See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/371469491

Models of Speaking and the Assessment of Second Language Proficiency

Chapter · June 2023

CITATIONS READS
0 153

1 author:

Peter Skehan
Institute of Education University College London
105 PUBLICATIONS 15,499 CITATIONS

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Language Aptitude: Pushing the Boundaries in Theory and Testing View project

Processing perspectives on task performance View project

All content following this page was uploaded by Peter Skehan on 10 June 2023.

The user has requested enhancement of the downloaded file.

Peter Skehan

Chinese University of Hong Kong

This is an earlier version of the article which appeared as: Skehan, P. (2009c). Models of speaking and the
assessment of second language proficiency. In A. G. Benati (Ed.), Issues in Second Language Proficiency
(pp. 202-215). London: Continuum International Pub Group.

Within applied linguistics, the assessment of oral second language proficiency has been

explored in a way which is surprisingly detached from developments in a range of related

areas, including second language acquisition, psycholinguistics, and our understanding of

what is involved in the speaking process. It is the purpose of this article to explore the

relevance for such proficiency assessment of theory in the area of first (and second)

language speaking, and empirical findings in task-based performance. It will be argued

that these two ancillary areas have much to contribute to how we design and evaluate oral

assessment procedures.

Models of Speaking

The dominant model of first language speaking is that of Levelt (1989, 1999, Kormos

2006). This model proposes that there are three general stages in speech production, as

shown in Table One.

Table One: The Levelt Model of First Language Speech Production

Conceptualisation (Message Level of Representation)

• involves determining what to say

• speaker conceives an intention and adopts a stance

1
 • speaker selects relevant information in preparation for construction of an intended

utterance

• the product is a preverbal message

Formulation

• involves translating the conceptual representation into a linguistic form

• includes the process of lexicalisation, where words that the speaker wants to say

are selected

• includes the process of syntactic planning where words are put together to form a

sentence

• involves detailed phonetic and articulatory planning

• includes the process of phonological encoding, where words are turned into

sounds

Articulation

• involves retrieval of chunks of internal speech from buffer

• involves motor execution

The first stage, Conceptualisation, is concerned with developing and organising ideas to

be expressed, in a particular situation, and with a particular emphasis and stance. The

second stage, Formulation, involves lexical selections to match the preverbal message,

triggereing lemma retrieval from the mental lexicon, and then syntax building from the

information retrieved through this retrieval. The third stage is concerned with

transforming this form of representation of ideas, lexis and syntax into actual speech.

2
Fairly obviously, if we are dealing with native language speech, certain assumptions are

made. Most central for present purposes is that the Conceptualiser delivers challenges to

the Formulator which the Formulator is able to meet, since the driving force for its

operation is the mental lexicon, which is extensive, well-organised, and contains rich

entries, with semantic, syntactic, collocational, and phonological information. It is the

capacity to retrieve such rich information very quickly that enables speech production to

be so smooth, without requiring consciousness, and with parallel operations. The

Conceptualiser can be working on one stage of operation, while the Formulator is dealing

simultaneously with previous Conceptualiser operations, as language is being assembled,

just as the Articulator can be working on previous Formulator operations.

When we turn to second language speaking, things are very different. The ‘driver’ for the

Levelt model is the mental lexicon, which is the repository of considerable information

about each lemma, information which underpins the production of native language

speech in real time. With second language speakers, this mental lexicon is:

 smaller, so that many lemmas required by the pre-verbal message, will not be

available

 incomplete, so that where many lemmas are part of the mental lexicon, they are only

partially represented, so that for example, limited semantic information may be

available, or limited syntactic implication information is available, or no collocational

information etc.

3
 less organised, so that connections between lemmas do not prime one another, or

alternative forms of expression are less available

 less redundantly structured, in that collocational chunks are less available, so that

speech production has to be done more often ‘from first principles’ on the basis of

rule-generated language

The result of all these omissions is that during speech production, the pre-verbal message,

however impressive, encounters considerable difficult at the Formulator stage. The

smooth process of speech production is disrupted, as second language speakers have to

try to find alternative methods of expressing their meanings, or to find ways of using the

resources that they have sufficiently quickly so that ‘normal’ communication can

proceed. This is likely to present serious difficulties for the modular, parallel operation of

the Levelt model, so that some of the time, this ideal set of interlocking processes is

severely impaired.

Task Research

The preceding discussion has been general, and data-free. It has assumed that second

language speakers, of different proficiency levels, will encounter difficulties, and that it is

meaningful to express these difficulties in terms of the Levelt Model. What is necessary

to take the discussion further is some empirical method of characterising difficulty. To

this end, we will next review findings from the literature on task-based performance.

This literature provides a wide range of generalisations regarding the way task

4
characteristics and the conditions under which tasks are done might have an impact on

different aspects of second language performance. These findings can take us beyond the

programmatic statements in the previous section.

Broadly, the task findings can be portrayed in terms of the information that tasks are

based on, the operations that tasks require to be performed on this information, and the

conditions under which tasks are done. These three headings organise the findings which

have emerged in the task literature. Regarding information, we can say that:

 the scale of the information (e.g. the number of elements, the number of participants)

has an impact on performance, with more information leading to performance that is

more complex (Michel, Kuiken and Vedder 2007)

 the type of information has a strong influence, with concrete information being easier

to handle than abstract information which in turn is easier to handle than dynamic

information (Brown et al. 1984), and also with there being a tendency for concrete

information to advantage fluency and accuracy in performance

 greater familiarity with the information is associated higher levels of accuracy and

fluency (Foster and Skehan 1996)

 greater organisation and structure in the task information leads to performance which

is also more accurate and more fluent (Tavakoli and Skehan 2005)

There are interesting theoretical issues which relate to this range of findings. For now, we

will simply say that these are a set of empirically based generalisations which have

relevance for any development of a model of second language spoken performance. In

5
any case, we move on to consider the operations which are carried out on the information

in tasks. Here another set of generalisations can be offered, contrasting simple and

complex operations. Simple operations, e.g. listing, enumerating (Willis and Willis

2007), or retrieving (Foster and Skehan 1996) or simply describing are associated with

higher fluency and accuracy in performance. In contrast, more demanding operations on

information, such as sequencing (Willis and Willis 2007) or reorganising (Foster and

Skehan 1996), or integrating information (Tavakoli and Skehan 2005), or reasoning

(Robinson 2001, Robinson and Gilabert 2007) are associated with greater language

complexity. As these different influences suggest, there is something of a contrast here

between the performance areas of complexity, on the one hand, and accuracy and

fluency, on the other. Complex operations seem to drive greater language complexity,

while simpler operations are more likely to be associated with higher fluency and

accuracy.

The final area to consider here is the conditions under which tasks are done. This has

provoked considerable research effort. Robinson (2001) reports higher levels of accuracy

in his there-and-then condition, i.e. input absence, contrasted to a here-and-now condition

(input presence). A whole series of studies has examined the effect of pre-task planning

on performance. The results suggest that such planning is consistently associated with

greater complexity and fluency, with these generating sizable effects, while accuracy is

generally advantageous, but not to the same degree (Ellis, in press, Foster and Skehan

1999). Yuan and Ellis (2003) have reported that opportunity for on-line planning, i.e.

more relaxed time conditions during performance, are associated with greater accuracy.

6
Skehan and Foster (1997, submitted) have proposed that requiring second language

speakers to do post-tasks which emphasise language form leads selectively to increases in

accuracy. Finally, Skehan (in press) proposes that tasks which contain unavoidable and

more difficult lexis are associated with lower levels of accuracy and complexity.

Applying the Levelt Model

The generalisations about second language task-based performance provide the basis for

examining the relevance of the Levelt model in accounting for the second language case.

An initial assessment would suggest that it has to be applicable. Essentially, the model

proposes that thought precedes language, and that the functioning of the model is

essentially concerned with how Formulator operations, and associated resources, such as

the mental lexicon, can enable the thoughts embodied in the pre-verbal message to be

expressed through externalised language. One is drawn into saying that anything other

than this would not be satisfactory on general, logical grounds. Communication requires

ideas, and the ideas, if they are to be transmitted effectively, require language.

The difficulty, of course, is that first language speakers are able to function in this way

because they have mental lexicons which are extensive, elaborate, analysed, and

accessible. Pre-verbal message demands can therefore be met, and ongoing speaking in

real time is possible. In particular, the different modules within the Levelt model can

function in parallel, i.e. the Conceptualiser may be working on the next pre-verbal

message, while the Formulator is working on the current one, even while the Articulator

7
is giving voice to the previous. This is possible because the different modules can handle

the demands which are placed upon them in real-time, and smoothly, such that successive

cycles of communication proceed, if not totally effortlessly, at least fairly so.

Problems occur when a second language speaker attempts the same ideas-to-language

mapping, but is equipped with a much more limited mental lexicon. The result is, as

Kormos (2006) argues, that parallel processing below the level of consciousness is

replaced by serial, effortful and conscious attempts to deal with pre-verbal message

demands on the mental lexicon – demands which are either beyond it, or which require

considerably greater allocation of attention. Messages may then be assembled slowly, or

communication strategies used, or messages may be modified or even abandoned (Faerch

and Kasper 1983). This is turn brings us back to the question of the applicability of the

model. What is argued here is that the model is still relevant (for the general reasons

proposed earlier), but that it is inappropriate to regard the model as an ‘all or none’ affair.

The generalisations from the previous section show that performance is systematically

influenced by a range of information, operation, and condition influences, and that these

impact upon complexity, accuracy and fluency aspects of performance. Broadly, it is

argued that what we have learned from this research is the conditions which facilitate the

relevance of the Levelt model and those that do not. So the question needs to be recast so

that we explore not whether there is a simple positive or negative answer to applicability,

but rather the influences which make the model a more useful explanatory account and

those which suggest that its potential for second language speakers is limited.

8
So far, in reviewing empirical work, we have presented research in the terms which are

typical for task-based performance, distinguishing between information, operations and

conditions, as independent variables, and complexity, accuracy, and fluency, as the

dependent variables. To examine the relevance of the Levelt model, though, it is

necessary to consider an alternative perspective, and to organise the findings which have

reviewed in a way more consistent with the model, and transparent in terms of its

functioning. This is shown in Table 2.

Table 2: The Levelt Model and Task Performance

Complexifying/Pressuring Levelt Model Stage Easing/Focussing

 complex cog. operations information

 complex info. type  less information

 easier cog. Operations

 non-negotiable task retrieval  dialogic

 heavy input pressure encoding  structured tasks

 monologic  dialogic

 post-task condition

Two introductory points are helpful here. First, the table is organised around the central

‘spine’ which shows the relevant stages of the Levelt model. There is a section which

focuses on the Conceptualiser, and then a separation between the two Formulator stages

9
of lemma retrieval (and access to the mental lexicon), and syntactic encoding (which is

driven by the information made available when lemmas are accessed). These three stages

organise the presentation of the information from the task-based performance literature.

Second, the outer columns are concerned with four general influences on performance,

two which indicate difficulty (complexifying, pressuring) and two which are more

focused on achievement (easing, focussing).

Complexifying focuses on performance influences which make the message to be

expressed more demanding, elaborate, or extensive. These influences are mostly relevant

for Conceptualiser operations. Pressuring concerns influences which make performance

more difficult, but are not connected with differences in the complexity of the message.

In contrast, they are concerned with the time under which a task is done, or the amount or

nature or inflexibility of the material which is involved. Easing is, in a sense, the reverse

of Complexifying, and entails the ways in which the pre-verbal message can be arrived at

in a more direct manner. Finally Focussing (which is not the reverse of Pressuring),

concerns the way in which performance conditions themselves introduce some level of

Focus-on-Form, thereby privileging performance areas such as accuracy and complexity.

This framework is useful for re-presenting the results from the task literature in ways

which bring out more clearly how the findings we have can illustrate the differing

degrees of applicability of the Levelt model to the second language speaker case. We will

explore each of them in more detail.

10
There are three clear influences on Complexifying. The first involves planning. At the

outset, it needs to be said that planning appears more than once in Table 2, reflecting the

different things that can happen during planning. Here the facet of planning time which

is focussed on is when preparation is associated with making the ideas in a task more

complex than they otherwise would be. It is clear (Skehan and Foster 1999, Ortega 2005,

Skehan and Pang, 2008: ms.) that this happens some of the time. Speakers use planning

time to explore the ideas in a task, and consequently approach the task as more

challenging than it otherwise would be. For example, in an Agony Aunt task, planning

may be used to generate more complex advice to the writer of a problem letter, or in a

narrativeit could be used to develop the connections in a picture story series.

A second influence here could be the more complex operations which a task inherently

requires if it is to be completed properly. These might be reorganise the input which is

provided, e.g. Foster and Skehan’s (1996) narrative task, where a story had to be invented

from a series of pictures, rather than a pre-existing storyline simply narrated. Or there

might be the need to integrate information to tell a story effectively, as in Tavakoli and

Skehan’s (2005) task where background and foreground elements had to be connected to

one another. There might also be a greater need to use reasoning, as in Michel et al’s

task (2007) involving selection of a cellphone where there are many features which

influence the eventual decision.

The third kind of influence concerns the type of information which is involved. Brown et

al (1984) showed that dynamic information, e.g. relating to a changing scene, is more

11
difficult to deal with abstract information, which in turn is more difficult to deal with than

static information. It seems that the greater difficulty in dealing with such information

types is more demanding of memory resources, and where there are also more complex

operations involved, this adds to task complexity. It is assumed that all three influences

here concern the Conceptualiser stage of speech production, and the nature of the

cognitive activity implicated in producing the pre-verbal message.

We turn next to a series of influences which increase the pressure on the operation of the

Formulator stage. Regarding the lemma retrieval stage, the first influence is the

infrequency of the lexis which is involved in a task. It appears that when less frequent

lexis is required, this has damaging implications for the complexity and accuracy of the

language which is produced (Skehan, in press). This contrasts with native speaker

performance, where lexical infrequency is positively correlated with language complexity

– less standard lemmas seem to drive more complex syntax in a harmonious way. So

tasks which push learners to need more difficult lexis seem to give them lemma retrieval

problems which spill over, because of their attentional demands, into other aspects of

performance. Lemmas are retrieved slowly and imperfectly, and the additional effort

required for this disrupts the parallel processing of the material for speaking.

Similarly, non-negotiable tasks (Skehan and Foster, 2007) also cause pressure. Native

speakers, when generating language, are able to draw upon a range of alternative choices

relatively effortlessly. So during speech production they can make a range of selections

as they are producing an utterance (Pawley and Syder 1983). Non-native speakers do not

12
have the luxury of such choices, and as a result are less able to adapt if a first lexical or

syntactic choice is unavailable in the mental lexicon. Where tasks are negotiable, such

speakers can adapt and revise the pre-verbal message so that it meshes more easily with

resources which are available. When tasks are non-negotiable, however, as with

narratives where the input is given, this is not possible, and disruption of performance

results.

These two pressure-inducing factors are concerned with the nature of the meanings that

are required for a task. The remaining pressuring influences are, in one way or another,

simply associated with time. Most obviously, requiring tasks to be done under timed

conditions is going to add to the speaker’s problems. Ellis (2005) has reviewed these

studies and shown that on-line planning is a possibility when time pressure conditions are

relaxed, with the result that greater accuracy is obtained. When there is greater time

pressure, in contrast, the result is that accuracy is lowered. But a related issue concerns

whether a task is monologic or dialogic. Monologic tasks are consistently more difficult

(with lower accuracy, and sometimes, lower complexity Skehan and Foster 2007). What

seems to happen here is that the speaker, being responsible for keeping the discourse

going, has to plan, execute, monitor and continue speaking without any respite. The result

is performance is lowered, because on-line planning opportunities are less frequent. In

contrast, a more dialogic condition does enable one speaker to have a break while

interlocutors are speaking. As long as the other speakers’ contributions are processed

sufficiently, there may spare capacity available while listening to enable a speaker to plan

upcoming conversational contributions, as well as potentially use something said by

13
interlocutors if it is appropriate to his or her own contribution. Monologic tasks do not

give any easy natural breaks. Dialogic tasks do and so reduce the pressure, overall

(Skehan, in press).

We turn next to factors which ease performance. Some of these can simply be dealt with

as the reverse of the effects we have already covered on complexifying or pressuring. For

example, regarding information, we can say that tasks based on concrete, static

information (Foster and Skehan 1996); tasks involving less information (Brown et al.

1984); and tasks which require simpler operations, such as listing (Willis and Willis

2007), are likely to be easier, and as a result, Formulator-based aspects of performance

(accuracy, fluency) are likely to be raised. Similarly, task conditions which reduce

pressure, such as a dialogic task, also make the task easier, as more time is available to

plan on-line. But there are other influences which are not mirror images of what has been

said before. For example, planning figures here, but now in different ways. If planning is

directed to organising what will be said, actual performance is eased (Pang and Skehan

2008: ms.). The planning, effectively, handles the wider plan of what will be said, so that

during actual performance, major ideas do not have to be developed and the speaker can

focus on the surface of language. Similarly, if a task is structured (Tavakoli and Skehan

2005), speakers are more likely to be able to exploit the macrostructure of the task, and

not need to engage in deeper planning. They too can focus on the surface, and are able to

mobilise Formulator resources more effectively, and thereby achieve higher levels of

accuracy and fluency. Returning to planning, in addition to the functions of extending

ideas, which complexifies, and organising, which eases, planning may be directed to

14
rehearsing language for actual performance (Ortega 2005). This too eases, but on the

assumption that what is rehearsed is both remembered and is actually useful during

performance. If these conditions are met, the result is that performance is eased.

So far, we have been looking at the interplay of ideas and their realisation through

language. Following much of the task literature, we have conceptualised performance in

terms of language complexity, accuracy, and fluency, and have tacitly assumed in

information processing perspective based on limited attentional resources. So the

assumption has been that speakers have limited capacities, and that task difficulty, as well

as eased performance conditions, influence performance priorities (Skehan 1998).

Performance, in this view, is a reflex of other influences. But there are also studies which

suggest that second language speakers, with their limited attentional capacities, may

choose to prioritise particular performance areas. We have seen this slightly with the way

planning time can be directed towards rehearsal, where speakers use the preparation time

in order to be ready with specific language, and as a result target accuracy, or sometimes,

particularly complex structures. But there are other ways in which the same selective

effect can occur, always with a focus on some aspect of form, either directed towards

accuracy or complexity. We have already seen how interactive tasks can help speakers

since they provide on-line planning opportunities (Foster and Skehan 1996; Skehan and

Foster 1997). But the effects of such tasks may be wider. Dialogic tasks make salient the

existence of an interlocutor, and it may be that speakers increase their focus on accuracy

precisely because they want to be understood and try to avoid misunderstandings by

selectively attending to accuracy and avoiding error (Pang and Skehan, 2008: ms.). In

15
this way, they prioritise attentional focus through awareness of their interlocutor’s

comprehension needs.

These claims, though, are based on interpretations of research studies with different

research foci. Two studies, though, specifically examine how speakers may have the

capacity to prioritise particular performance areas. Skehan and Foster (1997) showed that

using a post-task condition (requiring speakers, post-task, to engage in a public

performance) led to higher levels of accuracy on a decision-making task (although not on

narrative or personal information exchange tasks). It had been hypothesised that

foreknowledge of this post-task condition would lead learners to selectively prioritise

accuracy during their task performance. Subsequently, Skehan and Foster (submitted

2009) used a different post-task condition, hypothesised to be stronger. They required

learners to transcribe one minute of their task performance, a more personally-relevant

and language-focussed condition, and hypothesised that this would have a stronger effect

specifically on accuracy. This prediction was borne out, not only for a decision-making

task, as in Skehan and Foster (1997) but also for a narrative task. In addition, for the

decision-making task, complexity, too, was also significantly raised. These results

suggest that effective task conditions can lead speakers to focus on particular aspects of

performance, despite the attentional limitations which prevail.

Task Research and Proficiency Assessment

16
The analysis and review of research presented so far suggests that attentional limitations

are vital in understanding performance on second language tasks and that one can

categorise the range of different influences as follows:

 Complexification, in that factors such as extension planning, information type and

operations upon information push speakers to express more complex ideas. This set

of influences impacts on Levelt’s Conceptualiser stage

 Easing, with factors such as the reverse of the last set and which therefore simply

reduce the work the Conceptualiser has to do, coupled with other factors, e.g.

organising or rehearsal planning, dialogic conditions, or structured tasks which give

speakers clear macrostructure for what they want to say, or provide more on-line

planning opportunities. The consequence in all these cases is that Formulator

operations are eased

 Pressuring, where input conditions or performance conditions mean that speakers

have less opportunity to regroup while speaking, and are deprived of on-line planning

opportunities. This too is a Formulator influence.

 Focussing, where selective aspects of performance are privileged, and so a focus-on-

form is injected into task performance so that attention is directed in particular ways,

towards accuracy or complexity. Once again this is a Formulator influence.

These are a set of factors which influence what is going to be said and how it is going to

be said. What is central to this analysis is the balance between Conceptualiser pressure

(or lack of it) and Formulator pressure (or lack of it), and how these two sets of pressures

work themselves out during ongoing performance. Complexification gives a limited

17
attentional system more things to do, depriving other areas of resources. Easing has the

reverse effect, where ideas are manipulable or packagable more easily, thus releasing

attention for use in other aspects of speech performance. Pressuring has the general

effect of depriving all other areas of performance of time that would be useful. Finally,

Focussing operates differently in that it prioritises attention allocation, whatever is going

on elsewhere.

We now need to switch and try to consider how this analysis might have relevance for the

assessment of second language spoken proficiency. Here it is useful initially to revisit

some of the basics of language testing – that testing concerns the ways we use the

information elicited through tests in order to sample, to make judgements, and to

generalise about real-world performance. One approach to doing these things is to

provide learners with tasks to do, of increasing levels of difficulty, and then observe what

is the maximum level of difficulty that can be successfully handled. (This is like treating

language proficiency as similar to a weightlifting contest.) This approach makes the

unidimensionality assumption – that increasing proficiency means getting better along

only one dimensions (as when the bar gets heavier in weightlifting). Sampling, in this

view, means assessing learners along this one-dimensional scale of difficulty. But it can

be argued that anything interesting enough to be worth measuring is likely to be multi-

dimensional (Skehan 1984), with the result that sampling has to be directed at probing the

different dimensions that are important (whatever they may be) and then decisions have

to be made about how strengths and weaknesses in performance across these dimensions

can be combined to yield an overall judgement of proficiency.

18
The major insight from the analysis presented in this chapter is that for spoken language

performance we have to analyse test tasks first in terms of the influences covered in Table

2, and particularly in terms of the demands they make on Conceptualiser and Formulator

stages in speaking respectively. What influences the Conceptualiser appears to be

different from what influences the Formulator and so Conceptualiser difficulty is not the

same as Formulator difficulty. In a sense, it is what the Conceptualiser does that shapes

the overall difficulty of the ideas which are expressed in a task, and the influences on the

Formulator are then constrained by challenges set by the Conceptualiser, although this

stage too is affected by a range of influences. The problem is essentially one of

sampling, and what the above analysis does is clarify the basis on which sampling needs

to take place. It is a truism of testing that one-item tests are non-functional, and so if we

apply this to the assessment of spoken language performance, this means that a series of

tasks will be necessary for any effective assessment to be made. The matrix in Table 2

helps clarify how a range of tasks and a range of performance conditions can be sampled

as the basis for language testing. Different Conceptualiser influences and different

Formulator influences need to be drawn on if any rounded estimate of ability for use is to

be provided.

This approach interacts with how performance itself is measured (Pollitt 1990).

Language use, being complex and multidimensional, is not susceptible to simple

counting. There needs to be a more complex rating of performance, and this in turn

means that one has to decide what areas of performance should be represented in the

19
different rating scales which are used. We have seen that Conceptualiser work is reflected

in language complexity, and so this aspect of performance requires valid and reliable

rating in terms of what is often termed Range in language assessment, i.e. use of syntax

and vocabulary. But Formulator activity is associated with greater fluency and accuracy,

other areas where rating scales exist, with precisely these headings. In other words, a

Leveltian analysis of performance suggests that we need separate ratings of language

complexity (range), accuracy, and fluency if we are to obtain any satisfactory overall

assessment of the quality of a second language speaker’s performance. Only with such

information will we be able to make effective prediction about how second language

speakers will use language in the real world.

References

Brown G., Anderson A., Shillcock R. and Yule G. (1984), Teaching Talk: Strategies for

Production and Assessment, Cambridge: Cambridge University Press

Ellis R. (2005) (Ed.), Planning and Task Performance in a Second Language,

Amsterdam: John Benjamins Publishing Company

Ellis R. (in press), ‘The Differential Effects of Three Types of Task Planning on the Fluency,

Complexity, and Accuracy of L2 Oral Production, Applied Linguistics

Faerch C. and Kasper G. (1983) (Eds.), Strategies in Interlanguage Communication,

London: Longman

Foster P. and Skehan P. (1996), The influence of planning on performance in task-based

learning, Studies in Second Language Acquisition, 18, 3, 299-324

20
Foster P. and Skehan P. (1999), The effect of source of planning and focus on planning

on task-based performance, Language Teaching Research, 3 3, 185-215

Foster P. and Skehan P. (2009, submitted), The effects of post-task activities on the

accuracy and complexity of language during task performance

Kormos J. (2006), Speech Production and Second Language Acquisition, Mahwah, N.J:

Lawrence Erlbaum

Levelt W.J. (1989). Speaking: From intention to articulation, Cambridge, Ma: MIT Press

Levelt W. (1999), Language production: a blueprint for the speaker, In C. Brown and P.

Hagoort (Eds.), Neurocognition of Language (pp. 83-122), Oxford: Oxford

University Press

Michel M.C., Kuiken F. and Vedder I. (2007), The interaction of task condition and task

complexity in the oral performance of Turkish and Moroccan learners of Dutch,

Paper presented at the 2nd International Conference on Task-based Language

Teaching, Hawaii, U.S.A.

Ortega L. (2005), What do learners plan? Learner-driven attention to form during pre-

task planning, In Ellis R. (Ed.), Planning and Task Performance in a Second

Language, pp. 77-109, Amsterdam: John Benjamins

Pang F. and Skehan P. (2008: ms.), Using a model of speaking to explore second

language learners use of planning time, Chinese University of Hong Kong

Pawley A. and Syder F.H. (1983), Two puzzles for linguistic theory: nativelike selection

and nativelike fluency, In Richards J.C. and Schmidt R. (Eds.), Language and

Communication, London: Longman

21
Pollitt A. (1990), Giving students a sporting chance: assessing by counting and judging,

In C. Alderson and B. North (Eds.), Language Testing in the 1990s, pp. 46-59,

London: Macmillan

Robinson P. (2001), Task complexity, cognitive resources, and syllabus design: a triadic

framework for examining task influences on SLA. In Robinson P. (Ed.),

Cognition and Second Language Instruction, (pp287-318), Cambridge:

Cambridge University Press

Robinson P. and Gilabert R. (2007), Task complexity, the Cognition Hypothesis, and

second language learning and performance, Special Issue of International Review

of Applied Linguistics for Language Teaching, Berlin: Mouton de Gruyter

Skehan P. (1984), “Issues in the testing of English for Specific Purposes”, Language

Testing, 1,2, pp 202-220

Skehan P. (1998), A Cognitive Approach to Language Learning, Oxford: Oxford

University Press

Skehan P. (in press), Modelling Second Language Performance: Integrating complexity,

accuracy, fluency and lexis, Applied Linguistics

Skehan P. and Foster P. (1997), The influence of planning and post-task activities on

accuracy and complexity in task based learning, Language Teaching Research,

1,3, pp 185-211

Skehan P. and Foster P. (1999), The influence of task structure and processing conditions

on narrative retellings, Language Learning, 49, 1, 93-120

Skehan P. and Foster P. (2007), ‘Complexity, Accuracy, Fluency and Lexis in Task-based

Performance: A meta-analysis of the Ealing Research’, In Van Daele S., Housen

22
 A. and Kuiken F., Pierrard M. and Vedder I. (Eds.), Complexity, Accuracy, and

Fluency in Second Language Use, Learning, and Teaching, Brussels: Univ. of

Brussels Press

Tavakoli P. and Skehan P. (2005), Planning, task structure, and performance testing, In

Ellis R. (2005) (Ed.), Planning and Task Performance in a Second Language,

Amsterdam: John Benjamins

Willis D. and Willis J. (2007), Doing Task-based Teaching, Oxford: Oxford University

Press

Yuan F. and Ellis R. (2003), The effects of pre-task planning and on-line planning on

fluency, complexity, and accuracy in L2 monologic oral production, Applied

Linguistics, 24, 1, 1-27

23

View publication stats