i 71.1 



ii win:, ii ik i .1 i-m m. i 'i 



Ml Rl il'm SIOLOGY III 



munications engineering has produced results of the 



highest importance, little of it can at present be 

 titled into the context of traditional physiology. For 

 this reason, information theor) will not he considered 

 in the present chapter. Excellent accounts of its 

 contribution to psychophysiolog) are available else- 

 where (86, 87, 1171. 



SPEECH PERCEPTION 



The understanding of speech is to be regarded 

 properly as a problem in auditory physiology. Un- 

 fortunately, our knowledge of the central mechanisms 

 upon which it depends is so fragmentary that no 

 unified account can be given. None the less, a good 

 deal is known as to physical and psychological 

 variables which govern the intelligibility of speech 

 and some, at least, of this information may prove 

 relevant to neurophysiology. 



The intelligibility of speech is commonly assessed 

 by the so-called 'articulation testing procedures' (32, 

 37, 80). A standard scries of items (syllables, words 

 or sentences) is presented under controlled con- 

 ditions and recognition tested by the accuracy of 

 concurrent repetition (articulation score). The 

 material is commonly chosen to reflect the relative 

 frequencies with which the different phonemes occur 

 in evcrydav speech. It is usual to distinguish the 

 "threshold of deteetability,' i.e. the level of speech at 

 which the listener just hears the speech sounds, from 

 the 'threshold of intelligibility,' i.e. the level at which 

 50 per cent of the test items are correctly repeated 

 Standard articulation test scores as a function of the 

 sound pressure level of speech have been established 



(37). 



I In deteetability thresholds for individual con- 

 sonant sounds are found to be close to those for pure 

 tones in the middle frequency range but higher than 

 those for single vowels (131). Further, it appears 

 that the 'stead) state' portions of phonemic units are 

 nui in general sufficient for recognition and certain 

 transition properties must also be utilized (46). In 

 the recognition of vowel sounds, duration, onset 

 characteri tics and inflection are relevant charac- 

 teristics ii.'f',). Iii general, the intelligibility of simple 

 material is main!) .1 function of the deteetability 

 thresholds of the component sounds, but in the case 

 of more complex material grammatical structure, 

 familiarity, context and meaning take on increasing 

 importance Certain ol these variables, in particular 

 context, have been analyzed experimental!) with 



some success in terms of information theory (65, 77, 

 116). 



Effects of Stimulus Distortion 



The effects of frequency, amplitude and related 

 types of distortion upon the intelligibility of speech 

 have been widely studied, more especially from the 

 standpoint of communications engineering (37, 80). 

 Some of the findings arc of great theoretical im- 

 portance. Thus it has been established by experiments 

 involving frequency distortion that low frequencies 

 contribute surprisingly little to the intelligibility of 

 speech, despite the fact that they carry most of the 

 speech power. If, for example, all components <>f 

 speech below 1000 cps are attenuated by a high-pass 

 filter, speech power is reduced by about 80 per cent 

 but articulation score falls by only 10 per cent (39, 

 80). Equally, experiments involving amplitude 

 distortion have made it clear that the central portion 

 of the speech wave carries essential information for 

 recognition (78, 86). Whereas 'peak-clipping 1 has 

 surprisingly little effect on recognition, •center- 

 clipping' virtually destroys the intelligibility of 

 speech. In view of these findings, it his been suggested 

 that intelligibility is best defined as a function of the 

 intensity-frequency-time pattern of the stimulus 

 (80). The significance of this analysis for neuro- 

 phvsiological correlation ma\ well prove considerable. 



It has been pointed out by I'umphrcy (102) that 

 the transmission of speech through a number of 

 narrow bands, as in the Vocoder system, ma) effect 

 ,1 virtually complete separation of the emotional and 

 informative aspects of speech. Although intelligibility 

 as such is little, if at .ill, .if lei ted, the qualit) ill speech 

 is greatly altered and little trace ol emotional ex- 

 pression ma) be detected. In general, it would appear 

 that cues relevant to intelligibility are carried within 

 the regions of greatest energ) of the spectrum, whereas 

 cues relevant to e1nuin111.1l expression are carried 

 b\ changes in the fundamental frequency, I'umphrcy 

 is therefore led to suggest thai a partial dissociation 

 between the emotional and cognitive aspects of 

 speech may have been an import. ml factor in the 

 e\ uliiiiiin ul hum. in language. 



Masking »/ S/>, ,, /,- 



The problem ol masking, in so far as ii has spec i.il 

 reference to speech, has been reviewed b) a number 

 of authors (55, 7.', 841. In the case of masking b) 

 tones (pure or complex), the principal finding is that 





