1. Introduction

Gaze behavior studies in metaphor processing have a long history. In experimental psychology, the gaze effects and default interpretations of novel and entrenched metaphors have been vastly tested. Most commonly, they show that gaze costs contingent on entrenched metaphors are lower than on novel metaphors, which is explained by priming effects (Blasko & Connine, 1993; Blasko & Briihl, 1997), analogical reasoning (Bowdle & Gentner, 2005), lower ambiguity effects (Libben & Titone, 2008), the presence of biased priming context (Frisson & Pickering, 1999). At the same time, multiple experimental works produce sufficient evidence on individual differences in understanding metaphor and in ad hoc gaze behavior, for instance in terms of executive control (Columbus et al., 2015) which displays high contingency on gaze characteristics of higher and lower executive control readers depending on prior context. Experiments also tested the effects of selective attention on gaze behavior and metaphor comprehension (Klepousniotou et al., 2008), which have proved that in the cases of possible multiple literal and figurative meanings activation the participants with lower comprehension skills had higher gaze costs contingent on default interpretations. These effects have also been explored as contingent on reading models (Kliegl et al., 2006; Rayner et al., 2012). These and similar experiments are aimed at specifying the individual differences in metaphor processing.

Indeed, apart from testing the individual gaze behavior in processing novel and entrenched metaphors, there are other research directions which specify the metaphor character as stimulating these differences. Considering the fact that figurative language construal displays high diversity, for instance in participant construal in terms of their animated or non-animated character (Wårwik, 2004), event construal in terms of its achievement or agentivity (Langacker, 2015; Talmy, 2000), or in perspective construal in terms of its vantage point or observation path (Verhagen, 2007; Iriskhanova, 2013), we cannot leave these characteristics unattended when selecting experiment stimuli and expecting to test individual differences in metaphor processing. Several potentially significant construal factors have been advocated in the gaze studies of metaphor processing, with event construal specifics (Coulson & van Petten, 2002; Coulson, 2008; Thibodeau & Boroditsky, 2013), linguistic (especially syntactic) construal effects (Gibbs & Matlock, 2008), salience in metaphorical models (Giora, 2003) or pragmatic models (Giora et al., 2018), verbal and nonverbal categories distribution (Glucksberg, 2003), modulating construal in novel and entrenched metaphors (Kiose & Kharlamova, 2021).

In this study, we explore the construal effects of novel and entrenched metaphors onto individual gaze behavior. We hypothesize that despite the differences in metaphor types, figurative construal in terms of cognitive and linguistic construal may affect the gaze characteristics. To observe possible individual differences, we addressed the factor of working memory, since this factor is admitted as most significant in entrenchment effects in context (Gunter et al., 2003). This can be explained if we consider that in context novel metaphors may become entrenched through their repeated use and consequently lower gaze costs will be observed (Cardillo et al., 2012; Goldstein et al., 2012). In the previous study (Kiose & Kharlamova, 2021) we reported the results on the construal differences observed in novel and entrenched metaphors separately, however these results allowed to predict that there might be significant construal and gaze cost contingencies irrespective of metaphor novelty / entrenchment but dependent on other metaphor construal factors like agentivity or degree of abstractness.

2. Research methods and procedure

In cognitive linguistics, linguistic and cognitive construal are the key notions which are described through cognitive mechanisms and operations. In recent works, different construal mechanisms are considered through Event-modeling framework (Divjak et al., 2020; Hart & Queralto, 2021), which allows to explore the systemic effects of different construal patterns. Cognitive and linguistic construal patterns are explored in multiple works extending their application to metaphor models (Boroditsky, 2000; Graumann & Kallmeyer, 2002; Dancygier & Sweetser, 2012; Goatly, 2017). In our prior studies (Kiose et al., 2020) we proposed three types of cognitive construal patterns to explore nonverbal event construal, which are Referent Construal, Event Frame Construal and Perspective Construal patterns. Following J. Pustejovsky (1995), we distinguish four types of Referent construal patterns or ways of seeing the participants, Part-whole, Kind, Functional, Life-history, which allow to explore the participant in terms of their agentivity, complexity, referential integrity, etc. Following V. Demyankov (1983), to describe Event Frame Construal we consider such patterns as event completeness, instantness, evaluation, manageability, repeatability, etc. Following O. Iriskhanova (2013), we describe Perspective construal exploring the patterns of vantage point, viewpoint, distancing, observation path, etc. Verbal construal is assessed in the patterns of linguistic foregrounding described in Iriskhanova (2014). The complete list of construal patterns attested in the stimulus is given in Appendix B.

To explore cognitive and linguistic figurative construal, we address two types of metaphors, novel and entrenched. Following Schmid (2016) and Langacker (2016) we consider the metaphors which do not exploit earlier activated metaphoric models and correspondences as novel. The following sample may demonstrate the case; here the metaphoric nominal group новостройки (‘new buildings’) refers to large new waters:

– Тоже проблема: почему рыба из малых рек уходит в новые большие водоемы?  

– Проблема! А как вы думаете?.. Еще какая! У нас тут были целые рыболовецкие артели – крышка. Распускать! А у людей – образ жизни сложился, профессия…

– Назовите это: рыба уходит на новостройки – и дело с концом (V. Shukshin)

(fish-3SG go-3SG to new-PL building-3PL-ACC) 

(Translation) There is a new problem: why does fish leave small rivers for large new waters?

– It is a problem! What do you think? A pretty big one! We had whole fishermen’s artels here, all dead. Dissolve! And people developed a way of life, a profession ...

– Name it so: the fish goes to new buildings – and call it a day.

In the clause рыба уходит на новостройки the metaphor новостройки is novel, which is corpus-supported. Its use in the Russian National Corpus does not allow to find similar metaphoric correspondences with новостройка as a fish habitat, although the NCRL frequency use of the lexeme is 684 with новостройка – 43, новостройки – 224, новостройку – 53, новостройкам – 21, новостройке – 70, новостроек – 273. It employs one of the ontological metaphor models (ANIMALS are PEOPLE), still there are two metaphor correspondences, ANIMALS’ HABITAT is PEOPLE’S HABITAT which is not novel, see дом птиц (Е. Гончаренко), собачий дом (М. Горький), and  NEWLY-FORMED HABITAT FOR ANIMALS is NEWLY-MADE HABITAT FOR PEOPLE which is novel. Therefore, the metaphor is a novel one.

Importantly, entrenchment can be observed in context (Cardillo et al., 2012) and over time (Bowdle & Gentner, 2005). In this study, we address the case of textual entrenchment when one novel metaphorical nominal group is used repeatedly to refer to the same referent. These examples are not frequently present in short textual fragments being more common for larger textual spaces, however the latter will not suffice to be applied in a gaze study; besides, single repetition will not satisfy the needs of analysis since a reader may simply miss the nominal group in its first or second use, which will make us disregard his results. Therefore, for the experiment we had to select the stimulus which demonstrated several uses of a novel metaphorical nominal group in a small textual fragment. The selected text sample from Fazil Iskander’s short story “Thirteenth Feat of Hercules” has 5 uses of novel metaphoric nominal group, where the group “Отъявленный Лентяй” (F. Iskander) is used three times in the text referring to a boy who is not lazy, thus this is the example of metaphoric disanalogy (Bowdle & Gentner, 2005) and twice with the modification of the form “лентяй, добросовестный и послушный лентяй” (F. Iskander) bearing the same reference. The Stimulus and its translation and glosses are given in Appendix A. Below, we will consider the differences in the event construal in this stimulus. In its first use, in “Оказалось, что я держу носилки как Отъявленный Лентяй” (F. Iskander), the novel metaphor “ОтъявленныйЛентяй” (F. Iskander) is used to refer to a single person participant of agentive type which is the speaker himself; the event is of an incomplete type, displaying space location relations, bodily construal; it presents the inner observer type (the event participant’s perspective). In terms of linguistic construal, it displays graphic foregrounding (capitalization), syntactic foregrounding (syntactic rhematic position). The second use, in “Через некоторое время слухи об Отъявленном Лентяе дошли до директора школы” (F. Iskander), displays weaker syntactic foregrounding, still graphical foregrounding is present. Participant construal patterns are the same, however the event construal patterns differ; here the event is completed, it has time location, however, the participant’s perspective is vague. The third and the fourth uses display modifications in linguistic form in “Вскоре выяснилось, что никаких фокусов я не собираюсь выкидывать, что я, напротив, очень послушный и добросовестный лентяй. Более того, будучи лентяем, я вполне прилично учился” (F. Iskander). The fifth use, in “Так, доигрываянавязанныймнеобразОтъявленногоЛентяя, япришелкзолотоймедали” (F. Iskander), is still graphically foregrounded, however the event construal patterns differ since the event in the clause with the entrenched metaphor is of incomplete type, has no space or time location, displays no bodily construal.

The eye-tracking experiment was conducted to test the gaze costs (in terms of Fixation Duration, First Fixation Duration and Fixation Count) in 5 Areas of Interest corresponding to the zones of 5 metaphoric nominal groups. The experiment was preceded with an n-back test performed in Brain Workshop 4.8.4 (Jaeggi et al., 2008) which is customarily used in memory tests (Kane et al., 2007; Farvardin et al., 2014) to identify the groups of participants who display differences in the capacity of working memory. In the n-back task, the participants are presented with several visual stimuli. They have to report whether each stimulus matches a stimulus in trials before. The test assesses the number of mistakes and the relative execution time. In the eye-tracking experiment, 60 gaze probes were received with 12 participants, with the mean age of 22. All experiment participants were native speakers (Russian) and had corrected-to-normal or normal vision. Participants were asked to read the sample and answer the questions concerning the reference of the groups, however in this work we will consider only their gaze behavior. Pdf-formatted visual stimulus appeared for 90 seconds. In the experiment we used The SMI Red-x eye tracker (running at frequency 60 Hz and with operating distance of 60-80 cm). The eye movement data were next sampled (fixation duration, first fixation duration, fixation count in AoIs) and subjected to regression modeling.

3. Research Results

3.1 Data preparation

There were 5 AoIs and each AoI was manually annotated. The annotations served to explore contingency on the gaze characteristics. We regarded eye movement characteristics, distinguished in eye movement control models (Reichle et al., 2003; Richter et al., 2006). The accumulated data in 5 AoIs is presented in Table 1 for 3 gaze characteristics: fixation duration per sign in the Areas of Interest (FD), first fixation duration in Areas of Interest (FFD), fixations count (FC).

Table 1. Gaze characteristics data in 5 AoIs

K-means cluster analysis, performed in IBM SPSS Statistics 20, was used to define groups of participants based on n-back test scores. As 4-cluster solution did not provide the resultant solution, the sample was divided into two clusters: 5 subjects were categorized as “demonstrating better working memory solutions (BWM)”, 7 subjects were categorized as “demonstrating worse working memory solutions or worse working memory (WWM)”. The descriptive statistics of 12 participants’ exactness and relative execution time are as follows: the mean number of mistakes was 2.36 (Min. = 0.875, Max. = 9), the mean time per each answer was 1.69 s (Min. = 0.84, Max. = 3.47). The datasets – 5 AoIs annotations of non-verbal and verbal construal in the stimuli, and the data on 3 gaze characteristics for 2 participant groups, were sampled and subjected to contingency analysis and regression tests.

3.2 Construal effects

The gaze characteristics involved Fixation Duration per sign (ms), First Fixation Duration (ms) and Fixation Count per sign (ms). In Figure 1 we present the gaze characteristics for two groups.

Figure 1a. Gaze characteristics for Group 1 (BWM), ms

Figure 1b. Gaze characteristics for Group 2 (WWM), ms

As seen, the gaze characteristics display difference in two groups, with all three gaze characteristics being higher in duration and count with Group 2.

For data processing we used JAMOVI software. The two datasets were preliminarily submitted to contingency analysis. The first step was to explore the effects of cognitive and linguistic construal onto gaze behavior. Since the parameter values were of nominal type (46 cognitive and linguistic construal parameters with positive values) and continuous type (3 gaze characteristics), we applied the Student’s t-test. To verify the results significance, we applied Bonferroni-Holm correction (Holm's Sequential Bonferroni Procedure) which controls the familywise Type I error rate in a less conservative manner as compared with standard Bonferroni correction. The t-tests (276 trials) were conducted for two participant groups separately, Bonferroni-Holm corrections were further calculated.

Next, regression modeling was applied to reveal the regression models predicting the gaze behavior (3 gaze characteristics) with non-aliased construal parameters (4 parameters, which are Agentive participant, True event, Perception, Positive evaluation) and with non-aliased syntactic parameters (Clause-final position for neutral syntactic foregrounding, Position in subsidiary clause, Position in parallel structure, Position in elliptical clause) separately. The need to establish the regression model with syntactic parameters was stimulated by the collected evidence that these were the syntactic parameters that were mostly contingent on the gaze behavior with the participants of Group 2, while the participants of Group 1 demonstrated very few contingencies, overall. We will see which model suits better to describe the process of entrenchment within each of the groups. The model performance summary statistics will be presented for 6 models for 2 participant groups in 3.2.1, 3.2.2 and 3.2.3.

3.2.1 Construal parameters and Fixation Duration per sign

As explained above, the presence or absence of cognitive and linguistic construal parameter markers in 5 AoIs of the stimulus was previously established by the annotators. We received the data on non-verbal and verbal construal in binary format, with 1 for AoI presence and 0 for AoI absence which was further subjected to t-tests in two participants’ groups. Table 2 shows the results of t-tests with 46 construal parameters in two Groups, only the parameters displaying significant p-values are given. Bonferroni-Holm corrections are also presented.

Table 2. Construal parameters and Fixation Duration per sign in two groups

The analysis has revealed significant differences in two participants’ groups. The most obvious is the difference in the number of construal parameters which display significant effect on Fixation Duration. There are two parameters which lead to higher gaze costs with the BWM participants; they both display lack of dynamics and relations, at the same time dynamicity and visual foregrounding reduce the gaze costs. The situation with WWM participants is far more complicated; whereas agentivity and dynamicity play similar role, it is additionally strengthened by intensity and bodily and emotional experience. Higher costs are identified in the situations of spontaneity, repetition (possibly, since repeated use seemed unnatural), paradigmatic and syntagmatic foregrounding of various types. We may hypothesize that the WWM participants tend to pay more attention to different types of linguistic foregrounding in construal, presumably they serve as affordances stimulating the memory decisions. At the same time cognitive foregrounding (agentivity, dynamicity, etc.) is easier to follow, however this hypothesis needs further verification.

To establish the regression Model 1 (with non-aliased parameters), we conducted Linear regression analysis, which revealed only 4 parameters with non-aliased coefficients. Since linear regression was held during successive subjection of construal parameters, only one Participant (Agentive Participant) and three Event Frame parameters (True Event, Perception, and Positive evaluation) were included. Table 3 shows the regression models for two participants’ groups predicting Fixation Duration.

Table 3. Regression Model 1 with non-aliased coefficients predicting Fixation Duration in two groups

The results demonstrate that the predictability of Regression model for Group 1 is far lower (R² = 0.198), besides there are no good predictors. As opposed to it, the regression model with Group 2 has higher prognostic potential (R² = 0.356), and one (which may mean that there are several good predictors since there will be multiple aliased parameters) good predictor (with p = 0.005) which is the Bodily construal (displaying perception markers). Noticeable, that Perception also displayed significance as a single parameter contingent on lower gaze costs in terms of Fixation Duration.

Since there are multiple syntactic construal parameters which display contingency on Fixation Duration, we also conducted Linear regression analysis separately for this parameter group. There are 4 non-aliased syntactic parameters which are Clause-final position for neutral syntactic foregrounding, Position in subsidiary clause, Position in parallel structure, Position in elliptical clause. There are no good predictors in Group 1, however there are two predictors in Group 2 which are Clause-final position for neutral syntactic foregrounding (p = 0.006) and Position in elliptical clause (p = 0.019). Since both these parameters display lack of syntactic foregrounding, we may conclude that this factor plays a decisive role in gaze costs for the WWM participants increasing the gaze costs.

3.2.2 Construal parameters and First Fixation Duration

Since First Fixation Duration is not contingent on the number of signs in the AoIs, we considered its absolute values in two Groups and with all AoIs. Table 4 shows the results of t-tests with 46 construal parameters in two Groups, only the parameters displaying significant p-values are given. Bonferroni-Holm corrections are also presented.

Table 4. Construal parameters and First Fixation Duration in two groups

The results manifest that the distributions of First Fixation Duration with the participants of Group 1 are more contingent on individual reading styles and do not display any rigid contingencies on the construal parameters, whereas in Group 2 there exist several construal parameters which produce shorter first fixations and consequently lower gaze costs. These are the parameters which demonstrate realistic events, dynamicity, interpersonality, lack of evaluation. Higher costs are identified in the situations of event fictitiousness, although in terms of event stativity and evaluation in events there are no rigid contingencies. This might indicate that only gaze cost decrease effects are stable with Group 2, and higher gaze costs (similarly to the effects detected in Group 1) are more contingent on individual styles.

To establish the regression Model 1 (with non-aliased parameters), we conducted Linear regression analysis with 4 parameters with non-aliased coefficients, which are Participant (Agentive Participant) and three Event Frame parameters (True Event, Perception, and Positive evaluation).  Table 5 shows the regression models for two participants’ groups predicting First Fixation Duration.

Table 5. Regression Model 1 with non-aliased coefficients predicting First Fixation Duration in two groups

Model predictability is low in both Groups. At the same time, we may notice that in Group 1 the cumulative effects are higher than the effects of single construal parameters. There are two parameters which work as good predictors for the Model which are the realistic type of event and the bodily event construal (explicit perception); and agentivity of participant (however not displaying rigid contingency) also plays a role. In Group 2 there are no good predictors, which evidences in favor of stronger effects of individual construal parameters, in our case (see Table 2) they are realistic events, dynamicity, interpersonality, lack of evaluation. The results demonstrate that in terms of First Fixation Duration realistic event (True Event) is a good predictor in cumulative model with Group 1 (BWM), however as a single parameter with the same group it does not display contingency. At the same time, True Event is contingent on shorter first fixations with Group 2 (with worse memory) in terms of single effects, however it does not display any significant effect in a cumulative model. Presumably, realistic and fictious event construal does play a role in gaze costs in both groups, however this hypothesis needs higher specification and more detailed analysis. Noticeable, that similarly to Model 1 with Fixation duration per sign, Perception also displays significance; here it serves as a good predictor with Group 1, whereas it is a good predictor in terms of fixation duration per sign with Group 2 and stimulates longer fixation duration in terms of the effects of single parameters with Group 2. These observations suffice to prove that bodily construal effects are significant for both Groups and should necessarily be considered in event construal analysis.

The prognostic modeling of syntactic foregrounding effects which display contingency on First Fixation Duration (Linear regression analysis was conducted with 4 non-aliased syntactic parameters specified in 3.2) has shown that there is one good predictor in Group 1 which is Position in parallel structure (p = 0.049) and two good predictors in Group 2 which are Clause-final position for neutral syntactic foregrounding (p = 0.028) and Position in parallel structure (p = 0.028). Since syntactic construal appears to be more significant for the participants with worse working memory, we may conclude that it allows them to relocate their attention on more syntactically focal and consequently more important information.

3.2.3 Construal parameters and Fixation Count per sign

Since Fixation Count displays contingency on the number of signs in the AoIs, we considered its relative values in two Groups and with all AoIs. Student’s t-tests revealed that in Group 1 there are no results with significant p-values. Table 6 shows the results of t-tests with 46 construal parameters in Group 2, only the parameters displaying significant p-values are given. Bonferroni-Holm corrections are also presented.

Table 6. Construal parameters and Fixation Count in Group 2

The results seem demonstrative since they unanimously prove that gaze costs in Fixation Count similarly to other gaze characteristics are dependent on individual construal styles and not on single construal effects with the BWM participants. Meanwhile, the situation is different with the participants in Group 2. We may notice that semantic foregrounding, here expressed in agentivity, actionality, emotivity, graphic and syntactic foregrounding in most cases is contingent on lower gaze costs, whereas the lack of semantic foregrounding (displayed in lack of dynamics, lack of interactivity) and also unexpected construal effects or the effects which require the back-up information retrieval (spontaneity, repetition, paradigmatic novelties, position in parallel structure) are more demanding.

The established regression Model 1 (with non-aliased parameters) for Fixation Count per sign with 4 parameters with non-aliased coefficients is presented in Table 7.

Table 7. Regression Model 1 with non-aliased coefficients predicting First Count in two groups

Model predictability is low in Group 1. However, the situation is different in Group 2 where its prognostic potential is high and there is one good predictor which is Bodily construal (in terms of perception). Since bodily construal effects were also noticeable in other gaze characteristics with Group 2, we may conclude that this construal type displays high importance for the participants with worse working memory. The presence of perception in construal contributes to lower gaze costs in fixation duration and fixation count as single parameters contingent on gaze characteristics, and as model predictors, which advocates that the WWM participants may rely on pre-position bodily construal markers to develop a specific reading strategy.

4. Discussion

In this section, we will consider the results which allow scaling the construal parameters in terms of their significance for gaze costs within two groups, the participants with better and worse working memory. We will also contrast the t-test results in two groups and the prognostic model predictability within the two groups to see whether the effects revealed may really serve to identify several distinctive features in construal and gaze cost contingency.

To range the construal parameters significance, where possible, we will rely on 6 gaze identifiers deduced, t-values of parameter contingency on Fixation Duration per sign, First Fixation Duration, and Fixation Count, and predictor values of parameters on Fixation Duration per sign, First Fixation Duration, and Fixation Count. We will consider the construal parameters to demonstrate higher significance in case they have higher values acting as single predictors. In Table 8 we present the scaled results of construal parameters predicting higher gaze costs in two groups.

Table 8. Construal parameters predicting higher gaze costs in two groups

Table 8 displays multiple effects of higher gaze costs with Group 2 which appear in Event construal and Linguistic construal which proves the dominant role of the factor of working memory as affecting the gaze costs in creative language interpretation (Gunter et al., 2003). Higher costs result from 1) foregrounding effects in lexical and syntactic levels of linguistic construal, 2) lack of dynamicity and interaction in event construal, 3) unexpected event construal. In Group 1 higher gaze cost effects are more sporadic and result from lack of dynamicity and interaction in event construal. The results demonstrate that creative language use in its novel and entrenched variants still has several common effects onto its gaze perception which appear to be steady with the participants with worse memory. Therefore, the findings extend the eye-tracking experimental results obtained in (Cardillo et al., 2012; Goldstein et al., 2012; Kiose & Kharlamova, 2021) which state the differences only for novel and entrenched metaphors gaze costs. Even though the novel indirect name is used repeatedly, the participants face multiple challenges in the event construal in AoIs which affect their reading strategy. It presupposes the recognition of particular construal affordances where the participants intend to search for the most important information, in repetitions, parallel structures, lexical novelties. Meanwhile, the unexpectedness in the planned event construal or inability to discover these construal affordances also leads to higher gaze costs. These results prove the significance of foregrounding effects in the construal identified in (Iriskhanova, 2014), however they allow to scale these effects in respect to the readers with better and worse working memory.

Now we will address the construal parameters which are contingent on lower gaze costs. In Table 9 we present the scaled contingency results in two groups.

Table 9. Construal parameters predicting lower gaze costs in two groups

The results show that lower costs in Group 1 result from 1) visual foregrounding, 2) dynamicity of events. In Group 2 the effects are more complex. Lower costs result from 1) bodily event construal, 2) foregrounding and intensifying construal information, 3) dynamicity and interactivity of events, 4) emotionality in event construal. These findings additionally prove that the construal parameter typologies specifying the Referent and Event Frame construal parameters (Pustejovsky, 1995; Coulson & van Petten, 2002; Coulson, 2008; Langacker, 2015; among others), as well as the classifications of foregrounding parameters (Wårwik, 2004; Verhagen, 2007; Iriskhanova, 2013) are reliable and valid when assessing gaze contingency effects.

If we consider the best predictors for the regression models, we will see that in both groups they are Perception (Event type), and True Event in Group 1. This observation serves as empirical proof of the theoretical assumptions on the role of embodiment in metaphor construal (Boroditsky, 2000; Goatly, 2017). The prognostic role of models in two groups is different, although; there is only one model which has high prognostic value, it is the regression model for Group 2 predicting Fixation Count. In this model the construal parameter of Perception is a good predictor, which means that bodily event construal is significant for creative language use gaze perception with the participants with worse working memory.

5. Conclusion

The study has proved the efficiency of cognitive semantic analysis in testing the cognitive costs produced by metaphor interpretation effects. In the current work, we applied the oculographic procedure to experimentally assess the interpretational value of language construal parameters (in cognitive and linguistic construal). Most commonly, experimental studies explore the gaze costs contingent on entrenched and novel metaphors and reveal the differences in gaze costs affected by lexical priming, lexical ambiguity, the presence of analogy and disanalogy construal models. Extending these results, we hypothesized that irrespective of a metaphor type, figurative construal influences the gaze costs. In this case, these results should be considered when assessing the effects of priming, analogy and ambiguity as affecting the gaze costs in metaphor interpretation, both novel and entrenched.

In the experiment, we observed the cases of multiple use of a novel textual metaphor during which the metaphor is conventionalized and modified becoming novel again; in these cases, it is interpreted with decreased and again increased cognitive load, here assessed via gaze costs. We largely considered the role of the readers’ working memory which affects novel metaphor interpretation and also the process of its entrenchment. We verified the hypothesis that gaze behavior characteristics (fixation duration, their number, first fixation duration in the Areas of Interest) display contingency on the metaphor construal parameters and not singularly on metaphor types, novel and entrenched. The contingency analyses specified the gaze values of six groups of parameters, the cognitive parameters of Referent construal, Event Frame construal, Perspective construal, and the linguistic parameters of graphological, lexical and syntactic construal. The regression analyses established the construal parameter models of gaze behavior characteristics reveals several predictors which define the steady decrease and increase of the gaze costs within two participant groups. The study showed that the participants with higher working memory display higher gaze costs when the construal parameters Lack of action or dynamics and No interpersonal or interactive relations are active; whereas the participants with lower working memory are affected by multiple parameters including Clause-initial position for contrastive focus, Lexical synonyms and antonyms, Spontaneous or occasional event, Lack of action or dynamics, No interpersonal or interactive relations, and several others. We also identified the predictors which decrease the gaze costs; the participants with lower working memory are affected by Perception (Event type), Shifting / changing, Agentive participant, Intensifiers in pre-position, Action verbs. Therefore, the participants with lower working memory are highly sensitive to linguistic construal of metaphors, presumably, it creates additional affordances and stimulates metaphor construal, whereas linguistic construal does not display consistency with the participants with better working memory; at the same time lower dynamicity in the cognitive construal hampers metaphor interpretation in both groups. Overall, the study allowed to claim that irrespective of the metaphor type, several construal parameters produce consistent gaze costs, either increased or decreased.