Should We Teach Semantic Prosody Awareness?

Abstract

While considerable attention has been paid to collocation, and the development of the collocational competence of L2 learners in recent years, very little has been said about a related concept in teaching journals, namely semantic prosody, and L2 learner awareness of this phenomenon. In this paper the concept of semantic prosody is introduced, and then discussed, before an experiment is reported investigating the semantic prosody awareness of three different sets of participants (university undergraduate Arab students, Arab English teachers, and native speaker (NS) English teachers) for seven lexical items for which specific semantic prosody claims have been made by corpus linguists. The results show that the data from the NS Group (when considered as a group) concurred with corpus observed semantic prosody tendencies for all seven items, but for the other two sets of participants the results were mixed regarding agreement. It would appear that for some items non-native users of English of different language levels appreciated the typical semantic prosody. For other items this was clearly not the case, and it appears that semantic prosody awareness can be incremental, or may not actually develop. Given the mixed data, it is suggested that explicit instruction about semantic prosody in the classroom may be of some value to learners.

Keywords

Semantic prosody corpus data elicited data collocation pragmatics advanced learners lexical knowledge irony idiomaticity vocabulary acquisition

Introduction

In his seminal paper of 1976, Richards noted that lexical competence was multi-faceted and his ‘rich concept’ (1976: 88) approach to word knowledge listed eight assumptions to be considered in this area. The second part of Richards’ second assumption was: ‘For many words we also know the sort of words most likely to be found associated with the word’ (1976: 83). Today, such knowledge is typically called ‘collocation knowledge’, a type of knowledge which has, in the last two decades or so, received a considerable amount of attention – from corpus linguists, SLA researchers, and materials writers. Indeed, it would be fair to say that the development of L2 learner collocational competence is now a standard objective and learning goal in classrooms around the world.

Defining Semantic Prosody

However, an aspect of word knowledge which has received far less attention in the classroom is the phenomenon of semantic prosody. Richards himself did not use the expression semantic prosody, and it is unclear whether he had the concept in mind in writing his second assumption. There has been considerable debate concerning what semantic prosody entails – and this is not the place to review and critique all the discussion surrounding it (see Hunston, 2007; Morley and Partington, 2009; Partington, 2004; Stewart 2010 for recent insightful comment). At the risk of over-simplification, and for the purposes of this paper, I shall use the term to refer to a word or phrase’s discourse function in conveying attitude or evaluation – a function which the word (or phrase) in isolation is not typically recognized as playing.¹

Exemplifying Semantic Prosody

To contextualize the above it is worth considering an example. The adverb utterly is defined in the Longman Dictionary of Contemporary English (hereafter LDOCE) as ‘completely or totally’, and the example sentence provided is: ‘You look utterly miserable’. The word utterly is not defined ‘negatively’, and yet the example sentence, along with corpus data, indicate that it is a word which is very strongly connected with negative collocates. Louw (1993: 160), in his investigation of the word provided corpus examples of: burned, confused, demolished, and meaningless, among others. This seemingly ‘neutral’ word (at least according to LDOCE) is described by Louw as having ‘an overwhelmingly “bad” prosody’ (1993: 160). But how evaluative is utterly in the above, given the comment on evaluation in our definition (i.e. ‘…conveying attitude or evaluation…’)? One might argue that it is not very evaluative as the adjectives from the corpus data noted above for this word are not ones which are particularly desirable (under any circumstances). However, even here, the choice of the collocation utterly confused, as opposed to wonderfully confused (for example) hints at evaluation – the confusion is not desirable – and it is complete (cf. slightly confused).

On occasion, the evaluative function of certain words comes to the fore, and an additional example brings this into focus. Louw (2008) notes that advocates of euthanasia choose to talk about euthanasia bringing about death, rather than causing death. Corpus linguist research suggests that cause tends to be used with negative consequences (Stubbs, 1995a), whereas bring about is more often used with positive results (Xiao and McEnery, 2006: 115-20). Since death is ‘positive’, in the context of pain and suffering, (in the eyes of euthanasiasts), it is no accident that the verb bring about is used rather than cause in such a context – indeed such usage is very deliberate, and at the same time rather subtle – bring about seems to prime the reader to expect a positive result, hence the evaluative force. It is also possible for a language user to ‘break’ the typical semantic prosodic patterning of a word in a less subtle and more dramatic way to convey irony, and such usage is immediately evident to the linguistically proficient listener or reader, because of the departure from the typical usage, Louw’s (1993:164) example of ‘…bent on self improvement’, being oft cited in the literature (Biber, 2011; Partington, 2004, 2011; Zhang, 2009). It is important to note here, however, that non-typical prosodic usage need not, indeed should not, necessarily be viewed as ironic – such use may actually be highly conventional and idiomatic and this fact has, occasionally, been overlooked. For example, Sinclair, commenting on corpus data for the phrasal verb break out, stated that ‘it is bad things that break out, not good ones’ (Sinclair, 1990: xi). Although this is, generally, true, it is not absolutely so: laughter, for example also breaks out, and such a collocation is highly idiomatic. Indeed such exceptions, refusing to conform to generally observed semantic prosody patterns, pose something of a problem to the theorizing of semantic prosody – an issue outside of the focus of this paper.

In the last 30 years numerous words, expressions, or discourse stretches containing certain words/expressions have been classed as showing certain semantic prosodies – typically negative, though occasionally positive. As noted above, it is very rare for a word to be used in an exclusively polar sense, i.e. only used to convey a positive, or only used to convey a negative evaluative stance: for example bring about is used to speak of negative consequences, as any corpus search will verify, though it is more typically used to convey positive results from the viewpoint of the speaker /writer. It is the observation of such typicality, and the recording of statistical tendencies from corpus data, which have, in many ways, made the study of semantic prosody possible, and it is no accident that the term is most frequently encountered in corpus linguistics journals.²

The Hidden Nature of Semantic Prosody

Very little has been said about the hidden nature of semantic prosody, alluded to above in our definition: the idea that certain words/phrases’ evaluative roles are typically not appreciated in isolation. Louw’s (1993) claim that corpus linguists uncovered semantic prosody (for the first time), is generally understood to be exaggerated (see Partington, 2004: 155). However, dictionaries, whether corpus-based or not, have not always noted important semantic prosody information related to words. As Channell (2000) points out, the Oxford Advanced Learners’ Dictionary (OALD, 1995) defines right-on and par for the course without giving them a negative evaluation. Similar (negative prosody) dictionary omission observations have been made by Stubbs, for cause (1995a: 27) and cronies (2002: 72), and by Sinclair, for budge (1998: 16). The importance of recent corpus-based studies in enabling a better understanding of semantic prosody should, therefore, be acknowledged.

So, what is meant by the ‘hidden-ness’ part of our definition? Perhaps the best way to appreciate this is to compare semantic prosody with individual word connotation. As an example, skinny denotes thin, but it connotes thin negatively, whereas slim, again denoting thin, connotes thin positively. Skinny and slim, therefore, do not appear to be good candidates for words ‘conveying’ semantic prosody – as the connotation is inherent to the word when considered in isolation. A phrasal verb such as bring about though, is different: there seems to be no ‘inherent’ goodness in the phrase. It is only when we observe its typical collocates that we can observe that the phrase is typically (though not exclusively) used to describe causation when something desirable results from the point of view of the user of the word.

Although it is arguably a simplification to consider semantic prosody as collocational connotation (overly simplistic as some negative collocates do not necessarily convey negative evaluation e.g. alleviate suffering, where suffering is ‘negative’, but the overall collocation positive), semantic prosody is more akin to collocational connotation rather than individual word connotation. However, even this explanation, lacking as it does an overall discourse role, is in some ways inadequate (see Hunston, 2007; Morley and Partington, 2009; Stewart, 2010). Some writers (e.g. Channell, 2000: 51-53) have argued that a word itself may begin to be so ‘tainted’ by its typically negative /positive collocates, that it begins to connote in isolation – and there may well be a cline in this regard from words like commit (which almost seem to include the idea of not just doing something, but doing something bad) to seemingly neutral words like bring about. Allowing for a cline in this regard (from ‘less-hidden’ to ‘more-hidden’) seems to be the only way of maintaining a general ‘hidden-ness’ requirement in the definition, and this is important as Stewart (2010) is of the opinion that the whole concept of semantic prosody is called into question, if the hidden-ness element is not maintained.

The issue of hidden-ness is of particular importance to us in this paper. If a word has a particular connotation this can be noted by the language teacher: such knowledge is open to a teacher’s introspection and falls within his or her explanatory capabilities. Connotation is also recorded (clearly) in a dictionary – a student can note that slim is a more positive evaluation of someone’s figure (LDOCE – ‘someone who is slim is attractively thin’) as opposed to skinny (LDOCE – ‘very thin, especially in a way that is unattractive’), for example. As hinted at above, if words which can convey semantic prosody are not so open to a teacher’s introspection, and, as noted earlier, if dictionaries fail to make mention of semantic prosody tendencies associated with an item’s use, then it may be that L2 learners could become familiar with a word or phrase (e.g. cause, bring about, break out, utterly), and not be aware of the typical evaluative usage patterns – i.e. face a deficit in Richards’ second vocabulary knowledge assumption noted earlier – knowing the typical context within which the word occurs, and its pragmatic function. Having said this, one would expect the examples provided by teachers in the classroom to indicate the evaluative stance (and hence the semantic prosody implicitly), as would examples in a dictionary. Whether such exposure would suffice to alert a student to this issue remains to be seen.

The Language Learner and Semantic Prosody

If language learners are not aware of the typical semantic prosodies associated with the usage of certain words or phrases one might hypothesize over-extension in usage. For example, an L2 learner might treat break out as roughly synonymous to start (i.e. without the required awareness of the associated negative colouring, or its typically negative collocates). Such a usage would work with ‘the war broke out in …’ and indeed possibly attract the teacher’s commendation for depth of vocabulary knowledge; however, this commendation would quickly be negated, if the same writer went on to say that ‘Peace, having broken out, lasted for another 15 years’.³

But how likely is this to happen? Even if a learner’s attention was not specifically drawn to the (typically) negative set of grammatical subjects occurring with break out (for example), how likely is it for a learner to know such a word well enough to try to use it, without knowing the associated colouring – or evaluative meaning with which it is typically used – even if the learner’s attention has not been drawn to this fact (explicitly) in an instructional context? Such a case must be sharply differentiated from knowing the typical collocates of a word. This is, without doubt, far more complex, as collocability acceptability is highly arbitrary, and even semantic preference collocation sets are not entirely predictable. For example, some crimes are committed (e.g. murder) but not all crimes (e.g. theft, break in), and some actions not considered crimes in most countries are also committed (e.g. adultery). This seemingly arbitrary characteristic of collocability justifies direct teaching involvement – as is the case in classrooms around the world today. Semantic prosody, on the other hand, appears to be more ‘basic’ – simply ‘good’ or ‘bad’ (though note the comments previously about exclusivity) and so, one might assume, the L2 learner may pick up semantic prosody ‘overtones’, without much need for explicit focus in the classroom. But is such an assumption – that L2 language users ‘subconsciously’ pick up the general evaluative colouring or collocational connotation associated with certain words’ usage – a safe one? Stated more simply, is it worthwhile to consider explicitly teaching semantic prosody/semantic prosody awareness, or will it ‘look after itself’?

Research Goal

The research goal of this paper is to investigate to what extent non-native speakers of English use words with their associated semantic prosody in comparison to native speakers in a controlled elicited data experiment. Recognizing that semantic prosody appreciation may be incremental in nature, two non-native speaker groups – of two different language levels – were tested, along with a group of native speaker teachers.

Methodology

Seven lexical items for which corpus linguists have assigned specific semantic prosodies (be they positive or negative) were chosen for inclusion in the task. As the words were to be given to two different language level groups (in addition to native speakers), it was felt preferable to avoid less common words and phrases as stimuli (e.g. utterly, set in, sit through, break out – all of which have been discussed in the literature). The final list of words used in the task is given in Table 1.

Table 1.

Task Items¹

Lexical Item	Author	Prosody	Comments
Bring about	(Xiao and McEnery, 2006) and Louw (2008)	+
Cause	Stubbs (1995a)	–	Verb
Completely	Partington (2004)	–
Face	Tognini-Bonelli (2001: 95)	–	Only when used as a verb in an abstract sense
Potentially	Stubbs (2001:107)	–	Collocates listed are all adjectives
Provide	Stubbs (1995b: 247-48)	+
Regime	Channell (2000)	–

including reference to the claim, dominant prosody, and comments.

For each of the seven items the participants were asked to write up to a maximum of three dictionary-style example sentences (if they were able to do so), which would exemplify the word or phrase’s use in its typical lexical and grammatical environment. The requirement to provide dictionary-type examples was stipulated to reduce, as much as possible, idiosyncratic responses. The task is quite a high level vocabulary task (for the lower level language group), in terms of testing vocabulary knowledge (Wesche and Paribakht, 1996), as lexical item knowledge may only be partial, and indeed may not be sufficient to provide a sentence containing the word. It was decided to ask for a maximum of three sentences, rather than just one, as the interest in this paper is the type of associates that the word has in the minds of the participants, not necessarily its strongest associate (see De Groot, 1989: 824). In designing the task, rather than providing strict instructions on usage for every word (e.g. ‘Use face in an abstract sense’, or ‘Use potentially before an adjective’ etc.), no restrictions on usage were stipulated other than giving the part of speech of the word. In adopting this practice it was recognized that the data would need ‘trimming’ after being collected. Time to complete the task varied from between 10-25 minutes. The issue of semantic prosody was not referred to in the explanation of the task: participants were naïve as to the purpose of the task.

A potential weakness with the task design was that it was possible that a participant was aware (implicitly at least) of the typical semantic prosody of an item, but may not have provided a sentence showing this; however, the elicitation approach adopted was deemed preferable to a more explicit test – i.e. requiring the participants to ‘assign’ the items a positive or negative semantic prosody, as it is ‘tapping’ knowledge which is more implicit, and usage-based, as opposed to more abstract or metalinguistic. Such an approach has been used before in investigating aspects of word knowledge (see e.g. Gilquin, 2005, 2008; Nordquist, 2004; and Schmitt, 1998).

Participants

Group 1

The first group of participants consisted of undergraduate Arab students majoring in Translation at a Gulf Arab university. Thirty-seven students took part in the task – three males and 34 females. All the students are Omani, and their language level would, broadly speaking, be considered of an intermediate level.

Group 2

The second group consisted of male qualified Arabic mother-tongue teachers of English, working at the Language Centre of a university in the Arabian Gulf. All have an English language teaching qualification – whether certificate, diploma, degree or post-graduate degree, and all would be considered upper intermediate or advanced level users of English. Table 2, below, indicates the breakdown of nationalities within the group.

Table 2.

Arab Teacher Participants According to Nationality

GROUP 2	NATIONALITIES
Arab non-native speakers (31)	Egyptian (12), Jordanian (7), Sudanese (6), Syrian (5), Yemeni (1)

Group 3

The third group of participants consisted of 23 male native speaker English language teachers. All of these teachers worked at the same university as the non-native speaker teachers in Group 2, and all are similarly qualified. Table 3, below, lists the teachers according to their nationalities.

Table 3.

Native Speaker Participants According to Nationality

GROUP 3	NATIONALITIES
Native speakers (23)	British (13), American (7), South African (2), Nigerian (1)

All three groups were given the same task, containing the same lexical items. The task was administered by the author in the same setting for Groups 2 and 3, and in a similar setting for Group 1.

Results and Discussion

Trimming and Classifying the Data

After collecting the data, sentences were included or excluded for analysis. Sentences were excluded for four different reasons: where the wrong part of speech was used (e.g. face used as a noun, or potential used in a sentence instead of the required word potentially); where the usage had no bearing upon semantic prosody claims (e.g. face being used in a literal sense); where the same subject used the stimulus word with the same collocate (e.g. the collocation face difficulties being produced twice), in which case only the first example was retained for analysis; and finally when the word was used in a clearly incorrect way (semantically), or the sentence was not comprehensible, for example, ‘Please bring about this title more information’. Where there were grammatical mistakes in the sentence, but the prosody was clear, the sentence was retained for analysis (e.g. ‘The tree looks completely die’).

In the majority of cases it was not difficult to classify the elicited sentence as positive, negative or neutral, and where more evidence was needed to understand the evaluative force of the sentence, it was assigned a ‘neutral’ classification. For example, ‘He was completely surprised’ could be a positive or a negative evaluation (depending upon whether the surprise was a pleasant one or not, among other things). As there was no explicit indicator (either positive or negative) in the words around completely in the above example, the ‘default’ classification of ‘neutral’ was assigned to the sentence.

Tables 4, 5 and 6, below, provide the data for each word, including the number of analysed sentences for the stimulus items, and the percentage of sentences which were deemed to display positive, negative and neutral prosody for each of the three groups of participants. As it was technically possible for a subject to produce three sentences, the last column in each table indicates how many different participants contributed sentences of a certain type, indicating how widespread a certain usage was. It was noticed, in analysing the sentences, that it was by no means the case that a subject necessarily produced the same prosody if two or three different sentences were provided. Reasons for the occasionally large disparity in elicited sentences and analysed sentences are noted below.

Table 4.

Group 1 (Students) Data

Lexical item	Total no. of sentences elicited	Analysed sentences	Prosody of elicited sentences	Prosody attestation		No. of different participants
				No.	(%)
bring about (+)	13	0	+	0	(0)	0
			–	0	(0)	0
			N	0	(0)	0
cause (−)	54	45	+	3	(6.7)	2
			–	42	(93.3)	26
			N	0	(0)	0
completely (−)	52	35	+	19	(54.3)	16
			–	11	(31.4)	9
			N	5	(14.3)	5
face (−)	53	35	+	0	(0)	0
			–	32	(91.4)	24
			N	3	(8.6)	3
potentially (−)	11	1	+	0	(0)	0
			–	1	(100)	1
			N	0	(0)	0
provide (+)	58	45	+	45	(100)	24
			–	0	(0)	0
			N	0	(0)	0
regime (−)	42	2	+	0	(0)	0
			–	0	(0)	0
			N	2	(100)	2

N = 37

Table 5.

Group 2 (Arab Teachers) Data

Lexical item	Total no. of sentences elicited	Analysed sentences	Prosody of elicited sentences	Prosody attestation		No. of different participants
				No.	(%)
bring about (+)	34	22	+	5	(22.7)	5
			–	14	(63.6)	10
			N	3	(13.7)	3
cause (−)	67	59	+	2	(3.4)	2
			–	54	(92)	29
			N	3	(4.6)	3
completely (−)	48	44	+	12	(27.3)	10
			–	23	(52.2)	17
			N	9	(20.5)	8
face (−)	48	44	+	0	(0)	0
			–	44	(100)	22
			N	0	(0)	0
potentially (−)	34	10	+	4	(40)	4
			–	4	(40)	4
			N	2	(20)	2
provide (+)	48	46	+	44	(95.7)	23
			–	0	(0)	0
			N	2	(4.3)	2
regime (−)	54	30	+	5	(17)	4
			–	10	(33)	8
			N	15	(50)	15

N = 31

Table 6.

Group 3 (NS Teachers Data)

Lexical item	Total no. of sentences elicited	Analysed sentences	Prosody of elicited sentences	Prosody attestation		No. of different participants
				No.	(%)
bring about (+)	46	38	+	31	(81.6)	21
			–	5	(13.2)	5
			N	2	(5.2)	2
cause (−)	60	50	+	3	(6)	3
			–	42	(84)	22
			N	5	(10)	5
completely (−)	55	51	+	16	(31.4)	13
			–	29	(56.8)	18
			N	6	(11.8)	6
face (−)	59	36	+	0	(0)	0
			–	34	(94.4)	18
			N	2	(5.6)	2
potentially (−)	47	21	+	6	(29)	5
			–	15	(71)	14
			N	0	(0)	0
provide (+)	58	58	+	50	(86)	23
			–	0	(0)	0
			N	8	(14)	6
regime (−)	51	44	+	2	(4.5)	2
			–	25	(56.8)	17
			N	17	(38.7)	13

N = 23

As can be seen, for Group 1, only four of the seven items elicited sufficient data for analysis. Of these, three concurred with the semantic prosody observations in the literature: cause, face, and provide. For completely more sentences were produced in which the word was used in a positive sense, as opposed to a negative sense – the dominant sense in corpus data according to Partington (2004). With regards to Group 2 (the Arab teachers), the elicited data (when considered collectively) concurred with the corpus linguist claims for five of the seven words, but not for potentially, or bring about. For the third group (the NS teachers) there was general agreement between the elicited data and the overall semantic prosody observations from the corpus data for all of the words. It is evident that there is a cline of evaluative positivity/ negativity for these items for this group (cf. Morley and Partington, 2009: 151 and their comments regarding the cline of positivity or negativity associated with an item). For example, the percentage of sentences where provide was used by Group 3 in a positive context is greater than that for bring about, and cause appeared to be more evaluatively negative in the minds of the participants than potentially.

While the data from Group 3 might be considered to be unexceptional or ‘expected’, it should be noted that some corpus linguists have been highly skeptical concerning the ability of native speaker language users to produce typical language in a decontextualized setting (e.g. Sinclair, 1997: 31; Beaugrande, 1999: 247). There is some justification for this view: L1 language users, for example, do not seem to be aware of the most common collocates of words (Gilquin, 2005; McGee, 2009; Nordquist, 2004) or of the typical grammar patternings of verbs (Roland and Jufarsky, 2002). Here, however, there is general accord between observations in the literature about the semantic prosodies of these seven words and the native speakers’ ideas about typical usage elicited in a decontextualized setting. It would seem that Hunston’s (2002: 20, 21) claim about poor intuition in the area of semantic prosody would, in the light of these data, at best, be restricted to more abstract metalinguistic knowledge, until further tested.

Specific Comments

Group 1 almost exclusively used the word ‘regime’ with reference to a diet. This is most likely due to the effect of the ‘false’ cognate in Arabic (pronounced /reɪdʒim/ or /reɪʒim/ meaning ‘diet’/‘regimen’). No student provided an appropriate use of bring about as a phrasal verb, and only 1 student did so for potentially as an adverb – and very few attempts were made to use these words. As participants were asked to leave words for which they could not provide sentences, this was not considered to be problematic. Group 2 also, at times, used regime in a similar way to Group 1 and this data was also excluded from analysis. This was also very occasionally the case for Group 3, many of whom have some knowledge of Arabic. The reason for the quite large disparity between the elicited and analysed sentences for potentially for Groups 2 and 3 was that non-attributive uses were excluded, in line with the corpus claims.

In what follows I look briefly at each word’s treatment, providing example sentences when necessary to highlight areas of interest.

Bring about

Perhaps surprisingly, this phrasal verb was almost universally avoided by the students (Group 1) – who seem either not to have known this lexical item, or not known it well enough to use it. The Arab teachers, however, were confident enough to use the phrase, and the full set of data from this group is provided below in Table 7 along with the classification.

Table 7.

Group 2 (Arab Teacher) Elicited Data for ‘bring about’

They can	bring about	new ideas for solving the problem.	+
Hard times	bring about	the best of our morals.	+
Bees	bring about	honey.	+
Each government should	bring about	good education.	+
I will	bring about	everything.	+
Smoking	brings about	a lot of health problems.	–
He was very ignorant; which	brought about	many troubles.	–
His personal defects	brought about	his tragic downfall.	–
What you have done in the past	brought about	all these problems now.	–
The invasion of Iraq	brought about	a lot of disasters.	–
Wars	bring about	disasters.	–
Ignorance	brings about	loss in money and efforts.	–
Poverty	brings about	a lot of miseries around the world.	–
Misunderstanding	brings about	a lot of misunderstanding.	–
Poverty and deprivation	brought about	a lot of crimes in society.	–
The idea of teaching EFL	brings about	a lot of barriers.	–
Problems ..… are mainly	brought about	(by) the ill−thinking people.	–
The flood	brought about	much damage to the village.	–
Do not	bring that about	It’ll make him sad.	–
Joining this department	brought about	many problems and advantages.	N
The war	brought about	a lot of debate.	N
Clouds	bring about	rain.*	N

Classed neutrally as this may have been given as a scientific fact, and rain’s desirability or otherwise is culturally determined.

The majority of the Arab teachers used the word seemingly synonymously with cause, in the sense that the thing caused was, generally, negative, e.g. ‘Smoking brings about a lot of health problems’. Indeed, even some of the sentences classed as being positive (sentences 3, 4 and 5 in Table 7) do not seem particularly idiomatic – and while ‘bring about’ within these sentences seems to have a causation meaning, it does not always clearly do so – and would require some syntactic adjustment to do so, at times.

Cause

For this item there was a strong tendency for the word to be used with negative effects across all three groups. An example from each group is provided below.

His selfishness causes his failure. (Group 1)

Tampering with matches may cause fire all over the house. (Group 2)

Being careless causes failure. (Group 3)

However, instances where the word was used with positive consequences need to be differentiated: the three uses by group 1 seem awkward and not particularly idiomatic:

4. Her parents support causes her success.

5. His nice dealing with his customers causes crowd in his shop.

6. Studying hard cause the good grades.

The positive sentences provided by the NS teachers have a greater ring of idiomaticity about them, even though the prosody is not the dominant one:

7. He causes the sun to shine.

8. The news of the election results caused widespread celebration.

9. I like to cause others to think out of the box.

Completely

The treatment of this word was unique, in the sense that it was the only item for which the lower level language group provided a set of responses which, when considered collectively, indicated a preference for a different dominant semantic prosody to that of the more linguistically proficient Groups (2 and 3). The full data from Group 1, along with the classification, is provided below in Table 8.

Table 8.

Group 1 (Students) Data for ‘completely’

They have	completely	finished their work.	+
Do you understand	completely	what I have said already?	+
My friend finished her exam yesterday	completely.		+
I am ready for Eid	completely.		+
I am finishing studying to the exam	completely.		+
When every you’ll be asked for your duty, do it	completely.		+
I do my homework	completely.		+
I finished all the questions	completely.		+
This paint is	completely	perfect.	+
The work is	completely	done.	+
It is	completely	right.	+
I left everything there and it was	completely	clear.	+
Nothing in this life is	completely	correct.	+
Do your homework	completely	and then give it to me.	+
I do my work	completely	last night.	+
What I learn with this teacher were	completely	easy.	+
He do his homework	completely.		+
He does the work	completely	done.	+
I have done all my duties	completely.		+
Stolen money is	completely	wrong.	–
This tree looks	completely	die.	–
You are	completely	looks tired.	–
He was	completely	speechless when he asked to tell the truth.	–
Your homework is	completely	wrong.	–
He is	completely	wrong.	–
This person is	completely	crazy.	–
I am completely	completely	tired.	–
I think my answers in the last exam were	completely	wrong.	–
My phone fall into the water and it	completely	damaged.	–
That was	completely	crazy.	–
He speaks	completely	different from his parents.	N
The culture in Britain is	completely	different from Oman.	N
The good student has	completely	different ideas.	N
What I see is	completely	different.	N
We are	completely	different when we are in the formal situation.	N

It is not our intention here to comment on grammatically incorrect uses of the word (e.g. I am ready for Eid completely, or You are completely looks tired), as this is not our focus. What is important to note is that for verb modification, the sentence is always positive. For attributive adjective use, on the other hand, the negative overtones are more common – around twice as common as the positive classification – with examples including wrong, crazy, tired and damaged. For Group 2 the pattern of usage is different – the verbs used were both positive and negative (roughly equally so), though there was a preference (similar to Group 1) to provide a negative environment when the stimulus word was used in attributive adjective position (the case for 12 of the 14 sentences produced). Group 3’s responses are very similar, in terms of the general patterning, to that of Group 2’s.

The key difference, therefore, between Group 1 and 2’s responses is that group 1 not once uses the adverb with a ‘negative’ verb, whereas Group 2 uses such verbs as: hate, reject, refuse, destroy. The consequence of this, for Group 1 as a whole, is that the usage is more commonly positive than negative, which does not concur with the corpus-observed tendencies. The total absence of negative verb usage is significant – the students seem to have some understanding of the word and can use it acceptably in sentences; however, an important aspect of the word’s typical combinatory patterning seems to be under- or undeveloped. The fact that there is such an appreciation with Group 2, seems to indicate an incremental understanding (tied to proficiency) of the semantic prosody associated with this word.

Face

Regarding face, Tognini-Bonelli (2001: 95) notes that in a general corpus its abstract sense usage often has to do with, ‘confronting problematic areas: problems, problem, fear, serious’. Our elicited data concur with these kinds of grammatical objects (in terms of semantic prosody) – and most of the sentences were classed as conveying a negative prosody. This is, however, one of those cases where an overly simplistic view of semantic prosody begins to encounter problems. For many of the sentences, the overall discourse prosody was clearly negative, as in the examples below:

10. Our country is facing a lot of problems.

11. I have never faced such (a) number of problems and hardships.

12. I face this problem weekly.

However, in a number of the elicited sentences, although the objects were negative (and hence the sentence was classed negative), the participants gave the word face a positive aura: to face problems, rather than to run away from them, was positive.

13. One should face challenges otherwise he will be a slave.

14. I faced my illness with no fear.

15. You have to face your problems.

The above examples serve, inter alia, to highlight the problematic issue of where semantic prosody lies, alluded to at the beginning of the paper. These examples confirm the inadequacy of views which assign semantic prosody to a word, or even to a word’s collocates, and give extra weight to the importance of taking an ‘overall discourse’ stance when considering semantic prosody.

Potentially

The analysable data from Group 2 for this word was rather limited, as only attributive instances of usage were retained for analysis, as per the corpus data claims. Table 9, below, provides the full set of data, from 10 different participants.

Table 9.

Group 2 (Arab Teachers) Data for ‘potentially’

This grammar rule is	potentially	strong.	+
He is	potentially	skilful and can achieve the task.	+
He is	potentially	good in English.	+
This project is	potentially	recognized.	+
Living alone can be	potentially	uncomfortable.	–
This company is	potentially	poor.	–
He is	potentially	unable to do the task.	–
Put	potentially	dangerous materials away from children.	–
This research paper is	potentially	different from my paper.	N
He was	potentially	aware of what he was doing.	N

Around twice as much data was collected from the NS participants (Group 3), but unlike the group 2 data, the tendency was for the word to be used in a negative way – of the 21 uses, 15 uses were negative: dangerous (9), unethical, wrong, fatal, life-threatening, active⁴ and disturbing). One might hypothesize that because potential is something positive (e.g. He has potential), that the Arab teachers assumed that potentially carries with it this same positivity – i.e. that there is a false extension into the adverbial form. Although caution should be taken with these data, being limited, the responses to this word do indicate that Group 2’s ideas about the word’s typical usage is at variance with Group 3’s.

Provide

Provide was almost exclusively used in a positive sense by all three groups. The kinds of things ‘provided’ included: support, food, money, help and equipment – and are similar to the list provided by Stubbs (1995b: 247, 248). Provide is quite unlike its near synonym give, in the sense that it is not used to ‘give’ something undesirable. The LDOCE notes, for example, that one can give someone problems, trouble, difficulties and punishments, among other things. Even the lower level language group (Group 1) seemed to appreciate this difference – not one negative example for this item being elicited. The LDOCE defines ‘provide’ in the following way: ‘to give something to someone or make it available to them, because they need it or want it’. The second clause of this definition adds the required prosody – the giving is desired. In terms of usage, almost all participants who wrote a sentence seemed to appreciate, implicitly at least, that to provide is to give something good.

Regime

Regarding regime, Channell, on studying corpus data, noted that it was typically used to ‘invoke a critical stance’ (2000: 47). In analysing the data, many instances had to be classed neutral, as insufficient information was present in the sentence to be sure about the speaker’s stance. Three examples are noted below, one from each classification:

16. This is a just regime (+)

17. The regime in Israel is racist (-)

18. You read about Saddam Hussein’s regime in Iraq (N)

As noted earlier, there were problems in collecting data for this word (particularly so for Group 1); however, the word seemed to be more associated with negative political systems or governance rather than positive. An experimental design with far more context provided would be needed to investigate usage patterns for this word more thoroughly.

Conclusion

The data collected, analysed and discussed above suggests that semantic prosody appreciation may or may not be picked up by L2 learners. Three of the words (cause, face and provide) elicited similar data across the three language levels. This suggests that semantic prosody awareness is not, necessarily, dependent upon language level – but may be acquired relatively early on by the learner – though note the untypical sentences 4, 5 and 6 above from Group 1 for cause. The data collected for the word completely from Group 2, when compared to Group 1, suggest that semantic prosody appreciation may be incremental – and one can fairly safely assume that the incremental learning would have been incidental, i.e. through experience and usage, rather than through explicit instruction. However, the data collected for bring about, and, to a lesser extent potentially, suggest that it is possible for quite advanced proficiency learners not to be aware of the typical semantic prosody usage of an item, though knowing it well enough to use it. The data is mixed, but as it cannot be assumed that semantic prosody awareness is implicitly acquired, or even acquired incrementally, it would seem reasonable to recommend that teaching materials begin to address semantic prosody as they have recently begun to seriously address collocation. It is suggested that insights from corpus linguistics studies be incorporated into student learning materials, highlighting semantic prosody awareness where appropriate.

There are several possible reasons for why the data from Groups 1 and 2 were mixed, in terms of corpus agreement. First, as noted earlier on, it is difficult, if not impossible to discuss semantic prosody without allowing for a cline of hidden-ness – and it would seem that, with regard to usage, an item’s typical semantic prosody may be more or less ‘hidden’: for cause, provide and face, for example, it seems that the ability to use the word includes implicitly acquired semantic prosody knowledge – which is not hidden with regards to usage. However, for other words this is clearly not the case: the L2 learner may be able to use a word such as completely, bring about or potentially, but lack such an awareness – in such cases the word’s typical semantic prosody appears to be ‘really’ hidden from user awareness. Another possible explanation for the differences may be found in language transfer. A number of researchers looking at semantic prosody cross-linguistically have noted that similar words across different languages often carry similar prosodies (See Xiao and McCenery, 2006, Chinese-English; Sardinha, 2000, English-Portuguese; and Tognini-Bonelli, 2001:143, English-Italian). It may be then, as has been noted in collocation studies (e.g. Bahns, 1993), that cross-linguistic transfer may occur in this area, and such transfer may (given the above findings from cross- linguistic studies) be more successful than not. Having said this, cases have also been noted when semantic prosody transfer does not ‘work’ (as noted by Sardinha, 2000), and there may be instances where a language simply does not have a similar lexical item from which prosody can transfer. Third, and finally, an additional possible reason for a language learner’s lack of awareness of semantic prosody may lie in a lack of exposure to such words in their typical contexts, resulting in cases where a ‘not fully-developed’ appreciation of a word’s typical usage results.

Turning now to more theoretical issues, the fact that ‘opposite’ semantic prosody usage may be idiomatic, and non-ironic, needs to be given greater emphasis in discussion about semantic prosody. Partington’s observation (2004:153) that prosodies are rarely black and white is a valid one, but we might add that they can be both black and white. It would seem that there is a need for a more sensitive lexicogrammatical approach to semantic prosody which can, among other things, deal with a word like utterly being perfectly idiomatic in both negative (e.g. utterly miserable) and positive (e.g. utterly brilliant) environments, without concomitant irony, in the non-dominant prosody usage. Unless this occurs, the theorizing of semantic prosody may not survive critical assaults.

There are a number of limitations to this study. It could be argued that the experimental design used in this research is overly simplistic, for one does not start with a word and create a sentence: one starts with a message which one puts into words. Further, when one starts with an evaluative message, this emotional involvement is singularly difficult to capture in an experiment of the type described above. These are all valid points, and part of the problem with ‘test-tube’ sentences (Roland and Jurafsky, 2002: 334). Future studies might include a metalingusitic task: a simple rating task, where participants assign semantic prosodies to words, to further investigate Hunston’s (2002: 20, 21) claim about poor intuitive awareness for this aspect of language. In addition, the number of lexical items used in this study was rather small, and future studies could investigate many more items – though only with advanced level L2 language users. It is hoped that results from such research may help us further investigate this neglected aspect of language use and awareness, and that such data may inform theoretical discussion much more than has been the case to date.

Footnotes

Notes

References

Bahns

(1993) Lexical collocations: a contrastive view. ELTJ 47(1): 56-63.

Biber

(2011) Corpus linguistics and the study of literature: back to the future? Scientific Study of Literature 1(1): 15-23.

Channell

(2000) Corpus-based analysis of evaluative text. In: Hunston

Thompson

(eds) Evaluation in Text: Authorial Stance and the Construction of Discourse. Oxford: Oxford University Press.

De Beaugrande

(1999) Reconnecting real language with real texts: text linguistics and corpus linguistics. International Journal of Corpus Linguistics 4(2): 243-59.

De Groot

AMB

(1989) Representational aspects of word imageability and word frequency as assessed through word association. Journal of Experimental Psychology: Learning, Memory, and Cognition (15): 824-45.

Hornby

(1995) Oxford Advanced Learner’s Dictionary (5th Edition). Oxford: Oxford University Press.

Gilquin

(2005) To take or not to take phraseology into account. The place of multi-word sequences in corpus data and experimental data. In: Cosme

Gouverneur

Meunier

Paquot

(eds) Proceedings of the Phraseology 2005 Conference, Louvain-la-Neuve, 13-15 October 2005. Louvain-la-Neuve: Universitée catholique de Louvain.

Gilquin

(2008) What you think ain’t what you get: highly polysemous verbs. In: Lapaire

Desagulier

Guignard

(eds) (2008) Du fait grammatical au fait cognitive. From Gram to Mind: Grammar as Cognition. Volume 2. Pessac: Presses Universitaires de Bordeaux.

Hunston

(2002) Corpora in Applied Linguistics. Cambridge: Cambridge University Press.

10.

Hunston

(2007) Semantic prosody revisited. International Journal of Corpus Linguistics 12(2): 249–68.

11.

Louw

(1993) Irony in the text or insincerity in the writer? The diagnostic potential of semantic prosodies. In: Baker

Thompson

Tognini-Bonelli

(eds) Text and Technology. In Honour of John Sinclair. Amsterdam: John Benjamins.

12.

Louw

(2008) Contextual prosodic theory: bringing semantic prosodies to life. Texto! 13/1. Available at: http://www.revuetexto.net/docannexe/file/124/louw_prosodie.pdf. (Last accessed 2 January 2011).

13.

McGee

(2009) Adjective-noun collocations in elicited and corpus data: similarities, differences and the whys and wherefores. Corpus Linguistics and Linguistic Theory 5(1): 79-103.

14.

Morley

Partington

(2009) A few frequently asked questions about semantic – or evaluative – prosody. International Journal of Corpus Linguistics 14(2): 139-58.

15.

Nordquist

(2004) Comparing elicited data and corpora. In: Achard

Kemmer

(eds) Language, Culture and Mind. Stanford, CA: CSLI Publications.

16.

Partington

(2004) Utterly content in each other’s company. Semantic prosody and semantic preference. International Journal of Corpus Linguistics 9(1): 131-56.

17.

Partington

(2011) Phrasal irony: its form, function and exploitation. Journal of Pragmatics 43(6): 1786-1800.

18.

Richards

(1976) The role of vocabulary teaching. TESOL Quarterly 10(1): 77-89.

19.

Roland

Jurafsky

(2002) Verb sense and verb subcategorisation probabilities. In: Merlo

Stevenson

(eds) The Lexical Basis of Sentence Processing. Formal, computational and experimental issues. Philadelphia, PA: John Benjamins.

20.

Sardinha

(2000) Semantic prosodies in English and Portuguese: a contrastive study. Cuadernos de Filología Inglesa 9(1): 93-110.

21.

Schmitt

(1998) Measuring collocational knowledge: key issues and an experimental assessment procedure. International Review of Applied Linguistics: 119-20: 27-47.

22.

Seidlhofer

(eds) Principle and Practice in Applied Linguistics: Studies in Honour of H.G. Widdowson. Oxford: Oxford University Press.

23.

Sinclair

(1997) Corpus evidence in language description. In: Wichmann

Fligelstone

McEnery

Knowles

(eds) Teaching and Language Corpora (Applied Linguistics and Language Study). Harlow: Longman.

24.

Sinclair

(1998) The lexical item. In: Weigand

(ed.) Contrastive Lexical Semantics. Amsterdam: Benjamins, 1-24.

25.

Sinclair

(ed.) (1990) Collins COBUILD English Grammar. London: Collins.

26.

Stewart

(2010) Semantic Prosody: A Critical Evaluation. London: Routledge.

27.

Stubbs

(1995a) Collocations and semantic profiles: on the cause of the trouble with quantitative studies. Functions of Language 2(1): 23-55.

28.

Stubbs

(1995b) Corpus evidence for norms of lexical collocation. In: Cook

Summers

(ed) (2001) Longman Dictionary of Contemporary English, 3rd Edition. Essex: Longman.

29.

Stubbs

(2002) Words and Phrases: Corpus Studies of Lexical Semantics. Oxford: Blackwell.

30.

Tognini-Bonelli

(2001) Corpus Linguistics at Work. Amsterdam: John Benjamins.

31.

Wesche

Paribakht

(1996) Assessing L2 vocabulary knowledge: depth versus breadth. The Canadian Modern Language Review 53(1): 13-40.

32.

Xiao

McEnery

(2006) Collocation, semantic prosody, and near synonymy: a cross-linguistic perspective. Applied Linguistics 27(1): 103–29.

33.

Zhang

(2009) Semantic prosody and ESL/EFL vocabulary pedagogy. TESL Canada Journal 26(2): 1-12.