RSS feed

The automated interlingual mapping effect in trainee subtitlers

Mikołaj Deckert, University of Łódź


The paper reports and discusses the findings of an empirical study conducted with trainee subtitlers. As a result of the study, a decision-making pattern has been recognised which is attributed to a cognitive mechanism termed ‘automated interlingual mapping’ (AIM). The AIM effect is observed in English-to-Polish subtitling data and is postulated to occur when one prototype-based category, and then one verbal node that represents this category, is automatically activated and prioritised as an interlingual match in translation. As the paper seeks to show, the effect is especially salient and intriguing when that prototype/node gets used in the stead of a prototype/node which is more consistent with contextual evidence, and therefore a more optimum target variant.


Automaticity, conceptual structure, construal, conventionalisation, interlingual subtitling, language and cognition, linguistic representation

1. Introduction

Language use is characterised by high automaticity (cf. Pawley and Syder 1983, Bolinger 1979, Wray 2002). The linguistic signals we produce every day are hardly ever fully novel compositional configurations. Rather, we naturally re-use pre-existing building blocks and combinational patterns.  With subtitling as a case in point, the paper addresses automaticity in terms of Langacker’s (2007, 2008) ‘conventionalisation’ by discussing an effect that can be understood as rooted in, and then as a translational elaboration of, Langacker’s descriptive construct. Namely, the inquiry documented here identifies an associative memory pattern in trainee subtitlers which can be termed ‘automated interlingual mapping’ (AIM). An examination of English-to-Polish subtitling output focused on the case study of the construction ‘to like + somebody’ shows that in certain decision-making contexts trainee translators display a tendency to opt for a particular ST-TT construction mapping. The principal argument put forward here is that it is due to its high conventionalisation that this mapping is easy to access and its routinisation is such that it is employed in translation despite different types of evidence contradicting its use. My objective is to describe the AIM effect for the ‘to like + somebody’ construction in subtitling, suggest its motivation and, vitally, discuss some of its implications, both for the understanding of cognitive processing in subtitling and the didactics of Audiovisual Translation as well as translation in general.

2. Intralingual conventionalisation

Before we proceed with a translation-centred discussion, I will briefly talk about Langacker’s notions of entrenchment and conventionalisation as they are central to the argument. They were posited within Cognitive Grammar (Langacker 1987, 2007, 2008), a subfield of the functional framework of Cognitive Linguistics ─ approaches that have been developing since the late 1970s, in which “the formal structures of language are studied not as if they were autonomous, but as reflections of general conceptual organization, categorization principles, processing mechanisms, and experiential and environmental influences” (Geeraerts and Cuyckens 2007: 3).

Entrenchment is about the cognitive automation of a particular fragment of language structure in a language user, while conventionalisation is about the mechanism as it operates across a given language community. The more conventionalised a formation is, the more automatic its production and comprehension. To illustrate this on a very basic level, when confronted with a visual stimulus we have at our disposal prefabricated linguistic components to pair up with the fragments of conceptual structures that get activated. Therefore, when one wishes to call their interlocutor’s attention to an element in the surrounding environment, they are more likely to say “What a beautiful tree!” or “What a beautiful oak!” than “What a beautiful plant comprising a trunk, branches and leaves!” or “What a beautiful quercus!”, unless the utterance is crafted as ostensively marked or is part of a specialised discourse. That is to say, for the sake of minimising the addressee’s cognitive computing and to cut down on one’s own effort, the speaker accesses the mental inventory of pre-existing pairings of conceptual and linguistic material ─ by and large shared by interactants ─ and selects those portions of the material that, at least to the best of the speaker’s estimation, are communicatively most optimum.

Conventionalisation operates on different levels of linguistic and conceptual organisation. For instance, in line with our linguistic intuition and as is evidenced by corpus data, it would be more conventional to talk about ‘heavy rain’ than ‘strong rain.’ Yet, this will not work analogously for natural phenomena like wind which combines into a more conventionalised expression with ‘strong’ than with ‘heavy.’ Then again, those will function differently across languages and in Polish ‘rain(ing)’ would collocate with adverbs such as mocno [‘strongly’/’powerfully’] rather than ciężko [‘heavily’]. The notion of conventionalisation can be used here very productively because the difference between those pairs of expressions might not easily lend themselves to precise and systematic comparative description. Whether we say ‘strong rain’ or ‘heavy rain,’ arriving at an intended interpretation is ultimately feasible, if not equally easy.

On another level, adjustments along the continuum of ‘granularity’ ─ one of the parameters of how we (choose to) present conceptual content (cf. Langacker 2008: 55-57) ─ will notably condition conventionalisation. Therefore, (1) will be deemed more conventional than (2) when it comes as a response to a question such as “Are you busy?”:

(1) I am having breakfast.
(2) I am buttering my bread.

In addition to showing that conventionalisation depends on construal ─ in the above case the construal’s granularity ─ this shows that a given expression will be assessed as differently conventionalised depending on the context. Thus, while in the above example (1) would likely be more conventional, a more fine-grained expression such as (2) would be expected if the speaker were making an instructional cooking video in which case it would be unconventional to say “I am cooking.”

3. From conventionalisation to automated interlingual mapping

The AIM effect that this paper sets out to illustrate and discuss can be seen as a cross-systemic elaboration of conventionalisation as it operates within a single language system. The effect has been identified and will be addressed on the basis of the English construction ‘to like + somebody’ ─ or, to be more accurate, the English linguistic representation together with the concepts that it evokes ─ and its Polish subtitling counterpart(s). Because the lexeme ‘to like’ is among the most basic vocabulary items of the English lexicon and is acquired early on, it could be considered unproblematic in translation. However, in English-Polish transfer the ‘like + somebody’ construction constitutes a complex case as no one-to-one rendition will be universally functional.

While ‘to like’ can be rendered in Polish in a range of ways, two nodes around which constructions are built emerge as predominant – lubić and podobać się. Used with human objects, these two nodes (N1 and N2) – in the linguistic sense ─ correspond to two distinguishable prototypes (henceforth P1 and P2) ─ on the conceptual level (cf. Rosch 1975). The considerable difference between those prototypes is that the former profiles a favourable disposition devoid of an explicitly romantic element, while the latter signals physically-grounded attraction and therefore carries a host of sexually-oriented (if differently dormant) implications.

This is a very interesting case of conceptual-semantic asymmetry found between English ‘liking somebody’ and candidate Polish variants which inadvertently evoke a different, narrower, portion of conceptual content. In other words, the English-to-Polish subtitler has to select a target variant that points to a more distinct meaning than originally envisioned. What can be termed ‘incommensurability’ of language systems (cf. Lakoff 1987; Lewandowska-Tomaszczyk 1987, 2010) surfaces on a number of levels. To come back to the notion of granularity addressed above, it is conventionally coded in different ways in different languages. To take the English verb ‘to get’, a simple sentence like “He got a cup of coffee while waiting for the plane” need not be completely straightforward as far as the choice of verb in translation goes. While the English construal is coarse-grained with respect to the manner in which the agent obtained the coffee, a conventional Polish verbal construction would canonically need to specify whether he bought it or perhaps whether there was complimentary coffee at the airport. A similar shift might be effected by how much passivisation is conventionally employed by the users of languages involved, or with respect to nominalisation. In fact, the problem of gradable conceptual-semantic commensurability is addressed ─ if with the use of a different set of descriptive constructs ─ already in the classic works by Nida (1964) and Nida and Taber (1982) as they discuss hierarchical organisation of meanings and cases of overlap between semantic areas in English (cf. Hatim and Munday 2004).

4. The study
4.1. Design of the study

To identify patterns in translation behaviour, 58 subjects were asked to produce subtitles for a film passage featuring the ‘to like + somebody’ construction. A total of 57 participants considered here were native speakers of Polish taking part in an AVT course as part of their BA degree in English. Importantly, the ST fragment was not limited to the utterance containing the structure in question. In addition to providing context ─ whose role in subtitlers’ decision making I will address further on ─ this was to ensure that the subjects could identify a range of potential translation pitfalls (if any), the premise being that their behaviour with respect to a predefined translation problem would differ from their behaviour if no problem were postulated in advance. It is also relevant that because the subjects produced their translations in writing, on answer sheets, there is some lasting evidence of their decision-making process that can be used to complement introspective post-task feedback (cf. section 4.2.). It should also be mentioned that no explicit information was provided as to the maximum number of characters per line that could be used in the assignment. This made it possible to control ─ at least to some extent – for the ‘spatio-temporal constraints’ variable which could otherwise impact the target variant selection more considerably, thus distorting the results with respect to the examined variable.

The fragment under scrutiny comes from the 2013 film About Time written and directed by Richard Curtis. The scene relays a conversation that the film’s protagonists, Mary (Rachel McAdams) and Tim (Domhnall Gleeson), have in a London street. Their exchange follows a chance dinner they and their two friends had at a ‘pitch black’ restaurant where they met for the first time.

The table below presents the ST conversation that the subjects were requested to subtitle. The presentation method used here draws on the multimodality theory (cf. Baldry and Thibault 2006) to render, at least to some extent, the polysemiotic and dynamic nature of the material.


Source text


No dialogue

Mary emerges from the restaurant, slowly goes out and nearly trips stepping down to the sidewalk. She smiles. Tim waits outside. They walk towards each other uncertainly.

Tim: Hi.
Mary: Hi.

They speak with uncertainty. Mary laughs nervously and smiles at Tim.

Mary: Where… Where's...
Tim: Ahh… The… She
and Jay just...      

They speak nervously. Mary looks around for their friends. Tim gestures the direction they went.

Tim: She took him to... I… I don't know.

Tim speaks hesitantly, he shakes his head.

Mary: Right.

Mary smiles as she speaks.

Mary: Oh, well, I guess I'd better...

Tim: Would it be very wrong if I asked you for your number?

Tim shakes his head and frowns.

Mary: No.

Mary shakes her head and smiles.

Tim: Just in case I ever
had to call you about...

Tim reaches for his phone.

Mary: …stuff.

Mary says the word emphatically.

Mary: Okay.

Mary laughs nervously as she responds.

Tim: Would you...

Tim passes his phone to Mary.

Mary: It’s ‘Mary’.

Mary speaks in a whispering voice.

Tim: Mary.

Tim shrugs his shoulders and nods as he repeats the name. Mary laughs.

Mary: Okay.

Mary smiles and hands the phone back.

Tim: I thought this phone
was old and shit,

Tim speaks in a confident voice.

Tim: but suddenly it's my most valuable possession.

Mary: You really like me?

Tim’s hand holding the phone with Mary’s number displayed

Mary smiles at Tim and shakes her head.

Mary: Even my frock?

Mary looks down on her dress.

Tim: I love your frock

Tim looks at the dress and smiles.

Mary: And my hair?
It's not too brown?

Mary strokes her hair.

Tim: I love brown.

Tim shakes his head.

Mary: My fringe is new.

Mary places her hand on her forehead.

Tim: The fringe is perfect.
Fringe is the best bit.

Mary smiles and closes her eyes.

Table 1.

4.2. Data analysis

Let us now focus on the ‘to like + somebody’ construction featuring in Mary’s question “You really like me?”. As has been discussed above, the choice is between two prototypes ─ that of favourable – yet non-romantic/sexual ─ disposition (P1) and that of physical/aesthetic pleasantness (P2). The subtitler has to opt for either prototype, but once the choice is made, there is still a range of actual linguistic realisations clustered around the prototypes to be selected from. The distribution of different versions of the subtitle is presented below.

target variants with N11

literal back-translation into English

no. of occurrences


Naprawdę mnie lubisz?

Really you like me?



Lubisz mnie?

(Do) you like me?



(Na) serio mnie lubisz?

For real you like me?



Naprawdę mnie polubiłeś?

You really got to like me?



Table 2.

target variants with N2

literal back-translation into English

no. of occurrences


Naprawdę ci się podobam?

Really you find me attractive?



Podobam ci się?

You find me attractive?



Naprawdę ci się spodobałam?

Really you started to find me attractive?



Table 3.

The first interesting finding is that despite the range of available translations ─ given that in languages like Polish and English there are clearly innumerable ways to express a common concept such as the one at hand ─  the homogeneity is high both with respect to the choice of verbal node2 ─  with only two options across 57 individual texts ─  and with respect to the actual linguistic realisations based on those nodes, with relatively few variants and one variant appearing unchanged in 84.37% of N1-based TTs and in 82.60% of subtitles featuring N2. This lends credence to the hypothesis that translational behaviour is highly conventionalised, and invites questions as to the extent to which such conventionalisation (its nature and degree) overlaps with conventionalisation in authentic language use. That is to say, how likely ─ in the linguistic and interpersonal sense ─ is a norm-abiding (cf. Toury 1995) native speaker of Polish to ask a question using an N1 or N2 construction in analogous circumstances? Then, in turn, we have to ask how likely a native speaker of English is to ask the question we find in the original, which brings us to the overall issue of how much of an emulation of genuine language and social interactions the filmic language is (cf. Chaume 2001, 2004). Be that as it may, in the examined sample, the English construction, due to its lower granularity, will be less socially threatening than the Polish variant, especially a N2-based one.

As far as the other central point of the present discussion is concerned ─ i.e. the choice between P1 and P2 within the English concept of ‘liking somebody’ ─ the results are as follows. Out of the 57 translations, ‘N1’ was opted for 32 times and ‘N2’ 23 times (cf. Figure 1). With the N1-based constructions amounting to 56.14% and N2-based constructions to 40.35%, there is a clear tendency supporting the AIM effect hypothesis.

Figure 1.

The competition between P1 and P2, as the prototypes are textually represented by N1 and N2, is tangibly illustrated in the output of 5 subtitlers. In that dataset, 3 subtitlers first wrote down both an N1 and an N2 construction, and eventually – as a result of critically assessing the variant produced via AIM – decided to cross out the N1 construction and opt for a subtitle with N2. Further 2 subtitlers left two versions, in this way signalling their inability to decide, but all the more significantly testifying to the challenge of conceptual resolution and the meta-cognitive procedure involved in their translation process.

4.3. Discussion
4.3.1. The AIM effect against evidence

The findings indicate that the AIM effect can remain undefeated by different types of evidence. Those types can be overall referred to as ‘contextual’ in the broad sense but for ease of discussion they will here be further grouped into three interlinked and interdependent types: polysemiotic, situational, and co-textual.

Starting with the first type of clues, in subtitling, as a mode of Audiovisual Translation, the semiotic structure of messages is more complex than in some of the other types of translation. The text – i.e. the spoken linguistic material in the form of dialogues or monologues as well as written language – is but one of the components of a semiotic conglomerate. For instance, Tomaszkiewicz (2006) talks about ‘semiotic complexes’ rather than ‘texts proper’. In this vein, a recent and productive approach was postulated under the heading of multimodal analysis (Baldry and Thibault 2006). The premise for this line of research is that the non-linguistic strata have to be incorporated into the investigation of meaning. The need for considering multimodality in AVT analyses has already been discussed and demonstrated quite extensively (Taylor 2003, 2004; Chuang 2006, 2009; Pettit 2007, Perego 2009 and Bączkowska 2011).

In the scene under scrutiny the audio and visual clues can aid the subtitler – body posture and facial expressions, for instance, hint at the characters’ nervousness and uncertainty that conglomerate with the linguistic signal, creating a polysemiotic message. Utterances are combined with movements like gesturing to an article of clothing (‘frock’) or other elements of appearance (‘fringe’) while talking about it, which could additionally guide the subtitlers and can be viewed as evidence against the N1 variant that most subtitlers chose in the study.

The other source of input at the subtitlers’ disposal is context. This type of evidence will be inextricably bound up with the polysemiotic type as the viewers’ assumptions about parameters like the situational setting heavily depend on visual and sonic stimuli. What is meant by ‘context’ goes beyond ‘when’ and ‘where’ to include also what we know about the film’s characters, about the plot, about the location where action takes place, i.e. the entirety of our encyclopaedic knowledge and assumptions about the state of affairs in the film, incorporating what we know about social behaviour3. The events that took place before the sequence in question constitute an important element of the contextual evidence. Indeed, in the case of the fragment discussed here, the preceding scene must be considered to make a well-informed choice as far as the ‘like’-based construction goes. As has already been mentioned, the two characters met for the first time shortly before the conversation happens. The critical piece of evidence is that their first meeting took place in complete darkness and they were not able to see each other. Given the cheerful, jocular and relatively long conversation they had at the dark restaurant ─ the elapsing time is indicated on the screen by captions from ‘7:48 pm’ to ‘10:37 pm’ ─ at the point of leaving the restaurant they had already formed a kind of judgement about each other. Since they could only hear each other’s voices, that judgement would be primarily in terms of the non-visual, personality-based type of ‘liking’ (P1) singled out above. Therefore, it stands to reason that when they get to see each other for the first time, what Mary would be linguistically representing more plausibly is the aesthetic prototype of ‘liking’ (P2) serving to express her anxiety about whether the mutual positive personality-based disposition ─ ascertained by both already at dinner ─ is coupled with physical attractiveness in Tim’s view.

As regards the co-textual subtype of evidence, it is again convergent with other types mentioned, for instance because the linguistic component can de facto be seen as a subcomponent of the contextual type of evidence mentioned above. To elucidate the import of co-textual clues, let us take a closer look at the utterance that follows the ‘to like’ construction.

Mary: You really like me? Even my frock?

The second question does not explicitly reiterate the verb but because it refers to the earlier question, in the second one ‘to like’ is interpreted as combined with ‘frock.’ If the translator decides not to include a (different) verb in the second inquiry, the choice of verb in the first one has to be such that it can work with both the objects – ‘me’ and ‘frock.’ And this again is an argument against the N1 variant, at the same time legitimising N2 as a workable rendition.

In line with what was posited in section 3, lubię here points to a more psychologically-grounded positive disposition that takes longer to form. In turn podobać się, foregrounding more the surface properties of the entity commented on, can be used to reflect upon one’s instantaneous impressions. To employ N1 with ‘frock’ as in “Lubię tę kieckę,” where ‘kiecka’ is an informal word for a dress, implies having developed a more complex and long-lasting positive disposition, which could take some time, and therefore having seen the frock already prior to the occasion in question. In fact, the construction could typically be expected from the owner/user of the frock because she is the one who is most likely to have developed this kind of P2-based attitude towards the article of clothing, one that goes beyond mere aesthetics but possibly also incorporates properties like its diaphanous fabric which makes it practical on a hot day, non-iron quality or ease of combining it with other articles of clothing.

Whereas “Podoba mi się ta kiecka” ─ employing N2 ─ would canonically be expected as an aesthetically-motivated remark, for instance from someone who has just seen the piece. Some of the construction’s other variants, possibly even more conventional ones, would be a verbless phrase ‘Ładna kiecka’ [A pretty frock] which again highlights the surface aesthetic element. It has to be noted, though, that the N2 construction could well be utilised in the present tense to talk about a more permanent state of finding someone or something aesthetically pleasing in general, not just on a particular occasion. This versatility once more makes the N2 construction safer as a target variant.

For the cases where subtitlers did choose the N2 variant for “You really like me?”, a possible scenario is that once they got to deal with the second question (“Even my frock?”), and then the follow-up questions along the same lines, pertaining to the colour of Mary’s hair and the haircut – they realised N1 was not optimum there and were prompted to backtrack their decision and revise the previous subtitle. That in over 56% of cases this did not take place indicates again how hard the AIM effect can be to overcome ─  either the very moment the TT is cognitively constructed, or when it is written down, already in a more controlled manner, or even later when additional grammatical hints become available from the accompanying linguistic data.

4.3.2. Formation of associations

What motivates the mapping’s automation for the ‘to like + somebody’ construction is a major question. One possible response could be how the foreign language is acquired and learned. The fact that lexical items and constructions are often presented and then memorised as clear-cut pairings, for instance as lists of words with their translations, could be part of the explanation. In that vein, it is easier to note and then commit to memory a link between ‘to like’ and ‘lubić’ than the alternative ‘podobać się’, because the use of ‘lubić’ is more similar to ‘to like’ in terms of syntax and consequently attention distribution (cf. Deckert forthcoming). A sentence “Tim likes Mary” can simply be translated into Polish as “Tim lubi Mary” whereas a construction employing ‘podobać się’ ─ “Timowi podoba się Mary” [To Tim Mary is attractive] or “Mary podoba się Timowi” [Mary is attractive to Tim] – no longer construes ‘Tim’ as an agent. The similarity in spelling – the initial ‘l’ and comparable word length ─ can, too, contribute to the association between ‘like’ and ‘lubić.’

Likewise, the interlingual associative link between ‘to like’ and ‘lubić’ is clearly reinforced by some of the established translations. This is seen for instance in Facebook’s ‘Like’ option which is translated into Polish as ‘Lubię to!’ (cf. Figure 2.). Leaving aside the appropriateness of the button’s use in certain discursive situations ─ where it is intended as an expression of commiseration, solidarity etc. also with respect to posts whose subject matter one in fact ‘dislikes’ ─ this pairing again points to linguistic-conceptual asymmetry which nonetheless becomes less and less striking as the users are repeatedly exposed to it.

Figure 2.

The buttons ─ in the sense of the visual representation, a platform affordance they stand for, their function and  the perlocutionary effect they can bring about – enter into a relation of analogy which is carried over to the linguistic layer, thus establishing and reinforcing the cross-language match between ‘to like’ and ‘lubić’. The entrenchment of this match is visible in the noun ‘lajk’ ─ produced through reification and used in colloquial Polish expressions such as ‘mieć dużo lajków’ [to have a lot of likes].

5. Final remarks

The common, or even commonplace, character of ‘to like + somebody’ constructions in filmic language can mislead translators (-to-be) into assuming that those are semantically and conceptually uncomplicated. Yet, as I have attempted to demonstrate through the discussion of the AIM effect, the constructions hint at a complex problem which may be a source of translational difficulty, therefore necessitating sufficient sensitisation at an early stage of subtitler training. Speaking more broadly, instances such as that of ‘liking + somebody’ and its interlingual renditions prompt questions about whether the lack of one-to-one linguistic counterparts across languages equates with conceptual mismatches. In that respect subtitling quality assessment can be a point of departure for a discussion about the nature of links between language and cognition.

The subtitler chooses between linguistic representations which construe conceptual content differently ─ from one another but also from the ST ─ and so differently instruct the viewers in their process of constructing meanings (cf. Deckert 2013). The implications of that choice will vary but they can be far-reaching. Indeed, in the analysed example the construction marks the film’s linguistic-conceptual turning point by framing the relationship between the main characters, and the relationship itself is a superstructure of the entire movie, one onto which the film’s plot and other characters are fitted.

With respect to the AIM effect, a natural question is whether it would be observed in professional subtitlers, however fuzzy the boundaries of that category will be. It can be of some significance that the Polish subtitles available for the DVD release of ‘About Time’ use the N2 node. With the range of lexico-grammatical variants to choose from around P2, this confirms the hypothesis that translational behaviour tends to display high conventionalisation. On the other hand, this could also suggest that the AIM effect is endemic to trainee subtitlers. To verify that suggestion, however, a larger sample of professional subtitles would need to be analysed. Interestingly, though, if a more general AVT perspective is taken, ‘experience’/’professionalism’ fails as an explanatory variable because the same DVD has the N1 node in the Polish voice-over.

  • Baldry, Anthony and Christopher Taylor (2002). “Multimodal Corpus Authoring System: multimodal corpora, subtitling and phrasal analysis.” Mark Maybury and Jean-Claude Martin (eds) (2002) Multimodal Resources and Multimodal Systems Evaluation. LREC Congress, Post-conference Workshop Papers. Las Palmas, 45‑51.
  • Baldry, Anthony and Paul. J. Thibault (2006). Multimodal Transcription and Text Analysis. A multimedia toolkit and coursebook. London and Oakville: Equinox.
  • Bączkowska, Anna (2011). “Some remarks on a multimodal approach to subtitles.” Linguistics Applied 4, 47‑65.
  • Bolinger, Dwight (1976). “Meaning and Memory.” Forum Linguisticum 1(1976). 1-14.
  • Chaume, Frederic (2001). “Models for research in Audiovisual Translation.” Babel 48 (1), 1‑13.
  • (2004). “Discourse Markers in Audiovisual Translating.” Meta 49(4), 843‑855.
  • Chuang, Ying-Ting (2006). “Studying subtitle translation from a multi-modal approach.” Babel 52/4, 372‑383.
  • ─ (2009). “Subtitling as a Multi-Modal Translation”, Gilbert C. F. Fong and Kenneth K. L. Au (eds) (2009) Dubbing and Subtitling in a World Context. Hong Kong: Chinese University Press: 79‑90.
  • Deckert, Mikołaj (forthcoming) “Constrained meaning construction and attention re-allocation.” Meng Ji (ed) Empirical Translation Studies: Interdisciplinary Methodologies Explored. London: Equinox.
  • ─ (2013). Meaning in Subtitling: Toward a Contrastive Cognitive Semantic Model. Frankfurt am Main: Peter Lang.
  • Geeraerts, Dirk and Hubert Cuyckens (2007). “Introducing Cognitive Linguistics.” Dirk Geeraerts and Hubert Cuyckens (eds) (2007) Oxford Handbook of Cognitive Linguistics. Oxford: Oxford University Press, 3–21.
  • Hatim, Basil and Jeremy Munday (2004). Translation. An Advanced Resource Book. London and New York: Routledge.
  • Lakoff, George (1987). Women, fire, and dangerous things: What categories reveal about the mind. Chicago: University of Chicago Press.
  • Lewandowska-Tomaszczyk, Barbara (1987). Conceptual Structure, Linguistic Meaning, and Verbal Interaction. Łódź: Łódź University Press.
  • ─ (2010). “Re-conceptualization and the emergence of discourse meaning as a theory of translation.” Barbara Lewandowska-Tomaszczyk and Marcel Thelen (eds) (2010) Meaning in Translation. Frankfurt am Main: Peter Lang, 105‑147.
  • Langacker, Ronald W. (1987). Foundations of cognitive grammar. Vol. 1, Theoretical prerequisites. Stanford, CA: Stanford University Press.
  • ─ (2007). “Cognitive Grammar.” Dirk Geeraerts and Hubert Cuyckens (eds) (2007) Oxford Handbook of Cognitive Linguistics. Oxford: Oxford University Press, 421‑462.
  • (2008). Cognitive Grammar: A Basic Introduction. New York: Oxford University Press.
  • Nida, Eugene. A. (1964). Toward a Science of Translating. Leiden: E. J. Brill.
  • Nida, Eugene. A. and Charles R. Taber (1982). The Theory and Practice of Translation. Leiden: E.J. Brill.
  • Pawley, Andrewand Frances Hodgetts Syder (1983). “Two Puzzles for Linguistic Theory: Nativelike Selection and Nativelike Fluency.” Jack C. Richards and Richard W. Schmidt (eds) (1983) Language and Communication. London: Longman, 191-227.
  • Perego, Elisa (2009). “The codification of nonverbal information in subtitled texts.” Jorge Díaz Cintas (ed.) (2009) New Trends in Audiovisual Translation.Bristol: Multilingual Matters, 58‑69.
  • Pettit, Zoë (2007). “Translating verbal and visual language in The Piano.” Perspectives 15/3, 177‑190.
  • Rosch, Eleanor (1975). “Cognitive Representations of Semantic Categories.” Journal of Experimental Psychology: General, 104 (3), 192-233.
  • Sperber, Dan and Deirdre Wilson (1995). Relevance: Communication and cognition (2nd ed.) Oxford: Blackwell.
  • Taylor, Christopher (2004). “Multimodal Text Analysis and Subtitling.” Eija Ventola, Cassily Charles and Martin Kaltenbacher (eds) (2004) Perspectives on Multimodality. Amsterdam/Philadelphia: John Benjamins, 153‑172.
  • ─ (2003). “Multimodal Transcription in the Analysis, Translation and Subtitling of Italian Films.” The Translator 9(2), 191‑205.
  • Tomaszkiewicz, Teresa (2006). Przekład audiowizualny. Warsaw: Wydawnictwo Naukowe PWN.
  • Toury, Gideon (1995). Descriptive Translation Studies and beyond. Amsterdam/Philadelphia: John Benjamins.
  • Wray, Alison (2002). Formulaic Language and the Lexicon. Cambridge (GB); New York: Cambridge University Press.

About Time (2013). Dir. Richard Curtis. Translux and Working Title Films.


Deckert portraitMikołaj Deckert works as assistant professor in the Department of Translation Studies, Institute of English Studies, University of Łódź. His research is in translation, language and cognition, media discourse, linguistic pragmatics, as well as corpus linguistics. He is the author of a monograph Meaning in subtitling: toward a contrastive cognitive semantic model (2013) and a co-editor of volumes devoted to translation didactics as well as AVT.  He can be reached at


Note 1:
As they are noted here, the subtitling variants for N1 and N2 constructions introduce spelling uniformity for ease of illustration. For instance, ‘naprawdę’ was spelled as two words by some participants, which is not directly pertinent for the present discussion.
Return to this point in the text

Note 2:
The term ‘verbal node’ is used to integrate all occurrences of a verb in its different variants that will differ in aspect or tense. Therefore, for instance, ‘lubisz mnie’ [you like me] and ‘polubiłeś mnie’ [you got to like me] are both treated as centred on the ‘lubić’ verbal node.
Return to this point in the text

Note 3:
This formulation is close to that offered in Relevance Theory (Sperber and Wilson 1995).
Return to this point in the text