Slowed-down speech - the root of listening comprehension failure?

[reineke]

Slowing down is being defended here from non-existent pitchfork mobs. Yesterday, when I listened to a news item on the VOA slow English website, I got Hulk-mad and irritated. Today I tried something else, by a different reader.

While I'm not exactly in love with this type of material, what I heard sounded a lot better than what I heard yesterday. I still don't like this type of language. I cannot listen to slow English long enough to judge if it is a great resource. If I listened to this style of speech in an unfamiliar language I may even perceive it as useful. This could be for a variety of reasons. Maybe the French do it better.

VLC allows you to slow down at 0.9x speed. You can also go much lower than that. At 0.9x speed the sound is slightly crumpled in places but otherwise very much like the original. At 0.7x and below both picture and sound get affected too much for my liking. Like I said before, I didn't perceive slowing down in this manner as useful. However, I could also make such a casual observation about listening to things I don't understand (or I'm unable to catch). This I know to be untrue. I have plowed enough linguistic fields to know that if you hit a particularly hard stump it's wiser to keep going. You may come back to it tomorrow or next week and find out that the troublesome roots have loosened.

Objective speed is only one factor when considering listening difficulties so slowing down, if and when it works, is not a cure-all solution. It is also reasonable to conclude that slow speech is not an indispensable ingredient for successful language acquisition. Slowing down in a classroom setting is very different from consuming slow material. My main concern would be dependency. I see too many learners who are heavily dependent on language crutches such as subtitles.

[Cavesa]

Proving that listening to distorted English doesn't help you learn proper English isn't really newsworthy.

Learning distorted English means being able to communicate with vast majority of English speakers on the planet as they speak badly too. Learning proper English is sometimes actually an obstacle in communication, trust me.

True, but as there is no shared "learner English" dialect, the only real target model we have is good English...

The Myth of Good English

... (edited out a great but long part for practical reasons-

But Cavesa was talking about something else -- the difficulty in understanding learner English -- and suggested that learning native-like English was a potential barrier to comprehension. (This might have just been a joke, but it seemed like a serious comment to me.)

I would not be against teaching an "international standard learner English" if there was any such thing, but I don't think there is. ... (

First of all, to the roots of the discussion: I still believe that slowed down speech can but doesn't have to be useful, it depends on the situation. The most important part is progressing beyond it, which is where many learners get stuck. Getting stuck at slowed down speech, either mechanically or learner aimed stuff, is nearly as bad for real life listening comprehension as learning without audio completely, in my opinion.

I was just illustrating that English is often a wrong example, as the ESL situation is totally different from any other language on the planet. Many anglophones (including the researchers) automatically expect us to approach English almost religiously, striving for perfection that would balance out the bad luck we had by being born elsewhere (sarcasm). But that is not true, just a small part of English learners is that ambitious about it, just a tiny part of learns fills the FCE/CAE/CPE examination rooms. Most English learners do not care about "nativelike" accent, proficiency, having complex discussions with natives. They don't need it, they don't want to go beyond their school/employer basic requirements. Just like most chemistry or math learners all over the planet don't dream of getting a PhD in them.

I wasn't joking much, as Serpent clearly said. A few real life examples: my father speaks horrible basic English, mixed with a few words from German and Italian (usually all in one sentence). Yet, he had easier time understanding and getting a message across, when we met a Japanese native speaking bad English. Or a Greek native speaking bad English. Or an Italian native speaking bad English. I had trouble being understood by a Polish guy and had to dumb things down for him. Or by a few Spanish natives. But when it comes to talking to natives, my father doesn't have a chance, while I am ok.

No, my English is not native like. By far not. But I have left the usual learner level behind ages ago. I have spent a lot of time listening to natives, even though mostly through tv and such sources. But vast majority of learners either doesn't bother or doesn't have any means to do so. Their main (or only) resource is their teacher (which can have a really bad pronunciation, like the one my sister recently got and often cannot understand, because she is used to classes taught by natives) and the only other speakers they meet are their classmates. Hey, I've even witnessed a German class sharing mistakes, it was like an exchange market of pronunciation troubles. And this is even worse for English as most people you meet outside the classroom are not the natives either, while using German outside of class usually includes some.

Therefore using English as an example to demonstrate usefulness/uselesness of any kind of slowed down speech, that doesn't work well. The researches may be great within the ESL realm, but have very limited value elsewhere. Most people don't want to be as native like as possible. It's like discussing whether a fish needs cold or hot water to learn how to climb a tree.

The idea of learning "global English" at school is stupid, in my opinion. It totally makes sense to learn more the american or other variants in various regions. But learning mistakes on purpose is beyond crazy.
.........................................

Different languages are much more interesting for this discussion. I am now very grateful for the slow Assimil Finnish recordings. But I couldn't stand the slow Italian recordings for long (I think I stayed at that point for two weeks or perhaps one), I prefered to jump to tv series right away (talking about listening now). But those are two totally different situations.

It is an interesting idea though, to consider the difference betweens audio recordings we use for comprehension practice, and those we use for pronunciation practice, as those are usually not the same.

Pronunciation drills usually start with single words, at quite normal speed. Like the Forvo recordings. Than many of us use coursebooks with audio (usually slower and clear) for various kinds of activities, such as repeating after the dialogue or shadowing and so on. I use learner aimed material for practice, I pause and repeat the dialogues, therefore that is usually a slower pace, I do the same much later with full speed TV series, but much less than I should. I don't like mechanical slowing down, even though trying it more might surprise me.

But as far as comprehension goes, I think quantity beats any other aspect. Transcripts can help a lot at first, I totally believe it is possible to improve your comprehension fast even without ever using learner aimed material. Skipping something and returning later is a good option, I don't think obcessively looking up and consulting every unknown/misunderstood word is necessary (but some learners learn well this way, true). Listening extensively for dozens of hours does miracles.

It is rather the pronunciation practice, where our methods and opinions are very likely to be varied.

[Cainntear]

I was just illustrating that English is often a wrong example, as the ESL situation is totally different from any other language on the planet.
...
Therefore using English as an example to demonstrate usefulness/uselesness of any kind of slowed down speech, that doesn't work well. The researches may be great within the ESL realm, but have very limited value elsewhere. Most people don't want to be as native like as possible. It's like discussing whether a fish needs cold or hot water to learn how to climb a tree.

I think you've got your argument upside-down here.

The papers Reineke cited were about how learners' comprehension of native English failed to improve after listening to carefully pronounced, learner-targeted English.

Imagine that this was actually about Xhosa, for the sake of argument.

"The papers Reineke cited were about how learners' comprehension of native Xhosa failed to improve after listening to carefully pronounced, learner-targeted Xhosa."

You would probably then say that the findings of the study were not of relevance to the ESL/EFL realm because the real aim of most EFL isn't to understand native speech at all, but ELF (English as a Lingua Franca). I don't disagree with that, as I've already said. In fact, I am pretty staunch in my opposition to the idea of teaching "culture" in the EFL classroom precisely because the main function of English for most learners will be as a lingua franca.

But if the argument works that way for Xhosa, it works that way for English too -- i.e. even though the medium of the study was English, the conclusions aren't relevant to EFL, only to the teaching of other languages.

[reineke]

While the studies I cited do examine carefully pronounced, learner English, "the survey of literature" sections also mention poor results with mechanically slowed down speech.

Griffiths (1990) claims that speech faster than 200 wpm (3.8 sps) impairs comprehension for lower-level ESL learners but that comprehension at 100 wpm was not better than at 150 wpm. The average, normal speech rate for English speakers is about 150 wpm.

This 1997 study is still cited. The sound was slowed down mechanically and the researcher noted some positive results.

http://pages.uoregon.edu/itrummer/Zhao_1997.pdf


Barriers to Acquiring Listening Strategies for EFL Learners and Their Pedagogical Implications

http://tesl-ej.org/ej32/a2.html

[YtownPolyglot]

The nice people in Spain who totally confused me could have slowed down so that...they...were...speaking...like...this...and I wouldn't have understood them any better. There is a gap--a yawning chasm--between the Spanish many of us learn and what Spanish-speakers actually use. The same is true for most other languages, if not all.

There are plenty of people in every language group who don't enunciate as clearly as the learner tapes and audio-visual aids.

Particularly with German and Italian, I had trouble understanding dialects when I was in Europe.

Even a good language course can't give you all of the idioms and slang you might hear.

By the time I finished four years of high school French, I could understand my teacher and anyone in the class really well, but French people never spoke the high school French I studied.

[s_allard]

...
By the time I finished four years of high school French, I could understand my teacher and anyone in the class really well, but French people never spoke the high school French I studied.

A couple of points here. Firstly, this common idea of "four years of high school French" is extremely misleading. It's really more like two hours a week, 40 weeks a year for four years in a class of around 25 students. Let's say 320 hours of group instruction. And let's say this can get you to A2 in the best of cases. It's not bad but obviously this is very limited in terms of conversing with native speakers.

Secondly, and to split hairs a bit here, it is inaccurate to say that French people never speak the high school French one has studied. 320 hours of instruction covers the basics of French that French people use. It's just that they use the same French in more complex ways and they probably speak it quite quickly. For example, high school French will probably cover the 50 most common verbs in French. All speakers of French use some of these verbs all the time. You can't speak French for five minutes without using avoir or être or faire. This is certainly covered in high school French but French speakers use these verbs in many ways.

So French speakers speak high school French but they don't speak like high school learners of French.

But the more important issue here is the fact that this post is addressing a point that, in my opinion, is not the thrust of the OP. At this point in the thread, people are talking about slowing down fast speech to make it more understandable. As this post has pointed out, listening to slowed-down speech is of dubious value if you don't have the adequate vocabulary and grammar in the first place. Slowing down doesn't make incomprehensible words comprehensible. It does make them more intelligible on the other hand.

What the OP was talking about was the bad effects of spending too much time learning slowed-down and distorted speech in the early stages of learning a language. This would be the "root cause" of comprehension failure. I disagree totally with this "root cause" idea for reasons I have given in previous posts.

[tastyonions]

If all you learn is to read everything that is in four years worth of high school French textbooks and to comprehend the same things when they are spoken in relatively standardized, clear (but native) speech you will still be way further than most people who never move to a Francophone country get with the language.

[reineke]

The Potential Use of Slow-down Technology to Improve Pronunciation of English for International Communication

Abstract

The focus of this research is on oral communication between L1 (first language) and L2 (second language) English users - to determine whether an algorithm which slows down speech can increase the intelligibility of speech between interlocutors for EIC (English for International Communication). The slow-down facility is a CALL tool which slows down speech without tonal distortion. It allows English language learners more processing time to hear individual phonemes as produced in the stream of connected speech, to help them hear and produce phonemes more accurately and thus more intelligibly.

When L2 users hear L1 speech they frequently complain that it is too fast for them to catch what is being said. The slow-down facility is seen as a possible means of helping L2 English users to catch those elements of speech which are produced at speed by L1 English users, namely features of connected speech.

The slow-down software – the AOLA algorithm - was initially developed by a team of computer scientists and engineers with the aim of slowing down recorded music samples in real time without affecting the tempo (speed) (see Lawlor and Fagan, 1999, for a more detailed explanation of the technical aspects of the algorithm). There were other algorithms available at this time. However, the AOLA was deemed to be superior to these as it could slow down in real time, making it a more efficient algorithm.

The software uses TSM to slow down speech recordings without tonal distortion, so listeners can hear streamed speech segments, as they naturally occur in authentic speech, with more processing time to focus on how the sounds are actually being produced – including connected speech features, which are usually difficult for L2 English users to notice and process due to the speed of naturally occurring speech. Recordings can be slowed to any desired speed – in the following tests in this study, slowed speeds of 80%, 60%, 50% and 40% were applied.

Conclusion

"At the start of this study, aspects of speech reception difficulty were investigated to try to pinpoint where speech reception was hindered and reasons for this. From the initial tests into speech reception, it was decided that the focus of the study should change to investigate speech production, as less research has been done in this area. This investigation was coupled with testing the effectiveness of the speech software to determine whether slowing down speech without tonal distortion could be applied to a pronunciation training programme to increase subjects’ spoken intelligibility."

"Test subjects trained over a one-month period – hearing each lesson first at 100% speed, then 80% and then 60% - to help them hear how the targeted sounds are produced naturally by an L1 non-RP English-speaking model and for them to repeat what they hear – in an effort to increase their spoken intelligibility. While the slow-down software was not deemed much more effective when compared with Control subjects who underwent the same pronunciation training without the application of slow-down facility, it does not invalidate the usefulness of the slowdown in ELT."

"This study indicates that English language learning experience is a strong factor in subjects’ receptive abilities, as students from non-Germanic backgrounds, such as Chinese and Polish, performed very well in tests. These students almost always documented a long history of learning and communicating in English, either with L1 users or other L2 users of English."

"The results of the study also show that there is little correlation between transcription and ability to correctly answer comprehension questions based on a recording. Different cognitive processes are involved and students who perform well in one activity are not guaranteed to repeat the performance in the other."

"This study also found that for EIC communication, receptive intelligibility seems to be easier for Germanic L1 speakers than speakers from other language backgrounds. This may be due to the phonemic similarity between English and the Germanic languages, such as German and Dutch."

"Although it was expected that subjects with a Romance L1 would display superior receptive skills in English when the speaker was also from a Romance language background, this was not found to be the case. Similarity of L1 to English was found in this test to be a greater aid to receptive intelligibility than L1 similarity of both speaker and listener."

The study also "established that targeted phoneme practice alone is not sufficiently effective for improving the intelligibility of an L2 English speaker".

Limitations
Application of the Slow Down: as the slow-down algorithm was being developed during the course of this study, it was not possible at the time of testing for subjects to access the slow-down as a stand-alone tool. It was not possible for subjects to slow the recordings to speeds of their choosing. Previous tests on speech rate, such as those by Derwing and Munro(2001) show subjects have a preferred speaking rate, which may have a corresponding effect on intelligibility.

9.2. Implications for Pedagogy
The slow-down could be incorporated into ELT materials for oral/aural work or in a software package for independent learning, where users could choose how slow to hear items and how often, which could have a positive effect on learning.

http://arrow.dit.ie/cgi/viewcontent.cgi?article=1017&context=appadoc

[reineke]

This material from 2007 explains the theory behind the previous study.

User-defined distortion-free speech slow-down algorithm optimised for language learning

"An Irish research institute has adapted a time-scaling algorithm to run efficiently on a desktop computer, which allows users to alter pace of delivery of recorded speech without distortion, which significantly enhances the language learning experience of any non-native speaker.

"Language learners tend to learn an idealised version of a language and cannot immediately cope with a native-to-native speech speed. Using this speech slow-down technology, learners of any language can slow down recorded natural speech so as to study how it is produced and therefore better reproduce the sounds themselves. This allows students to cope with rapid native speech in a non-threatening, self-learning environment by gradually speeding up the voice recording until full native speech production has been attained. The voice recording may derive from live recording, radio, TV, the Web or CD-ROM. By slowing down native speech without pitch distortion the programme enables language learners to follow what has been said by native speakers and helps them to reproduce a natural flow of speech. The approach applies to all natural languages."

http://cordis.europa.eu/result/rcn/41915_en.html

DITCall-Slow: Slowing Native Speech for
Language Learners

"It is a common experience of many learners of a foreign language that native speakers (NSs) of that language speak too quickly for them to understand or imitate. Slowing down a segment of speech with older technology results in the familiar deepening of the voice as the pitch drops as well. The result is unpleasant and not particularly instructive. The DITCall-Slow tool slows recorded speech without tonal distortion, so that the learner has – literally – more time to hear what was said by the NS and, especially at slower playback speeds, can attend to the manner in which the sequence was spoken. While the tool is currently being commercialised for the study of English as a foreign language (EFL), it can be applied to any spoken language...

In the same way that fast photography makes the subtleties of a tennis serve or a golf swing perceptible to the learner tennis player or golf tyro, DITCall-Slow will ‘allow time to be spent with the signal’, to use Richard Cauldwell’s term (Cauldwell 2002)8. Studying the intonation patterns of slowED speech (as opposed to slow speech) and repeating the model at increasing speeds until a NS rate is reached, enables the learner to acquire a NS delivery, or at the very least render NS speech more intelligible.

"DIT internal research (Meinardi 2006) has produced evidence that students prefer a slowed-down speed of 80% as this rate gives the user sufficient time to process the NS speech signal while still retaining a high degree of naturalness in the recording. Slower speeds, however, while sounding ‘drunk’ or ‘exhausted’, highlight the natural prosody of the speech act and are useful for the learner to study native-speaker intonation patterns."

"Asian learners in particular can benefit from DITCall-Slow, as it allows them to listen to NS recordings at normal speed or slowed down to a practical limit of 40%, steplessly.

"At slower playback speeds some idiosyncrasies of spoken speech become apparent which are not evident at normal speaking rates. In an internal DIT recording the speaker says: ‘ ... that must be quite difficult’. Only when the recording was slowed to 60% did it become apparent to the speaker herself that she had, in fact, said ‘diff-cult’, not the perceived ‘difficult’. The fact that NSs tend to say ‘teM balloons’ instead of ‘teN balloons’ in informal, relaxed speech can only be made obvious by use of the slow-down algorithm."

http://arrow.dit.ie/cgi/viewcontent.cgi?article=1092&context=dmccon

[s_allard]

I couldn't wade through all these citations so I'll assume that they are interesting and somehow relevant. I'll assume that the main idea of using slowed-down speech as a learning tool is to give the learner a chance to acquire the target patterns and then bring these up to the desired speed. So, let's say your goal is 150 words per minute (wpm) of fluent and correct speech. So, maybe you'll practice some stuff at 50 wpm until you are comfortable and then crank it up to 100, then 125 until you hit 150. Seems pretty logical to me. I don't know if there's anything left to add.

What I would like to discuss is the role of transcripts that iguanamon mentions. I'm also a great fan of listening with transcripts. I would suggest that the transcript is in fact a way of "slowing" down recorded speech because we can read the transcript at any speed.

What happens is that the transcript eliminates the intelligibility problem. We don't have to strain our ears to figure out what is being said. Plus we have, not a phonetic transcription, but a standard alphabet transcription which is great for learning how to read and of course for studying grammar and vocabulary.

When we study or read a transcript without listening to the recording, or maybe just listening in our head, we are basically slowing down. And when we repeat passages many times, that's also a form of slowing down. You can't do this in a real-life conversation.