Saturday, December 17, 2016

August 13, 2015



August 13, 2015

Written by Maximus Peperkamp, M.S. Verbal Engineer


Dear Reader,

This writing is my thirteenth response to the paper “Talker-specific learning in speech perception” by Nygaard and Pisoni (1998).  The researchers are unknowingly honing in on what I call Sound Verbal Behavior (SVB), when they state “The crucial research question then becomes whether, given experience with the particular aspects of the speech signal relating to talker identity, it follows that listeners also become sensitive to talker-specific linguistic properties.” What is said in SVB is easier understood because of how it is said and how the speaker sounds determines whether what the speaker says makes sense to the listener. However, the opposite is also true for Noxious Verbal Behavior (NVB), in which “particular aspects of the speech signal relating to talker identity”, the sound of the speaker’s voice, because of its aversive nature, gets more attention than the “talker-specific linguistic properties.” It requires the listener's effort to relate these “aspects of talker identity” with “talker-specific linguistic properties”, if the speaker expresses NVB, but if the speaker has SVB, no such effort is needed. 


Unaware of the SVB/NVB distinction, the researchers interpret various investigations done by others by stating “Taken together, these practice effects with synthetic and compressed speech suggest that the speech processing system is capable of adjusting to a variety of distortions, both synthetic and natural, that occur in the acoustic signal.” Only in NVB there is, because of the aversive-sounding voice of the speaker, a need for the listener to adjust “to a variety of distortions”, but in SVB there is no need to adjust to the speaker’s “acoustic signal” as there is no aversive stimulation. 


In SVB the speaker’s voice only has an appetitive effect on the listener. Not capturing this affective quality of the speaker’s vocalization, these authors believe that “Each talker’s vocal style shapes the acoustic realization of linguistic constituents in different but systematic and predictable ways.” With knowledge about the SVB/NVB distinction, however, is is clear that  SVB speakers shape in a SVB way and NVB speakers shape in a NVB way. How the speaker sounds influences the negative or positive emotional experiences of the listener and indeed shape “the acoustic realization of linguistic constituents in different but systematic and predictable ways.”


Tossing out the emotional influence of how we talk is characteristic for NVB. In SVB, listeners can recognize this influence, because they can, as speakers, articulate and explore the positive or the negative emotions, which they experience because of how the speaker sounds. In other words, in SVB there is feedback from the listener to the speaker. Rather than the listener making an effort to adjust to the speaker, the SVB speaker effortlessly adjusts to the listener. Effortless adjustment of the speaker to the listener occurs because the speaker listens to him or herself while he or she speaks. 


As the speaker listens to him or herself and experiences him or herself, he or she estimates how listeners other than the speaker are experiencing him or her. As the speaker listens to him or herself more often, his or her ability to accurately assess whether his voice is having a positive or a negative effect on the experience of other listeners reaches increasingly higher levels of accuracy. As a consequence, SVB is continued, even under the worst circumstances. This writer has reached this point and believes that others can reach it too.  What used to be a big problem turned around and made his life enjoyable. He used to feel very troubled and upset by how others were sounding, but as he now recognizes NVB as NVB and SVB as SVB, he is able to withdraw from NVB and approach and attract SVB. The problems of NVB can be avoided only if we recognize NVB as NVB. For most of us  this is not the case since we also don’t recognize SVB as SVB. Once we discriminate this distinction, NVB will decrease and SVB will increase. Our ignorance about this distinction prevents SVB and perpetuates NVB.  

  
These researchers don’t know the SVB/NVB distinction and their reasoning is by default based on NVB. Since they are scientists, they find some pieces of the puzzle. “Nevertheless, perceptual adaptation to individual talkers’ voices, as mentioned previously, has traditionally been cast as a problem of eliminating variation due to individual differences in speakers’ voices from underlying linguistic constants, rather than as a perceptual learning process in which listeners become attuned to properties of the speech signal which subserve both talker identification and linguistic processing". This “perceptual learning process in which listeners become attuned” describes SVB. Note that the “problem of eliminating variation due to individual differences in speaker’s voices” is not the speaker’s problem, but the listener’s problem. It is, of course, also the speaker’s problem, but it doesn't seem that way as long as he or she can get away with forcing others to listen to him or her and not listening to him or herself. The speaker’s problem becomes the listener’s problem in NVB. When the speaker recognizes his or her own NVB, he or she changes his or her sound and influences the listener who is him or her self and other listeners positively. 


We keep going around in circles as long as we treat speakers and listeners as separate entities. In SVB, in which the speaker listens to himself while he or she speaks, the speaker is the listener. When the listener within the same skin is not listening to the speaker, the speaker produces NVB. It is not the listener outside of the skin of the speaker, who needs to become “attuned to properties of the speech signal which subserve both talker identification and linguistic processing,” but it is the listener within the same skin as the speaker, who needs to become “attuned.”


In the discussion section of the paper, the authors mention that “Individual listener performance across training groups ranged from 28% correct for the poorest learner to 97% correct for the best learner, after 9 days of training. This finding suggests that simple exposure to the set of voices over the 9-day period was not sufficient for perceptual learning of talkers’ voices to occur.” Much more is involved in real life when we the listener adjust to how the speaker sounds. Generally speaking, someone who has experienced more SVB in his or her behavioral history adjusts to someone with NVB less easily than someone with more NVB. However, someone with more NVB in his or her behavioral history will more easily adjust to someone who has more SVB in his or her behavioral history. The authors clearly have no clue about this. “Given that these listeners differed in their ability to learn the voices, it was possible to characterize some of the listeners as “good” voice learners and others as “poor” voice learners.” 


What is a “good” or a “poor” voice learner, depends from whether one reasons from a SVB or a NVB perspective. From a SVB perspective, the “poor” voice learner, although he or she may be impaired by the speaker, recognizes NVB as NVB and is therefore a “good” voice learner. In other words, failure in NVB implies success in SVB. Many things are upside down because of how we talk. Only from a NVB perspective, the person who is distracted by the NVB speaker is called a “poor” voice learner. 


Without the SVB/NVB distinction we are bound to draw many wrong conclusions. ““Good” learners improved to a much greater extent than did “poor” learners. This divergence suggests that through practice in categorizing and explicitly identifying voices, “good” learners become “attuned” to the fine acoustic–phonetic details that distinguish each talker’s voice.” From a SVB perspective, however, the fact that “poor” learners improved to a much lesser extent indicates that they were “attuned” to and upset about, distracted by and therefore negatively affected by the not-so-fine aversive acoustic-phonetic details of the NVB speaker’s voice. Such a distraction would never occur with a SVB speaker. 


Many people are classified as “poor” learners simply because they respond with fear, anxiety and stress to a NVB speaker. ““Poor” learners do not seem to acquire the same kind of perceptual sensitivity using these voice dimensions during this type of laboratory training task.” However, the NVB speaker doesn’t acquire “perceptual sensitivity” to the listener as long as or she can continue to force others listen to him or to her. Thus, the conclusion that““Poor” learners do not seem to acquire the same kind of perceptual sensitivity” is confounded by poor NVB speakers. Yet, there is hope, as the authors found “it appears that both talker-specific and listener-specific variables contribute to the eventual identification of a talker’s voice.”  
Moreover, the finding that “Perceptual learning of a set of novel talkers’ voices caused listeners to be better able to recover the linguistic content of the signal” experimentally demonstrates that “the perceptual mechanisms responsible for analyzing talker identity are not independent from the mechanisms responsible for extracting the lexical content of an utterance from the speech wave form.” Can we finally admit that how the speaker sounds always affects the listener’s ability to understand the speaker?
  

As SVB and NVB are universal subclasses of vocal verbal behavior within each language, the authors inadvertently make indirect references to this distinction. “One explanation of these results is that the “poor” learners did not receive sufficient training to “fine tune” or adjust their attentional mechanisms to the relevant talker-specific information in the signal.” This not receiving "sufficient training to fine tune" of course means they didn't engage often enough in SVB to be able to recognize NVB. Furthermore, the insufficient training also indicates that something is wrong on the speaker’s side and not on the listener’s side. In NVB, however, it is always the listener who is blamed for not understanding the speaker. Interestingly, the authors add “It should be noted that the “poor” learners did not necessarily have difficulty processing speech from a variety of talkers, but rather, when the perceptual system was taxed, as when words were presented in noise, they were unable to utilize their prior knowledge of each talkers’ idiosyncratic style of speech to help recover the phonetic content and lexical information in the signal.” They seem to be saying that in NVB speakers produce some kind of voice-noise, which taxes the listeners’ perceptual system with their “idiosyncratic style of speech.”

No comments:

Post a Comment