تويت
What makes a text difficult to translate?
Stuart Campbell and Sandra Hale
Language Acquisition Research Centre
University of Western Sydney Macarthur
Introduction
This paper reports on a project under way at the University of Western Sydney Macarthur, which is exploring the question of source text difficulty in translation. We begin by briefly reviewing the notion of source text difficulty, and then report on our research which explores how universal translation difficulties might be. Comprehension and production difficulties are distinguished, and we then go on to ask whether production difficulty can be compared across target languages. We present examples drawn from empirical work on translations from English into Arabic, Spanish and Vietnamese. We end with a brief discussion of the practical relevance and future directions of the research.
The problem of difficulty in translation
For translator educators and accrediting bodies, a key problem is assessing the difficulty of source texts used in tests and examinations. While experienced translator educators no doubt have reasonably reliable instincts on text difficulty, the construction of an expert system that mirrors those instincts is a goal not yet attained. A study by Hill (1997) makes some headway, aiming to assess the reliability of French, German and Russian texts used by the National Accreditation Authority for Translators and Interpreters, and identifying five "difficulty indicators", which she labelled "thematic, formal, stylistic, linguistic and syntactic" (Hill 1997:10). However the study's reliance on subjective perceptions for its assessment of cross-linguistic difficulty indicates the need for methodological progress in this area.
Unlike Hill's study, the project reported here focuses on English source texts, a choice dictated by the fact that English is probably the most widely translated language in the world, and that it is common practice to use the same source texts to examine translation into multiple languages. This practice raises two questions: Can the translation difficulty of English source texts be assessed? Does a given English text cause equal difficulty when translated into different languages?
Two possible sources of indirect assistance are available to us in considering text difficulty or complexity. The first is the field of readability, where numerous indices of text complexity have been proposed over the years. Measures such as the FOG index, for example, correlated perceived readability with fairly crude criteria such as sentence length and word length (See Chall 1958 for an overview of such measures), although such measures have been seriously criticised in recent years because of their lack of sophistication in the light of developments in text linguistics and discourse analysis (See Davison & Green 1988 for a recent reconsideration and critiques of readability).
A second source is research that links text types with their microlinguistic exponents. Works like Biber (1988) have catalogued the lexical and grammatical features of a large number of text types, and give us important clues as to what makes a text more complex. This kind of research has had important practical consequences for the Plain English movement, which has had successes in improving the readability of English texts (Charrow & Charrow 1979). It should be remembered, however, that the complexity of an English text for monolingual reading may not equate to its complexity in reading for translation. de Groot (1997), for example, has questioned whether reading, decoding, encoding and writing always occur in sequence in translation. It may be that translators read in rather a different way from monolingual readers (Danks & Griffin 1997) perhaps incorporating some aspect of pretranslation processing as they read.
In addition, the special nature of translation suggests that the source text is only one element of difficulty. In a paper recently submitted to Target, Campbell suggested that text difficulty is one of three variables to be considered in exploring translation difficulty, the others being translator competence and translation task type. The paper goes on to propose a cognitive model of translation difficulty which, in brief, holds that difficult items in a text are those that require more processing effort to resolve target language choices.
How universal are translation difficulties?
Our research on translations of English texts into Arabic, Spanish and Vietnamese finds that a considerable portion of text difficulty is common to translation into all three languages. This conclusion was reached on the basis of a study of the alternative translations provided by groups of translators translating the same texts into the three languages. Where subjects in one language provided different renditions of a chunk of text, that chunk was judged to be difficult on the basis of the processing effort needed to decide among the various choices available. If the chunk caused a similar level of difficulty in all three languages, then it was judged to be universally difficult. We concede that the term universal is ambitious here, but argue that since the three target languages are non-cognate, we have a strong case. The study, which focussed on lexis and grammar, identified several areas of difficulty:
• Words low in propositional content, ie. stretches of text where propositional content is distributed over several words, one of which - characteristically a verb - is apparently low in meaningfulness, for example in become free from all opioid use, which can be paraphrased as stop using opioids altogether, where become merely relates a previous state to a present state. Looked at from a discourse perspective, this type of difficulty is an example of grammatical metaphor, where a marked or longer route is used to achieve the same propositional content; according to Halliday, grammatical metaphor is typical of the register used in written language and makes text "inherently complex"(Halliday 1985:329). Indeed, some of the Spanish subjects translated such items into the congruent form (to use Halliday's term): to have schools and workplaces free from discrimination became conseguir erradicar la discriminaciر® ...de las escuelas (to eradicate discrimination from the schools) or librar de (free schools from).
• Complex noun phrases, ie. noun phrases where two or more nouns are concatenated, and there is difficulty in interpreting or encoding the relationships between the head and the other items in the string. An example is Methadone treatment, where methadone is in an instrument relationship with treatment.. Among the Arabic subjects in the study, two interpreted the relationship as patient; their renditions could be glossed as the treatment of Methadone. Among the Spanish subjects, there was an initial suspicion that some had also made a patient interpretation because of renditions like Tratamiento de Metadona rather than Los tratamientos con metadona. However, a close analysis of the coding of instruments indicated that there is a continuum of explicitness in the choice of preposition. The typical range includes:
inexplicit de
explicit con, por
highly explicit con m todo de, por medio de, por administraciَn de.
Thus, for the Spanish subjects, the focus of difficulty was on encoding the instrument relationship and deciding on the degree of explicitness. For the Arabic subjects, there was a degree of similar difficulty, but the choices were fewer, ie. between the simple preposition bi- "with" and the elaborated preposition biwasit`a "by means of"; the focus of difficulty was perhaps on interpreting the relationship rather than encoding it.
• Abstractness: Words like practice, contact and action were more difficult to translate than workplace, Department, school, and person, a finding in line with experimental research on monolingual and bilingual lexical retrieval. For example, Schwanenflugel, Harnishfeger & Stowe (1988) discuss a dual-representation model, which assumes that concrete words are more easily understood because they have both a verbal and an imaginable representation; they compare this with a context availability model that proposes that abstract words are harder to understand because of the difficulty in retrieving contextual information to aid their comprehension. de Groot's (1993) work in bilingual lexical representation proposes that concrete words have a single mental representation, while the representation of abstract words is language-dependent.
• Official terms: The problem with items Anti-Discrimination Board and contact officer may be associated with the widely known difficulty of finding equivalents for official titles. This category is one associated with both comprehension and production, and is largely a consequence of the dense semantic packing of such terms.
• Passive verbs: These evidently posed difficult decisions for translators, probably more in the area of encoding than interpretation. Both Arabic and Spanish have quite different grammatical strategies to deal with English passives; examples are given in a later section of this paper.
Work to identify more of these "universal" difficulties, as well as language-specific difficulties, is proceeding with the use of a database of translation examination scripts based on of various English source text types.
Comprehension versus production difficulties
One of our interests is to find out if translators working into different languages confront similar levels of difficulty. This challenge is expressed in practical terms by the question "How reliable is a translation test that uses the same English source text for different target languages?" Questions of language typology arise here. Hill (1997) asked the mirror image question - whether it was possible to construct translation tests in different source languages of similar difficulty, and claimed that French was easier to translate into English than Russian into English. But the justification for such claims needs to be considered with caution; how confidently can we claim that French is somehow structurally closer to English? And at any rate, our study found that there were English source text difficulties common to three unrelated languages - Arabic, Spanish and Vietnamese. This necessarily leads us to a consideration of comprehension and production, a matter we hinted at in the discussion of Methadone treatment. . What we suspect to be the case is that there are two loci of difficulty. The first locus is comprehension, and difficulty is likely to be fairly universal. The second locus is production, and here there are likely to be different levels of difficulty depending on the lexis and grammar of the target language. The example of Methadone treatment above began to tease out these separate loci of difficulty: The Arabic subjects may have found comprehension more difficult, and the Spanish subjects may have found production more difficult.
Is production difficulty measurable?
If, as we have claimed, there may be different levels of production difficulty depending on the target language, can the difference in difficulty somehow be measured, or at leats compared in a principled way? One of the tools we are developing to test this idea is what we call a choice network, in effect a psycholinguistic model of the options facing the translator, based on the evidence of the target texts of groups of subjects translating the same text. Examples of choice networks are given in Figures 1 and 2. Here we have analysed the translations of the chunk The patient will be closely monitored and drawn up the most parsimonious networks that will account for all the data pertaining to grammatical choices. Note that we have not included word choices in these networks. As can be seen, the networks are comparable in quantitative complexity, but quite different qualitatively. While both groups are faced with a three-way choice of sentence types, the Arabic subjects need to grapple with the option of expanding patient, and whether to use a future particle. The Spanish subjects' concerns are focussed on the adverbial, with choices of word class and position. We believe that models like these - based on empirical data- can also be used to explore lexical and pragmatic choices. Indeed they are used in Campbell (1998) to model stylistic choices in translators working from Arabic to English.
Figure 1: Choice network based on nine translations of The patient will be closely monitored into Arabic.
Sentence type
_________________|____________________
| | |
nominalisation passive active
| | |
______ |_____________ __|________ |
| | | | |
| | | | |
patient patient patient patient
expanded to not expanded expanded to not expanded
patient's | patient's |
condition | condition |
| adverbial= | |
| prepositional | __________|
source phrase | |
text | adverb
adverbial ____ |_____ omitted
shifted to | | |
verb future | |
| particle | |
| used | |
| | |
|_________________________ | |
| | |
no future particle
Figure 2: Choice network based on nine translations of The patient will be closely monitored into Spanish.
sentence type
___________________________|______________________ _
| | |
active true passive passive with se
| | |
| copula= |
| serل |
copula = | |
estarل ______|______ ______|______
| | | | |
| adverbial= adverbial= adverbial= adverbial=
| true adverb prepositional true adverb
prepositional
complement= | phrase | phrase
prepositional | | | |
phrase _______|__________ | _____ ___| |
| | | | | |
| | | | | ____________________|
| | | | | |
| adverbial adverbial
| precedes follows
| main verb main verb
|
__ |__________
| |
adjective adjective
prededes follows
noun noun
How practical are our aims?
Is it possible to create or select texts of a given level of difficulty? Our research so far suggests that the most profitable avenue is to weight items in a translation text. Some items would carry a weight of zero - the ones that we predict will cause no difficulty to any subject. A great deal of common vocabulary might be included here. Harder items would carry higher weights, possibly with different weights for different item types. These would include problematic words and grammatical structures. The total text difficulty would be based on summing the weights.
We are not suggesting that only the weighted items be scored when target texts are assessed. It is axiomatic that word class and rank shifts occur in translation, and we believe that holistic marking of a text is the only way of dealing with the redistribution of meaning that occurs when a message is recast in a different grammatical and lexical guise; it is after all a text that is being translated, and not a string of discrete words and phrases. The point of weighting texts would be to create materials to meet the needs of students and examinees at different levels of development or proficiency.
Future directions
Our plans are to undertake a larger empirical study of target texts in our database to find out more about what makes texts universally difficult to translators into different languages. This, we hope, will lead to a set of model graded texts along with guidelines that teachers and assessors can use to grade their own texts. At the same time, we plan to continue developing a theoretical underpinning to the work based on psycholinguistic modelling of the translation process. This work will be much informed by our findings about the production dimension in specific languages.
References
Biber, D. (1988) Variation across speech and writing. New York: Cambridge University Press.
Campbell, S. (in press) 'A cognitive approach to source text difficulty in translation', Target.
Campbell, S. (1998) Translation into the Second Language. New York: Longman.
Chall, J.S. (1958). Readability: an appraisal of research and application. Columbus: Ohio State University.
Charrow V. and R. Charrow (1979) 'Making legal language understandable: A psycholinguistic study of jury instructions', Columbia Law Review 79, 1306-.
Danks, J.H. and J. Griffin (1997) 'Reading and translation: A psycholinguistic perspective' in J.H. Danks, G.M. Shreve, S.B. Fountain and M.K. McBeath (eds) Cognitive processes in translation and interpreting . Thousand Oaks, California: Sage Publications, pp. 161-175.
Davison, A. and G.M. Green (eds) (1988) Linguistic complexity and text comprehension: Readability issues reconsidered. Hillsdale, NJ: Lawrence Erlbaum Associates.
de Groot, Annette M.B. (1997) 'The cognitive study of translation and interpretation: Three approaches' in J.H. Danks, G.M. Shreve, S.B. Fountain and M.K. McBeath (eds) Cognitive processes in translation and interpreting . Thousand Oaks, California: Sage Publications, pp. 25-56.
de Groot, Annette M.B. 1993. 'Word Production and the Bilingual Lexicon' in R. Schreuder and B. Weltens (eds) The Bilingual Lexicon. Amsterdam/Philadelphia: John Benjamins, pp. 27-51.
Halliday, M.A.K. (1985) Introduction to functional grammar. London: Edward Arnold.
Hill, M. (1997) Translation difficulty. Does it vary according to language? Australian Language Matters, April/May/June: 10-11.
Schwanenflugel, P.J. , K.K. Harnishfeger and R.W. Stowe (1988) 'Context availability and lexical decisions for abstract and concrete words'. Journal of Memory and Language 27: 499-520.
Note
This research was carried out as part of a project funded by a University of Western Sydney Macarthur internal research grant.
Stuart Campbell and Sandra Hale
Language Acquisition Research Centre
University of Western Sydney Macarthur
Introduction
This paper reports on a project under way at the University of Western Sydney Macarthur, which is exploring the question of source text difficulty in translation. We begin by briefly reviewing the notion of source text difficulty, and then report on our research which explores how universal translation difficulties might be. Comprehension and production difficulties are distinguished, and we then go on to ask whether production difficulty can be compared across target languages. We present examples drawn from empirical work on translations from English into Arabic, Spanish and Vietnamese. We end with a brief discussion of the practical relevance and future directions of the research.
The problem of difficulty in translation
For translator educators and accrediting bodies, a key problem is assessing the difficulty of source texts used in tests and examinations. While experienced translator educators no doubt have reasonably reliable instincts on text difficulty, the construction of an expert system that mirrors those instincts is a goal not yet attained. A study by Hill (1997) makes some headway, aiming to assess the reliability of French, German and Russian texts used by the National Accreditation Authority for Translators and Interpreters, and identifying five "difficulty indicators", which she labelled "thematic, formal, stylistic, linguistic and syntactic" (Hill 1997:10). However the study's reliance on subjective perceptions for its assessment of cross-linguistic difficulty indicates the need for methodological progress in this area.
Unlike Hill's study, the project reported here focuses on English source texts, a choice dictated by the fact that English is probably the most widely translated language in the world, and that it is common practice to use the same source texts to examine translation into multiple languages. This practice raises two questions: Can the translation difficulty of English source texts be assessed? Does a given English text cause equal difficulty when translated into different languages?
Two possible sources of indirect assistance are available to us in considering text difficulty or complexity. The first is the field of readability, where numerous indices of text complexity have been proposed over the years. Measures such as the FOG index, for example, correlated perceived readability with fairly crude criteria such as sentence length and word length (See Chall 1958 for an overview of such measures), although such measures have been seriously criticised in recent years because of their lack of sophistication in the light of developments in text linguistics and discourse analysis (See Davison & Green 1988 for a recent reconsideration and critiques of readability).
A second source is research that links text types with their microlinguistic exponents. Works like Biber (1988) have catalogued the lexical and grammatical features of a large number of text types, and give us important clues as to what makes a text more complex. This kind of research has had important practical consequences for the Plain English movement, which has had successes in improving the readability of English texts (Charrow & Charrow 1979). It should be remembered, however, that the complexity of an English text for monolingual reading may not equate to its complexity in reading for translation. de Groot (1997), for example, has questioned whether reading, decoding, encoding and writing always occur in sequence in translation. It may be that translators read in rather a different way from monolingual readers (Danks & Griffin 1997) perhaps incorporating some aspect of pretranslation processing as they read.
In addition, the special nature of translation suggests that the source text is only one element of difficulty. In a paper recently submitted to Target, Campbell suggested that text difficulty is one of three variables to be considered in exploring translation difficulty, the others being translator competence and translation task type. The paper goes on to propose a cognitive model of translation difficulty which, in brief, holds that difficult items in a text are those that require more processing effort to resolve target language choices.
How universal are translation difficulties?
Our research on translations of English texts into Arabic, Spanish and Vietnamese finds that a considerable portion of text difficulty is common to translation into all three languages. This conclusion was reached on the basis of a study of the alternative translations provided by groups of translators translating the same texts into the three languages. Where subjects in one language provided different renditions of a chunk of text, that chunk was judged to be difficult on the basis of the processing effort needed to decide among the various choices available. If the chunk caused a similar level of difficulty in all three languages, then it was judged to be universally difficult. We concede that the term universal is ambitious here, but argue that since the three target languages are non-cognate, we have a strong case. The study, which focussed on lexis and grammar, identified several areas of difficulty:
• Words low in propositional content, ie. stretches of text where propositional content is distributed over several words, one of which - characteristically a verb - is apparently low in meaningfulness, for example in become free from all opioid use, which can be paraphrased as stop using opioids altogether, where become merely relates a previous state to a present state. Looked at from a discourse perspective, this type of difficulty is an example of grammatical metaphor, where a marked or longer route is used to achieve the same propositional content; according to Halliday, grammatical metaphor is typical of the register used in written language and makes text "inherently complex"(Halliday 1985:329). Indeed, some of the Spanish subjects translated such items into the congruent form (to use Halliday's term): to have schools and workplaces free from discrimination became conseguir erradicar la discriminaciر® ...de las escuelas (to eradicate discrimination from the schools) or librar de (free schools from).
• Complex noun phrases, ie. noun phrases where two or more nouns are concatenated, and there is difficulty in interpreting or encoding the relationships between the head and the other items in the string. An example is Methadone treatment, where methadone is in an instrument relationship with treatment.. Among the Arabic subjects in the study, two interpreted the relationship as patient; their renditions could be glossed as the treatment of Methadone. Among the Spanish subjects, there was an initial suspicion that some had also made a patient interpretation because of renditions like Tratamiento de Metadona rather than Los tratamientos con metadona. However, a close analysis of the coding of instruments indicated that there is a continuum of explicitness in the choice of preposition. The typical range includes:
inexplicit de
explicit con, por
highly explicit con m todo de, por medio de, por administraciَn de.
Thus, for the Spanish subjects, the focus of difficulty was on encoding the instrument relationship and deciding on the degree of explicitness. For the Arabic subjects, there was a degree of similar difficulty, but the choices were fewer, ie. between the simple preposition bi- "with" and the elaborated preposition biwasit`a "by means of"; the focus of difficulty was perhaps on interpreting the relationship rather than encoding it.
• Abstractness: Words like practice, contact and action were more difficult to translate than workplace, Department, school, and person, a finding in line with experimental research on monolingual and bilingual lexical retrieval. For example, Schwanenflugel, Harnishfeger & Stowe (1988) discuss a dual-representation model, which assumes that concrete words are more easily understood because they have both a verbal and an imaginable representation; they compare this with a context availability model that proposes that abstract words are harder to understand because of the difficulty in retrieving contextual information to aid their comprehension. de Groot's (1993) work in bilingual lexical representation proposes that concrete words have a single mental representation, while the representation of abstract words is language-dependent.
• Official terms: The problem with items Anti-Discrimination Board and contact officer may be associated with the widely known difficulty of finding equivalents for official titles. This category is one associated with both comprehension and production, and is largely a consequence of the dense semantic packing of such terms.
• Passive verbs: These evidently posed difficult decisions for translators, probably more in the area of encoding than interpretation. Both Arabic and Spanish have quite different grammatical strategies to deal with English passives; examples are given in a later section of this paper.
Work to identify more of these "universal" difficulties, as well as language-specific difficulties, is proceeding with the use of a database of translation examination scripts based on of various English source text types.
Comprehension versus production difficulties
One of our interests is to find out if translators working into different languages confront similar levels of difficulty. This challenge is expressed in practical terms by the question "How reliable is a translation test that uses the same English source text for different target languages?" Questions of language typology arise here. Hill (1997) asked the mirror image question - whether it was possible to construct translation tests in different source languages of similar difficulty, and claimed that French was easier to translate into English than Russian into English. But the justification for such claims needs to be considered with caution; how confidently can we claim that French is somehow structurally closer to English? And at any rate, our study found that there were English source text difficulties common to three unrelated languages - Arabic, Spanish and Vietnamese. This necessarily leads us to a consideration of comprehension and production, a matter we hinted at in the discussion of Methadone treatment. . What we suspect to be the case is that there are two loci of difficulty. The first locus is comprehension, and difficulty is likely to be fairly universal. The second locus is production, and here there are likely to be different levels of difficulty depending on the lexis and grammar of the target language. The example of Methadone treatment above began to tease out these separate loci of difficulty: The Arabic subjects may have found comprehension more difficult, and the Spanish subjects may have found production more difficult.
Is production difficulty measurable?
If, as we have claimed, there may be different levels of production difficulty depending on the target language, can the difference in difficulty somehow be measured, or at leats compared in a principled way? One of the tools we are developing to test this idea is what we call a choice network, in effect a psycholinguistic model of the options facing the translator, based on the evidence of the target texts of groups of subjects translating the same text. Examples of choice networks are given in Figures 1 and 2. Here we have analysed the translations of the chunk The patient will be closely monitored and drawn up the most parsimonious networks that will account for all the data pertaining to grammatical choices. Note that we have not included word choices in these networks. As can be seen, the networks are comparable in quantitative complexity, but quite different qualitatively. While both groups are faced with a three-way choice of sentence types, the Arabic subjects need to grapple with the option of expanding patient, and whether to use a future particle. The Spanish subjects' concerns are focussed on the adverbial, with choices of word class and position. We believe that models like these - based on empirical data- can also be used to explore lexical and pragmatic choices. Indeed they are used in Campbell (1998) to model stylistic choices in translators working from Arabic to English.
Figure 1: Choice network based on nine translations of The patient will be closely monitored into Arabic.
Sentence type
_________________|____________________
| | |
nominalisation passive active
| | |
______ |_____________ __|________ |
| | | | |
| | | | |
patient patient patient patient
expanded to not expanded expanded to not expanded
patient's | patient's |
condition | condition |
| adverbial= | |
| prepositional | __________|
source phrase | |
text | adverb
adverbial ____ |_____ omitted
shifted to | | |
verb future | |
| particle | |
| used | |
| | |
|_________________________ | |
| | |
no future particle
Figure 2: Choice network based on nine translations of The patient will be closely monitored into Spanish.
sentence type
___________________________|______________________ _
| | |
active true passive passive with se
| | |
| copula= |
| serل |
copula = | |
estarل ______|______ ______|______
| | | | |
| adverbial= adverbial= adverbial= adverbial=
| true adverb prepositional true adverb
prepositional
complement= | phrase | phrase
prepositional | | | |
phrase _______|__________ | _____ ___| |
| | | | | |
| | | | | ____________________|
| | | | | |
| adverbial adverbial
| precedes follows
| main verb main verb
|
__ |__________
| |
adjective adjective
prededes follows
noun noun
How practical are our aims?
Is it possible to create or select texts of a given level of difficulty? Our research so far suggests that the most profitable avenue is to weight items in a translation text. Some items would carry a weight of zero - the ones that we predict will cause no difficulty to any subject. A great deal of common vocabulary might be included here. Harder items would carry higher weights, possibly with different weights for different item types. These would include problematic words and grammatical structures. The total text difficulty would be based on summing the weights.
We are not suggesting that only the weighted items be scored when target texts are assessed. It is axiomatic that word class and rank shifts occur in translation, and we believe that holistic marking of a text is the only way of dealing with the redistribution of meaning that occurs when a message is recast in a different grammatical and lexical guise; it is after all a text that is being translated, and not a string of discrete words and phrases. The point of weighting texts would be to create materials to meet the needs of students and examinees at different levels of development or proficiency.
Future directions
Our plans are to undertake a larger empirical study of target texts in our database to find out more about what makes texts universally difficult to translators into different languages. This, we hope, will lead to a set of model graded texts along with guidelines that teachers and assessors can use to grade their own texts. At the same time, we plan to continue developing a theoretical underpinning to the work based on psycholinguistic modelling of the translation process. This work will be much informed by our findings about the production dimension in specific languages.
References
Biber, D. (1988) Variation across speech and writing. New York: Cambridge University Press.
Campbell, S. (in press) 'A cognitive approach to source text difficulty in translation', Target.
Campbell, S. (1998) Translation into the Second Language. New York: Longman.
Chall, J.S. (1958). Readability: an appraisal of research and application. Columbus: Ohio State University.
Charrow V. and R. Charrow (1979) 'Making legal language understandable: A psycholinguistic study of jury instructions', Columbia Law Review 79, 1306-.
Danks, J.H. and J. Griffin (1997) 'Reading and translation: A psycholinguistic perspective' in J.H. Danks, G.M. Shreve, S.B. Fountain and M.K. McBeath (eds) Cognitive processes in translation and interpreting . Thousand Oaks, California: Sage Publications, pp. 161-175.
Davison, A. and G.M. Green (eds) (1988) Linguistic complexity and text comprehension: Readability issues reconsidered. Hillsdale, NJ: Lawrence Erlbaum Associates.
de Groot, Annette M.B. (1997) 'The cognitive study of translation and interpretation: Three approaches' in J.H. Danks, G.M. Shreve, S.B. Fountain and M.K. McBeath (eds) Cognitive processes in translation and interpreting . Thousand Oaks, California: Sage Publications, pp. 25-56.
de Groot, Annette M.B. 1993. 'Word Production and the Bilingual Lexicon' in R. Schreuder and B. Weltens (eds) The Bilingual Lexicon. Amsterdam/Philadelphia: John Benjamins, pp. 27-51.
Halliday, M.A.K. (1985) Introduction to functional grammar. London: Edward Arnold.
Hill, M. (1997) Translation difficulty. Does it vary according to language? Australian Language Matters, April/May/June: 10-11.
Schwanenflugel, P.J. , K.K. Harnishfeger and R.W. Stowe (1988) 'Context availability and lexical decisions for abstract and concrete words'. Journal of Memory and Language 27: 499-520.
Note
This research was carried out as part of a project funded by a University of Western Sydney Macarthur internal research grant.
تعليق