Introduction

According to the neurocognitive model of literary reading (Jacobs, 2011Jacobs, A. M. (2011) Neurokognitive Poetik: Elemente eines Modells des literarischen Lesens (Neurocognitive poetics: elements of a model of literary reading). In: Schrott, R. and Jacobs, A. M. (Eds.) Gehirn und Gedicht: Wie wir unsere Wirklichkeiten konstruieren (Brain and Poetry: How We Construct Our Realities). Carl Hanser Verlag, 492–520.), a text is composed of background and foreground. Background consists of familiar words that enhance fiction feeling and absorption into the story world, and foreground encompasses tropes (i.e., figurative language) that give an aesthetic feeling to the reader. While the former facilitates fluent reading (in eye-tracking experiments it is shown by e.g., short fixations, long and regular saccades, low affect rating); the latter slows down reading (e.g., long fixations, irregular short saccades, high affect ratings). Thus, figurative language requires more time to comprehend but elicits aesthetic appreciation. Our aim is to find out whether this holds true during second language (L2) processing. For the purpose of our study, we confined figurative language to metaphors because they are often used in literary texts and, to a large extent, constitute the very literariness of the work. In the scope of the paper, we aim to answer the following questions:

1. How do we perceive metaphors in L2?

2. Is the comprehension of metaphors in L2 cognitively demanding?

3. Do people understand literary metaphors in a foreign language?

4. Are literary metaphors aesthetically appreciated by L2 speakers?

By answering these questions, we will try to gain insight into L2 literary reading and the perception of figurative language in a foreign language. We will explore literary metaphors embedded in a literary text, which is a novel approach, especially in L2 research. Furthermore, studying the aesthetic appreciation of metaphors by non-native speakers is also novel to the best of our knowledge.

Theoretical accounts of metaphor comprehension

In metaphorical expressions, words yield non-literal meanings (Searle, 1993Searle, J. R. (1993) Metaphor. In: Ortony, A. (Ed.) Metaphor and Thought ( 2 ). Cambridge University Press, 83–111. ). The notion of metaphor can be explained through integration, that is when separate words are integrated to elicit more than the sum of the parts (Paivio & Walsh, 1993Paivio, A., & Walsh, M. (1993) Psychological processes in metaphor comprehension and memory. In: Ortony, A. (Ed.) Metaphor and Thought. Cambridge University Press, 2–307. ; Ortony, 1993Ortony, A. E. (1993) Metaphor and thought. Cambridge University Press. ). The term “metaphor” originates from Greek and means “transfer”. Metaphors transfer insight from the source (or base/ground) domain to the target (or topic) domain. According to the directional projection view, the meaning of the clear source domain is projected onto the less clear target domain (Wolff and Gentner, 2011Wolff, P., & Gentner, D. (2011) Structure-mapping in metaphor comprehension. Cognitive science 35(8), 1456–1488. ). For example, in the following nominal metaphor “some lawyers (topic) are sharks (vehicle)” (Ikuta and Miwa, 2021Ikuta, M., & Miwa, K. (2021) Structure mapping in second-language metaphor processing. Metaphor and Symbol 36(4), 288–310. ), the meaning of “sharks” is projected onto “lawyers”. Some theories of metaphor highlight the emergent commonalities view, according to which “a metaphor is a major vehicle through which words acquire new or broader meaning” (Hock & Joseph, 1996Hock, H. H., & Joseph, B. D. (1996) Language history, language change and language relationship: An introduction to historical and comparative linguistics. Mouton de Gruyter.: 228). In this paper, we will focus on the literary metaphor, which is a metaphor used in literary texts. It is believed that literary metaphors (e.g., “the curves of your lips rewrite history”) are richer than, for example, scientific metaphors (e.g., “energy flows through an ecosystem”) because the former has an expressive function whilst the latter has an explanatory function (Gentner, 1982Gentner, D. (1982) Are scientific analogies metaphors? In: Miall, D. S. (Eds.) Metaphor: Problems and Perspectives. Harvester Press, 106–132.; Steen, 1992Steen, G. (1992) Literary and nonliterary aspects of metaphor. Poetics Today 13(4), 687–704. ).

In the realm of cognitive science, metaphor can be explained by the conceptual blending theory (Fauconnier & Turner, 2002Fauconnier, G., & Turner, M. (2002) The way we think: Conceptual blending and the mind’s hidden complexities. Basic Books.). The emergent structure of conceptual blending describes the construction of metaphors. One of the most famous examples is “That surgeon is a butcher”. To comprehend this metaphor, we create two input mental spaces: one is for “surgeon” and another one is for “butcher”. The former contains such input elements as “healing” and “scalpel” whereas the latter includes “cleaver” and “severing”. By blending these two mental spaces, a new emergent meaning, “an incompetent, careless surgeon”, is created. However, if we take into account the directional view, then a category derived from the source domain “butcher” is applied to the target domain “surgeon”. The combination of emergent commonalities and directional view is closely related to the structure mapping process where the processing of metaphor involves finding commonalities and then projecting inferences (Wolff & Gentner, 2011Wolff, P., & Gentner, D. (2011) Structure-mapping in metaphor comprehension. Cognitive science 35(8), 1456–1488. ). In Lakoff and Johnson’s words, “the essence of metaphor is understanding and experiencing one kind of thing in terms of another” (Lakoff & Johnson, 1980Lakoff, G., & Johnson, M. (1980) Metaphors we live by. University of Chicago Press.: 5). Conceptual metaphor theory (Lakoff & Johnson, 1980Lakoff, G., & Johnson, M. (1980) Metaphors we live by. University of Chicago Press.) postulates that we understand the target domain through another concrete conceptual domain (or source domain). For example, during the comprehension of the conceptual metaphor “love is a journey”, we systematically organize information about “journey”, and then this knowledge helps us to comprehend “love”. In other words, the word “love” is conceptualized in terms of “journey”, lovers as travelers and relationships as roads (Ferreira, 2008Ferreira, L. C. (2008) A psycholinguistic study on metaphor comprehension in a foreign language. Revista Virtual de Estudos da Linguagem, 6(11), 1–23.). Taking into account the view of embodied cognition, embodied experience reflects our perception, actions and manipulation of objects systematically structuring linguistic behavior. These patterns are also known as “experiential gestalts” or “image schemas” which are constructed as a result of people’s relations manipulating objects or orienting in space and time (Johnson, 1987Johnson, M. (1987) The body in the mind: The bodily basis of meaning, imagination and reason. University of Chicago Press. ). Hence, such notions as “journey” are conventionally comprehended since they are connected with people’s embodied experiences.

Studies on the comprehension of metaphors in literature date back to the end of the twentieth century (e.g., Glucksberg et al., 1982Glucksberg, S., Gildea, P., & Bookin, H. B. (1982) On understanding nonliteral speech: Can people ignore metaphors? Journal of verbal learning and verbal behavior 21(1), 85–98. ), and we can identify two beliefs regarding the processing of metaphors; one argued that metaphors are processed the same way as literal sentences (e.g., Gibbs, 1984Gibbs, R. W. Jr. (1984) Literal meaning and psychological theory. Cognitive Science 8, 275–304. , 1987Gibbs, R. W. Jr. (1987) What does it mean to say that a metaphor has been understood? In: Haskell, R. (Eds.) Cognition and symbolic structures: The psychology of metaphoric transformations. Ablex, 31–48., 1994Gibbs, R. W. (1994) The poetics of mind. Cambridge University Press., 2001Gibbs Jr, R. W. (2001) Evaluating contemporary models of figurative language understanding. Metaphor and symbol, 16(3–4), 317–333. ; Blank, 1988Blank, G. D. (1988) Metaphors in the lexicon. Metaphor and Symbol, 3(3), 21–36. ; Way, 1991Way, E. C. (1991) Knowledge representation and metaphor (Vol. 7). Springer Science & Business Media. ; Cacciari & Glucksberg, 1994Cacciari, C., & Glucksberg, S. (1994) Understanding figurative language. In: Gernsbacher, M. A. (Eds.) Handbook of psycholinguistics. Academic Press, 447–477.; Goldvarg & Glucksberg, 1998Goldvarg, Y., & Glucksberg, S. (1998) Conceptual combinations: The role of similarity. Metaphor and Symbol 13(4), 243–255. ), while the other asserted that they need more time to be processed (e.g., Clark & Lucy, 1975Clark, H., & Lucy, P. (1975) Understanding what is meant from what is said: A study in conversationally conveyed requests. Journal of Verbal Learning and Verbal Behavior, 12, 335–359. ; Ortony et al., 1978Ortony, A., Schallert, D. L., Reynolds, R. E., & Antos, S. J. (1978) Interpreting metaphors and idioms: Some effects of context on comprehension. Journal of verbal learning and verbal behavior 17(4), 465–477. ; Gibbs, 1981Gibbs, R. W. Jr. (1981) Your wish is my command: Convention and context in interpreting indirect requests. Journal of Verbal Learning and Verbal Behavior 20, 435–444. ; Inhoff et al., 1984Inhoff, A. W., Lima, S. D., & Carroll, P. J. (1984) Contextual effects on metaphor comprehension in reading. Memory & Cognition 12(6),558–567. ; Janus & Bever, 1985Janus, R. A., & Bever, T. G. (1985) Processing of metaphoric language: An investigation of the three-stage model of metaphor comprehension. Journal of Psycholinguistic Research 14(5), 473–487. ). A more recent eye-movement study proposed that verbs with metaphorical meaning take longer to process than do the same verbs with literal meaning (Columbus et al., 2015Columbus, G., Sheikh, N. A., Côté-Lecaldare, M., Häuser, K., Baum, S. R., & Titone, D. (2015) Individual differences in executive control relate to metaphor processing: An eye movement study of sentence reading. Frontiers in human neuroscience, 8, 1–12. ). The latter view had an impact on the development of the three-stage model of nonliteral processing (Clark & Lucy, 1975Clark, H., & Lucy, P. (1975) Understanding what is meant from what is said: A study in conversationally conveyed requests. Journal of Verbal Learning and Verbal Behavior, 12, 335–359. ; Grice, 1975Grice, H. (1975) Logic and conversation. In: Cole, P. and Morgan, J. L. (Eds.) Syntax and semantics: Speech acts. Seminar Press, 41–58. ; Searle, 1979Searle, J. R. (1979) Metaphor. In: Ortony, A. (Ed.) Metaphor and thought. Cambridge University Press, 92–123.; Glucksberg, 2001Glucksberg, S. (2001) Understanding figurative language: From metaphors to idioms. Oxford University Press. ). According to this model (i.e., serial processing), the processing of metaphors requires the implementation of three stages. First, the word/phrase is interpreted literally. Second, the plausibility of the literal interpretation is assessed based on the context. Third, this interpretation is rejected because of its contextual inappropriateness, and a new metaphoric meaning is created (Clark & Lucy, 1975Clark, H., & Lucy, P. (1975) Understanding what is meant from what is said: A study in conversationally conveyed requests. Journal of Verbal Learning and Verbal Behavior, 12, 335–359. ; Grice, 1975Grice, H. (1975) Logic and conversation. In: Cole, P. and Morgan, J. L. (Eds.) Syntax and semantics: Speech acts. Seminar Press, 41–58. ; Searle, 1979Searle, J. R. (1979) Metaphor. In: Ortony, A. (Ed.) Metaphor and thought. Cambridge University Press, 92–123.). Janus and Bever (1985)Janus, R. A., & Bever, T. G. (1985) Processing of metaphoric language: An investigation of the three-stage model of metaphor comprehension. Journal of Psycholinguistic Research 14(5), 473–487. embedded the same phrase (e.g., “the fabric had begun to fray”) in one context where it had metaphorical meaning and in another where its meaning was literal and then compared the reaction time to the phrase. Their results showed that the former required more time to read than did the latter, thereby supporting the three-stage model of nonliteral processing. George and Wiley (2016)George, T., & Wiley, J. (2016) Forgetting the literal: The role of inhibition in metaphor comprehension. Journal of Experimental Psychology: Learning, Memory, and Cognition 42(8), 1324–1330. debate this view, suggesting that the longer processing time is caused by the implausibility of the context or morphological component of the metaphorical expression.

Kintsch and Bowles (2002)Kintsch, W., & Bowles, A. R. (2002) Metaphor comprehension: What makes a metaphor difficult to understand? Metaphor and symbol 17(4), 249–262. attempted to explain why metaphors are difficult to comprehend. It seems that the application of the three-stage model of nonliteral processing is more valid in the case of literary metaphors than ordinary metaphors because common metaphors are easier to understand (Kintsch & Bowles, 2002Kintsch, W., & Bowles, A. R. (2002) Metaphor comprehension: What makes a metaphor difficult to understand? Metaphor and symbol 17(4), 249–262. ). Literary metaphors are less familiar and used in literary texts while conventional (or ordinary) metaphors are more familiar and used in common speech. Kintsch and Bowles (2002)Kintsch, W., & Bowles, A. R. (2002) Metaphor comprehension: What makes a metaphor difficult to understand? Metaphor and symbol 17(4), 249–262. investigated how people interpret metaphors and assess their difficulty, combining qualitative and quantitative analysis (i.e., latent semantic analysis (LSA)). “LSA represents word meaning as a vector in a high-dimensional semantic space” (Kintsch and Bowles, 2002Kintsch, W., & Bowles, A. R. (2002) Metaphor comprehension: What makes a metaphor difficult to understand? Metaphor and symbol 17(4), 249–262. : 258). The results indicated that people interpret common metaphors in the same way and it is strongly correlated with the model vector. However, literary metaphors (or more difficult metaphors) are interpreted differently and it is less strongly correlated with the model vector. According to the prediction model, one possible explanation for the difficulty in understanding metaphors is the match between the argument and predicate (Kintsch & Bowles, 2002Kintsch, W., & Bowles, A. R. (2002) Metaphor comprehension: What makes a metaphor difficult to understand? Metaphor and symbol 17(4), 249–262. ). The more the meaning of the argument is matched with the closest neighboring meaning of the predicate, the easier it is understood (e.g., “my lawyer is a shark”).

Metaphor processing in a second language

According to Kecskes (2001)Kecskes, I. (2001) The ‘graded salience hypothesis’ in second language acquisition. In: Pütz, M., Niemeier, S. and Dirven, R. (Eds.) Applied cognitive linguistics: Volume 1. Mouton de Gruyter, 249–271. , native speakers process figurative meaning directly but non-native speakers first consider the literal meaning. It follows that contextual clues play a role in understanding metaphors in L2 (Kecskes, 2001Kecskes, I. (2001) The ‘graded salience hypothesis’ in second language acquisition. In: Pütz, M., Niemeier, S. and Dirven, R. (Eds.) Applied cognitive linguistics: Volume 1. Mouton de Gruyter, 249–271. ). However, Ferreira (2008)Ferreira, L. C. (2008) A psycholinguistic study on metaphor comprehension in a foreign language. Revista Virtual de Estudos da Linguagem, 6(11), 1–23. argues that there is a universal pattern that helps to comprehend metaphors without relying on contextual information. Ferreira (2008)Ferreira, L. C. (2008) A psycholinguistic study on metaphor comprehension in a foreign language. Revista Virtual de Estudos da Linguagem, 6(11), 1–23. supported the account that metaphors are understood through the embodied experience of foreign language learners. This universal pattern is related to the experiential domains of MOTION, CONTAINER, PLACE, and VISION. Interestingly, there is no significant difference in metaphor comprehension for foreign language learners at different levels of proficiency (Ferreira, 2008Ferreira, L. C. (2008) A psycholinguistic study on metaphor comprehension in a foreign language. Revista Virtual de Estudos da Linguagem, 6(11), 1–23.).

In a few qualitative studies, it was revealed that metaphors are difficult to comprehend for non-native speakers of English during university lectures (e.g., Littlemore, 2001Littlemore, J. (2001) The use of metaphor in university lectures and the problems that it causes for overseas students. Teaching in higher education 6(3), 333–349. ; Littlemore et al., 2011Littlemore, J., Chen, P. T., Koester, A., & Barnden, J. (2011) Difficulties in metaphor comprehension faced by international students whose first language is not English. Applied linguistics 32(4), 408–429. ). This might be explained by the lack of ”native speaker competence” which includes awareness of cultural conventions, cultural connotations and figurative language repertoire (Littlemore and Low, 2006Littlemore, J., & Low, G. (2006) Metaphoric competence, second language learning, and communicative language ability. Applied linguistics 27(2), 268–294. ). The notion of “metaphoric competence” is used to describe the ability to understand and use metaphor, and it was expanded into conceptual fluency by Danesi (1993Danesi, M. (1993) Metaphorical competence in second language acquisition and second language teaching. In: Alatis, J. E. (Eds.) Georgetown University Round Table on Language and Linguistics. Georgetown University Press, 489–500.; 2003Danesi, M. (2003) Second language teaching: A view from the right side of the brain. Springer Science and Business Media. ) to include metaformal competence, reflexive competence, and associative competence. Metaformal competence is needed for the appropriate use of concepts. Reflexive competence is used to transform concepts into language categories. Associative competence is necessary to connect concepts from a cultural perspective. All three of these competencies are essential for language acquisition (Danesi, 2008Danesi, M. (2008) Conceptual errors in second-language learning. In: De Knop, S. and De Rycker, T. (Eds.) Cognitive approaches to pedagogical grammar. Mouton De Gruyter, 231–258. ). According to another view, metaphorical competence is composed of such components as metaphor production, metaphor comprehension, the ability to deduce the meaning of a metaphor and the speed at which it is done (Littlemore, 2001Littlemore, J. (2001) The use of metaphor in university lectures and the problems that it causes for overseas students. Teaching in higher education 6(3), 333–349. ). Littlemore (2001)Littlemore, J. (2001) The use of metaphor in university lectures and the problems that it causes for overseas students. Teaching in higher education 6(3), 333–349. suggested the term ”metaphoric intelligence” to denote the ability to reason analogically and think divergently. This is believed to be essential for second language acquisition, especially for learners’ communicative competence in a foreign language (Hoang, 2014Hoang, H. (2014) Metaphor and second language learning: The state of the field. The Electronic Journal for English as a Second Language 18(2), 1–27.). According to the literature, second language learners experience difficulty with conceptualization in a target language resulting in literal understanding. These are conceptual errors that second language learners make when they mistake the source domain of a metaphor because of L1 interference. During metaphor processing, L2 speakers need to not only activate the source and target domains of metaphor in a target language but also suppress the same domains from L1 (Hoang, 2014Hoang, H. (2014) Metaphor and second language learning: The state of the field. The Electronic Journal for English as a Second Language 18(2), 1–27.).

In contrast to the view that comprehending metaphorical meaning is more difficult than literal meaning, is the assumption that metaphors are comprehended faster because of emotional arousal (Ifantidou & Hatzidaki, 2019Ifantidou, E., & Hatzidaki, A. (2019) Metaphor comprehension in L2: Meaning, images and emotions. Journal of Pragmatics 149, 78–90. ). Carston (2010)Carston, R. (2010) XIII – Metaphor: Ad hoc concepts, literal meaning and mental images. Proceedings of the Aristotelian Society, 110 (3), 295–321. believes that conventional and literal metaphors are understood using different processing modes. However, in any case, the comprehension of metaphors encompasses imagery (and emotion) as well as propositional meaning (Carston, 2010Carston, R. (2010) XIII – Metaphor: Ad hoc concepts, literal meaning and mental images. Proceedings of the Aristotelian Society, 110 (3), 295–321. ). Ifantidou and Hatzidaki (2019)Ifantidou, E., & Hatzidaki, A. (2019) Metaphor comprehension in L2: Meaning, images and emotions. Journal of Pragmatics 149, 78–90. highlight that emotions together with images as much as literal meaning contribute to metaphor comprehension in a foreign language. Ifantidou (2019) proposes that the high imagery and emotions evoked by metaphors are processed automatically by foreign language learners without the need to access their literal meaning. If English as a foreign language learners have difficulty in grasping literal meaning, comprehension occurs due to the imaginative construal of the world and not the literal construal of language (Ifantidou & Hatzidaki, 2019Ifantidou, E., & Hatzidaki, A. (2019) Metaphor comprehension in L2: Meaning, images and emotions. Journal of Pragmatics 149, 78–90. ).

Ikuta and Miwa (2021)Ikuta, M., & Miwa, K. (2021) Structure mapping in second-language metaphor processing. Metaphor and Symbol 36(4), 288–310. tested the generalizability of structure mapping theory (Wolff & Gentner, 2011Wolff, P., & Gentner, D. (2011) Structure-mapping in metaphor comprehension. Cognitive science 35(8), 1456–1488. ) in bilinguals’ metaphor processing using forward metaphors (e.g., “Some babies are angels”) and reversed metaphors (e.g., “Some angels are babies”). It was revealed that for non-native speakers forward metaphors are more comprehensible than reverse metaphors, supporting structure mapping theory. However, there were some differences in processing by native and non-native speakers that might be explained by the type of metaphors; novel metaphors are understood via similarity whereas familiar metaphors are comprehended via categorization. Hence, people may rely more on the similarity of noun pairs in L2. In addition, cultural differences also play a role in processing. Interestingly, Ikuta and Miwa (2021)Ikuta, M., & Miwa, K. (2021) Structure mapping in second-language metaphor processing. Metaphor and Symbol 36(4), 288–310. also suggest that it is not essential for literal processing to occur before metaphor processing. This view was supported by Vaid, Lopez, and Martinez (2015)Vaid, J., López, B. G., & Martínez, F. E. (2015) Linking the figurative to the creative: Bilinguals’ comprehension of metaphors, jokes, and remote associates. In: Heredia, R. R. and Cieślicka, A. B. (Eds.) Bilingual figurative language processing. Cambridge University Press, 53–86. who studied bilingual participants. According to them, the metaphorical meaning in L2 is activated immediately.

However, it should be noted that there may be differences in the processing of metaphors depending on the language. A study on the recall of metaphors and similes by Harris et al. (1999)Harris, R. J., Tebbe, M. R., Leka, G. E., Garcia, R. C., & Erramouspe, R. (1999) Monolingual and bilingual memory for English and Spanish metaphors and similes. Metaphor and Symbol, 14 (1), 1–16. demonstrates these dependencies. They found out that there is a preference for recalling both similes and metaphors in English as metaphors, whereas the reverse can be observed in Spanish. These results support the outcomes from the previous study by Johnson (1996)Johnson, A.T. (1996) Comprehension of metaphors and similes: A reaction time study. Metaphor and Symbolic Activity, 11, 145–159. and might be explained as “metaphors being more common or stylistically preferred in English and similes being more so in Spanish” (Harris et al., 2009: 14). However, the replication of studies by bilinguals showed that English metaphor preference did not maintain whereas Spanish simile preference did, suggesting the latter is a stronger effect. In a similar vein, it has been also found that distinct mechanisms are involved in processing metaphors in two entirely different languages such as Chinese and English: English metaphors require the activation of both hemispheres unlike Chinese metaphors and English/Chinese literal sentences (Chen, Peng & Zhao, 2013Chen, H., Peng, X., & Zhao, Y. (2013) An ERP study on metaphor comprehension in the bilingual brain. Chinese Journal of Applied Linguistics, 36 (4), 505–517. ). In another recent study (EEG) focusing on bilingualism, Jankowiak, Naranowicz and Rataj (2021)Jankowiak, K., Naranowicz, M., & Rataj, K. (2021) Metaphors are like lenses: Electrophysiological correlates of novel meaning processing in bilingualism. International Journal of Bilingualism, 25 (3), 668–686. revealed that cognitive processing of metaphors and similes in an individual’s first language (L1) and a highly proficient second language (L2) remains the same, and the comparison mechanism plays a pivotal role in advancing the comprehension of novel nominal metaphors, with a heightened significance observed in L2 during the phase of meaning integration. On the contrary, in the most recent electroencephalography (EEG) study conducted by Wang and Jankowiak (2023)Wang, X., & Jankowiak, K. (2023) Electrophysiological insights into the role of proficiency in bilingual novel and conventional metaphor processing. Linguistic Approaches to Bilingualism, 13 (2), 163–189. , disparities were observed in the processing of novel and conventional metaphors between L1 and L2, regardless of proficiency level. The difference emerged between intermediate and advanced proficiency levels in the duration of lexico-semantic access. Furthermore, additional cognitive processing costs in L2 were elucidated in an event-related potential (ERP) study investigating the processing of scientific metaphors in Chinese and English by Tang et al. (2022)Tang, X., Shen, L., Yang, P., Huang, Y., Huang, S., Huang, M., & Ren, W. (2022) Bilingual processing mechanisms of scientific metaphors and conventional metaphors: evidence via a contrastive event-related potentials study. Frontiers in Psychology, 13 , 894114. . This phenomenon was attributed to a reduction in cognitive automaticity and sensitivity in the L2 context.

Thus, the results of current studies on metaphor processing by bilinguals are far from conclusive. Nonetheless, a recurring pattern suggests additional cognitive costs in the processing of novel metaphors. Nevertheless, the existing body of research fails to inform us about L2 learners’ ability to comprehend literary metaphors.

Eye-tracking experiments on metaphor processing. State of the art

Despite the prevalence of metaphors in literature, relatively few eye-tracking studies have been carried out to accurately measure cognitive effort during their processing. Columbus et al. (2015)Columbus, G., Sheikh, N. A., Côté-Lecaldare, M., Häuser, K., Baum, S. R., & Titone, D. (2015) Individual differences in executive control relate to metaphor processing: An eye movement study of sentence reading. Frontiers in human neuroscience, 8, 1–12. studied the effect of executive control in L1 metaphor processing using an eye-tracking method, and the results suggest that people read verbs with metaphorical meaning more slowly than verbs with literal meaning. However, this difference was less pronounced for more familiar metaphors. Additionally, readers with higher executive control capacities read metaphors faster provided there was a prior context (Columbus et al., 2015Columbus, G., Sheikh, N. A., Côté-Lecaldare, M., Häuser, K., Baum, S. R., & Titone, D. (2015) Individual differences in executive control relate to metaphor processing: An eye movement study of sentence reading. Frontiers in human neuroscience, 8, 1–12. ).

Akkök and Uzun (2018)Akkök, E. A., & Uzun, I. P. B. (2018) Metaphor processing in Turkish: An eye-movement study. Dil ve Edebiyat Dergisi, 15(1), 105–124. conducted an eye-tracking study to reveal the discrepancies between the processing of literal sentences and metaphoric sentences in Turkish as L1. In particular, they focused on the prototypicality effect in the former and the familiarity effect in the latter. According to their results, the processing of metaphorical sentences is affected by the familiarity effect rather than the processing of literal sentences being affected by the prototypicality effect. It was interpreted in the scope of the literal first hypothesis, according to which, the processing of metaphorical meaning is preceded by the activation of literal meaning.

Another eye-tracking study on metaphors was conducted by Binghan and Hao (2018)Binghan, Z., & Hao, Z. (2018) Revisiting processing time for metaphorical expressions: An eye-tracking study on eye-voice span during sight translation. Foreign Language Teaching and Research (bimonthly), 50(5), 744–759. during sight translation (from English L2 to Mandarin L1). They concluded that each word in a metaphorical expression requires more processing time than does the whole text in general. Thus, metaphorical expressions are more cognitively demanding than are literal expressions during sight translation. This was also shown in a study by Płużyczka (2015)Płużyczka, M. (2015) Tłumaczenie a vista: rozważania teoretyczne i badania eyetrackingowe (Eng.: Sight translation. Theoretical considerations and eye-tracking research). Wydawnictwo Naukowe Instytutu Komunikacji Specjalistycznej i Interkulturowej. Uniwersytet Warszawski.. In sight translation, the areas on which the greatest fixations were recorded were metaphors. Also, the paragraph that contained metaphors registered the highest number of fixations, the longest fixation, and the highest number of revisits. This suggested that it was the most cognitively demanding for test subjects, i.e., it involved the highest cognitive load.

Olkoniemi et al. (2022)Olkoniemi, H., Bertram, R., & Kaakinen, J. K. (2022) Knowledge is a river and education is like a stairway: An eye movement study on how L2 speakers process metaphors and similes. Bilingualism: Language and Cognition 25(2), 307–320. conducted a comparative eye-tracking study to explore the difference between the processing of nominal metaphors (“education is a stairway”) and similes (“education is like a stairway”) in English as L2 by Finnish L1 speakers. It was found that people have more regressions while reading metaphors than they do reading similes. Hence, the results suggest that metaphor processing requires more effort than does simile processing. In addition, the familiarity and L2 proficiency effects are highlighted as modulating factors in metaphor processing.

Despite the scarcity of eye-tracking studies on metaphor processing from different domains, we see a pattern that shows processing costs associated with metaphors. Based on the existing research, we can hypothesize that the processing of literary metaphors will be more cognitively demanding for English L2 speakers than will the processing of the simplified version of metaphors written in plain language.

Experimental design

Participants

Twenty-eight participants (8 males and 20 females) were recruited for this study. Participants were between the ages of 18 and 34 (M _age = 22.9 years, SD _age = 2.94 years). Of all the people recruited, 22 qualified for analysis. We discarded test subjects with excessive visual impairment and recordings that highlighted some technical problems or inaccurate calibration. Since we intended to correlate the eye-tracking results with the results of the comprehension test and the appreciation test, we had to remove these subjects from other stages of the experiment. All participants were students of the Institute of Specialized and Intercultural Communication (Faculty of Applied Linguistics) at the University of Warsaw. They are studying English as their second language (level B2 and higher) and Polish is their native language.

In the experimental group (which read the original text with metaphors), there were 11 test subjects (3 males and 8 females). In the control group (which read the simplified version of the text without metaphors), there were 11 test subjects (4 males and 7 females). We did not adjust the groups for gender because earlier research (Płużyczka, 2013Płużyczka, M. (2013) Eye-tracking research into sight translation processes. Lapsological conclusions. In: Grucza, S., Płużyczka, M., & Zając, J. (Eds.) Translation Studies and Eye-Tracking Analysis (= Warschauer Studien zur Germanistik und Angewandten Linguistik). Peter Lang Verlag, 105–138.) had shown that gender did not correlate with the results of eye-tracking experiments.

All participants gave written informed consent before the experiment. All the data pertaining to participants had been anonymized before processing.

Materials

The text used in the experimental group was an excerpt from Chapter 20 of “The Picture of Dorian Gray” written by Oscar Wilde (see Appendix 1). The text consists of 403 words (2130 characters). The text was divided into four parts presented as different stimuli. This was done to make the results of the eye-tracking experiment as accurate as possible. There were 16 different kinds of metaphors in the selected excerpt; we chose a passage that contained a large number of metaphors to see how they would affect the subjects’ cognitive effort.

The text used in the control group was the simplified version of the text used in the experimental group. All 16 metaphors have been reformulated into non-metaphorical expressions. After we adapted the text replacing the metaphors with plain language for the purpose of the experiment, the text was sent to three independent experts (two linguists and one literary scholar) in the field of English. One of them is a native English speaker, the second is bilingual (L1: English, Polish), and the third is a specialist in teaching English as well as a translator of literary texts. After considering their suggestions and revisions, the text was approved by all three experts.

The text consisted of 2137 characters including spaces and the modified text consisted of 2073 characters. The selected excerpt has 406 total words and 207 unique word forms. The overall vocabulary density was found to be 0.510 and a readability index of 7.639. The average number of words per sentence was 17.7. Each participant read only one version of the text snippet, the assignment of which was random. Both versions are included in the appendix.

Experimental settings

The experiment was carried out in the Laboratory of Experimental Eye Tracking Linguistics22. http://​www​.lelo​.uw​.edu​.pl/ in the Institute of Specialized and Intercultural Communication at the Faculty of Applied Linguistics of the University of Warsaw. The Laboratory is located in a separate, acoustically isolated room. During the study, only the experimenter and individual participants were present in the room. The participants sat in front of the monitor at a distance of approximately 60 cm which was an optimal eye distance for the eye tracker that was used. There was no need to use a chin rest for head stabilization because the equipment had a high tolerance for participant head movements.

Apparatus

Eye movements were recorded with the remote eye tracker SMI RED 500 mounted under the monitor. The sampling frequency was maintained at 250 Hz. This sampling rate made the results very accurate and clear when presented on the scan paths in earlier experiments (Płużyczka, 2015Płużyczka, M. (2015) Tłumaczenie a vista: rozważania teoretyczne i badania eyetrackingowe (Eng.: Sight translation. Theoretical considerations and eye-tracking research). Wydawnictwo Naukowe Instytutu Komunikacji Specjalistycznej i Interkulturowej. Uniwersytet Warszawski.). It was, therefore, sufficient for this experiment, especially since the text was displayed in single paragraphs. The stimuli were presented on a 22-inch LCD screen. Participants were seated comfortably at a distance of 60 cm from the screen.

Procedure

Upon arrival at the laboratory, participants were informed that the experiment assessed reading for comprehension. At the beginning of the experiment, the function and use of the eye tracker were introduced, and the procedure of the experiment was explained. Participants also signed a consent form. Before the reading task, the eye tracker was set up and calibrated using a 5-point calibration screen. Participants read the text presented on a computer screen (font: Courier New, font size: 14, line height: 2.5). Participants were also instructed to read the text presented on the screen at their own pace and were told to press the spacebar on the keyboard to go to the next paragraph of the text. A total of four paragraphs were presented to each participant.

The reading task was followed by a comprehension test on a different screen. Additionally, each participant identified the specific metaphors in the text and indicated what they felt was the intended meaning. At the end of the experiment, each participant also filled in an appreciation test to assess the aesthetic values of metaphors as well as their equivalent simplified expressions. The appreciation test used the 5-point Likert scale. The whole experiment took 25–30 minutes for each participant to complete.

When analyzing research data in each paragraph (stimulus) the specified areas of interest (AOIs) were marked. These were metaphors in the text for the experimental group and the simplified expressions for the control group. Then, for AOIs, the values of the tested eye-tracking parameters were generated and extracted.

For the analysis, we chose eye-tracking parameters which have proven in previous studies to be the most universal and indicative of the cognitive load of the subjects. As shown by Płużyczka (2015Płużyczka, M. (2015) Tłumaczenie a vista: rozważania teoretyczne i badania eyetrackingowe (Eng.: Sight translation. Theoretical considerations and eye-tracking research). Wydawnictwo Naukowe Instytutu Komunikacji Specjalistycznej i Interkulturowej. Uniwersytet Warszawski., 2020Płużyczka, M. (2020) Tracking mental processes in sight translation. Neurobiological determinants of selected eye-tracking parameters. Translation, Cognition & Behaviour 3(2), 209–232. ), these were dwell time, fixation count, and revisits count.

Data analysis

The data was analyzed using IBM SPSS V29. Analyses were conducted for selected eye-tracking parameters: dwell time, fixation count, fixation time and revisits. The scores from comprehension and appreciation ratings were analyzed across two groups as well. First, the selected parameters were averaged by items, i.e., the averages of values were calculated per each item across all participants. In the analysis by items, statistical tests are conducted across all items (n = 16). Then, averaging by subjects was performed, i.e., the averages of values were calculated per each subject across all items. In the analysis by subjects, statistical tests are conducted across all subjects (n = 11). To see the overall effect of the group on eye-tracking parameters, MANOVA was performed followed by independent samples t-tests. Independent samples t-tests were also run with comprehension scores and appreciation ratings. Spearman correlation was conducted between dependent variables to see possible relationships. Before conducting statistical tests, assumptions were checked, and if needed transformation was performed to run parametric tests. The data was analyzed by items and by subjects to receive reliable results on how metaphors are processed by English as a second language speakers. Such a two-perspective analysis allowed us also to demonstrate the differences between the results specific to the foregrounding element – metaphor – and the results depended on the group of subjects who carried out this task. In the eye-tracking results, we included results on how one person processed one item on average (in the table indicated as mean), which best illustrates how we process a metaphor. These results will be the same regardless of whether it is an analysis by items or by subjects, because it is the mean of the averages, i.e. the difference will only be in the standard deviation. In addition, we have included the mean of sum, which indicates in the by-items analysis how the whole group processed one item on average, and in the by-subjects analysis how one subject processed all metaphors on average, which already gives us different results in these two types of analysis.

Analysis by items. Eye-tracking data

Descriptive statistics summarizes the mean and standard deviation of eye-tracking parameters such as dwell time, fixation count, fixation time and revisits across all items (see Table 1). In the analysis by items, MetR refers to metaphors and SimMetR can be regarded as their simplified versions. It can be noted from the table that the measures depicted are higher for all parameters in the experimental condition in comparison to the control condition.

First, the eye-tracking parameters were checked for assumptions of normality and homogeneity of variance. As eye-tracking parameters did not pass, they were transformed using Lg10 to perform parametric tests. To check the overall effect of the group (MetR vs SimMetR) on the eye-tracking parameters, MANOVA was performed with Lg10 transformed dependent variables that are highly correlated (fixation count, fixation time, and dwell time). The MANOVA test revealed a significant effect of group on the combined Lg10 transformed dependent variables (fixation count, fixation time, dwell time), Pillai’s trace = .245, F(3.000, 28.000) = 3.022, p < 0.05. Therefore, the individual eye-tracking parameters were examined further using independent samples t-tests.

Dwell time

Mean comparison between groups shows that metaphorical expressions (MetR) were read slower, one test subject read one metaphor on average 968.83 milliseconds (ms), while in the SimMetR group, one subject read the simplified version on average 632.45 ms. The result of the independent samples t-test confirmed that the difference is statistically significant, t(30) = 3.000, p < 0.05, two-sided. The statistical test showed a large effect size Cohen’s d = 1.061 95% CI[.310, 1.796]. These results demonstrate that the processing of metaphors needs more dwell time than the processing of their counterparts rewritten in plain language.

As for the total time (of the entire group) per metaphor, it was 10657.09 milliseconds on average, i.e. 10.65 seconds. One simplified version registered a total time of 6956.97 milliseconds on average, i.e. 6,95 seconds of dwell time across the SimMetR group.

Fixation count

The comparison between groups shows that metaphorical items (M = 4.07, SD = 1.61) were fixated more than non-metaphorical items (M = 2.78, SD = 1.28). The result of the independent samples t-test confirmed that the difference is statistically significant, t(30) = 2.869, p < 0.05, two-sided. The statistical test showed a large effect size Cohen’s d = 1.014 95% CI[.268, 1.745]. Hence, metaphors are indeed more fixated rather than their simplified versions.

As for the total count, one metaphor had a sum of 44.75 fixations on average and one simplified version had a sum of 30.56 fixations on average.

Fixation time

Outcomes show that metaphors (M = 866.94 ms, SD = 367.78 ms) had more fixation time rather than their simplified versions (M = 575.21 ms, SD = 277.43 ms). The result of the independent samples t-test confirmed that the difference is statistically significant, t(30) = 2.953, p < 0.05, two-sided. The statistical test showed a large effect size Cohen’s d = 1.044 95% CI[.295, 1.778]. Hence, the processing of metaphors needs more fixation time rather than the processing of their counterparts rewritten in plain language.

One metaphor had a sum of 9536.4 milliseconds of fixation time on average for the MetR group and one simplified version had a sum of 6327.313 milliseconds of fixation time on average for the SimMetR group.

Revisits

More revisits (M = .86, SD = .35) are observed for metaphors, whereas less revisits (M = .53, SD = .33) for their simplified counterparts. The result of the independent samples t-test with Lg10 transformation confirmed that the difference is statistically significant t(30) = 2.923, p <  0.05, two-sided. It is associated with a large effect size Cohen’s d = 1.034 95%CI[.286, 1.766].

One metaphor had a sum of 9.44 revisits on average for the MetR group and one simplified version had a sum of 5.81 revisits on average for the SimMetR group. Hence, metaphors are revisited more often rather than their simplified versions.

Exploration of individual AOIs on text with metaphors revealed that the AOI “his beauty had been to him but a mask” had the highest revisit of 1.54 followed by the AOI “made of ivory and gold” which produced 1.45 revisits. Such AOIs as “the curves of your lips rewrite history” and “worn its livery” produced the same number of 1.18 revisits. The AOI “sickly thoughts” had the least number of revisits – 0.27. Similarly, in the text with the simplified versions of metaphors, the AOI “his beauty had concealed his bad deeds” produced the highest number of revisits – 1.27 followed by “your beauty has a life changing potency” that revisited 1.18 times. The AOI that produced the lowest revisits of 0.09 times was “guilt-free”.

Thus, the effect of the items (MetR vs SimMetR) is observed on all parameters that have been considered suggesting that the processing of metaphorical expressions is different from the processing of non-metaphorical expressions. In all parameters analysed, values were significantly higher for MetR than for SimMetR. As the eye-tracking parameters taken into account are reported in the literature as indicators of increased mental effort and thus cognitive load, the results obtained may therefore suggest that the processing of metaphorical expressions is cognitively demanding.

Analysis by items. Comprehension test

The diagram shows the comprehension scores for the two groups (see Figure 1). The comprehension test consisted of 15 questions containing one metaphor each, and one question (question 2) containing two (i.e., he had filled his mind with corruption and given horror to his fancy). The scores are significantly higher for the SimMetR group for all the questions except the 8th (According to the writer of the letter, what role does his appearance play?) where the scores were 100%. We assume that this is because this question is straightforward and easy to answer (see Appendix 2 for the list of comprehension questions). The increased comprehension scores for questions on simplified expressions can be accredited to the reduced cognitive load required to process and understand them.

The descriptive statistics for comprehension test by items shows the mean and standard deviation of comprehension scores across items (see Table 2). The maximum score per item is 11 as there are 11 participants in total. 61.36% of participants from the experimental group and 86.90% of participants from the control group understood the expressions (n = 16) correctly.

As can be observed, metaphors were comprehended less (M = 6.75, SD = 2.21) rather than their simplified versions (M = 9.56, SD = 1.82). The data passed the assumptions of normality and homogeneity of variance, and independent samples t-test was conducted. It revealed a significant difference between two groups t(30) = −3.930, p < .001, two-sided, 95%CI[−4.274, −1.351] with large effect size Cohen’s d = −1.389, 95%CI[−2.156, −.604]. Hence, a smaller number of participants comprehend metaphors compared to the number of participants that comprehend their simplified version. As the results of comprehension scores are low, MANCOVA and ANCOVA were ran to check whether comprehension as covariate affects the significant difference between the two groups in eye-tracking parameters. MANCOVA was run with the group as an independent variable, comprehension as a covariate and Lg10 transformed eye-tracking parameters as dependent variables. The results of multivariate tests show that group affects eye-tracking parameters even after controlling for comprehension Pillai’s Trace = .360, F(3.000, 27.000) = 5.057, p < .05, η² = .360. Tests of between-subjects effects show that both factors such as group and comprehension do affect eye-tracking parameters p < .05. Also separately, comprehension affects eye-tracking parameters p < .05 as well as the group itself p < .05. ANCOVA results with revisits as an independent variable also confirmed this pattern: overall model p < .05, comprehension p < .05 and group p < .001. Thus, even if comprehension might affect eye-tracking parameters when considered alone and with respect to a group, the significant difference between the two groups might not be solely connected to comprehension difficulties, and there could be other factors influencing the difference in processing.

Analysis by items. Appreciation test

On the appreciation test, participants were asked to assess the author’s use of language based on their impressions (see Appendix 3). The appreciation ratings of all participants were aggregated to obtain their average ratings for the metaphors and their equivalent simplified expressions. It was noted that the average appreciation ratings ranged from 3.55 to 4.73 with an average rating of 3.9 for the metaphors. For the simplified expressions, the average rating was found to be 3.35 and ranged between 1.91 to 4.55. Figure 4 illustrates the average ratings for the individual metaphors across both groups.

The metaphor “the curves of your lips rewrite history” had the highest appreciation rating of 4.7 for all the participants while the metaphors “a monstrous moment” and “shallow moods” had the lowest appreciation score of 3.5. The appreciation scores for simplified versions were relatively lower than those for metaphors, with “your beauty has a life-changing potency” obtaining the highest appreciation score of 4.5 and “looking young” receiving the lowest appreciation rating of 1.9. Overall, metaphors were rated higher for aesthetic appeal than were their simplified counterparts. The diagram shows how 16 metaphors and their simplified versions were aesthetically appreciated using a 5-point Likert scale measurement, with 5 indicating the highest aesthetical appreciation (see Figure 2).

The descriptive statistics shows the mean and standard deviation of each group for appreciation ratings across items (see Table 3). From the table, it can be observed that metaphorical expressions are more appreciated (M = 3.90, SD = .35) rather than non-metaphorical expressions (M = 3.35, SD = .83).

The data passed the assumption of normality but not the assumption of homogeneity of variance, and therefore equal variances not assumed are reported from t-test: t(20.086) = 2.478, p < .05, two-sided 95%CI[.08831, 1.02533]. It shows a large effect size Cohen’s d = .876, 95% CI[.142, 1.597]. Hence, metaphors are more appreciated than their counterparts rewritten in plain language.

Analysis by items. Correlations

Due to violations of Pearson correlation assumptions, Spearman correlations were performed. Significant correlations were found between appreciation ratings and all eye-tracking parameters such as dwell time r = .380, p < 0.05, fixation count r = .454, p < 0.05, fixation time r = .388, p < 0.05, revisits r = .382, p < 0.05. It follows that those metaphorical expressions that were read and processed longer, were revisited more frequently and on which more fixations were recorded were also considered to be more aesthetically appreciated by test subjects. Although correlation shows neither causation nor directionality of correlation, it suggests that the increased cognitive load is associated with the increased appreciation. That is, the longer and more metaphors were cognitively processed, as evidenced by the eye-tracking parameter values, the more they were appreciated. This is a very interesting positive correlation and would need to be tested in the next stage of the research.

Analysis by subjects. Eye-tracking data

The descriptive statistics includes eye-tracking parameters across all subjects (see Table 4). It includes mean, standard deviation as well as group information. The MetR group read metaphors, whereas the SimMetR group read their simplified versions. Based on the table, it can be observed that the values for the experimental group (MetR) are higher than for the control group (SimMetR) for all eye-tracking parameters such as dwell time, fixation count, fixation time and revisits. Here the mean will be the same as for the analysis by items, as it is the average for one participant across all items, only it will differ in the SD score, as this time we are considering from the subjects’ perspective. On the other hand, the mean of the total score will differ, i.e. the results of how on average one subject fixated on all metaphors and simplified expressions corresponding to them.

As can be seen from the table above, the experimental group shows higher values for all eye-tracking parameters compared to the control group. First, the eye-tracking parameters were checked for assumptions of normality and homogeneity of variance. All eye-tracking parameters including dwell time, fixation count, fixation time and revisits passed the assumptions and therefore were used as untransformed. To check the overall effect of the group (MetR vs SimMetR) on the eye-tracking parameters, MANOVA was performed with three highly correlated eye-tracking parameters such as dwell time, fixation count and fixation time. The MANOVA test did not find a significant effect of group on the combined dependent variables, Pillai’s trace = .272, F(3.000, 18.000) = 2.246, p = .118. However, as partial eta squared showed a moderate effect size η² = .272 further examination of the dependent variables individually was performed.

Dwell time

Mean comparison between groups shows that metaphorical expressions (MetR) were read slower (M = 968.83 ms, SD = 290.95 ms) rather than non-metaphorical expressions (SimMetR) that were read comparatively faster (M = 632.45 ms, SD = 423.55 ms). The result of the independent samples t-test confirmed that the difference is statistically significant, t(20) = 2.171, p < 0.05 two-sided 95%CI[13.19150, 659.55850]. The statistical test showed a large effect size Cohen’s d = .926 95% CI[.032, 1.798]. As far as the average total dwell time score is concerned, one participant spent on average a sum of 15501.23 milliseconds to read all metaphors and a sum of 10119.23 milliseconds for all simplified versions. Hence, participants who read metaphors invested more dwell time to process than the participants of the control group who read their simplified counterparts.

Fixation count

The comparison between groups shows that metaphorical items (M = 4.07, SD = 1.06) were fixated more rather than non-metaphorical items (M = 2.78, SD = 1.45). The result of the independent samples t-test confirmed that the difference is statistically significant, t(20) = 2.378, p < 0.05, two-sided 95%CI [.15841, 2.42113]. The statistical test showed a large effect size Cohen’s d = 1.014 95% CI[.111, 1.895]. One participant had a sum of 69.16 fixations on average for all metaphors and a sum of 44.45 fixations on average for all simplified versions. Hence, the experimental group fixated on metaphors more than the control group on simplified expressions.

Fixation time

On metaphors (M = 866.94 ms, SD = 258.69 ms) we registered longer fixation time than on their simplified versions (M = 575.21 ms, SD = 397.50 ms). The result of the independent samples t-test did not find the difference as statistically significant, t(20) = 2.040, p = 0.055, two-sided 95% CI[−6.54934, 590.01979] with 95% Confidence Interval crossing 0. Although, the statistical test showed a large effect size Cohen’s d = .870 even if the 95% Confidence Interval crossed 0, 95% CI[−.017, 1.738]. On average one participant spent a sum of 14738.07 milliseconds of fixation time for all metaphors and a sum of 9203.364 milliseconds of fixation time for all simplified versions. As the result of the statistical test shows, the fixation time difference between the two groups was not consistent across all the subjects. In other words, there might be some individual variations in fixation time that affected reaching a statistically significant difference between the two groups shown previously in the analysis by items.

Revisits

According to the descriptive statistics, subjects revisited metaphors more (M = .86, SD = .45) rather than their simplified versions (M = .53, SD = .33). However, the result of the independent samples t-test did not confirm the difference to be statistically significant t(20) = 1.974, p = .062, two-sided, 95%CI[−.01871, .67780] with 95% CI crossing 0. Nevertheless, it is associated with a large effect size Cohen’s d = .842 95%CI[−.043, 1.707] but which 95% CI also crosses 0. Hence, there were some variations in revisits across subjects that prevented from yielding statistical significance. As regards the mean of sum one participant had on average a sum of 13.73 revisits for all metaphors and a sum of 8.45 revisits for all simplified versions.

The subjects who read metaphors spent more time and fixated more often (more dwell time and fixation counts) than subjects who read simplified equivalents. We also noted a greater fixation time and number of revisits in the metaphor-reading group, although these differences were not statistically significant. The observed non-significance in fixation time and revisits might be connected with a small sample size (n = 11 per group) that prevented obtaining a significant difference. Perhaps it could also be related to the specifics of metaphor processing, for example, the fact that subjects do not make as many revisits. This was also the benefit of using two types of analysis, both by items and by subjects, as during the by-items analysis, the difference in fixation times and revisits was statistically significant (there was n = 16), which would, however, indicate that the group in the by-subject analysis was too small (n = 11). In addition, we also see in the analysis by subjects large numbers for standard deviation, which shows that data points are spread out over a large range of values. This might have also had an impact. It can be attributed to individual differences within the participants. Another experiment should therefore be carried out on a much larger number of subjects.

Analysis by subjects. Comprehension test

The descriptive statistics shows the comprehension scores across subjects (see Table 5). It includes mean and standard deviation per group. The maximum comprehension score is 16 as there are 16 items read by each subject. Subjects of the MetR group correctly understood 61.37% of items and subjects of SimMetR approximately 86.94% of items.

As can be observed, the comprehension scores for MetR (M = 9.82, SD = 2.56) are lower than for SimMetR (M = 13.91, SD = 1.92). The data passed the assumptions of normality and homogeneity of variance, and an independent samples t-test was conducted. It revealed a significant difference between two groups with large effect size t(20) = −4.237, p < .001, two-sided, 95%CI[−6.105, −2.077], Cohen’s d = −1.807, 95%CI[−2.795, −.787]. Hence, subjects comprehend metaphors less rather than their simplified version. As the sample size is small, MANCOVA was not performed.

Analysis by subjects. Appreciation test

The descriptive statistics shows the appreciation ratings by subjects including such information as group, mean and standard deviation (see Table 6). It can be observed that metaphors are appreciated more (M = 3.9091, SD = .45101) than their simplified versions (M = 3.3523, SD = .50488).

The data passed the assumptions of normality and homogeneity of variance, and an independent samples t-test was conducted. It revealed a significant difference between two groups t(20) = 2.728, p < .05, two-sided, 95%CI[.13103, .98260] with large effect size Cohen’s d = 1.163, 95%CI[.242, 2.060]. Hence, subjects appreciate more metaphors rather than their counterparts rewritten in plain language.

Analysis by subjects. Correlations

Due to violations of assumptions of Pearson correlation, Spearman correlations were performed. Significant positive correlations were found between appreciation ratings and two eye-tracking parameters such as dwell time r = .461, p < 0.05 and fixation time r = .448, p < 0.05. But appreciation ratings are not correlated with fixation count r = .398, p = .067 and revisits r = .049, p = .830. Hence, the more participants appreciated the expressions, the slower they read but not necessarily fixated or revisited more often.

Thus, subjects read metaphors slower with more often fixations than their simplified versions, although not always revisited and prolonged their fixations. In a similar manner to the analysis by items, subjects comprehended less and appreciated more metaphorical expressions rather than non-metaphorical expressions, and significant positive correlations were found between appreciation ratings and such eye-tracking parameters as dwell time and fixation time.

Overall, both analyses by items and by subjects confirm that there is a significant difference between the two groups (MetR vs SimMetR) in eye-tracking parameters with some individual variations in fixation time and revisits. Furthermore, the difference between comprehension tests and appreciation ratings is statistically significant. Metaphors were processed differently as well as comprehended less and appreciated more rather than their simplified counterparts. Also, eye-tracking parameters, especially those that suggest slow reading, are positively correlated with appreciation ratings. Thus, the more participants appreciated the expression, the slower they read it.

Discussion

Based on the results, we can assume that metaphors in a foreign language are processed differently than simplified expressions. Our study supports the hypothesis of Columbus et al. (2015)Columbus, G., Sheikh, N. A., Côté-Lecaldare, M., Häuser, K., Baum, S. R., & Titone, D. (2015) Individual differences in executive control relate to metaphor processing: An eye movement study of sentence reading. Frontiers in human neuroscience, 8, 1–12. and others (e.g., Clark and Lucy, 1975Clark, H., & Lucy, P. (1975) Understanding what is meant from what is said: A study in conversationally conveyed requests. Journal of Verbal Learning and Verbal Behavior, 12, 335–359. ; Ortony et al., 1978Ortony, A., Schallert, D. L., Reynolds, R. E., & Antos, S. J. (1978) Interpreting metaphors and idioms: Some effects of context on comprehension. Journal of verbal learning and verbal behavior 17(4), 465–477. ; Gibbs, 1981Gibbs, R. W. Jr. (1981) Your wish is my command: Convention and context in interpreting indirect requests. Journal of Verbal Learning and Verbal Behavior 20, 435–444. ; Inhoff et al., 1984Inhoff, A. W., Lima, S. D., & Carroll, P. J. (1984) Contextual effects on metaphor comprehension in reading. Memory & Cognition 12(6),558–567. ; Janus and Bever, 1985Janus, R. A., & Bever, T. G. (1985) Processing of metaphoric language: An investigation of the three-stage model of metaphor comprehension. Journal of Psycholinguistic Research 14(5), 473–487. ) that metaphors require more time to be processed. The difference in processing could be related to the idea in line with that of Kecskes (2001)Kecskes, I. (2001) The ‘graded salience hypothesis’ in second language acquisition. In: Pütz, M., Niemeier, S. and Dirven, R. (Eds.) Applied cognitive linguistics: Volume 1. Mouton de Gruyter, 249–271. and Akkök and Uzun (2018)Akkök, E. A., & Uzun, I. P. B. (2018) Metaphor processing in Turkish: An eye-movement study. Dil ve Edebiyat Dergisi, 15(1), 105–124. that non-native speakers consider literal meaning first. We also favor the concept of a three-level model of nonliteral processing (Clark and Lucy, 1975Clark, H., & Lucy, P. (1975) Understanding what is meant from what is said: A study in conversationally conveyed requests. Journal of Verbal Learning and Verbal Behavior, 12, 335–359. ; Grice, 1975Grice, H. (1975) Logic and conversation. In: Cole, P. and Morgan, J. L. (Eds.) Syntax and semantics: Speech acts. Seminar Press, 41–58. ; Searle, 1979Searle, J. R. (1979) Metaphor. In: Ortony, A. (Ed.) Metaphor and thought. Cambridge University Press, 92–123.). Literary metaphors are not so familiar, especially to non-native readers, and therefore they might be first interpreted literally. Next, the readers try to understand them by relying on the context, and in the third stage, they reject the literal meaning and create a metaphoric one. However, to confirm this concept more studies are needed, especially online studies that will reveal the method metaphoric expressions are processed by test subjects. Here, the method of TAPs (think-aloud protocols) could be useful. Of course, the longer time required to process metaphors might also be connected with insufficient metaphoric competence in L2 or a general lack of linguistic competence in the second language (L2). Our results also partially reflect the assumption made by Hoang (2014)Hoang, H. (2014) Metaphor and second language learning: The state of the field. The Electronic Journal for English as a Second Language 18(2), 1–27. that difficulties might relate to the suppression of source and target domains from speakers’ first languages.

The results of the comprehension test confirmed our assumptions that metaphors in L2 would be less understandable than their simplified versions. The response accuracy score for metaphors was 61%, whereas the control group that read their simplified counterparts scored 87%. This shows that metaphors in a foreign language are a major challenge for learners, especially literary metaphors. The processing of metaphors may be influenced by the familiarity effect as noted in a previous study by Olkoniemi et al. (2022)Olkoniemi, H., Bertram, R., & Kaakinen, J. K. (2022) Knowledge is a river and education is like a stairway: An eye movement study on how L2 speakers process metaphors and similes. Bilingualism: Language and Cognition 25(2), 307–320. . Yet, looking at the control group’s score of 87 percent, we can conclude that some questions regarding simplified expressions were also challenging. This signifies insufficient competence in L2. In addition, it shows that a literary text is a trickier type of text when it comes to comprehension.

The results of respondents’ appreciation of metaphors also confirm our earlier assumptions. Ratings of metaphors ranged higher, from 3.55 to 4.73, but for simplified expressions from 1.91 to 4.55. Among the three most appreciated metaphors were: “the curves of your lips rewrite history”, “rose-white boyhood” and “to wear youth’s livery”. This appreciation might relate to the level of novelty and unconventionality of the metaphorical expressions as well as to their potential to stimulate the imagination, as they are very illustrative.

Significant results were obtained in correlation tests. Appreciation ratings were positively related to eye-tracking parameters. Whenever the eye-tracking parameters increased, the participants’ appreciation ratings went up as well. However, whereas fixation counts and revisits were positively correlated in the analysis by items, they were not in the analysis by subjects. A small group size may have played a role in the correlation tests conducted in the analysis by subjects, and therefore it is important to conduct the study in the next stage on a much larger group of people to check this assumption. However, dwell time and fixation time were consistently correlated with appreciation ratings. Thus, we can conclude from our results that appreciation is correlated with slower reading. This supports the neurocognitive model of literary reading proposed by Jacobs (2011)Jacobs, A. M. (2011) Neurokognitive Poetik: Elemente eines Modells des literarischen Lesens (Neurocognitive poetics: elements of a model of literary reading). In: Schrott, R. and Jacobs, A. M. (Eds.) Gehirn und Gedicht: Wie wir unsere Wirklichkeiten konstruieren (Brain and Poetry: How We Construct Our Realities). Carl Hanser Verlag, 492–520. that foregrounding is aesthetically appreciated which in turn leads to a slow reading.

Conclusions

We posed four research questions of particular interest at the outset, namely:

1. How do we perceive metaphors in L2?

2. Is the comprehension of metaphors in L2 cognitively demanding?

3. Do people understand literary metaphors in a foreign language?

4. Are literary metaphors aesthetically appreciated by L2 speakers?

Regarding the first research question, based on our results we can conclude that metaphors are perceived and processed differently than simplified expressions. First, the perception period is longer, and the subjects make many more and longer fixations as well as more revisits. The significant effect of the group on eye-tracking parameters while controlling for comprehension suggests that prolonged processing might be not solely due to comprehension difficulties. It might support the three-stage model of processing, according to which, non-native speakers process first the direct meaning, check this meaning based on the context, and accept the metaphorical meaning.

The second and third questions can be considered together as they are related. Answering the second research question, the results of selected eye-tracking parameters showed clearly that processing, that is comprehending metaphors, was cognitively demanding for test subjects. All values of eye-tracking parameters were much higher for metaphors than for simplified expressions. The question related to readers’ understanding of literary metaphors in a foreign language cannot be answered unequivocally, and more extensive research is needed. Yet, based on our research, test subjects understood metaphors far less than they did simplified expressions. In the case of metaphors, the results of comprehension tests were 61 percent, i.e., more than half of the metaphors were understood by the subjects. However, this is not a good result. It might be the case that participants did not reach the last stage of processing where the metaphorical meaning is created. In the case of simplified expressions, the accuracy of responses was 87 percent. But here, too, we see that comprehension is not complete. Hence, we can also conclude that the competence of test subjects in the L2 may have been insufficient to clearly understand a literary work from the 19th century.

The answer to the last question is positive. The respondents appreciated the aesthetic value of metaphors far more than they did that of simplified expressions. Thus, in a larger scope, our results support the neurocognitive model of literary reading proposed by Jacobs (2011)Jacobs, A. M. (2011) Neurokognitive Poetik: Elemente eines Modells des literarischen Lesens (Neurocognitive poetics: elements of a model of literary reading). In: Schrott, R. and Jacobs, A. M. (Eds.) Gehirn und Gedicht: Wie wir unsere Wirklichkeiten konstruieren (Brain and Poetry: How We Construct Our Realities). Carl Hanser Verlag, 492–520., which is that foregrounding that causes defamiliarization is more aesthetically appreciated than backgrounding expressed in non-figurative language. Furthermore, aesthetic appreciation is associated with a slow reading. It would be worth repeating this study on a larger number of respondents. As the analysis by items showed a correlation of appreciation with all eye-tracking parameters that are related to the cognitive effort invested in processing the items in question, it would also be necessary to check in the next stages whether appreciation influences increased eye-tracking values or conversely, whether our cognitive effort invested in the task increases our appreciation. The TAPs could provide valuable insights into this topic.