There seems no fixed definition of collocation in the literature. Nation (2001) viewed collocation as a term covers most multiword units, such as idioms and fixed expressions. Some other researchers defined collocations as a type of multiword expressions. For example, Howarth (1996) and Nesselhauf (2005) viewed collocations as a special type of recurring lexical combinations. Their definition of collocations were grounded on Cowie’s (1981, 1998a, 1998b) phraseological approach of view (Barfield & Gyllstad, 2009). Cowie’s (1981, 1998a, 1998b) divided word combinations into two broad categories, “composite” and “formulae.” The category of “composite” represents a continuum composed of four types of word combinations depending on the level of semantic transparency and the acceptable level of element substitution. Restricted collocations (e.g. perform a task) are characterized by possible but limited substitution of elements. In terms of meaning, the meaning of a restricted collocation is transparent and can be largely inferred from its components.
Laufer and Waldman (2011) also explained that collocations are characterized by a “restricted co-occurrence of elements” and “relative transparency of meaning.”
The former distinguish collocations (e.g. make decisions) from free combinations (e.g.
make sandwiches) where the elements can be replaced only if they follow the rule of grammar. The latter makes a distinction between collocations and idioms. A collocation’s meaning can be inferred from the elements that compose it while an idiom’s meaning can not be understood from their individual words.
Collocations are also distinct from lexical bundles or n-grams. While both terms refer to co-occurring combinations, collocations do not include all types
25
co-occurrences, they refer to “specific types of word combinations. ...generally syntactically regular in that they follow syntactic combinations (e.g. adjective + noun collocations)”(Crossley & Salsbury, 2011, p.3) Lexical bundles (e.g. kind of, I mean ), on the other hand, are determined by the frequency of co-occurrence and syntactic structure do not usually used to identify them.
The importance of collocations was highlighted by Altenberg (1993), “they tend to form the communicative core of utterances where the most important information is placed”( as cited in Nesselhauf, 2005, p.9). Wray (2000) noted that a good command of collocations can develop learners’ fluency and improve accuracy. For language learners, collocations are usually less salient in input than other forms of multiword expressions like idioms. This is so probably because they are usually semantically transparent (e.g., submit an application, quit a job, face a challenge), composed by frequent individual words and do not cause difficulty in comprehension. Hence, learners or teachers they may not notice them when they are encountered in input. On the other hand, learners may neglect the restrictions on acceptable substitutions of collocations when these restrictions are not pointed out or made salient to the learners.
Many studies have documented that learners tend to produce deviant collocations.
2.4 Learners’ insufficient knowledge and misuse of collocations
Empirical evidence is available that shows that EFL learners possess insufficient knowledge of collocations. They had a difficulty in producing acceptable collocations and often produce deviant collocations. Research also suggests that verb-noun collocations are especially problematic for EFL learners. (Bahns and Eldaw, 1993; Liu, 2002; Nesselhauf, 2005; Chen 2002; Lin 2010; Laufer & Waldman, 2011).
As early as two decays ago, Bahns and Eldaw (1993) conducted a study in responding to claims that says explicit instruction collocations is not needed since
26
learners will pick up collocations along with the acquisition of vocabulary. He recruited 58 advanced EFL learners, who had been learning for English for seven to nine years. A total of 15 verb-noun collocations were used to understand the participants’ collocation knowledge. Thirty-two participants were asked to do translation task and the other 24 participants did a cloze test. The results were graded by three native speakers. The researcher counted the total number of correctly and incorrectly translated lexical words produced for the translation task. He found that of the 2662 words, 2214 were correctly translated. Among the 448 incorrectly translated lexical words, almost half was found to be verbs in collocations. In terms of the translation of the target collocation items, half of the total number of items was not acceptably translated because of wrong collocates or semantically unacceptable paraphrases. He calculated the numbers of correctly chose items on the cloze test and showed that half of the items were incorrectly answered. The study concludes that the advanced EFL learners had an insufficient knowledge of collocation and it is not comparable to their vocabulary knowledge, suggesting that collocations may not be necessary acquired along with the vocabulary.
A further analysis of the paraphrase of collocations attempted by the learners who took the translation task showed that some of the collocation items were not correctly translated even when the participants could avoid using the verb-noun combination by paraphrasing the meaning, indicating the necessity of collocation knowledge. Items such as do damage, keep diary, reject proposal, achieve perfection seem less amenable to paraphrasing than other items.
Nesselhauf’s (2005) investigated around 2,000 verb-noun collocations produced by advanced German EFL learners in their essays. The collocations were manually extracted from the data of the German subcorpus of International Corpus of Lerner English (ICLE). Acceptability of learners’ collocations was determined by collocation
27
dictionaries, BNC corpus and native speakers’ judgments. The result showed that about one third of all the verb-noun collocations examined were unacceptable or questionable. What’s more these miscollocations were produced by 183 out of 207 learners and each did not produce more than 16 miscollocations. The high proportion of miscollocations can not be only a result of a few learners’ large amount of miscollocations. The study confirmed that verb-noun collocations cause problems for advanced learners.
Her study also analyzed the components which caused verb-noun miscollocations; namely, the verb, noun, determiner, structure or the whole collocation. The results showed a clear pattern that verb is the most frequent deviant elements. The researcher thought the result can be explained by that verbs are generally one of the most difficult elements to master for L2 learners and the verbs used in the verb-noun collocations (e.g. take a step but not do a step) are arbitrary restricted. The German EFL learners particularly misuse certain verbs. Nesselhauf’s (2005) summarized the simple verbs confused by at least two learners in her study.
The table is duplicated in Table 2. The misuse of light verbs were especially noticeable
Table 2. The simple verbs confused by at least two learners in Nesselhauf (2005)
verb used acquire consume disturb do do find follow appropriate verb develop take disrupt take perform get pursue
verb used gain get give give have make make
appropriate verb have have make provide gain have take verb used mount raise reach reject solve take undertake appropriate verb get increase achieve oppose end make take
28
According to Nesselhauf’s (2005), learners’ misuse of these high-frequency light verbs was well documented in the literature. Not only German-speaking learners, speakers of Swedish, Spanish, Polish, French and Chinese were observed to confuse these verb pairs as well.
The study found 18 verbs that were most frequently inappropriately used by at least three learners, including make, give, take, have, get, disturb, reach, do, destroy, gain, solve, find, consume, establish, acquire, follow, tell, and kill.
Nesselhauf’s (2005) further examined sources of producing inappropriate collocations, namely, elements from L2, learners’ L1 and factors other than these two.
For elements from L2, the study noted that semantically similar verbs or near synonyms may cause confusion, resulting inappropriate collocations. Examples were destroy –ruin, give-provide, tell-say, do-perform, raise-increase. A formal link between deviant verbs and appropriate verbs was also noted. The study identified cases of phonological link, such as take for make (e.g. take changes) and grow for go (go out of fashion) and morphological links, for example, solve for resolve and undertake for take. Nevertheless, the study found that formal links were more frequent observed in the cases of deviant nouns than deviant verbs. Another element of L2 that were displayed in learners’ miscollocations was blends of related L2 material. Learners seem take parts of different expressions and recombine them to create collocations.
The study highlighted a considerable influence of L1 on the learners’ production of collocations. It was found that among the 748 deviant collocations, 379 cases were possibly induced by learners’ L1. The exhibition of L1 influence on producing verb-noun combination was especially strong in verb choices.
Based on her study, Nesselhauf’s (2005) believed that the number of years that learners study English do not necessary enhance learners’ collocation use and the
29
length of exposure only has a light positive impact. Thus, she proposed that the most essential purpose of collocation teaching is to make learners aware of this language phenomenon. Teachers were recommended to use consciousness-raising activities.
Added to that, is explicit teaching of collocations. She thought explicit instruction was necessary in improving learners’ use of collocations.
Some principles for teaching collocations were suggested.
1. Be systematic. Teaching should focus on collocations or elements that learners often confused. Contrast appears useful. Teachers should contrast collocations that are similar in form or meaning or both (e.g catch a glance-catch a glimpse), stretch verb constructions and simple verbs (e.g. have a look-look ) and verbs or nouns that are often confused (e.g. meet-fulfill-satisfy). Another strategy is to teach various collocations with one certain verb simultaneously. For example, when teaching reach a conclusion, collocations using reach in a similar sense should also be taught (e.g.
reach a decision, an agreement, a goal).
2. A mere focus on form is not sufficient. Teacher should also present the exact meaning and the usage of collocations. Exercises based on concordance lines were recommended.
3. Contrasting L2 collocations with L1 collocations. This principle is especially important when the non-congruence is likely to cause miscollocations and when common L1 collocations do not have English correspondents. The researcher explained the importance of this principle with the following example. Teaching only have an experience or achieve an aim will not prevent German EFL learners from producing make an experience or reach an aim, “unless it is explicitly pointed out to them that the equivalents to the German expression can not be used on English (p.271).”
Laufer and Waldman (2011) also focused on verb-noun collocations. They
30
compiled a learner corpus consisting of free writing products of advanced, intermediate and beginner Israeli learners of English (the corpus size was around 300,000 words). A comparison of verb-noun collocations used in the learner corpus and a native corpus (LOCNESS) was made. The researchers extracted 220 most frequent nouns and their verb-noun collocations. Learners’ incorrect collocations were identified by consulting native speaker judgments, dictionaries of collocation and BNC corpus.
The results revealed that all three levels of learners produced much fewer verb-noun collocations and almost one third of the investigated collocations the learner attempted to generate were problematic. The study also observed that a significant growth in the occurrence of verb-noun collocations appears only in advanced learners’ subcorpus, though still less frequent than native speakers.
Wang & Shaw (2008) investigated the verb-noun collocations used in 200 English essays written by Chinese and Swedish learners of English. The researchers selected the most 20 frequent verbs and manually extract noun collocates of the verbs by analyzing concordance lines. Among the frequent verbs, have, do, make, and take were chosen to be further analyzed. The findings indicate that the most frequent error type in the attempt to produce verb-noun collocation was wrong choice of verbs. They further point out that the use of do and make (e.g. make damage / do a great effort) were especially problematic for both groups of learners.
The findings suggested that high-frequency verbs are repeatedly used incorrectly by advanced learners. Therefore, these verbs should not be neglected in instruction and especially highlighted in collocation teaching.
The suggestion seems to be further supported by Altenberg and Granger’s (2001) study examining French and Swedish advanced learners’ use the high-frequency verb make, though not specifically concerning collocation use. The learner corpus was the
31
subcorpora of International Corpus of Learner Englsih (ICLE). They found that the most frequent type of use of make was its delexical uses. Compared with a native corpus (LOCNESS), it was observed that while native speakers tend to use make to collocate with nouns expressing speech actions (e.g. comment, claim, remark, statement), learners tend to underuse this type of usage. What’s more, learners often misused the verb make as a delexical verb, resulting collocation errors like make a balance, make a poll, make research.
Liu (2002) investigated Taiwan EFL learners’ use of verb-noun miscollocations in their compositions which is extracted from a tagged corpus, English Twain Learner Corpus (ETLC). The researcher used the tags such as “word choice,” “problematic usage” and “wrong verb/word” to search for miscollocations. They study found that among the 265 miscollocations identified, 233 were verb-noun miscollocations.
What’s more, verb errors occurred in these verb-noun miscollocations were more common than noun errors.
The study further analyzed learners’ verb-noun miscollocations by identifying the semantic relationship between the wrong verbs used by learners and the corresponding appropriate verbs which should be used to express their intended meaning. He categorized the relationships into synonym (e.g. acquire/ gain), troponym and hypernym (e.g. create/ compose). The results showed that one third of the errors can be caused by misuse of synonyms. But, the study also found that almost a half of the wrong choice of verbs can not be classified as caused by synonym, troponym, or hypernym and can probably be better explained by L1 influences or word-for-word translation.
Focusing on Taiwanese and Chinese EFL learners’ error in verb-noun collocations, Lin (2010) built two learner corpora and collected the learners’ possible miscollocations that occurred at least three times in the same corpus via
32
semi-automatic extraction procedure. Manual check was done to further determine real miscollocations. The study found 210 different verb-noun miscollocations (1041 tokens) produced by Taiwanese learners and 268 different verb-noun miscollocations (2707 tokens) by Chinese learners. The study generated a complete list of these frequent miscollocations. The researcher reported that a few collocations were identical to those found in other studies examining Chinese learners’ miscollocations, namely, get knowledge, learn knowledge, study knowledge, teach knowledge etc.
The study also categorized the errors into ten types. It was found that around half of the miscollocations produced by both two groups of learners involved wrong choices of verbs and around ten percent of the miscollocations involved misuses of de-lexical verbs. The study again showed that the verb elements pose great difficulties for EFL learners to produce appropriate verb-non collocations. Possible causes of errors were also explored in Lin’s (2010) study. She classified the errors as induced by 11 types of possibilities. The results showed that around 50 percent of errors produced by both groups of learners could be induced by L1 transfer.
Above studies provide evidence showing that learners’ knowledge in collocations is insufficient. What’s more, due to limited collocation knowledge, learners tend to use deviant combination, resulting in awkward expressions. These studies also highlight the need of explicit collocation instruction. In the following section some empirical investigations on teaching collocations are examined.
2.5 Empirical studies on collocation instructions
Some empirical findings suggest that explicit collocation instruction is effective in promoting EFL learners' collocation knowledge. Among the explicit teaching methods under investigation, the use of web concordancer in classroom instructions received much attention (Sun & Wang, 2003; Chan & Liu, 2005). Some studies examined the effect of different tasks or learning conditions on collocation learning
33
(Webb and Kagimoto, 2009, Webb and Kagimoto, 2011; Sonbul and Schmitt, 2013).
There are also some attempts to investigate how the use of collocation dictionaries can benefit collocation learning (e.g. Laufer, 2011).
Sun and Wang (2003) used a concordancer-based teaching, aiming to determine the effectiveness of inductive and deductive approaches to learning four collocation patterns, namely, distinguish A from B/ distinguish between A and B; in excess of;
indignant with/ indignant at; the gulf between A and B. The participants were introduced to three web-based concordancing tools. The inductive group was asked to search for instances of using the target collocations, induce the underlying patterns, and use their findings to correct the sentences. The deductive group was given rules and example sentences and asked to correct the sentences.
Error correction was used to test the learners’ knowledge. The results seem to favor the inductive group. The researchers attributed the benefits to the discovering process involved in the searching and analyzing of concordance lines. Although the notion of easiness and difficulty was arbitrary, they further found that easy collocation patterns tend to be more effectively learned by using inductive approach than by deductive approach. There was no difference in different approaches in the case of difficult collocation patterns.
Recognizing that the literature generally points to learners’ sufficient knowledge of verb-noun collocations and that using appropriate verb-noun collocations is particularly difficult for learners, Chan and Liou (2005) especially focused on the learning of verb-noun collocations via web-based practices and concordancing. The researchers designed five web-based learning units, covering four common error types involving synonymous verbs, hypernymy and troponymy verbs, delexical verbs and English verb-noun collocations which do not have L1 equivalents. Concordancing was incorporated only in the first three units dealing with collocations involving
34
delexical verbs and English verb-noun collocations which do not have L1 equivalents.
The exercises used in the online units consisted of multiple-choice, sentence translation and gap-filling. The measurement test includes a total of 26 items selected from the 100 items taught in the five units and extra 10 items not taught in these units.
They found that the instruction significantly improved the learners’ collocation knowledge. The learner obtained around 10 out of 36 in pretests and around 19 in post tests. What’s more, a comparison between the pretest mean scores and delayed post-test mean scores indicated that the learning effects remain. They also analyzed the learning gains and retentions for the four types of collocations separately. It seems that collocations involving delexical verbs and English verb-noun collocations which do not have L1 equivalents especially amenable to instruction. It is not clear, however, if the benefits were related to the types of collocations or the type of learning activity (i. e. concordancing) involved. What’s more, the design of the tests do not capture a whole picture about how many of the item taught were actually learned by the learners.
Webb and Kagimoto (2009) used a reading task and a cloze task and compared the effectiveness of these two methods. In the reading task, 117 university students formed two experimental groups. One group was given a list of example sentences, with the target collocations glossed. Three example sentences were provided for each target collocation. The other group received cloze task where the same sets of sentences were used but the target collocations were replaced by blanks. The participants were given two sets of example sentences and two collocations. Their task was to fill in the blanks with appropriate collocations. The study used 24 target collocations.
To measure the effectiveness of the treatment, they used four tests to elicit the participants’ recognition of form, recognition of meaning, production of form and
35
production of meaning. The results suggested that both experimental groups outperformed the control group, showing effectiveness of explicit instruction on collocations. Both the reading group and the cloze group improved significantly after treatment on both receptive and productive tests, but the improvement in the two
production of meaning. The results suggested that both experimental groups outperformed the control group, showing effectiveness of explicit instruction on collocations. Both the reading group and the cloze group improved significantly after treatment on both receptive and productive tests, but the improvement in the two