Discourse analysis has become a powerful approach for understanding better how people use language to express and share their points of view, and how it allows them to build their social realities. Through discourse analysis we can detangle some of the entrenched relations of form and meanings, recognising how the form becomes an automated device to convey complex social meanings (Johnstone & Andrus, 2024). But this is a complex task, as it requires identifying the discursive places where significant forms and meanings get together. Metaphors have been considered as one of the most frequents cognitive and linguistic devices that allows speakers to do this by inserting new concepts through a projection process (Lakoff & Johnson, 1980; Kövecses, 2020). Because of that, speakers can conceptualise emerging realities and new concepts.

Some studies observe that adolescents are the age social group that create new meanings more frequently (Zimmermann, 1996). Studying how adolescents use metaphors in everyday contexts can help to improve not only theoretical inquiries —such as the relation between language and thought (Gibbs, 2008; Steen, 2023), but some linguistic practices —such as education activities (Cameron, 2003). However, it is a challenging task because it requires assuming theoretical guidelines with a methodological framework that allow gathering the contexts in which metaphors appear (Rubbo et al., 2025a).

Metaphors are a fundamental part of human language and cognition skills. There is a lot of evidence about its powerful relevance for learning processes (Martínez et al., 2001; Noyes, 2006; Petrie & Oshlag, 2012; Wegner et al., 2020). They are a device for expressing concepts, mainly those abstract or complex ones (Lakoff & Johnson, 1980; Kövecses, 2010, 2020). In that sense, the use of metaphors both enhance a concept with more information and makes it more accessible to interlocutors.

Although this description usually characterises the use of metaphors concerning abstract concepts, such as emotions (Kövecses, 2004). It is also observed that they convey expressiveness and feasibility to translate complex concepts into simpler ones, making the metaphor a bridge to explain and to narrate novel experiences. When they are assembled, it can increase the current state of knowledge (Ciapuscio, 2013). Also, by using them, speakers can express abstract and complex thoughts and the subtleties of their subjectivity (Kövecses, 2004). Because of these features, metaphors are a common device in language (Kövecses, 2004, 2010, 2020; Lakoff & Johnson, 1980; Steen, 2023).

The most widely method to study metaphor usage has been corpus studies (Sanford, 2010; Deignan, 2008) and experimental or quasi-experimental designs, with standardised tests that assess comprehension (Ferreres et al., 2007) and production (Kogan et al., 1980). Since 2020 quasi-experimental studies with remote test administration have increased, mainly due to the public policies of social isolation implemented during the COVID-19 pandemic. They tend to use a comparative and causal structure, which is “X is like... because…” (Aydoğdu & Gürsoy, 2021; Colak, 2022; Gök & Kara, 2022).

Nevertheless, there is evidence that the kind of task and the subjects’ expertise determine the linguistic outcomes (Nippold et al., 2005; Wallis & Westerveld, 2024), in the case of metaphors assessments, the tests tend not to consider them.

Regarding metaphor development, different stages have been described. Some studies agree that around 10-12 years old several changes can be observed (Vernarská & Zelenayová, 2018). The explanations posit differ causes: some studies consider they could be a correlate of the onset of productive use, and others sustain them as evidence of its mastery. The highest creativity peaks are identified at the end of adolescence (Cacciari & Padovani, 2012). In relation to gender differences, Fattahi and Nushi (2021) and Rocha et al. (2021) found no significant differences and no age effect (Willinger et al., 2019). On the other hand, Hussey and Katz (2010) and Vernarská & Zelenayová (2018) found differences. While the former attribute these to pragmatic cues, the latter adduce these to cognitive factors.

Summarising, although the amount of evidence reported, the characteristics of adolescent’s discursive metaphors usages have not been clarified precisely yet. And some of the characteristics recognised become from decontextualised studies which do not consider natural contexts. This study aims to describe adolescents’ spontaneous written discourse about the pandemic, analysing health metaphors elicited by a novel ecological test. It assesses adolescents in a writing task, very similar to a classroom activity. This task allows us to analyse and describe metaphors usages, considering differences in pragmatic requests and gender.

80 handwritten records were analysed, with a Mean of 9.3 utterances and 207.5 words per participant (753 utterances, Range3-24; 16,811 words, Range 53-496).

A mean of 2.8 metaphors per participant was identified (total of 225 metaphors, Range 0-10). Lexicalised metaphors were the preferred category, with a Mean of 1.8 (148, Range 0-8). Table 1 shows the analysis results in frequencies, percentages, and ranges.

Table 1. Table 1. Metaphors by type Source: Own elaboration.

The analysis by task has shown that the pragmatic request implies differences in the frequency of metaphor production (see Table 2).

Table 2. Table 2. Metaphors by question Source: Own elaboration.

The highest frequency of metaphor production can be found in question 3 (27.6%). This question asks examples about the pandemic. In second place is question 2 (21.8%), which requests an explanation for someone who is unfamiliar with the topic and context (the time-traveller). Among those questions which request an explanation of a personal experience (4, 5, and 6), question 4 (20%) was the most metaphorically productive.

The thematic analysis shows that different words are chosen to mention the pandemic and COVID-19. Figure 1 shows the lemmas used for the disease. The verb perder (“to lose”) was the most frequent, even more so than atacar (“to attack”), which is typical of the WAR domain. Words such as gripe (“flu”) and fiebre (“fever”) appear, too. It means that to refer COVID-19, students have metaphorized it as another illness or its symptoms.

Figure 1. Figure 1. COVID-19 metaphors by lemma Source: Own elaboration.

Figure 2 shows the items used for pandemic. Although we observe that some of them were included to refer COVID-19 —such as atacar (“to attack”), compared with Figure 1 we found more lexical variety there. The most frequently used lemmas were llegar (“to arrive”) and encerrar(“to lock [someone] up”). In the third place is the verb atacar (“to attack”).

Figure 2. Figure 2. Pandemic metaphors by lemma Source: Own elaboration.

We carried out a related-group difference analysis considering the ratio of thematic metaphor per subject. Assumptions of normality were analysed using the Shapiro-Wilk statistic, which showed that the distribution of the measures departed significantly from a theoretically asymptotic normal distribution (SW = .929, p < .001). For this reason, it was decided to use the Wilcoxon rank paired samples comparison statistic.

At the 5% significance level (p < .05), the test results showed that the ratios of COVID-19 and pandemic metaphors are significantly different, in favour of pandemic (p < .001). Table 3 shows the Mean, Median and Standard Deviation of the illness and pandemic metaphor ratios.

Table 3. Table 3. Metaphors by theme (COVID-19 and pandemic) Source: Own elaboration.

Regarding the use of metaphors by gender, an independent sample comparison analysis was carried out on the adolescents identified as female and male. It was decided to discard from this analysis 2 students who identified themselves as non-binary. The Shapiro-Wilk test suggested a violation of the normality assumption (SW = 0.937, p < .001) and the Levene’s test indicated that the variances were not homogeneous (p = .007). The U Mann-Whitney test identified a significant difference between groups (U = 450, p = .002) favouring the female adolescents. Descriptive results by gender are shown in Table 4.

Table 4. Table 4. Metaphors by gender Source. Own elaboration.

We observed that the test and the type of administration allowed us to elicit significant data. We value the high level of acceptance among the students. Some of them included written messages expressing their enthusiasm, such as the following:

5) Espero que ayude lo que les conté♥

“I hope it helps what I have told you♥”.

Also, they made an effort to be clear and informative in their answers. Furthermore, the students made suggestions to improve the test and the administration.

As we have pointed out, the designing of the test has involved theoretical and methodological decisions. Metaphors are not merely an analogical device, their nature is a complex symbolic one (Di Vincenzo, 2021). Because of that, the methods to study them cannot only test metaphors of an analogic nature. If so, the Conceptual Metaphor Theory assumptions of Lakoff and Johnson (1980) would be discarded because an analogical procedure is a translational one, and it does not consider mappings (Kövecses, 2020), nor the speaker’s point of view (Aydoğdu & Gürsoy, 2021). In fact, the first step when we design tests is to evaluate their theoretical framework and population adjustment levels (Iacobuzio et al., 2022).

The tests of Aydoğdu and Gürsoy (2021), Colak (2022), and Gök and Kara (2022) have shown highly effectiveness at metaphor elicitation. They display a syntactical structure making their insertion mandatory (X is like... because…). Though this structure can be useful requesting metaphor production, it does not consider that using them is motivated by a discourse function, such as to explain, to narrate or to give an example. Studying metaphor usage in written discourse also allows us to test other performance levels of skills and knowledge, such as Theory of Mind, syntactic knowledge, etc. In sum, writing always involves its functional use of the metaphor as a symbolic device (Halliday & Matthiessen, 2013). The data obtained in this research confirms the relevance of collecting metaphors by quasi-experimental settings because it allows us to consider them in their natural discursive context and function.

Regarding the analysis, we found that the judges’ agreement was higher than in a previous work (Rubbo et al., 2022). We attributed this difference to the improvement of the codifiers’ training method. The analysis requires exhaustive training because, unlike other devices in the study of language and cognition (i.e. grammatical features), metaphors require a higher level of metacognitive thinking (Di Vincenzo, 2021). This observation aligns with Rubbo et al. (2022), who suggest building a fine-grained metaphor definition —not only because it is a problematic issue, but also because the corpus identification requires a metacognitive process that can only be reached by practice.

We have found that the Mean index of MID is likeness with previous studies (Rubbo & Silva, 2024; Rubbo et al., 2022, 2023a, 2023b, 2024). These indices, which are around 0.6, could be a trend usage.

Regarding the frequency of metaphor’s categories, our results differ from previous ones (Rubbo et al., 2023a, 2023b). We consider the differences may be due to the discursive genre. We believe this design could break the unique metaphor mapping process throughout all the discourse, promoting the displaying of new metaphors in each answer. This could be caused, also, by the changes in the pragmatic requests —explanation, exemplification, and narration.

In other words, in opposition to the traditional metaphor tests, and in a similar way as other tests for assessing language and cognitive skills (Silva et al., 2016, see also Nippold et al., 2005 and Wallis & Westerveld, 2024), the ecological framework of this test has influenced the metaphors usages.

The results of MID index by task have shown a higher use frequency in question 3, which requests an example (¿Qué ejemplos le darías para que se imagine la situación? | “What examples would you give to help the time-traveller to imagine the situation?”). This category concentrates around a quarter of the total metaphors produced (27.6%). It seems that asking for an example after an explanation could be a fruitful activity for metaphoric production. In fact, question 2, which asks for an explanation of the pandemic for an imaginary interlocutor, produced a lower use (21.8%).

Considering the metaphor occurrence in explanatory sequences, we observed the first two questions show different frequencies. While question 1 requests a COVID-19 explanation (15.6%), question 2 asks for the pandemic taking into account a specific interlocutor (21.8%). Although the difference in frequency is not so much, it is possible that the students consider that the first answer did not exhaust the request, so the second is a better discursive niche to insert metaphors. We believe that the question theme and the mention of a specific interlocutor allow this through Theory of Mind applying.

We have observed another difference in those answers that request explanations from a personal experience point of view (questions 4, 5, and 6). Although all of them allow the insertion of a narrative sequence, question 4 is the most favourable for metaphors occurrence (20%). This requests the narration of the pandemic and the lock-down experience (¿Cómo viviste la cuarentena y el aislamiento por COVID-19? | “How did you experience the lock-down and isolation by COVID-19?”).

A significant difference in favour of girls was found regarding the frequency of metaphor by gender. This result is aligned with Hussey and Katz (2010) and Vernarská and Zelenayová (2018), but differs from Fattahi and Nushi (2021), Rocha et al. (2021), and Willinger et al. (2019). We believe it is needed to find the factors that could explain this difference and to explore deeper the gender differences in future studies.

The qualitative analysis shows different metaphorical lemmas for consolidated —or easy comparable— (COVID-19) versus not consolidated —or relatively new— experiences (the pandemic). In fact, in Figures 1 and 2 we can observe that students used more lexical diversity to metaphorize the pandemic. We can assume that a new or unfamiliar phenomenon —the global experience of pandemic— promotes more varied metaphorical expressions, while a highly familiar phenomenon —the experience of illness— promotes more established metaphors. Studies on grammaticalization and lexicalisation of metaphors point out that although this is a slow process, it is effectively grounded in the grammar and semantics (González, 2020; Paula & Coelho, 2015; Schiffman, 1999) based on usage (Haiman, 1997).

Finally, considering the impact of educational activities, we believe that our test has provided a productive task to put into words a collective trauma. In previous studies we identified that very few consider COVID-19 consequences on adolescents, do not give them opportunities to speak on emotions, in spite that they were one of the most vulnerable populations during this pandemic (Hoskins et al., 2023; Rubbo et al., 2024). Our results have shown a higher use of metaphors that display confinement and loneliness images. We believe that the rigorous process by which we designed the test allows responses that help students to symbolise meanings for processing emotions in the post pandemics. Also, it could be used for future traumatic collective experiences.

In sum, we consider that this study has made two main contributions. Methodologically, we have observed how an ecological test —that is, one that elicits different discursive genres— allows not only to obtain significant empirical evidence on a phenomenon — the emergence of spontaneous metaphors in discourse— but also offers opportunities for students to express themselves.

Regarding metaphor development, we have taken into account the frequency of use and its different categories considering the topic, the pragmatic request and gender. The results allow us to refine the description of contextualised adolescents’ metaphors usage.

Other issues such as the specific incidence of population factors (age and socio-economic class), as well as other theoretical factors (differences in the frequency of metaphorical domains, and the discursive and grammatical characteristics of each one) remain to be investigated. Replicating this test in public schools and rural environments would show the regional impact of the pandemics. Furthermore, a longitudinal study could show us linguistic and cognitive clues about metaphoric language and emotional conceptualisation, as well as the changes or mutation of a complex psycholinguistic and sociohistorical phenomenon. Finally, we must include a stronger gender perspectivity to better understand gender differences.

This work was supported by the National Scientific and Technical Research Council - Argentina under Grants RESOL-2021-2360-APN-DIR#CONICET and PIPKS0911220210100577CO; and the National Agency for the Promotion of Research, Technological Development and Innovation (Agencia Nacional de Promoción de la Investigación, el Desarrollo Tecnológico y la Innovación, Agencia I+D+i) under Grant PICT GRF T1 TI 00207.

As social researchers, we design our studies bearing in mind that every step of the research process involves epistemological decisions and, consequently, our lens on the data. By this, from the very first we clarify and debate why we decide to carry out the research. In this case, we should explain that we study Argentine adolescents’ language performance because all the co-authors are Argentine scholars specialised in language development and native Spanish speakers. One of us lives in the neighbourhood of the school, close to the community where the data was collected, thereby offering a valuable insight into the characteristics of the population. All authors worked together as a team, engaging in regular discussions to ensure that the study was guided by their collective cultural knowledge and expertise. This was a collaborative team project that was undertaken to guarantee that the study was both sensitive and appropriate to the context in which it was conducted.

In accordance with our research ethical commitment, we inform that María Florencia Tenaglia was employed by the Secondary school where the research was carried out. Due to this we displayed a protocol for dealing any potential conflicts arising. The other co-authors have no conflicts of interest to disclose.

The data that support the findings of this study are openly available at the CONICET Digital Institutional Repository http://hdl.handle.net/11336/252421