<table cellspacing="0" cellpadding="0" border="0" ><tr><td valign="top" style="font: inherit;"><div><br></div>Before making *any* claims about child speech here, I would very strongly<div>recommend also tracking down numbers for the obvious control: adult</div><div>conversational speech. It's well known in the speech community that bits</div><div>of adult speech are frequently unintelligible, or unintelligible without context</div><div>of varying amounts. Or they seem intelligible until you try to transcribe them</div><div>and try to pin down exactly which words were said. I've seen the same type</div><div>of demo done with adult speech.</div><div><br></div><div>I don't have a percentage handy. But I'm in the middle of a project syncing the two</div><div>transcriptions of the big Switchboard corpus and probably something like 10% of the</div><div>utterances have a significant difference between the
two transcriptions. In my case, </div><div>significant means a difference likely to change the syntactic parse, not just (say)</div><div>replacing "me" with "you". If you ask me in a few months, I might have a real number.</div><div><br></div><div>For adult speech, the rate depends massively on your standard for "intelligible",</div><div>the type of speech, and the type of utterance. Important content words from recent radio</div><div>news recordings probably have near 100% intelligibility. Filler words ("you know what</div><div>I mean") in conversational speech are frequently mashed into formless blobs of</div><div>audio. Parochial words (e.g. Kalman filter, slub, gorp, rennet, all proper names) </div><div>are likely to be mistranscribed if the transcriber doesn't have the right background</div><div>knowledge, even with the context of the rest of the
conversation.</div><div><br></div><div>It's also essential to match the dialects. Not only are some pairs of (decently-sized)</div><div>English dialects not mutually intelligible (e.g. Johannesburg and Newark New Jersey, </div><div>as my husband discovered) but smaller mismatches create errors. Recall that both</div><div>the quality and number of distinct vowels vary dramatically across the northeastern</div><div>US. This immediately creates a big population of content words that are ok in context </div><div>but unintelligible in isolation. Worse, the vowels are moving, so that an adult from</div><div>"the same" dialect doesn't pronounce them the same. Bill Labov is famous for demos </div><div>on these points.</div><div><br></div><div>Moreover, familiar words are compressed relative to less familiar ones. That's a</div><div>big effect which has to be controlled for.
It is unknown how much this effect is</div><div>purely created by the speaker's convenience and how much is their perception of</div><div>the hearer's requirements. Little kids have their own theories of the hearer's needs.</div><div>If their household eats Weetbix or dal, they may think it's familiar to the grad</div><div>student interviewing them where an adult might guess it's unfamiliar and articulate</div><div>the word more carefully.</div><div><br></div><div>Finally, it may be important to control for utterance length. Adult speech offers</div><div>long utterances that provide internal redundancy. The very fact that the children</div><div>have shorter utterances will immediately reduce intelligibility for the utterance in</div><div>isolation. Brief answers to questions (e.g. "some cruellers") that were originally said</div><div>in a strong context might provide a good parallel to the situation with
children. (For</div><div>those of you not from New England, where it's a common word, those are twist donuts.)</div><div><br></div><div>Margaret Fleck (U. Illinois)</div><div><br>--- On <b>Thu, 2/4/10, Nick Riches <i><nick.riches@googlemail.com></i></b> wrote:<br><blockquote style="border-left: 2px solid rgb(16, 16, 255); margin-left: 5px; padding-left: 5px;"><br>From: Nick Riches <nick.riches@googlemail.com><br>Subject: Re: Intelligibility of children's speech<br>To: "Info-CHILDES" <info-childes@googlegroups.com><br>Date: Thursday, February 4, 2010, 5:04 AM<br><br><div class="plainMail">Dear Mats<br><br>I attended a very interesting talk by Sara Howard, a clinical<br>phonetician at Sheffield university, in which she gave a demonstration<br>of a single child's utterance which was completely unintelligible<br>without a context, but completely intelligible with the context. In<br>fact it made the audience gasp! She might be a good
person to<br>approach.<br><br>Nick Riches<br><br>University of Reading<br><br>On Feb 3, 8:41 pm, Mats Andrén <<a ymailto="mailto:mats.and...@ling.lu.se" href="/mc/compose?to=mats.and...@ling.lu.se">mats.and...@ling.lu.se</a>> wrote:<br>> Dear list members,<br>><br>> A few years ago I read an article or a book where it was stated that X<br>> percent (a large number) of children's utterances became more or less<br>> completely unitelligible when the video is taken away. That is, removing<br>> the visible context and action/gesture et cetera and relying only on the<br>> audio signal as a source of interpretation. I can't recall who wrote it<br>> (it *could* have been Eve Clark or Jana Iverson). Neither am I able to<br>> remember whether it was a result of an experimental study of some kind,<br>> or some more sweeping approximation. I also do not remember the age of<br>> the children involved, but it should have
been some time during their<br>> second year of life (maybe earlier). Precisely due to all this<br>> uncertainity, I would be most grateful if any of you would be able to<br>> point me to information about these matters. Any publication that deals<br>> with this question is of interest.<br>><br>> Have a nice day!<br>><br>> Best regards,<br>> Mats Andr n,<br>> PhD student in General Linguistics,<br>> Centre for Languages and Literature/Centre for Cognitive Semiotics,<br>> Lund University,<br>> Sweden<br><br>-- <br>You received this message because you are subscribed to the Google Groups "Info-CHILDES" group.<br>To post to this group, send email to <a ymailto="mailto:info-childes@googlegroups.com" href="/mc/compose?to=info-childes@googlegroups.com">info-childes@googlegroups.com</a>.<br>To unsubscribe from this group, send email to info-childes+<a ymailto="mailto:unsubscribe@googlegroups.com"
href="/mc/compose?to=unsubscribe@googlegroups.com">unsubscribe@googlegroups.com</a>.<br>For more options, visit this group at <a href="http://groups.google.com/group/info-childes?hl=en" target="_blank">http://groups.google.com/group/info-childes?hl=en</a>.<br><br></div></blockquote></div></td></tr></table>
<p></p>
-- <br />
You received this message because you are subscribed to the Google Groups "Info-CHILDES" group.<br />
To post to this group, send email to info-childes@googlegroups.com.<br />
To unsubscribe from this group, send email to info-childes+unsubscribe@googlegroups.com.<br />
For more options, visit this group at http://groups.google.com/group/info-childes?hl=en.<br />