Summary
#e paper discusses associative meaning, i.e. one existing over and above the customary denotation, 
specifically the type arising from a text segment larger than a single word. #e idea is of fairly 
recent origin, focuses on negative and positive semantic effects, and stems from corpus-based 
findings. Dictionaries are uneven in their treatment of this aspect of meaning. It is suggested that 
research on this complex phenomenon of associative meaning might be conducted on any of 
three levels: single-word items (connotation), multiword items (semantic prosody), and broader 
if vaguer co(n)text (syntagmatic meaning). 
Key words: connotation, semantic prosody, syntagmatic meaning, collocational 
prosody, collocation, multiword lexical item
 Povzetek
Članek obravnava več vrst asociativnega pomena, tj. tistega, ki obstaja poleg standardnega 
denotativnega in se pojavlja v besedilih v kombinacijah vsaj dveh besed. Ideja o tem pomenu je 
sorazmerno nova; ukvarja se predvsem z negativnimi in pozitivnimi pomenskimi učinki, prihaja 
pa s področja korpusnih raziskav. Slovarji ta pomen obravnavajo neenotno. Avtor predlaga, da bi 
ta kompleksen pojav raziskovali na treh ravneh: v posamičnih enobesednih leksemih (konotacija), 
v večbesednih enotah (semantična prozodija) in v širšem – četudi manj jasno opredeljenem 
– sobesedilu (sintagmatski pomen). 
Ključne besede: konotacija, semantična prozodija, sintagmatski pomen, kolokacijska prozodija, 
kolokacija, večbesedna leksična enota
 DOI: 10.4312/elope.4.1-2.9-28

When looking at collocations as a pervasive phenomenon demonstrating and powerfully 
illustrating the functioning of a very significant aspect of the functioning of what has come to 
be known as the co-selection principle, one is struck by the fact that – unlike irregular verbs, 
tense usage, passives and relatives, reported speech, relative clauses, conditional sentences, and 
all that jazz – collocations often do not lend themselves to the familiar and charmingly simple 
right-or-wrong type of assessment. Indeed, it makes a lot of sense to consider the issue in relative 
terms, basically as one of lexical acceptability
1
 (Ball 1987, 188), meaning that few collocations 
can be firmly excluded as impossible, as they range from the unquestionably acceptable to the 
extremely unlikely, with context often being all-important.
2
 While collocations are all indicative 
of one type of varied patterns of mutual choice, illustrating the vagaries of combinability and its 
restrictions, it does seem that in collocability semantic factors are usually involved as well, even 
though they are sometimes quite slight or simply difficult to pinpoint. Accordingly, the search for 
collocational (a.k.a collocative) meaning regarded either as a distinct collocational contribution to 
lexical meaning recognized in single-word items or even as a discrete type of lexical meaning
3
 has 
resulted in several original suggestions arguing convincingly for the existence of such a meaning. 
#ese include the fairly restricted – more specifically, one restricted to collocations – concept of 
semantic tailoring (Allerton 1984), used to refer to the process in which the polysemy of the 
adjectival collocator is “narrowed down” or “trimmed” by the semantics of the base noun (e.g. an 
outstanding success [‘izjemen uspeh’] vs. an outstanding debt [‘neporavnan dolg’], or regular customer 
[‘reden gost’, ‘stalna stranka’], regular gas [‘navadni bencin’], regular duties [‘običajne dolžnosti’], 
regular heartbeat [‘enakomeren srčni utrip’], regular verb [‘pravilni glagol’], regular features [‘pravilne 
poteze’], regular army [‘poklicna vojska’]). Another recent idea revolves around a less clearcut but 
intriguing semantic concept usually dubbed semantic prosody, but also variously referred to as 
collocational prosody, discourse prosody, or as pragmatic prosody (cf. e.g. Stubbs 2002, 65−6, 


and Stubbs 1995a).
4
 In a pioneering paper, the concept has been defined as “the consistent aura 
of meaning with which a form is imbued by its collocates” (Louw 1993, 157). #is paper focuses 
on the latter phenomenon, thus disregarding other possible ways of exploring the diverse effects 
of very real – if somewhat elusive – semantic features of words arising from their being frequently 
used together in a phraseologically salient, or indeed syntagmatically significant, fashion.
5
 #e term semantic prosody was apparently first coined by John Sinclair (Partington 1998, 66−7)
6
 
to describe the phenomenon of a favorable or unfavorable connotation being contained not 
in a single item, but rather being expressed by that item in association with others, as e.g. in to 
happen and to set in, both of which are habitually associated with unpleasant events. It is “a kind 
of attitudinal or pragmatic meaning” (Sinclair 2004b, 23) that a lexical item has in addition to 
“the familiar classificatory meaning of the regular dictionary.” #e concept has been defined also 
along the following lines: “[A] word may be said to have a particular semantic prosody if it can be 
shown to co-occur typically with other words that belong to a particular semantic set” (Hunston 
and Francis 2000, 137), as “a feature which extends over more than one unit in a linear string”. 
Moreover, “discourse prosodies express speaker attitude” (Stubbs 2002, 65). Finally, Warren (2005) 
suggests that it is the combinatory restrictions of words that can also be seen in terms of semantic 
prosodies: “#at is to say, a particular word typically combines with words of a particular type of 
– normally evaluative – meaning which is not warranted by generalised meanings.” #e concept 
seems to have been introduced to the public by Bill Louw in 1993, with Sinclair having originally 
suggested it to Louw (Whitsitt 2005, 283). It is not yet to be found recorded in the standard 
English dictionaries of linguistics terms such as those by Crystal (2003) and Matthews (2007). 
Semantic prosody indicates the phenomenon of words combining not just with chosen other 
words, but with chosen meanings, thus displaying their semantic prosodies (cf. T ognini-Bonelli 
2001, 111−6) that appear to be mostly either positive or negative. In line with this observation, 
Partington (1998, 66−8) notes that to commit and rife both collocate with items of an unpleasant 
nature, so that the unfavorable connotation extends over the entire collocation. Similarly, to set 
in often signals that some undesirable process is being described. Semantic prosody “refers to 
the spreading of connotational colouring beyond single word boundaries” (ibid., 68) which, 
importantly, can only be found in certain words. In a broader framework, the term “reflects the 
[neo-Firthian] realisation that lexical items become infused with particular connotations due to 
their typical linguistic environment.” Partington (1998, 66) regards it as “one particularly subtle 


and interesting aspect of expressive connotation which can be highlighted by corpus data.” It 
can be a crucial aspect of an item’s lexical meaning, underscoring the contemporary conviction 
that meaning resides in typical combinations of lexical choices or “collocability” on the one 
hand, and typical combinations of grammatical choices or “colligation” on the other (Siepmann 
2006, 9). But does this mean that semantic prosodies are merely a matter of a single-word lexical 
item spreading its own connotative influence, in the process imposing certain logico-semantic 
restrictions, on to its surroundings? 
 #e concept of semantic prosody has been recently discussed at length, chiefly by post-Firthian 
corpus linguists following largely in Sinclair’s footsteps, as part of the awareness that “we 
do not communicate by stringing together individual words, but rather by means of semi-
prefabricated lexico-grammatical units” (Siepmann 2006, 9). #ese researchers include, 
chronologically and selectively, Louw (1993), Stubbs (1995b, 246 and 2002, 105−8, 198−206), 
Bublitz (1996, 11), Partington (1998, 65-78 and 2004), Rundell (2000), Cotterill (2001), 
Hunston (2001), Channel (1999), Schmitt and Carter (2004, 7−9, 20), Whitsitt (2005), and 
Dilts and Newman (2006, 233), with Hunston (2007) contributing a recent reassessment. 
Most of the work done on the topic has been in the monolingual mode, with English being 
for the most part heavily favored; however, there is an English and Chinese perspective on the 
phenomenon with reference to near-synonyms provided by Xiao and McEnery (2006), while 
Sardinha (2000) focuses on English and Portuguese. An early cross-linguistic study of this kind 
is Lewandowska-Tomaszczyk (1996). Louw’s own (1993) definition of the concept, not taken 
up by all later researchers,
7
 is that if several different words all sharing the same semantic trait 
are frequently used with another word, meaning will be passed, over time, from that group 
of words to the other word. Louw (1993) himself studied to be bent on something, which has 
a negative prosody (meaning ‘to be completely determined to do something’ but suggesting 
also ‘something bad’).
8
 #e key idea is that “constant proximity between words can lead to 
promiscuity wherein the meaning of one word or words will be ‘rubbed off’ onto another” 
(Whitsitt 2005, 284). Nowadays, the concept is so important that even certain textbooks treat 
it at some length (e.g. McEnery et al. 2006, 82−5, 148−51, passim). 
 For some reason, most authors seem to have identified a number of instances of “negative” 
prosody and far fewer cases of “positive” prosody, witness e.g. the frequently cited cases of 
to set in (Sinclair 1987, 155−6, passim), to cause (typically collocating with problems, trouble, 
damage, death, pain, and disease) as contrasted with to provide (typically collocating with facilities, 
information, services, aid, assistance, and money) (cf. Stubbs 1995a, also summarized in Schmitt 


2000, 78-9), dealings (Partington 1998, 72−4), and to happen and to slump as in slumped in front 
of the TV (Rundell 2000).
9
 #us e.g. the unusualness of the combination utterly content vs. perfectly 
content results from the fact that utterly typically restricts the choice of its collocates to words 
with some negative semantic content (hence the title of Partington’s [2004] paper). Similarly, if 
something is fraught with something rather than being full of it, we can expect something negative 
(problems, difficulties, risks, ambiguities, etc.) following the preposition. It might be relevant to 
try and identify, in a cross-linguistic framework of EFL writing/speaking, recurrent instances of 
inappropriate semantic prosodies, as for instance in to make an *unforgettable mistake. 
To take another look at the functioning of the semantic “prosodic constraint,” the verb to 
harbor (‘to keep/have’) is likewise largely restricted to something undesirable (such as doubts, 
fears, bad thoughts, and the like).
10
 #e same reasoning, but in a lot stronger version, applies 
to the verbs to wreak (‘to cause problems or damage’) and to lurk somewhere (not only ‘to wait 
there quietly and secretly’ but also ‘in order to do something wrong’). Next, something that 
is mounting is not merely ‘gradually increasing’ but is typically used about things that cause 
problems or trouble. Also, one is doomed to extinction/failure etc. but destined for a successful 
career. #is example, incidentally, suggests the possibility of semantic prosody being the result 
of grammatical (rather than lexical) collocability, or – to use alternative terminology – of 
colligational links, as in e.g. to reek of [something], ‘to have a strong bad smell’ in both literal 
and metaphorical senses. But then in such cases the role of the preposition may be difficult to 
determine in prosodic terms. Further, to arouse in one of its patterns (but not all!) demonstrates a 
close association with “negative” nominal heads such as hostility, anger, resentment, and suspicion. 
McEnery et al. (2006, 83−4) list a selection of items studied recently for their semantic prosody: 
happen, set in; personal price vs. personal and price individually; cause, commit, peddle/peddler, 
dealings, end up verbing, a recipe for, get oneself verbed, fan the flame, signs of, ripe for, underage 
and teenager(s), sit through, bordering on; provide, career. All but the last two – which do have 
a positive prosody – carry an unfavorable meaning. Could this mean that negative semantic 
prosody is more pervasive than its positive offshoot?
Again, while “there can be no doubt that good-bad evaluation is an important (and previously 
neglected) component of lexical analysis” (Hanks 1997),
11
 it is a fact that not all lexical items 
are assessable on such a scale. Hanks (ibid.) mentions twig and telephone directory as having no 
good-bad semantic value, and goes on to point out that, whereas to incite has a negative semantic 
prosody in English (you incite people to bad actions), the evidence of the British National Corpus 
suggests that to urge and to encourage are neither positive nor negative, but neutral.


Furthermore, there are less commonly adduced examples of semantic prosody to be found in 
the language, some of them illustrating instances of the “phraseological-only” semantic prosody, 
that is, one where the prosodic meaning is necessarily associated with a multiword item or a 
pattern rather than any of its constituents in isolation or a single-word lexical item. #ese include 
fixed expressions such as to blow your own trumpet (BrE)/horn (AmE), which, according to the 
Longman (Summers 2005) means ‘to talk a lot about your own achievements’ – but then comes 
a dash and a note of warning: used to show disapproval. #ere are also relevant phrasal items 
with “slots” to be filled, for instance one sense of the noun catalog, as employed in the pattern 
a catalog of ___, namely one that is virtually always associated with something undesirable. #e 
Cambridge Advanced Learner’s Dictionary (Walter 2005) has captured this sense, illustrating it 
nicely with the example a catalogue of disasters/errors/crimes/complaints. Similarly, the same old ___ 
(story, excuse, faces, etc.) has a negative prosody, connoting chiefly boredom. Unfortunately, even 
the best dictionaries are not always successful in capturing this rather elusive aspect of meaning: 
for example, the prosodically sensitive, as it were, advanced learners’ Longman Dictionary of 
Contemporary English (Summers 2005), defines what with sth as follows: ‘used to introduce a 
list of reasons that have made something happen or made someone feel in a particular way,’ 
thus failing to indicate its negative prosody. Admittedly, the example given does suggest it, but 
dictionary users surely cannot be expected to conclude from a single example – aside from the 
well-known fact that examples of use are often virtually ignored by many dictionary users – that 
the prosody illustrated with it is obligatory. 
 Importantly, as the above examples suggest, semantic prosodies do not necessarily constitute 
a semantic feature arising exclusively or chiefly from syntagmatic-collocational links. 
On the contrary – it is an element of meaning that need not be generated strictly on the 
phraseological, or more narrowly collocational, level, witness e.g. to cause, to provide, to harbor, 
to wreak, and utter(ly) as exemplified above. #is fact is indicated in some definitions of the 
concept, e.g. that the most common understanding of the term semantic prosody seems to be 
“that some WORDS, or WORD GROUPS, occur in contexts which are understood by the 
researcher to have ‘positive’ or ‘negative’ nuances, or prosodies” (Dilts and Newman 2006, 
233 [my emphasis]). In fact, the distinction is often made between the familiar connotation, 
a term usually used with reference to the associative/attitudinal/evaluative/emotive meaning 
of a single-word item, and the more recent semantic prosody referring to the associative/
attitudinal/evaluative/emotive meaning of a multiword item. #is type of analysis may include 
an entire set of semantically related single-word items: For example, among the adjectives 
expressing the concept of FULL, complete (and several “neutral-to-positive” synonyms such 
as absolute, total, whole, entire, thoroughgoing, unqualified) can be contrasted with the “utter 
group” comprising utter as well as downright, rank, arrant, consummate and unmitigated, in 
that the latter group will typically precede a noun indicating something undesirable (for more 
along these lines, cf. Gabrovšek 2005, 150, 174−5). 

Unless one observes the terminological distinction between connotation and semantic prosody, one 
can, then, legitimately observe that semantic prosody is observable, first, in certain single-word 
items, but at least occasionally as used in certain patterns only (as in a catalog of ___ referred 
to above). Secondly, one can find similar semantic features even in certain affixes such as the 
adjective-forming suffix -ridden, which refers to ‘something unpleasant or harmful’, as in crime-
ridden and guilt-ridden (Rundell 2007, 1280), or mosquito-ridden and disease-ridden (Summers 
2005, 1414). Also, the suffix -ish can be, inter alia, ‘the ending of some adjectives that show 
disapproval: selfish’ (Summers 2005) or clownish and mannish. #ird, semantic prosody is clearly 
present also in multiword sequences only, including those in which it is not to be found in the 
individual items making them up, e.g. day after day and day in day out mean not only ‘continuously 
for a long time’ but also ‘in a way that is annoying or boring’ (Summers 2005). Note, incidentally, 
that there seems to be no trace of such prosody in e.g. from day to day or day by day. Perhaps 
straddling the single-word and multiword-item situations, a look at the ways in which the concept 
of (COMPLETELY) FULL OF can be expressed in English, as shown in works such as Longman 
Language Activator
™
 (Summers 1993) and the online-only WordNet, reveals a number of options 
some of which display elements of semantic prosody (in the first place rife with, overrun with/by, 
and fraught with), but in this particular case (also) on the level of grammatical collocations: 
- (a tree trunk) alive with (ants) {´full of people, animals, or things that are moving´}
- (roofs) bristling with (antennas)
- (a garden) abounds with (flowers)
- (a place) teeming with (theater-goers)
- (a house) overflowing with (guests)
- (slums) rife with (crime)
- (a book) rich in (ideas)
- (air) thick with (snow)
- (an area) overrun with/by (locusts)
- (an incident) fraught with (danger)
- (silence) pregnant with (suspense)
- (a desk) flooded with (applications)
- (a book) replete with (diagrams)
- (a museum) swarming with (tourists)
- (a person) brimming [over] with (confidence)
- (a person) brimful of (ambition).
Admittedly, it is difficult to prove that in such cases semantic prosody is to some extent dependent 
also on the preposition following the adjectival or participial head. In fact, in such cases prosodies 
would rather seem to be part of the lexical meaning of the adjectival/participial head, and thus be 
regarded as connotation, meaning that the noun following the preposition is selected simply in 
logical terms, in line with (the restrictions imposed by) its meaning. Indeed, this type of analysis 
can be done also with single-word near-synonyms, such as to persist and to persevere, which may 
have similar cognitive meanings, but widely different prosodic behavior (Partington 1998, 77). 

It is quite hard even to suggest with any precision where the starting point, as it were, of this 
elusive semantic phenomenon might be located, that is, whether semantic prosody is to be 
regarded as a – largely context-free – feature arising either from 
a) the single-word lexical item itself or even an affix, 
b) its collocations, 
c) idiomatic combinations formed with the item in question,
d) its wider patterns/patterning, 
e) its typical contexts, 
or indeed from some conceivable combination of these factors. 
Different cases seem to call for different interpretations. #us e.g. never in all my life must be 
selected as a fixed expression whenever we wish to emphasize how bad something was (Summers 
2005, 1104). #is suggests that some prosodies only come out into the open, as it were, when a 
“neutral” item is used phraseologically. A similar phenomenon, by the way, exists in a grammar-
based framework: the Cambridge Grammar (http://www.cambridge.org/elt/cge/cge/cambridge_
international_corpus.asp) informs us that the passive voice with the verb to get is used much 
more often to convey ‘bad news’ than ‘good news’ (e.g. He got arrested. | We got charged 20 
pounds too much.).
12
 In any case, one can also find cases where the neutrality of the base item is 
at least questionable, as in to cause – and to some extent its synonym, to spark (off), although the 
researchers do not appear to have noticed the latter one, possibly because its negative prosody is 
not so absolute –, where the verb itself seems to trigger off something negative following: what 
something sparks (off) is characteristically a debate, an argument, fighting, riots, protests, or problems. 
No context needed really. In a similar vein, Mikhail (1994, 333−7) seems to be quite at ease in 
offering his lists of “good personal qualities” and “bad personal qualities”: #e good ones in e, for 
instance, are earnest, easy-going, effervescent, efficient, effusive, elegant, eloquent, eminent, emulous, 
energetic, enterprising, equable, equanimous, even-minded, even-tempered, expansive, experienced, 
and extrovert. Here are the bad ones beginning in e: edgy, egoistic, egotistic, envious, erotic [sic], 
evasive, evil, evil-minded, excitable, explosive, extravagant, extremist. While not everybody is likely 
to agree with the essential “goodness” or “badness” of the items, many do manage to perfectly 
connote either of the two on their own, without any supporting co(n)text or collocation. 
Whichever way you look at it, it is a fact that “the meaning of a word can often be illuminated by 
the other words which it tends to co-occur with” (Wierzbicka 1987, 21), so that e.g. comparing 
the adverbs which the verbs rebuke, reprimand and reprove tend to co-occur with, will yield 
important clues as to the semantic differences between them: rebuking tends to be done sharply 
whereas reprimanding tends to be done severely; only reproving can be done gently but cannot 
be done sharply, severely not being excluded though it is less likely to co-occur with reprove 
than with reprimand. #ese differences in co-occurrence support the following differences in 


the semantic formulae: While all three verbs refer to some “bad” behavior by the addressee, 
only rebuke contains the component ´I feel something bad towards you because of that´; hence 
the “sharpness” of a rebuke. Reprimand is official, not personal, and so its definition refers to a 
category of people subordinate to the speaker; moreover, it is meant to constitute in itself a kind 
of punishment – and punishments can always be severe (though not sharp or gentle). Reprove does 
not imply “bad feelings” toward the addressee, and its purpose is purely didactic, corrective, not 
punitive; there is no reason, therefore, why it should not be able to be done gently. Note that 
Wierzbicka’s analysis implies a semantically identifiable and stable single-word type of item itself 
associated with a prosodic value, which seems to be then only REFLECTED in the collocations 
on logico-semantic grounds.
 #e concept of semantic prosody has been criticized recently (Whitsitt 2005) as being unconvincing 
– and not only because it has been defined in at least three distinctly different ways that remain 
largely undiscussed. And there is at least one valiant attempt that was made not long ago to place 
it on an objective rather than subjective footing (Dilts and Newman 2006).
13
 
However that may be, a balanced view of this intriguing phenomenon can be found in 
Bartsch’s (2004, 156−8) analysis of semantic prosodies in collocations. Here are the results of 
her corpus-based study of adverb collocates of communication verbs and their positive (=P) or 
negative (=N) prosodies (ibid., p. 157):
categorically claim; assert; state N (strong rejection)
coldly enquire; query N (without emotion)
flatly reject; deny; state
N (complete rejection  
or blunt statement)
fluently speak; communicating; cajoling P (with great ease)
highly acclaimed; rated P (approbation)
strictly speaking P (stringent)
widely
acclaimed; recognized; 
acknowledged; reported
P (approbation of an event  
or achievement)
argue Cogently P (convincing)
declare Ruefully P (repentance)
talk incessantly
N (stretching the patience  
of the listener)
deny strenuously N (leaving doubt)


Sadly, dictionaries are far from being equally successful in capturing this rather elusive semantic 
feature. Likewise, they do not always address the issue in similar terms, and sometimes they even 
flatly disagree about the type of semantic prosody involved, or indeed about its very (non)existence 
– witness e.g. the treatment of the verb to glint in the sense that collocates with eyes functioning as 
the subject in some of the leading monolingual learners’ dictionaries
14
 of English:
Macmillan (Rundell 2007)
‘if someone’s eyes glint, they show  
a strong emotion such as anger’
Longman (Summers 2005)
‘if your eyes glint, they shine and  
show an unfriendly feeling’
1
Cambridge (Walter 2005)
‘when someone’s eyes glint, they look bright, expressing 
a lively emotion’: She smiled at him, her eyes glinting 
with mischief 
Oxford (Wehmeier 2005)
‘if a person’s eyes glint with a particular emotion, or an 
emotion glints in a person’s eyes, the person shows that 
emotion, which is usually a strong one’: Her eyes glinted 
angrily. ◊ Hostility glinted in his eyes.
By contrast, most of the leading native-speaker-oriented English dictionaries treat this sense – 
quite logically, at least to some, given that semantico-prosodic information typically forms part 
of native-speaker linguistic competence – much more concisely; witness e.g. the 2000-odd-
page American Heritage Dictionary (Pickett 2000), which defines the verb to glint simply as ‘to 
gleam or flash briefly’. One of them, however, namely the Oxford Encyclopedic, contradicts the 
learners’ dictionaries (as does, as is observed in footnote 10, an online version of the [learners’!] 
Collins COBUILD) as to the type of prosody involved: 
Collins (Butterfield 2003) ‘to gleam or cause to gleam brightly’
Merriam-Webster (Mish 2003) ‘to look quickly or briefly : GLANCE’
Oxford Encyclopedic (Hawkins and Allen 1991)
‘flash or cause to flash; glitter; sparkle; 
reflect’ (eyes glinted with amusement)
New Oxford (Pearsall 1998)
(of a person’s eyes) ‘shine with a particular 
emotion’: his eyes glinted angrily
Can semantic prosody be detected in compounds too? #ere is no reason in principle why 
it should not be, though a single example must suffice at this point: While the noun flame 
would seem to have almost none, Paul McFedries’s Word Spy web page
15
 contains, inter alia, 
the following entry: 


dictionary flame noun. A negatively-charged message that complains about a person’s spelling 
mistakes, word usage, or grammar.
Admittedly, perhaps this example is not overly typical. Nevertheless, and in a more serious vein, 
phrasal items can also show this semantic feature: If you put somebody through something, it has to 
do (but only if the entire sequence is used!) with making someone do or experience something 
difficult or unpleasant (Summers 2005, 1337). Likewise, much-vaunted seems to have a negative 
prosody on its own, meaning as it does ‘[of a plan, achievement, etc.] one that people say is very 
good or important, especially when this may not be true’ (Summers 2005, 1079), and so does, say, 
self-indulgent, ‘allowing yourself to have or do things that you enjoy but do not need, especially if 
you do this too often’ – used to show disapproval (ibid., p. 1489). Similarly, to fragment is not only 
defined (‘to break something, or be broken into a lot of small separate parts’); after a dash, there is 
also the comment used to show disapproval (Summers 2005, 639). Indeed, an earlier edition of the 
Longman (viz. Summers 1995, but not the current [4
th
] edition, Summers 2003/2005) treats the 
adjective utter in a “fully prosodic-collocational” manner: #ere is no customary decontextualized 
definition at all, but rather the sequence utter failure/rubbish/fool etc followed by the definition 
‘a complete failure etc’ indicating the semantic restriction of the common English pattern utter 
+ a “negative” noun, i.e. one indicating something undesirable.
16
 Finally, if something breaks out, 
and sets in, it just has to be something bad; likewise, in [of somebody] to be in for something, that 
something just has to be unpleasant – but again only in that particular combination.
On the other hand, a recent thesaurus (Jellis 2002) provides, inter alia, a “compare-and-contrast” 
boxed feature, where a selection of semantically related single-word items sharing a “core meaning” 
are given and their semantic specificities briefly discussed. #us e.g. the article on the “core 
meaning” TALKING A LOT (p. 893) first provides the entries (talkative, chatty, gossipy, garrulous, 
loquacious), and then goes on to briefly indicate the semantic differences between them:
talkative willing to talk readily and at length; chatty talking freely about unimportant things 
in a friendly way; gossipy talking with relish about other people and their lives, often unkindly 
and maliciously; garrulous excessively or pointlessly talkative; loquacious (formal) tending to 
talk a great deal.
Gossipy and garrulous are the likeliest candidates for negative prosody; moreover, in line with 
Wierzbicka’s treatment (cf. final section of 5.3), in many cases it appears to arise largely from the 
core meaning of the single-word item itself. #us a similar boxed feature (also p. 893) comprising 
the items talent, gift, aptitude, flair, bent, knack, genius, illustrating the core meaning of THE 
NATURAL ABILITY TO DO SOMETHING WELL, shows no trace – clearly on account of 
the core meaning itself – of negative prosody rearing its ugly head. Should these items, then, be 
regarded as all carrying positive prosody? Not likely: for isn’t the positive aspect of meaning rather 
(part of) denotation? After all, we need no collocation, phraseology, or co(n)text to establish it.


It is not uncommon for dictionaries to ignore this “added” semantic element, so that most 
dictionaries define, say, to put an end to something routinely as ‘to finish something,’ while few 
elaborate on that prosodically: ‘to stop an activity that is harmful or unacceptable.’ In general, 
learners’ dictionaries fare much better in this respect than do native-speaker-oriented works. 
#is, by the way, applies also to single-word items, so that e.g. Longman (Summers 2005), unlike 
most native-speaker dictionaries, notes that terse – unlike brief or concise – not only means ‘using 
very few words’ but also ‘often shows that you are annoyed.’ 
In any case, matters prosodic are not always interpretable in absolute terms: Aside from the 
conflicting dictionary evidence given above, I can offer some pertinent evidence coming from the 
Internet (Netscape, spotted on 21 October 2004), about actor Christopher Reeve’s widow saying 
that “it is COMPLETELY unfair, but life can be that way” (my emphasis). Similarly, Schmitt 
and Carter (2004, 8) report that bordering on carries the “prosodic” meaning of ‘approaching an 
undesirable state (of mind)’ in 57 instances of the 100 instances in the British National Corpus, 
whereas 27 instances refer to a physical location.
17
 It is also used to express positive evaluation, but 
only in nine instances out of the 100 (ibid., 20). However, other instances of semantic prosodies 
are easier to capture in absolute terms, e.g. to undergo and to experience in a 1990s version of the 
Longman: you typically undergo a change or something bad but experience an emotion, physical 
sensation, a situation, or a problem (Summers 1995). 
 Apparently, words can have a specific profile, either good and pleasant or bad and unpleasant; 
whenever such a word is uttered, it prompts hearers to expect a following word with a clear 
(un)pleasant sense – it “sets the scene” for a particular type of subsequent item. Yet it is often difficult 
to show convincingly the role of collocation in the creation and continued existence of semantic 
prosody. What linguists frequently refer to as connotation, or connotative meaning,
18
 is, after all, 
to be found in many single-word items and phrasal verbs (e.g. compact, lean, slim, lanky, skinny, 
notorious, to drone on, to show off).
19
 Second, one might wonder whether prosody is equally at work 
in colligational combinations, such as in league with?
20
 Also, is its creation related to onomatopoeia? 
Moreover, and perhaps more to the point, connotation itself is not quite as clearcut as one might 


suppose: It is defined in the standard sources as “an additional meaning” that is indicative chiefly 
of “emotional associations (personal or communal)” a lexical item has beyond its central meaning, 
usually referred to as denotation (e.g. Richards and Schmidt 2002, Crystal 2003). But does this 
really DEFINE the concept? For example, if December and child, selected in the two reference works 
just cited as illustrations of connotation, are good examples of items with connotative meaning, 
how is it that they are not labeled as such in any of the English dictionaries? Second, if the nouns 
argument and quarrel are connotation-free near-synonyms, is the closely related feud different in 
that it exhibits negative prosody, or is it rather that its special semantic features (“long,” “violent”), 
and/or its frequent use in the collocation a bitter feud and in the compound blood feud, show its 
intensity or duration without contributing anything semantically negative?
 More broadly, as Stubbs (1995a; cited in Schmitt 2000, 78−9) points out, words may habitually 
collocate with other words from a definable semantic set. #e words in these sets may carry 
either positive or negative connotations: e.g. to cause typically collocates with unpleasant 
things such as problems, trouble, damage, death, pain, disease, whereas to provide collocates 
mainly with positive things such as facilities, information, services, aid, assistance, money. 
Using work with the two words provides further illustration of the difference: to cause work is 
usually considered a bad thing, while to provide work is usually looked upon favorably. When 
examining items such as cause and provide, one is certainly tempted to conclude that semantic 
prosody
21
 indeed comes close to the more traditional semantic notion of connotation referred 
to in the preceding paragraph. Further, if e.g. to credit is “neutral” in semantic terms, to credit 
somebody with (doing) something invariably refers to something good and is thus “inherently” 
positive. In traditional terms, however, connotation has received attention almost exclusively 
as an element of the lexical meaning of individual single-word lexical items.
22
 
 Semantic prosody
23
 seems to be related to yet another recent notion, namely that of pattern. 


#us e.g. two patterns of the verb to claim, viz. claim + that-clause (e.g. He claims that he has 
discovered the ideal rock band) and claim + object noun (e.g. Several nations now claim linguistic 
independence), may be indicative of two different prosodies, the former rather negative (‘he may 
say so, but it is likely not true’), the latter neutral (‘they are calling for linguistic independence’). 
Moreover, it is difficult to determine whether prosodic meaning is to be related broadly to 
any string where it seems to apply, or more narrowly to collocation and perhaps some other 
phraseological word combinations, and indeed, as a consequence, whether it arises from 
(a) an item’s lexical meaning, in which case it can be regarded as being virtually synonymous 
with connotative meaning
(b) various types of phraseological units, notably collocations, or indeed 
(c) context (or rather co-text) in general terms, perhaps considered in a kind of pragmatic-
textual extralinguistic/experiential framework. 
 Even more to the point, even one and the same item is not always necessarily associated with a 
certain type of semantic prosody; that is, one and the same item can be prosodically different: 
Warren (2005) observes that to look forward to a meeting has a positive prosody, not because of 
the noun meeting, which is evaluatively neutral, but because “as a complement of look forward to 
a positive feature is coerced.” However, she adds that these constraints can be cancelled: It is e.g. 
possible to modify look forward to with the adverbial with mixed feelings, yielding Peter is looking 
forward to the meeting with mixed feelings, which brings about a change of the interpretation of 
meeting (ibid.). Next, Partington (1998, 77, citing Louw 1993, 171) observes that even different 
forms of the same lexical item may display different prosodic behavior: “to build up confidence” 
(transitive) is favorable, while “resistance builds up” (intransitive) is unfavorable. Hoey (2003), 
too, provides an example of the way that semantic association, defined as the tendency of a 
word to keep company with a semantic set or class (some members of this set or class will usually 
be collocates), works. Using consequence as an example, he notes that it has semantic associations 
with concepts of logic, with (un)expectedness, with negative evaluation, and with markers of 
(in)significance. Of the four, only the third is related to semantic prosody:
a] logic: unavoidable, inevitable, inexorable, inescapable, ineluctable, direct, ultimate, long-
term, immediate
b] (un)expectedness: likely, possible, probable, natural, unintended, odd, strange, planned-for
c] negative evaluation: awful, dire, appalling, sad
d] significance: serious, important, dramatic, enduring, prominent.

Many – but not all – of the adjectives in these semantic categories are also collocates of consequence, 
concludes Hoey (2003).
Not all lexical items display highly regular prosodies such as to set in or to peddle. #us Partington 
(1998, 72) refers to Sinclair’s work where the word happen is shown to have a general tendency 
to collocate with unpleasant events, but this characteristic is not binding, as happen occasionally 
collocates with neutral or even pleasant occurrences. 
Finally, semantic prosody remains a “contentious term”: Many writers use it to refer to the implied 
attitudinal meaning of a word, whereas Sinclair uses it to refer to the discourse function of a 
unit of meaning (Hunston 2007, 249). Moreover, there seems to be something of a foundational 
controversy here: While Sinclair (1996) refers to semantic prosody as the outcome of all the 
choices that a speaker or writer makes, Hoey (2005, 163) envisages the complete opposite of this 
– the initial impulsion to inform, contradict, praise, etc. If the semantic prosody matches the 
original intention, presumably the speaker/writer is satisfied. 
Let us, by way of conclusion, note briefly that semantic prosody is considered by linguists of the 
“narrower semantico-prosodic persuasion” to be only one of two kinds of semantic relationships 
obtaining between collocates, the other being semantic feature copying, which is the tendency of 
e.g. adjectives to collocate with nouns that they share a semantic feature with, e.g. [PHYSICAL] 
in physical attack or [SCIENTIFIC] in scientific study/experiment (Bublitz 1996, 6−10). In any 
case, many combinations are not easy to analyze in such terms, due largely to the interplay of 
semantics and combinability/usage factors. 
 Is semantic prosody real and worthy of scholarly interest? Definitely. And in language teaching? 
Another yes, but with a proviso: it is a part of advanced-level L2 skills. In any case, it is significant, 
being a component contributing to the overall meaning of certain multiword units. Yet there are 
(still) vexing questions: Is it really (and necessarily so) – and to what extent – collocational, or 
indeed more broadly phraseological, in nature, and does it really stand for a type of meaning that 
has “a life of its own,” existing over and above the “basic” denotative (also known as referential, 
conceptual, or cognitive) meaning of an item that one can actually dissociate it from? #is one 
– doubtless a key issue – is more difficult to answer in absolute terms. T o begin with, Hoey 
(2005, 23) points out that the claim made for semantic prosody that words are colored by 
their characteristic surroundings has been challenged. #us it is hardly surprising that semantic 
prosody has been associated 
(1) more or less absolutely with single-word items considered out of context for their 
attitudinal meaning existing besides denotation (e.g. utter, to cause, to provide), 
(2) sometimes with multiword items only (e.g. to break out, to cause uncertainty/concern, to 
go in for something, day after day), whose constituents on their own either do or do not 
contain an element of attitudinal meaning, or 

(3) occasionally more vaguely with a lexical item in co(n)text. In line with this view, the 
translator Taylor (1998, 326), for one, defines semantic prosody as ‘the semantic content 
or force of a lexical item in a given context.’ 
Furthermore, on another level, semantic prosody can be a kind of “additional,” “separable” 
meaning, that is, one existing separate from the denotation (e.g. to credit somebody with [doing] 
something, to put an end to something, to fragment), and yet in other cases this does not seem to be 
the case (much-vaunted, self-indulgent). But can this be determined in a foolproof manner? Hardly. 
However this may be, the analysis can be also carried out in a broader-based if vaguer fashion: 
#us e.g. the very term semantic prosody can be used to acknowledge the fact that the habitual 
collocates of the core unit of meaning naked eye – more specifically, their semantic constituency 
– are capable of coloring it; in this case, the coselection of the unit with verbs/adjectives related 
to the notion of ‘visibility’ activates a semantic prosody which suggests DIFFICULTY (González 
and Macarro 2000, 107). 
Significantly, semantic prosody does not always exist as a straightforward binary yes-no affair, 
witness e.g. to get as used in the passive, or the conflicting dictionary pronouncements about the 
emotion conveyed when one’s eyes glint. Moreover, it is unclear when exactly it exists over and 
above the denotation in a way that makes it “separable,” and when it is an inseparable part of 
the denotation. Furthermore, and more theoretically, one might wonder whether the existence 
of semantic prosodies in phrasal items is a signal that such units should be viewed as holistic 
units precisely because of that, or also because of that? Next, when and under what conditions 
can semantic prosody affect wider stretches of text? And finally, the situation is just as unclear 
when it comes to considering the provenance and “direction”: is semantic prosody a feature of a 
single-word item (connotation) that extends to its surroundings, in the process coloring some of 
the items in its vicinity, or is it the other way around (semantic prosody proper)? #e evidence 
suggests that the pendulum can swing both ways: cause and provide, for example, lend themselves 
to the single-word-item-spreading-its-influence type of analysis, while e.g. day after day and to 
put somebody through something do not. 
It seems wise to try to break down the topic into smaller and more manageable components, 
and thus to restrict oneself to concentrating on 
a) connotation as the “additional” meaning of single-word items, identifiable as their 
semantic property, whether or not considered in a given context
b) semantic prosody as the semantic property arising syntagmatically from certain multiword 
items (typically but not exclusively collocations
24
), where a distinction should be made 
between two subtypes: those where the prosodic feature is present in a constituent (to 
cause a fire) and those where it is not (day after day), and 


c) syntagmatic meaning as a semantic feature only brought into being in a given co(n)text, 
only observable in a specific string. #us Taylor (1998, 85) points out that in the utterance 
I would stay clear of that rat Jones!, “the listener would not conjure up an image of a rodent 
interlocutor, but of a mutual acquaintance reputed, at least by the speaker, to have a 
deceitful or vindictive nature.” 
Aside from varying the terminology (such as by introducing discourse prosody instead of 
syntagmatic meaning, and collocational prosody as a special subcategory; using semantic prosody 
as the cover term, as I have done in some parts of this survey paper), one might find it wise 
to try to come up with a catchall term for the three. Associative meaning? Or perhaps it is 
only the last two – semantic prosody and syntagmatic meaning – that are in need of a cover 
designation, in which case discoursal meaning could well fit the bill. Secondly, is this kind 
of meaning always necessarily negative or positive?
25
 Intuitively, one would be tempted to say 
no – but then the binary, dichotomous treatment (good or bad, left or right, rich or poor, big 
or small, and all the rest) seems to be ever so close to the human mind, and ever so efficient, 
especially when the categories involved are comfortably broad, as in this case. 
If anything, semantic prosody is a significant if complex phenomenon, showing that denotative 
and associative levels of lexical meaning are both real and diverse while being also strongly 
interactive, and that in syntagmatic terms the extent of semantic prosody (the stretch of text 
influenced by it) varies considerably. It is, then, only logical that it should also exist as such, and 
be treated as such. 






/