There are two main explanations of the results presented in 5.4:
In my opinion the methodological explanation of the results cannot be left out. The low percentage of incorrect answers (6.7) indicates that the test might have been too easy. This percentage is even lower when the answers given to the first ten questions of the experiment are excluded (5.6). These early answers probably reflect confusion in the confrontation with an unfamiliar task and have little value. There are too few incorrect answers to say anything certain about lateralization. Had the percentage of errors been significantly higher and there still had been no ear preference, I would have leaned towards explanation i), but in the light of the present results I find this indefensible.
How could the dichotic experiment have been changed so as to yield more suggestive results?
As far as I see, there are two types of answers to this question: One is to alter the dichotic listening technique; the other is to change the input. A solution of the latter type could, for example, aim at making the utterances more difficult to process and thus increase the percentage of incorrect answers. An instance of the technique-solution will be addressed in i) and the input-solution will be discussed in ii).
Instead of having the informants concentrate on one ear at a time, I might have conducted a non-informed non-directive experiment, asking the informants to simply report what they heard. Then I could have compared the number of answers reporting the illocutionary force of the input to the right ear to the number of answers reporting from the left ear. This way the respondents would not have the chance to answer correctly, since there would be no correct answers to give. The probability of getting a statistically significant tendency would consequently have been higher with this method.
A problem with non-directed non-informed approaches is improving the quality of the input so that the subjects perceive it as one perceptual unit (or at least close to one perceptual unit). The difficulty of this task increases with the number of syllables in the input item. As my input items are whole sentences, this would probably have been difficult to achieve, and it would certainly have required extensive digital editing. Another, perhaps more serious problem with non-directed non-informed approaches is that the subject when realizing that she will receive one specific type of stimuli, for example verbal stimuli, might automatically start to concentrate on the ear with which she perceives this type of stimuli more easily. This will create an ear bias even before the testing begins. The perceptual bias problem should be taken especially seriously if one adopts the attentional theory of Kinsbourne (1970,1973,1975) where an expectation of verbal stimuli serves to prime the language dominant hemisphere and makes it extra sensitive to stimuli (cf. section 4.1).
A way of making the respondents produce more mistakes would be to increase the distraction load. For example one could add noise to the tapes. Most probably the subjects would then have been forced to make more errors. The drawback of this technique is that what one is actually measuring in the experiment becomes more obscure. That the added sound would affect the processing of the input in the brain is certain, but to what extent and in what direction it would draw the result is less clear. Noise in the form of music or environmental sounds, for example, would activate the right hemisphere and therefore give an unwanted advantage to the left ear [BrydenBryden1988]. The best solution would probably be to pick a type of noise proven to be processed bilaterally.
In the present experiment, the sentence-pairs are presented with a 2 second interval. A shortening of this interval, for example by 1 second, would probably have contributed to bring about more mistakes. The 2 second interval seemed to give the informants more than sufficient time to react. This technique has been successfully used in a number of other experiments [Clark, Geffen, and GeffenClark et al.1988]. An example is a study by Sexton and Geffen (1979) where an increase in the rate of presentation from 1 to 2 word-pairs per second was shown to be sufficient to reduce the average score by approximately 30%. Again, however, such a technique might contribute to obscuring what the experiment is really measuring. Some informants, perhaps the younger ones, might deal with time pressure significantly better than the older ones. The result could be score differences which have little to do with the ability to process intonation. However, the lateralization effect would probably remain consistent throughout the informant community, even if it would be greater for some age groups.
Another option would be to filter the utterances so that all identifiable linguistic content except the prosodic contour would be rendered unintelligible. A number of experiments have shown that filtering of verbal input is an effective means to reduce, or completely eliminate, a right ear advantage [BehrensBehrens1985]. This tendency may be viewed as support for the 'functional load' hypothesis of prosodic lateralization, or the notion that the linguistic load of the input is a contributing factor in the determination of the laterality of the processing.
The results of the present study show no significant right ear advantage and it is therefore difficult to predict whether filtering would result in a reduced score for the right ear, an improved score for the left ear or neither of these. As mentioned in section 2.2, Blumstein and Cooper (1974), found a left ear advantage for the processing of intonation-based illocutionary force in a dichotic experiment where the input sentences were filtered and the response task was non-linguistic in nature (subjects were to select from a series of curves the one which depicted the intonational contour of the utterance). When filtering was not used, or when the response action was linguistic, there was no longer a significant ear preference.
No matter how filtering would have affected the results, it would undoubtedly have contributed to singling out what the experiment aims at measuring, namely intonation. In the present study, the other parts of the linguistic input are simply factors disturbing the validity of the measuring procedure and could profitably have been left out. Also, filtering might have contributed to a reduced average score as the informants are less used to hearing filtered than non-filtered input. Although it is uncertain how it would have affected the outcome, I conclude that using filtered input would have improved the experiment.