The research data are the film samples (fragments) with 5 male actors performing monologues in very popular Russian films: «Звонят, откройте дверь», А. Митта (“The doorbell rings, open the door” directed by A. Mitta), 1961, «Они сражались за Родину», С. Бондарчук (“They fought for the Country” directed by S. Bondarchuk), 1975, «Судьба человека», С. Бондарчук (“Man’s fate” directed by S. Bondarchuk), 1959, «Доживем до понедельника», С. Ростоцкий (“Let us live till Monday” directed by S. Rostotskiy), 1968, «Москва слезам не верит», В. Меньшов (“Moscow does not believe in tears” directed by V. Menshov), 1980. They feature the monologues of R. Bykov, Yu. Nikulin, S. Bondarchuk, V. Tikhonov and A. Batalov. To contrast the same actors’ function frameworks in less staged discourse, we selected 5 interviews taken in different environments, with the interviews with V. Bykov, Yu. Nikulin, and V. Tikhonov taken in the studio, the interview with S. Bondarchuk taken in his study, the interview with A. Batalov taken on the river embankment. All the monologues display realistic character; they are both descriptive and argumentative.

Since the research data were the samples of argumentation and description, to compile a list of functions we addressed the studies which looked into general discourse structuring and pragmatic aspects as well as the studies featuring argumentative and descriptive discourse strategies. Regarding argumentation, the studies account for its ability to express opinions and beliefs (Amossy, 2009), subjectivity and intersubjectivity (Willard, 1989), as well as argumentation schemes like Example, Cause to Effect and Effect to Cause, Practical Reasoning, Inconsistency (Cabrio et al., 2013) employing more fine-grained classification. All these features and schemes display pragmatic functions. Following these classifications and limiting their number to contrast the pragmatic functions in the selected cinematic and interview monologues, we explored 11 pragmatic functions (Opinion, (Emotional) assessment, Stating reasons, consequences, conditions, Contrast, Accusation, Agreement / Disagreement, Appeal to action, Promise, Threat, Comparison, Appeal to power). The studies of description mostly appeal to the topics which are considered in discourse and the way they are presented (Merlo & Mansur, 2004) as well as to the discourse components and types of discourse events including referent types and space construal specifics (Von Stutterheim & Klein, 1989; Longacre, 1996). Hence, we appealed to the pragmatic function of foregrounding the description event components and included their 9 types with three components specifying the type of action performed (Achievement, Process, and State) and six other components (Subject, Object, Action or State, Characteristics, Time, and Place). In terms of the second function, discourse structuring, we addressed the discourse construal theories where the patterns of information foregrounding were presented (Wårwik, 2004; Verhagen, 2007; Iriskhanova, 2014), the pragmatic theories of speech acts and performativity (Austin, 1962; Searle & Vanderveken, 1985; Green, 2000; Kearns, 2006, among others), but mostly to the theories of Rhetorical acts (Mann & Thompson, 1988; Kibrik & Podlesskaya, 2009; Holler et al., 2020). They name the patterns of activating information which are suitable for both argumentation and description, for this reason they are in most cases similar for the discourse stretches. The common functions are Emphasizing opinion or assessment / Emphasizing a discourse component, Self-correcting, Specification, Generalization, Intersubjectivity, Appeal to attention, Chain of arguments / Chain of events, Self-quote, Quoting others, and Figurativity; Rhetorical communication and Initializing communication are applicable only to argumentation.

In Table 1 we present the taxonomy of speech functions further employed for studying speech behavior in cinematic and interview monologues. We applied the coded annotation of speech and gesture functions using the codes 101–123 and 201–219 to annotate speech patterns.

To explore the gesture functions in the current work in both cinema and interview discourse, we will address their four basic communicative functions (with further specification): pragmatic, representational, deictic and adaptive (Kendon, 1995; Bressem, 2012; Cienki, 2017). Pragmatic gestures have various functions as these gestures are context dependent rather than form dependent, i.e., the meaning might change when their co-occur with different words, although the form can be the same. The main functions of pragmatic gestures are Discourse emphatic, Discourse structuring, Discourse representational, Expressing attitude/evaluation and Contact establishing (McNeill, 1992; Kreidlin, 2002; Calbris, 2011). As it is illustrated in the example below (Figure 1), pragmatic gesture with the function of expressing attitude is used in order to highlight the intensity of the laughter in the described event.

Representational gestures, also known as iconic gestures, are based on the idea of similarity between a hand form and / or its movement and the process or object which it refers to (Streek, 2008). There are several modes of representation which are distinguished in this study: Holding, Molding, Acting, Embodying and Tracing (Müller, 2014). In the example taken from the interview (Figure 2) with R. Bykov, we can see a Tracing gesture used to show the form of a small table in order to highlight its properties.

Another type, deictic gestures, are used to refer to people, objects, notions, places, events, etc. by creating axis in space which connects the speaker and the target of speech (Clark, 2003; McNeil, 2003; Cienki et al., 2014). We highlight two categories of deictic gestures: Pointing and Touching. The example of a deictic gesture (Figure 3) demonstrates its Pointing function. The gesture has a vector, it creates a line which links the speaker and the referent, the place where the action took place, thus attracting and guiding the attention of the listener.

The last type, adaptors, represent some movements, which can be self-oriented (Self-adaptors) such as rubbing one’s nose, adjusting glasses, fidgeting one’s fingers, etc., or they can be object-oriented (Object adaptors): touching the table in front of the speaker, moving a glass of water, trifling with a pen, etc. These gestures can be used in order to gain control of the situation when the speaker is in the state of distress (Ekman, 2004). As in the example given below (Figure 4), the speaker uses this gesture (touching his upper lip with his index and middle fingers repeatedly) while reflecting on the subject of writing and literature of the past.

In Table 2 we present the taxonomy of gesture functions further employed for studying gesture behavior in cinematic and interview monologues with the codes 301–314.

In modelling and processing the data, we will apply the method of variance and regression analysis in discourse profiles construal. The notion of discourse profiles suggested in construction grammar and structure building frameworks (Ariel, 2004), developmental studies (Singer, 2013) and since recently in multimodal studies (Iriskhanova & Cienki, 2018) and discourse studies (Kiose, 2021) is used here to assess the relative activity of speech and gesture functions in communication.

The study develops a two-stage procedure. First, it detects the variance in the speech and gesture function frameworks in more aesthetic cinematic discourse and less aesthetic interview discourse. Next, we proceed to contrastive analysis of individual specificity of each actor in stimulating this aestheticism, hypothesizing that despite the individual variance these function frameworks will display a steady character.

The main procedural questions were the selection of the unit of analysis and the annotation format. To select the procedural unit for analyzing speech and gesture complexes, we adopt the view that this unit should be able to manifest (and describe) both description and argumentation. Since the smallest unit capable of manifesting description is a word combination displaying predicate or attribute relations and the smallest unit capable of manifesting argumentation must necessarily be a proposition or its modal frame with either predicate or performative relations, we select the unit with a higher information potential which is the proposition or its modal frame.

For instance, Example (1) has 5 propositions which display different argumentation and description potential:

(1) Я никогда не был пионером // Но у нас во дворе был форпост // Но так как я никогда не был первым пионером // то я вам расскажу не про себя а про горниста // который жил у нас во дворе

I have never been a pioneer // But we had a fort-post // But since I have never been a first pioneer // I will tell you not about myself but about a bugler // who lived in our courtyard (R. Bykov, cinematic discourse)

Each of the units was annotated following the same procedure. In case we faced difficulties separating propositions in speech (for instance, when there were hesitations, hedges, interruptions) we adopted the following principle: a unit must necessarily involve either a proposition or a modal frame, therefore all the fragments which do not constitute a proposition or a modal frame are incorporated into the proposition or modal frame that was previously started and was not yet terminated. Example (2) illustrates the described case.

(2) Спасибо вам // что вы именно сюда приехали // потому что как… здесь снимались «Журавли» // Вот … эээ … ну я то … тут финал снимался … вот на этом месте буквально

Thank you // that you came right here // because it was … here… the “Cranes” was filmed // Right here … er… and I … here the final episode was made … at this very place (A. Batalov, interview discourse)

Units 3 (потому что как… здесь снимались «Журавли») and 4 (Вот … эээ … ну я то … тут финал снимался … вот на этом месте буквально) may have combined several propositions, however they are not completed in oral speech.

Annotation was performed in ELAN software, created by Max Plank Institute and used to annotate gestures (https://archive.mpi.nl/tla/elan). We chose it since it allowed to annotate the cinematic and interview shots considering their dynamicity. In Figure 5 we give the annotation example with the annotation tiers showing the role of speech and gesture.

The decision on the gesture function was adopted following the analysis of their form (e.g., the form of the hand: palm up / down, fist, finger extended, etc.; its movement: straight line, circle, wave, etc.; then the direction of it, as well as the space on which it occurred: horizontal or vertical axis, away or towards the speakers, etc. (see Bressem, 2013)) consistent with communicative functions. After that the attention was paid to the semantics of the gesture, which determines its type and corresponding functions in speech, as mere form analysis cannot be used to determine their role in speech due to the polysemantic nature of gestures (Calbris, 2011). The process of annotation included several steps. After uploading a video to ELAN, we created different tiers that represented the parameters that we analyze (speech functions, gesture functions). Figure 5 demonstrates the analysis of the interview with S. Bondarchuk, where the actor discusses the role of cinema and what cinema should look like. In the fragment, we annotated the following proposition in Example (3):

(3) Это слишком серьезное занятие

It is a very serious occupation (S. Bondarchuk, interview discourse).

First, we annotated speech functions in terms of argumentation and description. In the above given proposition, we can see that the actor uses Generalization as a pragmatic function of argumentation. S. Bondarchuk discusses the problems that exist in the cinema and summarizes the point by stating that cinema is a serious type of occupation. We pointed out two pragmatic functions of description: State (since this is a description of a state) and Accentuated characteristics (since the modifier serious is foregrounded). Next, we addressed the co-speech gestures, and specified the gesture function depending on the performed hand movements and speech. In the example given the actor is rubbing his hands which could indicate that he is using Self-adaptors.

We also employed the speech scripts with full annotations in txt-format. To process the data, we applied HETEROSTAT software (Kiose & Efremov, 2020) which allows to identify the annotated functions activity as well as their contingency.

In Figure 6 we show the HETEROSTAT window processing the data.

As it can be seen, the software checks the data for its consistency with the coded taxonomy, allows to select single or all tiers for further processing. To perform further processing to check the contingency of communication functions of speech and gesture in two discourse types and in individual discourse, we applied JAMOVI software (https://www.jamovi.org).

The analysis of functions in speech and gesture was performed with 10 samples, with 5 of them representing cinematic discourse and 5 representing interview discourse. Each sample lasted approximately 2–3 minutes (min 2:12, max 3:36). The number of annotation units (propositions and modal frames) varied significantly, with min 22, max 66 in cinematic discourse, and min 24, max 71 in interview discourse. The samples displayed finalized communicative events, for instance, in his monologue A. Batalov describes the way the film was made in the very place the interview is taken and presents his arguments on why the film has achieved great success. The total number of the units of analysis in cinematic discourse is 203, and 205 in the interview discourse respectfully, so the data are compatible.

The annotation procedure was carried out by three annotators with two annotators working with the interview discourse, and one annotator working with the cinematic discourse. Then the annotated samples were subjected to crosscheck and Cohen’s Cappa statistical coefficient (Landis & Koch, 1977) was applied (https://idostatistics.com/cohen-kappa-free-calculator/#risultati) to evaluate the agreement between the annotators to verify the validity of results.

We processed the Cohen’s Cappa separately for two discourse types. Since 56 functions of speech and gesture were annotated and the number of units was 203 and 205 correspondingly, we received a total number of annotation responses equal to 11,368 in cinematic discourse, and 11,480 in interview discourse. In terms of cinematic discourse, both groups of judges agreed to decide 1402 cases in the positive and 9866 cases in the negative with 90 cases decided in the positive by the first annotator and 439 by the other annotator group. The agreement coefficient is 95.52%, and Cohen’s k = 0.82, which is almost perfect agreement. In terms of interview discourse, both groups of judges agreed to decide 1478 cases in the positive and 9720 cases in the negative with 211 cases decided in the positive by the first annotator and 387 by the other annotator group. The agreement coefficient is 94.93%, and Cohen’s k = 0.8, which is also almost perfect agreement. We then voted for including the functions (since there were three annotators) and the final results are the following: the total activity of speech and gesture functions is 863 in cinematic discourse (728 in speech and 135 in gestures), and 1118 in interview discourse (826 in speech and 292 in gestures).

In Tables 3, 4 we give the results of contrastive function activity in speech and gesture in two discourse types.

The difference in all functions of speech and gesture seems significant but it is not statistically verified. With F(1, 14) = 0.085 at p = 0.775, we cannot reject the null hypothesis that there is no difference between the functions in cinematic and interview discourse. However, the variance in single functions (56 functions) is significant (F(1, 110) = 4.92, p = 0.027), which proves that specifying functions was an effective solution. It is noticeable that significant differences are observed in gesture only if we consider speech and gesture separately. With F(1, 26) = 7.14, p = 0.008 in gesture and F(1, 26) = 0.947, p = 0.33 in speech, we can claim that gesture distribution is of higher importance in more and less aesthetic discourse multimodal construal.

Therefore, we now move on to discussing single functions of gesture in two discourse types in more detail. Figure 7 gives the mapped gesture profiles of the gesture functions 301–314 which represent 4 basic functions.

Since these functions might be contingent on the speech functions, we performed regression modelling to reveal the predicting gesture functions in multimodal discourse. Regression modelling is an efficient method to cope with the problem of mixed effects of functions which is a typical case of construal in speech. However, we frequently deal with a problem of aliased coefficients. In the current study, however, these effects were surprisingly scarce which means that functional instrument works well for the needs of multimodal discourse analysis. We will present the model performance summary statistics for the most active functions, first in cinematic discourse, next for the interview discourse.

In cinematic discourse, two gesture functions with the highest activity are the pragmatic functions, Discourse emphatic (310), and Expressing attitude / evaluation (311). Their model performance statistics is given in Table 5. There were 36 non-aliased functions, in the table we give the statistics on the best predictors only.

As seen from Table 5, the best predictors of Discourse emphatic gestures are Specification, Accentuated time and Emphasizing discourse component which relate to description. In terms of Expressing attitude gestures, the best predictors are (Emotional) assessment, Threat, Stating reasons, consequences, conditions which relate to argumentation, and Achievement relating to description. The results show that the predictability of Regression model is quite high (R² = 0.341 and 0.359).

In the following example (Figure 8) the actor cites Ivan Karamazov from F. Dostoevsky’s “The Brothers Karamazov”, who denied accepting the fact that the language of arms and guns was the only means by which the dialogue with the Tsar could take place. Figure 8 illustrates two propositions introduced by the actor. In proposition (1) Все не верил ((He) still didn’t believe) the actor emphasizes the discourse component in description, whereas in proposition (2) the actor specifies the unwillingness to perform the action (Не хотел верить (He) didn’t want to believe). It is to point out that while introducing both propositions the actor used Discourse emphatic gestures to stress the importance of the ideas he was delivering to the pupils.

The best predictors for Expressing attitude / evaluation gestures are the pragmatic functions of argumentation and only one function of description. Figure 9 demonstrates an interesting example of the way the character (performed by Yu. Nikulin) introduces the phrase Не дай Бог ребята услышали бы (God forbid the fellows would hear (Yu. Nikulin, cinematic discourse), which has multiple pragmatic functions of argumentation (Emotional assessment and Stating reasons, consequences, conditions) and two pragmatic functions of description (Achievement and Accentuated action or state). The proposition co-occurs with two types of gestures: Expressing attitude / evaluation and Pointing, which on the one hand intensify the attitude of the character towards the message sent in the proposition, and, on the other hand, the character makes a reference to those who took part in the action.

The received data may provide evidence in favor of a specific functional feature of aesthetic multimodal discourse, which is the rigid (perhaps, because much practiced and rehearsed) correspondences of different gesture functions with different types of speech functions in argumentation and description. We will check this assumption contrasting the results with the model performance statistics of the interview discourse.

Three gesture functions with the highest activity are two pragmatic functions, Discourse representational and Expressing attitude / evaluation; and also one Adaptive function, Self-adaptors. We will present the results on only two functions, Expressing attitude / evaluation (311), which is the same with the cinematic discourse, and Self-adaptors (313), which is of a different function type not active in the cinematic discourse. Their model performance statistics is given in Table 6. There were 39 non-aliased functions, in the table we give the statistics on the best predictors only.

As seen from Table 6, results show that the predictability of Regression model is lower in both cases than the predictability in cinematic discourse (R² = 0.282 and 0.256). The best predictors are fewer; the best predictor for Expressing attitude/evaluation in gestures in interview discourse is Self-correcting, see Figure 10:

The sample given in Figure 10 demonstrates that the actor (Yu. Nikulin) expresses some level of uncertainty while proposing the argument that the troops were attacking and self-corrects the argumentation saying that it was the reconnaissance in force. While expressing the idea the actor uses Expressing the evaluation gesture which could symbolize his evaluation of the actions taking place, basing on the information he had (but this piece of information possibly requires verification). This is perhaps the least expected choice since it does not manifest high speech pragmatism to comply with the gesture pragmatism.

There are no other predictors, which makes us think that the interview discourse being less staged allows much more freedom in multimodal function frameworks. In terms of Self-adaptors, the situation is less peculiar. The best predictors are Opinion and Intersubjectivity, which are argumentation Pragmatic and Discourse-structuring functions, and the appearance of Self-adaptors in such situations was most expected.

The example shown in Figure 11 illustrates the gestural behavior of the speaker (S. Bondarchuk) expressing opinion and intersubjectivity.

As far as S. Bondarchuk introduces his own opinion, he is gaining control over the situation by using Self-adaptors.

The contrastive results presented here may suffice to deduce several structural types of function frameworks.

REPLICATION – SINGULARITY. This structure integrates several functions of the same type. It can have three types. Type 1 describes the framework of intensifying the same function in both modalities. In the present study we observed Type 1 with pragmatic function which can be demonstrated in both speech and gesture. Interestingly, in cinematic discourse the gesture functions (at least the ones we explored in terms of predictability) seem to comply with similar functions in speech as if to enhance them additionally, whereas in the interview we did not observe this effect. Therefore, cinematic (and more aesthetic) discourse mostly exploits REPLICATION function structure. Type 2 describes the framework of intensifying the same function within one modality. This happens when the same function is activated several times within one proposition or modal frame. Surprisingly, cinematic discourse shows less preference for multiple pragmatic function expression in description, with very scarce pragmatism of the subject (in contrast with the interview discourse). This result might be explained by the spontaneous nature of the interview discourse and multiple pragmatic patterns chosen by the speakers to describe the events and their participants. Type 3 describes the framework of intensifying multiple sub-functions of one function. The data have shown that this function framework is frequent in both discourse types. Therefore, cinematic discourse mostly exploits REPLICATION function framework, whereas interview discourse shows preference for SINGULARITY.

REINFORCEMENT – SELF-SUFFICIENCY. This structure integrates several contingent functions in multimodal discourse. As we have shown above, it has 2 types of realization. Type 1 describes the framework of reinforcing the function of one modality with the functions of the other modality which display rigid contingency. REINFORCEMENT is more frequently present in cinematic discourse, whereas interview discourse has fewer contingent functions. Interview discourse spontaneity may also account for it. Type 2 describes the framework of reinforcing the functions within one modality. We may notice here that neither discourse exhibits this type of REINFORCEMENT. The functions do not display alignment effects, therefore their individual input into the process of multimodal discourse construal needs no other evidence: these functions are self-sufficient in determining the communicative potential of the modalities considered.

LIMITATION – EXTENSION. This structure is of a different type; it describes the function framework of constraining and widening the number of functions which are manifested in multimodal discourse. Therefore, this structure can be detected only in contrastive analysis. Type 1 appears when the number of functions is significantly smaller in one of the modalities. Unexpectedly, this type is more typical of cinematic discourse which employs far fewer gestures of specific types, for instance, of Self-adaptors. They are mostly present in interview, since they are the most natural and uncontrolled movements, used in our speech when the speaker might feel more stressed and in need of exercising control over the situation (Ekman, 2004) or they can be markers of turn taking (Żywiczyński et al., 2017). Type 2 appears when this structure is manifested in both modalities. In this study, we revealed that cinematic discourse uses fewer pragmatic speech functions of description and fewer pragmatic functions of Expressing attitude / evaluation.

Next, we will find out whether these function frameworks appear in individual multimodal discourse of the 5 actors. This analysis might also reveal some other function frameworks relevant for assessing the aestheticism of cinematic discourse.

At the second stage we turn to the individual differences in multimodal discourse which may be found if we contrast speech and gesture in two discourse types, cinematic and interview, performed by the same actors. This procedure allows to detect the individual variations which may or may not fall within the function frameworks of either cinematic or interview discourse and consequently, may help specify them and show them in more detail. To proceed, we contrast the multimodal discourse profiles within each discourse. The idea that we entertain is that the profile differences will display some similar tendencies which will suffice to claim that they are more typical of more aesthetic discourse.

At the first step, we identify whether the differences in the functions distribution by each actor are significant and therefore dependent on individual multimodal discourse. In cinematic discourse, variance analysis in three sets, speech functions in argumentation, speech functions in description, functions in gesture revealed the following results: for argumentation (101–123) F(4, 22) = 19.2 at p < .001, for description (201–219) F(4, 18) = 23.9 at p < .001, for gestures (301–314) F(4, 13) = 15.7 at p = 0.003. In interview discourse, variance analysis revealed the following results: for argumentation (101–123) F(4, 22) = 17.9 at p = 0.001, for description (201–219) F(4, 18) = 36.4 at p < .001, for gestures (301–314) F(4, 13) = 19.9 at p < .001. The results suffice to claim that in both discourse types, the functions of speech and gesture display significant variance; interestingly, the highest variance values appeared in the functions of description, especially in the interview discourse. The lowest variance was in the functions of gestures in cinematic discourse, which specifies our earlier findings in the way that gesture functions are not only synchronized with similar speech functions but are also more restricted and allow fewer alternatives in their selection.

At the second step, we find out whether the speech and gesture functions distribution is similar with the actors. To find the answer, we contrasted the function distribution in three data sets (functions in argumentation, functions in description, and functions in gesture) in both discourse types, cinematic and interview. To do this, we introduced a grouping variable, 0 and 1 (for cinematic and interview discourse). In Tables 7, 8 we present the results for each of the three datasets and for each actor.

In terms of functions in argumentation, there is one actor, V. Tikhonov, whose speech behavior displays significant variance. This may happen only because his monologues display a different communicative character which is true since his speech patterns clearly display fewer opinion and emotional assessment in interview. However, in terms of description functions, all 5 actors displayed similar speech behavior. The most interesting results appear with the distribution of gesture functions. Three actors out of 5 displayed significant differences in gesture functions’ selection in cinematic and interview discourse. The highest variance is attributed to Yu. Nikulin and R. Bykov. The results point out that these are mostly gestures which manifest variance in multimodal discourse of actors in more and less aesthetic discourse and not the speech functions. In a way it supports our previous findings that the inventory of gesture functions in cinematic discourse is more restricted, and there exists higher alignment between gesture and speech functions in cinematic discourse.

Therefore, it is possible that these are particular / specific gesture functions that may serve as indicators of aestheticism of cinematic discourse. To find out which functions may fulfill this role and appear in gesture behavior of all the 5 actors, we now turn to analyzing the gesture functions in the actors’ discourse. To perform it, we again introduce a grouping variable, 0 and 1 (for cinematic and interview discourse). In Table 10 we present the variance results for each gesture function.

The results show that there are three gesture functions that have statistically different distribution with 5 actors and in two discourse types, they are Holding gestures, Contact establishing gestures and Self-adaptors; Discourse representational gestures also play a role. It means that other gesture functions work similarly in both discourse types, and the aesthetic specificity of gesture in cinematic discourse mostly lies within these four types.

The example given in Figure 12 illustrates gestural behavior of the actor introducing one proposition that has multiple pragmatic functions of description.

At the beginning of the sentence (1) while describing the situation, Yu. Nikulin uses Contact establishing gestures to attract attention of the audience to the case being described. Afterwards while giving the description of the fear (2) perceived by him (the German soldier) the actor uses Self-adaptors. At the end of the proposition the actor uses the Holding gesture to show the silence (3), in which the described situation took place. This example demonstrates the variety of co-speech gestures functions used in one proposition.

We selected a fragment employing two propositions with multiple pragmatic functions of description in cinematic discourse with the same actor (Figure 12) to contrast it with the situation given in Figure 11.

In comparison with interview discourse, Yu. Nikulin does not use many gestures while giving the description in cinematic discourse. In proposition (1) no gestures were identified, whereas in proposition (2) the actor uses one gesture with the function of expressing attitude / evaluation while describing the way a lot of people were forced to sleep in one house.

It is also noticeable that Holding gestures, Contact establishing gestures, Self-adaptors and Discourse representational gestures irregularly appear in cinematic discourse, being common only for the discourse of interview. At the same time, other gestures appear frequently in both discourse types, for instance, Discourse emphatic gestures. Considering this fact, we distinguish one more function framework structure revealed in contrastive analysis of individual difference – that is REGULARITY – IRREGULARITY. It describes the function frameworks with regular or irregular function distribution within the multiple samples of multimodal discourse of the same or contrasting types. At this stage we cannot account for providing a sufficient description of this structure since the research data need to be much extended to become relevant. Nevertheless, even with these data we still managed to detect the regular patterns of function distribution which may evidence in favor of this structure. The function distribution in cinematic discourse has shown more regularity since it was better predicted. As we have revealed, the lowest variance was observed with the functions of gestures in cinematic discourse and this variance displayed regularity among all 5 male actors.