Chapter 4 Methodological Framework and Experimental setup
4.4 Experimental Texts
To optimise the ecological validity of the data, the study used authentic texts shown in their entirety rather than presenting individual sentences one at a time. The four texts each consisted of approximately 150 words in English (the participants’ L2). This was the maximum length of text which could fit into the upper half of the Translog window without text scrolling. The individual sentences for each text were extracted from an English news magazine article22 and combined into an abbreviated version of the original article. No further modifications were made to the sentences. By using authentic texts, it was sought to avoid the pitfalls claimed to be made often in metaphor research according to Goatly. Goatly claims that these pitfalls include using only class of metaphor (either novel or conventional) and using constructed or recycled examples of metaphors presented without context (Goatly 1997: 5).
22
The experimental texts (see Appendix 1) were based on articles from the following sources: Text 1 from The Economist, Text 2 from Time Magazine, Text 3 from Cnn.com, and Text 4 from The Economist. All four articles were published and accessed in 2009.
107
According to the Flesch reading ease index scores, the difficulty level of the four texts was comparable as can be seen from Table 4. The index score range is from 0-100; the higher the score, the easier the text (www.editcentral.com). Although the Flesch reading ease index was developed for English native readers, the professional background of the participants was deemed sufficiently similar to make it possible to apply the scores to the reading of texts for translation by L2 readers. As can be seen from the table, the measures are relatively crude and may not reflect the cognitive effort required for comprehension and translation, as the measures do not take such factors into account as coherence, topic, and culture specificity. However, the measures provided by Editcentral give an idea of the objective comparability of the texts at least. It should be noted that Editcentral’s definition of complex words is a question of number of syllables rather than e.g., lexical frequency, which was instead controlled for in the statistical model.
Table 4 provides an overview of the objective characteristics of the four texts as identified by Editcentral.23
Table 4: Properties of texts 1-4
Text 1 2 3 4 Flesch reading ease score 68.6 79.9 74 68.7 Complex words 15 10 21 16 Words 154 149 147 147 Sentences 9 10 9 7
As all four texts were abridged versions of longer texts, the curtailed versions were subjectively evaluated by a third party to ensure that they read as coherent text units. This evaluation was done by asking the evaluator to read the text and comment upon any abnormalities in the text and evaluate whether or not the text made sense and could be read fluently. The individual sentences were not altered, but several intermediate sentences were eliminated by me to keep the
23
108
text length at the maximum of approximately 150 words. Attention was given to ensuring that the contexts of the metaphors were clear from the onset of the text. The subject varied from text to text although all four texts were on financial topics as financial texts often contain many metaphors.
The texts contained 37 linguistic metaphors in total, constituting 37 AOIs, ranging from six metaphors in Text 2 to 13 metaphors in Text 4. For the purposes of this experiment, the
boundaries of each individual metaphor were identified using the criterion that the constituents of the metaphorical expressions, i.e. the vehicle, must differ from the text in terms of semantic domain.
To illustrate what is meant by a different semantic domain, an example from one of the texts may be helpful: But the suitors must first gain the approval of regulators, who are sure to
supervise the courtship with care because of the size of the dowry (Text 3, see Appendix 1). The
subject of the text was the current high number of mergers between companies, and the overriding metaphorical semantic domain for the merger is marriage (and the parties and
processes involved in a marriage). Suitors is used as a metaphor for the two companies involved in a specific merger, and as suitors differs in terms of semantic domain from the text’s general semantic domain of business and mergers, suitors is identified as the linguistic metaphor in its entirety, and consequently constitutes the AOI to be analysed. The same applies to the other two metaphors in this sentence, namely courtship and dowry. This criterion is similar to Goatlys’ vehicle (1997) in that the constituents of the individual AOI are deemed to be the core of a metaphor carrying the metaphorical meaning. Furthermore, as Janus and Bever (1985) emphasised, it is preferable to investigate the processing time when the metaphor is first encountered rather than investigate whole sentences or end of sentences because of other integration processes which may interfere. They argued that “the greatest cognitive demands” will be at the point in the sentence at which the metaphor is first encountered (Janus and Bever 1985: 478). In other words, any differences in cognitive effort for processing literal expressions compared with metaphorical expressions will be more likely to be registered at constituent unit level than sentence level.
The majority of the metaphors were single words, but a few metaphors consisted of multiple words, and word length was therefore taken into account as a variable in the statistical analysis
109
(see section 4.6.1). For this study, ecological validity, i.e. using authentic texts rather than constructed texts, received higher priority than exact text comparability in terms of word length, AOI position, word frequency and other variables, which were instead included as variables in the statistical analysis.
For comparison purposes, 37 non-metaphorical AOIs were selected. Although Lakoff and Johnson (1980) claimed that everything is understood metaphorically, the selection criterion for the comparison AOIs was that the words were within the semantic domain of the text’s subject. In other words, the criterion was the opposite of the selection criterion for the metaphor AOIs. A few of the comparison AOIs such as “downturn” could perhaps be argued to be a conventional metaphor, but a lookup in a dictionary24 confirms that the word has lost its metaphoricity to the point of having a lexicalised definition. It is therefore doubtful that the word would be
interpreted metaphorically by the participants unless one takes the rather extreme views argued by Lakoff and Johnson.
The comparison AOIs were also chosen with a view to having a length comparable with the metaphor AOIs, and the single-word AOIs were all content words. Another selection criterion, applied to both types of AOIs, was the exclusion of start-of or end-of sentence AOIs, as research has indicated that gaze times may be longer for words at sentence-final positions caused by some form of wrap-up effect (Just and Carpenter 1980: 34) and it was assumed that sentence- initial positions could be influenced by some level of orientation taking place before the actual comprehension of the linguistic unit began. As the texts were authentic texts rather than
constructed texts, it was not possible to have comparison AOIs which were directly comparable with the metaphor AOIs in terms of word length and frequency, which were instead taken into account as variables in the statistical analytical model (see section 4.6.1).
The metaphors were also rated for their familiarity by a group of four evaluators, which consisted of three professional translators and one highly experienced interpreter. As it would not be feasible to have the participants in the experiment also evaluating the metaphors in the texts, it was necessary to have another group of evaluators not participating in the experiment do
110
this. The professional profile of the evaluator group matched the professional profile of the participants as a group, which was important in order to be able to reasonably draw inferences from the familiarity ratings of the evaluator group to the assumed metaphor familiarity of the participant group. Sentences containing metaphors were extracted from all four texts and
presented to the raters in randomised order. The metaphor vehicle, i.e., the AOI, was highlighted in red. If a sentence contained more than one metaphor, this sentence would be listed once for each individual metaphor and could therefore be listed up to three times. If a sentence contained more than one metaphor and was therefore listed more than once, these instances of the sentence would be grouped together to make the evaluator aware of the repetition. If the evaluator had encountered the same sentence again but after several other sentences, it is possible that s/he would have rated the metaphor as highly familiar because she had encountered the same sentence before in the evaluation. By grouping the instances of the sentences together and thereby making the evaluators aware of the repetition, any potential influence on the familiarity rating of each individual metaphor in the sentence was deemed minimal.
The evaluators were briefed that the metaphors were taken out of context and were asked to evaluate how familiar they perceived the individual metaphor to be on a scale from 1 to 7, 1 being not at all familiar and 7 being very familiar. The order of the sentences was different for each evaluator. An average was calculated for the ratings from all four evaluators (see Appendix 6) and included as a variable in the statistical model used for the inter-metaphor analyses 1.1 and 4.1.
The output of the reading and retyping task (see subsequent section) was not evaluated in any form as the product is identical to the ST with perhaps a few spelling mistakes. The output of the translation task, i.e., the TT, was not evaluated in terms of translation quality as this is very time-consuming and would merit its own study. Also, the focus of this thesis was not an investigation of metaphor translation quality or translation style. The produced TTs were analysed to determine the choice of translation options chosen for the metaphors. These translation choices were categorised according to Dobrzynska’s classification scheme (1995): use of an exact equivalent of the original metaphor (M-M), choice of another metaphorical phrase with the same meaning as the ST metaphor (M1-M2) or paraphrase (M-P) (Dobrzynska 1995: 595). This classification was done subjectively, but the benefit of using Dobrzynska’s classification scheme with only three categories, which are consequently quite comprehensive,
111
rather than a more elaborate classification scheme such as Newmark’s (1981) was that there were only a few instances of translation choices which could be classified in either one or the other category. Dobrzynska’s classification scheme is categorical and enables the experimenter to be more consistent in categorising the translation choices than with a more finely granulated classification scheme with more vague and overlapping definitions. Examples of translations from the experiment can be seen categorised in Table 5 to give an idea of the translations in each category.
112 Table 5: Examples of translations categorised by translation strategy
Translation Strategy M-M Translation Strategy M1-M2 Translation Strategy M-P Source text (AOI in italic) Securitization is radioactive
The market for securitization has
fallen off a cliff
The on-again, off-again plans of Porsche and Volkswagen to tie the
knot ...
Target text (AOI in italic) Sekuritiseringer er radioactive [securitization is radioactive] Markedet for sekuritisering er
kollapset [the market
for securitization has
collapsed]
Porche og
Volkswagens til tider vaklende planer om en
sammenlægning
[Porche and Volkswagen’s
sometime shaky plans of a merger]