Page 10 - ELG1911 Nov Issue 467
P. 10

El•Gazette 467.qxp_El•Gazette 467  23/10/2019  13:10  Page 10





              RESEARCH NEWS                      .


              However fast humans talk they deliver


              the same amount of data per second





              By Gillian Ragsdale
                                                                                                                        UNSPLASH
              Human speech tends to deliver
              the same amount of information
              per second regardless of the
              language being spoken, a major
              international study suggests.
              Native speakers adapt their
              average speed of speech according
              to the information density of their
              language, according to Christophe
              Coupé and colleagues.
                It has long been observed that
              some languages sound quicker
              than others. The researchers, who
              were drawn from universities in
              France, New Zealand, Hong Kong
              and South Korea, thought this
              might be related to how efficient
              different  languages  are  at
              delivering information.
                The study compared seventeen
              languages, from nine different
              European and Asiatic language
              families. There was substantial
              variation on several measures:
              Japanese, for example, uses only a
              few hundred syllables, compared
              to the almost 7,000 used in
              English.
                The  speaking  rate  was
              measured in syllables per second
              and recorded for each speaker.
              The researchers chose to use read-
              aloud texts to control what was
              being so that they could make   Researchers found that speaking rate varies, but amount of information delivered remains the same
              valid comparisons between the
              speakers.                   Information density also varied   pervasive  cultural  gender   through a sentence only to forget
                Ten native-speakers of each   by language: Basque, for example   differences play a role.   what it was they were talking
              language were each given a set of   delivers around five bits of   Despite these sex differences,   about.
              fifteen texts. They were asked to   information per syllable, while   the overall information rate,   The solution is for speakers of
              familiarise themselves with the   Vietnamese delivers eight.   however, appears to be stable   languages that deliver more
              text before reading them aloud so   By multiplying the speaking rate   across the wide range of languages   information per second to speak
              they could speak as naturally as   by the information density,   studied, suggesting that speakers   more slowly – and those delivering
              possible.                  researchers were able to arrive at   adapt their speaking rate to deliver   less information per second to
                The study found that the   the ‘information rate’ – the   information at a preferred rate for   speak more quickly. So, for
              speaking rate varied from four   number of bits of information   both speaker and listener – not too   example, Turkish speakers may
              syllables to nine depending on the   being delivered per second.   fast and not too slow.    sound faster than Vietnamese
              language. Some languages sound   Despite the variation in speaking   Speaking too quickly is hard   speakers, or slower than Spanish-
              faster because they are delivering   rate and information density, the   work for the speaker and increases   speakers, but the rate of
              syllables at twice the rate of   information rate was relatively   the cognitive strain on the listener   information received is pretty
              others.                    constant, at an average of 39 bits   – which is why teachers like myself   much the same.
                For each language, an estimate   per second.       often have to learn to speak more
              of information density was then   Across all languages, women   slowly.         REFERENCE
              calculated. The study used large   tended to speak more slowly and   Speaking too slowly leads to   n Coupé, C., Oh, Y. M.,
              written corpora to work out how   consequently  deliver  less   slower  turn-taking  in   Dediu, D. and Pellegrino, F.
              much information was being   information. The authors suggest   conversation, where the average   (2019) ‘Different languages,
              communicated as bits per syllable.   this may reflect the influence of   turn takes just two seconds. It also   similar encoding efficiency:
              This builds on research finding   biological constraints of the kind   requires the speaker to keep more   comparable information rates
              that the ability to process syllables   that also cause the average   information in their working   across the human
              is central to speaking and listening   women’s voice to be higher in   memory – a problem familiar to   communicative niche’, Science
              skills.                    pitch – but it is also possible that   anyone who has gotten half-way   Advances 5: eaaw2594.
              10                                                                                         November 2019
   5   6   7   8   9   10   11   12   13   14   15