You are on page 1of 18

See discussions, stats, and author profiles for this publication at:

https://www.researchgate.net/publication/249940005

Chunking in ELF: Expressions


for managing interaction

Article in Intercultural Pragmatics · May 2009


DOI: 10.1515/IPRG.2009.012

CITATIONS READS

30 133

1 author:

Anna Mauranen
University of Helsinki
50 PUBLICATIONS 1,281 CITATIONS

SEE PROFILE

All content following this page was uploaded by Anna Mauranen on 03 June 2016.

The user has requested enhancement of the downloaded file.


Author's Copy

Chunking in ELF: Expressions for


managing interaction

ANNA MAURANEN

Abstract

This paper adopts a linear perspective on analyzing language, based on Sin-


clair and Mauranen’s Linear Unit Grammar (2006), and complemented
with insights on recurrent patterning discernible in corpus data. The focus
is on elements of managing interaction and the way ELF speakers utilize
them in co-constructing successful discourse. Interactive phraseological pat-
Author's Copy

Author's Copy
terns range from very short and fixed expressions up to variable units of
around five words; the longer and more variable patterns are more suscepti-
ble to unconventional forms. ELF speech manifests approximations of con-
ventional forms. These tend to be close enough to the target to ensure com-
prehensibility, but at the same time they deviate from conventions with a
measure of regularity which suggests emergent patterning rather than ran-
dom errors.

1. Introduction

One of the most salient features of speech is its real-time nature. This is
normally invoked every time we outline the main di¤erences between
speech and writing. It is nevertheless symptomatic of the study of lan-
guage that we tend, still, to describe the consequences of real-time pro-
cessing in a somehow apologetic manner, as if in need of explanation or
justification. Speech tends to get described in terms of deviations from the
written language, which at least tacitly sets the standard. So we talk about
‘‘dysfluencies’’, ‘‘hesitations’’, ‘‘dislocations’’ using these and many other
negative terms for phenomena that are common features of normal
speech (see, for example, Biber et al. 1999: 1066–68). In this way, the
temporal character of speech is treated as one of the unavoidable hard-
ships of life, which can account for certain less desirable manifestations
of behavior, but is not worth serious attention in itself.

6 Walter de Gruyter
Intercultural Pragmatics 6-2 (2009), 217–233 1612-295X/09/0006–0217
DOI 10.1515/IPRG.2009.012

Author's Copy
Author's Copy

218 Anna Mauranen

The propensity to view the consequences of online speech as deviations


from good or normal language has a particularly unfortunate e¤ect on
how non-native speech is seen and evaluated. Because it is obvious on any
account that second language user or learner speech is very di¤erent from
polished prose or well-rehearsed monologues, our sense of what normal
speaking is like as opposed to problematic speech is apt to get blurred.
What I set out to argue in this paper is that if we take the real-time
nature of language seriously, it means a substantial change of approach
to linguistic description: we must account for the fundamental linearity
of language. In speech, linearity manifests itself as temporality. It is also
important that robust models of language are not limited to describing
native speech only (a criticism that Firth leveled at conversation analysis
as early as 1996). I outline here some basic principles of a linear model
of language, following Sinclair and Mauranen (2006), and apply it to
English spoken as a lingua franca (ELF). I show that ELF speech is fun-
damentally normal language use despite some surface deviations from
Standard English. The particular focus is on linguistic means used in
managing interaction, which is crucial for co-constructing successful
Author's Copy

Author's Copy
conversation.
What I also want to show in this paper is that ELF speakers’ deviations
from preferred conventions of Standard English are not just random
errors or idiosyncrasies. Emergent patterning, whether in terms of phra-
seological units or new distributions of existing patterns, is discernible
across speakers, situations and first languages. Patterning is rooted deep
in our linguistic capacity, and allows for both convention and creativity.

2. A linear view of language

Temporality is not entirely cast aside in many approaches to spoken dis-


course, including conversation analysis and interactional linguistics. They
often emphasize the significance of sequentiality in discourse: what has
been said constrains what can be said next and what interpretations it
lays itself open to (see in particular Auer’s work, e.g., 2005). Despite this
important observation, a more principled problem remains in that most
descriptive models fail to take the real-time sequential character of speech
seriously on board. If we think of language coming in online to a hearer’s
attention, it unfolds in rapid succession, perhaps something like this:
(1a)
yeahimeanoneunderlyingtendencyforthisthisbigdilemmamaybemightbethe
thiskindofobsessioninprogressthatthemodernwestatleastithinkhas

Author's Copy
Author's Copy

Chunking in ELF: Expressions for managing interaction 219

Or, if we process it one word at a time, as some models suggest, the fol-
lowing would be more like our online experience:

(1b)
yeah i mean one underlying tendency for this this big dilemma may be
might be the this kind of obsession in progress that the modern west at
least i think has

However, neither of these representations are what is normally encoun-


tered in analyses of spoken discourse. What we are usually faced with is
a ‘‘synoptic’’ view of a speech extract, looking at it all in one go, and in-
evitably analyzing it with hindsight, interpreting what was said first as it
were in the light of what we know happened later. It is very hard to see
discourse unfolding if it is available to the analyst all at once.
Even if it has not been taken much notice of, linearity goes back a long
time in linguistics. It was already taken up by Saussure (1916), who out-
lined two fundamental properties of language: the arbitrariness of the
sign, and the linearity of language. The former has been discussed to ex-
Author's Copy

Author's Copy
haustion, the latter virtually ignored. An innovative opening on a linear
description of the syntax of speech was made by Brazil (1995), who set
out to describe the way in which speakers proceed in real time towards
their communicative goals by what he called increments. His point of
departure was that utterances make meaning as they unfold in time, not
after they are finished.
In the rest of this paper I take the linearity of language as my point of
departure, following a model that we recently developed with John Sin-
clair, the Linear Unit Grammar (Sinclair and Mauranen 2006), and apply
this to English as a lingua franca. We share with Brazil the sense that lin-
earity is central in understanding how language works, and we adopt his
term ‘‘increment’’ for elements that make contributions of certain kinds
to meaning. In certain other ways we depart from Brazil’s approach.
One important di¤erence is that our theoretical model is more general: it
covers both spoken and written text, and dialogues as well as mono-
logues.
In this paper my primary concern is to highlight some interesting
phenomena in spoken ELF, and therefore I shall only give a very brief
outline of the Linear Unit Grammar (LUG) at the outset. I invoke
just a few of its basic features, such as the notion of ‘‘chunking’’, and
some fundamental distinctions in our analytical framework. I then focus
in on a type of element where linearity meets phraseology, and finally
bring in a corpus perspective to explore the phraseological aspect more
closely.

Author's Copy
Author's Copy

220 Anna Mauranen

3. Chunking on line

It is in principle possible that people process incoming language one word


at a time, as in Example (1b). This has been the point of departure in two
recent models, which seek to depict syntax from a ‘‘dynamic’’ perspective
(Kempson et al. 2001; O’Grady 2005). However, most cognitive evidence
seems to point towards somewhat larger chunks as the basis of our lan-
guage processing—along the lines of general information processing. It
has been suggested that the units of processing correspond to units familiar
from most descriptive models of language, such as clauses (Ellis 1996) or
intonation units (Du Bois 2003), but often it seems rather that processing
takes place in unit fragments (Levelt 1989; Muhkerjee 2001) and are fairly
short, just two or three words (Dabrowska 2004). Clearly, what constrains
all our information processing is the working memory, which comprises
the contents of our attention at any given moment. Its capacity is roughly
up to five units, plus or minus two, as already suggested in Miller’s (1956)
classic work. Chopping up the incoming speech stream into chunks seems
an e‰cient way of coping with it, and in line with our other perceptual pro-
Author's Copy

Author's Copy
cesses; it is not a learned skill but what people do naturally. The assump-
tion is, then, that we process speech in chunks of up to about five words.
Instead of linearity, we could equally well talk about temporality in the
case of speech. In written text, linearity is spatial and less obvious, be-
cause in principle we can move back and forth in text. In its two manifes-
tations, linearity nevertheless captures an underlying similarity in speech
and writing.
An important consequence of strict adherence to the principle of linear-
ity in linguistic analysis is that it throws into sharp relief the evolving
nature of context: as we move along constructing text, we change the
context at every step. This is particularly obvious in the case of co-text,
but we could argue that it occasions subtle changes in the context of situ-
ation as well.
In LUG, we take chunking to be an intuitive perceptual response to the
incoming speech stream. There is no reason therefore to expect it to di¤er
in L1 and L2 speech in principle; it is nevertheless obvious that a speaker
needs some threshold command of the language they are hearing before
they can chunk it up in a useful way. We all have experiences of listening
to foreign languages which sound like an uninterrupted stream of non-
sense if we have no knowledge of them. LUG takes chunking as a pre-
theoretical term, and we use our own intuitive capacity to perform chunk-
ing on stretches of transcribed speech. The following extract from the
ELFA corpus illustrates this. In (2a) we see a continuous stretch with
word spaces inserted but ignoring speaker changes:

Author's Copy
Author's Copy

Chunking in ELF: Expressions for managing interaction 221

(2a)
but that would allow er people more time to do whatever yeah but i in my
point of view that was really good because @yeah@ yeah (xx) that we are
an open an open market nowadays
The first step is then to chunk it up by inserting provisional unit bounda-
ries (PUBs) where a break seems natural. It is important to note that the
intuitive skill that is tapped here pertains to the placing of boundaries, not
attending to what is inside them. I have done this in (2b). Although it is
possible that some readers might have placed the boundaries di¤erently,
our workshop experience with di¤erent groups suggests that the variation
is not substantial. More importantly, the findings of Cheng et al. suggest
that tone unit boundaries can be predicted with a high success rate em-
ploying the analytical framework of LUG (Cheng et al. 2008: 78), and,
following Warren (2006), that backchannelling behavior is also systemat-
ically related to certain types of units in LUG (Cheng et al.: 80). I there-
fore hope that the reader will accept this as a working solution for the
time being.
Author's Copy

Author's Copy
(2b)
but |that would allow |er |people | more time to do whatever |yeah |but |i
|in my point of view |that was really good |because |@yeah@ |yeah (xx)|
that| we are an open |an open market nowadays
Once the boundaries are in place, we are left with a sequence of chunks
bounded o¤ from each other. The internal properties of these chunks
have not received any attention. This is where intuitive work ends; the
next steps are systematic and impose a model for analyzing the chunks.
LUG distinguishes two main types of element that all chunks fall into:
first, those concerned with that which is being talked about, the ‘‘content’’
of the text as it were, or, in LUG terminology, ‘‘incrementing shared
experience’’ (that was really good; we are an open market nowadays).
These elements are primarily oriented to the ‘‘message’’ that is being co-
constructed in conversation, and called ‘‘M’’ for short. The second major
category of elements comprises those concerned with discourse manage-
ment (oh yeah; mm; but; and ). They are oriented to managing and organ-
izing the discourse and called ‘‘O’’.
LUG distinguishes several categories of M, but the present discussion
concentrates on the O elements, which perform important pragmatic
functions. Here we can make another binary distinction: elements
concerned with interaction (‘‘interactive organizational elements’’, OI),
okay; well; oh dear, and those those concerned with text (‘‘text-oriented
organizational elements’’, OT), for example, because, then. O elements

Author's Copy
Author's Copy

222 Anna Mauranen

tend to be short, mostly two-three words, and we assume each type con-
stitutes something like a closed class.
Turning back to our example again, the categories outlined so far look
like this (2c):
(2c)
1 S1: but OT
2 that would allow M
3 er OI
4 people M
5 more time to do whatever M
6 S6: yeah OI
7 but OT
8 i M
9 in my point of view OI
10 that was really good M
11 [because] OT
12 S3: [@yeah@] OI
Author's Copy

Author's Copy
13 S5: [yeah] OI
14 S6: (xx)
15 that OT
16 we are an open M
17 an open market M
18 nowadays M
The OT elements (lines 1, 7, 11, and 15) express textual relationships. The
first is but in line 1, which sets up a contrastive relationship between the
previous turn and the one about to start, and the but in line 7 accom-
plishes the same kind of job. Because (line 11) sets up a causal relation-
ship, prospecting a basis for the evaluation in line 10. The actual basis
was partly inaudible on the tape (line 14), but part of it follows in the
last three lines, ushered in by that in line 15.
The OIs begin in line 3 (er). This is a hesitation. It helps manage the
interaction by indicating that the speaker intends to keep the floor. In
line 6, yeah relates positively to the previous speaker, indicating that she
has been heard. Simultaneously it serves to mark speaker 6’s taking the
floor. In line 9, the speaker prospects an evaluation: the element (in my
point of view) indicates that what follows is to be taken as his personal
opinion or stance. In lines 12 and 13, speakers 3 and 5 speak at the same
time, also overlapping with S6’s causal OT.
Language items are often at least potentially multifunctional, as can be
seen in lines 1 and 7: in both cases, but clearly serves a textual function by
indicating that a contrasting viewpoint is to follow. But in the latter in-

Author's Copy
Author's Copy

Chunking in ELF: Expressions for managing interaction 223

stance, there is a strong interactive overtone: while yeah relates positively


to the previous speaker, but anticipates a challenge. This capacity that
elements have for flipping into a di¤erent role is part of the flexibility of
linguistic resources and their sensitivity to context.
The world of shared experience, which gets co-constructed by the par-
ticipants increment by increment (the M elements), is essentially a virtual
world. Interaction nevertheless takes place in the real world. To mediate
between the two, the OIs serve as links between the virtual world and the
‘‘real’’ circumstantial world of interaction. An OI thus looks ‘‘outward’’
to the circumstances in which the discourse takes place, and serves to
accomplish the utterance of the text. In this way, OIs play the role of
second-level organizing with respect to the M increments. Speakers use
OIs in managing and maneuvering interaction. Each OI is a conversa-
tional move: it fills silence, it takes or keeps the floor, it indicates presence
and communicative activity.
Languages tend to possess a group of items specialized for performing
these tasks. In brief, then, the OIs have two major functions:
Author's Copy

Author's Copy
(1) Keeping the discourse going—keeping the floor, showing listener-
ship. This is a ‘‘convergent’’ category, one where participants indi-
cate their e¤ort and willingness to maintain the collaborative activity
of speaking.
(2) Shifting direction—making new openings, changing topic, challeng-
ing the previous state of the discourse. This frames a new opening,
and usually involves speaker change. This is a ‘‘divergent’’ use.
This brief sketch of LUG has hopefully given a general idea of the
model and of the interactive OI elements in particular. What we can see
from this example is that lingua franca speakers use O and M elements,
the fundamental building blocks of any language, and also both OT and
OI. They manage interaction with OI, bridging the virtual world of
shared experience with the real world of interaction. In these respects
they are indistinguishable from L1 speakers. In others, they take a di¤er-
ent route.

4. Corpus data

My data comes from the ELFA corpus, more specifically its first half,
ELFA(i), which consists of 0.6 million recorded and transcribed words.
The ELFA corpus, 1 million words in all, comprises spoken ELF dis-
course recorded in university settings (www.eng.helsinki.fi/elfa/). ELFA
is compiled from authentic speech situations where English is used as a

Author's Copy
Author's Copy

224 Anna Mauranen

vehicular language between speakers who do not share a first language.


English is not as an object of learning in any of the recordings, that is,
no classes of English have been included. Native speakers are occasion-
ally present in the speech events, but never as lecturers or in other domi-
nant roles such as examiners of doctoral theses; nearly all speech is by
non-natives, who come from over 50 di¤erent first language backgrounds.
For a closer description of the database, see for example Mauranen
(2006). The extract discussed so far comes from a seminar discussion ran-
domly picked from the corpus and subjected to a linear reading. The cor-
pus as a whole is used for checking the distributions of recurrent expres-
sions.
For comparisons with similar native speaker data, the MICASE corpus
(http://quod.lib.umich.edu/m/micase) has been used.

5. What can we expect from ELF?

The linear approach briefly outlined above sets out to describe language
Author's Copy

Author's Copy
use generally, not confining itself to a particular language, a particular
mode, register or participants’ native speaker status. Therefore, by and
large, we should expect the model to apply to people communicating in
a lingua franca just as much as to anyone else.
Thus, keeping our attention on OI elements, what can we expect from
people using English as a lingua franca? If OI elements are (a) crucial for
ensuring smoothly running interaction, (b) something of a closed class (c)
frequent, then we might expect LF users
(i) to use them, because they are necessary for ensuring successful com-
munication, (as in [a] above)
(ii) to get them mostly right, because they are frequent and a restricted
class, which makes them salient in their conventional form (as in [b]
and [c])
(iii) to show variation in form that ignores some conventions and pref-
erences of native English speakers, because English is a foreign lan-
guage to most of those who use it as a lingua franca.
It is easy to show that ELF speech contains a large number of OIs, and
that they are employed turn-initially (oh yes; yes but; yeah well; okay; I
think . . .) as well as turn-finally or medially (I don’t know; or something;
I mean . . .). What is more intriguing is the extra variation in the expres-
sion that is apparent in ELF discourse, and which nevertheless seems to
achieve communication just as well as elements that are indistinguishable
from the ‘‘Inner Circle’’ English convention.

Author's Copy
Author's Copy

Chunking in ELF: Expressions for managing interaction 225

In Example (2), what catches the eye is the longest OI in the extract, in
my point of view. It clearly looks like a phraseological unit, even though
slightly out of the common way. The meaning can be paraphrased as ‘‘in
my view’’, or ‘‘in my opinion’’; and in form, it looks like a blend: it con-
flates in my view and from my point of view.
It has become widely accepted as a fact that phraseology is where non-
native speakers go wrong even at advanced levels of proficiency (e.g.,
Pawley and Syder 1983; Nattinger and de Carrico 1992; papers in Schmitt
2004). Conventional turns of phrase, expressions, combinations and col-
locations which native speakers prefer instead of equally grammatical al-
ternatives constitute the ultimate stumbling blocks for foreign language
speakers. On the basis of such evidence, Wray (2002) has posited two dif-
ferent processing mechanisms for the native language and other, later
learned languages. Native speakers, on this account, have two language
processing mechanisms available to them: one holistic, which operates
with prefabricated units (‘‘formulaic sequences’’ as Wray calls them) and
which chiefly accounts for fluency and easy execution of speech. The
other mechanism is analytic or compositional. It operates with small
Author's Copy

Author's Copy
components and allows for flexibility and novelty in language use. In con-
trast to native speakers, non-native speakers are not supposed to have the
first type, holistic processing, available to them any longer. According to
Wray, they apply compositional processing to everything in the foreign
language, breaking it down to its smallest elements (or thereabouts, be-
cause Wray actually talks about individual words). In this way, L2 users
must construct all their expressions from scratch. Clearly, such a process
is much more error-prone than one which can utilize larger prefabricated
units; if no convenient formulae are readily at hand to reduce processing
e¤ort, a good deal of faltering is to be expected.
This reasoning seems to o¤er a convenient explanation to the deviant
form used by S6 in Example (2): as a non-native speaker he is prone to
error, which shows in his blunder as regards the form. But can this be
written o¤ as a foreign speaker’s idiosyncratic use or a downright error?
To get a handle on this, we need more data.

6. What ELF speakers do

One of the open questions with regard to O elements is to what extent


they constitute a closed class. Looking at the very brief one and two-
word OIs in any corpus they seem to range from vocalizations (er, mm,
uhuh, ha) to backchannelling or ‘‘minimal response tokens’’ in Carter
and McCarthy’s terms (Carter and McCarthy 2006) ( yeah, right, okay,

Author's Copy
Author's Copy

226 Anna Mauranen

oh yes, okay okay). There is little space for variability in such minimal ex-
pressions, and O elements on the whole resist being interrupted. As soon
as we go beyond the single-item element, the potential for variation
grows. At the same time, the borderlines between element types become
fuzzier. Interruptions such as insertions of modifiers into expressions typ-
ically functioning as OI (I don’t know) are more likely to be M (I really
don’t know). Context-dependence plays an increasing role, and expres-
sions of the same form perform di¤erent functions: I suppose, sort of or I
don’t know can act as M or OI elements depending on their position in
interaction. Further, a wide variety of expressions can assume OI roles
in context (I suppose, I remember, aren’t they . . .). These expressions are
not an entirely open category, but neither are they readily listable like
the most frequent items.
We can liken this to Winter’s (1977) concepts of Vocabularies 1, 2, and
3, where Vocabularies 1 (sentence internal conjunctions: but, and, al-
though . . .) and 2 (sentence linking connectors: however, therefore, never-
theless . . .) are essentially closed classes, while highly similar meanings are
performed by Vocabulary 3, in principle open-class items such as reason,
Author's Copy

Author's Copy
mistake, evidence, or opposite. In terms of grammaticalization, we could
say that a large number of OI are grammaticalized (or pragmaticalized),
whereas others draw on a wider pool of language resources, and are
highly context-dependent for their interpretation.
As variability increases in the expressions, we may assume that so does
the likelihood that second-language users di¤er from native speakers,
since it is reasonable to assume that conventionalized items are less
deeply entrenched in second language repertoires. One thing that appears
common to ELF speakers and second language learners is that the fre-
quency and distribution of expressions di¤ers from that of native speak-
ers, even if the expressions are used pragmatically in the same functions.
Typically one expression gets favored in a given function—I mean is a
rephrase-flagging OI that is used proportionally several times more by
ELF speakers than natives, as seen in comparing ELFA to native speak-
ers in the MICASE corpus (see, Mauranen 2007). On the other hand, like
learners, ELF speakers may ‘‘underuse’’ pragmatic expressions or just
show di¤erent preferences in their distribution. So for example with ex-
pressions involving right as part of an OI. Common expressions such as
all right/alright, um right/mhm right, and so right in ELF get used be-
tween 6% and 24% of corresponding NS frequency. More strikingly, the
expression that native users employ most commonly apart from alright,
namely repeated right (right right, right right right) is used only once in
ELFA. Similar non-use appears with fairly common expressions like oh
right and okay right (25 and 19 cases in MICASE respectively): one case

Author's Copy
Author's Copy

Chunking in ELF: Expressions for managing interaction 227

of oh right in ELFA(i) was by a NS, the other of an ‘‘unknown’’ language


status, and okay right was not found at all. On the other hand, yeah right
seems to have found its way to ELF repertoires better than the other var-
iants, and is used 67% of the corresponding MICASE frequency.
Slightly longer items such as I’m sure show how the OI function is re-
sistant to interruptions: I’m not sure can be an OI as in Example (3):
(3)
i suppose finland would be an ideal model in that we have lots of ICT but
then the e¤ects i’m not sure
Often I’m not sure is clearly an M element, as in (4):
(4)
. . . and they are rather vocal so i’m not sure if that is the, biggest erm
missing link
Again, if we insert more elements, as in I’m not quite sure, it is predomi-
nantly an M element which shows in ELF use as well (Example 5):
Author's Copy

Author's Copy
(5)
. . . in estonia you had a, erm swedish communities at la- at least i’m not
quite sure how big they were . . .
Quite sure has a clear negative semantic prosody in both L1 and ELF
speech, as can be seen in Example (5) On the whole, I’m (not)(quite) sure
seems to be used in very similar ways by speakers of English, whether it is
their native language or not.
Intriguingly, however, unlike L1 speakers ELF speakers use very sure
(1/100,000 words) as an alternative expression (Example 6):
(6)
modernizer of a sort but i’m not very sure (actually) exactly what from
er er 19th century but er i’m not very sure if er if er they were confusion
or the conflict mhm i’m not very sure you know a nigerian system of
helsinki er but, i’m not very sure about that er i read it in
The instances were uttered by speakers from di¤erent L1 backgrounds
(Finnish, Turkish, and Dagbani) on independent occasions. A search in
MICASE yielded two hits, one by a non-native and one by a near-native
speaker, but none among any of the native varieties. This would seem
to suggest that as we move away from very short, highly formulaic units,
native speakers’ conventionalized expressions begin to get chinks and
fractures in ELF speakers’ hands.
To focus in on one case in more detail, let us return to Example (2)
above, and the OI in my point of view. It can be viewed in formal terms

Author's Copy
Author's Copy

228 Anna Mauranen

as a structural blend of two expressions, which share a meaning. From


the perspective of ongoing discourse, or language in interaction, its func-
tion is to preface an utterance, signaling that the speaker’s opinion will
follow.
Contextually it functions as preparation for a divergent view. S1 and
S6 seem to hold di¤erent views on the event they both refer to deictically
as that (and which from the wider context turns out to be a piece of legis-
lation). S6 engages in preparatory work by starting out with yeah, but,
which is a two-part move towards acceptable divergence. The first part
shows solidarity with S1 ( yeah), the second part begins the new departure
(but) Then S6 orients to the interaction once more by prospecting the
kind of act he is about to perform, an expression of his personal view (in
my point of view).
Why is he engaging in such elaborate relational work? Consensus is a
powerful norm in face-to-face interaction, and any departure from that
requires preparatory acts. Academic discussions may permit, even en-
courage more divergent views than many other kinds of discourse, but
criticism and disagreement even in academia get expressed less directly
Author's Copy

Author's Copy
and openly than agreement or praise (see, e.g. Mauranen 2002, 2004). Al-
though academic discourse in principle encourages critical thinking and
divergent viewpoints as an essential rule of the game, the tendency to-
wards consensus seems to overrule much of it, presumably in the interests
of maintaining a friendly and cooperative atmosphere. It seems, then,
that speaker 6 in our extract is employing the OI in a way which con-
forms to the expectations of the discourse community; he gets the mean-
ing and the pragmatics right from this perspective, and presents himself as
an appropriately socialized member of the community. But the form of
his expression is unorthodox.
We can also invoke other conceptualizations of the expression. In cog-
nitive terms we can see this in terms of a non-native speaker’s failed at-
tempt to use a target expression (such as in my view or from my point of
view). In this way, as discussed in Section 5, it would count as support for
Wray’s claim that we only have a compositional processing mode avail-
able to us when speaking a second language. From a more sociolinguistic
angle, in my point of view can be seen as evidence of a speaker’s L2 idio-
lect. This would go along the lines of Mollin’s (2006: 44–45) claim that
ELF cannot be considered a variety for lack of su‰cient linguistic homo-
geneity—speakers’ errors are idiosyncratic.
Whether it is an individual’s processing error or a fossilized idiolectal
feature cannot be solved on the basis of this one occurrence alone—but
if it is a unique occurrence or specific to a single individual, this would
be compatible with both Wray’s and Mollin’s claims. If, however, this is

Author's Copy
Author's Copy

Chunking in ELF: Expressions for managing interaction 229

not a unique occurrence, neither are supported. I searched the ELFA(i)


corpus for possible further evidence of this or similar cases. The result is
seen as concordance lines in (7):
(7)
no er specific in-, interest erm in my point of view in in this kind of usage.
but in er example in my point of view daddies always cook what in my
point of [view] globalization in my point of view er but on the other
hand i would say yeah but i in my point of view that was really good
[because] it can be said and er on my point of view er i will say that the
second market regulation on my point of view are the third world coun-
tries or south conclusion we have where, on my point of view my humble
point of view i would say, intensifies, the world on my point of view if you
want we can open
As can be seen, it was not a unique occurrence, but appeared re-
peatedly (13 times/10,000 words). Each of the instances reflects the
‘‘opinion’’ sense. They come from five di¤erent speakers in four indepen-
dent speech situations, and the speakers represent four di¤erent first lan-
Author's Copy

Author's Copy
guages (Estonian, French, Flemish and Somali).
I have noted traces of in/on my point of view in other ELF texts. One
was a thesis investigating email discussion groups in an international
English-medium business program (Karhukorpi 2006). The study took
its data from ‘‘reflective essays’’ set as coursework to be done as email
group discussion. Here are two instances from a writer, whose L1 is not
identified but which is not Finnish, judging from the rest of her text:
(8)
In my point of view, she should talk with him about her feelings . . .
In my point of view the most important thing here is the good for the child
...
(Karhukorpi 2006: 110)
I also had access to a fragment of the VOICE corpus in Vienna
(www.univie.ac.at/voice/index.php), while acquainting myself with a
preliminary version of their software in July 2007. In the few searches I
made on this practice version, I found these two examples (9):
(9)
In my point of view 1
In your point of view 1
Even though it is clear that these instances do not constitute evidence
of a systematic ELF feature which has been adopted across the board,
the similarity of the form and function in these cases is striking. Their
contexts of occurrence are independent of each other, and they have

Author's Copy
Author's Copy

230 Anna Mauranen

been uttered by speakers of di¤erent first languages. They do not arise


from mutual accommodation in a given group of speakers; nor is there
any reason to assume that interference from the mother tongue could be
at stake, not even typological similarity, since the speakers’ first languages
were from distant language families.
In my point of view thus roughly indicates that the next increment is to
be taken as the speaker’s opinion. But the question arises why would
these speakers use a longer and arguably clumsier way of indicating this
than a more standard way, such as in my view or in my opinion? The an-
swer is they do not, always. In my view is also in use, albeit less often than
in/on my point of view, and from my point of view was found about
equally frequently as in/on my point of view. In my opinion is the com-
monest of them all. All of these ways of flagging speaker opinion are pro-
portionally more common than in a comparable native corpus (MI-
CASE), with from my point of view leading the way: it is eighteen times
more frequent in ELFA (18 vs. 1/10,000 words).
Such choices seem to suggest some preference for a longer form. This
relates to Ranta’s (2006) finding that the ing form of the verb is used
Author's Copy

Author's Copy
much more often in ELF than native-English discourse. She puts it
down to greater visibility and expressivity of the longer form, and some-
thing similar may be behind these opinion flags as well. The longer ex-
pressions would seem to carry more prominence, more weight.
Whatever the reasons for the preferences, this example can be described
as a blend of two forms close in meaning. This is a common enough
phenomenon in ordinary monolingual contexts of language internal
change. At a more general level, though, this could be seen as an instance
of ‘‘approximation,’’ the tendency of ELF speakers to latch on to sa-
lient features of a phraseological unit, which they use in its established
sense, but without exactly reproducing the standard form (see Mauranen
2005). In view of the present example it makes little sense to write this
o¤ as an inability of L2 speakers to master the target form in full, (let
alone a learner’s error), because the repeated use speaks against an error
interpretation.
In brief, then, these examples throw light on some aspects of ELF
speech where it departs from Standard English conventions. Specifically,
it makes new departures in phraseological units via ‘‘approximation.’’

7. Conclusion

This paper has looked at English spoken as a lingua franca from two
complementary perspectives: linearity and recurrent patterning. The lin-

Author's Copy
Author's Copy

Chunking in ELF: Expressions for managing interaction 231

ear view was based on the Linear Unit Grammar (Sinclair and Mauranen
2006). The analysis was applicable to ELF because the model is su‰-
ciently general and robust, and also because speaking a natural language
as a lingua franca is quite normal in human communication; it ought to
be covered by descriptive models. ELF data is therefore a good testbed
for models of language aspiring to generality.
With a linear perspective, we can get a new handle on how speakers
construct successful discourse. It helps us see how moment-to-moment in-
crementation of the shared world is achieved by speakers, and how they
organize discourse and manage interaction at the same time, with alter-
nating elements of di¤erent types.
The analysis focused on elements specialized in managing interaction
(OI), which play a crucial role in understanding how we achieve so many
di¤erent things by just using language; their role as mediators between the
situation and the text, the real and the virtual helps see how we create order
in the apparent chaos of real-time speech. These elements are vital for suc-
cessful communication, probably far more important than the lexicogram-
matical correctness that tends to steal the greatest attention. The snippet of
Author's Copy

Author's Copy
ELF analyzed here illustrated a skilful and appropriate use of OI elements
by the speakers in the discussion which progressed with no apparent com-
municative turbulence despite some non-standard forms of English.
OI elements are usually short and relatively invariable. The more de-
grees of freedom there are in the expressions, the more opportunity there
seems to be for ELF speakers to bend conventions. This is where OI
meet other phraseological units, whether of the M or the OT type: multi-
word units open up possibilities for deviating from existing conventions
and shaping new ones in social interaction.
We could posit that phraseology is at the interface of linguistic conven-
tion and creativity: it contains enough familiar material for the hearer to
go on to ensure comprehension, thereby allowing more freedom to the
speaker. In other words, by virtue of the conventional and fixed parts
which ensure recognition, phraseological units allow a measure of free-
dom for innovation or approximation in details without risking compre-
hensibility (see, e.g., Mauranen 2005).
The corpus perspective was able to show that ELF is not a chaos of
idiosyncrasies which arise at random and may work in their contexts but
which are not likely to have a bearing on English as a whole because
there is no regularity or systematicity.
In brief, then, both approaches used in this paper showed that English
spoken as a lingua franca is normal use of natural language, and at the
same time has its unique features which constitute an important part of
the pool of global English.

Author's Copy
Author's Copy

232 Anna Mauranen

References

Auer, Peter (2005). Delayed self-repairs as a structuring device for complex turns in conver-
sation. In Syntax and lexis in conversation, Auli Hakulinen and Margret Selting (eds.). 75–
102. Amsterdam: John Benjamins.
Biber, Douglas, Stig Johansson, Goe¤rey Leech, Susan Conrad, and Edward Finegan
(1999). The Longman grammar of spoken and written English. London: Pearson Educa-
tion.
Brazil, David (1995). A grammar of speech. Oxford: Oxford University Press.
Carter, Ronald and Michael McCarthy (2006). Cambridge grammar of English. Cambridge:
Cambridge University Press.
Cheng, Winnie, Chris Greaves, and Martin Warren (2008). A corpus-driven study of dis-
course intonation. Amsterdam: John Benjamins.
Dabrowska, Ewa (2004). Language, mind and brain: Some psychological and neurological
constraints on theories of grammar. Edinburgh: Edinburgh University Press.
Du Bois, John W. (2003). Discourse and grammar. In Michael Tomasello (ed.), The new
psychology of language (Vol. 2).: 47–87. Mahwah, NJ: Lawrence Erlbaum.
ELFA corpus www.eng.helsinki.fi/elfa/elfacorpus.htm
Ellis, Nick (1996). Sequencing in SLA. Phonological memory, chunking, and points of
order. Studies in Second Language Acquisition 18: 91–126.
Firth, Alan (1996). The discursive accomplishment of normality: On ‘lingua franca’ English
Author's Copy

Author's Copy
and conversation analysis. Journal of Pragmatics 26: 237–259.
Karhukorpi, Johanna (2006). Negotiating opinions in Lingua Franca email discussion
groups: Discourse structure, hedges and repair in online communication. Turku: Univer-
sity of Turku unpublished Licentiate thesis.
Kempson, Ruth, Wilfried Meyer-Viol, and Dov M. Gabbay (2001). Dynamic syntax: The
flow of language understanding. London: Blackwell.
Levelt, Willem (1989). Speaking. Cambridge, Mass.: MIT Press.
Mauranen, Anna (2002). A good question: Expressing evaluation in academic speech. In
Domain-specific English. Textual practices across communities and classrooms, Giuseppina
Cortese and Philip Riley (eds.). 115–140. Frankfurt: Peter Lang.
Mauranen, Anna (2004). Talking academic: A corpus approach to academic speech. In Dia-
logue analysis VIII: Understanding and misunderstanding in dialogue, Karin Aijmer (ed.).
201–217. Tübingen: Max Niemeyer.
Mauranen, Anna (2005). English as a Lingua Franca—an unknown language? In Identity,
community, discourse: English in intercultural settings, Giuseppina Cortese & Anna Dus-
zak (eds.). 269–293. Frankfurt: Peter Lang.
Mauranen, Anna (2006). A rich domain of ELF—the ELFA corpus of academic discourse.
The Nordic Journal of English Studies 5(2): 145–159.
Mauranen, Anna (2007). Hybrid voices: English as the Lingua Franca of academics. In Lan-
guage and discipline perspectives on academic discourse, Kjersti Flottum, T. Dahl, and T.
Kinn (eds.). 244–259. Cambridge: Cambridge Scholars Press.
MICASE corpus http://quod.lib.umich.edu/m/micase
Miller, George A. (1956). The magical number seven, plus or minus two: Some limits on our
capacity for processing information. Psychological Review 63: 81–97.
Mollin, Sandra (2006). English as a Lingua Franca—A new variety in the new expanding
circle? In The Nordic Journal of English Studies (Special issue: English as a Lingua Franca)
5(2). Anna Mauranen and Maria Metsä-Ketelä (eds.): 41–57.
Mukherjee, Joybrato. (2001). Form and function of parasyntactic presentation structures: A
corpus-based study of talk units in spoken English. Amsterdam: Rodopi.

Author's Copy
Author's Copy

Chunking in ELF: Expressions for managing interaction 233

Nattinger, James R. and Jeanette DeCarrico (1992). Lexical phrases and language teaching.
Oxford: Oxford University Press.
O’Grady, William (2005). Syntactic carpentry. An emergentist approach to syntax. Mahwah,
N.J: Lawrence Erlbaum.
Pawley, Andrew and Frances H. Syder (1983). Two puzzles for linguistic theory: Nativelike
selection and nativelike fluency. In Language and communication, Jack C. Richards and
Richard W. Schmidt (eds.). 191–227. London: Longman.
Saussure, Ferdinand de (1916). Cours de linguistique générale (compiled by Charles Bally
and Albert Sechehaye). Paris: Payot.
Schmitt, Norbert (ed.) (2004). Formulaic sequences: Acquisition, processing and use. Amster-
dam: John Benjamins.
Sinclair, John and Anna Mauranen (2006). Linear unit grammar. Amsterdam: John Benja-
mins.
Warren, M. (2006). A corpus-driven analysis of back-channels. Paper given at the Third
Inter-Varietal Applied Corpus Studies (IVACS) International Conference, University of
Nottingham 23–24 June, 2006.
Winter, Eugene (1977). A clause-relational approach to English texts. Instructional Science
(Special issue) 6(1): 1–92.
Wray, Alison (2002). Formulaic language and the lexicon. Cambridge: Cambridge University
Press.
VOICE corpus www.univie.ac.at/voice/index.php
Author's Copy

Author's Copy
anna.mauranen@helsinki.fi

View publication stats Author's Copy

You might also like