results

4.1. Tel‐Aviv Creativity Test (TACT)

To examine a possible correlation between baseline creativity and metaphor generation, as well as the effects of sleep on this relationship, we calculated the correlation between performance on the TACT and the number of novel metaphors (NM) generated (after a regular or partial SD). A significant positive correlation was found following regular sleep (r = 0.473, p = 0.016) attesting to the creative aspect of the metaphor generation task. However, no significant correlation was observed between TACT performance and NM productions following partial SD (r = 0.284, p = 0.175).

4.2. Metaphor Generation

To examine the potential effect of sleep duration on metaphor generation, we conducted two 2 × 2 mixed‐design repeated measures ANOVAs, analysing novel and conventional metaphors separately, as participants generated, on average, five times more novel than conventional metaphors. Both analyses were conducted with sleep condition (regular sleep, partial SD) as the within‐subject factor and order (regular sleep first, partial SD first) as the between‐subject factor. Sleep effect was found for conventional metaphors (F(1, 23) = 4.71, p = 0.041, ηp² = 0.170), but not for novel metaphors (F(1, 13) = 1.14, p = 0.305, ηp² = 0.081). More conventional metaphors were produced following partial SD (M = 3.40, SD = 2.35) than following regular sleep (M = 2.16, SD = 1.95). No order effect was found for both conventional (F(1, 23) = 0.16, p = 0.693, ηp² = 0.007) and novel metaphors (F(1, 13) = 2.63, p = 0.129, ηp² = 0.168), nor sleep and order interactions (for conventional metaphors: F(1, 23) = 0.50, p = 0.486, ηp² = 0.021; for novel metaphors: (1, 13) = 2.47, p = 0.140, ηp² = 0.159). The number of novel and conventional metaphors generated across the different sleep conditions can be seen in Figure 2.

FIGURE 2.

Averaged number of novel versus conventional metaphors after regular night sleep and after sleep deprivation. An increase in the number of conventional metaphors was found after sleep deprivation.

4.3. Stroop Test

A Pearson correlation analysis was conducted to examine the relationship between reaction times and accuracy under the sleep deprivation condition. The results showed an insignificant negative correlation between Stroop reaction times and accuracy (r(24) = −0.345, p = 0.099). In the regular sleep condition, the correlation between Stroop reaction times and accuracy was also negative and insignificant (r(24) = −0.101, p = 0.640). These findings indicate no accuracy‐reaction times trade‐off in both sleep conditions.

To look for partial SD effect on executive functions and specifically on attention and cognitive control, we conducted two repeated‐measures general linear model (GLM) one for accuracy and one for reaction time, with sleep condition (regular sleep, sleep deprivation) and congruency (congruent, incongruent) as within‐subject factors, and order (regular sleep first, partial SD first) as a between‐subjects factor. For accuracy, the analysis did not reveal significant main effects of sleep condition (F(1,22) = 1.2, p = 0.278, ηp² = 0.05), congruency (F(1,22) = 0.6, p = 0.45, ηp² = 0.03), or order (F(1,22) = 0.004, p = 0.953, ηp² = 0.000). Furthermore, no significant interaction was found between sleep condition and order (F(1,22) = 0.29, p = 0.59, ηp² = 0.01). The interactions between congruency and order (F(1,22) = 0.08, p = 0.78, ηp² = 0.004), sleep condition and congruency (F(1,22) = 0.02, p = 0.89, ηp² = 0.001) and congruency and order (F(1,22) = 0.081, p = 0.778, ηp² = 0.004) were insignificant, nor was there a three‐way interaction for sleep condition, congruency and order (F(1,22) = 0.09, p = 0.77, ηp² = 0.004). Descriptive statistics for Stroop accuracy across sleep conditions and congruency are presented in Table 1.

TABLE 1.

Descriptive statistics for accuracy across sleep conditions and congruency levels. While mean accuracy appears slightly higher in the sleep deprivation condition, statistical analyses revealed no significant main effects or interactions.

Condition	Congruent mean (SD)	Incongruent mean (SD)
Regular sleep	90.06 (27.18)	88.46 (27.46)
Sleep deprivation	96.63 (3.24)	96.15 (6.00)

For Stroop reaction times, there were also no significant main effects of sleep condition (F(1,21) = 2.91, p = 0.1, ηp² = 0.122), congruency (F(1,21) = 3.95, p = 0.06, ηp² = 0.16), and order (F(1,21) = 0.413, p = 0.527, ηp² = 0.019). However, a significant interaction of sleep condition and congruency was observed (F(1,21) = 6.95, p = 0.02, ηp² = 0.25), indicating that reaction times in the incongruent condition were slower following sleep deprivation (M = 831.89 milliseconds; SD = 41.18) compared to after regular sleep (M = 747.55 milliseconds; SD = 31.47). No significant interactions were found for sleep condition and order (F(1,21) = 0.47, p = 0.5, ηp² = 0.02) and congruency and order (F(1,21) = 3.36, p = 0.08, ηp² = 0.138), nor a three‐way interaction for sleep condition, congruency, and order (F(1,21) = 2.83, p = 0.12, ηp² = 0.12). Descriptive statistics for Stroop reaction times across sleep conditions and congruency are presented in Table 2.

TABLE 2.

Descriptive statistics for Stroop reaction times in milliseconds across sleep conditions and congruency levels. The interaction between sleep condition and congruency indicates that reaction times were longer in the incongruent condition after sleep deprivation compared to regular sleep.

Condition	Congruent mean (SD)	Incongruent mean (SD)
Regular sleep	757.97 (28.40)	747.55 (31.47)
Sleep deprivation	769.62 (26.25)	831.89 (41.18)

4.4. Wisconsin Test

To look for possible sleep duration effect on executive functions and specifically on shifting and attention, we conducted two repeated‐measures GLM Wisconsin test scores (accuracy and reaction times), with sleep condition (regular night, partial SD) as a within subject variable and order (regular sleep first vs. partial SD first) as a between subjects' variable. We found a main effect of sleep condition for accuracy scores (F(1,23) = 912.36, p < 0.001, ηp² = 0.98) indicating that participants were more accurate after regular sleep (M = 0.84, SD = 0.03, ηp² = 0.015) than after sleep deprivation (M = 0.82, SD = 0.06). No interaction between sleep duration and order was found (F(1,23) = 0.06, p = 0.8, ηp² = 0.003). For reaction times, no main effect for sleep condition was found (F(1,23) = 0.35, p = 0.56, ηp² = 0.02), but a significant interaction for sleep condition and order (F(1,23) = 14.73, p < 0.001, ηp² = 0.39). Reaction times were longer after partial SD when participants started with regular sleep first (Regular sleep: M = 1279 ms, SD = 347; sleep deprivation: M = 1464 ms, SD = 258) than when the order was of partial SD first (Regular sleep: M = 1475 ms, SD = 207; sleep deprivation: M = 1223 ms, SD = 240).

5. Discussion

The main goal of this study was to examine whether a reduction of 3 h from regular sleep, a partial SD, can affect language processing as measured by metaphor generation. Our findings revealed a significant increase in the number of conventional metaphors generated, while no such effect was found for novel metaphors. We also found that partial SD affected cognitive control (as assessed by the Stroop task). Participants were slower in processing the incongruent stimuli after partial SD than after regular sleep.

To better understand these results, it is important to first look at the neural processes underlying the production of each type of metaphor, and specifically at the mechanisms involved in generating novel metaphors versus familiar, conventional ones. The ability to generate conventional, familiar metaphors largely depends on vocabulary knowledge and meaning retrieval from the mental lexicon. These processes are bottom‐up processes that are associated with the DMN, which has been linked to internally directed attention and spontaneous cognition (Andrews‐Hanna 2012; Beaty et al. 2014; Mayseless et al. 2014). However, generating novel, creative metaphors involves verbal fluency (Kasirer and Mashal 2016), fluency of ideas (Kasirer and Mashal 2018), and fluid intelligence (Silvia and Beaty 2012). This process requires, in addition to bottom‐up processes, the engagement of top‐down cognitive control, including the evaluation of ideas, which is associated with executive functions like idea‐shifting, cognitive flexibility and evaluation. Recently, Sana et al. (2021) demonstrated that novel metaphor comprehension requires more inhibitory control, as individuals must suppress the literal meaning to compute the figurative one. The relationship between the DMN and FPN networks has been shown to be competitive, such that when one network is more dominant, the other becomes less active (Xin and Lei 2015; Uddin et al. 2009; Koshino, Minamoto, et al. 2023). Based on this reciprocal antagonism relationship, we propose that the partial SD in our study may have reduced the top‐down cognitive control (as was shown by slower response times in the incongruent stimuli of the Stroop task) typically associated with the FPN (Gevers et al. 2015; Kusztor et al. 2019; Sana et al. 2021). This reduction in FPN may have led to a compensatory increase in DMN activation and therefore heightened the bottom‐up retrieval process, which facilitated the generation of a greater number of conventional metaphors.

The number of novel metaphors generated was not influenced by partial SD. In essence, novel metaphor generation is a highly complex task that depends on a range of cognitive processes, including selective attention, cognitive control, mental flexibility, semantic retrieval from the mental lexicon, working memory, idea fluency, and fluid intelligence (Beaty and Silvia 2013; Kasirer and Mashal 2016; Kasirer and Mashal 2018; Menashe et al. 2020; Silvia and Beaty 2012). Thus, generating novel metaphors pushes the limits of various cognitive abilities involved in this type of creative output. Therefore, the challenge to the performance of this task may stem from disruptions in several cognitive functions, rather than solely from deficits in cognitive control. It may also be that only a reduction in sleep, as was the manipulation in the present study, rather than an SD of a full night (or even more), is not enough to affect the various cognitive processes involved in novel metaphors generated, as longer periods of SD, or a full night without sleep, might have led. Furthermore, a positive correlation between performance on the TACT and novel metaphor generation was found under the regular sleep condition, attesting to the creative aspect of generating novel metaphors process. This relationship was no longer observed following partial SD, suggesting that some part of creative metaphors generation is susceptible to partial SD and that SD may modulate the production of creative metaphors. To better understand why sleep deprivation did not affect the overall number of novel metaphors generated, future studies should dissect the effects of partial SD on the underlying processes of novel metaphor generation.

To gain a deeper understanding of the results and to further examine the potential effects of partial SD on executive functions—particularly attention, cognitive control, and cognitive flexibility, we assessed participants' performance on the Stroop test and the WCST. The Stroop test is designed to assess cognitive control, particularly the ability to suppress irrelevant responses (Gevers et al. 2015). Performance in this task reflects both bottom‐up and top‐down processes. The bottom‐up process involves recognising stimulus feature repetition to enhance performance (Mayr et al. 2003), while top‐down cognitive control mechanisms engage when conflict is detected. This triggers attentional reconfiguration, allocating additional resources to the relevant target dimension and reducing interference in subsequent trials (Botvinick et al. 2001). We observed longer reaction times for incongruent trials following sleep deprivation (SD) compared to regular sleep, suggesting that these prolonged reaction times reflect a reduction in top‐down cognitive control. This finding is consistent with previous research by Gevers et al. (2015), which showed that SD specifically disrupts top‐down adaptation during the Stroop test. Additionally, this finding is consistent with studies by Kusztor et al. (2019) and Sana et al. (2021), pinpointing that SD impairs executive functions, such as cognitive control and inhibitory processes, leading to slower responses, particularly in tasks requiring sustained cognitive effort.

The performance on WCST was reduced following partial SD, compared with the full night's sleep condition. Finding the sorting rule in the WCST, which changes randomly, relies on a process of trial and error. This process requires cognitive flexibility and the ability to adapt to new rules, both of which are high‐level executive functions sensitive to sleep deprivation (Alhola and Polo‐Kantola 2007; Forest and Godbout 2000; Killgore 2010). Interestingly, WCST RT was not affected by partial SD. This suggests that different mechanisms may underline RT and accuracy, with preserved RT in this specific task possibly reflecting automatic responses and indicating response fluency. In contrast, accuracy, which involves top‐down cognitive control, appears to be more susceptible to the effects of SD. Nevertheless, it should be noted that the influence of SD on the WCST test is inconsistent (Jones and Harrison 2001) and, as Killgore and Weber (2014) mentioned, the WISC sorting task may be more sensitive to detecting significant brain lesions in clinical settings and may lack the precision needed to identify the subtle cognitive effects of sleep deprivation.

There are several limitations to the current study. First, the study tested the effect of partial SD as reduction of the sleep duration is more common in modern daily life than full night SD. However, many studies cited above involved longer periods of sleep deprivation, ranging from 24 to 36 h (Bratzke et al. 2012; Cain et al. 2011; Kim et al. 2011; Sagaspe et al. 2006; Dai et al. 2020). Extended sleep deprivation may have resulted in more pronounced effects in our study. Furthermore, the study is cross‐sectional in design, examining the immediate effects of partial sleep deprivation. The long‐term effects of chronic partial sleep deprivation on cognitive flexibility and metaphor generation were not explored. Second, the lack of significant findings for novel metaphor generation might suggest task complexity, which could overshadow the influence of partial sleep deprivation. Further studies could break down the task to identify specific cognitive processes affected. Third, although the study accounted for order effects by randomising the sequence of regular and sleep‐deprived sessions, potential residual learning or fatigue effects between sessions cannot be entirely ruled out. Lastly, the study included 25 participants whose sleep duration and quality were monitored using smartwatches and actiwatches. Although the sample size is in line with previous studies and adequate for initial exploration, it may limit the generalisability of the findings. Future research could benefit from a larger sample size to ensure greater statistical power and should use polysomnography (PSG), a gold‐standard method for sleep assessment to broaden the applicability of the results.

6. Conclusion

Our study demonstrated the selective effect of partial SD on metaphor generation. Partial SD was found to selectively impair cognitive control processes, presumably due to a reduction in FPN activity. Consequently, it is possible that DMN activity was increased, facilitating the generation of conventional metaphors that rely on bottom‐up processes. Partial SD led to a greater number of conventional metaphor generations but did not affect the production of novel metaphors, which depend more heavily on complex processes and top‐down cognitive control. These findings underscore the involvement of distinct neural mechanisms in the production of different types of metaphors, with differential sensitivity to SD. It is important to note that our participants experienced only partial SD, with a reduction of 3 h of their average nightly duration. While this may have limited the understanding of how extended or full SD might influence metaphor generation and cognitive processes, it showed that conventional metaphor generation is a very sensitive to reduction in sleep. Finally, our study holds both theoretical and clinical implications. Theoretically, it pinpoints the distinct neurocognitive processes underlying metaphor generation and the specific effect of partial SD on each of them. Clinically, it offers insights into the impact of partial SD on common daily cognitive processes and tasks.

Author Contributions

Adi Lifshitz‐Ben‐Basat: conceptualization, investigation, writing – original draft, writing – review and editing, formal analysis. Leah Fostick: conceptualization, investigation, funding acquisition, writing – review and editing, supervision, formal analysis. Nira Mashal: formal analysis, supervision, writing – review and editing, conceptualization, investigation, methodology.

Conflicts of Interest

The authors declare no conflicts of interest.

Lifshitz‐Ben‐Basat, A. , Fostick L., and Mashal N.. 2026. “A Selective Effect of Partial Sleep Deprivation on Metaphor Generation Among Healthy Young Adults.” Journal of Sleep Research 35, no. 2: e70139. 10.1111/jsr.70139.

Data Availability Statement

The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.