Phraseology and Culture in English

12. Conclusion
In his book Words and Phrases to which I referred at the outset, Michael 
Stubbs (2001: 215–216) makes, inter alia, the following points: 
ʊ
Evaluative meanings are conveyed not only by individual words, but 
also by longer phrases and syntactic structures … . Repeated instances 
of a collocation across a corpus provide objective, empirical evidence 
for evaluative meanings. 
ʊ
Repeated patterns show that evaluative meanings are not merely per-
sonal and idiosyncratic, but widely shared in a discourse community. 
ʊ
Evaluative and attitudinal meanings are often thought to be due to con-
versational inferences (Grice 1975); however, many pragmatic meanings 
are conventionally associated with lexico-syntactic structures. 
ʊ
The over-emphasis on conversational inferences is probably due to a 
reliance on invented data, which have been stripped of markers of 
speaker attitude. 
I believe that these are all valid and constructive points. (For an
extended discussion of the misplaced faith in Gricean inferences, see 
Wierzbicka 2003b). Unfortunately, on the heels of these good points
come two further ones which in my view are neither valid nor construct 
ive:
ʊ
Semantics is inevitably circular, since some words are used to define 
the meanings of other words, and it is not yet clear what metalanguage 
should be used to describe evaluations. 
ʊ
If descriptions of evaluative meanings and cultural stereotypes can be 
made reasonably precise, it seems plausible that this will tell us about 
the important meanings expressed in a discourse community, but specu-
lations here are at an early stage. 
If semantics were “inevitably circular”, meanings could not be analysed 
in an illuminating way and the study of meaning would be doomed to fail-
ure. In the past, many linguists have thought that it is indeed doomed to 
failure and that it is best for linguists to stay away from it. 

---

Reasonably well
75
This is not Stubbs’ stance. On the contrary, the opening sentence of this 
book reads: “The topic of this book is words and phrases: how they are 
used, what they mean, and what evidence and methods can be used to study 
their meanings.” This is certainly a much more positive attitude to seman-
tics than that expressed, for example, in John Sinclair’s (1991) book Cor-
pus, Concordance, Collocation, where one reads, for example, statements 
like the following: 
Working with lexicographers made it clear that there are no objective criteria 
available for the analysis of meaning … More recently, the whole idea of 
discrete units of meaning is called into question. At the moment, the mood 
is somewhat negative. (p. 7) 
There is a broad general tendency for frequent words, or frequent senses of 
words, to have less of a clear and independent meaning than less frequent 
words or senses. These meanings of frequent words are difficult to identify 
and explain; and with the very frequent words, we are reduced to talking 
about uses rather than meanings. (p. 113) 
But if we didn’t have “discrete units of meaning” to work with, we could 
identify neither the different senses of the English word reasonable nor the 
differences between the English reasonable (and reasonably) and the French 
raisonnable (and raisonnablement); and we could not reveal the cultural 
underpinnings of these English key words. 
What lexicographers need is something more constructive than state-
ments along the lines of “You are right, there are no objective criteria for 
the analysis of meaning”; and it is good to see that Stubbs (2001), for one, 
rejects such an anti-semantic stand, and asks what methods can be used to 
study meanings effectively. 
Stubbs says that “it is not yet clear what metalanguage should be used to 
describe evaluations”. As I have tried to show in this paper, and as colleagues 
and I have tried to demonstrate in many other publications (see NSM Home-
page http://www.une.edu.au/arts/LCL/disciplines/linguistics/nsmpage.htm), 
evaluative meanings, like any other meanings, can be analysed very effec-
tively in the Natural Semantic Metalanguage, based on empirically estab-
lished universal human concepts such as, for example, 
GOOD
and 
BAD
.
Stubbs suggests that “speculations” about the important evaluative mean-
ings expressed in words and phrases “are at an early stage”. Although NSM 
researchers have tried to identify such meanings for years, and have pub-
lished scores of books and articles with that aim in mind, we do, of course, 
have a long way to go. But the hundreds of specific semantic analyses pre-

---

76
Anna Wierzbicka 
sented in this body of NSM work are not speculations; and while no doubt 
far from perfect, they are certainly being tested on empirical material. 
The availability of corpora has opened new perspectives for linguistics 
and has stimulated interest in the study of words and phrases, a study which 
was totally marginalised during the Chomskyan era. But the wonderful data 
available now through the corpora will not by themselves generate their 
own semantic analysis. Nor will such semantic analysis be generated by 
abstract semantic theories such as, for example, those associated with the 
term “formal semantics”. A practical and effective methodology for study-
ing the meaning of both words and phrases (or indeed, of any other aspect 
of language) is already available in NSM semantics. Using this methodol-
ogy, hundreds of meanings have already been described – not perfectly,
to be sure, but systematically, empirically, and, dare I say it, reasonably 
well.