Phraseology and Culture in English

partners while the socio-cultural category refers to casual conversations be-
tween friends. Interactions recorded at the workplace have been classified 
as professional and any type of “service encounter” was assigned to the 
transactional category. The pedagogic category sits somewhat uncomforta-
bly with the others but it still describes a particular context which is per-
haps more easily defined in terms of the contextual goal of the interaction 
than in terms of the relationship between the speakers. The CANCODE 

---

Definitely maybe: Modality clusters and politeness in spoken discourse 
261
categories tend to be mutually exclusive, but there is necessarily some de-
gree of overlap and embedding when we try to label any process as com-
plex as that of human interaction.
Nevertheless, the careful contextual annotation system that has been de-
veloped for CANCODE facilitates the extraction of patterns of modality in 
a set of different spoken contexts on a somewhat larger scale than many 
previous studies on politeness phenomena have been able to do. A concor-
dance search of modality markers in a large spoken corpus reveals not only 
information about the use of individual markers in different contexts, but 
also allows for an analysis of patterns in their co-text. This makes it possi-
ble to analyse additional properties of the modal marker, such as colloca-
tion, patterns of syntactic integration and semantic prosody (Sinclair 1996).
5. Identifying modality clusters 
A range of approaches have been proposed which aim to define the form 
and function of different types of clusters or recurrent sequences in language 
use (Aijmer 1996; Manes and Wolfson 1981; Moon 1998). The frameworks 
that have been developed to describe the nature of such sequences vary 
widely, both in the definition of a unit and in the methodology used to iden-
tify a specific unit (Read and Nation 2004). Some studies have relied heav-
ily on intuition in this process, sometimes accompanied by prior or subse-
quent corpus research (Nattinger and DeCarrico 1992). Others have used 
the criterion of frequency as the main starting point and produced lists of 
sequences according to their frequency ranks in a given corpus (Biber et al. 
1999). Wray (2002: 9) defines a formulaic sequence as a sequence that is 
“prefabricated” and that is “stored and retrieved whole from memory at the 
time of use, rather than being subject to generation or analysis by the lan-
guage grammar.” Similarly, Moon (1997: 44) argues that institutionalization, 
fixedness, and non-compositionality are the main criteria of what she calls 
multi-word items. 
The starting point of this paper has been the observation that a cluster of 
two consecutively used modal items can carry a particular function and 
meaning that is different to the meaning of the individual parts of this clus-
ter when used in isolation. Prime candidates for this type of exploration 
could be identified intuitively and would probably include sequences such 
as might well, couldn’t possibly, or definitely maybe. However, for the pur-
pose of this study, a couple of two-word modality sequences have been 

---

262
Svenja Adolphs
identified through reading a randomly selected stretch of spoken discourse 
in the CANCODE corpus: might just and could possibly. These two-word 
modality sequences form the basis for this study. 
5.1. Might just and Could possibly
Previous research on modal items and their function in discourse has rec-
ognised the co-occurrence of certain modal items (Blum-Kulka et al. 1989; 
Trosborg 1995), but has not tended to discuss these as sequences in their 
own right. Instead, they have been seen as a re-enforcement of a certain type 
of modality and discussed in terms of the accumulative hedging effect they 
have as part of a given speech act. 
Rather than describing these clusters as a random sequence of modal 
markers, the present study aims to explore whether or not particular modal 
combinations carry an additional meaning in their own right. The clusters 
identified above will thus be analysed in terms of their individual compo-
nents, as well as in terms of their properties as a sequence.
5.1.1. Might just 
While the word just can function as both an adjective and an adverb, it oc-
curs most frequently as an adverb and is more common in spoken English 
than in written. In the CANCODE corpus the item just occurs 29,020 times 
and is the 32nd most frequent item, accounting for 0.6% for the entire cor-
pus. In this corpus, just is used to express the following non-compositional 
meanings, listed by Carter and McCarthy (in press): simply often in declara-
tive clauses and often introduced by it’s, only and a very short time ago, 
recently. From a functional perspective, just is often used to minimise im-
positions and to downtone statements. Common collocates of the word just
are think, wonder and want which further reinforce this function and tend to 
combine with personal pronouns and the word just to form a particular type 
of modality cluster, such as ‘I was/am just wondering + conditional clause’. 
In terms of two word modality clusters, Carter and McCarthy point out that 
“the regular pattern of just with modals (usually epistemic modals) also 
shows how just helps a writer or (usually) a speaker to sound more tentative 
and polite.” They give the following example, which also includes the item 
might taken from the CANCODE corpus: I might just turn up at the party 

---

Definitely maybe: Modality clusters and politeness in spoken discourse 
263
after all. However, it seems that the cluster might just, in this example, adds 
a meaning other than that of increased politeness. In this particular instance 
a degree of certainty and determination is added to the utterance which is in 
keeping with the phrase “after all” at the end of the utterance. 
The modal item might is generally used as a downtoner in the CAN-
CODE corpus. It occurs 4,091 times and accounts for 0.09% of the overall 
corpus. It tends to be used as an epistemic modal item which hedges the 
speaker’s commitment towards the truth of a proposition, as in the follow-
ing utterance: “that might have something to do with it”. As a modal auxil-
iary verb, the item might often follows a personal pronoun. Common clus-
ters include: “I might have..” (140 times), “you might have…” (80 times), 
“he might have…” (47 times). These clusters are often followed by past tense 
verb phrases and denote uncertainty with regard to a past event (e.g. “I 
think I might have told you this one”). 
The cluster might just occurs 94 times in the CANCODE corpus while 
the combination just might only occurs 9 times. 
In 59 instances might just is preceded by a personal pronoun: I (28 
times), they (11), we (7), you (6), she (5), he (1), one (1). In a further 5 
instances there is an ellipsis of the personal pronoun, as in the following 
example: 
Now you can sit down if you want. 
Oh yeah. Might just do that actually. 
In the majority of instances might just denotes certainty about a decision, 
rather than fulfilling a minimising or downtoning function. The following 
example illustrates this particular meaning: 
Are you going anywhere? 
No. No well I was gonna up to Glossop. I was gonna go camping or 
something but since the weather’s so bad I think I might just hang 
round here and
Bake. 
Mm? 
Bake lots of nice cakes. 
Often the certainty is linked to a juxtaposition of different options, which 
tend to be made explicit by the speaker. The following concordance lines 
illustrate this tendency: 

---

264
Svenja Adolphs
There’s more more room I might just leave it like this and hope that 
They’re going out for a drink tonight so I might just pop out and see them 
em thirdly if I don’t get a job I think I might just go part-time at work and go back to 
year because I don’t really use it. Or I might just get it fixed. But my bike