You are on page 1of 17

This article was downloaded by: [University of Utah]

On: 22 November 2014, At: 23:40


Publisher: Routledge
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK

Psychotherapy Research
Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/tpsr20

Developments in task analysis: New methods to study


change
a b b
Antonio Pascual-Leone , Leslie S. Greenberg & Juan Pascual-Leone
a
Department of Psychology , University of Windsor , Winsor, Ontario
b
Department of Psychology , York University , Toronto, Ontario, Canada
Published online: 22 Sep 2009.

To cite this article: Antonio Pascual-Leone , Leslie S. Greenberg & Juan Pascual-Leone (2009) Developments in task analysis:
New methods to study change, Psychotherapy Research, 19:4-5, 527-542, DOI: 10.1080/10503300902897797

To link to this article: http://dx.doi.org/10.1080/10503300902897797

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the “Content”) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the
Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and
are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and
should be independently verified with primary sources of information. Taylor and Francis shall not be liable for
any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever
or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of
the Content.

This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions
Psychotherapy Research, JulySeptember 2009; 19(45): 527542

PSYCHOTHERAPY RESEARCH METHODS

Developments in task analysis: New methods to study change

ANTONIO PASCUAL-LEONE1, LESLIE S. GREENBERG2, & JUAN PASCUAL-LEONE2


1
Department of Psychology, University of Windsor, Winsor, Ontario & 2Department of Psychology, York University, Toronto,
Ontario, Canada
(Received 16 July 2008; revised 12 March 2009; accepted 16 March 2009)
Downloaded by [University of Utah] at 23:40 22 November 2014

Abstract
This article provides a conceptual overview of task analysis, which is an inherently multimethod approach. The authors
present the method as a step-by-step illustrative template for researchers who seek to develop qualitatively rich models of
change and quantitative measures that correspond to these change models. The current article provides an epistemological
framework to develop both descriptive and causal models of change. It also offers a comparison with other methods of
inquiry that are exclusively qualitative in nature and do not explicitly highlight the use of theory in model development. In
addition, the authors describe recent developments in task analysis for dynamic modeling. In tandem with this, they
articulate advances in the relationship between task analytic construct development and measurement development.

Keywords: task analysis; process; qualitative; models; measurement; change patterns

Future directions for studying the process of change provide an illustrative template for researchers
are often vaguely discussed among clinical research- who wish to blend qualitative and quantitative
ers as requiring a blending of qualitative and quanti- research methods within a program of research using
tative methods. However, the method by which this a lock-step fashion. As indicated, this blending of
might be achieved often remains unelaborated methods is something that is often recommended in
(Greenberg & Pinsof, 1986; Rennie & Toukmanian, the literature but for which there is little available
1992). Such an investigative approach involves a instruction. Task analysis is an inherently multi-
qualitative study of underlying structures, which then method approach, and although it has been success-
is somehow transposed into quantitative analyses fully used within single studies (e.g., Greenberg,
(J. Pascual-Leone, 1978). Thus, in a very real sense, 1984, 2007; Joyce, Duncan, & Piper, 1995; Safran
studying the process of change as it unfolds in-session & Muran, 2003), this may have led to some confusion
requires explanations that bridge descriptions of because it ultimately describes and requires a pro-
what happened with some deeper understanding of gram of research. The programmatic nature of this
why and how. Although there are several approaches approach may be one reason why task analysis can, at
to task analysis (see introduction to Greenberg, first, appear daunting. Even so, it is consistent with
2007), the formal method we outline here prescribes Lakatos’s (1970) notion that science progresses
how to blend methodological approaches and has through the development of research programs rather
been developed to study complex processes of than by the falsification of single studies.
change. In this article, we outline recent develop- Second, we articulate task analysis as a method to
ments in task analysis, addressing four conceptual develop both descriptive models as well as to infer
and technical issues not discussed in Greenberg’s causal models of change. Studying the process of
(2007) methodological guide. change in this way requires an epistemological
First, by presenting our extended method in the framework, which we describe for the first time
context of a step-by-step overview, we hope to with respect to task analysis as it is conducted in

Correspondence concerning this article should be addressed to Antonio Pascual-Leone, Department of Psychology, University of Windsor,
401 Sunset Avenue, Winsor, Ontario N9B 3P4, Canada

ISSN 1050-3307 print/ISSN 1468-4381 online # 2009 Society for Psychotherapy Research
DOI: 10.1080/10503300902897797
528 A. Pascual-Leone et al.

psychotherapy research. Moreover, we also offer a specific therapeutic change process. Moreover, the
few points of comparison with other methods of investigator must have sufficient expertise to be able
inquiry that are exclusively qualitative in nature and to recognize good client performance (in terms
do not explicitly highlight the use of theory in model of therapy goals) when it occurs. Even so, how
development. Although comparisons are not the the performance is produced or which critical
focus of our article, they help orient the reader components were coordinated to produce good
with respect to the purpose and scope of task task performance are, at least initially, unknown
analysis. (often to both client and investigator).
Third, we describe the recent development of a
new kind of study in task analysis, one that relates
complex sequences, or patterns of process, to out- Underlying Assumptions of Task Analysis
come. This innovation is more sophisticated than When we discuss how therapeutic change occurs, a
past task analytic validation studies, which have number of epistemological levels of discourse are
simply related the presence of process components implicated. The most obvious is that we may
to outcome. Seeking empirical support for an ‘‘observe’’ either overt or covert signs of change,
explicitly sequential pattern of process is a new which reflect a client’s subjective experience.
approach to studying change and is a development Although it is essential for researchers to accurately
that highlights the potential of task analysis as a describe these subjective experiences as components
method for the causal modeling of change.
Downloaded by [University of Utah] at 23:40 22 November 2014

of change using descriptive models, we posit that it is


Fourth, in tandem with advancements in model ultimately the dynamic and sequential coordination
development are advancements in the task analytic of descriptive components that brings about the
approach to measurement. Thus, we articulate the processes of change itself.
relationship between construct development and Furthermore, some paths in a descriptive model of
measurement development in a task analytic frame- subjective states/meanings/actions are optimal for
work and discuss the way in which categorical achieving some particular kind of change while
measures of subjective experience can be trans- others are not. This sets the initial parameters for a
formed into ordinal measures. The subsequent task. An account of the optimal flow between states
generation of ordinal (rather than simply categorical) (model components) toward task completion repre-
data allows researchers to further explore hypotheses sents a causal model. Such a model should recognize
about the patterns of process that help to explain those dynamic and sequential pattern(s) that are
therapeutic change. invariant across types of situations and individuals,
reflecting the nature of change at a higher level of
abstraction. Thus, causal models capture those
What is Task Analysis?
dynamic patterns of change (based on a descriptive
Task analysis in psychotherapy is an investigative model) that seem to hold true for clients in general.
strategy for illuminating steps in the successful Notice that recognizing characteristic patterns of
completion of a therapeutic task (e.g., a cognitive, change actually implies some organizing process, one
affective, or physical problem situation) and provid- that is hidden from direct observation and yet is
ing a detailed understanding of a participant’s apparent using a much broader scope of analysis
performance in completing a complex change task. (i.e., considering an array of situations, individuals).
Although many change tasks are not openly recog- Representations of invariants in goal-directed per-
nized as such, the goals to be obtained in a given formances are incipient causal models because they
moment are ultimately what define the task as it is speak to tacit characteristics of the client, as an
being conducted. Furthermore, the research partici- organism, productively engaging the internal/exter-
pants in a therapeutic task to be analyzed could be a nal world. These ‘‘client characteristics’’ (i.e., orga-
client, a therapist, or some relational dyad of people, nismic processes) are ultimately the true causal
depending on the identified task for study. Even so, mechanisms for the individual expression of internal
for the purposes of simplicity, we refer to client and experience and of outward behavior.
client tasks as the focus of study. Whatever the case Thus, we posit that personal change does not
may be, the main strategy of task analysis is the occur as a result of isolated states or units of
decomposition of tasks into performance compo- information but rather through an interplay between
nents, which are ordered in time (Greenberg, 2007). hidden (causal) properties and manifest (descriptive)
In essence, task analysis is a method of inquiry experiences. This means that to study the process of
designed to address the question, ‘‘How did the change, a method should describe those subjective
client do that?’’ The investigator must be able to client experiences relevant to some goal and, further-
identify and be familiar with the task at hand, a more, begin to articulate the principles by which
Task analysis 529

those client experiences unfold. A phenomenological uses evidence from the first two studies to support
study of experience and personal meaning, which development of an ordinal scale representing the
describes a range of client’s subjective ‘‘states,’’ can model of change. This scale can then be used as the
be thought of as a subjective analysis. One way of point of departure for a new generation of research
developing a qualitative study of this kind is to hypotheses about the process of change.
consider the goal-oriented purpose of selected
subjective states and to examine the emerging
‘‘pattern of states,’’ which seems to predict a process A Unique Approach
of change toward some goal: In our approach to task Model building in task analysis is a form of intensive
analysis, we call this is a dynamic process analysis. analysis distinguishable from other qualitative inqui-
Finally, although one cannot directly observe the ries in at least four important ways: (a) purpose, (b)
inherent qualities or ‘‘traits’’ of the participant, use of theory, (c) the nature of data collection, and
loosely defined, one comes to see how these char- (d) measurement development in the interest of
acteristics are implied through the emerging pattern predictive modeling. The first three are the conse-
of states. To model these hidden properties (e.g., quence of a different epistemological perspective.
traits) that are only indirectly observable is to make First, the purpose of task analysis is the dynamic
hypotheses about the participant as an organism. modeling of a change task; concretely, this means
This is organismic process analysis. Notice that sub- discerning the minimum number of distinct and
jective refers to the personal experience a client is
Downloaded by [University of Utah] at 23:40 22 November 2014

essential steps in a task’s resolution. Thus, the


implicitly or explicitly aware of, whereas organismic objective in this method is to identify only the most
refers to the inherent properties and attributes of the necessary and efficient ‘‘steps’’ toward the intended
client (e.g., individual differences in personality- goals (e.g., self-growth in emotional adaptation).
social characteristics, affective or cognitive processing Ultimately, task analysis aims to represent a goal-
styles). Thus, organismic process analysis examines the directed performance (which participants have or
hidden resistances and affordances that a client might produce) as efficiently as possible. Thus, the
brings to bear as a problem-solving organism. discovery phase of task analysis focuses on modeling
the process rather than the hermeneutic analyses of
meaning in a client’s experience per se.
Phases in Task Analysis
Second, task analysis begins with a declared set
In relating these questions to the process of change, of theoretical assumptions and then deliberately
all task analysis allows one to do is to (a) describe uses these to interpret and inform observations;
client experiences relevant to a task, (b) look for this is referred to as rational modeling. Thus, task
natural sequences in productive client process, (c) analysis dialectically makes use of both top-down
empirically predict these natural sequences, and (d) (rational) and bottom-up (empirical) modeling. This
eventually infer which organismic processes may use of theory as both an observational tool and as a
permit such natural sequences. This occurs over repository for observations bears some resemblance
several phases of inquiry. In task analysis, the first to Stiles’s (2007) method of theory-building case
study of a change process is the discovery phase, studies, although task analysis is not concerned
which uses qualitative methods that blend observa- with general case theory but rather with theories
tion and theory (i.e., synthesize by combining specifically about task resolution.
qualitative observations and their plausible theore- Third, task analysis also differs in the nature of its
tical meaning) to develop a specific model and an data collection, partly as a result of the two points
accompanying measure of clients’ therapeutic already indicated. The grounded theory form of
change. The second study, the validation phase, qualitative research, when applied to psychological
uses traditional methods of hypothesis testing to topics such as psychotherapy, primarily entails the
measure and test the newly created model on a grounding of understanding of a phenomenon in
separate sample. While validating the model, this reports of how it was subjectively experienced by the
phase of research also helps to validate a rich client. Accordingly, although the user of grounded
categorical measure of change. A typical project theory interprets these reports when conceptualizing
using task analysis is finished when both the dis- categories of experience, such interpretations are
covery and validation phases of research are com- based on what participants themselves are aware of
plete (Greenberg, 1984, 2007). Recent innovations and able and willing to address (Rennie, personal
by Pascual-Leone (2005, 2009) have indicated that a communication, January, 2009). In contrast, given its
task analytic model can be further extended to test different purpose, task analysis also relies on the
new hypotheses in a dynamic modeling phase. Thus, systematic observation of a client’s performance in a
the third study in a research program of this kind task, often done in the spirit of human ethology. Thus,
530 A. Pascual-Leone et al.

data used for task analysis may include both aspects of rationalist facts (patterns across observations), must
a change process: those that may be covert and are be treated as tentative and need to be independently
within the client’s awareness as well as the directly tested. Thus, the qualitative investigation is not
observed aspects outside the client’s awareness. consolidated until a validation study is done, usually
Fourth, models in task analysis are developed with in the form of a second, quantitative study. It stands
the understanding that they must be amenable to to reason that two phases of investigation, each
measurement and be empirically predictive. Thus, requiring a different data set, make for a lengthier
the creation of qualitative categories is contingent inquiry than a single qualitative inquiry; still, we
on their ultimately being codified into quantifiable believe that multimethod approaches such as task
terms. Moreover, measurement criteria are con- analysis can yield richer theoretical insights into
structed in parallel with the development of con- actual processes by which change occurs. Even so,
structs (categories), such that identifying criteria selecting between qualitative methods should be
actually impact the way in which constructs are done based on the research question at hand.
conceptualized. Thus, both model and measure When inquiring into the meaning of a phenomenon,
emerge from a dynamic process of continued refine- grounded theory may be the method of choice.
ment. This emphasis on model measurement and However, for questions related to how a particular
predictive validity is unique to task analysis. phenomenon unfolds in time and what causal
A distinct contribution of the task analytic discov- mechanisms might be implicated in that process,
ery method is that it deliberately makes use of existing
Downloaded by [University of Utah] at 23:40 22 November 2014

task analysis is the method of choice.


theoretical frameworks to tentatively explain qualita-
tive results of actual or possible goal-driven behavior.
It does this by instantiating theory in these data as a Discovery Phase
way of sharpening the investigator’s perception of This section describes the discovery phase of task
observed phenomena. This enhances researchers’ analysis, which involves construction of a rationally
ability to interpret qualitative findings. In contrast, based model and development of a related empirically
an approach that minimizes the influence of existing based model through systematic observation (Green-
theory is more or less bound to a single context berg, 1984, 2007). The researcher observes and
because it is contingent on concrete examples taken
discovers using top-down (rational) and bottom-up
from a given situation. As a result, qualitative
(empirical) approaches concurrently. The two-
approaches in which theory is either bracketed or
pronged dialectical process of observation and theo-
simply held in abeyance do not allow for theoretical
rizing emerges as part of a very dynamic process of
postulates (rational modeling) that are empirically
model construction. Steps of the discovery-oriented
inferred as functional invariants across types of
phase are represented schematically in Figure 1 and
situations (i.e., reliable interrelated patterns). Yet
described in the subsections that follow. Initial con-
such empirically grounded theoretical postulates,
jectures act as null hypotheses (see Figure 1, top
sometimes called rationalist facts (J. Pascual-Leone &
right), which are often based on the literature and on
Sparkman, 1980), could empirically ground the
synthesis of meaning across substantially different the investigators’ knowledge of the phenomenon. The
situations. One illustration of this type of rationalist discovery-oriented process subsequently involves two
fact, which is abstracted from a pattern across distinct efforts: the theory-based rational model (see
concrete situations, is a psychodynamic formulation Figure 1, top left) and a corresponding, observation-
(A. Pascual-Leone & Greenberg, 2006; of course, we based empirical model (Figure 1, middle), followed
hope it is clear that the yields from this type of by the integration of these two.
meaning abstraction are not necessarily psychody-
namic in nature). Task analysis recognizes and Step 1: Subjective Analysis of Phenomena
capitalizes on the fact that observations can rarely (if (Rational Modeling)
ever) be theory free (Kuhn, 1996). Thus, by encoura-
ging the use of existing theories of change when Inquiry begins with a subjective analysis of some
examining performances, the method allows investi- phenomena. This step makes use of existing theory
gators to observe sequential relationships both across as a way of guiding the observation of raw data at the
time and across different situations. Without using most molar level, whether in video-recordings, ses-
existing theory as part of the framework, the observa- sion transcripts, interviews, live observations, and so
tion of patterns of experience across different situa- on. In the aim of doing more than descriptive
tions is not possible. research, the phenomena to be examined should be
The inconvenience of task analysis is that models regarded as a dynamic process in time that captures
from the discovery phase, which are infused with some goal-directed behavior.
Task analysis 531
RATIONAL MODEL

Prior to model
Rational analysis building,
is based on HYPOTHESES
theoretical are generated based
postulates about on the literature,
the present task inquiry, and
environment and anecdotal
participant observation.
behaviours.

EMPIRICAL MODEL: Based on 3 good & 3 poor cases

Case 1 Case 2 ...Case 6

Observe Refine Compare Remove


theoretically
Downloaded by [University of Utah] at 23:40 22 November 2014

components & contrast non-essential


postulated components components
components

Add observed
components

SYNTHESIS:
Rational-Empirical Model

Next major phase of Findings:


research: • Model of
VALIDATION change.
PHASE, • Measure of
New, a hypothesis testing process.
larger follow-up
data set investigation of the
synthesized model.

Figure 1. The procedures of task analyses from discovery to verification.

How to choose a change task. The first issue in researcher’s conceptualization of the task must ad-
conducting this type of research is to identify some dress various specific questions: What signals the
implicit or explicit therapeutic task. The investigator initiation of a task? How might one know when a task
must identify some phenomenon that (a) seems has been successfully concluded? In doing this, the
therapeutically important, (b) relates to change, more loosely circumscribed a task, the more it will be
and (c) would benefit from being ‘‘mapped out.’’ global and appear in the form of some continuous
Moreover, (d) the task cannot be too global and process (as, e.g., practicing a skill). Finally, (f) tasks
must involve observable performance (e.g., using must also be explicitly formulated as occurring within
audio, video, a series of subjective reports across certain task environments or situations (e.g., in a
time). Notice that the observations of change used kind of therapy, relational context), whether this is
for model building may be direct or indirect and they broadly or narrowly defined.
do not necessarily have to pertain to a client’s Although interventions proposed by therapists are
subjective experience, although they often do. obvious examples of task environments and implied
Furthermore, defining a task, as such, has other tasks, it is not necessary that the task chosen for
demands for the researcher: (e) to concretely identify analysis be characterized or marked by an already
the start and end points of the task. Thus, a existing intervention. In fact, a task could consist of
532 A. Pascual-Leone et al.

spontaneously occurring (goal-directed) events out- emotional processing*serve as the beginning and
side of the client’s or therapist’s explicit awareness. end, start and finish, of this event. Following this
Because complex ‘‘internal’’ tasks have cues that are identification of a task’s boundaries, the basic
contextualized by a participant’s past experience, the objective is the creation of a model of change that
nature of a task is often, in a literal sense, within bridges the start and finish with specified intermedi-
the mind of the beholder. For example, resolving a ate steps. Because the goal of this method is to define
conflict is a task that is explicitly recognized by the client change processes, tasks are best mapped out
observing researcher and either implicitly or expli- from the perspective of the client. For this reason,
citly so by the experiencing client. Thus, the client’s the task analytic approach temporarily puts the
therapist, or even the chosen approach to treatment, therapist in the conceptual background so as to
does not have to recognize the client’s task in focus on optimal client process (Greenberg, 1991,
question. Examples of change tasks include the 2007).
resolution of interpersonal grievances, resolving Modeling states versus state changes. Although the
self-criticism, overcoming an identified emotional discovery phase of task analysis was developed
block, repairing an alliance rupture, or even the independently from grounded theory, in some sense,
cognitiveemotional processing that follows an iden- it is like a grounded theory analysis done against
tified cue (where a cue could be, e.g., momentary time, where the object of study represents some goal-
forgetting, emotional dysregulation, psychodynamic directed behavior. Once the preliminary rational
defense, disengagement, a dysfunctional belief).
Downloaded by [University of Utah] at 23:40 22 November 2014

analysis is done, our discovery phase will share an


Notice that many of these ‘‘tasks’’ could eventually initial moment with the model-building procedures
be examined across treatments, although it is ad- of grounded theory (in the next step). However, the
visable to restrict the initial inquiry within a single current step of rational modeling, which is based on
general approach. theory and expert insight, anticipates the empirical
A. Pascual-Leone and Greenberg (2007) deli- modeling against time. For this reason, some meth-
neated and studied an emotional processing task, odological contrasts will be useful here.
and this will serve as an illustration of task selection. In grounded theory method, ‘‘meaning’’ is taken
The ‘‘naturally occurring’’ event of transforming/ as the goal of investigation, in that the analysis has an
resolving emotional distress was identified as begin- emphasis on the structure of meaning as such
ning with a problem state of global distress (i.e., the (Rennie, 2000) rather than the moment-by-moment
task marker) and ending with a state of ‘‘resolution’’ process of meaning construction and its variation
or transformed emotional distress (i.e., the end in time. It follows then that, despite any associa-
marker). The task marker of global distress was tive relations that can be captured in a dendrogram
formally defined as (a) an aversive state of suffering, (a tree diagram illustrating the arrangement of
which is (b) high in expressive arousal, (c) has little or clusters), temporal discontinuities exist from one
no substantive meaning, (d) does not entail a clear category state to the other. Of course, researchers
and specific referent, and (e) lacks both agency and using grounded theory know that meanings evolve in
a clear sense of direction. Thus, states of global time, but their method, nonetheless, does not easily
distress (e.g., hopelessness, helplessness, resignation, allow for the expression of this understanding, and
self-pity, undifferentiated complaint), which are high indeed this is not its purpose. The models produced
in expressive arousal but low in meaningfulness and by grounded theory are insightful and clinically
are in response to personally sensitive themes or life useful descriptive models, which explain behavior
events, were identified as markers for the task of by elaborating complex latent meanings. In contrast,
emotional processing. Similarly, characteristics of the aim of task analysis is to capture the temporally
end-marker states representing advanced processing unfolding development of meaning in action and to
states also were defined to help ‘‘bookend’’ the task to produce predictive models of change.
be studied. This was defined as any of the following: Of course, qualitative methods such as grounded
(a) new feelings being freshly experienced in the theory method can also be used to explore processes
moment, expressed clearly in an integrative and of change, although they more easily explore the
affirmative fashion; (b) productive meaning making meanings of a change experience than the moment-
associated with the experience; (c) positive self- by-moment change itself. However, this difference
evaluation (either implicit or explicit); (d) an agentic may ultimately be a matter of emphasis. The in-
stance with an adaptive action tendency; and (e) tended emphasis in using grounded theory is to
identifying clear and specific objects of emotional provide a faithful description of subjective experi-
concern. ences (i.e., a subjective analysis). By contrast, task
Thus, global distress and the advanced process- analysis emphasizes the natural sequences of a
ing states*or, better said, unskilled versus skilled change event (i.e., a dynamic process analysis of
Task analysis 533

patterns in subjective experience). This is because Rational modeling: Making and managing
the purpose of a task analysis is ultimately to conjectures. A process of change study requires larger
pursue the causal determinants of these sequences rational inferences than those provided in more
through the articulation of change principles. grounded descriptive models of subjective meaning.
It should be clear by now that grounded theory This is because models of change are intended to
and our approach to task analysis in psychotherapy reflect observed patterns and predict future perfor-
research share a moment of subjective analysis mance. Thus, the rational modeling step in task
and modeling. Soon after, of course, the different analysis gives license to farther-reaching critical
objectives of each of these methods become appar- speculations. This is done for the purpose of
ent, as a task analytic program of research moves to explaining transitions between components and
another phase of inquiry. So task analysis begins with may even refer to underlying causal mechanisms
a description of subjective states, followed by an that would help explain the process of change.
identification of natural sequences of change, and Consistent with Lakatos’s (1970) notion of scientific
aims to eventually identify the organismic properties progress, it is also done with the understanding that
that allow for these natural sequences. the model will later be validated (i.e., by subsequent
Thus, once the rational development of categories empirical observations and again during the valida-
is underway in task analysis, those categories are tion phase).
sequenced. Concretely, sequencing is often repre- As a counterbalance to rational modeling, the
sented by arrows between the boxes in a flow chart empirical analysis that follows is an opportunity to
Downloaded by [University of Utah] at 23:40 22 November 2014

(e.g., see Figure 2, which is explained later). Arrows cull essential model components, rejecting any
represent the transformation (or evolution) of mean- excesses or vestige components of the rational model
ing in action from one moment to the next. More- that have not proven useful in describing the closely
over, the arrows represent the client’s shifting of observed change process, and to add any newly
attentional resources (e.g., shifting topics, focusing discovered components. In a very real sense, the
on, disembedding). So the model denotes psycho- researcher adopts opposing investigative attitudes
genetic, causal processes by which clients move for rational versus empirical modeling steps. During
forward. In short, if grounded theory is a method the rational analysis, the researcher assumes the role
that captures the deeper meaning of a process, then of theoretician, allowing for more liberal conjectures
task analysis captures*or better yet ‘‘tracks’’*the rooted in sound theory and inspired by expert
change and implications of a process. insight. However, during the empirical analysis, the

Study 1: Study 3:
Discovery Phase Dynamic Modeling Phase

Start
Global
1 Global Distress
Distress
Degree of Transformation Scale (DTS)

Rejecting Fear / 2 Rejecting Anger


Anger Shame 3 Fear/Shame

4 Negative Evaluation
Negative 5 Need
Need
Evaluation
6 Relief
Emotional Relief
Assertive Anger Grief / 7 Grief/Hurt
or Self-Soothing Hurt
8 Self-soothing
Assertive Anger
Acceptance
& Agency 9 Acceptance & Agency

Figure 2. A task analytically derived model: The emotional processing of distress. Note: In Study 1 (discovery phase), a model was deve-
loped to represent a sequential pattern of change (at left; modified from A. Pascual-Leone & Greenberg, 2007). Later, in Study 3 (dynamic
modeling phase), categories (emotion states) in the model are assigned ordinal values (based or findings from Studies 1 and 2; discovery and
verification phases). Thus, Study 3 begins with ‘‘flattening’’ the model into a 9-point ordinal Degree of Transformation Scale (right).
Copyright # 2007 by the American Psychological Association. Adapted with permission.1
534 A. Pascual-Leone et al.

researcher adopts a new attitude of skepticism when notion of primary versus secondary emotions, adap-
applying the initial rational model to the data. Thus, tive versus maladaptive emotions, and the relation-
the researcher takes on a new aim of searching for ships between these (Greenberg & Paivio, 1997).
differences between the initial model and the data in Using this interpretive framework, six video-recorded
an effort to refute conjectures and refine that model. cases, each beginning with a start marker of global
This change in tact helps to offset biases that may be distress, were selected and intensively examined.
based in the early rational model. Three cases were believed to demonstrate the pro-
ductive transformation of global distress (successful
task completion), whereas the other three distress
Step 2: Phenomena Are Organized into events remained unresolved (Figure 1, middle). This
Coherent Categories (Empirical Modeling) strategy of using three versus three prototypic case
As we have begun to describe, the phenomena are examples is often recommended as a manageable
eventually empirically modeled through some abduc- (and minimal) number of cases for study when
tive method that synthesizes key/essential facets of the developing the empirical model (Greenberg, 2007).
observed event into coherent categories. Abduction is In any case, as the researchers tracked each client’s
a form of inference in which one imaginatively creates process moment by moment, observations of the task
hypotheses that provide plausible explanations performance (e.g., thoughts, feelings, behaviors,
(sometimes causal explanations) for a given set of expressions, meanings) were chunked together into
meaning units. The most important level of abstrac-
Downloaded by [University of Utah] at 23:40 22 November 2014

findings. In essence, when one is confronted with


findings that cannot yet be explained, abduction is tion for these units is that which represents different
the process by which one works backward, inventing experiential states (cf. Greenberg, 2007).
possible explanations (i.e., hypotheses), which might In this fashion, empirical model building should
subsequently be supported (Haig, 2008; Pierce, aim at describing similarities in process among three
19311958). Thus, as categories are abductively good (i.e., successful) examples and at the same time
created from observations, they are constantly com- contrast them with the three poor (unsuccessful)
pared with facets of the initial rational model. The examples. Notice that including poor cases as points
result is an emerging rational-empirical model, of contrast to good cases is essential in task analysis,
one that is reiteratively examined for supporting which is a method to look for optimal elements and
evidence or redundancies and is eventually appraised patterns in subjective experience rather than describ-
for saturation (completeness) and stability. A number ing the diversity of subjective experiences. For
of these strategies for qualitative analysis are very example, unproductive sequences that, nonetheless,
similar to those described by Rennie (2000) in his may be commonplace are not represented in a task
approach to grounded theory. analytic model (unless they also served as initial
When used in task analysis, this step involves performance components in the good cases). More-
identifying and sorting many observed task perfor- over, if a researcher were only to examine good
mances, each one semantically or pragmatically cases, it would be impossible to know which model
distinct and idiosyncratically different from the components were unique to good cases (i.e., opti-
others. The researcher’s struggle in this process, mal), as opposed to those that might also have been
aided by the use of a theoretical lens, is to sort and to found among poor cases.
find invariant patterns (i.e., patterns recurrent across Observe and describe: A dynamic and reiterative
types of situations) in the essential task performances process. In this discovery-oriented phase, an investi-
that contribute to task completion. In the end, the gator constructs the description of a phenomenon.
qualitative approach that is necessary here is to Construction is understood here in the dialectical
identify and elaborate essential aspects of the phe- sense of formulating a coherent representation or
nomena’s teleological structure until such a model understanding that is consistent with a given phe-
reaches stability. Thus, as the data are empirically nomenon and embodies its resistances (the reality
modeled, they inform the rational model such that a constraints it imposes on the subject) and affordances
synthesis of rational and empirical models (see Figure (the possibilities of action or representation that
1, bottom right) is created. This is done by adding, the phenomenon in question offers; see Greenberg,
modifying, or deleting rationally based model com- 1984; Greenberg & Pascual-Leone, 2001; J. Pascual-
ponents in accordance with empirical observations Leone, 1991). Thus, the investigator collects exam-
and by similarly accommodating empirically ob- ples of ill-defined events that are recognized as
served steps within the model. relevantly contrasted (i.e., three vs. three contrasting
In the emotional processing study of A. Pascual- cases). From this, the spiraling method of task
Leone and Greenberg (2007), the rational model was analytic discovery begins with an initial fuzzy con-
based on prior clinical theory that introduced the cept. At first, this event/concept could be defined in
Task analysis 535

any number of ways, which would result in different level of processing (see Figure 2, middle), articulation
conceptualizations of the phenomenon. However, of a core negative self-evaluation is contrasted to an
once the investigator finds an initial way of (a) existential need: This tension serves as a pivotal
suitably defining/measuring the concept (e.g., a given change step that elicits more adaptive emotional
step toward the transformation of distress), the reactions and occasionally produces a sense of relief.
examples are, in turn, reexamined to ensure that Following this is a fourth level of more advanced and
they properly express relevant aspect of the adaptive states in emotional processing (see Figure 2,
data across instances. This allows investigator to lower part). In it, on one hand, clients often
subsequently (b) refine the fuzzy concept, which enter a state of grief/hurt, in which they are able to
further clarifies the phenomenon via a repeated acknowledge personal losses without complaint or
dialectical process of (re)conceptualization and ob- self-pity. On the other hand, clients often become
servation. As the conceptualization of the phenom- mobilized through assertive anger, in which they
enon is increasingly refined, criteria for measurement proactively affirm healthy entitlement to experiences
are developed (Greenberg, 1984, 2007 ). This can be of competence, worth, or connection with others.
viewed as constructive measurement, where the Similarly, clients may enter a state of self-soothing as
phenomenon and its measurement are simulta- a way of directly attending to their existential
neously being coconstructed (‘‘discovered’’) and the needs rather than through explicit assertion. Clients
phenomenon itself is increasingly defined by its in grief/hurt can move to assertive anger or self-
measurement. soothing, but they may just as easily move back again
Downloaded by [University of Utah] at 23:40 22 November 2014

Although there will be large individual differences into grief/hurt; Eventually, a synthesis of these
regarding associations and personal resources, some advanced emotional states leads to the most resolved
of these emotional/meaning pathways become reliable state of acceptance and personal agency (Figure 2,
patterns, invariant across different data instances. So, bottom).
describing common processes of change entails cap- This example of a task analytic model captures
turing these invariant patterns, which in fact, are descriptive phenomenological facets involved in the
facets of an underlying causal model. Thus, modeling change process, which are the result of a subjective
of this kind answers questions such as, ‘‘What do the analysis. However, the general pattern of subjective
three good cases have in common in the way they experiences is also represented, and this representa-
unfold? How are they different from poor cases?’’ For tion suggests certain principles of change. Moreover,
example, the sequence secondary hopelessness (i.e., the model refers to the directionality of (health-
global distress) followed by shame, followed by producing) optimal sequences and points to tension
empowering anger constitutes a good, important between facets of subjective experience, tension that
change sequence in depression. To model these canons seems to propel the client through the change
of change is one major objective of psychotherapy process. These dynamic states of the client ‘‘as
process research. organism’’ suggest causal features in the model (see
In the example of research on transforming global A. Pascual-Leone, 2005).
distress, the synthesis of rational and empirical In the end, through the reiterative process of
models has produced a richly descriptive state- identifying and progressively articulating observa-
transition model of key processes. In their first study, tions, the discovery phase of research yields two
A. Pascual-Leone and Greenberg (2007) describe findings: first, a model of change; second, a catego-
clients who enter a state of global distress (i.e., highly rical measure that helps operationalize critical as-
aroused distress with low levels of idiosyncratic pects of that model using specified criteria. In this
meaning making) as part of initially working through way, for example, A. Pascual-Leone and Greenberg’s
reactive and depressogenic emotions that are undif- (2007) study also produced a lengthy document, the
ferentiated and poorly processed. For the purposes of Classification of Affective-Meaning States (CAMS),
illustration, this model is presented in Figure 2 (left). which provided criteria for each process step in the
Following the global distress marker, fear and shame model. The Appendix offers an overview of each step
represent subsequent, second-level states in the addressed in the discovery phase and outlines the
model of processing. These are characterized by goals for the phases of research to follow.
deep, enduring, painful yet familiar states that are
highly idiosyncratic and often anchored in generic
Step 3: Qualitative Categories Are Quantified
autobiographical narratives. Clients at a comparable
(Measurement)
level of processing alternatively enter into a state of
rejecting anger, for which the reactive action ten- After what amounts to a liberalized phenomenolo-
dency is either to angrily create distance from, or gical analysis (i.e., the rational/empirical model-
destroy, the source of distress. Further on, at the third building process) all potentially relevant categories
536 A. Pascual-Leone et al.

are quantified. This is where task analysis moves tic change. In short, successful psychotherapy is a
beyond qualitative research as it is most often complex process in which the intrapsychic context
practiced; qualitative categories must ultimately be changes, and this change in context is ill represented
codified into quantifiable terms. However, the new using standard psychometric scales, which often
quantitative relations must preserve the relevant oversimplify the process of change.
qualitative relations. Interobserver reliability and The second reason why task analysis develops
consensual classifications are two approaches for customized measures is that developing categorical
this that have been used in psychotherapy research. measures actually helps to operationalize dynamic
Psychometric scale versus model measurement. intrarelational aspects of the model. Standard psy-
Rather than using more generic psychometric coding chometric scales reflect increases in quantity on a
systems, coding in task analysis is developed speci- single dimension (e.g., ‘‘What is the degree of A in
fically to measure relevant-task constructs. There are the client’s performance?’’). Model measurement,
two chief reasons for this: The first is that measure- however, relies on categorical measures, which make
ment in task analysis must be highly context specific. qualitative distinctions between different sets of
In contrast, generic scales often index single vari- affective-cognitive performance features (e.g., ‘‘Is
ables (e.g., depth of experiencing, emotional arousal, the client doing A, B, or C?’’). This form of
client cognitive progress, self-disclosure, strength of measurement promotes the identification of qualita-
alliance) in order to be reliable and valid. However, tive patterns across time and forms the basis for the
construction of a model of change. A task analytically
Downloaded by [University of Utah] at 23:40 22 November 2014

although simple psychometric scales allow one to


measure progress on one or more dimensions, derived measure organizes observed categories and
usually on ordinal or interval scales, the inherent their relationships to one another (e.g., A and B are in
assumption is that such measurements represent opposition, are alternatives, or are conditions of one
qualitatively comparable changes that do not each another). Model measurement also relates categories
require elaborate contextualization. Model measure- across time (e.g., A precedes B). This organization of
ment, on the other hand, presupposes that a measurable model categories provides an ordered set
participant’s performance in complex tasks can (i.e., a ‘‘continuum’’) of cognitive-affective categories
only be understood through the interpretive lens of (or structures). A continuum such as this gives a
a theoretical model. Thus, model measurement is much richer description of performance than any
the qualitative measurement of participants’ pro- single quantitative dimension.
gress within the contextual framework of a model Thus, measurement is relevant to the discovery
that can be empirically related to change. phase of task analysis in that it is a product; but also
A classic example that speaks to the importance of in that the development of measurement criteria
model measurement is when observing developmen- itself is an integral part of the method’s process.
tal performance on Piagetian conservation tasks. Obviously, this procedural significance of measure-
Given these context-sensitive tasks, children demon- ment would not be possible using an already
strate choppy development, developmental décalage, available, generic psychometric measure. Although
in acquiring abilities that would otherwise appear as measurement development is part and parcel of the
pertaining to a homogeneous area of competence. task analytic method, this development of qualitative
From the perspective of psychometric scale mea- findings could be similar to extensions of other
surement, these choppy, U-shaped patterns of devel- qualitative approaches. Levitt and Frankel (2004),
opment represent a transitionary regression or lapse for example, began with findings from a grounded
in measured ability. However, when the aberrant theory analysis of in-therapy silences and then used
performance is considered in the context of a model those findings as coding criteria for observation.
of change/development, such fluctuations in perfor- Thus, other qualitative approaches may also provide
mance are anticipated by the model (see J. Pascual- the means for a rich categorical measure.
Leone & Johnson, 2005). Similarly, Stiles (2006) has Figure 2 shows measurable model components
described a ‘‘sawtoothed’’ pattern of client change that concretely illustrate complex notions of primary
that is consistent with his context-sensitive and adaptive, maladaptive, and secondary emotions
theoretically informed model of successful therapy (Greenberg, 2002) as they appeared in therapy.
as an assimilative process. Moreover, aberrations Such constructs are highly context sensitive and
may actually reflect key change events. For example, would not have been measurable using generic
in working through complex grief, clients often go psychometric scales. Moreover, early renditions of
through a stage of anger before they are able to fully the model identified anger, sadness, and self-soothing
experience sadness and mourn. Similarly, an alliance as each having important (albeit highly complex)
rupture may serve as a paradoxical step on the path roles in the change process. However, as specific
toward greater interpersonal intimacy and therapeu- criteria were being articulated for the accompanying
Task analysis 537

measure (i.e., the CAMS), this influenced how the Confirming process and relating it to outcome. The
described states were functionally understood in their second major phase of a task analytic research
interaction. Thus, the process of constructing mea- program is represented at the bottom left of
surement criteria in parallel with a model of change Figure 1. Given that the model is theory laden and
serves to identify a suitable dynamic relational interpretive (albeit with support from multiple case
network among model components. Unlike a stan- analyses), it must be tested. Thus, the validation
dard psychometric scale, a categorical measure can phase is the second part of a task analysis. In this
describe qualitatively different self-organizations, phase, the qualitative model already developed in the
and when used in conjunction with a model of discovery phase is tested by applying empirical
change, it can capture nonlinear relationships among criteria to a new, larger sample of cases. This phase
uniquely identified model components. of research establishes predictive and external valid-
When developing measurement criteria, descrip- ity of the synthesized model using in-session exam-
tors of each state (i.e., each model component) ples that were not part of the original model
could make use of qualitative and quantitative (cut- building. In other words, the validation phase must
both validate the components of the model and
off) criteria. Moreover, coding criteria may be drawn
relate the described process (and patterns) of change
from original work by the researcher, inspired by
to outcome.
existing psychometric measures, or may come from
To this end, a larger selection of available cases
combining or dismantling existing measures. In the
must be reviewed for the identified task marker.
Downloaded by [University of Utah] at 23:40 22 November 2014

CAMS, for instance, coding criteria for each cate-


Then, using criteria that are independent from the
gory entailed an assemblage of descriptors about
model measure, this larger sample must be sorted
subjective meaning, nonverbal behavior, vocal qual- into those who attained productive in-session end
ity, and so on (see A. Pascual-Leone & Greenberg, states and those who did not (i.e., good vs. poor
2005). To accomplish this, new categorical measures event outcomes). It is advisable for researchers to
in task analysis often assimilate concepts and specific have some hand in ensuring that both types of
observational criteria from a wide range of process outcomes are represented in this newly collected
measures, ‘‘using them for parts’’ and reconstituting sample, so that a contrast between subgroups will be
them for a customized purpose. This approach possible. As a case in point, in their second study A.
permits one to capitalize on new observational Pascual-Leone and Greenberg (2007) collected 34
insights, as well as existing research, to produce a video examples of clients in global distress. They
measure that highlights structural patterns of change then used ‘‘outside’’ indices of task outcome (i.e., the
(Rice & Greenberg, 1984) rather than simple Experiencing Scale and clinical judges) to sort these
degrees of change. cases into two subgroups.
Next, independent observers, who are blind to
both the tentative model and the event outcomes,
The Validation Phase
use the model measure to identify components/states
Step 4: Quantitative Analyses on Identified represented in the model. For A. Pascual-Leone and
Structures of the Model Greenberg (2007), this meant that blind observers
rated clients’ emotional states using the CAMS (i.e.,
Quantitative analyses are done on those qualitative
the categorical model measure). The resultant data
structures (model components) that were first iden-
set must allow researchers to (a) establish interrater
tified through the model-building process. Given a
reliability of the model measure, (b) show that good
suitably practical classification system, the researcher
versus poor event outcomes could be distinguished
is now in a position to do quantitative inquiries on a statistically using the presence or absence of different
larger and relatively independent data set. This model components, and (c) use statistical analyses to
research step is used to verify earlier qualitative demonstrate that for good treatment outcomes key
findings about process and demonstrate their rela- states (e.g., of emotion, meaning, behavior) are likely
tionship to outcome. This has been referred to as the to emerge in a sequential pattern described by the
validation phase of a task analysis (Greenberg, 2007). model.
Notice that confirmatory empirical controls are built It is worth highlighting here that this last aim*
into the task analytic program of research: Although that is, seeking empirical support for an explicitly
the synthesized model embodies a number of theore- sequential pattern of process*is a new methodolo-
tical and rationally based findings, these cannot be at gical development in task analysis. This development
the expense of their empirical grounding because highlights the potential of task analysis as a method
now, in the verification phase, the model must show for the causal modeling of change. Overall, such
predictive validity. findings provide validation for a process of change
538 A. Pascual-Leone et al.

model in its relevance to positive therapeutic effects verified, this gives legitimacy to an associated measure
and in its sequential structure. Findings of this kind that is ordinal rather than simply categorical. So,
also validate the relevant model measure (e.g., the based on the verification offered by quantitative
CAMS) that reflects the modeled process. Task findings and supported by a theory of change, a
analytic studies have traditionally stopped here after categorical measure can be transposed onto an ordinal
the validation of discovery-oriented findings. quantitative measure. In this way, one can use the new
In a different approach to the quantitative valida- ordinal measure (ordered set of categories) for other
tion of qualitative findings, Levitt and Frankel’s studies that examine new, more complex, patterns
(2004) nominal measure (based on grounded theory across time.
findings) of different kinds of client in-therapy Thus, once categories are quantified and shown to
silences was applied to an independent sample and be generalizable beyond the initial handful of cases
related to outcome (Frankel, Levitt, Murray, Green- that were used in the discovery phase, this new
berg, & Angus, 2006). This method of validation generation of hypotheses and quantitative analyses
was useful in demonstrating the relevance of cer- can extend the applicability of the model and
tain productive silences to good overall treatment measure. Moreover, this new set of more theoretically
outcome and the reliability of a set of observable refined hypotheses are ones that were not initially
criteria. However, as discussed earlier, grounded available to either purely qualitative or quantitative
theory models entail discontinuous categories of inquiries alone: Qualitative inquiry was limited in the
breadth of scope it could encompass, whereas quan-
Downloaded by [University of Utah] at 23:40 22 November 2014

meaning, and so the categories/codes in the model


of client silences need not (and cannot) be validated titative inquiry was limited in the complexity or
with respect to their sequential order. Although this richness of measurement it could allow. The dynamic
is a moot point regarding validation, it does highlight modeling phase organizes model components into an
the difference in scope and purpose of grounded ordered set (a ‘‘continuum’’) of structures, which
theory models compared with task analytic models gives an opportunity to further explore ordered
of subjective experience. The former provides a patterns that seem to be invariant within a given
change process. Specific hypotheses might address,
comprehensive understanding of a subjective phe-
for example, complex temporal micropatterns within
nomenon, one that may be inclusive of change but
a session, subtle yet systematic macropatterns across
will not be as focused on it. In contrast, the purpose
treatment, idiosyncratic patterns related to individual
of the latter (task analysis) is to focus on a specific
differences, or patterns of model processes across
change process. This more delineated focus of task
treatment approaches. This methodological step (not
analysis is in the aim of developing a theory of
represented in Figure 1) is a new contribution beyond
sequential process, one that may allow causal me-
the task analytic method typically presented by
chanisms of subjective experience to become more
Greenberg (1984, 2007) and used to date in clinical
apparent through observed patterns of change.
psychology.
In the validation phase of our task analytic
example, the CAMS measure was used to demon- Extending the use of model measures. To illustrate
strate that for productive in-session events eight of how model measures can be developed and used for
the emotion categories are likely to emerge in the study of dynamic patterns among qualitative
sequentially ordered patterns of at least four phases states, we return to the program of research on
(A. Pascual-Leone & Greenberg, 2007). As we shall transforming global distress. A. Pascual-Leone
see, this finding will provide an empirical rationale (2005, 2009) predicted quantifiable dynamic pat-
for the next phase of research to convert the model terns within a qualitative model, and this is, to our
measure (e.g., CAMS) from a nominal into an knowledge, the only known example of this metho-
ordinal scale. dological approach in psychotherapy research. Qua-
litative and theoretical models of emotion-focused
therapy (EFT) provided support for situating
Dynamic Modeling Phase CAMS codes onto an ordinal scale (i.e., a ‘‘con-
Step 5: New Quantitative Analyses to Explore tinuum’’ of structures) of emotional transformation.
Sequential Structures So, in the dynamic modeling phase (A. Pascual-
Leone, 2009), emotion codes of the CAMS were
Until this point in a research program, the model transposed onto a 9-point ordinal scale. This was
consisted of a set of structures and had developed in done by taking categorical coding (i.e., emotion
tandem with a categorical measure, which, in turn, codes in the CAMS; Figure 2, left) and assigning
has been used to verify the model empirically. How- ordinal ratings to them depending on their position
ever, when a sequentially ordered model of process in the sequence, which was hypothesized in the
(one with ordinally fixed categories) is empirically discovery phase and subsequently validated. Notice
Task analysis 539

that the measure itself was not substantially changed quantitative analyses such as the statistical modeling
in doing this, except for the fact that categories were of dynamic, nonlinear patterns of change according
now understood as indexing progress in an organis- to the specified model. For example, using the DTS
mic process of change. Of course, criteria for (i.e., an ordinal rendition of the model measure)
identifying model components in the ‘‘new’’ measure A. Pascual-Leone (2009) was able to formulate
remained the same, and the number of categories/ hypotheses about productive client performances as
components in the measure again remained largely occurring through a sequential (albeit nonlinear and
unchanged. In short, the only significant change to cyclical) pattern progressing forward and then
the instrument was the addition of hierarchically dropping backward to earlier levels of processing
ordered values assigned to each emotion state. As a (i.e., a ‘‘sawtoothed’’ pattern, as also described by
result, each state now had a relative position Stiles, 2006). Based on a sample of 34 cases (one
(validated in the previous phase) as a step on the session per client), results of univariate and boot-
path toward resolution (i.e., emotional transforma- strapping analyses showed that effective emotional
tion). This theory upgrade permits codes to be processing was simultaneously associated with an
understood as expressing an ordinal measure: overall steady improvement according to both the
the Degree of Transformation Scale (DTS; see model (as measured by DTS) and the increased
Figure 2, right). emotional range of clients. The study was also able
Thus, the complex nonlinear model of emotional to raise questions about (and track) the dynamic role
processing was ‘‘flattened’’ into a string of categories of microrelapses (i.e., momentary collapses or re-
Downloaded by [University of Utah] at 23:40 22 November 2014

paired to ordinal numbers: Global distress, the least gressions) that occur within a session and the
resolved state, is represented by 1 and is followed by relationship of this clinical phenomenon to progress.
progressively more processed states, such that fear/ In the end, the data were rich enough to support
shame 2, rejecting anger 3, negative evaluation  these new hypotheses (Figure 3).
4, need 5, relief6, grief/hurt7, assertive an- Second, an ordinal model measure also allows one
ger 8, self-soothing 8, and the most resolved state to study a given task as it is progressively engaged
of acceptance and agency 9. That the DTS nu- across multiple treatment sessions. Although task
merically conveys model ratings allows analyses in analysis is initially designed to model exemplar
the new phase of research to exploit the ordinal performances based on observations in a circum-
properties of the model measure. This data-sensitive scribed set of events, the quantitative (in this case,
ordering, which tracks progress in the clients’ ordinal) measure provides a means to track the same
microdevelopment, reflects a continuum of cogni- detailed process using a longer perspective (e.g., the
tive-affective process moments, moments that can be entire course of a treatment). With this sort of
investigated through a new generation of hypotheses. methodology, many less-than-ideal processes of
Some of these new hypotheses, however, may not change that are believed to be reiterative and cumu-
have been conceivable during the original model- lative can now be examined. For example, using four
building phase (e.g., hypotheses regarding micro- or case studies, A. Pascual-Leone (2005) examined the
macrosequential patterns). entire course of each treatment (1214 sessions) for
Another example of the quantification of qualita- markers of global distress and rated relevant emotion
tive model-based findings can be found in Stiles’s events on the DTS. For each case, these data across
Assimilation of Problematic Experiences Scale sessions were then subjected to hierarchical linear
(APES), which is an ordinal scale of the type being modeling (sessions nested within a given course of
described here. The APES provides an ordered treatment). The purpose of these analyses was to
continuum of qualitative structures (Stiles, 2006) capture patterns of change relevant to the emotional
that have been derived mainly from theory-building processing model, which might only occur across
case studies (a kind of rational-empirical method treatment sessions (i.e., the gradual development of
that has an affinity to task analysis). As such, the emotional resilience). The series of case analyses
APES effectively allows for detailed tracking of client provided evidence for treatmentwise patterns by
process in time. However, assessing assimilation which clients develop emotionally in psychotherapy.
levels cannot be based on a single moment of It is important to notice that micro- and macro-
observed performance, demanding instead a detailed patterns such as those referred to previously (i.e., A.
understanding of a case in its entirety. Pascual-Leone, 2005, 2009) must be theoretically
Testing new hypotheses. Returning to the task ‘‘invented,’’ using a rational process to go beyond
analytic method, once one has upgraded to an straight qualitative findings, before they can be
ordinal model measure, new hypotheses and ana- quantitatively ‘‘discovered’’ (J. Pascual-Leone &
lyses may address at least two sorts of issues. First, Sparkman, 1980). This is to say that hypotheses in
an ordinal model measure allows one to do detailed the dynamic modeling phase often cannot be explored
540 A. Pascual-Leone et al.

Degree of Transformation Scale (DTS)


9
8
7
6
5
4
3
2
1
0
1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 46 49 52 55 58 61 64

Time (30 second intervals)

…Durations of “collapses”
4.12 min 3.75 min 0.94 min

Figure 3. Using an upgraded measure to test new hypotheses: plotting the Degree of Transformation Scale against time. Note: In Study 3
(dynamic modeling phase), microtemporal patterns were examined. This plot shows codes of emotional improvements and collapses for a
typical good outcome. One of the new hypotheses tested in this way (bottom) was about the changing duration of momentary emotional
Downloaded by [University of Utah] at 23:40 22 November 2014

collapse in a good outcome case. Figure modified from A. Pascual-Leone (2009). Copyright # 2009 by the American Psychological
Association. Adapted with permission.1

or even concretely articulated during the initial Next, this model is skeptically appraised and enriched
discovery phase. The reason for this is because such using qualitative empirical observations taken from
hypotheses require a level of detail (e.g., nonlinear contrasting cases, completing the discovery phase.
micropatterns of change) or a scope of observation Then, the rational-empirical model is used as a
(e.g., an entire set of treatments sessions) that goes far hypothesis generator and subjected to rigorous test-
beyond the analytical capacity of a process researcher ing in the validation phase, thereby affirming the
using qualitative empirical methods alone. None- credibility of the model. In prior research, this has
theless, at the same time, such hypotheses are based been done by relating the presence of key components
on qualitative process models of change, the richness in the model to good task outcome. Here, we have
of which cannot be captured by the sole use of presented a modified verification method, which goes
standard psychometric measures. further by relating a sequential pattern of processes to
outcome.
Finally, the most innovative development pre-
Conclusions sented here is to add a new phase: In the dynamic
Task Analysis: A Guide for Programmatic modeling phase, one explores the sequential struc-
Research on Change ture that underlies a given model of change. We have
described how categorical measures based on quali-
Task analysis is best conceptualized as a method that tative findings can be revised into ordinal measures,
systematically unfolds over a program of research by opening the way to a new generation of inquiries that
way of complementary methodological phases (cf. fuse both qualitative observations and quantitative
Lakatos, 1970). We believe that by presenting the predictive modeling. This task analytic approach to
method in this way, as a template for programmatic model building lends itself to developing causal
research, it will be more accessible to researchers models of change by discerning those organismic
who hope to develop models and measures about a recurrent characteristics (i.e., invariants) of the
process of change. In this article, we have also change process that likely index clients’ organismic
described the way in which task analysis differs in processes that induce change.
scope and purpose from other qualitative methods
and that task analytic models may yield both
descriptive and causal accounts of change. More- Future Directions
over, we have provided a step-by-step account of task Challenges to making causal models. There are
analysis, which emphasized recent developments in some tangible limits to task analysis and other
task analysis (also see the Appendix). methods seeking to sketch causal models of change
Modeling a process of change in task analysis in psychotherapy. Although modeled patterns of
begins with a rational model that draws on the change allow one to identify invariant patterns in
literature and on one’s expert clinical experience. the process, we can only infer that these patterns
Task analysis 541

reflect causal mechanisms to the extent that more goal-directed behaviors could likely be extended using
fundamental research has articulated relevant pro- task analytic methods to eventually yield quantita-
cesses as constructs (e.g., hidden organismic opera- tively predictive models of change. In short, if reliable
tors). Existing theories of functioning in clinical and start and end markers were established for these or
personality psychology are not yet coherent enough other tasks of interest, diverse theories (cognitive,
to allow this to be done more than tentatively. Until behavioral, psychodynamic, and relational theories)
then, we must be content with inferring that certain could be used to interpret key processes that lead to
client characteristics (i.e., organismic dispositions) change. In the end, models of moment-by-moment
exist, without having a clear idea of what they are. change help to explain how and why clients eventually
Conceptual challenges for using task analysis. Aside improve in treatment and can have important
from the ambitious aims of causal modeling, task implications for practice.
analysis has not been widely used in the field of
psychotherapy research. One of the reasons for this
Note
is likely a lack of clarity in the types of questions that 1
might be best suited to the method. Indeed, See reference list for official citation to be used in referencing
this material. The use of APA information does not imply
although there have been some noteworthy excep- endorsement by the APA.
tions, task analysis as applied to psychotherapy has
generally been associated with EFT and with mar-
Downloaded by [University of Utah] at 23:40 22 November 2014

ker-based interventions in particular. We shall in- References


dicate future possible applications where task Frankel, Z., Levitt, H. M., Murray, D. M., Greenberg, L. S., &
analysis could be a useful strategy but is not Angus, L. (2006). Assessing silent processes in psychotherapy:
currently used. The fact that clients’ productive An empirically derived categorization system and sampling
strategy. Psychotherapy Research, 16, 627638.
processes are not always self-evident and, moreover,
Greenberg, L. (1984). Task analysis: The general approach.
that they are not always conceptualized as reflecting In L. Rice & L. Greenberg (Eds), Patterns of change: Intensive
some specific goal-oriented behavior poses an im- analysis of psychotherapy process (pp. 126184). New York:
portant conceptual challenge. Thus, a prerequisite Guilford Press.
to extending the application of task analysis is to Greenberg, L. (1991). Research on the process of change.
explicate the nature of clients’ implicitly goal- Psychotherapy Research, 1, 1424.
Greenberg, L. S. (2002). Emotion-focussed therapy: Coaching clients
oriented efforts (i.e., what are often tacit tasks). to work through their feelings. Washington, DC: American
For example, how a client eventually manages to Psychological Association.
generate evidence to challenge automatic dysfunc- Greenberg, L. S. (2007). A guide to conducting a task analysis of
tional thoughts can be seen as an important client task psychotherapeutic change. Psychotherapy Research, 17, 1530.
that requires several tacit steps in cognitive-affective Greenberg, L. S., & Paivio, S. C. (1997). Working with emotions in
psychotherapy. New York: Guilford Press.
information processing. Moreover, although these Greenberg, L. S., & Pascual-Leone, J. (2001). A dialectical
steps have been theorized by means of cognitive constructivist view of the creation of personal meaning. Journal
behavioral (rationalist) models, they have not been of Constructivist Psychology, 14, 165186.
systematically studied using a moment-by-moment Greenberg, L. S., & Pinsof, W. M. (1986). Process research:
Current trends and future perspectives. In W. M. Pinsof & L. S.
discovery-oriented phase of research. On another
Greenberg (Eds), The psychotherapeutic process: A research
note, psychodynamic insights often occur by coordi- handbook (pp. 320). New York: Guilford Press.
nating processes that are within as well as outside the Haig, B. D. (2008). Scientific method, abduction, and clinical
client’s awareness, which suggests the client’s task of reasoning. Journal of Clinical Psychology, 64, 10131018.
gaining insights may be more amenable to task Joyce, A. S., Duncan, S. C., & Piper, W. E. (1995). Responses to
analyses than to other qualitative methods (e.g., see dynamic interpretation in short term individual therapy.
Psychotherapy Research, 5, 4962.
Joyce et al., 1995). For task analysis to be used more Kuhn, T. S. (1996). The structure of scientific revolutions (3rd ed.).
broadly, aspects of forming and maintaining a ther- Chicago: University of Chicago Press.
apeutic alliance must also be conceptualized as goal- Lakatos, I. (1970). Falsification and the methodology of scientific
directed efforts (i.e., as interpersonal tasks). For research programmes. In I. Lakatos & A. Musgrave (Eds),
instance, just as Safran and Muran (2003) have Criticism and the growth of knowledge (pp. 5158). Cambridge,
UK: Cambridge University Press.
identified resolving relationship ruptures as a task in Levitt, H. M., & Frankel, Z. (2004). Pausing Inventory Categor-
and of itself, the initial development of a strong ization System, revised. Unpublished manuscript, University of
therapeutic alliance could also be conceptualized as Memphis.
an interpersonal task. Similarly, terminating therapy Pascual-Leone, A. (2005). Emotional processing in the therapeutic
is again a task in which two participants strive to reach hour: Why the only way out is through. Unpublished doctoral
thesis, York University.
a defined end marker. Pascual-Leone, A. (2009). Dynamic emotional processing in
In short, many qualitative studies that can be experiential therapy: Two steps forward, one step back. Journal
understood as reflecting some (implicitly or explicitly) of Consulting and Clinical Psychology, 77, 113126.
542 A. Pascual-Leone et al.
Pascual-Leone, A., & Greenberg, L. S. (2005). Classification of nature of successful task performance (i.e., reso-
Affective-Meaning States (CAMS). In A. Pascual-Leone (Ed.),
lution, end marker).
Emotional processing in the therapeutic hour: Why the only way out
is through (pp. 289367). Unpublished doctoral thesis, York . Explicate the investigator’s perspectives and as-
University. sumptions (e.g., emotion theory).
Pascual-Leone, A., & Greenberg, L. S. (2006). Insight and . Specify the task environment and its potency
awareness in experiential therapy. In L. G. Castonguay & C. (e.g., EFT for depression).
E. Hill (Eds), Insight in psychotherapy‘‘ ). Washington, DC:
American Psychological Association (pp. 3156).
. Conduct rational analysis and build a model (i.e.,
Pascual-Leone, A., & Greenberg, L. S. (2007). Emotional review the literature, speculate).
processing in experiential therapy: Why ‘‘the only way out is
through. Journal of Consulting and Clinical Psychology, 75, 875 Step 2: Empirical Model and Synthesized Model (Figure
887. 1, middle, bottom right)
Pascual-Leone, J. (1978). Compounds, confounds, and models in
developmental information processing: A reply to Trabasso and . Contrast good and poor performances to discern
Foellinger. Journal of Experimental Child Psychology, 26, 1840.
essential steps (i.e., fuzzy concepts).
Pascual-Leone, J. (1991). Emotion, development, and psy-
chotherapy: A dialectical constructivist perspective. In J. Safran . Merge rational and empirical models. Then
& L. S. Greenberg (Eds), Emotion, psychotherapy and change reiterate modeling and check for stability.
(pp. 302335). New York: Guilford Press.
Pascual-Leone, J., & Johnson, J. (2005). A dialectical constructi- Step 3: Measurement
vist view of developmental intelligence. In O. Wilhelm & R. W.
. Develop a categorical measure with as many
Downloaded by [University of Utah] at 23:40 22 November 2014

Engle (Eds), Handbook of understanding and measuring intelli-


gence (pp. 177202). Thousand Oaks, CA: Sage. categories as components in the model.
Pascual-Leone, J., & Sparkman, E. (1980). The dialectics of
empiricism and rationalism: A last methodological reply to
Trabasso. Journal of Experimental Child Psychology, 29, 88101. Validation Phase
Peirce, C. S. (19311958). Collected papers (Vols. 18, C.
Hartshorne, P. Weiss, & A. Burks, Eds.). Cambridge, MA: Step 4: Quantitative Analysis (Figure 1, bottom left)
Harvard University Press.
Rennie, D. L. (2000). Grounded theory methodology as metho- . Establish interrater reliability of the categorical
dical hermeneutics: Reconciling realism and relativism. Theory measure.
& Psychology, 10, 481501. . Validate the modeled processes by relating their
Rennie, D. L., & Toukmanian, S. G. (1992). Explanation in
psychotherapy process research. In S. G. Toukmanian & D. L.
presence to positive in-session effects.
Rennie (Eds), Psychotherapy process research: Paradigmatic and . Relate the model processes to final treatment
narrative approaches (pp. 234251). London: Sage. outcome. This can be done directly through an
Rice, L. N., & Greenberg, L. S. (Eds). (1984). Patterns of change: outcome study or indirectly by drawing on exist-
Intensive analysis of psychotherapy process. New York: Guilford
ing outcome literature.
Press.
Safran, J. D., & Muran, J. C. (2003). Negotiating the therapeutic
alliance: A relational guide to treatment. New York: Guilford
Dynamic Modeling Phase
Press.
Stiles, W. B. (2006). Assimilation and the process of outcome: Step 5: Explore Sequential Structures
Introduction to a special section. Psychotherapy Research, 16,
389392. . ‘‘Upgrade’’ to an ordinal measure by assigning
Stiles, W. B. (2007). Theory-building case studies of counselling values to the existing categorical measure.
and psychotherapy. Counselling and Psychotherapy Research, 7,
122127.
. Generate and test new hypotheses about the
model’s micro- or macropatterns of change.

Appendix: Step-By-Step Goals in Task Analysis


Discovery Phase
Step 1: Rational Model (Figure 1, top left)
. Define the nature of a task of interest: Articulate
its starting point (task marker) and specify the

You might also like