Speech Rate

UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

What makes listening difficult?
Factors affecting second language listening comprehension
Amber Bloomfield, PhD
Sarah C. Wayland, PhD

Purpose—To establish what is currently known about factors that affect foreign language listening comprehension, with a focus on characteristics of the listener, passage, and testing conditions.

Elizabeth Rhoades, MA
Allison Blodgett, PhD

Conclusions—Research on second language (L2) listening comprehension strongly supports the importance of a number of factors, for example, a listener’s working memory capacity and the number of ideas in a passage.
Much of the research, however, reports weak or inconclusive results, leaving many factors and complex interactions among factors unresolved and in need of further investigation.

Jared Linck, PhD
Steven Ross, PhD

Relevance—Identifying the factors that affect L2 listening comprehension will help Defense Language Institute Proficiency Test (DLPT) designers anticipate how qualities of selected authentic materials will impact listening comprehension. Executive Summary
Purpose
The U.S. Government administers the
Defense Language Proficiency Test
(DLPT) to military linguists and other government personnel to assess their listening and reading comprehension in a number of foreign languages, including critical languages such as Mandarin, Modern Standard Arabic, Egyptian
Arabic, and Persian Farsi. The DLPT is updated every 10 to 15 years, and the most recent transition—from DLPT IV to DLPT5—included a greater emphasis on testing listening comprehension with authentic materials. In turn, this has led to a growing interest in the factors that make second language (L2) listening difficult.
To examine these factors, CASL reviewed the current scientific literature and summarized the characteristics of listeners, passages, and testing conditions. The review targeted features of particular interest to stakeholders at the
Defense Language Institute (DLI). The long-term goal of the project is to support the selection of authentic listening
© 2010 University of Maryland. All rights reserved.

TTO 81434 E.3.1 | CDRL A017 | DID DI-MISC 80508B | Contract No. H98230-07-D-0175

materials that accurately reflect different proficiency levels.
Conclusions
Although the available research on L2 listening comprehension is limited,
CASL’s literature review identified several factors that affect listening comprehension. These factors are summarized below and in Tables 1, 2, and 3.

1

Characteristics of the listener

Understanding a foreign language taps several general cognitive abilities. For example, listeners with greater working memory capacity— that is, those who are most efficient at attending to, temporarily storing, and processing incoming information— understand more of what they hear when they are listening to their nonnative language.1 Further, listeners who effectively use metacognitive strate-

gies—that is, those who are aware of and use effective strategies, such as avoiding mental translation—demonstrate better L2 listening comprehension.2
In addition to these general cognitive abilities, a number of factors pertaining to experience with the L2 influence listening skill. These factors include the amount of prior exposure to the language; familiarity with and an ability to understand the non-native language’s phonology; vocabulary size; and background knowledge about the topic, text, structure, schema, and culture. Familiarity with the L2 changes the extent to which the L2 listener uses top-down or bottom-up strategies in listening. For example, expert listeners use both types of strategies: They are able to accurately make sense of the speech signal (bottom-up information)3 and integrate this information with
April 2010 i

What makes listening difficult?
Table 1. Effects of listener characteristics on L2 listening comprehension
Working memory

Greater working memory capacity correlates with better comprehension. Metacognitive strategies The use of metacognitive strategies improves listening comprehension. L2 proficiency and experience

As proficiency increases, the listener’s ability to correctly use bottom-up information (including deciphering the L2 phonology and vocabulary) improves. Background knowledge enables the use of top-down strategies to compensate for mishearing or encountering unfamiliar words, which can improve comprehension.

Anxiety

Anxiety negatively impacts L2 listening comprehension.

background knowledge (top-down information).4 By contrast, non-expert listeners attempt, often unsuccessfully, to use background knowledge to compensate for failure to understand speech sounds.5,6
Vocabulary size also impacts the extent to which L2 listeners will comprehend a spoken message, but this effect of vocabulary may be related to other more general qualities of listeners, such as their experience with the
L2.7
Listeners’ anxiety can also impact their ability to understand what has been said. If a listener is anxious or in some other way distracted and unable to pay attention, it will be more difficult to accurately determine what was said. of redundancy (e.g., exact repetition, paraphrase).13,14 Passage complexity also may affect
L2 listening comprehension. A higher number of negatives15 and the pres-

Table 2. Effects of passage characteristics on L2 listening comprehension
Length

• Overall length—Longer length increases listening difficulty, but the effect is weak and inconsistent across studies.
• Information density—A large number of ideas in a passage has a negative effect on listening comprehension.
• Redundancy—Repetition of information consistently improves comprehension, but whether the listener benefits depends on the type of redundancy (e.g., exact repetition, paraphrase) and listener proficiency.

Complexity

• Syntactic features—Simplifying sentence structure does not consistently improve comprehension. Negatives and infrequent vocabulary have a detrimental impact.
• Directness and concreteness—Passages with implied meaning can be more difficult to understand. Research in reading comprehension suggests that texts with more concrete objects or entities may be easier to comprehend, but little research has examined this factor in L2 listening. • Pragmatic information—The inclusion of L2 pragmatic constructs such as idioms and culturally specific vocabulary decreases comprehension. Organization

• Orality—Passages with higher orality—that is, ones more like unscripted conversations—have greater redundancy, more disfluencies, and simpler syntax. They are easier to understand than passages with less orality.
• Coherence—Overall coherence of a passage seems to have little effect, but only a few studies have examined its effects. Further, coherence may be difficult to define and measure objectively.
• Discourse markers—Words and phrases that signal the relationship between adjacent propositions and the overall structure of the passage improve comprehension. However, this effect depends on the type of marker. • Position of relevant information—Information is most easily recalled when it occurs near the beginning or at the end of a passage.

Auditory features • Speaker accent—Familiar accents are easier to understand than unfamiliar accents.
• Hesitations and pauses—Disfluencies, like hesitations and pauses, generally aid comprehension, especially for more proficient listeners.
• Noise and distortion—The presence of noise or distortion in the speech signal interferes with comprehension.
• Speech rate—How quickly someone talks can hurt comprehension, but slower speech rates do not necessarily help. L2 listeners may mistakenly attribute difficulties caused by other factors to a too-fast speech rate.

2

Characteristics of the passage

Studies directly examining the effects of passage length on L2 listening comprehension find little evidence that this factor alone affects comprehension difficulty.8 However, these studies have often explored a limited range of lengths9 or have confounded length with other factors.10 Information density (i.e., the number of ideas in the passage) and redundancy (i.e., the extent to which passage information is repeated), which are correlated with passage length, have more consistent effects. Information density increases listening difficulty consistently across studies,11 even when this factor is measured using different methods.12
Redundancy improves comprehension, but the effect depends both on the proficiency of the listener and the type

© 2010 University of Maryland. All rights reserved.

ence of infrequent vocabulary16 may increase difficulty. Further, simplifying the syntax of a passage does not consistently aid L2 listening comprehension.17 Indirect passages, which include more implied information, can also be more difficult for L2 listeners to comprehend,18 and the ability to cope with this type of information improves with L2 proficiency.19 Concreteness, or the extent to which a passage refers to concrete objects or entities, has rarely been explored as a factor affecting L2 listening comprehension,20 though it does affect L2 reading comprehension.21 Finally, L2 listeners have a harder time understanding passages

April 2010 ii

What makes listening difficult?

that contain culturally specific words22 and idioms.23
Several dimensions of passage organization affect comprehension, including orality—that is, the extent to which passages are similar to spoken language. Passages with higher orality have simpler syntax, greater redundancy, more hesitation markers (e.g.,“um” and “ah”), and more pauses;24 these types of passages are easier for L2 listeners to comprehend.25
Coherence is an additional dimension that can be characterized as the appearance of logicality26 in a passage or the extent to which ideas introduced at the beginning of a passage are carried through until the end.27 Research examining the effect of coherence on
L2 listening comprehension is sparse and generally inconclusive, and there are potential issues with defining this factor in a way that can be measured objectively.28 Discourse markers, which help to establish relationships between adjacent utterances (e.g., “yet”—a micromarker) and the overall structure of the passage (e.g., “the first point is”—a macro-marker), improve L2 listening comprehension.29 That said, there is some evidence that macro-markers make a passage more comprehensible, while micro-markers do not consistently help L2 listeners.30
Another feature of organization that affects listening comprehension is the position of the information that is necessary to answer a test question.
Information toward the beginning or at the end of a passage is more easily recalled than information from the middle of the passage.31
Several auditory features impact L2 comprehension, including the familiarity of the speaker’s accent. Accent familiarity affects passage comprehension for both first language (L1) and
L2 listeners, though more so for L2 listeners.32 The level of experience with an accent required to completely remove the detrimental effect seems to be extensive,33 though L1 listeners show partial adaptation after very brief

© 2010 University of Maryland. All rights reserved.

exposure.34
Disfluencies such as hesitations and pauses aid L2 listening comprehension in most studies examining these factors.35 Pauses give L2 listeners additional processing time and act as cues about the speaker’s upcoming utterances.36 Further, some evidence shows that disfluencies that occur in the L2 must be learned before filled pauses like “um” can be useful in listening comprehension, so listeners of different proficiency levels may benefit differentially from filled pauses.37
Noise or distortion in the audio signal interferes with listening comprehension for L138 and L239 listeners, though the effect is larger for L2 listeners. For L2 listeners, noise that most closely resembles the signal
(e.g., babble noise when listening to speech) presents the greatest challenge for listeners. Other types of distortion (e.g., white noise, filtering out high-frequency information [as often happens in telephone calls], and timecompression) differentially affect the perception and processing of speech.40
Faster speech rates, whether computer-manipulated or naturally produced, tend to have a negative impact on the comprehension of L2 listeners,41 even advanced listeners.42
However, some research suggests that L2 listeners will tend to feel the speech rate of aural materials is too fast when comprehension difficulty is caused by factors unrelated to speech rate.43 Further, while a faster speech rate may be detrimental to L2 listening comprehension, a slower speech rate

is not necessarily beneficial44 or even preferred by listeners.45

3

Characteristics of the testing conditions Research shows that imposing time limits makes cognitive tasks more dif-

ficult.46 However, very little research has directly examined the effect of time limits on performance in L2 listening comprehension testing, though imposing time limits on any test is likely to affect response behavior and test-taking strategies.47
Performance on L2 listening tests may be unaffected by increasing time limits if the examinee has the option of pausing or replaying the passage.48
In general, listening to a passage multiple times improves comprehension.49
The improvements may be greater for lower-proficiency listeners than higher-proficiency listeners,50 but only if they have the lexical and syntactic knowledge needed to comprehend the passage.51 When listeners are given control over the number and timing of hearings of the passage, they will choose to replay the passage more often when the passage is difficult due to factors like rate of presentation.52
For L1 listeners, note-taking is an effortful activity that introduces time pressure due to the difference between speaking rate and writing rate,53 and for L2 listeners, note-taking is even more cognitively effortful.54 Notetaking can be damaging to L2 listening comprehension when listeners are urged to take notes.55 Other factors that impact difficulty (e.g., speech rate)

Table 3. Effects of testing conditions on L2 listening comprehension
Time limits

Time pressure generally makes cognitive tasks more difficult, but time limit effects on L2 listening test performance are largely unexplored. Increasing response time may not improve comprehension unless examinees can pause or replay the passage.

Multiple hearings

Comprehension improves with additional hearings of a passage if the listener has the L2 knowledge to understand the information.
Examinees are more likely to replay a passage if other factors present difficulty.

Note-taking

Note-taking is particularly effortful when listening to L2 passages. It may benefit L2 listening comprehension, but only if the participant is able to make good decisions about when to take notes.

April 2010 iii

What makes listening difficult?

affect whether taking notes in the L2 benefits comprehension and recall.56
Overall, the literature indicates that if
L2 listeners are able to successfully employ a metacognitive strategy for determining when to take notes and when not to take notes, note-taking can benefit listening comprehension.
The effects of item type (e.g., multiple-choice, free response) and the task associated with listening (e.g., comprehension versus transcription) were not covered in the literature review.

8
9
10
11
12
13
14
15
16
17
18
19
20
21

Relevance
The current Defense Language Proficiency Test (DLPT5) includes a greater emphasis on authentic materials than prior versions of the test. This review of the scientific literature suggests that during test development and the selection of authentic spoken passages, it is possible to anticipate some of the ways in which passage, listener, and testing condition factors will influence L2 listening comprehension scores. The report provides an initial framework for assessing features of authentic spoken passages in relation to their impact on L2 listening comprehension.

22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39

Endnotes

40

For the full citations, see the technical details section of this report.
1 Harrington and Sawyer (1992)
2 Vandergrift et al. (2006)
3 Tsui and Fullilove (1998)
4 Field (2004)
5 Goh (2000)
6 Tyler (2001)
7 Nation (2001)

42

© 2010 University of Maryland. All rights reserved.

41

43
44
45
46
47

48

e.g., Rupp, Garcia, & Jamieson (2001)
e.g., Nissan, DeVincenzi, & Tang (1996)
Moyer (2006)
Rupp et al. (2001)
Rupp et al. (2001)
Chang & Dunkel (1992)
Chaudron (1983)
Nissan et al. (1996); Kostin (2004)
Kostin (2004)
e.g., Pica, Young, & Doughty (1987)
Taguchi (2008)
Taguchi (2005)
Nissan et al. (1996)
Freedle & Kostin (1992, 1993)
Sasaki (2000)
Kostin (2004)
Tannen (1982)
e.g., Kiany & Jalali (2006)
Odlin (1989)
Freedle & Kostin (1992)
Ying-hui (2006)
e.g., Jung (2003)
Chaudron & Richards (1986)
Freedle & Kostin (1996, 1999)
Major et al. (2005)
Weil (2003)
Floccia et al. (2009)
e.g., Blau (1991)
Watanabe et al. (2008)
Blau (1991); Watanabe et al. (2008)
e.g., Adank et al. (2009)
e.g., Lecumberri & Cooke (2006)
Aydelott & Bates (2004)
Griffiths (1990, 1992)
Rosenhouse, Haik, & Kishon-Rabin
(2006)
Moore et al. (2007)
Griffiths (1990)
Derwing & Munro (2001)
Siemer & Reisenzen (1998)
Bejar (1985); Verhelst, Verstralen, &
Jansen (1997)
Buck (2001)

49
50
51
52
53
54
55
56

e.g., Cervantes & Gainer (1992)
Field (2008)
Chang & Read (2006)
Zhao (1997)
Piolat, Olive, & Kellogg (2005)
Barbier & Piolat (2005)
Hale & Courtney (1994)
Carrell, Dunkel, & Mollaun (2002);
Lin (2004)

Corresponding Author and Reprints: Sarah Wayland,
PhD, University of Maryland Center for Advanced Study of Language, [phone], [e-mail], www.casl.umd.edu. Funding/Support: This material is based upon work supported, in whole or in part, with funding from the United
States Government. Any opinions, findings and conclusions, or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the University of Maryland, College Park and/ or any agency or entity of the United States Government.
Nothing in this report is intended to be and shall not be treated or construed as an endorsement or recommendation by the University of Maryland, United States Government, or the authors of the product, process, or service that is the subject of this report. No one may use any information contained or based on this report in advertisements or promotional materials related to any company product, process, or service or in support of other commercial purposes. The Contracting Officer’s Representative for this project is Shannon Salyer, PhD, Senior Research
Scientist—Research and Analysis Division, Defense Language Institute Foreign Language Center, [phone], [e-mail]. Acknowledgment: We wish to thank the following people from the DLIFLC: Donald C. Fischer, PhD, Provost; John
A. Lett Jr., PhD, Dean—Research and Analysis; Shannon
Salyer, PhD, Senior Research Scientist—Research and
Analysis; Mika Hoffman, PhD, Dean—Test Development;
James Dirgin, Director—Test Review and Education; Ruth
Mehr, Test Developer; Gerd Brendel—Test Development and Education Specialist; Nhon Le, Senior Project Manager—Test Development; and William Eilfort, Project Manager—Test Development. We also wish to thank Barbara
Forsyth, PhD, Area Director for Performance & Assessment at CASL, for her calm guidance, as well as her careful and thoughtful review of this document. Catherine J. Doughty,
PhD, Area Director for Second Language Acquisition at
CASL, was also invaluable, providing information about issues pertaining to second language acquisition in addition to reviewing this document. Last but not least, we are grateful to Michael Bunting, PhD, Associate Research
Scientist at CASL, for his insightful comments on an earlier version of this review.

April 2010 iv

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

Technical Details
TTO 81434 E.3.1



CDRL A017



DID DI-MISC 80508B



Contract No. H98230-07-D-0175

What makes listening difficult?
Factors affecting second language comprehension
Amber Bloomfield, PhD, Sarah C. Wayland, PhD, Elizabeth Rhoades, MA, Allison Blodgett, PhD, Jared Linck, PhD,
Steven Ross, PhD

L

istening is an important skill for students of a second language (L2) to master as part of learning to effectively communicate in that language (Dunkel, 1991; Lund, 1991; Rost, 2002; Rubin, 1994). It is tested directly in one

component of the Defense Language Proficiency Test (DLPT), which the U.S. Government administers to military linguists and other government personnel. The current report summarizes the scientific literature on three main characteristics argued to affect L2 listening comprehension: (1) characteristics of the listener, (2) characteristics of the passage, and (3) characteristics of the test-taking conditions. For each characteristic, the review targets particular

factors of interest established in discussion with stakeholders at the Defense Language Institute Foreign Language
Center (DLIFLC).
In most cases, only a handful of studies have explored the role of these factors in second language

Main characteristics

Factors of interest

listening. As a result, this review distinguishes between

Listener

Working memory, L2 proficiency, exposure to L2, metacognitive strategies, anxiety Passage

Authenticity, length, complexity, type and organization, auditory features Test-taking conditions

Time limits, number and control over hearings, note-taking

areas that have received enough attention to justify firm conclusions about their role in L2 listening comprehension, and those that have not. The review further highlights connections between the described factors. Ultimately, the research literature suggests that during test development and the selection of spoken passages, it is possible to anticipate some of the ways

in which listener, passage, and test-taking condition factors are likely to influence L2 listening comprehension.

© 2010 University of Maryland. All rights reserved.

June 2010 1

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

TABLE OF CONTENTS
Introduction ........................................................................................................................................................................ 3
Interpreting This Targeted Literature Review .................................................................................................................. 4
Characteristics of the Listener ......................................................................................................................................... 6
Working memory ............................................................................................................................................................. 6
Working memory and L1 comprehension ................................................................................................................... 8
Working memory and L2 comprehension ................................................................................................................... 9
Proficiency and experience with the second language .................................................................................................. 12
Vocabulary size ........................................................................................................................................................ 12
Phonological and grammatical information ............................................................................................................... 13
Background knowledge about the topic, text, structure, schema, and culture .......................................................... 14
Metacognitive strategies................................................................................................................................................ 16
Anxiety .......................................................................................................................................................................... 17
Characteristics of the passage ....................................................................................................................................... 18
Authenticity .................................................................................................................................................................... 18
Passage length and related factors ............................................................................................................................... 19
Length of passage .................................................................................................................................................... 20
Redundancy ............................................................................................................................................................. 23
Information density ................................................................................................................................................... 26
Differences between authentic and created texts in passage length, information density, and redundancy ............ 31
Working memory and passage length-related factors .............................................................................................. 31
Overall summary of passage length and related factors........................................................................................... 32
Passage complexity ...................................................................................................................................................... 32
Syntactic complexity ................................................................................................................................................. 33
Concreteness ........................................................................................................................................................... 35
Directness of text ...................................................................................................................................................... 36
Infrequent words ....................................................................................................................................................... 36
Culturally specific vocabulary and idioms ................................................................................................................. 38
Passage type and organization ..................................................................................................................................... 39
Passage topic ........................................................................................................................................................... 39
Passage type ............................................................................................................................................................ 40
Rhetorical structure .................................................................................................................................................. 42
Passage type, passage organization, and working memory ..................................................................................... 44
Coherence and relevance ........................................................................................................................................ 45
Auditory features of the passage ................................................................................................................................... 50
Speaker accent......................................................................................................................................................... 50
Distortion and noise .................................................................................................................................................. 54
Hesitation and pause ................................................................................................................................................ 56
Speech rate .............................................................................................................................................................. 60
Characteristics of the Testing Conditions ..................................................................................................................... 69
Time limits ..................................................................................................................................................................... 69
Number of and control over hearings ............................................................................................................................ 71
Note taking .................................................................................................................................................................... 75
Conclusion ....................................................................................................................................................................... 79

© 2010 University of Maryland. All rights reserved.

June 2010 2

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

INTRODUCTION

What makes listening difficult in a second language? Of the two receptive language skills, listening to a foreign language is understudied relative to reading (Johnston & Doughty, 2006; Lund, 1991; Osada, 2004; Rubin, 1994;
Shohamy & Inbar, 1991; Thompson, 1995). Most research concerned with the evaluation of second language (L2) comprehension has focused on reading rather than listening because the process of reading is more easily observed and manipulated (Osada, 2004). That said, research findings regarding reading comprehension often fail to map fully onto the processes involved in listening comprehension (Schmidt-Rinehart, 1994). For example, listeners have a worse memory for spoken information than readers do for written information, with proportionally more details recalled by readers and proportionally more main ideas by listeners (Lund, 1991). Further, characteristics of the listener influence listening performance differently than they influence reading performance (Park, 2004). In addition, there are factors that are important to listening that are not relevant for reading, like coping with a fast speech rate and disfluencies. In short, it is important to recognize that listening is a distinct skill from reading. Listening involves real-time processing, generally without the option of going back to earlier sections of the passage the listener may have missed
(Buck, 2001; Flowerdew, 1994). While slow readers can alter their reading speed without damaging comprehension, slow listeners may well miss information that cannot be recovered. Once the information is lost, it can be difficult to understand the rest of the passage (Buck, 2001). Further, while most reading involves complete control of the rate at which text is received (i.e., readers may read a text quickly or slowly, as they wish, unless some time constraint is applied), control over the speed of delivery for listeners varies much more widely (Osada, 2004). In a conversation, listeners may be able to exert some control over the speech rate of their interlocutor, while listening to a radio program provides no opportunity for control over the speed of delivery, and attending a professional lecture does so only with certain restrictions and considerable

Readers typically…

Listeners typically…

…reread text as they please …hear speech that disappears quickly

…control their own reading rate

…adapt to how quickly the speaker is talking

…encounter spaces between words

…have to figure out where words start and stop

…read text that obeys grammar and spelling rules …encounter disfluencies, false starts, and pronunciations that may or may not resemble written forms effort.
Aurally presented material also involves a number of phonological and lexical features that are not present in written material (Flowerdew, 1994). In some written languages, though not all, word boundaries are marked with clearly visible spaces. In spoken language these boundaries must be inferred from a variety of lexical and phonological cues (Cutler, Dahan, & van Donselaar, 1997; Leeser, 2004; Weber & Cutler, 2006). Also, written text tends to involve more planning and editing before it reaches an audience, whereas spoken information is often spontaneous (Richards, 1983). Speech, particularly in spontaneous conversations, contains irregular pauses, false starts, disfluencies such as um, and intonation patterns that may affect comprehension (Gilmore, 2007; Shohamy &
Inbar, 1991). The pronunciation of words may also differ greatly from the way they appear in print and may be

© 2010 University of Maryland. All rights reserved.

June 2010 3

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

affected by the words with which they are presented (e.g., assimilation results in the word ten being pronounced tem in the phrase ten bikes, Crystal, 2003; reductions result in the phrase I’m gonna go instead of I am going to go, Ito,
2001). Because of factors like reduction, spoken language may contain less lexical information than printed language, and these reduced forms have been shown to interfere with listening comprehension for non-native listeners (Ito,
2001). Differences between spoken and written texts may lead to a partial dissociation between reading skills and listening skills in the L2 learner (Lund, 1991; Song, 2008).
Given the differences between reading and listening, the current literature review focuses primarily on research in
L2 listening comprehension. Separate sections cover Characteristics of the Listener, Characteristics of the Passage, and Characteristics of the Testing Conditions, in that order.
INTERPRETING THIS TARGETED LITERATURE REVIEW

While reading this literature review, there are several important issues to keep in mind. First, because the literature review is targeted, not all possible factors affecting L2 listening comprehension are addressed here. For example, while we understand the importance of test question construction for assessing comprehension, we do not cover that topic in this review. Decisions about the factors to review were made through discussions between the authors at CASL and DLIFLC stakeholders. The absence of any particular factor in this literature review should not be construed as a dismissal of its importance by the CASL authors or DLIFLC.
In addition, this review is intended to present a language-general and high-level discussion of the impact of various factors on L2 listening comprehension. Thus, there are several caveats to consider when interpreting the research reviewed here:
1. The studies reviewed in each section often used very different types of tasks to measure comprehension. That is, while some studies measured comprehension using multiple-choice questions targeting the main topic or details of the passage (analogous to the method used in the DLPT), other studies evaluate comprehension through the use of assessments like the cloze test 1, recognition of statements from the passage, analysis of the listeners’ notes, in addition to many other dependent measures. A detailed investigation of the influence of task type on L2 listening comprehension and its interactions with other factors was outside the scope of this review. However, because the purpose of this literature review is to discuss factors that impact L2 listening difficulty in general, and the comprehension measures used in the studies reviewed here are diverse, the reader may well find that not all results apply to a comprehension task like that presented by the DLPT.
2. One of the most eclectic collections of measures in the studies reviewed here are those assessing L2 listening proficiency. Some studies defined proficiency by a score on an accepted test, such as the TOEFL, while others used experimenter-designed proficiency tests or simply the language course level of the student participant. Because of

1

A cloze test presents the test-taker with a short text with blanks where some of the words should be and asks them to fill in the blanks.

© 2010 University of Maryland. All rights reserved.

June 2010 4

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

these differences, it is very difficult to make generalizations about the effect of different levels of a factor (e.g., different speech rates) on the listening comprehension of listeners with varying L2 proficiency. It is also extremely difficult to make conclusions about what the literature indicates with regards to what factors will affect a listener at a particular Interagency Language Roundtable (ILR) level. The review notes any finding where a factor affected listening comprehension differently for listeners of differing proficiency levels, but these results must be interpreted with caution due to the aforementioned issues.
3. The level of difficulty of materials used in the studies varied, and factors affecting the difficulty level of a passage were often not fully described (e.g., the presence of discourse markers in a passage might be described in detail, while other qualities, such as vocabulary level, are not fully explicated). While this is not necessarily an issue for the investigation of the factor(s) of interest in the study, it does create some difficulty in interpreting the impact of the factor(s) on listeners with differing L2 proficiency. For instance, if introducing paraphrasing (a type of redundancy) into a passage has no effect on comprehension for lower-proficiency listeners, but does improve comprehension for higher-proficiency listeners, this may be because paraphrasing has no effect on the comprehension of lower-proficiency listeners. Alternatively, the lack of an effect of paraphrasing for lower-proficiency listeners could be due to the presence of other, uncontrolled qualities of the passage that increase its difficulty beyond the language abilities of these listeners (e.g., a fast speech rate or infrequent vocabulary). This example shows why findings indicating an interaction between proficiency level and another factor affecting listening difficulty must be interpreted carefully.
4. Finally, while the review is intended to provide a language-general discussion of how factors influence L2 listening comprehension, there are several factors reviewed here whose effects are very likely to differ across languages. In most cases, the factors reviewed here have been explored in very few languages, so there is no empirical evidence for such differences. Nonetheless, it is important to keep in mind that the predominant effect of a factor on L2 listening comprehension, as described in this review, may not hold for all languages. Further research is needed to address this issue.

© 2010 University of Maryland. All rights reserved.

June 2010 5

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

CHARACTERISTICS OF THE LISTENER

While a large number of individual difference factors may affect both L2 listening comprehension and general test performance, this review covers only the subset of factors deemed by CASL and the stakeholders as relevant to the question of difficulty of listening passages on the DLPT5. The factors discussed here include working memory capacity, proficiency and experience with the L2, the use of metacognitive strategies, and anxiety.
Understanding a foreign language taps general cognitive abilities, as well as knowledge acquired as the result of dedicated study and exposure to the non-native language (L2). General cognitive abilities known to affect L2 comprehension include working memory and metacognitive strategies such as planning, prediction, monitoring, evaluation, mental translation, personal knowledge, and directed attention. Working memory is correlated with, and maybe even central to language learning aptitude (Bowles, Linck, Koeth, Mislevy, Campbell, Annis, Jackson,
Doughty & Bunting, 2009); if we hope to measure language proficiency as separate from aptitude, it may be important to understand the interaction between working memory and passage difficulty. In addition, researchers generally assume that listeners use metacognitive strategies when listening to their non-native language; Vandergrift
(2006) offered this explanation when he found that native language listening ability accounted for 14% of the variance in L2 listening ability, with L2 proficiency accounting for 25% of the variance. In addition to the metacognitive strategies identified by Vandergrift and his colleagues, Rubin and Roberts (1987) found that L2 listening comprehension scores were influenced by exposure to literary works in the L2, critical thinking, and recall abilities. In addition to general cognitive abilities, a number of linguistic factors influence listening skill. These factors include (a) familiarity with and ability to understand the phonology of the non-native language, (b) vocabulary size, and (c) background knowledge about the topic, text, structure, schema, 2 and culture. The mental state of listeners can also have an impact on their ability to understand what has been said. That is, if a listener is anxious or in some other way distracted and unable to pay attention, it will be more difficult to process what was said. These factors are described in more detail in the sections below.
Working memory

Working memory refers to a cognitive system that is crucial to the processing, storage, and retrieval of information in memory (e.g., Baddeley & Hitch, 1974). Working memory

Working Memory: A set of cognitive processes that all listeners use—with varying degrees of efficiency—as they attend to, temporarily store, and process incoming speech in L1 or L2.

is thought to comprise a storage component and an attentional control component known as the central executive. According to Baddeley’s original model of working memory

2

In this context, schema refers to a cognitive framework or concept that helps the listener to organize and interpret information. Schemas can be useful, because they enable the listener to interpret a vast amount of information within the context of an existing framework.

© 2010 University of Maryland. All rights reserved.

June 2010 6

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

(Baddeley & Hitch, 1974), the short-term storage component is subdivided into the phonological loop (for auditory information) and the visuospatial sketchpad (for visual and spatial information), which serve as buffers where modality-specific information is maintained in a highly active and readily accessible state. Any mental manipulation or processing of information stored in these buffers is overseen by the central executive. Thus, the central executive is believed to play an important role in guiding processing within working memory.
More recent theoretical models of working memory emphasize the role of the central executive in exerting cognitive control over the contents of working memory, and also as the primary determiner of individual differences in working memory (e.g., Engle, 2002). Engle’s controlled attention view of working memory places a strong emphasis on the role of the central executive in efficiently managing available attentional resources. These contemporary models are motivated in part by a body of research demonstrating that working memory is related to an individual’s ability to successfully and efficiently resolve conflict or ambiguity, such as when recovering from inappropriate interpretations of ambiguous phrases in so-called garden path sentences (e.g., Novick, Trueswell, &
Thompson-Schill, 2005). 3 As briefly reviewed below, there is a rich literature demonstrating the relationship between working memory and L1 comprehension (e.g., Daneman & Merikle, 1996).
Although reading and listening occur in different modalities, the underlying cognitive tasks required for their performance involve some of the same processes: comprehending an incoming stream of information, integrating this information online, constructing a mental representation of the syntactic structure and semantic content (i.e., meaning), resolving any ambiguities or overcoming interference from conflicting representations, etc. Because most of the research on working memory and language has been conducted with written materials, this portion of the review will describe both reading and listening research to highlight the available data on working memory’s role in comprehension, while drawing particular attention to data from listening tasks where available. Note that when studying comprehension of spoken passages, one cannot return to earlier points in the input stream as is possible during reading comprehension, which may impose additional load on working memory. 4
Further, because L1 and L2 comprehension engage an overlapping set of cognitive processes, the next section briefly reviews the literature on working memory and L1 comprehension before focusing specifically on the available research on working memory and L2 comprehension. These parallel literatures provide an overall framework for understanding the role of working memory in L2 listening comprehension.

3

John knew the answer was wrong is an example of a garden path sentence. Readers initially misinterpret the answer as the direct object of knew (as in John knew the answer). They must reinterpret the noun phrase as the subject of the embedded sentence the answer was wrong.

4

Of course, if the listener is engaged in a dialogue with the speaker (a.k.a., participatory listening), the listener can ask the speaker to repeat any information the listener missed.

© 2010 University of Maryland. All rights reserved.

June 2010 7

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

Working memory and L1 comprehension

In a seminal study, Daneman and Carpenter (1980) introduced a new measure of working memory—the reading span test—that stressed both the storage and processing components of working memory. They found that individual differences in this processing-plus-storage measure of working memory were strongly related to measures of L1 reading and listening comprehension (correlations ranged from .42 to .90, with an average correlation of .66). Their study inspired a surge in research on working memory and L1 comprehension, much of which has supported the claim that working memory is related to L1 comprehension (e.g., Just & Carpenter, 1992; Engle, Cantor, & Carullo,
1992). In a meta-analysis of this literature, Daneman and Merikle (1996) examined the research findings from 77 studies combining data from over 6,000 participants. They found that, when using working memory measures that involved both processing and storage, the estimated population correlation between working memory and L1 comprehension ranged between .30 and .52. These correlations were notably stronger than the corresponding correlations between storage-only measures of working memory (range = .14 to .40), indicating that complex, processing-plus-storage measures of working memory are better predictors of L1 comprehension. 5
A survey of the literature highlights the fact that there are strikingly few studies of the relationship between working memory and L1 listening comprehension in adults. In one study, Baddeley and Hitch (1974) imposed a working memory load on participants by visually presenting sequences of three (low load) or six (high load) digits while having them listen to a prose passage. Participants were required to recall the digit sequence before answering questions about the passage. The results indicated that taxing working memory resources by imposing the high load, but not the low load, significantly impaired comprehension of the prose passage relative to a control condition. Other evidence of the role of working memory in listening comprehension comes from an investigation of the cocktail party phenomenon. This phenomenon refers to a situation in which an individual selectively focuses attention on one stream of auditory input while ignoring other input (i.e., the unattended stream), but when meaningful input (e.g., the listener’s name) is presented in the unattended stream, the individual’s attention may be captured (Moray, 1959). In a contemporary investigation of this classic phenomenon, Conway, Cowan, and Bunting (2001) found that individuals with relatively low working memory capacities more frequently experience the cocktail party phenomenon than individuals with higher working memory capacities. They argued that this phenomenon occurs because these lowworking memory individuals have difficulty blocking out or inhibiting distracting information during listening comprehension. 5

Studies examining individual differences in working memory often refer to the measured construct as short-term memory, particularly when the task at hand measures only storage capacity (as opposed to both storage and processing capacity).
However, other studies claiming to examine working memory also use tasks that measure only storage capacity (e.g.,
Carrell, Dunkel, & Mollaun, 2002). This review refers to both of these as working memory, and distinguishes between the use of measures of storage capacity only and those which measure both storage and processing capacity.

© 2010 University of Maryland. All rights reserved.

June 2010 8

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

The available evidence from listening-specific comprehension tasks suggests that the literature demonstrating working memory effects on reading extends to listening contexts. Indeed, there are clear theoretical arguments to suggest that working memory plays a critical role in listening (e.g., Engle, 2002). Working memory has been found to be related not only to reading comprehension, but also to other higher-level cognitive processes including reasoning ability (e.g., Jonides, 1995) and multitasking performance (e.g., Konig, Buhner, & Murling, 2005), and it has been argued that working memory is a general mechanism that underlies performance on many complex tasks (e.g.,
Conway, Cowan, Bunting, Therriault, & Minkoff, 2002). Given that these types of cognitive processes contribute to reading comprehension, it is clear that working memory is an important factor for L1 comprehension.
Working memory and L2 comprehension

Miyake and Friedman (1998) reported one of the few examinations to date of the role of working memory in L2 listening comprehension. They described a study performed by Miyake, Friedman, and Osaka (1998; cited in Miyake
& Friedman, 1998, p. 348) exploring the causal relations between working memory and L2 listening comprehension.
The original study revealed both a direct and an indirect impact of working memory on syntactic processing: participants with higher working memory capacity were able to make better use of syntactic information when comprehending the L2, and demonstrated a level of sensitivity to particular syntactic cues that was near native listener levels. The authors concluded that working memory span contributes to listening comprehension, both through influencing the ability to learn what cues are important in the L2 and through influencing how well the learner can make use of these cues during listening.
McDonald (2006) used a measure of working memory presented in the L2 of the participants 6 and found that performance on this L2 working memory measure correlated significantly with the accuracy of grammaticality judgments of spoken L2 sentences. In a second experiment, McDonald imposed a high working memory load on L1 listeners by requiring participants to maintain 7-digit numbers in memory. The study found that, when L1 listeners with an induced working memory load were asked to make grammaticality judgments about L1 sentences, they showed selective impairments in their judgments. The effect directly paralleled the performance of L2 learners on the same task (in the absence of an induced working memory load), suggesting that processing in the L2 imposes a load on working memory resources.
Other investigators have employed L2 reading comprehension measures to study the relationship with measures of working memory. Harrington and Sawyer (1992) assessed participants’ L2 working memory processing and storage capacity as well as L2 storage capacity alone (i.e., digit and letter span measures completed in the L2). Using

6

Participants in this study were late L2 learners who self-rated their L2 proficiency as M = 3.82 on a scale of 1-5 (the range of self-ratings was 2.5-5), with 1 indicating poor and 5 indicating excellent. They also self-rated their current frequency of L2 use as nearly equivalent to their current L1 use.

© 2010 University of Maryland. All rights reserved.

June 2010 9

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

TOEFL Grammar and Reading sections as well as a cloze test 7 as their measures of L2 comprehension, they found that performance on the L2 working memory test of both processing and storage capacity, but not the test of L2 storage capacity alone, significantly correlated with L2 reading comprehension scores, suggesting that the attentional control aspect of working memory is critical to L2 reading comprehension.
It is important to note that in both the McDonald (2006) and
Working memory measures that use L1 may avoid confounds with L2 proficiency.

Harrington and Sawyer (1992) studies, the measures of working memory were conducted in the L2, which may confound to some

extent individual differences in working memory with L2 proficiency. Although this issue is worth considering when selecting working memory measures, Osaka and colleagues (Osaka & Osaka, 1992; Osaka, Osaka, & Groner, 1993) found that L1 and L2 working memory measures in language learners were strongly correlated (r’s ranging from the mid-0.70s to mid-0.80s). L1 measures of working memory may provide a purer measure of working memory
(particularly in the context of L2 research); nonetheless, the results of Harrington and Sawyer provide important evidence of the importance of working memory in L2 comprehension.
There also is evidence that the effect of differences in working memory resources may be moderated by features of the input. Leeser (2007) had participants read texts about familiar or unfamiliar topics. Results showed that working memory was related to performance, but this effect differed depending on the task and the familiarity of the topic. On a recall measure of comprehension, greater working memory was related to better performance for familiar, but not unfamiliar, topics. However, on a verb form recognition measure (which compared recognition for verbs that have the same future tense they had in the passage and distractors having other forms [e.g., imperfect subjunctive]), the benefits of greater working memory capacity were instead found with unfamiliar topics. Walter (2004) found that working memory correlated with the transfer of L1 reading comprehension skills to an L2 comprehension task only in the more difficult condition where a pronoun and its referent were separated by multiple clauses (versus the easy condition, where the two occurred in the same clause). These findings suggest that the difficulty of the passage
(determined by factors like topic familiarity) and the difficulty of the task determine how large a role working memory capacity plays in level of performance.

7

A cloze test is a form of assessment consisting of a portion of a text with certain words removed which the test takers are asked to replace.

© 2010 University of Maryland. All rights reserved.

June 2010 10

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

It is worth noting that not all researchers are convinced working memory is critical to on-line L2 processing. In a study of reading comprehension involving native speakers of Japanese, Chinese, and Spanish, Juffs (2004) presented garden path sentences in the L2 (English) to participants using a self-paced moving window paradigm. This methodology tracks the amount of time a participant spends reading each word in a sentence. Juffs analyzed reading times at the disambiguating region, where the garden path is resolved by the

Source of characteristic Factor of interest Consensus in the literature

presentation of a word that disambiguates the previously ambiguous word. He found

Listener

Working memory Impacts L2 listening comprehension, particularly when other conditions, such challenging characteristics of a spoken passage, pose additional demands that several measures of working memory, including reading span and listening span, did not predict reading times in L2. He

concluded that other researchers’ claims of the importance of working memory processing capacity in L2 processing should be tempered in light of his findings. However, it is critical to note that Juffs also examined L1 comprehension with these same participants and failed to find any working memory effects. Given the robustness of working memory processing capacity effects on L1 comprehension across a variety of contexts and a wide range of populations (see Daneman & Merikle, 1994), this null effect raises questions regarding the stability of Juffs’s null result for working memory effects on L2 comprehension. Looking across the current literature on working memory processing capacity and L2 comprehension (both reading and listening), the pattern of results indicates that working memory is likely to impact L2 listening comprehension, and that these effects will be particularly strong in conditions that impose additional demands on working memory.

© 2010 University of Maryland. All rights reserved.

June 2010 11

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

Proficiency and experience with the second language

Although language proficiency is a variable in many studies examining L2 listening comprehension, it is defined in so many different ways that it is difficult to compare results across studies. One reason for this difficulty stems from the lack of standardized tests for determining proficiency level across languages. Most studies measure proficiency using age of acquisition, teacher judgment, course level, or performance on a non-standard test. In turn, the definition of high vs. low proficiency can vary from study to study, even when the same variable is measured; in one study, beginning language learners might be defined as students in their second year of study, whereas in another study, the same second-year students might be called intermediate learners. Although standardization of listening proficiency tests is still evolving, tests like those developed by the American Council of the Teaching of Foreign
Languages (ACTFL) and the Defense Language Institute could serve as a way to evaluate listening comprehension levels, though these would need to be widely available to language researchers. The lack of a standard definition of high and/or low proficiency is an important caveat for interpreting findings across the L2 comprehension literature.
The literature on L2 proficiency focuses on three types of knowledge—vocabulary size, phonological and grammatical information, and background knowledge.
Vocabulary size

An obvious factor that can influence comprehension of a spoken passage is the overlap between the listener’s vocabulary knowledge and the vocabulary of the passage. Nation (2001) makes a compelling case that listeners must have an adequate vocabulary to understand a passage in another language. Adequate vocabulary might be estimated by the number of words a listener needs to know to understand a representative sample of texts (a.k.a. text coverage).
The 5,000 most frequent words yield a coverage of 90 to 95 percent of the word tokens in an average passage in many languages, including Russian (Steinfeldt, 1965), French (Guiraud, 1954; Sciarone, 1979), English (Bongers, 1947;
Carroll, Davies, & Richman, 1971; Hirsh & Nation, 1992; Johnson 1972; Nation 1993; Palmer, 1931), and Dutch
(Nieuwbourg, 1992; Ostyn & Godin 1985; Sciarone, 1979; Vannes, 1952). Furthermore, Hirsh and Nation (1992) have argued that in order to understand all the main points in a text, readers need to be familiar with 95 percent of the words therein. There is no similar measure of the coverage required to understand a spoken passage; but we assume that if listeners know more than 5000 vocabulary terms, they are likely to have a good chance at understanding what has been said.
Nation (2001) also argued that vocabulary size is an indirect measure of other variables also known to influence listening comprehension ability, including world knowledge. These other factors are discussed in more detail below.

© 2010 University of Maryland. All rights reserved.

June 2010 12

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

Phonological and grammatical information

Research on the impact of phonological and grammatical information in spoken language comprehension has focused on whether high-ability and low-ability listeners use top-down and bottom-up processes 8 differently. Because words are not heard in isolation, but in specific contexts, both L1 and L2 listeners will use top-down processing strategies such as inferencing and elaboration to help make sense of a passage, particularly when they do not recognize every word in the input (Goh, 1998a, 1998b). Voss (1984) had users transcribe spoken passages in their non-native language and found that many listeners were extremely dependent on top-down information. In fact, the listeners who relied most heavily on bottom-up information made the most errors (Voss, 1984). Voss argued that less experienced L2 listeners rely even more heavily on top-down information because they are less familiar with the nonnative phonology, vocabulary, and syntax. To compensate for this lack of familiarity, they use higher-order cues to comprehend what they heard. Field (2004) describes a series of experiments by Koster (1987, cited in Field, 2004, p.
366) in which participants were asked to discriminate between words and non-words in their non-native language.
Koster found that words were correctly identified more often when preceded by a closely associated word. 9 The effect was biggest when the listener had lower proficiency in the language (as opposed to advanced or native proficiency).
In a second experiment, participants performed a word/non-word decision task. Non-native participants were faster to identify words that were preceded by a closely associated word, while native speakers showed the same effect, but to a lesser degree. It was as if the native listeners did not have to rely on context to help them decide whether the word was real or not, whereas non-native listeners relied more heavily on the context provided by the preceding word.
Tsui and Fullilove (1998) also observed that less-skilled L2 listeners are less able to monitor their top-down impression of a passage and modify it as necessary using incoming bottom-up information than are more-skilled L2 listeners. They analyzed the answers given by 20,000 people in Hong Kong who had to answer comprehension questions about different types of listening passages. Some of the passages were constructed to allow listeners to rely on a consistent schema introduced at the beginning of the passage as they listened, while other passages required revisions of the initial schema as more information was presented. The less-skilled listeners answered fewer questions correctly for the passages that required modification of the initial schema than did the more-skilled listeners. The authors concluded that this finding was due to less-skilled listeners relying more on top-down information (the schema activated initially) and failing to use bottom-up information (new information contradicting the initial schema) to modify the top-down information they applied in listening.

8

In this review, the term top-down processes refers to the use of information from the highest conceptual levels (e.g., inferencing, elaboration, integration, etc.) to fill in missing details at the lower levels. The term bottom-up processes refers to decoding information at the lowest level (e.g., acoustic-phonetic information) and using that information to progressively build higher-level representations.

9

A closely associated word is a word with a higher score on a “word association” test, wherein a group of participants respond to a given word with the first word that comes to mind. The more participants that generate a particular word, the more highly associated that word is to the given word.

© 2010 University of Maryland. All rights reserved.

June 2010 13

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

Goh (2000) asked 40 language students to describe the processes they used to understand the spoken L2, as well as the problems they experienced. She found that the more familiar the students were with the non-native language’s phonology, the more they relied on low-level (bottom-up) information. Indeed, she found that when L2 students do not know how to pronounce a word, they default to their native pronunciations. This is particularly evident with stress pattern differences in the native and non-native languages (hosTEL vs. HOStel), which, in addition to pauses, wordonsets, and perceptual salience, are used to segment the incoming auditory stream into words (Sanders, Neville, &
Waldorf, 2002). This complements the findings of Voss (1984) in that the more familiar the listeners are with the non-native phonology, the more they rely on phonological cues (bottom-up information).
The results so far are consistent with Cutler’s (2001) observation that L2 listeners tend to segment words in continuous speech on the basis of their usual L1 segmentation procedures. Indeed, when Weber and Cutler (2006) asked proficient German users of English as a second language to listen to nonsense sequences and respond whenever they heard an English word, they found that while the proficient L2 listeners used the permitted sound sequences of a language (the L2 phonotactics), there was still interference from the L1 phonotactics.
Interestingly, the inability of listeners to produce a phonological segment does not impair their ability to hear the phonology of their non-native language. Yamada and Tokura (1992) showed that while Japanese speakers can be trained to hear /r/ vs. /l/ many of them cannot produce this distinction.
Background knowledge about the topic, text, structure, schema, and culture

Listeners’ background knowledge about a passage can have a profound impact on their ability to understand what has been said. Without a schema, understanding a passage can be extremely difficult. A classic demonstration of this phenomenon was first reported by Bartlett (1932). He asked
Schema: A framework that helps the listener to organize and interpret incoming information.

participants to listen to passages in their native language. These passages (the most famous of which is War of the Ghosts)

described a sequence of events that seemed logical, but were often slightly illogical, with several subtle nonsequiturs. 10 Most people found it extremely difficult to recall the story exactly even after repeated readings; where the elements of the story failed to fit into the listener’s schemata, they were omitted or changed into more familiar forms.
This phenomenon demonstrates that the degree to which information in the passage conforms to the listener’s existing knowledge base determines how easy it is to understand.

10

Here is an example from the War of the Ghosts. “But presently the young man heard one of the warriors say, ‘Quick, let us go home: that Indian has been hit.’ Now he thought: ‘Oh, they are ghosts.’….” In this context, a non sequitur is a statement that does not follow logically from what preceded it.

© 2010 University of Maryland. All rights reserved.

June 2010 14

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

In the literature, authors often note the importance of shared knowledge

Source of characteristic Factor of interest Consensus in the literature

between the speaker and listener

Listener

Proficiency and experience with the second language

Most studies do not define proficiency in the same way, yet researchers agree that:

(Churchland, 1999). Shared knowledge is important because listeners are pragmatic by nature and, whenever

•

L2 listeners are more likely to rely on top-down processing when they have a weaker command of the phonology and grammar, factors that enable accurate bottom-up processing. •

L2 listeners may attempt to understand spoken passages using whatever background knowledge they may have, e.g., of the topic, genre, culture, and schemas, even when their knowledge is not complete or accurate.

possible, will situate themselves in terms of the topic, the setting, the event, the speaker, and the purpose for listening (Rost, 2005; Lantoff, 1999).
Vandergrift (2007) noted that L2 listeners will use prior knowledge

(including topic, genre, culture, and other schemas in long term memory) to build a conceptual framework against which they interpret what they hear. Context, non-verbal information, world knowledge, cultural information, and common sense are all used to build this framework. Bodie, Worthington, Imhof, and Cooper (2008) reported that other factors specific to a particular listener (specific knowledge about the topic, world knowledge, memory span, motivation, listening capacity) interact with aspects of the speaker (interactivity, status, power, role) and the passage
(objectives, purpose) to predict listening comprehension scores.
In his well-known “SPEAKING” model, Hymes (1972) identified eight situation-bound features of a message
(written or spoken) that require background knowledge on the part of the L2 listener; each message has the potential to be culturally specific. Presenting material that is incongruent with the listener’s expectation will result in confusion.
These features include:
•
•

the Participants in the interpretive community 11 (as described in Lakoff, 2000),

•

the Ends, or purposes of the communication (e.g., the desired outcomes and goals),

•

the sequence of communication Acts (which can be dictated by the message’s format and context),

•

the Key (register – formality, politeness, power relations),

•

Instrumentalities (channel, forms of speech),

•

Norms, and

•

11

the Setting/scene,

Genre (e.g., passage type).

Lakoff’s term “interpretive community” refers to a group of like-minded individuals who share similar assumptions about how a text should be understood.

© 2010 University of Maryland. All rights reserved.

June 2010 15

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

As described above, listeners with rich background knowledge use it to compensate for misunderstandings, unclear speech, and a lack of local or specific context from earlier parts of the passage (Goh, 2000). This use of background knowledge can be detrimental, however, because listeners that rely too heavily on prior knowledge, prior conversational units, and relational history with the speaker may be unduly biased by this information (Bodie et al.,
2008). That said, the advantages likely outweigh the disadvantages; using background knowledge to help understand a passage frees up attentional resources to be devoted to other aspects of the listening task (Tyler, 2001).
Metacognitive strategies

Metacognition is both self-reflection and self-direction. Reflecting on one’s thinking while listening, for example, can help listeners to adopt more appropriate ways to listen effectively. A number of studies have demonstrated the impact of metacognitive strategies on learning (Boekaerts & Niemivirta, 2000; Bolitho, Carter, Hughes, Ivanic,
Masuhara, & Tomlinson, 2003; Eilam & Aharon, 2003; Mokhtari & Reichard, 2002; Palmer & Goetz, 1988;
Schoonen, Hulstijn, & Boosers, 1998; Victori & Lockhart, 1995; Winne, 1995; Zimmerman & Schunk, 2001) and a few on second language reading skills. For example, Schoonen et al. (1998) described the positive effect of three types of metacognitive knowledge, including self-knowledge, task knowledge, and strategic knowledge, on the L2 reading comprehension of 6th, 8th, and 10th grade students. They found that the 8th and 10th grade students who reported using more metacognitive strategies had better L2 reading comprehension scores. Sixth grade listeners didn’t use as many metacognitive strategies. This may have been because they were struggling to understand the language and didn’t have the resources to effectively implement metacognitive strategies.
Vandergrift, Goh, Mareschal, and Tafaghodtari (2006) developed and validated a listening questionnaire (the
Metacognitive Awareness Listening Questionnaire or MALQ) that assesses the metacognitive awareness of second language listeners as well as their (reported) use of metacognitive strategies when listening to a spoken passage. Their work revealed five distinct factors that predicted listening comprehension scores on the University of Ottawa’s
Placement Test (Wesche, Paribakht, & Ready, 1996):
•

Problem-solving strategies: Strategies listeners use to make inferences and monitor them (e.g., using known words to infer the meanings of unknown words, using experience and general knowledge when interpreting the text; Rubin & Roberts, 1987).

•

Planning and evaluation strategies: Strategies listeners use to prepare themselves for listening, and for evaluating whether their efforts are paying off (e.g., having a plan for listening, keeping a goal in mind while listening; Hinkel, 2006; Richards, 1990; Vandergrift, 2003).

•

Avoiding mental translation: Beginning-level listeners often will mentally translate a passage as they hear it (Eastman, 1991), but they must overcome this strategy in order to become skilled L2 listeners
(Vandergrift, 2003).

© 2010 University of Maryland. All rights reserved.

June 2010 16

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

•

Self-knowledge: These strategies involve understanding the difficulty of the L2 listening task and being aware of one’s confidence levels and anxiety level when listening to the second language (Chemers, Hu,
& Garcia, 2001; Sparks & Ganschow, 2001; Wenden, 1991; Yang, 1999; Zimmerman & Schunk, 2001).

•

Directed attention: Strategies that listeners use to stay on task. These include recovering concentration when one’s mind wanders, focusing more when understanding falters, and not giving up (Rost, 2002).

However, just being aware of these strategies is not enough; Deci and Ryan (1995) observed that a listener’s motivation will affect whether he or she is able to use metacognitive strategies to help improve listening performance.
Anxiety

Listener anxiety can have a profound

Source of characteristic Factor of interest Consensus in the literature

Listener

Metacognitive strategies Promote comprehension during
L2 listening tasks through directing attention or avoiding unsuccessful methods, like mental translation

Anxiety

Can negatively affect comprehension, but may be ameliorated by testing conditions (e.g., the ability to take notes)

effect on comprehension abilities. When listeners are concerned that the message is too complex or that they will not be able to understand it, their ability to concentrate falters, and comprehension declines; this is true even in the native language. Listeners are more anxious when in a new situation,

listening to new information, trying to sort conflicting information, listening to seemingly illogical passages (Clark;
1989; Priess & Wheeless, 1989; Wheeless, 1975), or when they think their performance reflects their abilities or intelligence (Hussey, Teubner-Rhodes, Dougherty, & Novick, 2010). Indeed Hussey et al. (2010) found that anxiety had a direct impact on listeners’ abilities to resolve ambiguities in their native language in real time. Factors related to the testing conditions, such as the ability to take notes, may decrease anxiety during L2 listening comprehension
(Chang & Read, 2008). Ability to take notes and other characteristics of the testing conditions are discussed below.
Summary of research findings for listener characteristics
Factors with strong effects or convincing evidence
Beneficial to listeners: Greater L2 proficiency

Factors with sparse or inconsistent evidence
None in current review

Greater working memory capacity
Use of meta-cognitive strategies
Difficult for listeners: Anxiety

© 2010 University of Maryland. All rights reserved.

June 2010 17

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

CHARACTERISTICS OF THE PASSAGE

Although there are many passage-based factors that one could discuss in determining what makes L2 listening comprehension difficult, this review covers only a portion of them. The factors discussed include authenticity, passage length and related factors, passage complexity, passage type and organization, and auditory features.
Authenticity

Authenticity of aural materials can be defined in many ways, involving the speaker, the listener, the context, and the message (Breen, 1985). One prevalent way of defining authenticity for a passage itself, and the definition that this report adopts, is that an authentic passage is a piece of real

Authenticity has been defined in different ways. This review adopts the following definition:
A piece of real language that is created by a real speaker for a real audience in order to convey a message of some sort (Gilmore, 2007; Thanajaro,
2000).

language created by a real speaker for a real audience in order to convey a message of some sort (Gilmore, 2007; Thanajaro, 2000). 12 This characterization is designed to exclude passages that are created to exemplify some aspect of the language (e.g., the use of the future tense or speech acts such as apologies) rather than to convey an actual message. The characterization does include other types of passages such as speech from a native speaker to a non-native speaker and scripted television programs (Gilmore,
2007). This definition is similar to that currently employed by the Defense Language Institute, which defines authentic texts as “those which are produced by users of the target language and which are intended to be read [or heard] by other users of the target language in the target-language culture.” (Defense Language Institute Foreign
Language Center, 2010, p. 23). Note that “users of the target language” indicates that the receivers of the passage may be native or non-native listeners. Despite the fact that L1 speakers delivering a spoken message to non-native listeners may alter their speech in a number of ways, there are still likely to be differences between passages that are authentic and those that are created. For instance, Flowerdew and Miller (1997) noted several important differences (e.g., use of discourse macro-markers) between an authentic lecture delivered to L2 students and a taped commercial listening passage from an English for Academic Purposes textbook.
The push to use authentic materials in teaching second-language listening skills began in the 1970s (Gilmore,
2007). Apart from more general concerns that using created passages rather than authentic passages robs the L2

12

The quality of authenticity with regards to a second language listening passage may be better termed genuine
(Widdowson, 1976, as cited in Long & Ross, in press). A passage may be a genuine example of the L2 (e.g., a recorded telephone conversation between two friends), but most activities performed with the passage (e.g., responding to comprehension items after listening) are not authentic uses of the passage, so the task itself (listening and responding to items) cannot be termed authentic (Long & Ross). Although this review acknowledges the distinction between authentic and genuine, it adopts the term authentic to refer to a quality of the passage rather than of the task to accord with how it is often used in the literature.

© 2010 University of Maryland. All rights reserved.

June 2010 18

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

learner of experience with many elements of the L2 (e.g., lexical reductions like can’t, a realistic speech rate, disfluencies; Breen, 1985; Cobb, 2004; Crossley, McCarthy, Louwerse, & McNamara, 2007; Flowerdew & Miller,
1997; Long & Ross, 2009; Rings, 1986; Rogers & Medley, 1988), experimental evidence shows that L2 learners benefit from experience with authentic materials (Herron & Seay, 1991; Kienbaum, Russell, & Welty, 1986). Of primary concern to the current literature review, however, is just how authentic passages might differ from created passages in terms of passage-based and context-based factors of interest, and the impact these differences are likely to have on the difficulty of the passage for an L2 listener. Throughout this section, any research uncovering differences between authentic and created (e.g., textbook) listening passages will be discussed in the section describing the relevant passage-based factor.
Passage length and related factors

One factor of concern in L2 listening comprehension is passage length and the extent to which listeners can cope with the amount of information that is presented for processing (Alderson et al., 2006; Bejar, Douglas, Jamieson,
Nissan, & Turner, 2000; Carroll, 1977, cited in Dunkel, 1991, p. 440; Rost, 2006). Unlike reading, listening comprehension occurs in real time. Listeners may not have the option of going back to something they failed to comprehend (unless they can rehear the passage, see the section on Number and Control Over Hearings). Instead, the result of such a failure will be an inability to attend to new information as the listener invests additional time in attempting to understand what they missed, or an inability to comprehend later information because it relies on the understanding of earlier information (Goh, 2000; O’Malley, Chamot, & Kupper, 1989). Longer passages may be more likely to disrupt comprehension due to overwhelming listeners’ working memory storage capacity (Henning,
1990). In addition, the longer a passage is, the more information listeners could miss after encountering information they do not understand.
There is also reason to predict that longer passages will have a
Passage length has been defined with a number of measures, including syllables/second, duration (in minutes or seconds), and number of words or sentences.

greater impact on the listening comprehension of lowerproficiency listeners. Lower-proficiency L2 learners often try to understand a passage on a word-by-word basis; because word

comprehension in the L2 is slow and effortful, this strategy is generally maladaptive for lower-proficiency L2 learners
(Vandergrift & Tafaghodtari, 2010). As lower-proficiency listeners fixate on a particular word they missed, they may fail to attend to the continuing stream of information (Field, 2004; O’Malley et al., 1989; Vandergrift, 2003).
Researchers have used a number of measures to quantify passage length, including duration, number of syllables, number of words, and number of sentences, with inconsistent results at best. The lack of consistent results for passage length may be due to its relationship to other more predictive variables like redundancy (when information is presented more than once through repetition, elaboration, or other methods) and information density (the number of distinct ideas in a passage). Redundancy and density, in turn, may interact with passage authenticity and demands on working memory. In the sections that follow the review summarizes findings related to passage length, redundancy of information, and information density.

© 2010 University of Maryland. All rights reserved.

June 2010 19

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

Length of passage

Studies examining passage length generally find that passage length alone does not affect L2 listening comprehension difficulty. Thompson and Rubin (1996) reported that students complained about segments longer than
2.5 minutes as being too long for them to maintain concentration. However, Thompson and Rubin found that passage length interacted with other qualities of the passage: longer dramatic segments were better tolerated than longer news reports, suggesting that number of speakers, structure of the passage, or topic may interact with length. Nissan,
DeVincenzi, and Tang (1996) found no effect of overall passage length in seconds on listening comprehension item difficulty, and Kostin (2004) found no significant effect of the total number of words in the passage on comprehension difficulty. However, the passages examined in both these studies were fairly short (< 20 seconds in length). It is possible that the range of lengths was too narrow or the length of examined passages as a group too short overall to produce a reliable effect of length on item difficulty. Moyer (2006) found no significant difference between performance on comprehension questions corresponding to long passages (2–3 minutes) and those for short passages
(2–4 sentences) for either non-native or native speakers. However, passage length in Moyer’s study was fully confounded with passage type, since long passages were all informal dialogues and short passages were all formal news reports or announcements. The effect of passage length cannot be evaluated independently from the effect of passage type in her results.
Rupp, Garcia, and Jamieson (2001) explored the effects of passage word count and average sentence length on
L2 listening comprehension. An initial analysis indicated that longer average sentence length predicted more difficult listening comprehension items, and a second analysis found that both overall word count of the passage and sentence length contributed to item difficulty. However, the results for word count must be interpreted with caution, as analyses were done for reading texts and listening passages combined together rather than each modality separately.
Combining findings across the modalities may have led some factors to look as though they were important in determining both listening and reading comprehension item difficulty, when in fact they were important for only one modality. Rupp et al. (2001) also argued that average sentence length likely increased item difficulty due to the greater syntactic complexity of longer sentences, not because longer sentences simply provided more information to be processed. The syntactic complexity of passage sentences is another factor that likely correlates with overall passage length, though the two qualities are separable: a passage containing many short, simple sentences can be equivalent in length to another passage containing a smaller number of longer and more syntactically complex sentences. Greater syntactic complexity may increase listening comprehension difficulty for L2 listeners (Chaudron, 1983), but this may not always be the case if other features, like redundancy, are present (Long & Ross, 2009; see the section discussing Redundancy below).
In summary, studies examining overall passage length, whether measured by duration, word count, or number of sentences, do not generally find that this factor predicts listening comprehension difficulty. However, materials in these studies confounded factors like passage type and length, or examined only a narrow range of passage lengths with the average length being quite short. Rupp et al. (2001) did not perform analyses separately for listening and
© 2010 University of Maryland. All rights reserved.

June 2010 20

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

reading materials, leaving open the possibility that their findings obscured differences between the modalities.
Passage length may also be a better predictor of difficulty when the range of

Source of characteristic Factor of interest Consensus in the literature

Passage

Length

Effects of this factor alone are weak and inconsistent

considered lengths is wider, when longer passage lengths are examined, and when other important factors, such as the syntactic complexity of the passage, are controlled.
Working memory and passage length

There is reason to predict that working memory ability would interact with the effects of passage length on L2 listening comprehension. Demands on working memory are higher when processing is less automatic (Baddeley,
2007), and listening comprehension is less automatic for non-native listeners than for native listeners (Tyler, 2001).
This should result in L2 listeners generally experiencing greater demands on working memory processing when listening to an L2 passage. Further, the more information that must be held in working memory, the greater the strain on working memory storage (Baddeley & Hitch, 1974); thus, a passage containing more information should pose a greater challenge for working memory.
Henning (1990) examined individual differences in working
The scientific literature does not currently address whether passage length interacts with both working memory storage and working memory processing capacity.

memory storage capacity (measured using digit span 13) and passage length for effects on L2 listening comprehension. He found no relationship between working memory and the effect of

passage length on comprehension item difficulty; though he did find that longer passage lengths (30 words compared to 10 words) were associated with more difficult listening comprehension items. This latter result was statistically weak, however, and Henning’s study confounded passage length with the number of test items corresponding to a passage and with item type, making it difficult to draw conclusions about the effect of passage length on comprehension or its potential interaction with working memory.
Carrell, Dunkel, and Mollaun (2002) examined the effects of passage length, L2 listening proficiency (measured by the Institutional TOEFL listening comprehension section), and working memory storage capacity (using the digit span test) on L2 listening comprehension. Although they found no interaction of working memory with passage length on performance for listening comprehension items, they did find an interaction between L2 listening proficiency and passage length on performance. Higher-proficiency listeners (TOEFL score ≥ 49) did significantly better for items corresponding to short passages (~2.5 minutes long) than those for longer passages (~5 minutes long), while lower-proficiency listeners (TOEFL score < 49), who performed worse than higher-proficiency listeners overall, showed no effect of passage length. This suggests that lower-proficiency listeners were overwhelmed by even the shorter passages.

13

The term digit span refers to the number of randomly ordered digits a person can remember in the presented order.

© 2010 University of Maryland. All rights reserved.

June 2010 21

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

The findings regarding the relationship between differences in working memory and passage length suggest that, to the extent that longer passage lengths hurt L2 listening comprehension, it is not likely to be due to an increased working memory load. However, as noted above, measures of working memory that focus on storage capacity alone
(such as digit span) may simply not be the appropriate measure of individual differences in working memory for language comprehension (Daneman & Merikle, 1996). Research on L2 listening comprehension with measures of working memory storage and processing capacity is needed to address this issue.
Further, the findings of Carrell et al. (2002) point to another important consideration: passage length may only affect L2 listening comprehension once listeners have reached a particular level of proficiency. When L2 listeners are lower proficiency, they may be overwhelmed in trying to process even shorter passages, especially as the “shorter” passages

Passage length is related to other factors that can be separately manipulated, including:
•

Syntactic complexity of sentences

•

Length of material surrounding itemrelevant information

in Henning [1990], and < 20 seconds in Kostin [2004] and

•

Redundancy

Nissan et al. [1996]).

•

Information density

examined by Carrell et al. were quite long compared to those examined in earlier studies (~2.5 minutes compared to 10 words

Length of passage material surrounding item-relevant information

The degree of effort required to comprehend and integrate a particular piece of information is also likely to determine how likely it is to be comprehended. Buck and Tatsuoka (1998) found that L2 listening comprehension items were more difficult if there were more words surrounding the critical information (>8.3 words). Similar results were found by Brindley and Slatyer (2002). These results are consistent with the finding that information in the middle of a passage (which is both preceded and followed by additional words) is more difficult to comprehend and recall than information at the beginning or end of a passage (Freedle & Kostin, 1996, 1999), which will be discussed in the section on Discourse and Rhetorical Structure. However, this method of measuring the amount of information surrounding key information may have captured syntactic complexity of the sentence containing the key information rather than the amount of information, as Rupp et al. (2001) argued for their measure of sentence length. Although little research has explored the effects of length of the material surrounding key information on listening comprehension difficulty, this factor may be more predictive of difficulty for comprehending particular information in the passage than is overall passage length.
Summary: Length of passage

Studies examining the effects of passage length on L2 listening comprehension suggest that overall passage length is not a strong factor in determining passage difficulty, but this may be due to imprecise operationalization of length. Studies that do find a relationship find one that is weak (Henning, 1990) or qualified by interactions (Carrell et al., 2002). However, problems exist in terms of confounding other aspects of the passage with length (Moyer, 2006) or examining passages with a very limited range of lengths (e.g., Nissan et al., 1996). In addition, the range of lengths compared varied greatly across studies (e.g., 2.5 versus 5 minutes in Carrell et al., 2002, and 10 words versus 30 words in Henning, 1990). Several studies have also uncovered other factors more specific than overall passage length

© 2010 University of Maryland. All rights reserved.

June 2010 22

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

that affect difficulty, such as the length of information surrounding the item-relevant information (Buck & Tatsuoka,
1998).
Redundancy

Redundancy involves repeating key information through exact repetition, paraphrase, and elaboration (Chaudron,
1983). Across a variety of passage types (e.g., both conversations and lectures), speakers will circle back to previously introduced points to check for comprehension in their listeners, or simply to reiterate information they believe is most important (Field, 2008). Perhaps counterintuitively, redundancy is often classified as a form of simplification of input because it involves re-presenting information, thus giving the listener another chance to comprehend the information, sometimes in a form that is easier to process or retain (Chaudron, 1983; Parker &
Chaudron, 1987; Oh, 2001). Further, redundancy is argued to be a superior manner of simplification compared with syntactic simplification, which involves modifications like restricting the passage to canonical word order (e.g., Subject-Verb-Object is the canonical word order in English; Crystal, 2003) or using simpler

Redundancy: When information is re-presented through exact repetition, elaboration, or other methods

syntactic forms (Cobb, 2004; Parker & Chaudron, 1987; Pica, Young,
& Doughty, 1986).
Types of redundancy and L2 listening proficiency.

Redundant information (e.g., presenting a synonym of a previously given word, elaborating on an earlier point) may be expressed in several different ways (Chaudron, 1983). These forms differ in the complexity of their syntax and the psychological salience of their semantic relationship with the initial information. Simple repetition, using the exact same word(s) as stated earlier, is the least complex and most salient form, while giving a synonym is more complex and less salient (Chaudron, 1983; Chiang & Dunkel, 1992). The complexity and saliency of the redundancy form can affect how redundancy impacts L2 listening comprehension.
Chaudron (1983) examined the effects of five types of redundancy, of differing syntactic complexity and psychological salience, on L2 passage comprehension. He found that type of redundancy determined its impact on comprehension and interacted with L2 proficiency of the

Chaudron’s (1983, pp. 441-443) five types of redundancy in increasing order of syntactic complexity Initial utterance:
They are selling beer at the picnic
1. Simple noun

The beer tastes terrific.

2. Repeated noun

The beer…the beer tastes terrific lower-, medium-, and higher-proficiency groups (MTAC

3. Synonym

The brew tastes terrific.

scores of 30–52, 53–67, or 68–82, respectively). All

4. Topicalizing rhetorical question

What about the beer? It tastes terrific.

5. If-clause

If you afford the beer, it tastes terrific.

listener (measured by performance on the Michigan Test of
Aural Comprehension). Participants were divided into

participants showed the lowest level of performance on

© 2010 University of Maryland. All rights reserved.

June 2010 23

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

comprehension items 14 when topic redundancy was presented in the least salient form (synonyms). Higherproficiency participants showed greater comprehension than did medium or lower-proficiency participants when hearing more complex forms of redundancy (rhetorical question and if-clause), with lower-proficiency listeners benefitting most from repeated noun (high salience and low complexity). These results indicate that not all forms of redundancy are equally beneficial to listening comprehension for all listeners.
Later studies provided additional support to the findings of Chaudron (1983). Gainer (1997) found that dialogue passages where key information presented by the first speaker was echoed by the second speaker (Speaker 1: He was born in 1955. Speaker 2: Born in 1955.) yielded superior comprehension 15 for both higher- and lower-proficiency listeners, compared to an unmodified version of the same passage. This finding is consistent with Chaudron’s (1983) finding that redundancy in the form of exact repetition (low syntactic complexity and high psychological saliency) is beneficial for both lower- and higher-proficiency listeners. Chiang and Dunkel (1992) explored the effects of redundancy presented in the form of paraphrasing (“The food of the Pennsylvania Dutch Country is very hearty and delicious. Hearty and delicious food is nourishing and tasty,” p. 354), which is more complex than exact repetition
(Chaudron, 1983), and found that this redundancy improved comprehension 16 for higher-proficiency listeners
(Comprehensive English Language Test score 20–35) but not lower-proficiency listeners (CELT score 8–18). Again, this finding is consistent with Chaudron’s conclusion that more complex forms of redundancy are less likely to benefit lower-proficiency listeners.
The results for type of redundancy and its interaction with proficiency indicate that more transparent types of redundancy (e.g., exact repetition) are beneficial for lower-proficiency listeners, while higher-proficiency listeners can also benefit from more complex forms of redundancy like paraphrase. Lower-proficiency listeners may experience an increased working memory processing load if redundancy is not transparent, as they try to understand this information independently from what was previously given (Blau, 1991; Field, 2008; Rubin, 1994). Similar differences in the benefit of redundancy have been found between younger versus older children who are native listeners (Sonnenschein, 1982, cited in Anderson & Lynch, 1988, p. 51). This further bolsters the idea that less

14

Comprehension was measured in this study through performance on two types of items: (1) topic-related recall and (2) topic-related recognition. Recall items were verbatim sentences from the passage with a key word clozed (i.e., missing), to be filled in by the participant. Recognition items were statements for the participants to identify as having been presented by the passage or not.

15

Gainer (1997) measured comprehension through the number of correct responses to verbatim statements from the passage with one or two key words clozed, to be filled in by the participant.

16

Chiang and Dunkel (1992) measured comprehension with multiple-choice items. These items may have targeted the comprehension of main ideas, implied information, or details, but the authors did not specify which. It is possible that this comprehension measure was less dependent on the recall of exact details than were the measures used by Gainer
(1997) and Chaudron (1983).

© 2010 University of Maryland. All rights reserved.

June 2010 24

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

experienced listeners benefit less from redundancy (at least redundancy that is more complex and less salient than the simple repetition of topic nouns) than do more experienced listeners.
Interactions between redundancy and other passage factors

As will be discussed throughout this review, particular qualities of a passage, of the listener, or of the testing conditions can make the passage more or less difficult to comprehend. Some types of redundancy may be too complex for lower-proficiency listeners to benefit from (Chaudron, 1983). However, the presence of other qualities in the passage that alleviate comprehension difficulty may increase the chances that lower-proficiency listeners can more effectively use more complex types of redundancy.
Kelch (1985) investigated the effects of reduced rate of delivery and redundancy on listening comprehension.
Redundancy in this study involved presenting synonyms, hypernyms (e.g., animal is a hypernym of cat), and paraphrases of important ideas. Redundancy improved recall of equivalent-meaning words (i.e., words synonymous with the exact words from the passage), but only when the speech rate of the passage had been slowed to 2/3 its unmodified rate. As synonyms and paraphrase are more complex forms of redundancy according to Chaudron (1983), participants in Kelch’s study may have needed a slower speech rate to allow them to comprehend and make use of the more complex forms of redundant information.
In another investigation of redundancy effects and speech rate, Teng (2001) found a general effect of redundancy through paraphrase, but failed to find an interaction with speech rate. Paraphrase is generally described as more syntactically complex than other types of redundancy (Chaudron, 1983; Chiang & Dunkel, 1992), but the term is used somewhat loosely. Other studies have characterized paraphrase as the exact repetition of earlier-presented words in the same order (Gainer, 1997). Because Teng (2001) did not describe the method of paraphrase used, it is difficult to determine what was meant by redundancy, and Teng’s failure to find an interaction between speech rate and redundancy may be due to the use of a simpler form of redundancy. Another issue in comparing Teng and Kelch’s studies is that the speeds that were classified as fast and normal were not the same: Teng’s (2001) normal rate passages were considerably slower than those of Kelch (1985), and thus likely to be more easily comprehended by the participants. The issue of defining fast, slow, and normal speech rates in L2 listening comprehension is covered in greater depth in the section on Speech Rate.
Although findings in this area are not completely consistent, more complex forms of redundancy may become accessible to lower-proficiency L2 listeners when other qualities of the passage, like speech rate, make the passage easier to comprehend. The potential interaction between redundancy and speech rate should be addressed in future studies with more consistent speech rates.
Redundancy of item-relevant information

In all the studies described above, redundancy was manipulated through repeating, in some fashion, main or key ideas in the passage. Other research has directly explored the relationship between whether a particular idea is redundant in the passage and whether a test item for which that idea is relevant is easier or more difficult. The results of this research are quite consistent. In all cases, hearing information more than once when it is important for
© 2010 University of Maryland. All rights reserved.

June 2010 25

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

responding to a particular comprehension item decreases item difficulty (Buck & Tatsuoka, 1998; Freedle & Kostin,
1996, 1999; Ying-hui, 2006). Unfortunately, in these studies the forms of redundancy are not described in enough detail to determine their complexity or salience. Nonetheless, these results indicate that the comprehension of a piece of heard information, and its later recall, will be improved if the relevant information is repeated in some form in the
L2 passage.
Summary: Redundancy

Overall, the research examining the effects of redundancy on listening comprehension suggests that repetition or paraphrasing of information in the passage improves comprehension for that information. This increase in comprehension is true both for item-relevant information and other information in the passage. However, it is important to take into account how forms of redundancy with different complexity and salience affect comprehension across listeners of higher and lower proficiency (e.g., Chaudron, 1983). Lynch (1988; as cited in Rubin, 1994, p. 203) noted that the usefulness of redundant information may depend on its being recognized as a repetition of previously given information (see also Blau, 1990, and Field, 2008); in lower-proficiency listeners, more complex types of redundancy may simply add to the processing load (Anderson & Lynch, 1988;
Chaudron, 1983). In considering redundancy, it is also important to take into consideration its potential interaction with speech rate: redundant information in passages that are too fast for L2 listeners to comprehend may

Source of characteristic Factor of interest Consensus in the literature

Passage

Redundancy

Benefits of redundancy depend on the method used to represent information, and how this interacts with the proficiency of the L2 listener, and other characteristics of the passage

not benefit comprehension.
Information density

One rationale behind examining passage length as a factor in L2 listening comprehension is the belief that a greater processing load is introduced by a longer passage (Carroll, 1977, cited in Dunkel, 1991, p. 440; Henning,
1990; Rost, 2006). In service to this concern, the amount of information in a passage may be a more predictive factor for comprehension difficulty than is overall passage length. Information has been defined in the literature in several ways: content words (e.g., a noun, verb, adjective, or adverb, Nissan et al., 1996), the related concept of words with independent meaning (e.g., mother) as opposed to those without independent meaning (e.g., a, Gilmore, 2004), and propositions (the smallest unit of knowledge that can stand alone as a separate true-false statement, Dunkel, Henning,
& Chaudron, 1993). Measures of information density involve dividing the number of pieces of information in a passage by the total number of words or the duration of the passage. Sometimes measures of density only include those pieces of information that have not been previously given in the passage in the numerator (i.e., they control for redundancy, Aiken, Thomas, & Shennum, 1975), and so these measures directly capture the density of unique information in a passage. While information density will tend to be highly correlated with passage length in authentic passages, such that more dense passages will tend to be longer than less dense passages, the two factors could be

© 2010 University of Maryland. All rights reserved.

June 2010 26

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

independently manipulated and may vary separately in authentic speech depending on how information density is defined. Like passage length, greater information density is believed to make higher cognitive demands of L2 listeners, which may increase the effort involved in listening comprehension (Gilmore, 2004).
Information density as content word density

One way of defining a piece of information is as a content word (e.g., a noun, verb, adjective, or adverb; Nissan et al., 1996). Nissan et al. examined content word count alone as a measure of amount of information (i.e., they did not examine information density) and found no relationship between content word count and item difficulty. However,
Buck and Tatsuoka (1998) found that the average number of content words per idea unit in the area surrounding itemnecessary information predicted item
Some measures of information density include counting the number of propositions or content words and dividing this count by the total number of words in the passage.
Other measures take into account some types of redundancy (e.g., ratios of type/token or number of unique propositions/total propositions), though this may be misleading because the listener may not be aware that the information is redundant.

difficulty. 17 One potential reason for the difference between the results of Nissan et al. and those of Buck and Tatsuoka is that the passages investigated by Nissan et al. were short in length overall (5–20 seconds), and so

necessarily represented a limited range of content word counts. Buck and Tatsuoka reported a range of 4-20 content words per information unit; greater variation in content word counts could have increased the potential for finding an effect of this factor. In addition, Buck and Tatsuoka examined content word count in the area surrounding itemnecessary information, as opposed to an overall count of content words in the passage. They argued that the average amount of information in the portion of the passage containing the item-relevant information may be more important in predicting item difficulty than the overall amount of information in the passage.
In addition to the number of content words/idea unit in the area around the item-necessary information, Buck and
Tatsuoka (1998) also examined the proportion of content words to all words surrounding the item-necessary information (a measure of information density). They found that this factor was also a significant predictor of item difficulty, but they made no direct comparison of its predictive power and that of the content words/idea unit factor.
Nonetheless, their finding for the ratio of content words to all words surrounding the item-necessary information suggests that when information is defined as content words, item difficulty increases as information density increases.

17

Buck and Tatsuoka did not describe how they define idea unit, so it is difficult to determine if their count of content words involved entire sentences, clauses, or some other level of analysis. The authors cite Chafe (1985), who defined this term as “the amount of information a person can comfortably pay attention to and verbalize” (pp. 106). Because it is unclear how Buck and Tasuoka defined idea unit, it is possible that the length of the idea units, in words or duration, varied between passages. For this reason, their measure of content words/idea unit may not have been a pure measure of information density, in that it may not have controlled for length or duration.

© 2010 University of Maryland. All rights reserved.

June 2010 27

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

Rupp et al. (2001) employed another measure of information density using content words: type/token ratio. This is the number of unique content words divided by the total number of words in the passage. To estimate the number of types in the numerator, for example, the appearance of the words dog and dogs would be counted only once. In

Type/token ratio of a passage is defined as the number of unique words that are not from the same word family (e.g., dog and dogs are from the same word family) divided by the total number of words in the passage.

the analysis of L2 listening comprehension by Rupp et al.
(2001), type/token ratio emerged as a significant predictor of item difficulty, with test items for passages with larger type/token ratios being more difficult for the listener. This method of measuring information density provides an advantage over Buck and Tatsuoka’s (1998) measure of information density, in that type/token ratio controls for one type of redundant information: exact repetition of words. Because of the findings regarding the impact of redundancy on listening comprehension (e.g., Chaudron, 1983; Gainer, 1997), it is important to consider redundancy in addition to information density. As described below, other researchers have also taken redundancy into account when measuring information density.
Information density as propositional density

A proposition is the smallest unit of knowledge that can stand alone as a separate true-false statement (Dunkel,
Henning, & Chaudron, 1993). For example, The cat ate the meat can be expressed in multiple forms (e.g., The meat was eaten by the cat) that convey the same proposition (Crystal, 2003, p. 377). Further, a single sentence can convey multiple propositions (Those nice red apples cost a lot expresses the propositions the apples cost a lot, the apples are red, and the apples are nice; Crystal, 2003, p. 377). Some methods of defining propositions in the literature are more specific: for instance, Rupp et al. (2001) operationalized propositional density as the number of phrases in a passage containing a noun + attributive adjective + prepositional phrase (e.g., the fluffy cat on the table or that was a good suggestion you made in the meeting). A proposition is often the information that listeners remember from a text even when they cannot recall the exact wording of the presented utterances (Eom, 2006). Propositional density can be calculated by counting the number of propositions in a passage and dividing by the total number of words or the duration of the passage (Bejar et al., 2000; Rupp et al., 2001).
In Rupp et al.’s (2001) analysis, propositional density (categorized as beginner, intermediate, and expert based on the number of propositions per 100 words) emerged as an important factor in determining item difficulty for listening passages and reading texts, with greater density indicating greater difficulty. However, Rupp et al. did not examine the effects of propositional density separately for listening and reading comprehension, so it is possible that the strength of the relationship between density and item difficulty differed for reading and listening test materials.
Another interesting finding uncovered in this study was that the reading texts had significantly higher propositional density compared to the listening passages. While this finding may be due to their particular sample of passages and texts, it has been argued that speakers’ cognizance of the need for their listener to process information online generally leads spoken messages to be less propositionally dense (Shohamy & Inbar, 1991). If this finding applies to spoken and written texts generally, it suggests one reason why converting written texts into listening materials may be problematic: higher propositional density in a passage increases listening difficulty, and this type of density will tend

© 2010 University of Maryland. All rights reserved.

June 2010 28

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

to be higher for written texts presented aurally than passages originally delivered aurally (see also Gilmore, 2004).
This possibility requires more investigation in future studies.
Other methods of calculating propositional density also refer to redundancy of information in a passage like the type/token ratio mentioned in the previous section. Bejar et al. (2000) described two ratios capturing different aspects of propositional density: the ratio of unique propositions to the count of total propositions in the passage and the ratio of total propositions to passage duration. Because of the way the first ratio is calculated, if a passage includes the sentences The capital city of the state of Colorado is Denver. So, you see, Denver is a state capital, the proposition containing the information about Denver being a state capital, given in both sentences, would be counted only once towards unique propositions, and twice towards total propositions.
The two ratios proposed by Bejar et al. (2000) provide far more information about a passage than its total number of propositions (as in Rupp et al., 2001). However, they are better characterized as separate measures of density and redundancy than as both describing propositional density (though they are both described as measures of propositional density by Bejar et al.). For instance, a small ratio of unique to total propositions would indicate a large amount of redundancy, a characteristic likely to make the passage easier to comprehend (e.g., Chaudron, 1983). If this were combined with a large ratio of total propositions to passage length, then there would be a large amount of redundancy in a high-density passage. The ratio of total propositions to passage length in seconds provides a global measure of how quickly information is being presented. The ability to take into account redundancy (unique propositions divided by total propositions) as a separate factor from amount of information presentation (total number of propositions divided by passage length) makes the use of these two ratios preferable to other measures of propositional density that do not take redundancy into account.
A comparison of Bejar et al.’s (2000) ratios for four passages of the same length is shown in Table 1. Though
Passage A has lower propositional density than does Passage C, Passage C contains the same number of unique propositions as Passage A due to considerable redundancy. Passage C thus might not exceed Passage A in difficulty, though exactly how information density and redundancy trade off to determine passage difficulty is a topic that should be explored in future studies. Further, despite their equivalent level of density, Passage A is likely to be easier to comprehend than Passage B due to its higher level of redundancy.

© 2010 University of Maryland. All rights reserved.

June 2010 29

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE
Table 1. Using Bejar et al.’s (2000) measures of propositional density and redundancy, passages can have the same overall duration, but different levels of density and redundancy.
Density
measure:

Relative density Redundancy measure: Relative redundancy Passage label Unique propositions Total propositions Duration
(seconds)

Propositions/
Duration

Unique/ Total propositions A

5

10

20 s

1 prop/2 s

Lower

1 unique
/2 total

Higher

B

10

10

20 s

1 prop/2 s

Lower

1 unique
/1 total

Lower

C

5

20

20 s

1 prop/1 s

Higher

1 unique
/4 total

Higher

D

10

20

20 s

1 prop/1 s

Higher

1 unique
/2 total

Lower

While it may seem desirable to combine the ratios described by Bejar et al. (2000) into one by dividing the number of unique propositions by the duration of the passage, there may be good reason to separate redundancy and information density as factors. As discussed above in the section on Redundancy, not all types of redundancy may appear to be redundant information for all listeners. In particular, more complex types of redundancy, like paraphrase, may seem redundant to higher-proficiency listeners but not lower-proficiency listeners (Chiang & Dunkel, 1992). A measure of propositional density that includes only non-redundant propositions should be used only in those situations where there is good reason to believe that the listeners can appreciate the redundant information as such.
For example, if propositions that are exactly repeated are treated as redundant, dividing the number of propositions that are not exact repetitions by the duration of the passage may provide a better measure of propositional density than does dividing the sum total of all propositions by the duration of the passage.
Summary: Information density

Existing research on how information density impacts L2 listening comprehension indicates that greater density results in greater difficulty (e.g., Rupp et al., 2001). There are several methods of calculating information density in the literature, including some measures that account for the redundancy of information in a passage. Because redundancy tends to result in lower comprehension difficulty (e.g., Buck & Tatsuoka, 1998; Chiang & Dunkel, 1992), but can be confounded with information density in some measurements, future investigations of information density and L2 listening comprehension should be sure to use methods of measuring density

Factor of interest Consensus in the literature

that tease apart redundancy and density of

Source of characteristic information. Further, because information

Passage

Information density A large number of (unique) ideas in a passage has a negative effect density, redundancy, and passage length are heavily inter-connected, any examination of one of these factors should take the others into account.

© 2010 University of Maryland. All rights reserved.

June 2010 30

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

Differences in passage length, information density, and redundancy between authentic and created passages

One study investigated the extent to which passage length, information density, and redundancy differ for authentic and created aural materials. Gilmore (2004) examined transcripts of seven listening exercise dialogues from various language training textbooks, all involving service encounters (e.g., requesting a car from a car rental shop).
To collect authentic dialogues, he took questions from the information receiver (e.g., the person requesting the car) in these transcripts and used them as a basis for encounters in real situations with service providers (e.g., the clerk at the car rental agency).
Gilmore (2004) found that authentic dialogues were considerably
Qualities of Gilmore’s (2004) spoken dialogues

longer than textbook dialogues (M = 2,764 vs. M = 1,283 words). A

Authentic

Created

principle reason for the difference in length was that the authentic

More words

Fewer words

passages contained a great deal more repetition of information than did the

Higher redundancy Lower redundancy textbook passages (M = 24.14 vs. M = 1 occurrence of repetition).

Lower lexical density Higher lexical density According to the findings reported above (e.g., Chaudron, 1983; Chiang &
Dunkel, 1992; Gainer, 1997), this quality of the authentic dialogues would make them easier to comprehend than the textbook dialogues.

In addition, Gilmore (2004) found that authentic passages contained lower lexical density than textbook passages.
In this case, lexical density was the ratio of words with independent meaning (e.g., mother has an independent meaning, whereas a does not; Gilmore, 2004, p.367) to the total number of words in the dialogue (similar to the measure of density of content words used by Nissan et al., 1996). The finding that textbook passages contained greater lexical density than authentic dialogues is consistent with the finding of Rupp et al. (2001) that written texts had greater information density than spoken passages, and with the finding of a corpus analysis from Flowerdew
(1993) that planned passages (e.g., a radio news story) are more dense than unplanned passages (e.g., a phone conversation between friends). Thus, Gilmore’s (2004) findings indicate that textbook passages may inadvertently differ from authentic passages in ways that make L2 listening comprehension more difficult for the aurally presented textbook passages.
Working memory and passage length-related factors

The impact of the listener’s working memory capacity on listening comprehension is likely to be affected by the factors discussed in this section. In general, the more information that must be held in working memory during a task, the greater are the demands on working memory (Baddeley & Hitch, 1974). Greater information density should thus increase working memory demands during listening comprehension and leave less working memory processing capacity for other tasks, such as noting a particular character’s name. Redundancy, because it reinforces already-given information, may decrease working memory demands, particularly if the form of redundancy has low complexity and high salience or if the listener has higher L2 listening proficiency. These possible interactions between density and redundancy and working memory demands provide another reason why these factors should be considered in conjunction with passage length.

© 2010 University of Maryland. All rights reserved.

June 2010 31

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

Overall summary of passage length and related factors

Few studies present evidence that passage length itself increases L2 listening comprehension difficulty. Before discounting the role of passage length in listening difficulty, however, it is important to note that work in this area is limited in some ways. For instance, the range of passage lengths examined is often narrow, and some studies have uncovered differences in the effects of passage length on listening comprehension for lower- and higher-proficiency listeners. Some studies that do find evidence of a relationship between listening difficulty and passage length find that the amount of information surrounding the item-relevant information (in a testing context) has an effect, whereas overall passage length does not.
This section also explored two factors with which passage length is highly likely to be confounded: redundancy of information and information density. In the discussion of redundancy and its effects on L2 listening comprehension, it is clear that the effects are generally positive. More redundancy of the information in a passage decreases listening comprehension difficulty. This is true both when redundancy is examined for item-relevant information only, and when it is examined more broadly for key ideas or main points in a passage. These results suggest that, to the extent that a passage is longer because it contains more redundant information, the passage should not increase listening comprehension difficulty. However, the type of redundancy (synonyms vs. exact repetition of words), the L2 proficiency of the listener, and other factors that affect the difficulty of the passage (e.g., speech rate) should be taken into consideration, as these affect how redundancy influences difficulty.
The second factor confounded with passage length is information density. In contrast to redundancy, the findings for information density indicate that when density is higher, L2 listening comprehension is more difficult. To the extent that a passage is longer and has greater density, the passage will result in greater difficulty in listening comprehension. Further, while several measures of information density have been employed and suggested in the literature, those that take redundancy of information into account are likely to be preferable, particularly since redundancy and information density have opposing effects on difficulty.
Finally, a study examining differences between authentic and textbook dialogues shows that all the factors discussed in this section (length, information density, and redundancy of the passage) may differ depending on whether the passage is authentic or created. Care should be taken that L2 listening comprehension is not made more difficult due to decreased redundancy or increased information density in created passages.
Passage complexity

While the section above focused on factors concerning the amount of information a listener must process to comprehend a passage, the factors in this section relate to how challenging the information in a passage is to process.
While two passages may share the same degree of redundancy and level of information density, they may differ a great deal in other factors that can impact L2 listening comprehension, such as their demand for pragmatic knowledge, the concreteness of the information described, and the syntactic structure of the utterances.

© 2010 University of Maryland. All rights reserved.

June 2010 32

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

Syntactic complexity

One way of measuring the complexity of a passage is to consider structural elements of the phrases and sentences of the passage, or its syntactic structure. The studies that have investigated the impact of the syntactic structure of a passage have done so by considering the degree of subordination (Blau, 1990; Cervantes & Gainer, 1992; Pica

Measures of passage complexity refer to such dissimilar properties as syntactic structure, concreteness, and word frequency.
Additional measures appeal to the extent to which a listener must use pragmatic knowledge (e.g., culture, context). et al., 1987), the number of negatives (Kostin, 2004; Nissan et al., 1996; Yanagawa & Green, 2008; Ying-hui, 2006), the number of dependent clauses (Kostin, 2004; Ying-hui, 2006), or the number of references (Kostin, 2004; Ying-hui,
2006) in the passage.
Sentence structure

Blau (1990) investigated whether simplifying syntax or including surface clues for more complex sentences would affect L2 listening comprehension and found no significant effect of these manipulations. These results imply that modifying sentence structure (in terms of simplifying syntax and including cues to underlying structure) of aural passages does not impact second language learners’ listening comprehension. Pica, Young, and Doughty (1987) also explored the effect of syntactic modification on listening comprehension for scripted instructions. The results revealed that participants hearing scripts with lower syntactic complexity, but without interaction with the speaker, did not perform better than those exposed to unmodified scripts and the opportunity to interact with the speaker.
Some studies have found that L2 listeners benefit from syntactic
Simplifying sentence structure does not consistently improve comprehension.

simplification, however. Cervantes and Gainer (1992) found that listeners hearing a syntactically simplified version of a lecture scored

significantly higher on a recall test than did listeners hearing a more complex version of the lecture. In a second study, they replicated their first findings, but found a similar improvement in comprehension when a more complex version of the lecture was played twice. While these results indicate that syntactic simplification can improve listening comprehension, they also suggest that simplifying the syntax of a passage may not be necessary if listeners can hear a passage more than once.
Negatives, dependent clauses, and referentials

Several studies have investigated the impact of additional features related to syntactic complexity on the difficulty of listening comprehension test items. These studies suggest that negatives (e.g., negative markers like not and negative prefixes like un-) may play a role in listening comprehension, but suggest less of a role for features like dependent clauses or referentials.

© 2010 University of Maryland. All rights reserved.

June 2010 33

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

Nissan et al. (1996) found that the difficulty level of an item was significantly higher when the number of negatives in the corresponding passage was greater than one. 18 Kostin (2004) also explored the effect of negatives in a passage on item difficulty, in addition to the effects of referentials and dependent clauses, and distinguished between number of negatives in the first speaker’s utterance and the number in the second speaker’s utterance for dialogue passages. 19 She found that the presence of two or more negatives in the entire passage increased item difficulty and that a greater number of negatives in the utterance of the second speaker, but not in the utterance of the first speaker, was associated with difficulty for dialogue items. However, she included so few dialogue passages in the analysis (Kostin, 2004) that this result might be due to the particular dialogues she examined. Kostin (2004) did not find a relationship between dependent clauses or referentials in the passage and item difficulty.
Other, similar analyses of test passages failed to uncover any overall relationship between negations in the passage and item difficulty (Yanagawa & Green, 2008; Ying-hui, 2006). However,

Negatives may have a detrimental impact on L2 listening comprehension, but the effect is unlikely to be strong.

Yanagawa and Green (2008) did find that negatives affected item difficulty for certain items, such that items where the correct answer contained many of the same words as the passage were more difficult when there were more negatives in the passage, and items where an incorrect answer contained a lot of the same words as the passage were easier when there were more negatives. 20 Possibly listeners understood that some information in the passage had been negated, but were unsure as to what information the negation applied to, and so avoided answers having a lot of overlap with the passage. In addition to negatives, Ying-hui (2006) also explored how dependent clauses and referentials in the passages affected item difficulty, but found no relationship. 21
Summary: Syntactic complexity

The results from the existing literature on the effect of syntactic complexity on listening comprehension are mixed. Regarding overall syntactic complexity, Blau (1990) and Pica et al. (1987) both concluded that simplifying the syntactic structure of an aural passage does not improve second language learners’ listening comprehension.
Cervantes and Gainer (1992) found that learners hearing a syntactically simplified passage performed better on a recall test than those hearing an unmodified version of the passage, but also that hearing the unmodified passage a second time improved comprehension as much as hearing the simplified version. It is important to observe, however,

18

Data for Nissan et al.’s (1996) analyses were taken from TOEFL test results, so examinees would be expected to represent a range of proficiency levels.

19

Kostin’s (2004) data were taken from several sets of post-1995 TOEFL test results, so examinees would be expected to represent a range of L2 listening proficiency levels.

20

Participants for Yanagawa and Green’s (2008) study were recruited from the Test of English As International
Communication examinee pool. People from this pool should represent a range of proficiency levels.Yanagawa and
Green used prior TOEIC scores as a covariate in their analyses to control for L2 listening proficiency.

21

Data for Ying-hui’s (2006) analyses were taken from National College English Test of China (CET) test results for 1000 randomly-selected examinees. These examinees likely represented a range of L2 proficiency levels.

© 2010 University of Maryland. All rights reserved.

June 2010 34

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

that the studies exploring the effects of syntactic complexity on L2 listening did not explore how increasing the complexity of a passage (from some baseline text) impacted comprehension, but rather how simplifying syntax affected comprehension. Simplifying syntax may not be a good strategy to make passages more comprehensible for
L2 listeners, but increasing the complexity of syntax may nonetheless hurt their comprehension.
Negatives in the passage may increase item difficulty, but the effect is unlikely to be strong. While the results from Nissan et al. (1996) and Kostin (2004) imply that the presence of two or more negatives may increase item difficulty, and those of Yanagawa and Green (2008) suggest that negatives may increase the difficulty of particular kinds of items, Ying-hui (2006) failed to find any effect of negatives on difficulty. None of the studies investigating referentials or dependent clauses found a relationship between these factors and item difficulty, so there is no evidence to date that these factors will impact L2 listening comprehension.
Concreteness

Another characteristic of a passage that contributes to its complexity and may influence item difficulty is whether a passage is concerned with concrete entities or objects (concreteness). In the

Concreteness refers to whether a passage or text is concerned with concrete entities or objects versus abstract concepts.

existing literature, a passage has been considered to be concrete if there is a concrete object in the passage (Nissan et al., 1996) or if the main idea of the text and its development are concerned with concrete entities (Freedle & Kostin, 1992, 1993).
In reading comprehension, Freedle and Kostin (1992) found that texts in which the main idea and its development are concerned with concrete entities rather than abstract entities lead to easier inference and explicit statement comprehension items for L1 readers. In a similar study with L2 readers, Freedle and Kostin (1993) again found that greater concreteness in a text made its items easier.
However, not all conceptualizations of concreteness indicate that concreteness should make comprehension easier. Nissan et al. (1996) proposed that passages that involved references to concrete objects may be more difficult because L2 listeners may need to recognize the existence of the object in the setting of the passage in order to comprehend the passage. However, although they found a slight trend for passages that did not refer to concrete objects or entities to have easier items, there was no significant difference between those passages that referenced a concrete object in the speakers’ shared environment and those that did not. Further, their method of defining concrete was less about concrete versus abstract objects, and more about whether the listener had to understand that there was a particular object in the speakers’ context to make sense of what was being said. This study’s findings thus may not bear directly on how concreteness affects L2 listening comprehension difficulty.
Although there is almost no research on the effects of concreteness in L2 listening comprehension, findings in reading comprehension for L1 and L2 readers suggest that more concrete passages and texts should be easier to comprehend. Further, other research indicates that concreteness of words might impact comprehension through demands on working memory. The translations of more concrete L2 words are easier to recover from memory because they are usually easier to define and to contextualize (e.g., De Groot & Poot, 1997). Comprehending concrete
© 2010 University of Maryland. All rights reserved.

June 2010 35

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

words should thus require fewer working memory resources, and passages that contain more abstract L2 lexical items will impose additional working memory demands on the L2 listener, possibly injuring comprehension.
Directness of text

The level of directness of a passage is another factor that may impact its overall complexity and, therefore, the difficulty of the passage. Passages that are indirect tend to contain more implied information, requiring listeners to make inferences in order to comprehend the meaning of the passage. Conversational implicatures are one form of indirect communication that require the listener to infer what the speaker thinks and feels in order to arrive at the correct interpretation of the passage (Grice, 1975). There have been several studies investigating the comprehension of conversational implicatures by second language learners (Garcia, 2004; Taguchi, 2005, 2008).
Ability to comprehend implied meanings seems to be a quality of
Passages with implied meaning can be more difficult to understand, particularly at lower proficiency levels, as acknowledged in the ILR scale.

advanced L2 proficiency. Garcia (2004) compared the comprehension of conversational implicatures (e.g., turning down an invitation by describing a conflicting obligation rather than outright refusal) by

higher and lower proficiency L2 listeners and found that higher-proficiency learners performed significantly better than the lower-proficiency learners. Taguchi (2005) similarly found that L2 proficiency was a significant predictor of performance on a conversational implicature comprehension task. This result implies that the ability to comprehend conversational implicatures may be related to the level of the listeners’ second language proficiency. In another study,
Taguchi (2008) again found a significant correlation between L2 listening proficiency and accuracy in comprehending conversational implicatures. Further supporting the idea that experience with the L2 improves the ability to comprehend implicature, findings from this study revealed that participants were more accurate and faster in responding to more conventional implicatures (refusals) than less conventional implicatures (opinions, which vary more in how they are expressed). Taguchi argued that refusals, which include more common patterns of discourse within the L2, may be easier to comprehend because they require less processing effort compared to opinions.
The results from the existing literature suggest that L2 proficiency may be related to a learner’s ability to comprehend the implied meaning in indirect forms of communication such as conversational implicatures. This conclusion is further supported by findings that more conventional implicatures (which language learners are more likely to have exposure to) are easier to process than are less conventional ones. Although little research has specifically examined the impact of directness of a passage on comprehension, the relationship between directnessrelated factors and overall L2 proficiency suggests that lower-proficiency listeners may have difficulty when a passage contains a large number of implicatures or other indirect forms of communication, particularly if they are non-conventional. Infrequent words

The occurrence of infrequent words in a passage contributes to its complexity and may impact item difficulty.
Infrequent words in a passage may impact listening comprehension item difficulty because examinees are less likely

© 2010 University of Maryland. All rights reserved.

June 2010 36

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

to be familiar with low-frequency words, and so they may need to infer the meaning of any low-frequency words in a passage (for a more in-depth discussion of this process, see the section on Proficiency and Experience with the
Second Language). The studies that have investigated the impact of infrequent vocabulary on item difficulty have utilized lists such as Berger’s (1977) list and JACET 8000 (2003) to determine word frequency.
Nissan et al. (1996) found that the frequency of the words
(determined using a list including 100,000 words compiled by Berger,

Infrequent words have a negative impact on L2 listening comprehension.

1977) in a passage was related to item difficulty: the difficulty of dialogue items corresponding to passages that contained words not on the word frequency list was greater than the difficulty of the items for passages that only contained words on the list. This result implies that infrequent vocabulary may impact the difficulty of dialogue items. Further, the authors noted that certain high-frequency words that would be used on a university campus, such as semester and textbook, were not included on Berger’s list, which may have actually weakened the relationship between word frequency and item difficulty. Kostin (2004) considered both word frequency (using the same measure as Nissan et al.) and whether knowledge of the meaning of the infrequent word was necessary in order to answer the item correctly. The results of this study showed no significant relationship between the presence of an infrequent word in the passage and item difficulty, but there was a significant correlation between item difficulty and the presence of an infrequent word relevant to answering the item correctly.
These results suggest that it may not be the mere presence of an infrequent word that impacts the difficulty of a dialogue item, but whether the infrequent word is relevant to responding correctly to the item.
Some evidence shows that less-frequent words decrease the difficulty of a passage, but this is likely due to the way frequent is defined. The impact of the presence of infrequent vocabulary (determined using JACET 8000,
2003 22) in a passage on item difficulty was examined in Yanagawa and Green (2008). Their results showed that the presence of infrequent vocabulary in a text affected item difficulty, but unlike the findings from Nissan et al. (1996), it was found that less-frequent vocabulary in the passage was associated with easier items. This result is counterintuitive and the authors suggested that some of the words that were classified as being infrequent based on the JACET 8000 word list might have actually been more familiar to the test takers in their study than were the morefrequent words.
Research also points to a relationship between word frequency and working memory, at least for L1 listeners: listeners take longer to process low-frequency words relative to high-frequency words (e.g., Ferreira, Henderson,
Anes, Weeks, & McFarlane, 1996). Thus, if a passage contains a number of low-frequency words, this may impose additional demands on L2 listeners’ working memory as they attempt to recognize the low frequency word (i.e., access the lexicon). Past research shows that a greater working memory processing load will complicate listening comprehension (e.g., Baddeley & Hitch, 1974).

22

JACET 8000 is a word frequency list for Japanese learners of English.

© 2010 University of Maryland. All rights reserved.

June 2010 37

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

The bulk of research on word frequency and L2 listening comprehension suggests that infrequent vocabulary interferes with listening comprehension (Nissan et al., 1996). Whether understanding an infrequent word is important to answering a test item correctly has been found to have a significant relationship with difficulty when overall vocabulary frequency did not (Kostin, 2004). Thus, it is possible that earlier studies examining the relationship between vocabulary frequency and item difficulty were inadvertently measuring the effect of infrequent information that was necessary for answering the item rather than overall frequency of words in the passage. Yanagawa and Green
(2008) uncovered conflicting results, but the authors themselves argued that this finding might have been a result of the word list they used to determine frequency. However, all the studies discussed in this section have the limitation of using word lists to determine word frequency that may not be as reliable as originally expected; this limitation should be taken into consideration when interpreting the findings.
Culturally specific vocabulary and idioms

Another characteristic of a passage that may have an impact on its level of difficulty is whether the passage contains any culturally specific vocabulary or idioms, requiring the examinee to have some pragmatic knowledge in order to understand the passage and respond correctly to the associated items. Several studies have investigated this factor, by either considering the number of culturally unfamiliar words (Nissan et al., 1996; Sasaki, 2000) or the number of idioms in a passage (Kostin, 2004; Ying-hui, 2006).
Sasaki (2000) found that L2 readers completing cloze tests containing culturally familiar words (e.g., names more common in the culture of the L1) showed correct understanding of the key terms more often, tried to solve more items, and generally understood the text better than readers completing cloze texts with culturally unfamiliar words (e.g., names more common in the culture of the L2). L2 readers completing the cloze tests with culturally familiar words also performed better on the items requiring within-sentence information (e.g., information provided by the clause or sentence in which the item appeared). This result did not seem to extend to comprehension beyond the sentence level, however (e.g., information provided by the context of the paragraph containing the item or the entire text). These results indicate that replacing culturally unfamiliar words in a text with culturally familiar words increases examinees’ understanding of within-sentence information.
Research has also been conducted on the effect of culturally specific words or idioms on listening comprehension.
Nissan et al. (1996) found that the presence of culturally specific vocabulary in a listening passage could not be used to predict difficulty of the associated items. However, only a few items required comprehension of culturally specific vocabulary. The authors suggest that this is because test designers are sensitive to the fact that examinees may not be familiar with certain aspects of the L2 culture, and so they often include other clues in the passages.
Kostin (2004) explored the effect of idioms in the passage on listening comprehension. The American Heritage
Dictionary (2000) defines idiom as “an expression consisting of two or more words having a meaning that cannot be deduced from the meanings of its constituent parts.” An example would be snake in the grass—an expression that cannot be understood even if the L2 listener is familiar with the meanings of grass, in, and snake. In Kostin (2004), analyses uncovered a positive correlation between whether a passage contained an idiom, the meaning of which was

© 2010 University of Maryland. All rights reserved.

June 2010 38

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

central to answering the test item, and item difficulty. Again, this finding highlights that it is not simply the information in the passage that determines item difficulty, but whether understanding

Source of characteristic Factor of interest Consensus in the literature

Passage

Complexity

This factor corresponds to several distinct features, e.g., syntax, directness, concreteness, and word frequency. that information is important for answering a test item.
Although few studies have examined the effect of culturally specific vocabulary or idioms on L2 listening comprehension, results suggest that L2 listeners have a harder time comprehending a passage containing these types of features.

Directness and word frequency have the strongest effects.
Passages are harder when they require inferencing from the listener, as with indirect speech or unfamiliar vocabulary.

Passage type and organization
Passage topic

Topic of the passage is another characteristic that may affect how well L2 listeners comprehend the passage. In general, passages about familiar topics are easier for L2 listeners to comprehend than are passages about unfamiliar topics (Sadighi & Zare, 2006; Tyler, 2001). Exposure to information about a topic prior to listening to a passage about that topic improves comprehension for higher- and lower-proficiency L2 listeners (measured through the Test of English for International Communication; participants with scores ≥ 40 were classified as higher proficiency, those
≤ 39 were classified as lower proficiency; Chang & Read, 2006).
Another factor that may affect L2 listening comprehension is whether a passage is on an academic or nonacademic topic. The relationship between this factor and L2 listening comprehension difficulty has been explored in two studies (Buck & Tatsuoka, 1998; Ying-hui, 2006), but only Buck and Tatsuoka found a significant relationship between topic type and difficulty (non-academic topics were associated with easier items). However, it is difficult to say from this study’s findings exactly what about academic topics might make them more difficult than non-academic topics for L2 listeners. Many factors believed to constitute differences between passages covering academic and nonacademic topics are covered in other sections of this review: required background knowledge, ability to distinguish between relevant and irrelevant information, amount of implied meaning, ability to cope with long passage lengths, and note-taking demands (Ferris & Tagg, 1996). The key to the effect of academic versus non-academic topic on listening comprehension may be one of these factors rather than a benefit provided simply by the non-academic topic itself. In terms of more general differences between passages addressing different topic matter (e.g., humanities vs. mathematics lectures), differences of structure have been the focus of the relevant literature. For instance, Coulthard and Montgomery (1981) analyzed university science lectures to determine their structure and found that lectures are

© 2010 University of Maryland. All rights reserved.

June 2010 39

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

composed of free-standing informative syntactic clauses, which are combined into sequences, indicated by prosodic phrases. These sequences are combined to make transactions; each lecture is composed of a number of transactions.
While that structure seems to hold across a number of scientific disciplines, Rounds (1987) analyzed mathematics lectures and found a different discourse structure in which major points were named and explicitly marked by the teacher as being relevant for evaluation. Cohesion in mathematics lectures is developed using repetition and links to previous concepts, and topic changes are explicitly marked. Topics are presented in a very organized fashion using persuasion, with a question and answer format incorporated at appropriate points in the discourse. Further, the structures of scientific and mathematic lectures are quite different from those in the humanities where topics are developed from a variety of different perspectives with different interpretive frames (Strodt-Lopez, 1991).
Passage type
Lectures and recorded conversations

Structural differences can also be found between different types of passages. The research on passage types has focused, for the most part, on lectures and conversations, which have

Different types of spoken passages have different structure, which can make them easier or more difficult for L2 listeners to understand. very different structures. In a conversation, participants are allowed to ask for repetition and clarification, they follow turn-taking conventions, and they often do not distinguish between relevant and irrelevant information (Flowerdew, 1994). Furthermore, conversations do not necessarily require specialized knowledge or an understanding of implied and indirect speech acts (Flowerdew, 1994). They are by nature informal, contextualized, and involved (Biber, 1988). Furthermore, conversations have more repairs, negotiation of meaning, confirmation checks, and back-channel cues, all of which improve communication
(Chaudron, 1988).
On the other hand, when attending to a lecture, listeners hear long stretches of uninterrupted speech, without the opportunity to take turns or clarify, and they must be able to distinguish between relevant and irrelevant information
(Flowerdew, 1994). Furthermore, lectures generally require specialized knowledge, though they do not require that the listener be able to understand implied and indirect speech acts (Flowerdew, 1994). They are by nature formal, elaborated, decontextualized, and detached (Biber, 1988). Furthermore, lectures have more complicated syntax including that clauses, subordinate clauses, subordinate conditional clauses, first and second person pronouns, contractions, and the pronoun it (Tyler, Jeffries, & Davies, 1988), not to mention that the speakers often assume that listeners have prepared for the lecture by reading relevant material in advance.
However, even lectures can vary dramatically in style. Dudley-Evans and Johns (1981) identified four types of lectures, including (1) formal lectures, which are read from written copy (also called reading-style lectures); (2) less formal conversational-style lectures; (3) rhetorical lectures in which the lecturer acts more as a performer using a wide intonational range, many digressions, and shifts in key and tempo; and (4) participatory lectures (described by
Benson, 1989), in which the lecturer interacts with the audience, asking questions and soliciting input. The first three

© 2010 University of Maryland. All rights reserved.

June 2010 40

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

types of lectures are monologues in which the listener is not encouraged to participate; the last is processed by the speaker in real time and not composed off-line completely ahead of time.
Another finding that bears mentioning here is that comprehension of academic lectures is best when there is explicit discourse structuring indicated with terms like First let’s look at.... or What I will do now is.... (Camiciottoli,
2004). As mentioned in the discussion of discourse markers, discourse macro-markers such as these enhance comprehension (Chaudron & Richards, 1986). Further, these types of markers are more common in college-level lectures than in college-level textbooks, likely due to a high need for organization and structure in this complex type of spoken communication (Biber et al., 2004). This may be why reading-style lectures are more difficult to understand: Coulthard and Montgomery (1981) observed that reading-style lectures have fewer of these types of markers. Much effort has been devoted to characterizing the structure of lectures in different disciplines. Sadly, little research has examined whether one style is easier to understand than the other. Some studies, however, have examined the impact on listening of factors correlated with the different lecturing styles. This work is discussed below. Orality

One quality found to differ between some passage types (e.g., dialogues versus monologues) is the degree of orality. Orality is

L2 listeners have less difficulty understanding passages that are more oral. Such passages have simpler syntax, more disfluencies, and greater redundancy.

the extent to which a passage contains features of spoken language as opposed to features typical of written language, with highly oral passages tending to contain more disfluencies and redundancy, and simpler syntax (Tannen, 1982).
This factor differs between spoken passage types: a spontaneous dialogue between friends would be highly oral, whereas a formal lecture would be less oral (Inbar, 1988; Shohamy & Inbar, 1991).
Orality also tends to differ between authentic and created passages: authentic passages tend to have more oral qualities, in that they tend to contain more aspects of spontaneous language (e.g., ellipses, redundancy, pauses;
Shohamy & Inbar, 1991). The two terms should not be understood as interchangeable, however: both a spontaneous dialogue and a formal lecture are authentic by the definition adopted here. Further, a section from an authentic text read aloud can be highly authentic, in the sense that the text was created by a real writer/speaker for real listeners/readers to convey a real message (Breen, 1985; Rings, 1986), and also be low in orality, in that it would have few characteristics of spontaneous spoken language (Brindley & Slatyer, 2002).
Greater orality in a passage seems to improve L2 listening comprehension. Shohamy and Inbar (1991; see also
Inbar, 1988) explored the effect of passage orality on listening comprehension. Passages in this study included the same basic factual information, but differed on a number of factors relevant to orality: the extent to which they were planned (less planning = higher orality), potential for interaction between the speaker and the listener (more potential for interaction = higher orality), redundancy (more redundancy = higher orality), disfluency (more disfluency = higher orality), and sentence complexity (less complexity = higher orality). Degree of orality affected comprehension,

© 2010 University of Maryland. All rights reserved.

June 2010 41

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

with the passage with lowest orality (a news broadcast)
Feature

Higher orality

Lower orality

Planning

Less

More

the medium-orality passage (a lecture), and the high-

Potential for interaction with speaker More

Less

orality passage (a dialogue) producing the highest overall

Redundancy

More

Less

Disfluency

More

Less

producing the lowest overall performance, followed by

performance, although there was no significant difference between comprehension of the dialogue and the lecture.
In contrast, Brindley and Slatyer (2002) examined performance with a monologue passage (lower orality)

Sentence complexity Less

More

compared with a dialogue passage (higher orality) and found no difference between the two, although the authors

stated that comprehension of the dialogue may have been complicated by a higher relative speech rate. Kiany and
Jalali (2006) found significantly superior comprehension for a dialogue compared with a monologue and attributed the difference to the greater orality of the dialogue passage. It is important to note, however, that dialogues and monologues also differ in the number of speakers. Thus, it may be that the improved comprehension observed in
Kiany and Jalali (2006) was due to differences in the number of speakers and not due to the differing oral qualities of the dialogues. However, Shohamy and Inbar (1991) found no significant difference in performance between participants listening to a lecture and those listening to a dialogue; if the number of speakers determined their effects, comprehension of the dialogue would have been superior to that of the lecture. Further, Brown and Yule (1983) found that when there are more speakers it is more difficult to understand a passage. It will be important for future studies to control the number of speakers when investigating the effect of orality.
In relation to the discussion of lectures in the previous section (Dudley-Evans & Johns, 1981), orality is likely to differ across different lecture types. Formal lectures delivered from written material would have lower orality than would the other types of lectures (e.g., rhetorical or participatory) because they are more planned and involve fewer qualities of spontaneous language (Shohamy & Inbar, 1991). Further, conversations will be more oral than lectures of any type, because of their greater potential for interaction between the speaker and the listener and lower formality
(Biber, 1988; Flowerdew, 1994). Findings regarding the effects of orality on L2 listening comprehension suggest that non-native listeners would have greater difficulty with lectures delivered from written material than they would with lectures that were less planned, and greater difficulty with lectures than with conversations.
Rhetorical structure

Even though researchers have identified a number of different types of rhetorical structures (Meyer & Freedle,
1984), few studies have been devoted to determining whether one type is easier for listeners to comprehend than another. There are a few exceptions. Meyer and Freedle (1984) studied listening passages on the TOEFL and identified five types of rhetorical structures. The structures fell along a single dimension of degree of organization, with each succeeding structure incorporating qualities of the preceding ones. Listed in order of degree of organization, these are description, collection, causation, problems/solution, with the final structure, comparison, having the

© 2010 University of Maryland. All rights reserved.

June 2010 42

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

potential to vary in its degree of organization (Meyer & Freedle, 1984). Meyer and Freedle (1984) found that L1 listeners recalled the fewest number of idea units from a passage presented as a description, the least organized of the rhetorical structures. While their findings suggest that passages with less organized rhetorical structures will be more difficult for L2 listeners to comprehend, it is possible that L1 listeners were sensitive to differences in organization to which L2 listeners would not be sensitive.
Studies examining the relationship between rhetorical structure and L2 listening comprehension do not provide strong evidence that particular structures are easier to comprehend than others for non-native listeners. Ying-hui
(2006) examined passages with description, comparison, and causation structures, but did not find any relationship between rhetorical structure and difficulty. Freedle and Kostin (1996) found that items associated with less organized structures (according to Meyer and Freedle’s [1984] characterization) were actually easier: items for passages with a list structure were easier than those for other structures, and items for passages with a comparison or problem/solution structure were more difficult. However, Freedle and Kostin (1996) did not directly manipulate the rhetorical structure of the passages or their associated comprehension items (they were taken from a selection of testlets), leaving open the possibility that other factors, such as type of test item, were confounded with rhetorical structure. For example, list passages may have had fewer inference test items, which have been found to be more difficult than main idea or detail items (Nissan et al., 1996). Thus, it is difficult to conclude based on Meyer and Freedle (1984), Ying-hui
(2006), and Freedle and Kostin (1996) exactly how, if at all, L2 listening comprehension is affected by the rhetorical structure of the passage. However, other studies have found that certain rhetorical structures are associated with a greater number of syntactic features like negations. For example, causation passages contained more negations than did the comparison passages (Ying-hui, 2006). As there is some evidence that negations increase listening comprehension test item difficulty (see the section on Syntactic complexity), certain types of rhetorical structures which include more negations may be more difficult.
While evidence that particular rhetorical structures are easier to comprehend than others is weak, another possibility is that familiar structures are easier for L2 listeners to comprehend than are unfamiliar structures. Yang
(2007) presented one of two passages to a group of L2 listeners: one with a structure common in their L1, and the other with a structure common in the L2. Analyses of the listeners’ notes revealed no differences for the group hearing the familiar structure and the one hearing the less familiar structure. However,the author pointed out that requiring the participants to take notes (as this was the source of the dependent measure) may have interfered with their comprehension of both passages, preventing the discovery of an advantage for the familiar structure. This is in line with the findings of Hale and Courtney (1994), who found that participants compelled to take notes actually showed inferior listening comprehension relative to when they were not allowed to take notes at all (see section below on Note-Taking).
Position of item-relevant information

Some research supports the idea that position of the information in a passage predicts comprehension difficulty for that information. Freedle and Kostin (1996, 1999) found that when the information required by comprehension items occurred either early in the passage or in the last sentence of the passage, the item was generally easier.
© 2010 University of Maryland. All rights reserved.

June 2010 43

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

Conversely, when item-relevant information was located in the middle of the passage, items tended to be more difficult. Rupp et al. (2001) failed to find the same relationship between position and difficulty, but suggested that this may have occurred because they allowed their examinees to listen to each passage as many times as they wished.
Examinees could have approached additional hearings of the passage with the specific goal of attending to the information they had failed to comprehend the first time, removing any effect of position on comprehension. Yang
(2007) also found a significant effect for the location of the proposition in the passage, with participants being more likely to recall a piece of information if it occurred near the end of the passage than if it occurred in the middle or at the beginning. Yang’s findings are consistent with classic research showing a higher probability of recalling recent information (i.e., the recency effect) than information that occurred first (i.e., the primacy effect), though both types of information have a higher chance of being recalled than information occurring in the middle (Murdock, 1962).
Passage type, passage organization, and working memory

Some research suggests passage organization or type may impact listening comprehension because of an effect on working memory load. Presenting information in a more organized fashion makes this information easier to encode and maintain in working memory (Anderson, 2004; Baddeley, Lewis, Eldridge, & Thompson, 1984). Further, the relationship between working memory capacity and tasks involving reading comprehension or recognition differs depending on whether the topic is familiar or unfamiliar (Leeser, 2007). Findings such as these indicate that the role of working memory in listening comprehension is likely to be affected by the organization of the passage and its topic. When the passage topic is

Passage organization is likely to interact with working memory in L2 listening comprehension.

unfamiliar or its content is less organized, listening comprehension may be more difficult.
Summary: Passage type and organization

The effects of passage type on L2 listening comprehension are as diverse as the conceptualizations of passage type itself. Findings regarding orality and listening comprehension show that, when there is a difference, more oral texts like dialogues are easier for L2 listeners than more literate texts like news reports. The results regarding the influence of rhetorical structure on listening comprehension are less consistent. While L1 listeners show the best recall for information from more highly organized rhetorical structures (e.g., causations rather than lists; Meyer &
Freedle, 1984), L2 listeners either fail to show any
Factors such as coherence and the use of discourse markers may be better predictors of listening comprehension difficulty than rhetorical structure.

difference in listening comprehension with different rhetorical structures (Yang, 2007; Ying-hui, 2006) or show better performance for less organized structures,

such as lists, than more organized structures, such as comparisons (Freedle & Kostin, 1996). Other factors described below, such as the coherence of the passage, possibly including the number or type of discourse markers, may serve as better predictors of listening comprehension difficulty than the rhetorical structure of the passage. Indeed, Freedle and Kostin (1993) argued that examining the effect of the rhetorical structure of a passage on comprehension

© 2010 University of Maryland. All rights reserved.

June 2010 44

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

indirectly assesses the influence of different types of discourse markers on comprehension because specific discourse markers occur for each structure (e.g., lists use and and then; comparisons use however and yet).
Coherence and relevance

Some factors have been described in the literature examining language comprehension which attempt to capture how coherent or cohesive a particular passage is, and how this factor relates to comprehension difficulty. Carroll
(1986) argued that a series of meaningful sentences can nonetheless be combined in a way that makes no sense to the listener or reader if there is no overarching coherence (cited in Dunkel & Davis, 1994, p. 56). Coherence involves the appearance of logicality and relevance in a passage (Odlin, 1989). A passage will seem less coherent to the extent that it lacks strong, logical relationships between its propositions, and this may also be construed as the passage containing many propositions that seem off-topic or tangential (Odlin, 1989).
L1-L2 differences and coherence

Comprehension difficulties associated with coherence can arise from L1-L2 differences. Kaplan (2001) argued that chronic discourse organization differs markedly between English, Russian, Asian, and the Romance languages.
Differences in the typical manner of organizing speech could present issues for an L2 listener who comes from a language background with a different typical discourse organization. For instance, in Japanese, texts may be presented in ki-shoo-ten-ketsu form, which involves a shift away from the main topic to introduce a subtopic, while
English texts are generally more linear, so reading a Japanese text in ki-shoo-ten-ketsu form may be difficult and confusing for a reader accustomed to linear texts (Odlin, 1989). Similarly, the preferred style of speaking in Chinese is to put the topic at the end of an expository text, while English speakers tend to state the topic first (Yang, 2007).
Thus, while coherence is a factor that may differ between texts or passages (Freedle & Kostin, 1992; Ying-hui, 2006), it may also arise from L1-L2 differences in discourse patterns (Odlin, 1989).
Coherence effects and L2 listening comprehension

Research examining the relationship between coherence and L2 listening comprehension is sparse, and the findings are mixed. Ying-hui (2006) examined the effects of coherence on L2 listening comprehension using a coding method developed by Freedle and Kostin (1992) for reading texts: coherence was defined as a relative rating of to what extent the elements of the first sentence of the passage were represented in the rest of the passage, as compared with the other passages in Ying-hui’s sample (scored as 1 = minimal coherence, 3 = maximal coherence). In Yinghui’s study, higher coherence in a passage was
A passage is coherent when it has logical connections between its propositions. Due to differences in discourse organization norms across languages, a coherent passage could appear incoherent to L2 listeners.

associated with easier test items. These results suggest that the overall coherence of a passage might play a role in listening comprehension.

Nissan et al. (1996) examined what they referred to as local coherence, the explicitness of the connection between the speakers’ utterances. Passages containing explicit lexical links like repetition (e.g., Speaker 1: What time are you planning on leaving for the airport? Speaker 2: I’m leaving for the airport at 5:30.) or structural links like

© 2010 University of Maryland. All rights reserved.

June 2010 45

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

anaphora (e.g., Speaker 1: I hope Henry managed to catch his flight! Speaker 2: He just made it.), were predicted to be easier than those with more implicit links (e.g., Speaker 1: I heard it’s going to snow tomorrow. Speaker 2: Oh, no! I’m supposed to fly out at 7pm.). However, no significant relationship between coherence and item difficulty was found in their study.
The studies examining coherence do not provide strong evidence that coherence affects L2

Local coherence: The explicitness of the connection between adjacent utterances

listening comprehension difficulty. However, there

Global coherence: The cohesiveness of the entire passage as a unit

are methodological issues with both studies, as well as in making generalizations across them. In

Ying-hui’s (2006) study, the operationalization of coherence is highly subjective, and ratings of coherence were made relative to other study passages rather than some general definition of coherence. In Nissan et al.’s (1996) study, the operationalization of coherence did not distinguish between the types of links (e.g., lexical vs. structural). This is an issue because other research in the literature has described the difficulty presented by referentials like he and there for second language readers (Freedle & Kostin, 1993; Leow, 1993). Some links may have enhanced coherence and comprehension, but others may have made the passage more complex.
Further, the studies above examined two different types of coherence: local coherence and passage-level coherence. It may not be justifiable to equate these two factors. It is easy to imagine a passage that exhibits high local coherence and little overall coherence, such as a recording of a question and answer session with a celebrity. Every individual question and its answer will be strongly connected, but neighboring question and answer pairs may be completely disparate. In future research examining the impact of coherence on L2 listening comprehension, it will be important to distinguish between these two characterizations of coherence.
Discourse markers

An alternative way of examining coherence in passages is through studying the presence of discourse markers.
Discourse markers signal the rhetorical structure (e.g., comparison vs. list; Freedle & Kostin, 1993) of a passage, as well as highlight connections between adjacent propositions (Chaudron & Richards, 1986). The category of discourse markers includes macro-markers like my first point is and in conclusion, which provide clues about the overall structure of the passage, and micro-markers like yet, because, and in fact, which establish links between adjacent utterances (Chaudron &

Certain types of discourse markers, which help establish the structure of a passage and the links between adjacent utterances, may enhance coherence.

Richards, 1986; Dunkel & Davis, 1994;
Flowerdew & Tauroza, 1995).
Macro-markers are also a type of lexical bundle, a frequently occurring sequence of words with widespread use
(e.g., occurring 40 times or more per million words) that is not idiomatic (e.g., do you want to is a lexical bundle, while kick the bucket is an idiom; Biber et al., 2004). Lexical bundles include stance expressions that express attitudes or assessments (e.g., I don’t know what), referential bundles that directly reference a physical or abstract entity (e.g.,

© 2010 University of Maryland. All rights reserved.

June 2010 46

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

one of the things), and discourse organizers, which reflect relationships between prior and upcoming discourse (i.e., macro-markers such as on the other hand; Biber et al., 2004). In a corpus analysis by Biber et al. (2004), discourse organizers (i.e., macro-markers) were found to occur more frequently in college-level teaching than in conversation or textbooks, likely due to a high need for organization and structure in this complex type of spoken communication.
These findings indicate that particular types of passages may have more macro-markers than others, and spoken formal passages may tend to have more than written formal texts.
Research suggests that L2 listeners can recognize lexical bundles. Nekrasova (2009) found that, while native speakers recalled more lexical bundles verbatim from a passage than did lower-proficiency L2 listeners, higherproficiency L2 listeners recalled even more than native speakers, likely due to the rote memorization of lexical items demanded in the L2 students’ language courses (participants were assigned to proficiency groups based on their enrollment status as degree-seeking students of an Intensive English Program). Further, discourse-organizing bundles
(i.e., macro-markers) were recognized more frequently than were referential bundles for all three participant groups.
These findings show that macro-markers are salient to L2 listeners, and this increases with proficiency, though this salience may diminish as the listener approaches native-like proficiency.
Discourse markers have been found to improve comprehension of aural materials for L1 listeners (Hron,
Kurbjuhn, Mandl, & Schnotz, 1985, cited in Jung, 2003, p. 563). These markers appear to benefit L2 listeners as well.
Jung (2003) found that L2 listeners who heard a passage containing discourse markers recalled more information than participants who listened to a passage in which most of these markers were removed. However, other research finds different effects for different types of markers. Chaudron and Richards (1986) found that lower- and higherproficiency participants who heard a version of a passage with macro-markers showed superior performance to those who heard a baseline version without added micro- or macro-markers, a version with only added micro-markers, or a version with both types of markers added; the latter three conditions did not differ from each other. They suggested that macro-markers make a passage more comprehensible, but adding micro-markers may make a passage seem less organized, increasing the listeners’ cognitive load without providing useful information.
Dunkel and Davis (1994) contrasted the listening comprehension of native speakers with non-native speakers for lectures which were intact (evident condition) or had discourse markers like first and in contrast removed (nonevident condition). No difference in recall was found for native speakers or non-native speakers between those participants listening to the two versions. Although these results seem to contradict the findings of Chaudron and
Richards (1986), Dunkel and Davis (1994) did not distinguish between discourse macro- and micro-markers, a distinction found to be important by Chaudron and Richards (1986).
The usefulness of micro-markers is not a closed issue, however. Flowerdew and Tauroza (1995) argued that
Dunkel and Davis (1994) included too few micro-markers compared to what occur in authentic passages and too many markers associated with written rather than spoken language. They also pointed out that Chaudron and Richards
(1986) specifically inserted micro-markers in such a way as to minimize the semantic information they conveyed (i.e., discourse micro-markers in their study could only be acting as filled pauses [a.k.a. hesitation markers], see

© 2010 University of Maryland. All rights reserved.

June 2010 47

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

the Hesitation and Pause section for a more in-depth discussion). Flowerdew and Tauroza (1995) conducted their own study and found that recall and comprehension was lower for a passage from which discourse micro-markers had been deleted compared to one which had not been modified. Flowerdew and Tauroza’s (1995) results provide evidence that micro-markers, when allowed to contain semantic information (in contrast to Chaudron and Richard’s
[1986] study) and when presented in appropriate number and type (in contrast to Dunkel and Davis’s [1994] study), improve comprehension for L2 listeners.
Authenticity and coherence

Type and frequency of discourse markers may be one way in which textbook passages differ from

Authentic passages may contain a greater variety and more natural use of discourse markers than created passages.

authentic passages. Flowerdew and Miller (1997) analyzed a series of authentic passages and found frequent use of a variety of discourse micro- and macro-markers.
They argued that textbook passages are often too short to include some of the more global discourse macro-markers, such as those that refer to segments across long sections of a passage, and that the number of discourse markers included in textbook passages can appear too dense and unnatural when an effort is made to include markers
(Flowerdew & Miller, 1997). Authentic spoken materials may contain a greater variety of discourse markers and more natural use of these markers than materials attempting to emulate authentic passages.
Authentic and textbook passages may differ in other ways that are related to coherence. Authentic and simplified reading texts have been found to differ in causal cohesion (the extent to which the elements of the text are connected causally), and the density of logical operators, with authentic texts having greater cohesion and more logical operators
(Crossley et al., 2007). These results provide additional evidence that authentic materials may have greater coherence than created materials.
Relevance

Relevance, a factor related to coherence, can be defined as the extent to which the propositions contained in a passage are relevant to the main topic of the passage (i.e., textual relevance, van Dijk, 1978). Alternatively, relevance could be defined more narrowly as the proportion of propositions in the passage that are relevant to the test item at hand (if this item targets the main idea of the passage,
Relevance: An understudied area related to coherence and redundancy – It is the extent to which the propositions in a passage relate to or bear upon the main topic or particular test item

these two definitions of relevance should be interchangeable). The former operationalization of this factor should be strongly related to coherence as it has

been described above, while the second should be strongly related to redundancy of item-necessary information as examined by Freedle and Kostin (1996).
Relevance and its impact on comprehension is important to consider for listening in particular, as speakers are less explicit about connecting information to a central point or theme than are writers and often rely on the context in which the message is conveyed to provide this information (Smiley, Oakley, Worthen, Campione, & Brown, 1977).

© 2010 University of Maryland. All rights reserved.

June 2010 48

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

However, a review of the literature failed to uncover any studies directly examining the role of relevance in L2 listening comprehension.
Despite a probable strong relationship between coherence and relevance, the factor of relevance is also likely to be partially independent of the established definitions of coherence. Coherence is related to whether initial topics are carried through the entire passage without regard for whether these are central topics (Ying-hui, 2006), or the degree of connectedness between adjacent propositions (Odlin, 1989). Relevance, on the other hand, takes into account the overarching theme of a passage. However, the presence of discourse markers should have an enhancing effect on both coherence and relevance, as these markers can establish the relevance of a piece of information in a passage to what has already been stated (e.g., on the other hand to introduce a counterpoint) and build coherence by highlight the overall structure of the passage (e.g., my first point is).
The effect of relevance on L2 listening comprehension is likely to be related to the listener’s proficiency. In the same way that lower-proficiency listeners may have difficulty in perceiving that redundant information in a passage actually is redundant (Blau, 1990), these listeners may also have difficulty

Source of characteristic Factor of interest Consensus in the literature

Passage

Type and organization Rhetorical structure, one measure of passage organization, does not show strong effects on listening comprehension. recognizing that irrelevant information is actually irrelevant. Distinguishing between irrelevant and relevant information is of high importance for listening comprehension, particularly for academic lectures
(Flowerdew, 1994). For these reasons, relevance is an important factor to consider in listening comprehension research in general, and that in the L2 in particular.

Spoken passages are easier to understand when they are more oral, more closely resembling spoken rather than written language. Discourse markers play a major role in establishing organization and coherence.

Summary: Coherence and relevance

Little research has directly examined the role of coherence in L2 listening comprehension. Overall coherence seems to be associated with improved comprehension, but the operationalization of this factor is a problem. A great deal of research, however, has explored the effects of discourse markers on L2 listening comprehension. These markers have been found to enhance comprehension in English, but there is some disagreement about the extent to which different markers are helpful. 23 Authentic lectures have been found to contain more discourse markers than textbook lectures. In estimating the difficulty of L2 listening passages, it may be important to take the presence of

23

It should also be noted that Flowerdew and Tauroza (1995), Dunkel and Davis (1994), and Chaudron and Richards
(1986) all explored the effect of discourse markers on comprehension for L2 learners of English; it is possible that the usefulness of discourse markers for L2 listeners will vary with the language being learned as well as with the type (micro versus macro) of marker.

© 2010 University of Maryland. All rights reserved.

June 2010 49

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

discourse markers into consideration: more markers (particularly macro-markers) should make the passage more coherent and easier to comprehend and recall, though this might only be true in English.
Auditory features of the passage
Speaker accent

Comprehending spoken language involves adapting to the idiosyncrasies of a particular speaker (e.g., speaking rate or the pitch of voice). In general, listeners are quite good at this skill, but adaptation becomes considerably more challenging when the speaker has a different accent than the listener (Weil, 2003). Accented speech has been found to affect both the extent to which listeners successfully retrieve a speaker’s message and the effort involved as listeners identify particular words in the message (Floccia et al., 2009). In the case of a speaker with a different accent, a listener must cope with variation arising from both the speaker’s own idiosyncrasies and additional variation the speaker shares with others from the same linguistic background (Weil, 2003). Such difficulties in adaptation are further exacerbated when the spoken language is not the listener’s native language, particularly when proficiency in that language is low.
Accent versus dialect

Oftentimes, the terms accent and dialect are used interchangeably in the literature to describe phonological differences in speech. For example, General American English may in the same discussion be described both as a different accent than Southern American English or as a different dialect (e.g., Clopper & Bradlow, 2008). Although the two terms are often used synonymously, Crystal (2003) defined dialect as a regionally or socially distinctive variety of language, identified by a particular set of words and grammatical structures (p. 136, emphasis added), while defining accent as the cumulative auditory effect of those features of pronunciation which identify where a person is from, regionally or socially (p. 3, emphasis added). In this section, all studies controlled for the words and grammatical structures presented to the listeners in a way that makes describing their manipulations as accent more appropriate than describing them as dialect. For this reason, the term accent

Accent: Distinct from dialect, it refers to features of pronunciation that identify where a person is from regionally or socially.

will be used exclusively in this section.
Accent and L1 comprehension

Even when listening to native speakers of their own language, listeners can have difficulty if the speaker has an accent that differs from their own. Ikeno and Hansen (2006) examined the effect of native accents of varying degrees of familiarity on transcription accuracy for native speakers and found that more unfamiliar native accents led to lower accuracy. Recent data suggest that difficulty for native speakers with unfamiliar accents may be overcome through repeated exposure. Findings examining word recognition show that the first presentation of an accent to native speakers triggers a delay in word identification, indicating lower comprehensibility of the accented words (Floccia
Goslin, Girard, & Konopczynski, 2006). This initial delay is followed by a subsequent adaptation across a brief series of additional trials with the same accent (Floccia, Butler, Goslin, & Ellis, 2009). However, this adaptation is not

© 2010 University of Maryland. All rights reserved.

June 2010 50

 UNIVERSITY OF MARYLAND CENTER FOR ADVANCED STUDY OF LANGUAGE

complete: even after several trials, words with unfamiliar accents are still responded to more slowly than words in a familiar accent (Floccia et al., 2009). This suggests that, to the extent that full adaptation is possible, more extensive exposure is needed. Other results suggest that familiarity with a particular speaker is as important to determining comprehensibility as familiarity with a particular accent. Gass and Varonis (1984) found better performance on a listening comprehension measure when aural materials were presented in a familiar accent and the highest level of comprehension when the particular speaker was familiar to the listener. Their findings suggest that the less familiar the accented speech (e.g., unfamiliar non-native vs. familiar native) and the less familiar the speaker (novel vs. familiar speaker), the more difficult speech will be to comprehend.
If familiarity with an accent is the key to improving the comprehensibility of accented speech, it is possible that training listeners with a particular type of accented speech will improve comprehension in a way that generalizes to other speakers with the same accent. Weil (2003) examined the effect of training on comprehension of foreignaccented L1 speech for native listeners. Trained listeners showed superior performance relative to untrained listeners when the speaker at test was the same speaker as during training, indicating that participants did adapt to that person’s speech, as found by Gass and Varonis (1984). As to whether receiving training with speech from a speaker with a particular accent improved comprehension more generally for speakers with that same accent, the results were mixed.
Tasks involving single words did not show an advantage of training with a different speaker with the same accent, while those involving full sentences did show an advantage of this type of training. These results suggest that training with foreign-accented L1 can improve comprehension for other speakers with that accent for longer stretches of speech. Recent research on the acquisition of unfamiliar phonetic contrasts suggests that the proficiency of the learner may influence whether they will be able to generalize their experience with an accent across different speakers. Lee et al. (2007) trained participants to recognize pitches from a language in which they had no proficiency. Participants completed this initial training with a high level of performance (e.g., >70% accuracy at identifying pitches) or a low level of performance (e.g.,