Meaningfulness, the unsaid and translatability: Instead of an introduction

Abstract The present paper opens this topical issue on translation techniques by drawing a theoretical basis for the discussion of translational issues in a linguistic perspective. In order to forward an audienceoriented definition of translation, I will describe different forms of linguistic variability, highlighting how they present different difficulties to translators, with an emphasis on the semantic and communicative complexity that a source text can exhibit. The problem is then further discussed through a comparison between Quine's radically holistic position and the translatability principle supported by such semanticists as Katz. General translatability — at the expense of additional complexity — is eventually proposed as a possible synthesis of this debate. In describing the meaningfulness levels of source texts through Hjelmslevian semiotics, and his semiotic hierarchy in particular, the paper attempts to go beyond denotative semiotic, and reframe some translational issues in a connotative semiotic and metasemiotic perspective.


A linguist's translation studies
Aim of the present paper is to provide a theoretical framework for the collection of papers hosted in this Topical Issue of Open Linguistics. Their focus is the description and analysis of the translation techniques adopted by translators in contexts culturally, geographically and chronologically distant from the contemporary Western milieu. Special attention is drawn to translation practices in East Asia.1 Readers could notice that the following papers do not, or not exclusively, deal with what could normally be expected or at least assumed to be part of a collection of translation studies. My goal here is, indeed, to show how things as different as the translation of a sacred text, an ancient grammatical tradition and a philosophical discourse can be considered to be sides of one and the same phenomenon.
As a linguist, rather than a specialist in translation, my intention in this pages is to sketch out some linguistics-oriented considerations on the recurring questions that translation scholars of the last decades, as well as translators of all times, have faced.
Regretfully, the history of the relationship between linguistics and translation studies is one of unrequi-ted love. These two fields deal with similar problems and describe the same underlying realities: meaningfulness, knowledge, information and the process of communication. It was therefore quite natural for translation theorists to expect that linguistics would support this field when it emerged as an independent research program: "any theory of translation must draw upon a theory of language -a general linguistic theory" (Catford 1965: 1). Linguists, however, failed to meet these expectations. All too often their conclusion was that "the fundamental assumptions of modern linguistic theory imply that translation is at best incapable of methodical control and at worst impossible" (Beaugrande 1987: 2). This became especially true after the rise of Chomsky's formalist approach to linguistic analysis. The idea of a "deep structure", supposedly common to all languages, was seen as the philosopher's stone that would definitely solve the problem of translation, as the intermediate invariant to which translators could refer while "decomposing" their source texts before "recomposing" them in the target language.2 However, Chomsky has always maintained that his theory is not meant to have practical applications, translation not being an exception: "The existence of deep-seated formal universals […] does not […] imply that there must be some reasonable procedure for translating between languages" (Chomsky 1965: 30). Personally, I would not be so pessimistic about the possible use of linguistics in a theoretical approach to translation. If not in improving the efficiency of translation itself, linguistics can be helpful in defining its goal and in describing its functions. In the next paragraphs I will discuss the various levels of meaningfulness of a text, as a prerequisite for the understating of translation processes. Such a topic lies on a crossroad of different disciplines, from linguistics and philosophy of language to semiotics and translation studies, and it will thus be analysed with reference to these multiple paradigms.

The mystery of the unsaid
A sensation common to translators of all times is the impossibility of conveying in their target language each and every meaning inherent to their source text. Yet, practice shows that the translators are usually quite successful in achieving a satisfying degree of translational equivalence. Some mental skill allows us to reformulate in the target language what is said in the source with acceptable approximation, that is "to say almost the same thing" (Eco 2003). To assess this degree, we should begin by defining the notion of what is said. Modern philosophy of language uses this locution in a technical sense, to denote the actual referential content of the message resulting from the communicative intentions of its author and the objective knowledge gain conveyed through it. Some philosophers define what is said precisely as the content preserved throughout the process of translation (more about this further below). However, this philosophical category is probably too restrictive to account for all the facets of meaningfulness that a translator has to deal with.
To show how vast the content inherent to a source text can be, seen from the point of view of the translation practice, instead of what is said, I prefer to use a different term: the unsaid. By this I mean every nontrivial component of a message besides those that are stated by its literal wording and those, still trivial, that can be automatically deduced from it (such as anaphor resolution, for example). The key element of a good translation lies therefore, in the equivalence of what is unsaid, not less than that of what is effectively said. In other words, with the proviso that what is said in the source text has been accurately rendered by the translators, the quality of their work is then judged on the ground of translational equivalence of the unsaid component.
It is worth noting that in many different historical contexts translation theorists have indeed attempted to define the exact nature of this unsaid component of source texts. The answers given by different authors vary terminologically but not in their essence. Thus, St. Jerome prescribed for good translators to render sensum de sensu 'sense with sense' instead of doing a merely literal translation (verbum e verbo 'word by word'). The same distinction is expressed by Nida's dichotomy between formal vs. dynamic translational equivalence (just to mention one well-known current definition). Jerome's sense and Nida's dynamic equivalence are both referring to the unsaid component of the text.
In my opinion, this unsaid should be "demystified", i.e. rationalised and formally described. If primary speakers understand all of the non-trivial components of a message, theorists too ought to be capable of inferring them and analysing them formally. The only condition of such a capability is to consider the source text not as a flat structure, but as a complex product of human semiotic and metasemiotic activity, based on the same rules that govern communication itself.

Dimensions of linguistic diversity
It is normally thought that translation is needed because languages are different. Below I will argue that a better definition should focus on the hearers -or readers -of the message, rather than on the language. But firstly, I wish to survey the possible dimensions of natural language diversity, not all of which are obvious. In this respect the Saussurean notion of arbitrariness, in all of its manifestations, plays a key role. To Saussure, arbitrariness means that there is no logical or natural motivation in the diachronic shaping or the synchronic form of natural languages, at all the possible levels of analysis.
The most recognisable -though not necessarily the most important -manifestation of arbitrariness is phonological diversity. Languages sound different: they select arbitrary sets of phonemes and distinctive features in order to form the signifiers of linguistic signs.3 The partitioning of articulatory and acoustic space is arbitrary. Therefore, cross-linguistically, similar signifieds (meanings) are paired with different signifiers (phoneme strings). However, the real difficulty of translation does not lie in this difference: translation is never a simple permutation of signifiers between signs with the same signified. Mostly because there is no such a thing as the "same signified": finding in the target language a word that exactly matches the meaning of a given word in the source language is quite unlikely in practice.

Lexical and grammatical anisomorphism
Another greatly relevant manifestation of arbitrariness involves "meaning boundaries": that is, the fact that different languages can have unpredictably different and cross-linguistically mismatching lexical meanings. This means that not only do words sound differently across languages (the "signified-to-signifier" arbitrariness mentioned before), but they also -and more importantly -mean different things, because of their semantic and structural anisomorphism. Word meanings are, so to say, shaped differently in different languages. In the more rigorous Hjelmslevian terminology we can say that the content's substance is formed differently in different languages. The undetermined quality of these boundaries depends on the fact that the mental purport of lexical meanings -we can call it thought -is "[…] only a shapeless and indistinct mass.
[…] Without language, thought is a vague, uncharted nebula. There are no pre-existing ideas, and nothing is distinct before the appearance of language" (Saussure 2011: 111-112). Hjelmslev also reached similar conclusions: "[…] if we think without speaking, the thought is not a linguistic content " (1969: 49).
A myriad of more or less exotic and striking examples from all the languages of the world can be easily made. One such a case is the well-known use of the term game in the later work of Wittgenstein (1953). In English several different words correspond to the German word Spiel, covering different subsets of its semantic area: game, or match (e.g. in football), or play (e.g. in theatre), or also gambling.4 Wittgenstein patently exploits the polysemy of the German term in his work, thus posing a serious challenge for his English translators.
The arbitrariness of the grammatical categories and categorial meanings also lies in this aspect of language arbitrariness. In his seminal paper Jakobson (1959) introduced the issue of grammatical categories and discussed the limits of their translatability. Following Boas, he defined grammaticality as a set of obligatory choices within sets of essentially semantic -therefore, arbitrary -values (e.g.: masculine or feminine, past or present, definite or indefinite etc.). Grammatical anisomorphism can be a source of astonishment, especially for those who approach a new language for the first time. For example, an English-speaking learner is easily baffled in discovering that Japanese has no grammatical way to distinguish the future from the present tense or the third person of the verb from the others, while on the other hand categories such as honorifics must be expressed in every utterance. However, the communicative burden of grammatical values and categories must not be exaggerated. Only rarely do they bring new information (as when we use the grammatical gender to distinguish a he-cat from a she-cat), and in the majority of cases, they only serve the needs of the agreement among sentence constituents. Thus in most cases, they do not represent a real problem for translators (contra Osimo 2011: 58).
As Quine (1960Quine ( /2013 observes, "translation between kindred languages, e.g., Frisian and English, is aided by resemblance of cognate word forms. Translation between unrelated languages, e.g., Hungarian and English, may be aided by traditional equations that have evolved in step with a shared culture". The greater is the genetic, geographic and typological distance between two languages, the larger is their mutual incommensurability and the more uneven is the process of translation. A paradigmatic case in such respect is that of Kupwar, an Indian village where several dialects of both Dravidian and Indo-Aryan origin (possibly up to four) have coexisted in close contact for such a long time that their grammatical and lexical structures have been shaped into some almost parallel forms; see Gumperz and Wilson (1971).
It is worth noting thatthis kind of diversity does not imply any kind of Whorfian determinism. Those who do not know or do not have in their language a word that specifically expresses certain content can manage the same results using a phrase or other linguistic means.5 The absence of a certain grammatical category, such as tense, does not imply that the speaker cannot discern between past, present and future.

Diversity in linguistic norms and behavioural patterns
At a higher level of analysis we find "linguistic habits" or "behavioural patterns", which also can vary quite unpredictably. By habits I mean all the discursive conventions established by speakers of a certain language. In modern linguistics there is no generally accepted term referring to such categories. The most appropriate suggestion seems to be the notion of linguistic norm proposed by Coseriu (1970) as an intermediate level between linguistic system and linguistic usage. Inspired by Coseriu, Toury introduced this notion in the field of translation studies. In Toury's definition the norm is a "regularity of [linguistic] behaviour in recurrent situations of the same type" (Toury 1998: 55).
From the perspective of the content plane, we can think of norms as expressing cognitive frames, i.e. the standardised mental representations of recurring and ritualised real situations, see Fillmore (1976).6 Norms are crucial for translation because very often, if not always, it is they that represent the real unity of meaning, not the single words. What is stored in the mental lexicon is not only a list of meanings paired with phonological forms, but also -and more importantly -a list of textual norms paired with the cognitive frames that they refer to.
In this context a good example would be the set of academic writing norms (and of "good style" generally speaking). In English-speaking academia, for example, authors are encouraged to be assertive and direct, to avoid distancing and other elusive discourse strategies and directly state their most important ideas.7 Contrariwise, in Japanese academic writing it is customary to be extremely polite and tentative in promoting one's own ideas. This results in a great deal of post-verbal phrases, such as to mo ieru 'it could be also said that' or to omoeru 'it could be thought that', that are supposed to attenuate and mitigate the assertiveness of the main clause. Thus, speaking about an unpleasant phenomenon, one can easily say in Japanese: (1) Sukosi ranbou da to mo ieru.
Literally: 'One can also say that it is a little rough'. Pragmatically: 'It is quite rough'.
The attenuation of the verbal assertiveness through the introduction of a dubitative post-verbal phrase in the example (1) is neither imposed by the referenced reality nor by a grammatical necessity, but is only added on the basis of Japanese discursive conventions. This can indeed be deleted in translation since this Japanese sentence does regularly and pragmatically correspond to quite a simpler sentence in English with no attenuation elements.

Informational asymmetry 2.3.1 Presuppositions and implicatures
Another kind of linguistic diversity relates to the way in which information is conveyed by words. A theoretical premise is needed here. Informational asymmetry is an essential part of human communication, which effectively consists of continuous attempts to fill the gap between the speaker's and the hearer's inner states, that is to overcome a pre-existing asymmetry in their respective mental representations of reality. What makes communication possible is the almost negligible size of this gap, when compared to the huge amount of information that the participants of the conversation share beforehand. Hearers concentrate their interpretative efforts on a very limited portion of what is effectively uttered, since sentences are full of redundancy and reference to the sizable bulk of information that is supposed to be known to all the members of the dialogue situation. Communication is, therefore, incremental in nature. An interesting question is where this shared knowledge comes from. It seems that four possible sources of such information can be envisaged (see Keidan 2008 for a more detailed discussion). a) "Biological" and "physiological" information shared by all human beings as a species. Thus, the awareness of body parts, cardinal directions, and basic perceptual schemes belong to a common mental background shared by all the speakers; according to cognitivists, "human language and thought are structured by, and bound to, an embodied experience" (Lakoff and Johnson 1999: 233). Such categories as time and space, self-awareness and the notion of otherness, etc. belong here too.8 b) "Objective" information, shared by those who belong to a community of people with the same history and/or language; that encompasses all the products of the conscious intellectual labour (i.e., production of encyclopaedic knowledge), as well as of the unconscious cognitive activity (i.e., arbitrary shaping of the mental lexicon).9 c) "Co-textual" information, i.e. every piece of information that is acquired by the audience within a 7 To cite a classic: "Any one who wishes to become a good writer should endeavour, before he allows himself to be tempted by the more showy qualities, to be direct, simple, brief, vigorous, and lucid" (Fowler and Fowler 1906: 11). 8 The scope of these natural predispositions has been questioned in many ways. Cf. the criticism of all kinds of semantic universals in Fintel and Matthewson (2008). Most likely, we should view them as tendencies rather than obligatory mental structures. 9 Many theoreticians of translation would use the term culture in a similar context. Thus, Osimo (2011: 57) speaks of "mother culture" alongside the "mother tongue" that an individual has to learn. As a linguist, I suggest to see culture as a set of meanings, and avoid the term culture itself, as too vague and interpretable. communicative event. Such elements are referenced by the mechanism of anaphora. Following Halliday (1999: 3) I denote, by the term "co-text", verbal environment only, not the physical context in a broader sense. d) "Perceptual" information, i.e. what is self-evident to all the participants of a communicative event through the direct observation of the circumstances of utterance. Such elements are referred to by the deictic elements and categories (such as adverbs and bound morphemes indicating space and time).
Point a) and especially point b) build up the corpus of discourse presuppositions, i.e. everything constituting the shared knowledge that is taken for granted by the speaker when s/he addresses the hearer. Communication itself can be defined as a process by which the speakers help the listeners to update their mental presuppositions.10 A particular type of presuppositions has been called by Grice (1981) discourse implicatures. When the utterances made by dialogue partners refer to something apparently unrelated to the main topic of conversation, yet fully yield a communicative goal, discourse implicatures come into play. To make a literary example let us quote a famous verse from Dante's Paradiso, XVII (in Longfellow's English version): (2) Thou shalt have proof how savoureth of salt the bread of others.
What is effectively meant here is Dante's exile from Florence. To be able to understand this we must take into account a series of cultural and situational assumptions. First, one must remember that in Florence bread was -and still is -baked without salt, which is a cultural presupposition. Second, Dante also was assuming a situational implicature: though born in Florence, he had eventually been exiled for political reasons, into a land where the bread was made with salt. So, 'to eat salty bread' becomes an equivalent of 'to live in exile'.
Together with linguistic behavioural patterns, presuppositions belong to a highly and unpredictably variable component of linguistic communication that cannot be ignored during the translation process. Flattening metaphors and neglecting presuppositions often makes the target text almost incomprehensible. For example, Whitney openly defended a strictly literalist approach to the translation of ancient Indian literature (such as Vedas, Upanishads, etc.). As a linguist, he did not pay much attention to the referential meanings of the Sanskrit expressions under consideration, or their aesthetic value, but was mostly interested in grammatical and lexical morphology.11 Rather than reconstructing incorrectly the presuppositions of an ancient text, he thought it was preferable to avoid their restoration altogether. However, his own translations result almost unusable and barely intelligible. The opposite approach is generally adopted nowadays. Thus, Witzel (1996) defends the necessity of "entering the Vedic mind" in order to render correctly the hidden meanings of a Vedic text. For example, he claims that the phrase iyaṃ gauḥ, recurring throughout ancient Indian religious and philosophical books (specifically in Vedas and Brāhmaṇas), cannot be translated with its literal meaning 'this cow' but is rather to be understood as a metaphor for 'the earth is this one here'.

Word level presuppositions: meaning history
An important type of presuppositions rests on what we could call "inner memory" of the words. The history of preceding utterances of a word is sedimented in its meaning. Sometimes it becomes evident through the formal etymology. Some other times it involves long abandoned metaphors and sense shifts. So, a word does not only mean what it means, but does also communicates the way it comes to mean that meaning, i.e. the cognitive path leading the speakers to elicit that referent. This inner informational structure of words is a factor of even greater cross-linguistic variability. In short, while one language uses certain words that have previously had certain meanings in order to denote some referents, another language refers to the same reality using words with a different semantic history; and this semantic history is always presupposed when a word is used.
As Potebnja puts it, words are not images of reality, but rather images of images of reality (see Kokochkina 2000). This is to say, again, that words do not represent things directly and simply through their meanings, but rather, denote things through the invocation of other images.
As a concrete example, let us consider a case from the Greek New Testament together with its rendering in Old Slavonic. The Old Slavonic krĭstiti 'to baptise' was used as the translational equivalent of Greek βαπτίζω 'to baptise'. The two terms are co-referential in most contexts. However, etymologically they differ in that the former is clearly linked to krĭstŭ 'cross' (a loanword from Old High German krist 'crucifix', 'Christ'12); the latter, on the other hand, is a metaphorical extension of the classical verb that meant 'to dip'. Therefore, these two verbs invoked completely different mental presuppositions, albeit obliterated by translational equivalence. A passage from Mark 7:4 reads βαπτισμός ποτηρίων, in its literal meaning of 'washing the vessels' (before eating). An inadvertent Slavonic scribe writes krĭštenie stĭklěnicamŭ, literally 'baptism of the vessels' (Codex Zographensis, ed. Jagić, p. 58), without realizing the difference in the inner form of the Slavic word and, being therefore oblivious of the modified denotation.
Theology is, generally, a good source of such cases. Thus, when the canons of the Council of Trent (1545-63) established the dogma of transubstantiation, claiming that "[…] a conversion is made of the whole substance of the bread into the substance of the body of Christ our Lord, and of the whole substance of the wine into the substance of His blood" (Waterworth 1848: 78), their understanding of "substance" was quite different from what is meant by this term in present days. At that time substantia was a philosophical term coming from medieval scholasticism and referred to something abstract, metaphysical and intangible. In modern use, on the other hand, this word -or, better, its vernacular equivalents: English substance, Italian sostanza etc. -possesses new meanings, accumulated during the development of physics and chemistry over last centuries. This is to say that, regardless of the notion of transubstantiation per se, today's substance cannot be considered a translational equivalent of the XVI Century's substantia, because of a radical change in the meaning history of this word.13

Higher order semiotic diversity
Human communication is characterised by what could be termed a "pervasive meaningfulness": every constitutive part of the communication process can become a sign with a meaning, or befit itself the meaning of a sign. Such process is still subject to Saussurean arbitrariness and therefore to cross-linguistic variability.
This semiotic manifoldness is pregnantly described in Hjelmslev's (1969: §22) terms. With some simplification, we can say with him that natural language is a denotative semiotic, i.e. a set of signs resulting from the combination of an expression (i.e. a sequence of phonemes) and a content (an abstract semantic representation referring to some external reality). Yet, a denotative semiotic as a whole can also become either the content or the expression of a higher order semiotic. The former is termed the connotative semiotic (since it generally captures the connotations), while the latter is the metasemiotic.
12 A German loanword in Slavonic religious terminology possibly points to a Western Christian mission to the Slavs that predated that of St. Cyril; cf. Gusmani (1995). 13 Cf. "[…] the language of substance and accident in which the notion of transubstantiation is formulated belongs to a metaphysical world-view that does not square with reality as we now understand it" (Stebbins 1990: 225). See also Alfieri (2006) for the history of the term substance in grammar and philosophy.

Connotative semiotic
Not only do the words convey their meaning, but even the way we speak can be meaningful. Phonological variants can be meaningful, the lexical selection (among possible synonyms) can be so too, and the selection of certain linguistic behaviours patterns, or norms, can also be relevant for the comprehension of a text. Therefore, when someone chooses the most appropriate words for communicating some content, his/her word choice is usually meaningful, rather than neutral. They choose a specific style, register or genre when they speak (literary, formal, colloquial, familiar, slang, etc.).
The expression plane of a connotative semiotic is a denotative semiotic, while its content is a certain stylistic qualification, or connotation,14 see Figure 1. The connotations, therefore, form a well-established semiotic code, which is usually taken intentionally into consideration by the speakers. As an example let us quote a passage from Homer's Iliad XI, 556, together with its first English translation by Alexander Pope. In it Ajax is compared to a stubborn donkey. As remembered by Adams (1972), Pope argued that this simile involved no irony in ancient times, since the donkey was perceived as a strong, patient and noble animal, "upon which Kings and Princes could be seen with dignity". On the other hand, "in English, for a modern reader, the connotations of "ass" and "donkey" are strongly comic" (Adams 1972: 85). Pope, consistent with his interpretation, eliminated in his rendering every mention of the donkey and reworded it as a "slow beast with heavy strength", thus preserving most of the meaning without implying any inaccurate stylistic colouring.
Contrariwise, present days' criticism supports the reverse interpretation, which rejects the presumed seriousness of Greek epic poetry: "Homer […] shares with Shakespeare […] an absence of what more fastidious neo-classical ages -and many later epic poets -were to see as essential decorum" (Silk 2004: 55). Homer's likening Ajax to a donkey is thus viewed as a simile that "being drawn from the concrete world of nature […] do not directly affirm permanent ethical traits" (Kirk and Hainsworth 1993: 284). This makes it superfluous to ennoble Homer's donkey in translation.15 But also a language as a whole can bear a connotative value. That the Emperor Charles V allegedly spoke in different tongues to God, women, men and horses could be difficult to explain to those who now are respectively native speakers of Spanish, Italian, French and German. James Joyce's Finnegans Wake is extremely hard to translate also because its wide use of foreign words; while they are used to convey determinate connotative content to English-speaking readers, they do not do the same for speakers of other languages. Leo Tolstoy's War and Peace is a Russian novel, but it opens with two pages written entirely in French, by which the author intended to convey some important information to the reader. Monty Python's A Fish Called Wanda presents a character who is sexually excited by Italian language. When the film was distributed in Italy, it was very carefully translated and dubbed; but since Italian was the new base lan-guage, the trigger of sexual excitement needed to be changed; indeed, it became Spanish in this version. 16 The comprehension of the connotative semiotic level is, therefore, a sensible component of the original message, and it can be, if necessary, preserved in translation.17

Metasemiotic
Not only do we use words for expressing some content, but also words themselves can become the content of a message. Language as a whole can be the content of a higher order semiotic. We can speak about the way we speak. In Hjelmslev's terms, this is called the metasemiotic, i.e. a semiotic whose plane of content corresponds to a denotative semiotic (i.e. a natural language). The most obvious instance of such phenomenon is represented by linguistics. When we describe the grammar of a language what we are actually doing is a metasemiotic operation, since the content of our discourse is, in itself, a semiotic system. See Mentioning grammatical traditions as cases relevant of Translation Studies is not something that is normally done by scholars. Yet, historically the grammatical traditions of many languages have been established on the ground of those already existing for some other languages. Since languages are different, a metasemiotic description of a certain language can easily be inconsistent with another one; that makes grammar creation an issue of translation as well, starting from grammatical terminology. Adapting pre-existing grammatical terms to a new language often required a rearrangement of the theory. Also, a degree of misinterpretation becomes unavoidable. The history of grammars is, indeed, full of such misinterpretations.
For example, Latin grammatical categories were derived from the Greek ones. The latter however were initially developed in an Aristotelian milieu, of which Latin translators were not aware. Consequently, they interpreted this terminology as referring to simple iconic representations rather than to the categories of the Aristotelian scientific reasoning. For example, the names of nominal cases in Latin drifted away from the original intentions of the Greek grammarians. While calquing Greek terminology, Latin grammarians trivialised the terms by mistaking their abstract denotation and obscuring their etymology. Thus Accusative (from Greek αἰτιατική) was not originally meant as the case of "accusation" but that of "cause"; Genitive (from Greek γενική) was not the case of "parents" but that of "genus", i.e. a scientific classificatory category; eventually, as shown by De Mauro (1971), the Dative (from Greek δοτική) must have been thought, by the Greeks, as the case of "the given", i.e. "condition", rather than that of "giving".18 For a modern example, on the other hand, I wish to cite a case in the translation of a popular book by 16 Even a single letter can have connotations, in the case of "satirical misspellings". Thus, the deliberate "Teutonic" replacement of C with K, in such words as Amerika and the like, is used within the leftist milieu -first in the US, and later also in some Latin countries in Europe -in order to emphasise a critical view of the American lifestyle and foreign policy. It is explainable as a connotative allusion to Nazism, or to the symbolism of the letter K in Kafka's novels. 17 This obvious conclusion is far from being unknown to scholars of translation, see e.g. Osimo (2011). 18 Seuren's (1998: 21fn) (1988) programmatically translated also the Spanish examples, when possible, since -as he stated in the preface -the original audience was meant by Chomsky to speak Spanish natively.20 However, Spanish syntax is not as parallel to the Italian one as one can intuitively think. Thus, the closest literal translation of (3) sounds ambiguous in Italian, notwithstanding the opposite claim of G. Cinque who worked out the adaptation of Chomsky's book in Italian; see example 16 in Chomsky (1991: 21): (4) Chi i fece radersi i/j Gianni j ? Who made shave-REFL Gianni. a. 'Whom did Gianni have shave himself?' b. 'Who had Gianni shave himself?' The twofold semantic interpretation in (4) is due to the fact that, differently from Spanish a quién, which is openly marked as direct object, Italian chi can mean either the subject 'who' or the direct object 'whom', thus allowing for two different semantic readings. Therefore, the speaker's grammatical intuition works differently in the Italian edition compared to the original text. Note that this invalidates Chomsky's primary data and makes his reasoning less accurate in an important way, which is a good example of how a translational issue can have consequences on the functioning of a metasemiotic theory.

Audience-oriented approach
In the preceding sections we have presented an incremental model of linguistic communication. The meaning of a sentence can be reached by the audience only provided that the hearers share with the speaker a certain amount of pre-existing background information, going far beyond the simple linguistic knowledge (meant as lexicon plus grammar). Two different audiences can share the same language -granted that we know how to tell whether two languages are different or still the same -yet have different informational background, due to social factors, age, education level, and the like. Therefore, the speaker(s), the hearer(s) and the background are more relevant components of the communicative circuit than the code itself. This leads to an audience-oriented, rather than a code-oriented, definition of translation: Def. 1 Translation is a process that modifies the sources text to the extent that is needed in order to make it readable by a different audience.
Note that this modification does not necessarily consist in the change of the code. Such activity as 19 In my examples (3) and (4)  commenting or paraphrasing an ancient text in order to make it comprehensible for modern readers should be also considered a case of translation. Note that with the audience-oriented definition of translation Jakobson's (1959) distinction between "intralingual" and "interlingual" translation appears unnecessary, for it is not the language but the audience that has to be stressed.21 For example, Dante's Italian is, more or less, the same language as the one spoken in Italy today (unlike Ancient French, which is a totally alien language for the present-day French speakers). However, Dante's poems need to be heavily commented upon in order to be understood by contemporary readers, simply because the presumed knowledge changed substantially, and what at Dante's times was just a self-evident discourse implicature or a common cultural presupposition, is today a highly refined historical knowledge which is acquired only through special training.
This kind of audience-oriented definition of translation is preferable to the more traditional one based on the language shift. According to Toury (1995: 55) "translation is a kind of activity which inevitably involves at least two languages and two cultural traditions". In my opinion, "two cultural traditions" is a sufficient condition for a making a translation, while "two languages" is not "inevitable". I follow Nida (1964: 159), according to whom the ideal translator "is not so concerned with matching the receptor-language message with the source-language message, but […] that the relationship between receptor and message should be substantially the same as that which existed between the original receptors and the message".

Arbitrary and invariant in the translation process
With Quine and other authors, in the 60s, a period of scepticism about meaning in philosophy of language began. Everyday communication seemed deprived of any positively delineated objective meanings, objective enough to be accessible to all the speakers of a language community (e.g. all the native English speakers). Even worse was the case of two different languages put into comparison. Quine (1960Quine ( /2013 suggested a translation metaphor for communication, the so-called radical translation: any communicative act is considered by Quine an act of translation. The meanings of such simple and obvious words as apple, well, go can still differ slightly even between my personal usage and that of a sibling, regardless the fact that we share most of our family, historical and cultural background. The more so, in case the speakers engaged in conversation are farther apart, and definitely so if they speak two different tongues. This is because no speaker can ever know with full certainty what is going on in someone else's head, given that mental activity is individual and essentially incommunicable. For this reason, and also in light of the multiform diversity of language exposed in the preceding sections, there is neither such thing as "sameness of meaning", nor words that can be considered the only and exact terms for certain referents. This is valid as much for the intralingual communication as for the interlinguistic translation. Every communicative act is only an attempt to find the most appropriate words -not the exact ones -to express a certain content.22 But what determines the success of a communicative act? In case of translation, when and how does it reach its goal? How can we tell that the source text has been correctly rendered in the target language? This is the ever-present question of the translational equivalence. Now, the pars destruens of Quine's sceptical discourse, his semantic pessimism, can be tempting. However, one must also observe that, in most of the 21 Jakobson's classification is sometimes blamed as "frequently quoted but highly artificial" by Lambert and Robyns (2004: 3604), however the importance that it has had in translation studies is still undisputable. 22 Nietzsche's Aphorism 55 "The Realistic Painter" from Gay Science (here quoted in A. Del Caro's translation) is a nice allegoric representation of this state of affairs. The reader should only substitute "painter" with "speaker" and "painting the nature realistically" with "saying truth about reality": To cases, communication is not as hopeless as he claims. It is rather undeniable that communication usually achieves its goal, might it be sooner or later.23 Truth can be unattainable in its totality, but a good approximation to it is, nonetheless, usually at hand. Indeed, we behave as if we had a full semantic comprehension of each other, and this stance seems functional enough for most practical purposes. Quine himself explained it in terms of stimulus and response: the speaker gradually elicits a reliable, i.e. still not guaranteed but well working, theory of the meanings of another speaker from the observation of behavioural patterns correlated to recurring speech acts. In order to counter Quine's pessimism, we should accept the idea that, between the source text and the target or -in terms of audience -between the original hearers of a message and its new ones, there is an objective intermediary: the objective knowledge. This knowledge represents the invariant to which both speaker and audience must refer in order to achieve the communicative goal. To define such knowledge in its essence goes probably beyond the means and the scope of linguistics. But we have to presuppose its existence as a condition for a successful use of language. Only if the audience succeeds in activating the same portion of the objective knowledge as the one that was intended by the speaker the message is effectively transferred.24 Positing an extra-linguistic denotative invariant excludes a direct link from speaker to audience, or from one language to another. All the links are mediated by the invariant, see Figure 3.

Objective) knowledge
Speaker Audience The dashed line here shows that the link between the speaker and the audience is only an indirect one. Rather than "putting" some notions into the hearer's head, the speaker explains to the hearer the right path to the intended portion of the objective knowledge. The hearer is then, hopefully, in the condition to refer to such portion and thus to understand the message. There is no objectivity in words per se. Language does not contain knowledge, but only activates it. The opinion, shared by some scholars, that during the process of translation the intended meaning is simply "unclothed" from its original encoding and subsequently "dressed up" with new garments (i.e. expressed in the target language), is hardly tenable. I am arguing against claims such as this: "[…] translation seems to be an instance wherein a signified can persist for a time outside a bond with a signifier" (Beaugrande 1987: 3). The linguistic signified does not exist without its formal enclosure in the language. The tertium comparationis during the translation process is, therefore, not linguistic at all. It seems appropriate here to quote Popper (1980: 156), according to whom "the objective thought content is that which remains invariant in a reasonably good translation. Or more realistically put: the objective thought content is what the translator tries to keep invariant, even though he may at times find this task impossibly difficult". Popper's "objective thought" is what I have termed here "objective know-23 A famous example of the breaking of some apparently insurmountable communication barriers is the enterprise achieved by D. L. Everett, the discoverer of Pirahã, the oddest native Amazonian language, see Everett (2005). The differences between Pirahã and English are as striking as in the wildest of Quine's fictional examples. Yet, Everett eventually managed to learn this language, to successfully communicate with the native speakers and also to produce a grammatical description of Pirahã. 24 In a more Saussurean fashion we could say, with Koller (1979: 185), that translational equivalence exists at the level of language use (parole) rather than on that of a language system (langue). This because reference, as such, is manifested only in a concrete act of language, rather than at the level of the abstract code. ledge". In developing Frege's (1918-19) notion of Drittes Reich 'the third realm', Popper called this domain of objectivity World 3 (while Worlds 1 and 2 are, respectively, the universe in and of itself, and the subjective mental representations thereof).25 We cannot say anything certain about World 3, but, as a matter of fact, we need to postulate it, and to act as if there is one.

Translatability
The question of translatability has been widely discussed by scholars of natural language semantics, who presented themselves as "rationalists" and opposed Quine's sceptical stance labelled as "empiricist".26 Rationalists support the assumption that the semantics of natural languages does not vary to the degree that puts a real obstacle to their inter-translatability. This is how Katz (1976: 39) formulates it: Def. 2 For any pair of natural languages and for any sentence S in one and any sense σ of S, there is at least one sentence S′ in the other language such that σ is a sense of S′.
However, Katz's "translatability principle" appears as much idealised and hyperbolic as Quine's "meaning holism" is pessimistic. Katz refers to a flat and oversimplified model of semantics, with words that simply and directly mean things, and with no space for any semantic indeterminacy. Fintel and Matthewson (2008: 144) correctly and quite realistically emend Katz's "translatability principle" into a weaker form that better accounts for the cross-linguistic anisomorphism: Def. 3 What language X expresses simply is also expressible in language Y, but at the price of some complexity.
The focus of translation studies must be concentrated on the last element of the Def. 3: the complexity. How can a text be translated while preserving its original levels of meaningfulness and without reaching an unnecessary level of complexity? The translators are well aware of the answer to this practical question: some choice, and therefore, some sacrifice, has to be made. The translator selects the most prominent communicative content of the source text, and tries to make it understandable for a new audience, while some other features of the original are necessarily lost.27 A proposal for a definition of the notion of "translation technique" can be, therefore, the following:  Jakobson's (1981: 751) definition: "the dominant may be defined as the focusing component of a work of art: it rules, determines, and transforms the remaining components. It is the dominant that guarantees the integrity of the structure". Cf. also Torop (1995: 160). Metlen (1933: 544). Yet, Ulfilas can hardly be blamed of being a dull writer, since his cultural background and personal qualities were reportedly of a very high level. As argued in Keidan (2005), the true interpretation of Ulfilas' approach should be that he was indeed an excellent translator who was able to combine a close adherence to the original wording with an attempt to preserve and recreate in Gothic the stylistic connotations that the text of Greek Bible had acquired by that time.28 Trying to maintain both the mysterium of wording and the connotative component makes Ulfilas an even more skilled translator than St. Jerome. It is important not to mistake the choice of the prominent level of meaningfulness. Not everything that is said is communicatively relevant: much is redundant and can be safely sacrificed in translation. On the contrary, examples of disproportionate relevancy estimation are ridiculed in a widely known passage from Adams (1972: 12): "Paris cannot be London or New York, it must be Paris; our hero must be Pierre, not Peter; he must drink an aperitif, not a cocktail; smoke Gauloises, not Kents; and walk down the rue du Bac,not Back Street".29 It is worth noting, however, that this estimation may vary depending on time, fashion, stylistic conventions and theoretical background of the translator. For example, the proper names calques and adaptations condemned by Adams were considered fully acceptable by others. A curious case concerns the Russian translation of Pinocchio. It first appeared in 1906 under the title Priključenija Fistaški, literally: 'The adventures of Pistachio'. Apparently, for the translator, pistachio sounded more familiar to a Russian reader than the so far unheard Pinocchio. In all the subsequent translations the original name was restored. The issue with the names of characters in the translations of Lewis Carroll's Alice in Wonderland is also well known. Interestingly, Vladimir Nabokov in his 1924 translation, changed Alice's name into Anja (a Russian diminutive of Anna), rather than borrowing the original name; however, forty years later he himself condemned similar adaptations in his theoretical writings on translation, see Friedberg (1997: 86).

A conclusion. The unsaid demystified
Within the present days' translation studies discussion on translational equivalence, the dualistic, oppositional approaches à la Nida are gradually fading, with the rise of more complex models including several levels of equivalence. For instance Koller (1979; distinguishes five types of equivalence -denotative equivalence; -connotative equivalence; -text-normative equivalence; -pragmatic equivalence; -formal equivalence. My claims in the preceding sections of this paper are generally consistent with this contemporary tendency. Even on a purely terminological level, the adjectives used by Koller to describe the different types of equivalence are not too far from concerning the levels of meaningfulness included in my discussion. One improvement that I suggest here consists in simplifying Koller's list (and other similar lists) by making it conceptually homogeneous. There is no need, in my opinion, to postulate equivalence types of different nature, since the equivalence phenomenon is essentially the same for every level of meaningfulness. It is always the same mechanism, reproducing itself at different scale, as a fractal structure. There is always a certain linguistic object that serves as the expression of a certain content (to use Hjelmslev's versatile terminology). The mechanism is recursive: what is already a bilateral linguistic entity can become, in its turn, either the content or the expression plane of a higher rank semiotic entity.
To briefly summarise, on the expression and the content planes we can find, among others, the following pairings of expressions and contents: -phonetic features or segments ~ can express social connotations; -phonetic patterns (such as rhymes, alliterations) ~ express poetic quality; -words ~ express referential, denotative content; -phrases ~ descriptively express referential content; -linguistic behaviours or norms ~ express cognitive frames; -language as a whole ~ connotatively expresses nationality and the like; -the usage history of a word ~ expresses the cognitive path of the denotation; -the grammar ~ expresses a language (by describing it).