Chapter 2: Theoretical Background
3.2 Research into the effects of cognitive task complexity on L2 production
3.2.5 Modes of performance
As can be seen from the summaries, written tasks are relatively under-researched in TBLT and even fewer studies compare L2 performance across modalities. Table 13 summarizes L2 writing studies by grouping them according to their research focus and their respective findings. Some studies have begun to explore the impact of task complexity on L2 writing. These studies have used narrative and argumentative tasks to examine resource-directing complexity variables such as reasoning, the number of elements to consider, and one resource-dispersing variable: planning time. The findings are rather mixed, but the emerging pattern seems supportive of the CH (Robinson, 2003a). The Cognition Hypothesis predicts that resource-directing task variables have the potential to improve learners’ performance by directing their attention to linguistic features, while resource-dispersing variables have an opposite effect
as they consume learners’ cognitive resources by taking their attention away from linguistic aspects of language, and as a result the level of learners’ linguistic performance decreases.
More recently task-based researchers have become interested in exploring how task design affects oral and written L2 production differently (see Groups 2 and 3 in Table 13). Argumentative and narrative tasks have been used to elicit L2 output. As the summary of the results (see Table 14) indicates, the findings are varied and inconclusive, in fact, no two studies seem to have reached the same conclusions based on the results they documented.
Table 10. Research into cognitive task complexity
STUDIES ASPECTS OF TASK COMPLEXITY
OTHER VARIABLES
METHOD
PARTICIPANTS DESIGN TASK TYPES MEASURES OF TASK PERFORMANCE
MAIN FINDINGS
Robinson, (1995)
+/- Here-and-Now none 12 adults,
aged 19-25, intermediate-level learners of English
each participant narrates both simple and complex stories
oral narrative based on cartoon strips
Linguistic complexity:
· S-nodes per T-unit
· propositions per utterance Lexical complexity:
· percentage of lexical words Accuracy:
· target-like use of articles Fluency:
· number of pauses (nonphonation)
· number of words per pausally defined unit
Significant difference for lexical complexity. Trend for target-like use of articles and utterance length.
Complex tasks are performed less fluently but more accurately than simple tasks.
Ortega (1999) +/- planning time (pre-task planning)
none 64 advanced learners
of Spanish in a FL setting
each participant narrates two stories under no planning and planning conditions (no planning vs. 10 min. planning time + note-taking allowed)
oral narrative based on pictures
Lexical complexity:
· number of words per
utterance
· type-token ratio
Accuracy:
· target-like use of articles
· target-like use of noun modifiers Fluency:
· pruned speech rate
Available planning time increases accuracy, fluency and lexical complexity except type-token ratio. Retrospective interviews revealed that planning facilitated learner-driven focus on form which also freed up attentional resources for on-line improvisation during task completion.
Skehan &
Foster, (1999)
+/- planning time task structure (structured vs.
unstructured)
47 adults, low-intermediate
2x4 (2 task structures, 4 planning time conditions)
oral narrative based on video viewings
Linguistic complexity:
· index of subordination Accuracy:
· number of error-free clauses
· number of words Fluency:
· number of repetitions
· number of false starts
· number of reformulations
· number of replacements
Significant effect of task structure on fluency.
Significant effect of planning time conditions on complexity Significant effect of the interaction of task structure and planning time effects on accuracy
Structured task leads to more fluent production, planning time leads to more complex production, planning time AND structured task leads to more accurate production.
Bygate et al., (2001)
Task repetition Effects of practising
48 ESL learners Participants are assigned to groups: two experimental, one control.
Oral tasks:
Narrative interview
Linguistic complexity:
· number of words per T-unit Accuracy:
· number of errors per T-unit Fluency:
· number of filled pauses per T-unit
Task repetition led to increased complexity on both task types. It led to decreased! Fluency in the interview task.
Task practice had a very weak effect.
Wigglesworth (2001)
+/- task structure (prompts vs. no prompts) +/- familiarity +/- planning time (no planning vs. 5 min.)
Native vs.
non-native interlocutor
Altogether 400 ESL learners at two levels of English in a testing situation.
Each task type with manipulated task characteristics and task conditions was administered to 20 learners.
5 oral task types (3 at level 1 and 2 at level 2).
Oral task types:
Giving instructions Asking information Negotiating for goods and services Telephone enquiries Negotiating a problematic exchange
Ratings of
· grammar
· fluency
· cohesion
· communicative effectiveness
Structured tasks appeared to be easier for learners. Also, when performed with a non-native speaker interlocutor, the tasks were easier. The familiar task was also easier when there was no planning time. Planning time adversely affected both structured and unstructured tasks.
Iwashita, et al.
(2001)
+/- here-and-now +/- planning time (3 min.) +/- perspective of storyteller +/- adequacy (tell the story from a set of pictures vs. invent the story based on pictures with two of them missing)
none 36 young adults,
intermediate-higher int. level of English (TOEFL scores)
each participant narrates all the stories
Oral narrative based on a sequenced set of pictures
Linguistic complexity:
· number of clauses per number of c-units Accuracy:
· number of error-free clauses Fluency:
· number of repetitions
· number of false starts
· number of reformulations
· number of hesitations and pasuses per total amount of speech
Significant effect of the +/- here-and-now condition on accuracy was found. The other measures and conditions did not yield significant differences or tendencies in learners’
performance.
Yuan & Ellis (2003)
+/- planning time (no planning vs. pre-task and on-line planning)
None 42 learners of
English, lower intermediate level (TOEFL scores)
Each participant narrates a story in one of the task conditions.
Oral narrative based on a sequenced set of pictures.
Linguistic complexity:
· ratio of clauses to T-units
· number of different grammatical forms Lexical complexity:
· mean segmental type-token ratio Accuracy:
· number of error-free clauses
· target-like use of verb forms Fluency:
· number of syllables per minute
· number of meaningful syllables per minute
Planning had a significant effect on fluency (a greater number of meaningful syllables per minute) and complexity (more clauses per T-unit). Increased accuracy was found in the on-line planning condition.
Ellis & Yuan (2004)
+/- planning time (no planning vs. pre-task and on-line planning)
None 42 learners of
English, lower intermediate level (TOEFL scores)
Each participant writes a story based on a set of 6 pictures under one of the planning conditions.
Written narrative based on a sequenced set of pictures.
Linguistic complexity:
· ratio of clauses to T-units
· number of different grammatical forms Lexical complexity:
· mean segmental type-token ratio Accuracy:
· number of error-free clauses
· target-like use of verb forms Fluency:
· number of syllables per total number of minutes used to complete the task
· number of dysfluencies (crossed out or changed words per total number of words
Online planning had a significant effect on accuracy (increased), and fluency (decreased, but with fewer dysfluencies) while it had no significant effect on complexity. Pre-task planning significantly increased fluency and resulted in fewer dysfluencies and a marked increase in syntactic complexity.
Gilabert (2005) +/- here-and-now (telling a story in the present with the possibility of looking at the picture prompts vs. telling a story in the past with no picture prompts) +/- planning time (no planning time vs. 10 min.)
Affective variables:
difficulty, stress, confidence, interest, motivation (measured by a 5-item questionnaire)
48 learners of English (lower-intermediate)
2x2 design Oral narrative
based on a sequenced set of cartoon strip
Linguistic complexity:
· S-nodes per T-unit Lexical complexity:
· Percentage of lexical words
· Ratio of lexical words to function words
· Guiraud’s Index Accuracy:
· number of error-free T-units
· target-like use of articles
· percentage of self-repairs
· ratio of repaired to unrepaired errors Fluency:
· number of syllables per minute
· number of pruned syllables per minute
Planning time had a significant effect on fluency, lexical complexity but not on structural complexity and accuracy.
Cognitive task complexity had a significant decreasing effect on fluency while accuracy increased on some of the measures (repairs). In terms of fluency and complexity learners benefitted more from planning time on the simple task. In terms of accuracy they did so on the complex task.
The affective variables of difficulty, stress and confidence correlated with fluency measures while interest correlated positively with self-repairs.
Kuiken &
Vedder (2007a)
+/-reasoning demands and +/- number of elements
L2 proficiency 91 adults (learners of Italian, beginner level) + 76 adults (learners of French, lower-intermediate)
cross-sectional, two complexity levels of tasks, two groups of participants based on language proficiency.
Written (letter to a friend) based on task prompts in L1.
Linguistic complexity:
· number of clauses per number of T-units
· dependent clause ratio Lexical complexity:
· Type-token ratio
· Alternative TTR (number of word types per square root of two times the number of tokens)
Accuracy:
· number of errors per T-unit (graded errors)
Significant effect of task complexity on accuracy (but not on the other measures). Significant effect of proficiency level on accuracy and lexical complexity
No interaction of task complexity and proficiency that is, the effects of cognitive complexity are not related to language proficiency.
Predictions of the Cognition Hypothesis partly supported, but no evidence was found for either Skehan’s or Robinson’s model.
Kuiken & +/-reasoning L2 proficiency 84 adults (learners of cross-sectional, repeated written Lexical complexity: The significant effect of task complexity on accuracy is
elements based on lg proficiency. L1. Accuracy:
· number of errors per T-unit (graded errorswith respect to Grammar, Lexicon, Orthography and Appropriateness)
explain.
Questions regarding Skehan’s and Robinson’s task complexity framework s are raised.
Michel, Kuiken
& Vedder (2007)
+/- number of elements (task complexity) +/- monologic (task condition)
None 44 L2 learners of
Dutch at intermediate level
2x2 design, 22 participants performed in a monologic and 22 in a dialogic condition
Oral tasks:
leaving a message on voicemail (monologic) or discussion on the phone (dialogic)
Linguistic complexity:
· number of clauses per AS-unit
· subordination index Lexical complexity:
· Percentage of lexical words
· Guiraud’s Index Accuracy:
· number of errors per AS-unit
· number of lexical errors per AS unit
· number of omissions per AS unit
· percentage of self-repairs
· ratio of repaired to unrepaired errors Fluency:
· number of syllables per minute
· number of pruned syllables per minute
· number of filled pauses per 100 words
Both complexity and condition had a significant effect on accuracy and fluency while task condition also had an effect on linguistic complexity. The complex task generated more accurate and fluent speech while the dialogic task triggered more accurate, more fluent, but structurally less complex output. No significant combined effects of complexity and condition were found.
CH partially confirmed.
Gilabert (2007) +/- here-and-now (narrating in the present with pictures prompts vs.
narrating in the past without picture prompts) +/- number of elements and perspectives +/- reasoning
Affective perception of difficulty, stress, confidence, interest, motivation L2 proficiency
42 lower and higher-intermediate learners
3x2 design Oral tasks:
Narrative Instruction-givin g Decision-making
Accuracy:
· number of errors per AS-unit
· ratio of errors per words
· number of error-repairs per AS-unit
· ratio of error-repairs per words
· all repairs per AS-unit
· ratio of all repairs per words
· percentage of self-repairs
· ratio of repaired to unrepaired errors
· corrected ratio of repaired to unrepaired errors
Task complexity has an overall significant effect across task types with the strongest effect being observed on the instruction-giving task.
No differences can be found in self-repair behaviour between low and high level learners.
Regarding affective variables, significant effects can be observed for difficulty, stress and confidence, the main difference being among the complex versions of the task types.
Tavakoli &
Foster (2008)
Storyline complexity (foregounded events and +/- background elements) +/- structured task (tightly structured storyline vs. loose structure)
Learning environment (ESL vs. EFL)
40 learners of English (London) 60 learners of English (Teheran) Intermediate level
2x2 design Oral narratives
using picture prompts.
Linguistic complexity:
· ratio of clauses to AS-unit
· words per AS-unit Lexical complexity:
· D value Accuracy:
· Percentage of error-free clauses Fluency:
· number of reformulations
· number of false starts
· word replacement
· repetition
· mid-clause pauses greater than 0.4s
· end-clause pauses greater than 0.4s.
Narrative structure had a significant effect by increasing accuracy in the tightly structured condition. Background elements generated more complex syntax. Also learners in London produced more complex language and more diverse vocabulary than learners in Teheran.
Foster &
Tavakoli (2009)
as in Tavakoli &
Foster (2008)
Native speaker performance compared to non-native in ESL (London) and EFL (Tehran) settings.
40 native speakers of English
as in Tavakoli & Foster (2008)
as in Tavakoli &
Foster (2008)
as in Tavakoli & Foster (2008) Storyline complexity generated more subordinated language in both NS and NNS. Narrative structure does not affect NS fluency.
Table 11. Studies on the interplay of cognitive task complexity and ID effects on L2 narrative performance
STUDIES ASPECTS OF TASK COMPLEXITY
OTHER VARIABLES
METHOD PARTICIPANTS DESIGN TASK FEATURES
MEASURES OF TASK PERFORMANCE
MAIN FINDINGS Niwa, Y. (2000) +/- reasoning demands
+ dual task (participants had to decide on the sequence of picture prompts and tell the story accordingly)
intelligence aptitude WM
22 adult learners of English
2x3 design (1 task complexity factor and 3 IDs)
Oral narrative based on 4 picture strips from Wechsler Adult Intelligence Scale (based on pictures depicting characters performing different actions, and the correct chronological sequence has to be found by inferring motives, intentions of the characters).
Linguistic complexity:
· S-nodes par T-unit Lexical complexity:
· Type-token ratio Accuracy:
· Percentage of error-free T-units
Fluency:
· Time on narrative
· Words per second
· Seconds between pauses
· Words per pause
· Words per T-unit
As tasks increase in their complexity, IDs in cognitive abilities increasingly differentiate performance in the area of fluency.
Higher working memory capacity and aptitude are associated with less fluency.
Albert & Kormos (2004)
+ planning message content
Creativity 35 young adult learners of English Intermediate level
The same narrative task is repeated twice with a two-month interval between them.
Participants have to invent a story based a on a set of pictures.
Oral narrative based on pictures.
Linguistic complexity:
· Clauses per AS-unit Lexical complexity:
· D-value
· Number of words Accuracy:
· Number of error-free clauses per total number of clauses
Narrative structure:
· Number of temporally ordered independent clauses per number of AS-units
Creativity has a moderate but significant effect on the quantity of talk produced (negative correlation) and the narrative structure (positive correlation). Creativity was found to account for about 10-15% of the variance in the students’ performance.
Albert (2008) +/- planning content L2 proficiency Creativity Aptitude
41 adult learners of English at upper-intermediate level
Participants performed two oral narrative tasks, completed a C-test and a TOEFL test, a standardized creativity test and the Hungarian version of the language aptitude test (HUNLAT)
Oral narrative based on picture cues
Linguistic complexity:
· Clauses per AS-unit Lexical complexity:
· D-value
· Plex lambda
· Number of words Accuracy:
· Number of error-free clauses per total number of clauses
Narrative structure:
· Number of temporally ordered independent clauses per number of AS-units
· Number of temporally ordered independent clauses
The more complex task resulted in less complex, less fluent and more accurate production which is probably due to the synergic effects of this task drawing simultaneously on both resource-directing and resource-dispersing dimensions of complexity.
IDs in aptitude were correlated positively with accuracy on the simple task and fluency on the complex task and lexical variety on both tasks.
The CH was not confirmed. Creativity affected the more complex task regarding originality and relative flexibility and measures of fluency, quantity of talk and lexical variety.
Robinson (2007b) +/- reasoning demands (on three levels)
Input/processing /output anxiety
42 adult learners of English
2x3 design Oral narrative task based on the Wechsler Adult
General CAF measures and Use of state terms
No significant effect of task complexity was found with the general measures.
complexity as well, low output anxiety is associated with more complex expression.
Guará-Tavares (2008)
+/- planning time Working
memory
50 learners of English at intermediate level
Two tasks performed under planning and no-planning conditions.
Results compared with a control group.
Oral narratives based on picture cues
Linguistic complexity:
· Clauses per c-unit Accuracy:
· Percentage of error-free clauses
· Number of errors per 100 words
Fluency:
· Words per minute (unpruned)
· Words per minute (pruned)
· Number of silent pauses per c-unit
· Percentage of total pausing time
Significantly more accurate and complex performance under the planning condition.
Working memory (as measured by the speaking span test) correlated significantly with fluency and complexity in the planned task.
Révész (in press b) +/- reasoning +/- number of elements
Language use anxiety Self-perceived communicative competence Linguistic self confidence
43 ESL learners 2x2 design Participants perform two oral argumentative tasks and complete questionnaires regarding the other variables
Linguistic complexity:
· Clauses per AS-unit Lexical complexity:
· D value Accuracy:
· Ratio of errors to AS-units
· Ratio of error-free AS-units to AS units
· Proportion of self-repairs to total number of errors
· Target-like use of conjoined clauses
· Target-like use of developmentally advanced conjoined clauses Interactional measures: number and types of language-related episodes (LREs)
Task complexity affected linguistic complexity negatively, but induced greater accuracy, lexical diversity and more specific developmentally advanced constructions. IDs in linguistic self-confidence only slightly affected performance with regard to lexical diversity.
Trebits (2010) +/- inventing content Aptitude Working memory
30 EFL learners, B2 level
2 TC factors, 2 ID variables
Participants perform two written narratives
Linguistic complexity:
· Ratio of subordinate clauses
· Ratio of relative clauses
· Clause length Lexical complexity:
· D value Accuracy:
· Ration of error-free clauses
· Ratio of error-free relative clauses
· Target-like use of past tense verbs
Lexical and syntactic complexity correlated positively with higher working memory scores.
3 specific and two general measures of accuracy and complexity correlated positively with different components of language aptitude. Ch only partially confirmed, L2 output on the complex task was not affected by IDs to a greater extent than on the simple task.
Table 12. Summary of the research focus of studies investigating the interplay ID and task complexity variables
Task complexity factors
L2 proficiency
Intelligence Working memory
Aptitude Creativity Anxiety
+/-here-and-now
Gilabert (2007) – oral tasks
+/- planning time
Guará-Tava res (2008) – oral tasks +/- reasoning Kuiken &
Vedder (2007a, 2007b) – written tasks
Niwa (2000) – oral tasks Robinson
(2007b) –oral tasks Révész (in press b) – oral tasks
+/-conceptualizing content
Albert (2008) – oral tasks
Albert &
Kormos (2004) – oral tasks
Albert (2008) – oral tasks
+/-conceptualizing
Trebits (2010) – written tasks
content
Table 13. Summary of L2 writing studies
Group 1: Studies on the effects of task complexity on written production
Study Task complexity variable Task type
Ellis & Yuan (2004) +/- planning time narrative Kuiken & Vedder (2007a) +/- reasoning
+/- number of elements
Argumentative Kuiken & Vedder (2007b) +/- reasoning
+/- number of elements
Argumentative
Gökgöz (2009) +/- reasoning
+/- number of elements Argumentative Trebits (2010) +/- conceptualizing content Narrative Group 2: Studies comparing L2 production across modalities
Martínez-Flor & Fukuya (2005)
None Argumentative
Grandfeldt (2008) None Narrative
Alcón-Soler (2009) None Argumentative
Bulté & Housen (2009) None written task: argumentative
oral task: narrative
Ferrari & Nuzzo (2009) None Narrative
Group 3: Studies on task complexity effects on L2 production across modalities
+/- number of elements
Table 14. Summary of results comparing L2 production across modalities– Groups 2 and 3 Performance
superior in written mode
Performance superior in oral mode
No effect of modalities on performance Lexical complexity Bulté & Housen
(2009)
Grandtfeldt (2008)
Alcón-Soler (2009) Kuiken & Vedder (2009)
G r a m m a t i c a l complexity
Ferrari & Nuzzo (2009)
Martínez-Flor &
Fukuya (2005)
Grandtfeldt (2008) Kuiken & Vedder (2009)
Accuracy Martínez-Flor &
Fukuya (2005)
Ferrari & Nuzzo (2009)
Grandtfeldt (2008)
Alcón-Soler (2009) Kuiken & Vedder (2009)