See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/228377785

Collocations in Translated Language: Combining Parallel, Comparable and

Reference Corpora

Article · May 2012

CITATIONS READS

7 649

1 author:

Silvia Bernardini
University of Bologna
40 PUBLICATIONS   2,020 CITATIONS   

SEE PROFILE

All content following this page was uploaded by Silvia Bernardini on 07 August 2015.

The user has requested enhancement of the downloaded file.

Collocations in Translated Language:
Combining Parallel, Comparable and Reference Corpora

Silvia Bernardini 1

1. Introduction

This paper describes an attempt at investigating some collocational properties of

translated language. The notion of collocation is one of the cornerstones of corpus
linguistics, and has been the subject of substantial speculation and empirical research
(section 2.1). Within translation studies, few works have tackled this issue, partly
because of methodological conundrums (section 2.2). Yet collocations – and
idiomaticity in general – would seem to be relevant to the “elucidation of the nature of
translated text as a mediated communicative event” (Baker, 1993: 243), and thus
central to corpus-based translation studies.
The paper claims that the research questions regarding collocations in
translated language posed so far need to be reframed in order to avoid the
methodological problems faced by previous studies. The method devised to answer
these questions is described in some detail (section 3), and a case study is presented
(section 4) of a single phraseological pattern in translated and original Italian fiction
texts (Noun preposition|conjunction Noun). Section 5 concludes the paper and makes
suggestions for further research.

2. Background

2.1 Collocations

The search for collocations is one of the driving forces behind corpus linguistics. The
notion is traditionally associated with the work of J.R. Firth, who promoted “the study
of key-words, pivotal words, leading words, by presenting them in the company they
usually keep” (Firth 1956:106-107). Scholars inspired by his work have attempted to
make the notion less obscure and more operational, and to pursue its study through
the use of text corpora. Jones and Sinclair (1974) describe significant collocation as
the “regular collocation between items, such that they occur more often than their
respective frequencies and the length of the text in which they occur would predict”.
In recent years, several definitions of collocation have been proposed, usually
falling within one of two general approaches (Nesselhauf 2005). Phraseological
approaches attempt to tell collocations apart from free combinations on the one hand,
and from other lexical restriction phenomena on the other. A typical phraseological
definition is Howarth's (1996: 37), who describes collocations as “fully
institutionalised phrases, memorized as wholes and used as conventional form-
meaning pairings”. Clearly, collocations here are viewed as abstract entities with
instantiations in texts: the main focus is on the language user's competence.
Frequency approaches focus less on classifying collocations, and more on identifying
them in texts. Compare Kjellmer's (1987: 133) definition: “A sequence of words that
occurs more than once in identical form and is grammatically well-structured
1
School for Translators and Interpreters, University of Bologna, Italy
e-mail: [email protected]
(Kjellmer, 1987: 133). This definition (like Jones and Sinclair's above) steers clear of
criteria for collocativeness such as commutability of elements, semantic opacity and
figurativeness, which are often called upon by phraseology scholars to delimit the
notion from a theoretical point of view, and focuses instead on the parameters needed
to automatise the extraction of collocations from corpora.
The present work falls within the frequency approach. It makes no attempt at
distinguishing e.g. semantically-motivated combinations from (arbitrary) collocations,
or restricted collocations from idioms. This follows from the view that any
“lexicalised expression” - i.e. resulting from the operation of the idiom principle
(Sinclair 1991) - is potentially relevant to our analysis (see sections 3-4 below). We
shall therefore adopt Manning and Schütze's (1999: 151) rather general definition of
collocation as “an expression consisting of two or more words that corresponds to
some conventional way of saying things”, and focus on 2-word collocations only.

2.2 Corpus-based Translation Studies

Following Baker's seminal paper (1993), a large body of research in translation has
adopted a corpus-based methodology to try and shed light on the “features which
typically occur in translated text rather than original utterances and which are not the
result of interference from specific linguistic systems” (Baker, 1993: 243). This way
of envisaging translation, not as an individual act of transfer of a source text into a
target language, but rather as a socio-culturally regulated communicative event in the
target language community, has its roots in the work of translation studies theorists,
particularly Toury (1995). Toury's notion of “translation norms”, i.e., of socio-cultural
constraints regulating the behaviour of professional translators and leaving traces in
translated texts, has been the object of substantial research, along with the more
controversial notion of “translation universals”. Several features (whether universal or
not) have been isolated, that would seem to characterise translated language with
respect to other kinds of text production, and to point at norms of translational
behaviour. These are (the list is not exhaustive) explicitation/explicitness,
simplification, disambiguation, homogeneity, conventionality,
normalisation/sanitisation and so forth (see Laviosa (2002) and Olohan (2004) for
more exhaustive discussions).
These studies have been conducted using two kinds of resources: the more
traditional parallel corpora, made of originals in language A and their translations into
language B, and the innovative monolingual comparable corpora, made of originals in
language A and comparable translations into language A. Sometimes these resources
are combined, i.e., to form bidirectional corpora, made of originals in languages A
and B, and the respective translations in languages B and A. Reference corpora of the
languages under analysis are also sometimes employed as benchmarks. By limiting
their scope to the target language, studies of monolingual comparable corpora are
unaffected by language system-specific differences, an important variable in the
parallel approach. Therefore, they have been used extensively to compare overall
textual features such as sentence length, lexical variety, ratio of content words to
function words (Laviosa, 2002), or more specific patterns of use of (semi-
)grammatical (Olohan, 2001; Olohan and Baker, 2000) and lexical features
(Tirkkonen-Condit, 2004; Mauranen, 2000).
Parallel corpus approaches are more appropriate for the analysis of local shifts
and strategies. Studies following this approach have focused, e.g., on explicitating
shifts (Øverås, 1998), on normalising/sanitising shifts (i.e., the tendency to select
habitual target language expressions to render creative turns of phrases in the source

2
text; Kenny, 2001) and on translator choices with implications for a description of
translator's style (Malmkjær, 2004; Marco, 2004).
Turning specifically to collocation, few studies have tackled the issue.
Identifying collocations in (monolingual) corpora is far from straightforward, and
researchers adopting a parallel paradigm may have felt that adding a bilingual
dimension would render the task a daunting one. Advocates of the monolingual
comparable corpus approach, for their part, have typically tended to focus on aspects
that could be identifiable via automatic procedures; this requires some ingenuity in the
case of collocations, as we shall see. Yet the issue is central to an understanding of
strategies and norms for dealing with lexicalisation and creativity in translation.
Kenny (2001) and Øverås (1998) provide some evidence of normalising shifts
affecting collocations, i.e., in Toury's (1995) terms, of a tendency for translators to
produce repertoremes (lexicalised target language collocations) in place of textemes
(creative source text coinages). Neither method can be applied to a systematic analysis
of collocations in translation, though: in the case of Øverås, because the finding is
incidental, and in the case of Kenny because the starting point is the (manual)
identification of the creative combinations formed around a single node word in the
source text.
Attempts at analysing collocations in translated language systematically have
been made by Baroni and Bernardini (2003) and Danielsson (2001). The former
approaches the issue of collocations in translation from the target perspective, using a
monolingual comparable corpus of Italian original and translated articles from a single
geopolitics journal. All bigrams from the translated sub-corpus and from the original
sub-corpus were ranked according to their log-likelihood ratio value. The bigrams
most representative of the translated subcorpus (i.e., infrequent in the original
subcorpus) and those most representative of the original subcorpus (i.e., infrequent in
the translated subcorpus) were extracted for manual comparison. The authors report
that translations in the corpus show a tendency to repeat structural patterns and
strongly topic-dependent sequences, whereas originals show a higher incidence of
topic-independent sequences, i.e., the more usual lexicalised collocations in the
language. This work has the merit of proposing an original method for identifying
collocations in translated language, that relies on a mono-source monolingual
comparable corpus and goes beyond local observations of single cases selecting
candidates on statistical grounds. However, the results are rather difficult to interpret.
This is a common problem with quantitative studies of monolingual comparable
corpora, since the general tendencies observed are difficult to pin down and
interpretation is often not straightforward.
Danielsson's (2001) is an attempt at identifying “units of meaning” (~
collocations) in two monolingual corpora (one English, one Swedish), with the
ultimate aim of finding “units of translation” (i.e., bilingual collocation pairings) in
parallel corpora. Based on a frequency list of all the words in the corpus, word-forms
occurring 200 times of more are extracted for further analysis. Upward and downward
collocates (cut-off point: ≥5) are searched for and the evidence is combined to
produce citation forms. As she moves on to search for units of translation,
Danielsson’s work is plagued with data-sparseness problems. In the source text
component of her parallel corpus of fiction texts translated from Swedish into English
(~400,000 words per component), she finds that only 2 units of meaning (of the
12,099 previously identified) occur five times or more. Similar results are obtained
from the English target text corpus. Danielsson is well aware of the limits of her
method when it gets to the translational perspective, and acknowledges the need for
much larger corpora. Unfortunately, parallel corpora are costly to assemble and tend
to be small (unless one contents oneself with some widely-available text types, such

3
as EU parliament proceedings). Therefore a method such as Danielsson's, which starts
with units of meaning in reference corpora and then proceeds to look for units of
translation in parallel corpora (rather than the other way round), despite its obvious
value from a monolingual perspective, is bound to result in a substantial amount of
precious evidence being wasted in the parallel phase.
To obviate this problem, a change of perspective is needed. Rather than
identifying collocations based on the frequency and/or relatedness of word
combination tokens in a monolingual comparable or parallel corpus, we extract word
combination types from the corpus under study (however small), and obtain their
frequency and relatedness in a (large) reference corpus. In other words, we are using
reference corpora to approximate “the collective linguistic experience of a language
community” (Howarth, 1996: 72), and thus bypass the data sparseness bottleneck
inherent in the corpora currently available to translation scholars. Section 3 below
describes the method in more detail.

3. Studying collocations in translated language

3.1 Research questions

This study addresses the following research questions:

1. Are translated texts more/less collocational than original texts in the same
language?
• i.e., the collocation types they contain are more/less frequently attested
and/or significant than the collocation types found in originals?

2. If any difference can be identified, is it likely to be a consequence of the translation

process?
• i.e., can we isolate shifts (less-to-more collocational or more-to-less
collocational) that can point us towards possible reasons for the observed
differences?

Question 1 requires a monolingual comparable corpus and a reference corpus of the

source and target languages, while question 2 requires a parallel corpus.

3.2 Corpus resources

Two tiny parallel corpora are used for this study, one containing extracts from novels
and short stories in original and translated English (source language: Italian), the other
containing similar extracts in original/translated Italian (source language: English).
Details of these corpora, referred to below as the LIT corpora, are provided in tables
1-2. 2

2 A corpus containing open source software manuals was also analysed, to check whether the same
tendencies would be observable in literary as well as technical translation. For reasons of space,
results regarding this second corpus are not discussed here.

4
 Sample Sample
Author Title (IT) Translator Title (EN)
size size
F. Camon La malattia 16,230 J. Shepley Sickness called 18,074
chiamata uomo man
G. Celati I narratori delle 19,144 R. Lumley Voices from the 20,903
pianure plains
C. Comencini Le pagine strappate 23,219 G. Dowling The missing 27,199
pages
Luther Blissett Q 16,295 S. Whiteside Q 18,247
D. Maraini Donna in guerra 17,669 D. Kitto, E. Woman at war 19,531
Spottiswood
G. Pontiggia Il giocatore 12,408 A. Cancogni The invisible 14,962
invisibile player
G. Tomasi di Il Gattopardo 22,275 A. Colquhoun The Leopard 23,816
Lampedusa
Total 127,240 142,732 269,972

Table 1: Composition of the LIT corpora: the Italian=> English sub corpus

Author Title (EN) Sample size Translator Title (IT) Sample size
M. The handmaid's tail 15,647 C. Penati Il racconto dell'ancella 15,184
Atwood
M. Cat's eye 15,146 M. Papi Occhio di gatto 15,134
Atwood
M. Cruz Gorky Park 10,863 P. F. Gorky Park 10,181
Smith Paolini
C. Red bride 12,350 S. Bini Nozze di sangue 12,566
Fowler
N. My son's story 13,999 F. Storia di mio figlio 14,897
Gordimer Cavagnoli
G. The tenth man 11,916 B. Oddera Il decimo uomo 12,284
Greene
D. A place I've never been 15,010 A. Cossiga Un luogo dove non sono 15,476
Leavitt mai stato
R. Kissing the gunner's 14,329 H. Brinis Oltre il cancello 14,284
Rendell daughter
Total 109,260 110,01 219,266

Table 2: Composition of the LIT corpora: the English=> Italian sub corpus

We might describe this resource as a very small and opportunistically built

bidirectional corpus (Johansson, 2000), i.e., a combination of parallel and
monolingual comparable corpus resources. Yet the texts included in each parallel sub-
component differ considerably from each other, such that doubts about their
comparability are not unwarranted. Italian novels in translation tend to be more
highbrow and to have been published by niche publishers, while English originals,
with some exceptions, typically belong to more low-brow, mass-market fiction. These
characteristics reflect real-world tendencies in the translation market, and cannot be
swept under the carpet. They should be kept in mind when attempting to relate the
results of the comparable corpus analysis and of the parallel investigation of
translation shifts to the wider socio-cultural norms regulating translation – an aspect

5
relevant to theoretical (more than descriptive) translations studies, and beyond the
immediate concerns of this paper.
The reference corpora used in this study are:3

1. The British National Corpus (BNC) for English (100 million words from
various sources)
2. The Repubblica Corpus for Italian (340 million words from a single
newspaper)

These corpora are a) not comparable with the study corpora, i.e. they are not made of
fiction texts and b) not comparable with each other (one being a “balanced” corpus,
the other a single-source corpus). These should not constitute major problems for the
present purposes. With regards to a), the point is to use the reference corpus as a
repository of collocations that language users would recognise as well-established,
filtering out sequences produced by the operation of the open-choice principle
(Sinclair 1991); a fiction corpus, being potentially rich in creative combinations,
might actually be detrimental. The non comparability of the two reference corpora (b)
could constitute a potential problem at the stage where we try to draw conclusions
about the universality of the claims, i.e. whether the shifts we observe apply to
English and Italian to the same extent, and therefore could be candidates for
“universal” status. Yet at the stage where we compare original and translated texts in
the same language, no bias is inserted due to the choice of the reference corpus.

3.3 Corpus preparation

The reference corpora were already available (tagged, lemmatised and indexed with
the Corpus Work Bench (CWB, Christ 1994)). The LIT corpora were:

1. scanned in from the paper sources

2. tokenised
3. tagged
4. lemmatised
5. indexed
6. sentence aligned

Steps 2-4 were carried out by the tree-tagger, a freely-available language-independent

tagger pre-trained on English and Italian, as well as a few other languages (Schmid,
1994). Steps 5-6 are taken care of by CWB. At this point the different sub-corpora
could be searched using the Corpus Query Processor (CQP), the interrogation
companion to CWB.

3.4 Extraction of candidate collocations

For the present purposes, and in order to make the data set manageable, the object of
study was arbitrarily restricted to collocations made of two lexical words that are
either contiguous or separated by at most two function words. POS patterns matching
these criteria that are likely to yield lexical collocations were then obtained from the

3 Data were also collected from the Web through automatic queries to Google. These will be used in
follow up studies that attempt to evaluate the effects of a massive scaling up of the reference corpus
size on the results obtained.

6
available literature (Benson et al., 1997; Dzierżanowska and Kozłowska 1999; Oxford
collocations dictionary for students of English 2002; Jezek 2005; Voghera 2004).
Examples of the patterns selected for the study are listed in table 3.

Italian English
Adj-Noun Adj-Noun
contiguous Noun-Adj Noun-Noun
Verb-Noun Verb-Noun
1 intervening Noun-prep|conj-Noun Noun-prep|conj-Noun
function word Verb-prep-Verb Adj-conj-Adj
2 intervening Noun-prep-pron-Verb Verb-pron-pron-Noun
function words

Table 3: Example patterns retrieved from the LIT corpora

3.5 Evaluation

All the sequences matching a given pattern are retrieved from the LIT corpora, and
frequency information about the combination and about its constituents is obtained
from the relevant reference corpus (see table 4).

Original (BNC) Translated (BNC)

W1 W2 Fq1 Fq2 Fq1-2 W1 W2 Fq1 Fq2 Fq1-2
absence game 5625 13978 1 act deception 11021 642 1
absence pain 5625 6690 2 act foundation 11021 2118 1
absence people 5625 113684 6 activities rules 11091 9624 1
aches pains 162 1007 69 admission guilt 1998 1547 17
act adultery 11021 248 5 admission order 1998 31665 1

Table 4: Sequence types matching the N prep|conj N pattern and their BNC frequency data

The next step consists in calculating Mutual Information (MI; Church and
Hanks, 1990) values for each sequence using the UCS toolkit (Evert, 2004-2006).

Original (BNC) Translated (BNC)

MI W1 W2 Fq1 Fq2 Fq 1-2 MI W1 W2 Fq1 Fq2 Fq1-2
6.1765 Uncles aunts 3 222 1 7.0731 Snakes Ladders 52 13 8
5.9863 narcissi hyacinths 24 43 1 5.9385 constancy inconstancy 72 16 1
5.8184 aunts uncles 222 219 32 5.7706 knives forks 609 181 65
5.7604 uncles aunts 219 222 28 5.7590 scribes Pharisees 95 110 6
5.7180 frogs toads 393 151 31 5.5936 forks spoons 181 169 12

Table 5: Sequence types matching the N prep|conj N pattern, ranked according to their MI values

7
 Once the results are ordered as in table 5, all sequences with an MI>2 and a
fq>1 are selected for further analysis. These are arbitrary cut-off points that take into
account the relatively small size of the reference corpora. Further work is required to
find empirically the most appropriate cut-off point relative to size/specialisation of the
reference corpus, as well as to evaluate the effects of other statistical measures of
collocativeness. As will be seen, however, the combinations selected for further
analysis would in general appear to match intuitions about
institutionalisation/lexicalisation. For our purposes, as long as the same cut-off point
is used for both translated and original language, no bias is inserted.
The final step of the monolingual analysis consists in comparing the
distribution of MI and frequency values in each matching original/translated ranking.
The Mann-Whitney significance test is used for this purpose. If the test returns a
significant (p ≤ .05) or near-significant difference, descriptive statistics are obtained
for the relevant ranking pair to determine which of the two rankings
(original/translated) displays higher/lower values for either MI or (LOG)FQ. Table 6
summarises these data for the Noun prep|conj Noun pattern in Italian. The difference
between the two MI distributions is statistically significant (p-value = 0.007834),
while that between the LOG(fq) distributions is not (p-value = 0.8724).

MI
MI
statistic original translated
6,5
6
min 2.001 2.000
5,5 q1
q1 2.381 2.425
5 min
4,5
median median 2.736 2.853
4
3,5 max
q3 3.392 3.590
3 q3
2,5 max 5.757 6.059
2
original translated mean 2.954 3.069

Table 6: Descriptive statistics for the N prep|conj N pattern in Italian (MI)

This result may be partly due to the different number of candidates in each set
(691 original vs. 855 translated). Notice though that this difference in size between the
two sets is not an artifice of the research setup, i.e., it is not a consequence of a
difference in size between the two sub-corpora. Prior to the comparisons, the number
of token bigrams for each pattern in the larger file was artificially reduced to the
number of occurrences in the smaller matching file (through random sampling), thus
ensuring that any subsequent difference was not due to this initial bias. Furthermore,
significant or near-significant differences were reported for rankings with the same or
very similar numbers of observations (e.g. N prep|conj N in English).
Based on this monolingual comparison, we can hypothesise that Italian
translators tend to make use of N prep|conj N established sequences (potential
collocations) more than Italian authors do. To confirm this hypothesis, we need first
of all to ascertain that the difference observed is indeed relatable to the process of
translation, and not to other variables, e.g. to the differences in corpus set-up
described in 3.2 above. To investigate this issue, we need to look at parallel

8
concordance data.

4. Case study: N prep|conj N sequences

1,061 parallel concordance lines were browsed and clear cases of shifts that might
affect the collocativeness of the pattern observed in the TT were extracted. Even
though this procedure has the methodological limit of not accounting for all the data,
it does have the advantage of providing a conservative estimate: in case of doubt, the
concordance line was assigned to the “irrelevant” set. Altogether, 127 shifts (11.9
percent of concordance lines) were found and grouped into sets according to the
strategy that might have motivated the shift. Tentative labels were then provided for
these sets, following a bottom-up approach. Clearly, the data available only allow us
to make informed guesses about translator motivation; other analytical tools would be
needed to proceed in this direction (i.e., think-aloud protocols, corpora including
different drafts of the same translation and so forth).
Table 7 summarises the shift types observed and the number of times they
occur in the data. For reasons of space, only a single example per shift will be
discussed (4.1-4.5).

Shift type n.
creative Æ institutionalised 7
institutionalised Æ institutionalised 7
(different meaning)
free Æ institutionalised 11
more explicit 86
more formal/precise 16
total shifts observed 127
total concordance lines analysed 1,061

Table 7: Shifts leading to increased institutionalisation: A quantitative estimate

4.1 Creative Æ institutionalised

In a few cases the institutionalised noun phrase in Italian would appear to result from
a process of normalisation or “sanitisation” (Kenny, 2001), in which an unusual or
stylistically marked expression in the source text is rendered with a more common,
recognisable phrase in the target text. Cf.:

[1] TT: l’odore della terra smossa , il <senso di pienezza> che davano le forme tonde dei bulbi
chiusi nella mano
(the smell of the turned earth, the sense of fullness that gave the round shapes of the bulbs
held in the hand)
ST: the smell of the turned earth, the plump shapes of bulbs held in the hands, fullness
The handmaid’s tale

In example [1], the noun fullness in the source text, isolated form the surrounding text
by two commas, follows and comments on the scene just described (the plump shapes
of bulbs held in the hands). In the target text, the syntax is normalised and made more

9
explicit (lit.: “the sense of fullness that gave the round shapes of the bulbs held in the
hand”), and coherently the rather vague noun “pienezza” is resolved as “il senso di
pienezza”, using a familiar Italian phrase (14 occurrences in Repubblica).

4.2 Institutionalised Æ institutionalised (different meaning)

This set groups those cases in which the tendency is not towards greater explicitness
or normalisation in translation, but rather towards preserving a level of
“collocativeness” similar to that of the source text. When an expression preserving
both the denotative meaning and the collocative value of the source text expression is
not immediately available to the translator, s/he may decide to favour the latter at the
expense of the former.

[2] TT: Fa collezione di <cartine di sigarette> con disegni di aeroplani, e ne conosce tutti i nomi.
ST: He collects <cigarette cards> with pictures of airplanes on them, and knows the names of
all the planes.
Cat’s eye

Cigarette cards “were issued by tobacco manufacturers both to protect the cigarettes
by stiffening the pack, and also to gain customer loyalty to their particular brand of
cigarettes. The cards […] are considered […] one of the first collectibles”. 4 Cigarette
cards did not exist in Italy, though cards did: they accompanied products like Lavazza
coffee and Liebig stock cubes. The latter are commonly known as “figurine”. By
analogy with these products, a possible translation for cigarette cards could have been
“figurine delle sigarette”. This option would select a free combination as a translation
for an institutionalised phrase (see table 8 below). The translator instead has chosen to
employ a more institutionalised phrase in Italian, i.e. “cartine di sigarette”, which
however has a very different meaning, namely rolling papers. While this choice may
create a misunderstanding if one stops to think about it, it is very likely that readers do
not stop at all. One of the effects of using run-of-the-mill expressions is exactly that
they do not draw attention to themselves.

Expression Attestedness (occurrences) Meaning

Cigarette cards BNC: 16 Collectible cards found in
Google: 491,000 cigarette packs
Cartine da/per/di/delle Repubblica: 3 Rolling papers, i.e. small sheets
sigarette Google: 726 of paper which are sold for
rolling one's own cigarettes
Figurine da/per/di/delle Repubblica: 0 (by analogy with other
sigarette Google: 2 products) collectible cards
found in cigarette packs

Table 8: cards and “cartine” compared (Google data obtained on 29 June 2007)

4.3 Free Æ institutionalised

Examples in this category are clear instances of normalisation, i.e. cases where the
translator has used an institutionalised Italian collocation to render an English free

4 http://en.wikipedia.org/wiki/Cigarette_card (accessed: 29 June 2007)

10
 combination. A rather straightforward case is [3]:

[3] TT: decorazioni di <spicchi d'aglio>, si rende conto che

ST: handpainted by Alex with purple garlic bulbs, she sees that
A place I’ve never been

Here the English garlic bulbs was rendered as “spicchi d’aglio” (garlic cloves)
in Italian. There is no question that the translator may have got the term wrong, as
bulb is “bulbo” in Italian, or more appropriately in the case of garlic, “testa” (head).
Besides, serving dishes are typically decorated with bulbs, not cloves of garlic in Italy
(and indeed, the scene is set in Tuscany). A tentative explanation for this choice could
be that the translator (rightly) felt that “bulbi d’aglio” and “teste d’aglio”, while
perfectly acceptable Italian phrases, would not be as familiar to the reader as “spicchi
d’aglio”. It is interesting at this point to consider the relative frequencies of the
different phrases in English and in Italian (table 9, frequency data from the Internet
are given as some of the phrases have no occurrences in the BNC/Repubblica).

Expression Attestedness Attestedness Expression

Garlic 34,300,000 (100%) 2,580,000 (100%) aglio
garlic bulbs + bulbs of garlic 109,600 (0.31%) 1305 (0.05%) bulbi d’aglio + bulbi di aglio
garlic heads + heads of garlic 59,300 (0.17%) 612 (0.02%) teste d’aglio + teste di aglio
garlic cloves + cloves of garlic 2,207,000 (6.43%) 229,100 (8.87%) spicchi d’aglio + spicchi di aglio

Table 9: bulbs and “spicchi” compared (Google, 29 June 07)

Taking these data to somehow approximate the perception of speakers of the

two languages, we might conclude that the Italian translator did not have at her/his
disposal an expression at exactly the same level of familiarity/collocativeness as the
one used in the ST. Faced with the choice to either employ a slightly less familiar
phrase (“teste d’aglio” o “bulbi d’aglio”) with the same denotational meaning, or a
much more familiar phrase with a different though related meaning, s/he may have
opted for the second.

4.4 Explicitation

4.4.1 Straightforward explicitation

[4] TT: l’avrei schiacciato sotto il <tacco della scarpa>, seppellito.
ST: ground away under my heel, buried along with it.
My son’s story

In example [4], an of-construction in Italian (“N di N”) is used to make the

meaning of the English ST more explicit: while English has heel, Italian specifies
exactly what heel, (i.e., the heel of the/one’s shoe). Notice that the Italian head noun
has exactly the same denotation as its English counterpart: with no further
specification, they would be taken to refer to the same objects.

11
4.4.2 Partitives

Very often the text is made more specific (and more familiar) through the insertion of
(quantity) partitive nouns. In example [5], a partitive construction is formed with a
noncount noun and a “typical” partitive that is restricted in its combinability:
[5] TT: Non riuscivo a prendere sonno , così sono sceso a bere un <sorso d'acqua> . [gulp of
water]
ST: I couldn’t sleep , so I came down for water.
The tenth man

4.4.3 Specificity of head nouns

A somewhat more subtle form of explicitation is also observable. In several cases,

Italian and English use a similar phrase, but the head noun in Italian (being more
specific or explicit than its English equivalent) could alone express the meaning of the
whole English expression. Since exactly the same level of explicitness could not be
obtained, the translator has opted for the more explicit solution, which is also an
institutionalised phrase in the language. Cf. example [6]:

[6] TT: anello giallastro e parecchi <mozziconi di sigarette> che galleggiano

ST: yellowish-brown ring around the inside of the bowl and several floating cigarette butts
Cat’s eye

In the BNC, butt collocates with cigarette, cigar and especially water and
rifle. In Repubblica, “mozzicone” in its non figurative use collocates with “sigaretta”
(cigarette), “sigaro” (cigar) and “candela” (candle). Both cigarette butt and
“mozzicone di sigaretta” are institutionalised sequences in the two languages, and
dictionary equivalents of each other. However, if we look up the corpora for
occurrences of butt not accompanied by cigarette (or any other noun), and
“mozzicone” not accompanied by “sigaretta” (or any other noun), we notice some
differences (concordances 1-2).

12
1 new building. Stubbing out <his butt,> Stan shudders at the
2 to " get Central Office off <its butt,> and put the Governme
3 w status as the comedians ' <favourite butt.> There were so
4 oys taking snapshots of his <own butt,> and loses his dad’s
5 h close simultaneously by a <direct butting,> as distinguis
6 er, were eager to be out at <the butts.> Elizabeth herself h
7 ow by late June with a lush <green butt,> with heads formed b
8 still public land, known as <The Butts.> The library itself
9 greenheart. With a spike in <the butt.> For sticking into th
10 starting at 1.30 p.m. from <The Butts.> The race, organised by
11 werful as it bends towards <the butt.> It was designed for li
12 corner of my eye. I grab <the butt,> with my fingers curled a
13 rifle furniture -- stocks <and butts.> There was only one ma
14 ered, and the hole through <the butt,> which was counterbored
15 bolt, the head to the foot <of butt,> and held the oil bottle
16 ptain’s " Let's go back to <the butts,> gentlemen " is the sign
77 n the joystick and go kick <some butt.> Chris Nicholls from P
18 rom inexperienced dealers. <No butts...> I have a five year ol
19 factory they 're actually <half butts,> for no other reason
20 the shape and condition of <the butt,> Klondyke would expect t
21 Verlaine with a rocket up <his butt --> hardly the sound of a
22 s all the time. " You see <her butt?> What a bitch of a butt,b
23 her butt? What a bitch of <a butt,> boy! " Sooner or later on
24 Gambling is a big pain in <the butt,> " he says. " A lot of c
25 inished it and tossed away <the butt,> will add to the rubbish

Concordance 1: BUTT followed by punctuation and not preceded by a noun (BNC)

1 del sesto è rimasto " un <mozzicone ">. Le rimesse dei liba

2 si a gettare per terra un <mozzicone,> però, e un pezzo di
3 cipale responsabile, il " <mozzicone ">, che ha un alto cont
4 erà 50.000 lire chi getta <mozziconi,> lattine e cartacce s
5 tacenere conserva un solo <mozzicone,> nessun cuscino non è
6 traccia, scatole vuote, <mozziconi,> ogni genere di rifiuti
7 ltsin -- mi ha risposto a <mozziconi,> ha detto che Gorbaci
8 a manipolato in seguito i <mozziconi )>, ma scientificament
9 ri, portaceneri pieni di <mozziconi,> e libri gialli; un co
10 di cui si raccolgono solo <mozziconi "> Basta... Anch' io sono
11 il tabaccaio ritirava il <mozzicone,> lo spegneva e lo rivende
12 volontariamente ". Questo <mozzicone,> è possibile -- continua
13 mmacolate -- non c’era un <mozzicone -->, immacolati la scuola
14 te, il piattino pieno di <mozziconi;> e impepate di cozze, me
15 , terreno mitragliato di <mozziconi,> si pregano gli spettator
16 on mano una candela; o un <mozzicone,> un moccolo della medesim
17 n la sabbia senza trovare <mozziconi ">, ha detto l’assessore
18 che il giallo parte da un <mozzicone,> ma in questo episodio i
19 ri che la colpa era di un <mozzicone,> di una stufetta, la cla
20 sua casa non restano che <mozziconi,> per la strada non c’è n
21 ndo per terra bottiglie e <mozziconi,> come in una finale di Co
22 izzato in michette... al <mozzicone (> multa di 850 mila lire )
23 che una campagna contro i <mozziconi,> le cicche, quei pezzett
24 rigano, sono in fondo un <mozzicone --> un plastico, un ricordo
25 a Lagorio: " Miliardi di <mozziconi,> graffiti sui muri. Non
26 to la Fenice a meno di un <mozzicone:> imputati di strage e di
27 per un incendio basta un <mozzicone,> i bianconeri si sentono

Concordance 2: "MOZZICONE" followed by punctuation (Repubblica)

Out of 58 occurrences of the lemma BUTT followed by punctuation and not

preceded by a noun in the BNC, only 2 lines refer to the smoking sense (Concordance
1 shows a selection of the hits found). If we compare a concordance from La
Repubblica, we find instead that two thirds of the concordance lines refer to cigarette
butts. In other words, while butt does not imply cigarette (i.e., a butt is not necessarily

13
understood as a cigarette butt, but more likely as a body part or part of a weapon),
“mozzicone” does imply “sigaretta” (i.e., “un mozzicone” is almost certainly to be
understood as “un mozzicone di sigaretta”). Therefore, while cigarette butt and
“mozzicone di sigaretta” may appear to be obvious decontextualised equivalents of
each other, the decision to translate cigarette butt with “mozzicone di sigaretta” in
fact results in a target text that is more explicit than its source text, and one that adopts
an established restricted combination in Italian where there is no strict need for it.

4.5 More formal/precise

This set groups cases where the expression used in Italian a) has an alternative more
informal or more vague rendering that would be equally adequate and (possibly) more
acceptable in context, and/or b) displays a concern with exactness of expression that
does not follow from the ST.

[7] TT: Spostando col piede i <capi di vestiario> sul pavimento , non trovò traccia della prova
incriminante. [items of clothing; instead of “vestiti” (clothes)]
ST: Kicking around among the clothes on the floor, he found no trace of the incriminating
article.
Red bride
[8] TT: Da un bel pezzo, un’eternità, non provava più un vero e proprio <senso di nausea> davanti
a spettacoli del genere. [a sense of nausea]
ST: It was a long time, an age, since he had felt actual physical nausea at such sights.
Kissing the gunner’s daughter

5. Concluding remarks

This paper has described a method for comparing use of collocations in original and
translated language that relies on monolingual comparable corpora and reference
corpora. Combining this monolingual perspective with a parallel focus, a case study
has been subsequently presented of the Noun preposition|conjunction Noun pattern in
original and translated Italian. Going back to our research questions, and relying only
on the analysis of a single collocation pattern, we can tentatively suggest that
translated texts would seem to be more collocational than original texts in the same
language, and that there is some evidence that this is a consequence of the translation
process. 5
Overall, collocations following this pattern have higher MI values in translated
than in original fiction texts. Over a thousand parallel concordance lines were
analysed, and almost 12 percent of these showed shifts towards greater
collocativeness. In many cases, the increase in collocativeness was accompanied by
other phenomena, i.e. normalisation and explicitation (the quantitatively most
prominent set). It is difficult to tell what was the principal driving force behind these
shifts. Yet the very relationship we observed between (some of) the regularities of
translation behaviour observed in the literature (cf. 2.2 above) and the (higher) level of
institutionalisation of the language used in translated texts is in itself an intriguing
finding, that deserves further investigation.
Several directions for further study can be taken at this point. To shed light on
the “universality” of the shifts described above, this method could be applied to other
mediating contexts (e.g. technical translation, interpreting) as well as to other

5 Analyses of all the significantly different ranking pairs (not reported on here) are consistent with this
claim.

14
language pairs. To confirm the hunches about translator motivation, corpus-based
(product-oriented) studies could be combined with process-oriented approaches to
translation research. Lastly, from a methodological point of view it would be
important to experiment with different statistical measures, frequency cut-off points,
and reference corpora.

References

Baroni, M. and S. Bernardini (2003). A Preliminary Analysis of Collocational

Differences in Monolingual Comparable Corpora. In D. Archer, P. Rayson, A.
Wilson and A. McEnery (eds.) Proceedings of Corpus Linguistics 2003, pp.
82–91. Lancaster: UCREL.
Benson, M., E. Benson and R. Ilson (1997). The BBI dictionary of English word
combinations. Amsterdam: Benjamins.
Christ, O. (1994). A Modular and Flexible Architecture for an Integrated Corpus
Query System. COMPLEX ’94, Budapest 1994.
Church, K.W. and P. Hanks (1990). Word Association Norms, Mutual Information,
and Lexicography. Computational Linguistics 16,1: 22–29.
Danielsson, P. (2001). The Automatic Identification of Meaningful Units in
Language. Unpublished PhD Thesis. Göteborg University.
Dzierżanowska, H. and C. Kozłowska (1999). LTP Dictionary of Selected
Collocations (edited by J. Hill and M. Lewis). Hove: LTP.
Evert, S. (2002–2004). The UCS Toolkit. Available online from
http://www.collocations.de/software.html (accessed: 29 June 2007).
Firth, J.R. 1956. Descriptive linguistics and the study of English. in F.R. Palmer (ed.)
(1968). Selected papers of J.R. Firth 1952-1959. London and Harlow:
Longman. 96–113.
Howarth, P. (1996). Phraseology in English Academic Writing. Tübingen: Niemeyer.
Jezek, E. (2005). Lessico. Classi di Parole, Strutture, Combinazioni. Bologna: Il
Mulino.
Johansson, S. (2000). “Reflections on Corpora and their Uses in Cross-linguistic
Research”, in F. Zanettin, S. Bernardini and D. Stewart (eds.) Corpora in
Translator Education, pp. 135–44. Manchester: St Jerome.
Jones, S. and J. McH. Sinclair (1974). English Lexical Collocations. Cahiers de
Lexicologie 24: 15–61.
Kenny, D. (2001). Lexis and Creativity in Translation. Manchester: St. Jerome.
Kjellmer, G. (1987). Aspects of English Collocations. In W. Meijs (ed.) Corpus
Linguistics and Beyond. Proceedings of the seventh international conference
on English language research on computerised corpora, pp. 133–40.
Amsterdam: Rodopi.
Laviosa, S. (2002). Corpus-based Translation Studies: Theory, Findings,
Applications. Amsterdam: Rodopi.
Malmkjær, K. (2004). Translational Stylistics: Dulcken’s Translations of Hans
Christian Andersen. Language and Literature 13,1: 13–24.
Manning, C. and H. Schütze. (1999). Foundations of Statistical Natural Language
Processing. Cambridge, MA: MIT Press.
Marco, J. (2004). Translating Style and Styles of Translating: Henry James and Edgar
Allan Poe in Catalan. Language and Literature 13,1: 73–90.
Mauranen, A. (2000). Strange Strings in Translated Language: A Study on Corpora.
In M. Olohan (ed.) Intercultural Faultlines. Research Models in Translation
Studies 1: Textual and Cognitive Aspects, pp. 119–41. Manchester: St.

15
 Jerome.
Nesselhauf, N. (2005). Collocations in a Learner Corpus. Amsterdam: John
Benjamins.
Olohan, M. (2001). Spelling out the Optionals in Translation: A Corpus Study. In P.
Rayson, A. Wilson, A. McEnery, A. Hardie and S. Khoja (eds.) Proceedings of
Corpus Linguistics 2001, pp. 423–32. Lancaster: UCREL.
Olohan, M. (2004). Introducing Corpora in Translation Studies. London: Routledge.
Olohan, M. and M. Baker (2000). Reporting that in Translated English: Evidence for
Subconscious Processes of Explicitation? Across Languages and Cultures 1,2:
141–58.
Øverås, L. (1998). In Search of the Third Code: An Investigation of Norms in Literary
Translation. Meta 43, 4: 571–88.
Oxford Collocations Dictionary for Students of English (2002). Oxford: Oxford
University Press.
Schmid, H. (1994). Probabilistic Part-of-Speech Tagging Using Decision Trees.
Proceedings of International Conference on New Methods in Language
Processing. September 1994.
Sinclair, J. McH. (1991). Corpus, Concordance, Collocation. Oxford: Oxford
University Press.
Tirkkonen-Condit, S. (2004). Unique items – Over- or Under-represented in
Translated Language?, in A. Mauranen and P. Kujamäki (eds.) Translation
Universals. Do they Exist?, pp. 175–84. Amsterdam: John Benjamins.
Toury, G. (1995). Descriptive Translation Studies and Beyond. Amsterdam:
Benjamins.
Voghera, M. (2004). Polirematiche, in M. Grossman and F. Rainer (eds.) La
formazione delle parole in italiano. Tübingen: Niemeyer.

16

View publication stats