chapter 10 - Division of Speech and Hearing Sciences

advertisement
Titze & Verdolini, © 2002
10-1
CHAPTER 10
MOTOR LEARNING PRINCIPLES
Roger was a 57-year-old engineer, who had a vocal fold paralysis caused by an
upper respiratory tract virus. Although he was able to produce relatively good voice on
some occasions, most of the time his voice was diplophonic and difficult to understand.
He tried, but just failed to stabilize his “good voice” production in everyday
circumstances. Soon, he accepted sounding “old,” and reconciled himself to limiting
contact with the public.
Beverly was a 37-year-old teacher. Her voice fatigued horribly by the end of
every work day, and she returned home to her family not wanting to speak. Her
physician recommended voice therapy. Although therapy instructed her about the value
of using a non-pressed laryngeal configuration in speech, she saw no results in her
everyday life. She stopped teaching because of the pain. Her family budget was
imperiled. Eventually, they moved into a smaller apartment because they could not meet
the mortgage on only one income.
Janet was a 20-year-old voice major at a prestigious conservatory. She had
received exceptional acclaim for her vocal prowess as a teen-ager performing in
community productions, and had been offered a full scholarship as a vocalist at the
conservatory. When she arrived, she was assigned a voice teacher who had studied voice
science. She anticipated improving in her vocal capabilities. Somehow, she failed to
meet the new challenges associated with auditions in a competitive environment. She
seemed to be “trying too hard.” Both she and her teacher were frustrated.
All of these scenarios represent situations in which knowledge of specific motor
learning principles might have changed the fate of the individuals involved. In Postulate
#7 (Chapter 7), we emphasize the importance of knowledge about motor learning applied
to voice training. In this chapter, we review important theories and principles of motor
learning. Consistent with the way that we define motor learning (below), we emphasize
cognitive, or “software” mechanisms in this chapter. We address physical, “hardware”
changes in tissue composition due to exercise in Chapter 9, on exercise physiology.
3/10/2016
Titze & Verdolini, © 2002
10-2
We begin with a definition of motor learning. We then proceed with a review of
historical models of motor learning and control, which will provide a context for much of
the discussion that follows. Next, we discuss critical issues in recent studies of motor
learning, focusing on (a) “controlled” versus “automatic” processes, (b) intention, and (c)
laws of practice. At the end of the chapter, you will be asked to make a bulleted list of
the important principles of motor learning reviewed in the chapter, and to give examples
of how you might apply them to voice training in your setting.
Motor Learning Defined
A clear definition of motor learning is important from the outset, as it will
influence the ways that we think about this phenomenon and the ways that it is studied.
We model our definition after one proposed by Schmidt and Lee (1999). According to
this definition, motor learning is a process, which is inferred rather than directly
observed, which leads to relatively permanent changes in the potential for motor
performance, as the result of practice or exposure. The italicized aspects of this
definition are key to understanding the phenomenon at hand and how it is studied.
The first issue is that motor learning is a process not a structure. It is not a “box”
nor a “bit” of information sitting in the head. Learning refers to changes in the activation
potential of various cognitive processes. Thus, learning’s essential nature is dynamic.
The second issue has to do with our inability to observe learning directly. We can
only infer it from observation of performance changes. This tenet derives from our view
of learning as essentially cognitive. It is an abstract “emergent property” of neuronal
activation. Even if neuronal activation changes are viewed in correspondence with motor
3/10/2016
Titze & Verdolini, © 2002
10-3
learning, as for example with functional Magnetic Resonance Imaging, such activation
signals physical aspects of learning but is not isomorphic to it.
A third tenet is that temporary performance shifts, which may occur during
training sessions, do not imply learning. Learning brings about stable changes in average
performance over time. As such, studies that look at performance changes within
individual training sessions cannot make any inferences about learning. Only studies that
incorporate a relatively long-term “retention test” can actually study learning. Later in
this chapter, we will see that this issue is crucial: many manipulations that enhance
immediate performance depress long-term learning, and vice versa. Thus, the
consideration of learning as a long-term change in average performance is critical.
The fourth tenet is that learning does not involve a monotonic increase in
performance over time. Temporary states, such as distraction, illness, arousal, etc., can
shift performance in one direction or another. Sometimes, a temporary downturn in
performance may even indicate an instability that signals learning, as a new, fledgling
behavior destabilizes an old, existing one. Thus we infer learning from general trends in
performance potential, over time.
Finally, motor learning is inferred only when performance changes are traceable
to practice or exposure. If an individual’s motor performance improves due to the use of
a drug, for example, or worsens because of an accident or fatigue (“negative”
performance change), we cannot attribute the changes to learning.
Historical Context: Theories of Motor Control and Learning
Early research
3/10/2016
Titze & Verdolini, © 2002
10-4
The world wars of the 20th century stimulated much attention and research around
perceptual-motor learning. A special impetus for this activity was the goal of
maximizing human military performance in airplanes and with telegraphs. One of the
most researched topics had to do with optimal schedules of practice and rest in motor
learning. Numerous studies pointed to the conclusion that spaced practice (training
which alternates practice and rest) tends to produce greater motor learning than massed
practice (training which involves extensive practice, with little rest) (cites).
Several hypotheses were proposed to explain the “spacing” effect. Predominant
ones were the “forgetting hypothesis” and the “consolidation hypothesis.” According to
the forgetting hypothesis, rest enhances learning because incorrect solutions are released
during non-practice periods. Thus the learner comes to subsequent practice having
released the hold of a negative bias in control operations. According to the consolidation
hypothesis, information is cognitively and neurally consolidated (processed and stored)
during rest, and is rendered stable during that period. It was not excluded that the
consolidation period also might involve covert problem-solving (“We learn to swim in
the winter and skate in the summer,” William James, 1890). Some level of evidence has
been brought to bear on both hypotheses to explain the spacing effect. Additionally, new
evidence has suggested a third, “contextual interference hypothesis,” similar to the
forgetting hypothesis, described later in this chapter. Regardless of the interpretation, the
spacing effect is empirically robust, especially for continuous as opposed to discrete-type
skills.
Stage models
3/10/2016
Titze & Verdolini, © 2002
10-5
Numerous models have suggested that motor learning occurs in stages. One
representative model suggest that a typical progression in motor learning involves
cognitive, associative, and autonomous stages, respectively (Fitts, 1964; Fitts & Posner,
1967). The cognitive stage is described as an essentially conscious, verbal stage, in
which the learner identifies “what” to do. That is, the learner identifies which body parts
to move where, and when. An example in voice might be learning that one needs to align
oneself well in space and expand the abdomen during inspiration, and then focus on
keeping the larynx “free” and creating a vocal tract constriction to produce a strong
sound. According to traditional stage models, verbal, mechanical instructions should be
useful in this stage. The associative stage is characterized by a focus on “how” to
actually accomplish the skill. During this phase, verbal instructions become increasingly
irrelevant. The final, autonomous phase is characterized by the development of
automaticity, i.e. skilled performance without attentional allocation.
Stage models of this type are appealing because they match our anecdotal
impression of how motor learning actually occurs. We are all familiar with the
phenomenon of receiving verbal instructions from our trainers early in learning, and
repeating those instructions back to ourselves during practice and performance (“Stand
up straight, release your jaw, breathe deeply, relax your throat, now…..speak!” or
“sing!”) Unfortunately, there is little if any support in the data for the notion that verbal
coaching emphasizing mechanical principles is useful for learning. In fact, there is
considerable evidence to the contrary. We discuss such evidence shortly in our review of
controlled and automatic processes in motor learning.
Open-loop models of motor control
3/10/2016
Titze & Verdolini, © 2002
10-6
Early in the 20th century, Lashley (1917) described a gunshot wound patient who
exibited deafferentation to the lower limbs (no sensations), but intact efference (motor
output). The remarkable observation was that the patient could position his leg with
“surprising accuracy,” similar to normal controls. Such observations pointed to “openloop” mechanisms of motor control, in which a central “motor program” generates
cognitive commands that lead to accurate movement, in the absence of sensory feedback.
It turns out that such models are pertinent to some types of motor control for
existing skills. However, they are poor models of motor learning. Skilled behavior,
especially for rapid movement, clearly can be produced via open-loop control. But it is
virtually impossible to account for motor learning without considering sensory feedback.
Bizzi is credited with providing critical empirical evidence to this effect, by
demonstrating that deafferented primates who were deprived of sensory feedback failed
to acquire new motor patterns, although they readily produced previously learned ones
(e.g. Bizzi, year).
The relevance of observations about open-loop control is that they strongly
suggest the presence of “motor programs” that contribute to movement regulation. Much
debate has existed over the nature of such programs, and their role in motor learning. We
shall review such debates next, in our discussion of cybernetic, “closed-loop” theories,
“schema theory,” and “dynamics theory” of motor learning.
Closed-loop (“cybernetic”) models of motor control and learning
Closed-loop models of motor control and learning are based on “man-machine” or
“cybernetic” models of human performance. Such models postulate the presence of a
mental “template” which specifies environmental goals for movement, and programming
3/10/2016
Titze & Verdolini, © 2002
10-7
parameters predicted to achieve them. Upon movement execution, sensory feedback
from the output is compared to the template. If differences exist between the intended
and achieved output, an error signal is generated. Program parameters are adjusted to
improve the outcome for subsequent trials.
An example of a closed-loop model of speech production was proposed by
Fairbanks (1954) (Figure 10-1). According to this cybernetic model of learning, a motor
command is issued through a mixer to an effector unit, and output occurs. Sensory
information from the movement then ascends to a comparator, which generates an error
message indicating any discrepancy between the original command and the command’s
sensory consequences, including those derived from the environment. The error message
is relayed to the mixer for on-line modification of running commands, as well as to a
storage unit for more permanent modification of stored commands. By these
mechanisms, performance improves over successive trials, and learning is said to occur.
Although closed-loop models improve on open-loop ones by addressing motor
learning in some fashion, several problems are noted. A first major problem regards the
“storage” issue. If, as implied by the models, humans have a separate template or
program for every possible movement, the amount of central neural storage required
would be unfathomable. A second major problem is that neither open- nor closed-loop
models can explain how people can perform novel movements competently on the first
trial, without corrections. Arguably, every movement that we produce is a new one, even
if the general domain is familiar. Simple versions of cybernetic models cannot account
for competent performance on the first trial of any behavior. Yet, skilled individuals
3/10/2016
Titze & Verdolini, © 2002
10-8
routinely demonstrate good performance under novel conditions. Schema theory
(Schmidt, 1976) addresses both storage and novelty problems in motor learning.
Schema theory
Schema theory (Schmidt, 1975, 1976) was developed to account for the storage
problem and the novelty (generalization) problem in prior theories of motor control and
learning. Before we describe what schema theory is, and how it explains motor learning,
we need to introduce its component parts. Schema theory distinguishes three types of
“schemata,” or mental representations relevant to motor control and learning: the motor
program schema, the recall schema, and the recognition schema.
The motor program schema. A general definition of a motor program is a mental
representation of movement that is activated prior to movement, and that causes it. Some
of the strongest evidence for the existence of motor programs is documentation of
previously acquired skilled behavior in individuals deprived of sensory feedback (e.g.
Lashley, year; monkey deafferentation studies; cite). Such observations strongly imply
some sort of central representation of movement.
A key question has regarded the contents of the motor program. What does is
represent? One of the ways that this question has been addressed has been by searching
for aspects of movement which tend to be invariant, across many exemplars in a class of
movements (e.g. what is consistent about a given person’s writing, tennis stroke, or
vocalization pattern?). Various inquiries have shown that very few movement parameters
are constant within a given class of behaviors. The primary ones appear to be: (a)
relative timing, or phase, of movement components, and (b) force ratios across
3/10/2016
Titze & Verdolini, © 2002
10-9
component muscle groups (cites). Based on this evidence, it has been argued that these
factors are essential aspects of the motor program.
Recall schema. The recall schema is conceived as an abstract relation among
“initial conditions,” “response specifications,” and “outcomes” associated with prior
actions (Schmidt, 1975, 1976). Initial conditions are those which exist prior to action
(e.g. the acoustic environment for a teacher of singer). Response specifications are
parameters applied to the motor program, which modulate it (e.g. muscle selection,
movement magnitude, overall movement force, etc.) Outcomes are the physical
consequences of a movement in the environment. The recall schema itself can be
conceptualized as a type of three-dimensional “regression plot” in the head, which relates
these parameters (Figure 10-2). The idea is that the more data points the regression plot
accumulates in practice, and the greater the distribution of those data points is, the
stronger and more accurate the regression line—or recall schema will be. We will return
to the relevance of this point for motor learning shortly.
Recognition schema. The recognition schema is conceived as the abstract relation
between past outcomes and past sensory consequences or feedback, given specified initial
conditions (Figure 10-3). This rule-based abstraction allows for interpolation of what
expected sensory consequences should be for novel initial conditions, given a desired
outcome. Again, the strength of the recognition schema should be related to variability of
sensory experience within a given class of movements.
Schema theory of motor control. Schema theory accounts for both open- and
closed-loop control mechanisms. Rapid movements, which are too fast to respond to
feedback control (e.g. < 100 ms), are accounted for by recall schemata. On-line
3/10/2016
Titze & Verdolini, © 2002
10-10
corrections which occur during slow movements (> 100 ms) are guided by recognition
schemata.
Schema theory of motor learning. How, then, does schema theory account for
motor learning? The answer lies with mechanisms proposed to modify both recall and
recognition schemata, and their interaction. The basic conceptualization is that motor
programs are parameterized for each trial via response specifications specifying aspects
of movement which vary from trial to trial. Such aspects include muscle selection (e.g.
Klapp, 1977), movement time and speed of response (Shapiro et al, 1981; Viviani &
Terzuolo, 1980; Summers, 1975; Shapiro, 1977, 1978; Armstrong, 1977), movement
magnitude (Merton, 1972), and overall movement force (Schmidt, 1981). For each trial,
the learner covertly registers the response specifications, and their relation to initial
conditions and outcomes, to progressively update the recall schema (Figure 10-2). With
increasing practice trials, a “regression line” is built up in the head. (Add how
recognition schema interacts.) It stands to reason that with increasing numbers of initial
conditions (variable practice), the data plot will allow increasingly accurate selection of
response specifications to meet outcome objectives, for a variety of novel initial
conditions. Stated differently, variable practice should enhance generalization of
competent performance to novel conditions. Another prediction is that information about
performance outcomes is critical for motor learning. Both of these predictions are borne
out by empirical data, as reviewed in our section on “laws of practice” later in the
chapter.
To relate these points to the situation of voice training more specifically, the
following findings should obtain. Learners who practice “yawn-sigh” in a large number
3/10/2016
Titze & Verdolini, © 2002
10-11
of contexts, including different phonetic and linguistic contexts, in different physical
environments (therapy room, cafeteria, restaurant, classroom, stage), should be able to
generalize “yawn-sigh” to new phonetic and linguistic contexts and physical
environments situations better than people who practice yawn-sigh in limited contexts
(few verbal stimuli, in a therapy room). Moreover, learners who receive valid feedback
about their performance in on “yawn-sigh” phonation should learn it better than learners
who do not.
In recent years, schema theory has accounted for paradoxical findings that both
variable practice and infrequent augmented feedback (feedback provided by the trainer or
some device) tend to reduce performance gains during practice, but enhance learning
seen at long-term retention. We will return to a discussion of this issue as well under
“laws of practice,” below. At this juncture, we will simply say that it appears that the
effort that the learner puts into the development of mental schemata depresses immediate
performance, but enhances learning. Variable practice, infrequent knowledge of results
and other factors appear to induce effortful processing in the learner, thus engendering
the paradoxical results.
Finally, it should be noted that although schema theory addresses higher-order
cognitive aspects of motor learning, the theory allows for lower-order mass-spring
(reflexive) aspects of motor control emphasized by dynamical theories of motor control
and learning discussed next.
Dynamical theory of motor control and learning
Dynamical theory of motor control and learning (e.g. Kelso, 1995; Wallace, 1996)
is linked to physical theories about self-organizing properties of organisms and systems
3/10/2016
Titze & Verdolini, © 2002
10-12
(e.g. Haken, 1977; 1983a, 1983b). In the dynamical approach, movement is not
necessarily seen as an ultimate goal. Rather, humans are seen as self-organizing systems
in which movement occurs as an “emergent” property arising from interactions of
“coordinative structures,” the environment, and task requirements. Our Postulate #3
(Chapter 7) addresses such concepts by describing examples of probable coordinative
structures in voice. These are sets of structures, which combined, tend to produce stable,
efficient performance or “attractor states” during vocalization, due to the characteristics
of the structures’ reciprocal physical interactions.
Coordinative structures have been a focus of interest in speech science for
decades (cites; examples). Recently, specific structures, or attractor states have been
described for voice, as well. Examples are vowels produced with low-frequency, speech
(chest register) voice quality following glottic onset, versus high-frequency, falsetto voice
quality following aspirate onset (Steinhauer, 2000). These patterns are appear to be more
readily learned, and demonstrate greater stability than some other patterns (for example,
mid-frequency, mixed register, simultaneous onset voice; Steinhauer, 2000).
Undoubtedly, other attractor states exist for voice that have not yet been formally
identified.
Dynamical theory is clear that attractor states may be caused by combinations of
physical properties of tissues as well as past experience. Moreover, as noted, in
dynamical theory, attractor states are considered to be subject to environmental
influences and task demands (Zanone & Kelso, 1992).
Much of the evidence surrounding dynamical theory has come from studies of
bimanual control. In such studies, subjects are asked to perform one of several possible
3/10/2016
Titze & Verdolini, © 2002
10-13
coordination tasks. Some tasks tend to be inherently stable and subjects achieve accuracy
for them virtually immediately. These include 0-degree or “in-phase” tasks (in which
homologous muscles on the left and right are used synchronously: tap fingers on both
hands at the same time), and 180-degree or “anti-phase” tasks (in which homologous
muscles are completely out of phase: tap fingers on left and right hands in temporally
symmetric alternation). Other tasks tend to be inherently unstable, and require
considerable practice, if they are acquired et all (e.g. 90-degree phase tasks, in which
tapping in one hand lags behind tapping in the other by ¼ of a cycle; see for example
Yamanishi, Kawato, & Suzuki, 1980).
One of dynamical theory’s predictions about motor learning is that the
establishment of new action patterns which are far from current attractor states involves
characteristic phases of instability (Zanone & Kelso, 1992). Learning is not a gradual,
incremental process in such cases. Rather, learning appears to occur abruptly—or “nonlinearly,” as new attractor states are established. Although introduced as if a somewhat
novel notion in the context of dynamical theory, in fact notions about abrupt versus
incremental learning are anything but new in the learning literature (e.g. Tolman, year;
versus Skinner, year; Bryan & Harter, 1897; Fuchs, 1962; Lee & Swinnen, 1993).
Regardless, dynamical theory makes the point that learning new attractor patterns
characteristically involves periods of performance destabilization, in which existing
attractor states are perturbed (see for example, Lee, 1998). During those periods, subjects
learn how escape from the existing patterns, and start approaching the new one. Average
error for the new pattern may decrease, but performance variability increases temporarily
(Lee, Swinnen, & Verschueren, 1995).
3/10/2016
Titze & Verdolini, © 2002
10-14
Dynamical theory indeed has the ability to explain much about motor control, and
has emphasized issues that had been quite overlooked in earlier models: physical
properties of neuromuscular systems themselves, and their interactions. Where
dynamical theory falls short is in its relative disregard for what is seen to be a vital role of
cognition in motor control and learning. Dynamical theory ignores the notion that there
is any necessary existence of mental representations of movement (motor programs), or
memorial representation of intentions established by past behavioral information. In fact,
numerous phenomena of motor learning are difficult to explain without such constructs.
Examples are reviewed by Lee (1998; in press). One example has to do with elderly
adults, who generally failed to learn a 90-degree phase bimanual coordination task at all,
despite learning of it by younger adults (Cunningham, Wishard, & Lee, 1998). Yet,
when the elderly subjects were provided specific strategies for the task, and augmented
feedback regarding their performance, their learning improved (Wishart, Lee, &
Murdock, 1998). Such findings are difficult to explain without invoking memorial
processes and mental representations of movement (Lee, 1998).
This type of finding suggests what few trainers and therapists would argue:
Cognition clearly is involved in motor control and learning. But what is the nature of
such mechanisms? We consider this question in the next section.
Recent Studies in Motor Learning
Attentional versus automatic processes
Recently, an ancient distinction between two types of mental processes has come
to the fore again, pertinent to motor learning (for review, see Epstein, 1994). The
distinction of interest in the present context has to do with what are called “controlled”
3/10/2016
Titze & Verdolini, © 2002
10-15
versus “automatic” processes. Classically, “controlled processes” are defined as
attentional.1 They involve conscious awareness, and are regulated by an extremely
limited-capacity, slow, serial processor (7 + 2 items capacity, processing time of
approximately 100? ms per item). Although slow, controlled processes are flexible.
They can either facilitate or inhibit other cognitive functions. In contrast, automatic
processes are defined as non-attentional and non-conscious. They are characterized as
regulated by a virtually unlimited-capacity, rapid, parallel processor. More rigid and
stereotypical than controlled processes, automatic processes can only facilitate other
cognitive operations (e.g. Posner & Snyder, 1975). Considerable experimental data exist
consistent with these two general classes of mental processes (e.g. Neely, 1977; other
cites).
The specific relevance of controlled versus automatic processing for motor
learning is that much of motor learning can be seen as a dance between them. The wrong
application of one or the other process in training, at the wrong time, probably can
undermine the most brilliant biomechanically-based intervention model.
Role of automatic versus attentional processes in motor learning. Evidence
suggests that controlled processes are irrelevant to some aspects of motor learning. One
example is elderly individuals, who tend to have reductions in controlled, attentional
resources. Yet they show motor learning (cites). Assuming that some sort of cognitive
processes are involved in motor learning, perhaps non-conscious, automatic processes are
responsible, at least in these individuals
1
Attention is defined in different ways in the cognitive literature. A general definition is that attention is a
limited-capacity mental resource that is directed towards some objects or events, and away from others,
increasing the likelihood that the selected items will be further processed. Thus, attention is seen as both a
filter and a limited-capacity cognitive quantity. It is often equated or at least coupled with consciousness.
This coupling may be a simplification. We will set ignore the simplification in the present chapter.
3/10/2016
Titze & Verdolini, © 2002
10-16
Other evidence more powerfully suggests that conscious attention not only is
irrelevant, but actually harmful when directed to certain aspects of learning. That
evidence indicates that, contrary to popular techniques in voice pedagogy and athletics,
directing learners’ attention to the biomechanics of motor tasks reduces both motor
performance and learning.
One example of this principle was reported by Wulf and Weigelt (1997). These
authors studied people learning ski-like motions on a ski simulator. In their Experiment
1, subjects who received mechanical instructions about how to perform the ski simulation
task—and thus whose attention was directed to task mechanics, had poorer training
performance and learning than subjects who were given no instructions at all (Figure 104).
A second example was reported by Hodges and Lee (1999). These authors
studied learning for a difficult, non-preferred bimanual coordination task involving a 90degree phase difference across the limbs. Also subjects in this study had the best training
performance and learning—and the least bias towards preferred patterns—when they
were given no instructions about how to do the task, in comparison to subjects who
received either general or specific instructions emphasizing mechanics (Figure 10-5).
A third example was reported by Verdolini-Marston and Balota (1994), who
looked at learning for a manual tracking task. In Experiments 2 and 3 of their studies,
two groups of subjects received instructions that were supposed to enhance learning. One
set of instructions directed subjects’ attention to the perceptual aspects of the task
(“Concentrate on the rotating target.”) The other set of instructions directed subjects’
attention to mental “images” similar to those widely used in voice pedagogy, which are
3/10/2016
Titze & Verdolini, © 2002
10-17
supposed to enhance the mechanics of learning (e.g. “Imagine that your arm is like a
locomotive wheel”). Both sets of instructions should have increased subjects’ attention
to the mechanics of learning. A third group of subjects did not receive any instructions
on the mechanics of the task. Figure 10-6 shows the results for Experiment 3. Subjects
who received mechanically-related instructions had poorer learning than subjects who
received no instructions at all. This finding extended to instructions involving
metaphoric images, which are supposed to circumvent “mechanical” approaches.
A similar example was reported by Verdolini and colleagues, for a voice task
(Verdolini et al., in preparation). In that study, subjects were given the task of learning to
initiate vowels with a target amount of pre-phonatory airflow (100 ml/sec), which
generally corresponds to “simultaneous voice onset.” One group of subjects received
metaphoric images that were derived from a broad survey of voice therapists and voice
teachers2 and pilot data indicating subjects’ preferences for various images proposed by
the professional trainers. The image selected based on these procedures was: “Imagine
that your voice is a never-ending wave.” The results were similar to those for the manual
tracking study (Verdolini-Marston & Balota, 1994). Subjects who received the imagery
instructions had poorer learning than subjects who were not given any instructions on
how to do the task (Figure 10-7). In both investigations, this result occurred despite the
fact that subjects rated the image as “quite useful” for learning. The interpretation is that
even metaphoric images used in voice training, which are supposed to circumvent
“mechanical” training, actually invoke attention to task biomechanics and thus reduce
learning.
2
(American Speech-Language-Hearing Association Special Interest Division 3: Voice and Voice Disorders
listserve, and Boston Chapter, National Association of Teachers of Singing)
3/10/2016
Titze & Verdolini, © 2002
10-18
The combined findings suggest that instructions directing attention to
biomechanical aspects of training are not useful. In fact, they appear outright harming for
motor learning. This conclusion extends to the case of metaphoric images which are
often favored in voice training due to their “non-mechanical nature.”3
Cautions about an emphasis on biomechanics in training are echoed by some
high-level athletic trainers. One example is the tennis pro, Timothy Gallwey. Gallwey
(1997) notes that a good way to undermine your opponent’s game is to ask him casually,
while switching courts, what he’s doing today to have such a good forehand (Gallwey,
1997). The notion is that by directing his attention to the mechanics of his game, his
performance will be disrupted.
The findings are perplexing and even troublesome in light of typical paradigms
used in voice training and therapy. Such paradigms tend to favor the direction of
controlled, attentional processes to the biomechanics of performance (“Stand up straight,
breathe deeply, drop your jaw.”). One interpretation is that “discovery learning” works
better than imposed learning strategies in the motor domain (Hodges & Lee, 1999).
Another interpretation is that the biomechanical degrees of freedom to be solved in motor
tasks simply are too great to “fit” into the limited-capacity, slow, serial processing
mechanisms associated with consciousness (e.g. Neely, 1977), and are instead best
handled by automatic processes. Another, complementary interpretation resonates with
anecdotal impressions for many performers and athletes: too much attention to one’s
performance—or “trying too hard”--can mess it up. We will consider this interpretation
3
Note that literal mental imaging of one’s performance, as occurs in common athletic visualization
procedures, is different from metaphoric imaging of the type we have studied here. Visualization may be
useful for both motor performance and learning. We will return to a discussion of this issue later in this
chapter.
3/10/2016
Titze & Verdolini, © 2002
10-19
more fully in the next section of our chapter sub-section on “intention” in motor learning.
Regardless of the explanation, the findings strongly suggest a detrimental role of
attention to the biomechanics of the task, for motor learning.
Interestingly, other literature has indicated that conscious attention directed to
other aspects of training is relevant for motor learning. The relevant literature has
examined the effect of locus of attention in learning. An extensive review, which guided
much of the summary in the following pages, is provided by Wulf and Prinz (2001).
Effects of instructions prompting internal versus external locus of attention in
motor learning. A series of studies have examined the influence of instructions
promoting an internal versus external locus of attention on motor learning. An internal
locus involves attention directed towards movements themselves (e.g. arm swing in golf).
An external locus involves attention directed towards the movements’ environmental
effects (e.g. club swing in golf). Again and again, studies have shown that attention does,
indeed, appear relevant for motor learning. The catch has to do with where attention is
directed. An internal locus of attention, involving a focus on intrinsic biomechanics,
appears to depress both performance and learning. This finding is consistent with those
just reviewed, which indicate a negative effect of instructions directing people’s attention
to the mechanics of learning (see preceding paragraphs). In contrast, an external locus of
attention—or more precisely an external locus directed at movements’ effects, appears to
enhance both performance and learning. These findings contradict almost every vocal
pedagogues’ and therapists’ preference for an internally-oriented, mechanical focus in
training.
3/10/2016
Titze & Verdolini, © 2002
10-20
A review of some of the critical literature is as follows. Wulf, Höß, and Prinz
(1998) looked at learning on a ski simulator. One group of subjects was instructed to pay
attention to the force of their feet on the simulator platform (internal focus), whereas
another group was instructed to pay attention to the force that their feet produced on the
wheels underneath the platform, directly below their feet (external focus). A control
group received no instructions about attention. The external focus group demonstrated
better performance and learning than either of the other two groups. The internal focus
group had no better learning than the group who received no instructions at all. In the
same report, a second experiment examined learning for a stabilometer (balancing) task.
One group of subjects received instructions to focus on maintaining a horizontal foot
position for the task (internal focus), while another group was instructed to focus on
maintaining a horizontal orientation in two markers attached to the platform directly in
front of their feet (external focus). Results are shown in Figure 10-8. At a 1-day
retention test, the external focus group demonstrated better learning (Wulf et al., 1998).
Although the actual distance between the attentional foci was minimal across the two
groups, a difference in learning was attributable to the object of attention: internal (poorer
learning) versus external (better learning).
Similar principles were tested in a more real-world situation involving golf. In a
study by Wulf and colleagues (Wulf, Lauterbach, & Toole, 1999), golfers were asked to
pay attention either to their arms movements during golf swings (internal locus) or the
swing of the club (external focus). The external focus groups achieved greater accuracy
in their shots during practice and also during a later retention test. Maddox, Wulf, and
Wright (1999) reported similar findings for tennis players who were instructed to pay
3/10/2016
Titze & Verdolini, © 2002
10-21
attention to the mechanical dynamics of their swing (internal focus) versus the ball’s
trajectory and its landing point (external focus). Wulf, Shea, and Park (in press)
determined that superior learning for an internal versus external locus of attention is
unrelated to subjects’ personal preferences, but is rather a general characteristic of the
human motor learning process.
Wulf and colleagues (Wulf, McNevin, Fuchs, Ritter, & Toole, 2000) attempted to
determine whether the advantage for an external locus of attention is due to the external
locus, per se, or whether it is due to not paying attention to one’s movements. The study
involved novice tennis players. One group was instructed to pay attention to the
approaching ball (antecedent condition), while another group was instructed to pay
attention to arc of the ball as it left the racket (effect condition). Thus, both conditions
favored an external locus of control. The difference across the groups had to do with
whether the external locus emphasized antecedent conditions or movement effects.
Results showed that subjects in the “effect” condition had better performance during
training and better learning as seen at retention test, than subjects in the “antecedent”
condition (Figure 10-9). This finding demonstrated that distraction from one’s
movements alone is not sufficient to enhance learning. Rather, an external focus aimed
at movements’ effects seems to be critical.
Having established as much, Wulf and colleagues asked whether an increasing
distance between body movements and observed effects would increase the advantage of
an external, “effect-oriented” attentional locus for learning. A review of some of their
prior studies indicated that it might (for review, see Wulf & Prinz, 2001). Indeed, an
additional study by McNevin, Shea, and Wulf (2001) found evidence consistent with this
3/10/2016
Titze & Verdolini, © 2002
10-22
hypothesis, by showing that markers placed progressively further from the feet, up to 20
cm, maximized performance and learning for the stabilometer task, in comparison to
markers closer to the feet (Figure 10-10). However, a further experiment found that the
effect is not monotonic. Golfers were instructed to focus either on the swing of the club,
or on the ball’s trajectory and its target. In this case, the more remote locus of attention
(trajectory and target) produced worse learning than the nearer locus (swing of the club)
(Wulf, McNevin, Fuchs, Ritter, & Toole, 2000). Thus, a slight external locus of attention
facilitates learning compared to a more proximal focus. But an extremely remote external
locus again degrades learning. A possible explanation is that a proximal external focus
keeps subjects’ attention on the movements’ effects, but is close enough to the body that
some cognitive resources “spill over” to the mechanics and assist learning. We will
discuss the links to voice training, which are quite exciting, in a summary of this chapter
section. For now, the implication is that changes in movement patterns do involve some
level of cognitive resources applied to mechanical processes. The evidence suggests that
those “resources” are automatic and non-conscious.
Effects of internal versus external locus of attention in augmented feedback in
motor learning. In addition to studies looking at the locus of attention at the time of
movement, other work has concentrated on the locus of attention in augmented feedback
which is critical to much learning. Augmented feedback is information provided by a
trainer or by some devices, which adds to learners’ intrinsic sensory feedback (Schmidt &
Lee, 1999). In one study, Shea and Wulf (1999) used the stabilometer task to address this
issue. Subjects in two groups were presented the identical feedback displays regarding
their ongoing performance on the stabilometer. The only difference across the groups
3/10/2016
Titze & Verdolini, © 2002
10-23
was the interpretation that subjects would have of the displays. One group was told that
the displays represented their own movements (internal focus). The other group was told
that the displays represented the stabilometer’s displacements. Two other groups were
told to pay attention to their feet (internal focus) and to markers on the stabilometers
(external focus), respectively, as in prior experiments already described. Retention
testing showed that not only an external focus relative to performance itself, but also an
external focus relative to augmented feedback enhanced both performance and learning
more than internal foci (Figure 10-11).
Similar findings were reported by Hodges and Frank (2000). These authors
looked at the effect of an internal versus external locus of attention, manipulating
augmented feedback for a bimanual coordination task involving a non-preferred pattern,
specifically a 90-degree phase difference in timing across the limbs. Two groups
received instructions that were considered to engender an external focus by viewing
augmented feedback while another subject performed the task, as seen on videotape, prior
to initiating the task themselves. One group was instructed to observe the augmented
feedback alone, and the other group was instructed to observe the feedback in relation to
the individual’s movements. A third group was considered to have an internal focus, as
they were simply given demonstrations of the movements prior to initiating practice. As
in previous studies, the groups with the external focus demonstrated better performance
and learning than the internal focus group.
In fact, one argument supporting the use of biofeedback in training is that it may
engender an external locus of attention. A focus on the effects of movements as seen on
oscilloscopes or other displays theoretically should enhance learning. A caution comes
3/10/2016
Titze & Verdolini, © 2002
10-24
with withdrawal of the feedback. If biofeedback adds information above and beyond the
subject’s own internal feedback, such information may become incorporated with the task
in the subject’s head. The danger is that when this information is withdrawn during later,
no-biofeedback, “real-world” trials, performance decrements may result (see Verdolini &
Krebs, 1999). Wulf and Prinz (2001) have suggested that biofeedback that reinforces
intrinsic feedback, i.e. is redundant with it but brings the focus outside the body, does not
bring with it the same concerns about withdrawal phenomena.
Finally, a key, “real-world” experiment that may relate to voice training was
described relative to the volley-ball serve. In that experiment, some volley-ball players
were provided augmented feedback after each trial from coaches, which directly
addressed the players’ movements (shift your weight from the back to the front leg when
you hit the ball; internal locus). Other players received augmented feedback with similar
content, but was “translated” to emphasize an external locus (shift your weight towards
the ball when you hit it). The latter group had better serves both during practice and at 1wk no-feedback retention test (Shea & Wulf, 1999). We will return to the relevance of
this finding for voice in this next, summary subsection of this chapter.
Summary and discussion of attention and automatic processes in learning, and
implications for voice training and therapy
Recent research in motor learning has produced evidence contrary to
preferred training styles in many voice clinics and studios. The evidence
strongly underlines the principle that knowing “what” to train does not
translate directly to “how” to train it. Knowledge of biomechanical and
acoustic principles of voice production is insufficient to train those same
principles effectively.
One of the suggestions that emerges from the motor learning literature is
that directing attention to biomechanical maneuvers in voice production
generally deteriorates both performance and learning. Directing attention
to the effects of the maneuvers seems to enhance them. Possibly, the
3/10/2016
Titze & Verdolini, © 2002
reason has to do with learners’ attempts to acquire a new coordinative
pattern. When attention is directed to movements, the effect seems to be a
stabilization of existing, “old” stable patterns (e.g. Hodges & Lee, 1999;
Hodges & Frank, 2000). Focus on movements’ effects may direct
attention to the emergent, abstract properties of the new coordinative
pattern that is the focus of learning, and away from existing patterns (Lee,
in press).
A critical question for voice is: What are “effects” in voice training?
What is the analogue of the club swing in golf, or the racket swing in
tennis? Are we talking about acoustic or auditory-perceptual effects?
Perhaps, but not necessarily. And certainly not solely. First, as suggested
in our Postulates (Chapter 7), the “effects” we are going for in voice
generally have to do with some sort of efficiency or power transfer.
However formulated, these constructs have to do with output-input
relations. Thus, the relation of physical action and acoustic output are
essential “effects” in voice training. How might such relations be
manifested, as the optimal foci of attention in learning in voice training?
Part of the answer may lie with findings about a U-shaped curve function
of distance between production structures and the attended “effect,” and
performance and learning gains. This function is shown in Figure 10-12.
Although attention to movements’ effects may useful, its utility seems to
be maximized for some intermediate distance. Utility seems to increase as
the attentional locus becomes progressively farther from production
structures (e.g. golf club as opposed to arm swing), but only to a point.
Extremely remote attentional foci (e.g. where golf ball lands) result in
declines of performance and learning, relative to optimal (McNevin et al.,
2001; Wulf et al., 2000). Perhaps voice trainers’ bias to “not listen to
yourself” is a reflection of this principle. Perhaps the auditory signal
simply is too far from the effectors to be the best locus of attention. What,
then, might represent an intermediate locus in voice training—equivalent
to the club swing in golf training?
Perhaps sensations produced by the effectors, but not isomorphic to them,
are an optimal solution in voice training. This solution actually
corresponds to one used by many voice trainers. An example is seen in
Lessac-Madsen Resonant Voice Therapy, modeled after work by voice
pedagogues Lessac and Madsen (Berry et al., 2001; Verdolini, 2000). In
that training, learners’ attention is directed to easy vibratory sensations in
the oral cavity during voicing. This parameter reflects both the strength of
the acoustic spectrum, and simultaneously, vocal fold adduction (see
Chapter 1+ for discussion; see also Berry et al., 2001). More important,
this percept reflects various output/input relations—or “emergent
properties” that we consider valuable in voice, including vocal efficiency
[output/(subglottic pressure x flow)] and vocal economy [output
3/10/2016
10-25
Titze & Verdolini, © 2002
10-26
intensity/impact intensity] as discussed elsewhere in this text (Chapter 1+;
Chapter 7, Postulates). By focusing on easy oral vibratory sensations in
voice production, attention is clearly directed towards movements’ effects
rather than the specific mechanical maneuvers used to achieve the effects.
But what do we do if we feel that our only recourse is to address directly
biomechanics, because other approaches have failed? Perhaps we can take
a lead from the study on the volley-ball serve, described earlier (Shea &
Wulf, 1999). Perhaps we can address technical considerations in voice by
emphasizing their interaction with “effect” variables. For example,
instead of saying, “compress your abdomen upwards into the ribcage as
you hit the high note,” we might say, “compress your air into the tone as
you hit the high note” (if that is our bias). Although the verbal
differences are small across the two instructions, evidence suggests that
they may be important.
Another part of the answer may lie with the manipulation of “intention.”
This is the topic of the next section.
The role of intention in motor control and learning
For practical purposes, we can liken “intention” to a racheted-up version of
“attention.” One can attend to something without intending to act on it. For the purposes
of this review, we will consider “intention” to occur when attention is fueled by
motivation to achieve an intended goal.
We initiate our discussion of intention by referencing observations of motor
learning in individuals with anterograde amnesia. These individuals demonstrate motor
learning despite the fact that they typically fail to recall prior practice episodes. Thus
they have no “intention” of demonstrating learning derived from those episodes, at
retention test. Yet, they demonstrate motor learning normally or near-normally (e.g.
Eslinger & Damasio, 1986; Kaushall, Zetin, & Squire, 1981; Martone, Butters, Payne,
Becker, & Sax, 1984; Milner, 1962; Tranel, Damasio, Damasio, & Brandt, 1994; Vakil,
Grunhaus, Nagar, Ben-Chaim, Dolberg, Dannon, & Schreiber, 2000; Yamashita, 1993;
Zola-Morgan, Squire, & Mishkin, 1982).
3/10/2016
Titze & Verdolini, © 2002
10-27
This observation has prompted many researchers to distinguish two types of
learning and memory systems, or processes. One type, which depends on the
hippocampus, amygdala, and certain thalamic structures, is impaired in amnesia. It is
variably called “declarative” or “explicit” memory. The other type does not require the
integrity of these structures, and is called “procedural” (non-declarative) or “implicit”
memory (e.g. Squire, 1986; Graf & Schacter, 1985).
Numerous studies have examined the cognitive characteristics of these two
different memorial phenomena (for reviews, see Epstein, 1994; Roediger, 19xx; Schacter,
19xx; Verdolini, 1997; Verdolini & Lee, in press). Table 1 summarizes the findings.
Here, we focus on differences in “intention” for explicit versus implicit memory.
Key findings were reported by Schacter and Graf (1986), in a study of amnesic
individuals. Individuals with amnesia were given a list of words to study (e.g.
“telegraph,” “bandana,” “radiator”…..; get actual words they used). In a subsequent test,
subjects were given word stems (e.g. “tel___;” “ban___;” “rad___”…..;) and asked to
complete them with the first words that come to mind. As for numerous other studies
before and since, subjects tended to complete the stems with previously studied words at
a rate that was far greater than chance. Thus, they demonstrated “implicit memories” for
the words. In another test, subjects were given the same words stems, and asked to use
them purposefully to try to recall previously studied words. They failed. In fact, most of
them did not even recall that they had studied the words before. In sum, implicit
memories were shown when they were triggered by non-intentional processes, in
response to environmental stimuli. When memory-impaired individuals attempted to
retrieve those same stimuli intentionally, they failed to produce them. This type of
3/10/2016
Titze & Verdolini, © 2002
10-28
observation has led to the conclusion that implicit memory—possibly including motor
learning--depends on spontaneous, triggered processes.
If findings for implicit memory in the verbal domain apply to implicit memory in
the motor domain (i.e. motor learning), we might conclude that motor learning may
belong to a memorial system that does not work well under “intentional” conditions.
“Trying” or “intending” to evoke motor memories driving performance may limit the
likelihood of success. This conclusion fits with anecdotal impressions about problems
associated with “trying too hard” in athletics and voice training. It also fits with formal
reports about “trying to hard” in the scientific literature. One example was provided by a
study looking at video game playing. Experimenters in that study first covertly
documented the performance of people playing video games, who did not know they
were being observed. Then the experimenters asked the players to try for their best
possible scores. Players decreased their scores by an average of 25% under the
“intentional” condition (Baumeister, 1984, cited in Wulf & Prinz, p. 648).
Together, the observations are consistent with the notion that “trying too hard” is
a problem in the motor domain. “Intending” to achieve a given performance seems to
produce the paradoxical effect of reducing it, and perhaps reducing learning as well.
We might be happy with this conclusion, and proceed to figuring out how to
apply it to voice training, except for one problem: the literature as well as practical
experience provide abundant evidence that intention does play a role in motor
performance and learning. Some of the most striking findings come from a speech study.
Cole and Abbs (cite) examined healthy subjects who were required to produce repeated
strings of bilabial consonant-vowel syllables (e.g. /ba ba ba ba ba/). Bilabial contact was
3/10/2016
Titze & Verdolini, © 2002
10-29
virtually wholly accounted for by lower lip activation; upper lip muscles did not activate.
On randomly selected trials, the experimenters perturbed the lower lip, preventing it from
rising to achieve closure with the upper lip. Subjects achieved lip contact nonetheless.
They did so by activating the upper lip—which they had never done before, bringing it
down to the lower one. Thus, subjects’ “intentions” to achieve bilabial closure clearly
dictated their actions.
Interestingly, although their “intentions” may have been conscious, the
mechanisms that mediated intentions’ achievement appeared to be non-conscious, and
automatic. This conclusion was based on electromyographic (EMG) recordings showing
that the upper lip muscle initiated its corrective action about 70 ms after the lower lip had
been perturbed. This is considerably slower than the fastest possible reaction involving
consciousness (about 100 ms), and faster than low-level, reflexive corrective actions
(about 35 ms). The 70-ms interval between perturbation to (EMG) response is consistent
with what the authors called a “transcortical response.” This hypothetical response
involves the travel of the sensory, perturbation signal to the cortex, where it makes
contact with the subject’s “intention,” in some form. Corrective calculations must have
been carried out covertly (non-consciously), because conscious calculations would have
taken much longer.
The findings are quite interesting, and illustrative regarding the role of
consciousness in intentional actions. The study suggests that consciousness may play a
role in establishing an intention. However, mechanical corrections needed to achieve the
intention appear handled by non-conscious, automatic processes. An important
remaining question is: How does this conclusion relate to motor learning?
3/10/2016
Titze & Verdolini, © 2002
10-30
Lee provides an excellent overview of the role of intentions in motor learning
(Lee, in press). An example is the following. Consider again learning for bimanual
coordination tasks. It is clear that certain patterns (e.g. in-phase and anti-phase patterns)
are inherently stable, and people perform then accurately with few if any learning trials
(e.g. Kelso, 1984, 1995; Swinnen, 20002; Turvey, 1990). Yet, when people are given the
task of learning new, unstable patterns, they do improve with practice if they intend to do
so (e.g. Hodges & Frank, 2000; Hodges & Lee, 1999; Yamanishi et al., 1980). Clearly,
“intention” is relevant to learning in such cases. Otherwise, people would not learn the
tasks. Examples abound in voice studios, clinics, and athletic fields daily.
How do we reconcile reports that intentionality can compromise motor
performance and learning (e.g. Baumeister, 1984, cited in Wulf & Prinz, p. 648; Schacter
& Graf, 1997), with clear evidence that intention is critical for both (e.g. Cole & Abbs,
year; Hodges & Frank, 2000; Hodges & Lee, 1999; Yamanishi et al., 1980)? The most
straightforward reconciliation parallels arguments for the role of attentional and
automatic processes in motor learning. Perhaps the answer lies with where intentions are
directed, and the cognitive mechanisms optimally involved in it. Conscious intentions
directed towards movements’ goals, or effects, seem required for learning difficult motor
tasks (e.g. Hodges & Frank, 2000; Hodges & Lee, 1999: Yamanishi et al., 1980). It
seems that the computation of mechanical solutions requires to achieve intentions are best
left to non-conscious, automatic processes (e.g. Cole & Abbs, year). An example of this
principle in voice training would be intending to achieve easy oral vibratory sensations in
voice, leaving solutions about how to achieve them to automatic, non-consciousness
processes. Moreover, perhaps intentions directed towards relatively proximal effects
3/10/2016
Titze & Verdolini, © 2002
10-31
(e.g. vibratory sensations in voice training) are more effective than intentions directed
towards extremely remote effects (e.g. in the extreme, winning the Met auditions).
Stated differently, conscious intentions probably not only appear useful, but are
nearly required in skilled motor performance and motor learning. Problems appear to
arise when people direct their intentions to the biomechanics of the task, rather than task
effects. Data on intentionality in motor control suggest that biomechanical solutions are
best left to automatic processes outside the learner’s awareness.
Another part of the utility of intentionality for motor learning may lie with the
quality of intention. A well-known “arousal” curve, called the Yerkes-Dodson curve
after its founders (cite), shows that in a large number of cognitive and physical domains,
performance is enhanced by an intermediate level of arousal as assessed by Galvanic
Skin Response (sweating), heart rate, and other means (Figure 10-13). Both low and high
levels of arousal tend to decrease performance relative to some intermediate “optimum”
level (cites). Although we are not aware of any data that have assessed arousal levels
specific to motor learning, it is reasonable to think that intermediate arousal levels favor
learning as well. Thus, it is predicted that a lack of intention to learn (low arousal) as
well as extreme anxiousness around learning (high arousal) should produce poorer
learning than an intermediate motivational level. It could be posited that low arousal
levels fail to bring the needed automatic cognitive resources to the task. High arousal
levels might divert resources away from the task at hand, towards a type of “fight or
flight” response. In the intermediate arousal state, the learner has the right level of
needed cognitive resources appropriate to the job.
Summary and discussion of the role of intention in motor learning, and implications for
voice training and therapy
3/10/2016
Titze & Verdolini, © 2002
The paradoxical finding is seen that although “intention” can sometimes
depress skilled performance and presumably learning, without some form
of intention, learning for difficult tasks can be expected to fail altogether.
We posit that the direction of the intention, its particular cognitive
mechanisms, and its quality, are crucial. Directing conscious intentions
towards a movement’s effects, or goal, should favor learning more than
directing conscious intentions towards mechanical solutions. Such
solutions appear best left to non-conscious, automatic processes.
Moreover, intentions that focalize relatively proximal effects (e.g. easy
oral vibration sensations in voicing) may be more beneficial than
intentions that cathect on extremely remote effects (e.g. winning some
audition). With respect to the quality of intention, it seems reasonable to
think that intentions fueled by intermediate motivational or arousal levels
should lead to better learning than lack of motivation, or extreme anxiety.
What might be further, specific applications in training? What should we
do it we feel we have to address technical, mechanical considerations?
We can borrow some insights from Sports Psychology. Gallwey (1997)
provides an excellent source. In his chapter on “discovering technique,”
Gallwey gives examples of traditional technical instructions, in
comparison to the type of instructions more consistent with the literature
we have reviewed here. An example of a traditional technical instruction
might be: “Hit the ball with your arm fully extended.” An alternative
instruction more consistent with motor learning principles is: “Notice the
degree of bend in your elbow at the moment of impact with the ball.” In
the second instruction, the environmental goal—the ball—is included
along with attention to the mechanics. Also in the second instruction, an
inherent conscious intention is evoked (meeting the ball) without “trying”
to use consciousness to solve the biomechanical problem. Moreover, there
is no “wrong answer.” This should limit negative anxiety.
Possible examples for voice training are given in Table 2. A traditional
instruction might be: “Stand up straight and take a deep breath.” A more
effective instruction might be: “Feel the position of your shoulders and
your breath as you initiate the sound.” Another traditional instruction
might be: “Expand your throat.” A more effective instruction might be:
“Notice the size of your throat as you sing.” Note that examples of
effective instructions do not require the learner to describe his or her
findings, a practice used in many studios and clinics (e.g. “Tell me how
your throat felt on that phase.”) Nor are learners even asked questions
about mechanics (e.g. “How big was your throat on that phrase?”) Both of
these approaches are misguided attempts to induce sensory awareness,
which may be useful in some cases. The approaches are misguided
because they invoke verbal responses, whether covert of overt. There are
no data that we are aware that any type of verbal descriptions of
3/10/2016
10-32
Titze & Verdolini, © 2002
10-33
movement are helpful in voice or any other kind of physical training.
Words simply represent the wrong code for motor learning. One of us
(KV) has used an approach to voice training for nearly 20 yr which
incorporates the types of instructions used in Table 2. Anecdotally, there
is strong support for it among her clinical and pedagogical caseloads.
Sports Psychology brings us another example of a way that intentions can
be used to benefit motor performance and learning. This approach has to
do with “visualization,” which is a strong form of an “intention exercise”
commonly used in athletic training. Visualization involves mental
exercises, which evoke literal mental images of future intended
performance. Imagined performance is not restricted to the visual domain,
but extends to auditory, olfactory, kinesthetic, and gustatory modalities as
well. In the exercises, the learner imagines future successful performance,
in detail, the way he or she would like it to occur. It should be emphasized
that such images are mental analogues of the physical world, not
metaphoric images discussed elsewhere in this chapter as potentially
harmful to motor learning (e.g. “imagine your arm as the wheel on a
locomotive;” or “imagine your voice as a never-ending wave;” VerdoliniMarston & Balota, 1994; Verdolini et al., in preparation). Evidence
suggests that appropriate visualization exercises can strongly enhance both
performance and learning (cites; for examples, see Baum & Trubo, 1999).
The reason probably has to do with the fact that in visualization exercises,
learners’ “intentions” regarding movement effects are exercised in a very
accurate way. Central nervous system physiology is similar for imagined
as compared to real physical actions (e.g. Kosslyn, year). Also, subtle
neuromuscular activation is seen during visualization exercises, similar—
but smaller intensity relative to activation seen during actual performance.
Thus, visualization exercises allow learners to practice critical motor
“intentions” effectively, while minimizing physical fatigue, injury, and
error.
There is every reason to think that “visualization” exercises should be as
effective for voice as it is for athletics. Several books are available that
provide excellent examples of visualization exercises for athletics (e.g.
Baum & Trubo, 1999; other cites).
Laws of practice in motor learning
Laws of practice have to do with ways that teachers and learners can structure
training sessions to enhance learning. An important principle in several laws of practice
has to do with what Bjork (1998) called “desirable difficulties.” These involve
challenges that are placed in the learner’s path that may depress performance gains
3/10/2016
Titze & Verdolini, © 2002
10-34
during training sessions, but enhance learning as seen by long-term retention. Such
challenges may be induced by at least three manipulations: (1) infrequent, terminal
augmented feedback; (2) variable practice; and (3) random distribution of to-be-practiced
items. An additional variable, (4) part versus whole practice, may be relevant for voice
training in particular, and (5) consistent mapping appears relevant for the automization of
new skills. We discuss these training manipulations, and evidence surrounding their
value, next.
Augmented feedback. As already noted, augmented feedback refers to feedback
about performance, in addition to feedback which occurs spontaneously from sensory
receptors. Examples of augmented feedback include verbal information provided by a
trainer, or instrumented feedback from an oscilloscope, videoendoscopy, or other devices.
There is no question that augmented feedback is useful to learning (see for
example, Adams, 1987; Swinnen, 1996). Augmented feedback can be manipulated in a
variety of ways. The frequency of feedback can be manipulated (it can be provided after
each trial or after a set of trials), and its timing can be manipulated (it can be provided online during performance, immediately after it, or following a delay). Data indicate that
although augmented feedback assists learning, even a lot, somewhat paradoxically, “less
is more.” A variety of researchers studying different motor tasks have shown that
infrequent augmented feedback depresses performance gains during practice, but
enhances performance during long-term no-feedback retention tests (e.g. Bilodeau,
Bilodeau, and Schumsky, 1959; Lee, White, & Carnahan, 1990; VanderLinden,
Cauraugh, & Greene, 1993; Winstein & Schmidt, 1990).
3/10/2016
Titze & Verdolini, © 2002
10-35
The explanation appears to lie with cognitive effort. Learners who are provided
with a lot of information from their trainers or machines during practice sessions
experience performance enhancements, attributable to the support that the information
provides. However, learners are not forced to generate their own solutions. When they
leave the training arena, they do so without the solutions. In contrast, learners who are
forced to exert cognitive effort during training have poorer immediate performance, but
they “own” the solutions that they have generated. They retain such ownership, continue
to develop it, and demonstrate it at a later time when augmented feedback is not
available.
Another important finding regarding augmented feedback has to do with the
timing of feedback. Concurrent, “on-line” augmented feedback is used in many clinical
situations, in which the learner is provided information about his or her performance from
an oscilloscope or videoendoscope during actual phonation trials. As we have discussed
in previous sections of this chapter, such information may promote an external, “effectoriented” locus of attention, thereby having the potential for encouraging learning.
However, interesting results are reported in the literature regarding this practice.
One study was described in an unpublished dissertation (Armstrong, 1970; see
Schmidt & Lee, 1999, pp. 316-317). In that study, subjects were asked to learn a 4-sec
upper limb flexion and extension task, matching time and space characteristics of a
template. Different types of augmented feedback were provided to three different groups.
One group received guidance in spatial trajectories of their movements, and thus had to
learn temporal characteristics only (guidance group). Another group received no
mechanical guidance, but were given a visual trace of their movements on a computer
3/10/2016
Titze & Verdolini, © 2002
10-36
monitor, which could be compared to a template of the correct spatial and temporal
characteristics of the movement (concurrent augmented feedback group). The third
subject group was provided similar augmented feedback as for the concurrent feedback
group, only it was displayed after each movement had been completed (terminal feedback
group).
Results for a no-augmented-feedback retention test conducted after 3 days of
practice are shown in Figure 14. The left side of the figure shows that subjects in the
guidance condition had very little error during practice sessions. For the concurrent
feedback group, spatial error reduced rapidly during the first day of practice, although it
never reached the performance levels achieved by the guidance group on the second and
third days of practice. The terminal feedback group had the poorest levels of
performance during practice. In this group, errors were reduced during the whole
practice period but were still double at the end of practice compared to the concurrent
feedback group.
The situation was dramatically reversed at the retention test. The right side of
Figure 14 shows that subjects in the terminal feedback group maintained the level of
performance that they had achieved in practice. However, when guidance and concurrent
feedback were withdrawn in the two groups who had previously received it, performance
regressed to levels that were similar to those for the terminal feedback group during early
practice trials. These findings are similar to those reported by Schmidt and Wulf (1997)
in a later study on (what?). The results have also been replicated for a voice quality study
involving the learning of nasalance (Steinhauer & Grayhack, 2000). The implication is
3/10/2016
Titze & Verdolini, © 2002
10-37
that concurrent augmented feedback can have strong effects in reducing error during
training, but it actually degrades learning seen at long-term follow-up.
Again, cognitive effort in the learner is implicated. Information that supports
performance during training, but limits learners’ cognitive effort, appears to reduce
learning.
Interestingly, there appear to be other methods to enhance learners’ efforts around
interpretation of augmented feedback, and thus enhance their learning. One example is to
require learners to rate the success of their performance before giving them augmented
feedback (e.g., Swinnen, Schmidt, Nicholson, & Shapiro, 1990).
Variable practice. Variable practice is practice that requires a consistent response
to numerous stimuli during training. An example in voice training is consistent use of
easy “belt” quality on different vowels, in different parts of the pitch range, in different
songs. Non-variable practice is practice that requires a consistent response to a single
stimulus. An extreme example would consistent use of easy “belt” quality on a single
vowel and note. Considerable evidence suggests that variable practice degrades
performance during training sessions, but enhances generalization to new stimuli on
retention and transfer tests later on.
An example was reported for an experiment on bean-bag throwing. In that study,
subjects tossed a bean-bag into a target zone either from a single location, or from
variable locations. Subjects in the single-location condition performed with the greatest
accuracy during initial training. However, when all subjects later were required to toss
the bean-bag into the target zone from a novel location, subjects with previous variable
practice performed the best (Carson & Wiegland, 1979).
3/10/2016
Titze & Verdolini, © 2002
10-38
Another example was reported by McCracken and Stelmach (1977). These
authors studied learning for an arm displacement, in a criterion time of 200 ms. One
group of subjects practiced the task over a constant distance (non-variable practice
group). Another group of subjects practiced over a variable distances (variable practice
group). The results showed superior performance (in mean timing error) for the nonvariable practice group during practice trials. However, a transfer test assessing
generalization to a new movement distance for both groups showed superior learning for
the variable practice condition.
In these experiments, again we observe the principle of “desirable difficulties”
(Bjork, 1998). Learners who are presented with the challenge of variable practice during
training tend to have poorer performance gains during training, but greater generalization
to novel targets later, compared to learners who receive “easier,” non-variable practice.
Probably the best explanation for the variable practice effect comes from schema
theory (e.g. Schmidt, 1975, 1976). In review, schema theory posits that the process of
motor learning involves the generation of three-dimensional mental “regression plots,”
which relate initial conditions, response specifications, and movement outcomes (recall
schema; Figure 2). It stands to reason that variable practice provides the learner with
more “initial conditions” in training (y-axis, Figure 2). Thus the learner generates more
data points in the regression plot, than for non-variable practice. The result should be an
increased ability to generate valid response specifications on successive trials (z-axis,
Fgure 2), under novel conditions. Consequently, generalization of skilled action should
be enhanced when novel conditions are encountered (x-axis, Figure 2).
3/10/2016
Titze & Verdolini, © 2002
10-39
A caution is that variable practice effects are most robustly seen when practice
items are randomly interspersed. That is, if you are working on producing “opera
quality” or “easy voice onset” in training, randomly practicing the target on different
words and phrases (random practice) should produce greater generalization to new words
than practicing one word over and over and then practicing the next word over and over
(blocked practice). Some evidence suggests that variable practice which is blocked loses
much or all of the variable practice effect, for speech and other tasks (cites).
Distribution of practice. Another variable that appears to influence cognitive
effort, and motor learning, has to do with the distribution of practice trials over time. In
voice training, the issue can be framed as whether to work on alignment in a series of
trials, then breath, then glottal-supraglottal impedance matching (for example), or
whether to mix the trials up in some random order. Assuming for a moment that this type
of “component part practice” is reasonable in voice, evidence suggests that the latter
approach may reduce performance gains during practice but enhance learning seen at
long-term follow-up.
Shea and Morgan (1979) reported some of the critical evidence. These authors
studied two groups of subjects who had the task of making specific patterns of sequenced
arm movements as quickly as possible following a visual cue. Subjects performed 18
repetitions each of three different patterns of movements, each one being paired with
specific colored lights cues. Each group practiced the three tasks using either blocked or
random practice. The dependent variable was time for completion of each movement.
The results were striking. The group receiving blocked practice performed better (i.e.
faster) almost immediately. This advantage was preserved over the whole practice
3/10/2016
Titze & Verdolini, © 2002
10-40
period. However, when long-term retention and transfer tests were examined the same
day and after 10 days, subjects who had received random practice performed
considerably better than the blocked group.
The combination of superior performance during practice for the blocked
condition, together with superior performance at later retention and transfer test for the
random condition, reflects what is called the contextual interference effect. This effect
has been seen in many experiments assessing both laboratory tasks and everyday living
activities (reviews are provided by Magill & Hall, 1990: Marley, Ezekiel, Lehto, Wishart,
& Lee, 2000; Schmidt & Lee, 1999).
An important question regards the reason for the effect. Again, we return to a role
of cognitive effort in motor learning. Theoretically, the random practice schedule
introduces effort. An example can be seen in speaking a dialogue in a theatre rehearsal.
You bring up the memorized words into consciousness, and prepare to speak, vocally.
You produce the dialogue. Let’s say that the director wants to listen to the dialogue
again. If you produce it immediately, many of the operations needed to perform the task
are still activated. You can thus by-pass calling them up again, and proceed to run off a
repetition of the dialogue. In this scenario, which corresponds to “blocked practice,”
some of the task’s cognitive demands are limited or avoided altogether. Stated
differently, “the repetition of the same action does not necessarily require the repetition
of the same processes which resulted in that action” (Lee & Magill, 1983, cited in
Verdolini & Lee, in press). In another scenario, perhaps the director’s lunch arrives at
just that moment, and the director stops to pay the delivery person and chat for a moment.
You might busy yourself with other activities, such as checking your costume, talking
3/10/2016
Titze & Verdolini, © 2002
10-41
with other cast members about the party last night, etc. After several minutes, your
director is ready to go again. At that point, you repeat the dialogue (random practice).
You will probably have to re-activate all the prior operations. Essentially, you have to
“fire them all up” again. A view is that repeated activation of cognitive traces, from
baseline, increases their strength. Similar principles should apply to neuronal traces as
well (for physiology of synaptic changes with nerve firing, see Aidley, 1998).
Part-whole practice. There is evidence that practice of component parts of a
global behavior is useful when the component parts will be run off serially—one after the
other—in subsequent performance. Examples include (CITES).
Interestingly, part-practice may have questionable value when component
movements are executed in parallel (at the same time) in the target behavior. Practice of
the behavior as a whole seems to be more effective. An example is (CITE).
Fortunately or unfortunately, voice production is an example of a largely parallel
production system. Clearly, we first establish and alignment and inspire, and then
establish some form of tone and adduction in the vocal folds prior to phonation.
However, from that point on until the end of the phrase, alignment, breathing, vocal fold
positioning, and vocal tract manipulations all occur in parallel. Data are not yet available
regarding the effectiveness of “part-practice” in voice training. In the meantime, our
emphasis on training coordinative states across phonation substructures, rather than
performance on single structures, leads us away from part practice and towards practice
of the vocal whole, in voice training.
Consistent mapping. Consistent mapping is defined as consistently responding in
a given way to a given stimulus, or perhaps a general class of stimuli. Variable mapping
3/10/2016
Titze & Verdolini, © 2002
10-42
is defined as responding in different ways to the same stimulus over trials. Variable
mapping should not be confused with what has been called variable practice. The
difference has to do with the number of stimuli and the number of responses involved.
Figure 15 provides a schematic. Variable mapping involves numerous responses to
single stimuli (e.g. speak with pressed voice, resonant voice, breathy voice in response to
the stimulus: “Speak!”) Variable practice, as previously reviewed, involves single
responses to multiple stimuli (e.g. “Speak with resonant voice in response to any of the
stimuli:” “Speak!” “Sing!” “Shout!”). Studies in perceptual learning have indicated that
consistent mapping enhances the ability to perform new skills automatically—i.e. without
attentional resources, in contrast to variable mapping, which tends to inhibit the
development of automaticity.
A rather complicated experiment was the following. Schneider and Fisk (year,
correct citation) showed subjects 20 frames each containing 4 letters, in rapid succession.
After the frames had been flashed, subjects were to respond (yes/no) whether they had
seen a given letter at any point during the set. Some subjects received a “consistent
mapping” condition, in which a given letter (e.g. “M”) was always a target, requiring a
“yes” response. All other letters were consistently foils, requiring a “no” response, in
that condition. Other subjects were given a variable mapping condition, in which a given
letter (e.g. “M”) sometimes appeared as a target (“yes”) and sometimes appeared as a foil
(“no”). Similarly, other letters sometimes appeared as targets and foils. Two other
intermediate groups had degrees of consistent mapping between 100% (as for the
consistent mapping group) and 0% (as for the variable mapping group) requirements.
3/10/2016
Titze & Verdolini, © 2002
10-43
The results showed that both groups improved equivalently with practice on the
perceptual task, in terms of both speed and accuracy of responses. However, differences
in group performance were seen when a distractor task was introduced in addition to the
primary task. The distractor task required subjects to (what), as they continued to
perform on the perceptual task. In this case, subjects who had received consistent
mapping during practice continued to achieve their prior level of performance. Subjects
who had received variable mapping during practice had decrements in performance
levels. The performance decrement was worse as the amount of variable mapping
increased, in prior practice.
Assuming that results are similar for motor learning as for this simple perceptual
learning task, the findings indicate that people can learn new tasks when they respond
inconsistently to a given stimulus (e.g. use resonant voice in response to the stimulus
“Speak!”). The problem may come when they are asked to perform other tasks at the
same time (e.g. “Think about what you’re saying and actually communicate!”). Under
such conditions, people who have not experienced consistent mapping earlier (“Use
resonant voice all the time when you speak!”) may experience a decline in performance
under the distraction conditions.
Summary and discussion of the laws of practice, and implications for voice training and
therapy
The laws of practice are relatively straightforward to apply to routine
voice training. Trainers should provide augmented feedback about
learners’ performance, either verbally (“good!”), gesturally (nod), or by
way of instrumented devices. However, augmented feedback should be
limited (not every trial). It has been suggested that more augmented
feedback is useful early in learning, and less augmented feedback is more
effective later on (cites). One method of establishing an effective amount
of feedback is called the “bandwidth method.” In this method, augmented
feedback is provided only when learners are relatively far from target
3/10/2016
Titze & Verdolini, © 2002
(cite). Early on, such feedback might be needed relatively frequently.
Later, augmented feedback will spontaneously decrease as learners
improve. The effectiveness of augmented feedback may be increased if
learners are asked to rate their own performance on selected trials.
Devices which provide augmented feedback constitute biofeedback. In
voice, numerous types of displays have been used, including fundamental
frequency traces, intensity traces, spectrograms, and spectral displays (e.g.
Nair, 1999). If biofeedback devices are used in voice training, certain
cautions apply. One caution, seen in a previous section of this chapter, is
that the devices should replicate intrinsic feedback as much as possible.
Another caution is that biofeedback information is best provided after
trials have concluded (terminal feedback), as opposed to concomitantly
with trials (concurrent feedback). Also, systematic, gradual withdrawal
trials should be used to limit decrements in performance from device
withdrawal.
Physical guidance is an intuitively appealing intervention modality. An
example in voice training is the trainer placing his or her hands on the
back or the neck to help the learner “feel” correct alignments. Guidance
may be useful in some portions of training, especially when learners are
not “getting it,” to help learners establish a template of the intended action.
However, a caution is that guidance may enhance performance but limit
learning. This is probably because it draws attention to biomechanical
aspects of learning, which have been seen to be a generally ineffective
target of training. Moreover, learner “effort” is minimized by guidance
techniques. Care should be taken to use guidance parsimoniously, and to
withdraw it as early as possible in training, if it is used.
Practice on numerous stimuli, using numerous phonemes, words, phrases,
conversations, songs, and vocal performance in a variety of physical and
emotional environments should enhance generalization of target behaviors
to novel situations in the future. An important principle is that variable
practice of this type, and practice on different voice goals, may be most
effective when the stimuli are randomly distributed intermixed with other
trials rather than practiced in blocked fashion.
There is caution about extensive practicing of component parts of voice
production in isolation. Evidence suggests that training of the whole voice
production system, all at once, may facilitate learning for voice, which
largely constitutes a task in which component parts are executed in
parallel. Our emphasis on training combined equilibrium states
(Postulates #3-4, Chapter 7) are consistent with this recommendation.
Finally, consistent use of a new voice pattern may enhance its stabilization
as an automatic, “default” pattern. This may be one of the greatest
3/10/2016
10-44
Titze & Verdolini, © 2002
10-45
challenges in voice training, for speech, in particular. Week after week,
clinicians and trainers may note in their records, “Poor carry-over from
last week. Good progress within session. Question of compliance
problem.” That is, it may appear that individuals who improve within
training sessions, but fail to carry the improvements over to the following
week’s session, have not done their exercises and are not motivated. This
may be the case. However, another possibility is that learners’
performance during sessions is good because they are able to allot near
full attention to their voice during the sessions. As soon as they leave the
training room, they revert to their old patterns as soon as they say goodbye to the receptionist because they have to think about what they’re
saying. Such thinking constitutes a “secondary task” which reduces
competence of the not-yet automatized target behavior. Essentially,
“variable mapping” has been invoked which interferes with automization.
The same problem does not usually exist to the same degree in singing
training. This is because usually, when a person sings, he or she can
dedicate near full attention to the task. The individual ends up being able
to perform relatively consistently each time he or she sings (at least more
consistently than occurs in speech). Eventually, this type of “consistent
mapping” allows for new skills to be automatized. At some point, learners
who are acquiring a new voice pattern in speech, which encounters more
attentional challenges, may need to make a commitment to use the new
voicing pattern in speech consistently, each time, every time. Whereas
errors may be very useful for initial phases of skill acquisition, in which
plasticity and change are the main concern, consistency, and minimum
error may be most useful for later stages, in which automaticity is the
concern. As noted by the famous and ubiquitous psychologist William
James (1890): (consistency quote).
Concluding Comments
We have reviewed numerous principles of motor learning. We list them in bullet
form in Table 3, giving examples of how they may be applied in voice training and
therapy. Some of the principles are consistent with much clinical and pedagogical
practice. Some are contradictory to much practice.
In the next chapter, we consider Sports Psychology. This chapter reviews
practical techniques and suggestions that illustrious trainers use to enhance athletic
performance. Many of the suggestions should be highly relevant for the vocal performer
3/10/2016
Titze & Verdolini, © 2002
10-46
as well. In reading the chapter, it should be possible to trace many of the trainers’
suggestions to the basic science reviewed in this present chapter. One of the particularly
intruiging concepts has to do with cognition versus meta-cognition: the art of becoming
“one with the thing.”x
3/10/2016
Titze & Verdolini, © 2002
10-47
List of Tables and Figures
Table 1:
Characteristics of explicit and implicit memory
Table 2:
Traditional technical instructions compared to “discovery” instructions
(based on Gallwey, 1997)
Table 3:
Chapter Summary: Bullet points (see below)
Figure 1:
Fairbanks’ cybernetic model of motor control and learning in speech
Figure 2:
Recall schema
Figure 3:
Recognition schema
Figure 4:
Wulf & Weigelt (1997), results from ski simulator task, Experiment 1
Figure 5:
Hodges & Lee (1990), results from bimanual coordination task
Figure 6:
Verdolini-Marston and Balota (1994), results for manual tracking task,
Experiment 3
Figure 7:
Verdolini-Marston et al. (in preparation), results for voice airflow task
Figure 8:
Wulf, Hoess, & Prinz (1998), results for stabilometer learning, internal
versus external locus of attention
Figure 9:
Sulf, McNevin, Fuchs, Ritter, & Torle (2000), results for tennis,
antecedent versus effect locus of attention
Figure 10:
McNevin, Shea, & Wulf (2001), results for stabilometer task,
effectiveness of increasing locus of external attention
Figure 11:
Shea & Wulf (1999), results for stabilometer task, internal versus external
locus for augmented feedback
3/10/2016
Titze & Verdolini, © 2002
Figure 12:
10-48
U-function, learning versus distance of locus of external attention relative
to body
Figure 13:
Yerkes-Dodson curve
Figure 14:
Results from Armstrong dissertation (possibly change to Steinhauer &
Grayhack, 2000)
Figure 15:
Schematic, variable practice versus variable mapping (and non-variable
practice versus consistent mapping)
3/10/2016
Titze & Verdolini, © 2002
10-49
PBLs for Chapter 10:
(1) Attempt to write your own brief summary for the subsections of the chapter,
“historical context,” “attentional versus automatic processing,” “intention,”
“laws of practice,” and “chapter summary.” In your sections, recapitulate
the main points of the section, and draw some potential applications for the
type of voice training situation that you are involved with.
(2) Be prepared to conduct a 5-min “voice lesson” or “voice therapy session”
with a person selected at random from the class, incorporating what you
consider three important principles of moto r learning derived from the
chapter.
(3) Make a “bullet list” summarizing important principles of motor learning
that were reviewed in the chapter. Highlight the principles that you would
like to incorporate in your training regimens, either as a learner or trainer.
3/10/2016
Titze & Verdolini, © 2002
10-50
Table 3. Bulleted Summary of Main Practical Points in Chapter on Motor Learning

Learning involves shifts from existing to new physical “attractor states.” Some
are easy to acquire, some are difficult and require attention used in a special way.

Attention to biomechanics in training (internal locus of attention) harms learning,
even if “metaphors” are used.

Allow non-conscious automatic processes to solve mechanical processes in
learning.

Attention to movements’ effects (external locus of attention) is helpful to learning

The effectiveness of an external locus of attention is not carried simply by simply
not paying attention to one’s body. Rather, attention specifically to movement
effects seems to be critical (for example, an external locus on conditions
antecedent to movement is not so effective).

An intermediate distance between one’s body and the locus of external, effectoriented attention is optimal.

An external locus of attention in augmented feedback is more effective than an
internal effect.

(Conscious) intention is critical to learning difficult new patterns. However, only
certain types of conscious intention are effective: directed towards movements’
effects; not directed towards biomechanics; intermediate arousal (motivational)
level.

Visualization exercises may be a useful intention exercise. Presumably all the
same principles apply as for real-life intention (above).

Augmented feedback is extremely useful in learning. However, less is more.
3/10/2016
Titze & Verdolini, © 2002

10-51
The correct amount of augmented feedback can be established using a
“bandwidth” method.

Asking learners to rate their own performance before providing augmented
feedback may enhance the feedback’s effectiveness.

Physical guidance in training may enhance immediate performance, but depress
learning, thus should be used sparingly.

Practice on numerous stimuli, in numerous environments, depresses immediate
performance but enhances later generalization to new situations.

Practice of the whole behavior may promote better learning for parallel tasks such
as voice, more than component practice.

Consistent use of a new behavior promotes its development as a default habit,
requiring little attentional resources.
Technical instructions, if required:

Should incorporate movement effect.

Should not induce people to solve mechanical problems consciousness, but rather
simply “notice” mechanics

Should promote an intermediate arousal (motivational) level.
Biofeedback:

Should be redundant with intrinsic feedback.

Should be provided after the performance, not during it.

Systematic withdrawal trials should be incorporated.
3/10/2016
Titze & Verdolini, © 2002
3/10/2016
10-52
Titze & Verdolini, © 2002
3/10/2016
10-53
Titze & Verdolini, © 2002
10-54
INGO’S STUFF: PUT BACK IN?
PUT THIS IN NEW SECTION ON BIOFEEDBACK. PERHAPS UNDER
“APPLICATIONS.”
There has been a growing interest in biofeedback and body awareness in recent years.
New methods of observation have allowed us to examine, if not directly control, various
physiological processes of which we were previously quite unaware. Blood pressure,
heart rate, isolated muscle contraction, and neural impulses can now be monitored and,
more importantly, altered by an individual if presented to him as a sensory imput. This
leads to new questions about the importance of sensory feedback in vocal control and
about the potential usefulness of additional feedback that might be obtainable for the
vocalist through instrumentation.
Control theorists have long distinguished programmed tasks as being either openloop or closed-loop. In both cases there is a desired performance and an actual
performance, but the manner in which the actual performance is monitored and altered
during execution of the task differs. In a closed-loop system, the program is dynamic.
Corrections can be made as unforeseen events attempt to alter the previously determined
course. A simple ballistic released from a cannon, and a guided missile heading toward
its target, provide examples of open-loop systems. Unforeseen disturbances, such as
changing winds or small particles in the bath of the projectile, may cause significant
deviation from the desired course, such that only a precise error correction will restore
the previous conditions.
A further consideration in control theory is whether the
feedback is instantaneous or delayed. A small amount of delay may be tolerable, but if
the delay is too long, the necessary correction may put demands on the system that are
3/10/2016
Titze & Verdolini, © 2002
10-55
physically unattainable. It is also possible that the performance has progr3essed to the
point that the opposite correction is needed by the time the feedback arrives. This may
not be the case when an automobile “fishtails” as the driver applies what he considers to
be the appropriate correction for a skid, but the combined reaction times of his body and
the car make it quite appropriate when the car has progressed through the skid.
If the feedback is delayed to the extent that it comes after completion of the task,
we are dealing with a system that is open-loop for a single execution, but may be closedloop for repeated executions. Performance will generally improve if the system is
adaptable, i.e., if it can modify its program on the basis of performance history. Much of
the feedback given to vocalists is of this type. Positive reinforcement from the audience,
the voice teacher, juries, colleagues, or personal evaluation of tape recordings fall into
this category. Most vocalists have recognized the importance of continual monitoring of
their performance on the basis of human reaction of this kind. After all, their livelihood
depends on it.
For some purposes, however, “listening” to one’s own body may be as valuable as
relying on reactions of others. This may be particularly the case when preservation of th
vocal mechanism is at stake. Internal feedback is usually immediate and directly related
to what goes on in the body. Some forms of visual, tactile, auditory, and kinesthetic
feedback are available to every normal person for monitoring of voice production. The
use of a mirror, for example, in examining facial distortions, checking the position of the
larynx or uvula, or evaluating over-all posture are well established. The benefits derived
from these simple visual feedback techniques are hardly detectable. More sophisticated
visual techniques include laryngeal fiberscopy (Sawashima and Hirose, 1968),
3/10/2016
Titze & Verdolini, © 2002
10-56
electromyography (Hirano, et al, 1970), electroglottography (Fourcin and Abberton,
1971), and Appleman’s vowel meter. Some of these are being introduced into voice
clinics and speech research laboratories for diagnostic and remediation purposes. Their
merits for practice and self-monitoring with limited supervision deserve careful
evaluation. Other visual feedback systems that have gained some acceptance for speech
training of the deaf include real-time spectrographic displays and speech-feature
detectors. The former show the acoustic power in all of the overtones, while the latter
extract specific features, such as fundamental frequency, formant frequencies, or intensity
as a function of time (Risberg, 1968).
Tactile and kinesthetic feedback is available to every normal individual through
various sensory receptors located throughout the entire body. Tactile receptors monitor
contact with the skin, while kinesthetic receptors monitor the relative position of various
moving parts of the body. Thus tongue position, tongue contact with the lips, teeth, or
palate, and jaw position are monitored continuously during speech or singing. If the
afferent (returning) pathways to the nervous system in the facial region are desensitized
(as in dental repair), the effect on articulatory movement can be quite pronounced. On
the contrary, some tactile sensations can be enhanced by rather simple means. The
sensation of airflow in the mouth and throat, for example, seems to be amplified with the
use of something as simple as a mint candy or chewing gum. It is possible that deaf
persons, deprived of much of their auditory feedback, may compensate by making greater
use of oro-tactile feedback, in particular with regard to airflow. Some vocal pedagogues
(Lilli Lehmann, 1972) have claimed the sensation of “whirling currents” of air in their
3/10/2016
Titze & Verdolini, © 2002
10-57
own sound productions. It is more likely, however, that the latter sensations were
misinterpreted and should be credited to vibrations in the tissue.
Tissue vibration is a major component of acoustic feedback. Pathways to the
inner ear, where auditory signals are detected and converted to neural signals, may be
entirely within the body or may include acoustic links through the air. The composit
signal, derived by superposition of signals that arrive from internal and external
pathways, does not always bear a simple relationship to the sound that is detected by
listeners on the outside. An interesting paradox exists here. From a production point of
view, the vocal instrument is inside the performer, but from a perception point of view,
the performer is inside the instrument. It is as though he has placed himself (or at least a
microphone) within the body of a grand piano to determine the acoustic radiation
characteristics on the outside. As difficult as that may be, vibrational and acoustic cues
are used effectively by virtually every singer. Some find it useful to experiment with
these cues by cupping their hands behind the ears, practicing with headphones,
alternating between large and small practice rooms (or live and dead ones), or otherwise
modifying the mixture of sound received by the inner ear.
In conclusion, it is obvious that numerous feedback mechanisms are available to
every singer. It is not clear that every singer uses them optimally, that the balance of the
sensations derived from this feedback is the same across singers, or that singers as a
whole will always benefit from additional feedback derived through artificial means. It is
conceivable, however, that dynamic information about vocal fold movement, muscle
activity, airflow rate, and various other physiologic processes may eventually be
highlighted so clearly through instrumentation and multi-sensory means that there is little
3/10/2016
Titze & Verdolini, © 2002
10-58
doubt as to what the vocal mechanism is doing. Whether or not this will help teachers
produce better singers or good singers faster, is still a lingering question. Science has
provided the tools. Now the responsibility for evaluation of these tools rests on both the
vocal teacher and the scientist.
Perhaps the most compelling ones come from individuals with anterograde amnesia.
People with this condition cannot remember post-traumatic events. Stated differently,
their ability to remember things consciously is impaired. Yet, they demonstrate motor
learning normally or near normally (CITES).
3/10/2016
Download