Content uploaded by Louise Sundararajan
Author content
All content in this area was uploaded by Louise Sundararajan on Mar 04, 2016
Content may be subject to copyright.
In: Semiotics Theory and Applications ISBN 978-1-61728-992-7
Editor: Steven C. Hamel © 2010 Nova Science Publishers, Inc.
Chapter 3
LANGUAGE, EMOTION, AND HEALTH:
A SEMIOTIC PERSPECTIVE ON THE WRITING CURE
Louise Sundararajan1, Chulmin Kim2, Martina Reynolds3 and
Chris R. Brewin4
1Rochester Regional Forensic Unit, NY
2Rochester Institute of Technology
3Brunel University
4University College London
ABSTRACT
The writing cure, otherwise known as expressive writing, is widely accepted as an
effective intervention. Hundreds of studies have shown that writing about one‘s thoughts
and feelings for 3 days, with at least 15 minutes a day, has beneficial effects on physical
and mental health. Yet, after more than two decades of research, there remains a large
gap between evidence and explanation for the phenomenon. The problem, we suggest,
lies in the general neglect to gain a deeper understanding of the basic building blocks of
the writing cure, namely language. This vacuum can be filled by Peircean semiotics.
Peirce‘s triadic circuitry of the sign is explicated and applied to the development of a
taxonomy of expressions of self and emotions. This taxonomy has been implemented by
a pattern matching language analysis program, SSWC (Sundararajan-Schubert Word
Count) to test our theory-based predictions of the health consequences of language use.
Two empirical studies of the writing cure that utilized SSWC for textual analysis are
presented as demonstration of the heuristic value of applied semiotics.
The writing cure has had an impressive track record since its first introduction by
Pennebaker (Pennebaker, 1985; Pennebaker and Beall, 1986) in the eighties. For the past
two decades, hundreds of studies have shown that writing about one‘s thoughts and
feelings has beneficial effects on physical and mental health (Frattaroli, 2006). But why?
What is it about language that its utilization for emotion expression has consequences for
health? This question has never been addressed by the extant theories of the writing cure
(e.g., Bootzin, 1997; King, 2002; Pennebaker, Mayne, and Francis, 1997). An
explanation that seems to have the most empirical support (Frattaroli, 2006) is emotion
exposure theory (Sloan and Marx, 2004), which by considering language use as an
The exclusive license for this PDF is limited to personal printing only. No part of this digital document
may be reproduced, stored in a retrieval system or transmitted commercially in any form or by any means.
The publisher has taken reasonable care in the preparation of this digital document, but makes no expressed
or implied warranty of any kind and assumes no responsibility for any errors or omissions. No liability is
assumed for incidental or consequential damages in connection with or arising out of information contained
herein. This digital document is sold with the clear understanding that the publisher is not engaged in
rendering legal, medical or any other professional services.
Louise Sundararajan, Chulmin Kim, Martina Reynolds et al.
66
instance of exposure therapy tells us more about the latter than language per se. Another
widely accepted explanation is narrative structure (Smyth, True, and Souto, 2001), which
claims that verbal expression facilitates the transformation of experiences and memories
into a structured ―story‖ (Pennebaker and Seagal, 1999). But Graybeal, Sexton, and
Pennebaker (2002) found no correlation between narrativity and health benefits. The use
of different types of words has also been investigated (Campbell and Pennebaker, 2003).
The finding is that the use of emotion words was not consistently correlated with self-
reported emotionality, and that ―style words‖-- such as function words and pronouns--
were more relevant to health status. Not based on any linguistic theory, such ad hoc
distinctions of language use seem arbitrary, albeit empirically supported. To date,
expressive writing remains a black box, in the words of Laura King: ―First, expressive
writing has health benefits. Second, no one really knows why‖ (King, 2002, p. 119). The
problem, we suggest, lies in the general neglect to gain a deeper understanding of the
basic building blocks of the writing cure, namely language. This vacuum can be filled by
Peircean semiotics.
The exposition of Peircean semiotic consists of five sections. The introduction sets
the stage by casting the language and health equation in the context of Shannon‘s ideal
code, which is informationally the most complex and energetically the least costly.
Peirce‘s triadic circuitry of the sign is subsequently introduced as an algorithm of
complexity that extends Shannon‘s information theory. Next, we introduce a language
analysis program, SSWC (Sundararajan-Schubert Word Count), which implements a
proposed taxonomy, derived from Peircean semiotics, of different types of language use
with varying degrees of complexity. The penultimate section presents two empirical
studies that showed how language analysis by means of SSWC can shed some light on
the language and health connection across different conditions. The conclusion discusses
the potential contributions of Peircean semiotics to theory and research on the writing
cure.
INTRODUCTION
Consistent with Heidegger‘s dictum that ―Man lives in language, as language‖ (cited in
Ott, 1972, p. 169), Charles Sanders Peirce claims that the sign user and the sign have
coalesced at a deeper level: ―the word or sign which man uses is the man himself . . . . Thus
my language is the sum total of myself‖ (Peirce, 1931-58, Vol.5, paragraph 314, emphasis in
the original). It is this semiotic perspective that provides solid theoretical grounding for the
language and health equation, rendering efficiency of the sign and health status intimately
related. In the following investigation, we are guided by three insights that may be derived
from Peircean semiotics: a. language is a sign, in the present context, a representation of
emotion information; b. the quality of the sign matters for the sign user, in the present
context, the quality of emotion representation has health consequences. Lastly, the quality of
emotion representation can be modeled by the triadic circuitry of the sign.
The quality of the sign is a central concern of Charles Peirce (Sundararajan, 2008,
Colapietro, 1989), but he did not spell out clearly the ramifications of this for the language
and health equation. To investigate this question, we may situate the language and health
connection in a larger context—the relationship between information and energy. The
intimate connection between information and energy is suggested by Metcalfe and Mischel
(1999), who have postulated two systems of emotion—―a cool, cognitive ‗know‘ system and
a hot, emotional ‗go‘ system‖ (p. 3)--corresponding to information and arousal, which,
Language, Emotion, and Health
67
according to both Berlyne (1960) and Estes (1972), constitute two essential functions of any
stimulus. Metcalfe and Mischel (1999) claim that the cool system is ―complex‖, whereas the
hot system is ―simple‖ (p. 4). This point can be further elaborated by the inverse relationship
between information and energy.
Energy is governed by the law of conservation, whereas information is concerned with
successful transmission, which requires complexity, in terms of order and organization, to be
successful. The inverse relationship between the two has been suggested by a number of
writers. According to Shannon (Campbell, 1982), information with a high degree of order and
organization renders energy useful or efficient, analogous to the cool system of Metcalfe and
Mischel (1999), whereas entropy (disordered information) renders energy costly analogous to
heat or the hot system of Metcalfe and Mischel (1999). David Bohm (1994) envisions a
progression toward the optimal display of meaning in a representation that is informationally
the most complex (satisfying the condition for transmission) and energetically the least costly
(satisfying the principle of conservation). Consistent with the finding that
psychophysiological arousal was associated with language disturbance as measured by
reference errors (Burbridge, Larsen, and Barch, 2005), Metcalfe and Mischel (1999) claim
that there is a compensatory and curvilinear relationship between level of activation/arousal
and the degree to which complex, integrated behavior is possible. Extending this hypothesis,
Labouvie-Vief and Marquez (2004) propose that dysregulated strong emotional activation
results in ―degradation‖ of complex representations (see also Labouvie-Vief, 2003). The
causal chain can go either way: Higher degree of order or complexity in representation may
either contribute to or result from regulated activation of emotion. Conversely, loss of
complexity in representation may either contribute to or result from dysregulated activation.
For instance, Zinken, Sundararajan, Butler and Skinner (2006, August) found a positive
correlation between anxiety/depression and degradation of syntax in the writings of the
clinical population.
Cast in the information and energy framework, the language and health connection
becomes a testable equation: language representations that are informationally complex can
be expected to be associated with the energy efficient cool system, whereas loss of
complexity in language representation, the energy costly hot system. To test this hypothesis,
we need to be able to measure the degree of complexity in language representations.
Complexity in language representation can be understood in terms of Shannon‘s ideal
code (Campbell, 1982), which consists of an optimal blend of two opposite tendencies of
information--variety and accuracy—resulting in the notion of redundancy as reliable variety.
But Shannon‘s ideal code lacks specificity. For an algorithm of complexity that maps out
explicitly the dynamisms involved, we turn to the semiotic notion of the sign, according to
Charles Peirce.
PEIRCEAN SEMIOTICS
What is a sign? ―A sign is an object which stands for another to some mind,‖ says Peirce
(cited in Fisch, 1982, Vol. 3, p. 66). Central to Peircean semiotics is the claim that a
representation is always representation to a mind, which generates interpretations referred to
as ―interpretant.‖
Louise Sundararajan, Chulmin Kim, Martina Reynolds et al.
68
Thus a sign consists of three elements: object, the sign proper, and interpretant. The
interaction among these terms constitutes sign relation, which is a complex structure known
as the triadic circuitry of the sign.
The Triadic Circuitry of the Sign
A triadic circuitry is implied in the following definition of the sign as ―Anything which
determines something else (its interpretant) to refer to an object to which itself refers (its
object) in the same way, the interpretant becoming in turn a sign, and so on ad infinitum‖
(Peirce in Hoopes, 1991, p. 239). This can be illustrated by the sunflower.
A sunflower is not a real sign, but a proto-sign or representamen: ―If a sunflower, in
turning towards the sun [object], becomes by that very act fully capable . . . of reproducing a
sunflower [interpretant] which turns in precisely corresponding ways toward the sun, and of
doing so with the same reproductive power, the sunflower would become a Representamen
[proto-sign] of the sun‖ (Peirce, 1961, 1: 274). This formulation seems to have anticipated
what we know of DNA today. But the function of DNA is only a proto-sign. To be a full
fledged sign, the interpretant has to be a mental, rather than a biological process.
Nevertheless, the sunflower scenario has summed up the basic triadic structure of the sign, in
which the relation between interpretant and object is an equivalent translation of the original
sign-object relation. The Peircean notion of equivalent representation (Parmentier, 1994) is
compatible with Shannon‘s notion of redundancy as reliable (equivalent) variety
(translations). But reliable variety in information or equivalent representation in signs is an
achievement, not a given as is the case with DNA, because it is a process that entails a
dynamic integration of opposites.
The notion of integration in representation (Bucci, 1995; Teasdale and Barnard, 1993),
and health (Krystal, 1988) has had a long tradition in psychology. The unique contribution of
Charles Peirce (Hoopes, 1991) lies in making it clear that integration is a far more dynamic
process than simply the combination of opposites. Integration is best understood in terms of
complexity, which in dynamical systems theory is characterized by bipolar feedback (Sabelli,
2005). According to Sabelli (2005), a bipolar (both positive and negative) feedback which
generates information is characterized by coexistence or alternation of synergy and
antagonism.
The bipolar feedback in signs may be understood in terms of that between two opposite
tendencies of information—accuracy and variety. This interplay of accuracy and variety is
manifest in Peirce‘s claim that the semiotic process involves ―two infinite series, the one back
toward the object, the other forward toward the interpretant‖ (Peirce cited in Parmentier,
1994, p. 10). Parmentier (1994) explains:
. . . the sign relation is constituted by the interlocking of a vector of representation
pointing from the sign and interpretant toward the object and a vector of determination
pointing from the object toward both sign and interpretant. (p. 25)
These two movements of the sign--one feeding forward generating an infinite series of
interpretants; the other feeding backward pointing toward the object--can be graphically
illustrated:
Language, Emotion, and Health
69
Figure 1a. The triadic circuitry of the sign.
In the feeding forward movement, the sign gives rise to the interpretant, which in turn
acts like a sign to influence the next interpretant, ad infinitum. The feeding back loop is
referred to by Wiley (1994) as a ―reflexive undertow‖ (p. 27), which is manifest in the
reentrant loops from the interpretant to the sign and the object. Corresponding to these two
movements are two important functions of the sign: effector and sensor.
The effector function of the sign serves the purpose of variety by generating a potentially
endless series of interpretations. By contrast, the sensor function of the sign monitors the
accuracy of representation. This is done by the reflexive feedback loop that makes sure that
the relation between interpretant and object is an equivalent translation of the original sign-
object relation. These two movements of the sign that serve the two fundamental
requirements of information--variety and accuracy—are hierarchically structured: The
feeding forward movement is ―upshifting‖; the reflexive loop is ―downshifting‖ (Lee,1997,
pp. 131-132).
The upshifting movement to a symbolic level is experience distant, as the interpretant is
one step removed from the object--a price it pays to capitalize on the generation of variety
through interpretations. The downshifting movement is experience near, as it privileges
fidelity to experience at the risk of rigidity or frugality in cognitive elaborations. Optimal sign
function requires proper coordination or integration of these opposing tendencies. But
integration may fail. For instance, the upshifting movement of interpretation may generate an
increasingly experience distant interpretants, when the process becomes de-coupled from the
reflexive movement back to experience. This is illustrated in the default processing of Figure
1b.
Note. In bold: default processing.
Figure 1b. Lack of integration in default information processing.
Cognate Ideas in Psychology
Complexity as modeled by the triadic circuitry of the sign is consistent with Don
Tucker‘s (2007) core to shell formulation of the neural structure, which is summed up
succinctly by Johnson (2007):
Object Sign Interpretant
Object 1
Figure 1a: The Triadic Circuitry of the Sign.
Object Sign Interpretant 1 Interpretant 2 Interpretant 3
Object 1
Figure 1b: Hierarchical structure of information processing
Note. In bold: default processing.
Louise Sundararajan, Chulmin Kim, Martina Reynolds et al.
70
The limbic core, with its dense interconnections and emotional valences, would present
us with a holistic, feeling-rich, emotionally nuanced grasp of a situation. The more modular
and highly differentiated sensory and motor regions of the shell (cortical) structure would
permit the discrimination and differentiation that we call conceptualization. (pp. 100-101).
The integration between the subsymbolic limbic core and the symbolic-cortical systems
is referred to by Tucker (2007) as a ―vertical integration‖ which is defined as a ―recursive
processing‖ (p. 223) that consists of movements in two opposite directions: limbifugal and
limbipetal.
Limbifugal movement refers to Core to Shell connection: This is the feedforward
movement toward increasing differentiation into specific and concrete forms.
Limbipetal movement refers to Shell to Core connection: This is the feedback, reentrant
loop toward integration and self-modification.
Together, limbifugal and limbipetal movements constitute one cycle of the recursive
processing referred to as vertical integration: The result of neural network patterns traversing
in both directions is the emergence of meaning. The connection between the two systems is
not necessarily smooth and automatic. As Tucker (2007) points out, the relationship between
the two systems is dialectical (as is characteristic of bipolar feedback):
The consolidation process across the linked networks from shell to core is dialectical in
that an inherent opposition of structural forms—fused versus separated—exists between the
core and shell. . . . Each wave in the cycle of abstraction traverses this conflict in some way.
In those rare optimal instances of the human mind, the dialectic is extended, recursive, and
progressive. (pp. 224-225)
The notion of vertical integration is consistent with the hierarchical, staged model of
memory (Conway and Pleydell-Pearce, 2000), which suggests that a retrieval strategy of
―moving across rather than down the memory hierarchy‖ (Williams, Barnhofer, Crane,
Hermans, Raes, Watkins, and Dalgleish, 2007, p. 136) constitutes failure in integration, as is
illustrated in the default functioning in Figure 1b. A case in point is the truncated search of
overgeneral retrieval of autobiographical memory that individuals suffering from depression,
PTSD, or related disorders are found to be especially prone to. These individuals tend to
capitalize on categorical memories (Birthdays make me happy) at the expense of event
specific details (contextual details of a particular birthday). Cast in the framework of the
triadic circuitry of the sign, the truncated search strategy of these individuals is a case of the
lack of integration between the two movements of the sign: the experience-distant symbolic
mode characteristic of the interpretant is running on overdrive, un-constrained by the
reflexive undertow (Wiley, 1994) back to experience.
Theory Based Predictions on Language and Health
To recapitulate, the triadic circuitry of the sign consists of a bipolar feedback (Sabelli,
2005), characterized by mutual synergy and antagonism, between two opposite movements of
the sign--generation of variety through symbolic, experience distant interpretation, on the one
Language, Emotion, and Health
71
hand; and monitoring of accuracy through the reflexive undertow (Wiley, 1994) back to
experience, on the other. The result of this dynamic integration of two opposite movements is
a fully developed sign, which according to Peirce consists of three modes of representation—
icon, symbol, and index--each contributing uniquely to the overall efficiency of the sign
(Deacon,1997). The icon embodies a relationship of contiguity between the representation
and its object to ensure fidelity in representation; the symbol is one step removed from the
object of representation to facilitate further elaboration through interpretations; and the index
is a reference loop that counterbalances the abstract tendency of the symbol by calling
attention to the object of representation. Equipped with these three modes of representation, a
fully developed sign is therefore capable of integrating its multiple functions of
representation--the concrete expression of experience (a function of the icon), understanding
through elaboration and interpretation (a function of the symbol), and validation of subjective
experience (the indexical function that calls attention to the object of representation).
Consistent with this model of representation is Clore and colleagues‘ (Clore, Ortony, and
Foss,1987) structural definition of bona fide emotion concepts in terms of three referential
foci—internal, mental, and affective: ―the best examples of emotion words would be ones that
refer to internal (as opposed to external) conditions, those that refer to mental (as opposed to
physical) conditions, and those that have a significant focus on affect‖ (p. 752). The affective
expression is iconic; the mental representation is symbolic; the internal focus is reflexive.
Integration of these multiple referential foci in bona fide emotion concepts approximates a
fully developed sign, according to Peirce, or the ideal code, according to Shannon, or to give
the screw another turn, what is referred to by Pennebaker (1989) as high level thinking.
However, as indicated by the foregoing analysis, optimal representations are an achievement
rather than a given.
Since optimal representation of experience is dialectic at its very core, it requires the
integration of two antithetical types of language use--one experience near (A), the other
experience distant (B). The mutual constraint, characteristic of bipolar feedback, of A and B
results in proper distance from experience (Scheff, 1979), which consists of the following
types of language use:
A. attention to affect;
B. facilitative mental distance from experience.
Less than optimal representations of self and emotions are hypothesized to be
symptomatic of a lack of integration of the two movements of the sign, with each going to
extreme due to lack of mutual restraint, resulting in:
C. under-distance from experience;
D. over-distance from experience.
Based on the foregoing analysis of the inverse relationship between complexity in
information and energy cost, the shifting balance between cool and hot systems in different
types of language use is predicted to be as follows: Informationally complex, optimal
representations of self and emotions (A and B) are predicted to be dominated by the cool
system; loss of complexity in less than optimal representations of self and emotions (C and D)
Louise Sundararajan, Chulmin Kim, Martina Reynolds et al.
72
is associated with activation of the hot system. We hypothesize that when the cool system is
in dominance, the hot system is neutralized, resulting in the following effects:
a) Health benefits.
b) Information for free: An efficient sign is one that processes information at minimum
energy cost. With complete neutralization of the hot system, emotion can be
processed as simply information, with minimum arousal. This hypothesized low
energy cost of the cool system is consistent with Damasio‘s (1999) notion of feelings
as involving the ―as if body loop,‖ which bypasses the body proper, partially or
entirely, a mechanism that ―saves both time and energy‖ (p. 281).
By contrast, the hot system is hypothesized to have varying degrees of health costs,
depending on its regulation by the cool system:
a) Partially neutralized activation: The hot system is neutralized to some extent by the
cool system, resulting in a reduction of activation.
b) Regulated activation: Activation is evident, but with benefit outweighing the cost.
This hypothesis is consistent with the notion of integration of hot and cool systems
according to Metcalfe and Mischel (1999), who claim that the former can be
harnessed in the service of the latter, for instance regulated activation of the hot
system may boost working memory.
c) Dysregulated activation: When the hot system dominates and the cool system is
inhibited, resulting in health cost.
SSWC: TOWARD A TAXONOMY OF SELF AND EMOTIONS
Based on the above formulation of the sign, a taxonomy of verbal expressions of self and
emotions has been implemented by a pattern matching language analysis program, SSWC
(Sundararajan-Schubert Word Count)(for a study of construct and external validity, see
Sundararajan and Schubert, 2005). This program consists of fifteen categories of verbal
expressions of self and emotions. The reason why representations of the self are included in
our taxonomy of affective lexicon is because any emotion expression invariably involves the
self (Lambie and Marcell, 2002).
We further propose that representations of the self are not confined to the first person
pronoun ―I‖ but extend to pronouns in general (it, they, you, and so on). This assumption is
supported by the neuroimaging results which showed that self-relatedness evaluation involves
a wide neural network, which relates any represented object to the representing subject
(Legrand and Ruby, 2009).
For categorization of affective lexicon, we have consulted Lane (1991), and Clore, et al.
(1987). The fifteen categories of self and emotions are grouped into the above mentioned four
types of language use:
Language, Emotion, and Health
73
Optimal Representation of Emotion
Categories of language use that constitute optimal representation are expected to be under
the sway of the cool system.
A. Attention to Affect
The following categories of expressions show an integration of three foci of reference in
representation: mental, internal, and affective.
Affect Focal (happy/sad): These are the bona fide emotion terms. Dictionary for this
category is based primarily on the affect-focal terms of Clore et al. (1987).
Valence focus (miserable/pleasant): This category indexes the valence dimension of
emotions. Dictionary for this category consists of the word list of pleasant and unpleasant
affect in Barrett and Russell (1998). Also included are word lists with highest scores on the
Depth and Evaluation dimensions in Averill (1975, p. 17).
B. Facilitative Mental Distance from Experience
The basic premise of the Peircean semiotics is that the relationship between any two
terms is always mediated by a third term. The inclusion of the third term--the other-- is what
creates a mental distance from experience, which is necessary for the proper regulation of
emotions. An element of the other is present in all of the following categories:
Detached Self (someone, they): This category is an index of the third person perspective,
which reflects a detached intentional stance toward personal experiences.
Reflexive Self (ourselves, itself): The reflexive self has a triadic structure of self-other-
self, which is a self to self recursiveness looping through the other. This triadic self-other-self
recursivity entails the integration of two lower dimensional structures of self representation:
self as identity (―I‖ and ―me‖) and self as other (―they‖).
This category consists of two types of expressions: One is expressions of self-
referentiality such as ―itself.‖ ―Myself‖ however is excluded from this category, because
representation of the self in this category is not an atomic self (an ―I‖), so much as an
extended self that includes the other, a ―we‖ (Wiley, 1994). The category of Reflexive Self,
therefore, includes expressions--such as ―our own,‖ or ―each other‖—that evince a looping of
the self through the other, resulting in an extended self.
External Attribution (sexy, wonderful): This category is based primarily on terms referred
to by Clore et al. (1987) as ―external conditions.‖ The referential focus of these words is on
the external attributions of the emotional states, a mode of processing which is hypothesized
to constitute a facilitative mental distance that reduces the intensity of affect.
Less Than Optimal Representation of Emotion
Categories of language use that constitute less than optimal representation are expected to
be under the sway of the hot system.
Louise Sundararajan, Chulmin Kim, Martina Reynolds et al.
74
C. Under-Distance from Experience
This refers to a state of immersion in experience (Lambie and Marcel, 2002), resulting in
representations that are deficient in mental reflection (Frijda, 2007).
Affected Self (me, ―making me . . .‖): This category indexes the perception of the self not
as doer, but as being done to. This passive self is enmeshed in the experience, lacking the
mental distance for reflection.
Somatic (headache): The dictionary for this category is based primarily on the word list
of ―physical and bodily states‖ in Clore et al. (1987). Words under this category are instances
where the referential focus is primarily on the physical rather than mental or psychological
conditions. Such representations are what Peirce refers to as indices. ―An index is a sign
which would, at once, lose the character which makes it a sign if its Object were removed, but
would not lose that character if there were no Interpretant‖ (Peirce cited in Hoopes, 1991, p.
239). Peirce gives the example of the bullet-hole as a sign of a gun shot to show that indices
have direct physical connections to the signified, a connection independent of an Interpretant:
―for without the shot there would have been no hole; but there is a hole there, whether
anybody has the sense to attribute it to a shot or not‖ (Peirce cited in Hoopes,1991, p. 240).
By the same token, the assumption behind somatic complaints such as ―hungry‖ is that they
are indications of some physiological change, a condition that ―truly‖ exists, regardless of
whether it is recognized/interpreted as such or not. From the Peircean perspective, this alleged
independence from the Interpretant explains why words denoting somatic concerns are
usually deficient in their impetus for symbolic elaborations.
Violent Words (kill, rape, swear words): This category is composed of words of violent
action and obscenities, which may be considered verbally acting out behaviors. This type of
language use is deficient in mental reflection.
Emotion as action (love/hate, used as verb): Dictionary for this category consists of
twenty-one ―noncausative verbs‖ (such as love, hate, used as verb, in active, not passive,
voice) from Clore et al. (1987, pp. 763-765). When emotion is represented as action, self
reflexivity is missing. Peirce has noted that when a child wants to move a table, he is likely to
be so absorbed in what he wills as to be oblivious to himself: ―Does he think of himself as
desiring, or only of the table as fit to be moved?‖ (1931-58, Vol. 5, paragraph 230). The same
applies to expression of emotion as action. For instance, in ―I hate him,‖ the emoter is not
self-reflexive, as his or her attention is absorbed by the qualities of the person as ―fit‖ to be
hated, no less than a table as ―fit to be moved.‖ Frijda (2005) makes a similar observation
concerning infatuation as an instance of the first-order experience, where one is enthralled by
a person‘s attractiveness, and considers ―I love her‖ as an objective fact. Frijda goes on to say
that in the immersed consciousness of the first-order experience, no subjectivity, no reference
to the self is involved.
Suffering (devastated, traumatized): This category consists of verbs in passive voice--
words that designate the extreme pole of the victim stance, which suggests a lack of
psychological distance from the experience.
High Activation (excited, nervous): Words in this category are hypothesized to be
indications of direct activation, unmitigated by any mental distance from experience.
Dictionary for this category is based primarily on the word list of activated affect in Barrett
and Russell (1998).
Language, Emotion, and Health
75
D. Over-Distance from Experience
Over-distance is defined as a defensive mental distance that compromises the fidelity of
emotion representations by limiting the access to or the scope of experience. It may be part
and parcel of the evaluative emotional processing, which has been found to be associated with
reactivity (Low, Stanton, and Bower, 2008).
Focal Self (I, myself, my own): The linguistic use of ―I‖ is indicative of focal attention to
the self system, which is referred to by Kihlstrom, et al. as ―an organized knowledge structure
that stores what one knows about oneself. This would include semantic knowledge about
one‘s physical and personality attributes, social status, and the like‖ (Kihlstrom, Mulvaney,
Tobias, and Tobis, 2000, P. 67, note 1). Antithetical to the experiencing self, this
representation of the self is hypothesized to entail direct access to semantic memory (Forgas,
2001), which detracts resources from online processing of emotional information.
Denial (―doesn‘t bother me‖): This category is composed of expressions that indicate a
distancing strategy that minimizes or represses the emotional impact of the experience.
Affect Non-Focal (cry, understanding): Dictionary for this category is based primarily on
the affect non-focal terms in Clore et al. (1987). It consists of representations in which the
referential focus has shifted from affectivity to cognitive and behavioral components of the
experience. Also included in this category are clichés, such as ―depressed‖ (Lane, 1991). It is
hypothesized that a preponderance of Affect Non-Focal terms are the result of heuristic
information search strategies and motivated processing as an attempt to control and limit the
scope and impact of one‘s affective experience (Forgas, 2001). This hypothesis finds
supportive evidence in one study (Sundararajan and Schubert, 2005), in which factor analysis
revealed high loading of Affect Non-Focal terms on a factor called Emotional Management.
Low Activation (bored, drowsy): Dictionary for this category is based primarily on the
word list of deactivated state in Barrett and Russell (1998). High frequency use of this type of
expressions could be an indication of apathy or withdrawal as a result of the cool system
coming de-coupled from the hot system, according to Metcalfe and Mischel (1999).
All the above categories are presented in terms of percentage, out of the total word count,
of words that fall into a specific category. In addition, SSWC computes three global
categories:
Word Count: The raw score that serves as an index of the length of the text.
Core Affect: The percentage, out of word count, of the sum total of words that fall into the
following categories: Valence Focus, High Activation, and Low Activation.
Expressions of Self and Emotions (E): The percentage of the sum total of words used in all
the SSWC categories minus Core Affect. Here we are following the advice of Russell
(2003) to treat E and Core Affect separately.
EMPIRICAL STUDIES
To test this semiotic model of language and health, we re-analyzed two empirical studies
of the writing cure. According to Metcalfe and Mischel (1999), the balance between the cool
and hot systems is determined by stress and developmental phase, in addition to the
individual‘s self-regulatory dynamics. The first study which used provoked stress to measure
Louise Sundararajan, Chulmin Kim, Martina Reynolds et al.
76
reactivity can shed some light on the shifting balance between the two systems in the
expressive writing of adults. The second study which consisted of children‘s writing
addresses the developmental implications for our proposed model of language and health.
Group difference has been the main focus in studies of the writing cure, but this approach
masks the individual differences in language use, as Fivush, Marin, Crawford, Reynolds, and
Brewin (2007) point out rightly. To go beyond this well beaten path, we selected for our
analysis two studies which had null results in terms of group differences—both control and
expressive writing groups improved at follow up. The null results help to cast the issue of
language and health into one urgent and sharply focused question: What good does expressive
writing do? As Fivush, et al. (2007) took the individual differences approach to follow up on
this question (see Study 2 below), we go one step further by situating the question of
language use in the context of the information and energy trade off. From this perspective, the
instruction set of expressive writing that urges the participants to write about their deepest
thoughts and feelings can be expected to promote complexity in information, or in
Pennebaker‘s (1989) term, high level thinking, and thereby enhance the cool system of
emotion. This hypothesis is put to test in the following two studies.
Study 1
Study 1 is a reanalysis of an unpublished study (Graybeal, 2004, Study 2), which
recruited 86 college undergraduates whose parents were divorced and who were randomly
assigned to a control or experimental group (n=43 each). The former was instructed to write--
on two occasions, 30 minutes each--about time management; the latter, their deepest thoughts
and feeling about their parents‘ divorce. Participants were also interviewed about the most
upsetting aspects of their parents‘ divorce, both before and after writing, in order to assess
their reactivity to provoked stress. The hypothesis was that the Expressive Writing group,
relative to the controls, would show decreased reactivity to stress at the final interview, one
month post writing. This was not supported empirically. Results showed that both groups
improved after the writing exercise--they were less distressed, improved their mean
performance on the working memory task, and exhibited fewer psychological symptoms. To
shed some light on this conundrum, we used SSWC to re-analyze the data.
Outcome Measures
To measure the participants‘ reactivity to provoked stress, a comprehensive
battery of tests were used in the original study, including measures of physiological
arousal (such as heart rate, skin conductance, and blood oxygen level), self reports of
emotional upset (such as questionnaires and mood scales), measures of physical and
psychological health (self reports of illness, and symptom checklist), and measure of
cognition (working memory tests). From this battery of tests, the following measures were
selected because of their robustness (Graybeal, 2004):
max HRd (maximum level of heart rate difference from the baseline),
Language, Emotion, and Health
77
restact (Degrees to which activities been restricted due to illness in last 2 weeks),
SUDSpk (Subjective Units of Distress, peak score),
sick2m (frequency of being sick for last 2 months),
WM (working memory).
All these measures were taken during or at the end of the interviews, pre- and post-
writing.
Besides the above measures used in the original study, we added physician‘s visits from
Health Center data, which are coded as follows:
Dr2m (frequency of health center visits within 2 months post writing).
Dr12m (frequency of health center visits post writing, two months to a year).
RESULTS
Over View
Language use: As shown in Figure 2, the two groups wrote very differently, as measured
by the SSWC variables.
Word Count: No significant difference was found between the two groups, neither in
terms of the sum total of words produced across the writing days (Figure 3, lower left panel),
nor in terms of any significant correlation between length of text and outcome measures
(Figure 4a). Furthermore, judging by the difference score of word count between the writing
days, both groups wrote less on the second day of writing (Figure 3, upper left panel).
Note. *p<.05, **p<.01, ***p<.001. Ns=Not significant at .05. In=Invalid comparison due to low
baseline.
Figure 2. Study 1 (N=86), group comparison, based on weighted mean across writing days, on variables
of SSWC (Sundararajan-Schubert Word Count), by writing task.
SSWC Variables
SSWC Variables (with tokens)
Expressive Writing group
Mean (SD)
Control group
Mean (SD)
Significance
Significance
Attention to Affect
Affect Focal (happy/sad) 1.33 (0.45) 0.33 (0.24) ***
Valence Focus (miserable) 0.30 (0.18) 0.03 (0.05) ***
Facilitative Distance from Experience
Reflexive Self (ourselves) 0.47 (0.19) 0.19 (0.20) ***
Detached Self (someone) 1.70 (0.80) 0.44 (0.30) ***
External Attribution (sexy) 1.22 (0.43) 0.60 (0.30) ***
Under Distance from Experience
Affected Self (me) 2.49 (0.82) 1.33 (0.64) ***
Violent Words (swear) 0.07 (0.08) 0.01 (0.04) ***
Somatic (headache) 0.02 (0.05) 0.14 (0.11) ***
Suffering (traumatized) 0.02 (0.03) 0.00 (0.01) In
Emotion as Action (love/hate, used as verb) 0.46 (0.26) 0.14 (0.16) ***
High Activation (excited) 0.09 (0.09) 0.06 (0.18) *
Over Distance from Experience
Focal Self (I, myself, my own) 6.29 (1.03) 7.18 (1.30) ***
Affect Non-Focal (cry, understand) 0.72 (0.33) 0.36 (0.22) ***
Denial (“doesn’t bother me”) 0.05 (0.06) 0.01 (0.03) ***
Low Activation (bored) 0.03 (0.05) 0.05 (0.06) Ns
Figure 2: Study 1 (N=86), group comparison, based on weighted mean across writing days, on
variables of SSWC (Sundararajan-Schubert Word Count), by writing task.
Note. *p<.05, **p<.01, ***p<.001. Ns=Not significant at .05. In=Invalid comparison due to low baseline.
Louise Sundararajan, Chulmin Kim, Martina Reynolds et al.
78
Note. diff=difference score between the writing days (Day2–Day1). sum=weighted mean across the
writing days (Day1+Day2). *p<.05, **p<.01, ***p<.001.
Figure 3. Study 1, group differences in global variables of SSWC (Sundararajan-Schubert Word
Count), by writing task.
Expressions of Self and Emotions (E): As shown in Figure 3, the Expressive Writing
group had significantly (p<.001) more output of E, in terms of weighted mean across writing
days, than the Control group. The two groups also differed significantly in terms of difference
score between the writing days, with the Expressive Writing group increased, while the
control group decreased, their output of E on the second day of writing.
As shown in Figure 4a, both groups benefited from higher percentage of E: The higher
percentage of using E, the less likely for the individual in the Control group (r=-.36, p<.05) to
show heart rate increase at the final interview, indicating decreased reactivity. Not so for the
Expressive Writing group, in which higher percentage of E was associated with increase in
heart rate at follow up (r=.41, p<.05). However, the salutary effect of self and emotion
expressions (E) was also evident for the Expressive Writing group, when difference score
between the writing days is examined. Higher percentage of E on the second day of writing
was correlated with a decrease in self reported stress (SUDS, r=-.40, p<.05) at the final
interview, and a decrease in the frequency of health center visits (r=-.43, p<.01) from 2
months post writing to a year. This pattern is consistent with the observation (Pennebaker and
Beall, 1986) that expressive writing increases stress in the short term, but produces health
benefits in long term.
Core Affect: As shown in Figure 3 (far right, lower panel), the Expressive Writing group
had significantly (p<.001) more output of Core Affect than the Control group. The two
groups also differed significantly (p<.05) in terms of difference score between the writing
days, with the Expressive Writing group increasing, while the control group decreasing, their
output of Core Affect on the second day of writing.
-150
-75
0
75
150
Word Count (diff)
MEAN
-0.50
-0.25
0.00
0.25
0.50
Representations of Self and
Emotions (diff)
***
MEAN
-0.12
-0.06
0.00
0.06
0.12
Core Affect (diff)
Control
Expressive
Writing
*
MEAN
0
500
1,000
1,500
2,000
Word Count (sum)
MEAN
0
4
8
12
16
Representations of Self and
Emotions (sum)
***
MEAN
0.00
0.12
0.24
0.36
0.48
Core Affect (sum)
Control
Expressive
Writing
***
MEAN
Figure 3: Study 1, group differences in global variables of SSWC (Sundararajan-Schubert Word Count), by writing task.
Note. diff=difference score between the writing days (Day2–Day1). sum=weighted mean across the writing days
(Day1+Day2). *p<.05, **p<.01, ***p<.001.
Language, Emotion, and Health
79
Note. diff=difference score between the writing days (Day2–Day1). sum=weighted mean across the
writing days (Day1+Day2). SUDSpk=subjective units of distress, peak score. maxHRd=difference
in maximum level of heart rate from baseline. Dr2m=frequency of doctor‘s visit within 2 months
post writing. Dr12m=frequency of doctor‘s visit post writing, two months to a year. *p<.05,
**p<.01, ***p<.001.
Figure 4a. Study 1, partial correlations, between outcome measures and global variables of SSWC
(Sundararajan-Schubert Word Count), for (top) Control and (bottom) Expressive Writing groups.
As shown in Figure 4a, weighted mean across writing days had no significant correlations
with outcome measures for both groups. However, for the Control group only, change scores
between the writing days were positively correlated with the frequency of health center visits
in the two months after writing (r=.40, p<.05), as well as two months to a year post writing
(r=.34, p<.05).
Results for components of optimal emotion representation—attention to affect and
facilitative distance from experience--are shown in Figure 4b.
Attention to Affect
Affect Focal (happy/sad): As Figure 4b shows, no significant correlations of this variable
were found with outcome measures for both groups.
Valence focus (miserable/pleasant): As Figure 4b shows, no significant correlations of
this variable were found with outcome measures for the Expressive Writing group. For the
Control Group, increased use of Valence focus from day 1 to day 2 of writing was related to
decreased reactivity at the final interview, as measured by self-reported stress (SUDS, r=-.40,
p<.05).
maxHRd
-0.362*
0.338*
Dr12m
0.397*
Dr2m
-0.6
-0.3
0.0
0.3
0.6
WC(diff)
WC(sum)
Representations of Self
and Emotions (diff)
Representations of Self
and Emotions (sum)
Core Affect (diff)
Core Affect (sum)
Control
MEAN
SUDSpk
-0.401*
0.405*
maxHRd
Dr12m
-0.432**
-0.6
-0.3
0.0
0.3
0.6
WC(diff)
WC(sum)
Representations of Self
and Emotions (diff)
Representations of Self
and Emotions (sum)
Core Affect (diff)
Core Affect (sum)
Expressive
Writing
MEAN
Figure 4a: Study 1, partial correlations, between outcome measures and global variables of SSWC
(Sundararajan-Schubert Word Count), for (top) Control and (bottom) Expressive Writing groups.
Note. diff=difference score between the writing days (Day2–Day1). sum=weighted mean across the writing
days (Day1+Day2). SUDSpk= subjective units of distress, peak score. maxHRd=difference in maximum level
of heart rate from baseline. Dr2m=frequency of doctor’s visit within 2 months post writing.
Dr12m=frequency of doctor’s visit post writing, two months to a year. *p<.05, **p<.01, ***p<.001.
Louise Sundararajan, Chulmin Kim, Martina Reynolds et al.
80
Note. diff=difference score between the writing days (Day2–Day1). sum=weighted mean across the
writing days (Day1+Day2). SUDSpk=subjective units of distress, peak score. sick2m=frequency of
being sick for the last 2 months. Dr12m=frequency of doctor‘s visit post writing, two months to a
year. *p<.05, **p<.01, ***p<.001.
Figure 4b. Study 1, partial correlations, between outcome measures and Components of Optimal
Emotion Representation in SSWC (Sundararajan-Schubert Word Count), for (top) Control and (bottom)
Expressive Writing groups.
Facilitative Distance from Experience
Reflexive Self (ourselves, itself): No significant correlations of this variable emerged for
the Expressive Writing group. For the Control Group, increase in the sum total of this
variable across the writing days was associated with both decreased self-reported frequency
of being sick for the last 2 months post writing (r=-.36, p<.05), and decreased health center
visits (r=-.37, p<.05), two months to a year post writing.
Under Distance from Experience
Affected Self (me, ―making me . . .‖): For the Expressive Writing group, increase in the
sum total of this variable across the writing days was associated with more reactivity at the
final interview, as evidenced by a positive correlation with heart rate (r=.49, p<.01).
However, progressive increase of this type of expression from day1 to day 2 of writing was
negatively correlated (r=-.39, p<.05) with self-reported frequency of being sick for the last 2
SUDSpk
-0.399*
-0.6
-0.3
0.0
0.3
0.6
Affect
Focal
(diff)
Affect
Focal
(sum)
Valence
Focus
(diff)
Valence
Focus
(sum)
MEAN
sick2m
-0.362*
Dr12m
-0.372*
-0.6
-0.3
0.0
0.3
0.6
Reflexive
Self(diff)
Reflexive
Self(sum)
Detached
Self(diff)
Detached
Self(sum)
External
Attribution(diff)
External
Attribution(sum)
Control
MEAN
-0.6
-0.3
0.0
0.3
0.6
Affect
Focal
(diff)
Affect
Focal
(sum)
Valence
Focus
(diff)
Valence
Focus
(sum)
MEAN
-0.6
-0.3
0.0
0.3
0.6
Reflexive
Self(diff)
Reflexive
Self(sum)
Detached
Self(diff)
Detached
Self(sum)
External
Attribution(diff)
External
Attribution(sum)
Expressive
Writing
MEAN
Attention to affect Facilitative distan ce from experience
Figure 4b: Study 1, partial correlations, between outcome measures and Components of Optimal
Emotion Representation in SSWC (Sundararajan-Schubert Word Count), for (top) Control and (bottom)
Expressive Writing groups.
Note. diff=difference score between the writing days (Day2–Day1). sum=weighted mean across the
writing days (Day1+Day2). SUDSpk= subjective units of distress, peak score. sick2m= frequency of being
sick for the last 2 months. Dr12m=frequency of doctor’s visit post writing, two months to a year. *p<.05,
**p<.01, ***p<.001.
Language, Emotion, and Health
81
months. This category had no significant correlations with outcome measures for the Control
group.
Note. diff=difference score between the writing days (Day2–Day1). sum=weighted mean across the
writing days (Day1+Day2). maxHRd=difference in maximum level of heart rate from baseline.
restact=Degrees to which activities been restricted due to illness in last 2 weeks. SUDSpk=
subjective units of distress, peak score. sick2m=frequency of being sick for the last 2 months.
Dr12m=frequency of doctor‘s visit post writing, two months to a year. WM=working memory
(reversed scale such that improvement is shown as downward bar; impairment as upward bar).
*p<.05, **p<.01, ***p<.001.
Figure 4c. Study 1, partial correlations, between outcome measures and Under Distance categories of
SSWC (Sundararajan-Schubert Word Count), for (top) Control and (bottom) Expressive Writing
groups.
Violent Words (kill, rape, swear words): For the Expressive Writing group, an increase,
from day 1 to day 2 of writing, of this variable was positively correlated with improvement in
working memory (r=.34, p<.05), post writing. No significant correlation of this variable was
found for the Control group.
High Activation (excited, nervous): For the Expressive Writing group, increase in the sum
total across writing days of this variable was strongly and positively correlated (r=.48, p<.01)
with self-reported extent to which activities had been restricted due to illness in the last 2
weeks. On the other hand, it was not without health benefits, as evidenced by a negative
correlation (r=-.36, p<.05) with the frequency of health center visits two months to a year post
writing. For the Control group, increase in the sum total across writing days of this variable
was positively correlated with increased heart rate (r=.43, p<.05). Progressive increase, from
day 1 to day 2 of writing, of expressions of high activation was positively correlated with the
health costs of increased heart rate (r=.35, p<.05), and self reported stress (r=.34, p<.05), on
the one hand, and with the health benefit of improved working memory post writing (r=.35,
p<.05), on the other.
0.350*
maxHRd
0.428*
maxHRd
0.343*
SUDSpk
-0.6
-0.3
0.0
0.3
0.6
MEAN
-0.6
-0.3
0.0
0.3
0.6
Control
WM
0.350*
MEAN
sick2m
-0.385*
0.489**
maxHRd
0.482**
restact
Dr12m
-0.362*
-0.6
-0.3
0.0
0.3
0.6
MEAN
-0.6
-0.3
0.0
0.3
0.6
Expressive
Writing
WM
0.339*
MEAN
Under distance from experience
Figure 4c: Study 1, partial correlations, between outcome measures and Under Distance categories of SSWC
(Sundararajan-Schubert Word Count), for (top) Control and (bottom) Expressive Writing groups.
Note. diff=difference score between the writing days (Day2–Day1). sum=weighted mean across the writing days
(Day1+Day2). maxHRd= difference in maximum level of heart rate from baseline. restact=Degrees to which activities been
restricted due to illness in last 2 weeks. SUDSpk= subjective units of distress, peak score. sick2m=frequency of being sick for
the last 2 months. Dr12m=frequency of doctor’s visit post writing, two months to a year. WM=working memory (reversed
scale such that improvement is shown as downward bar; i mpairment as upward bar). *p<.05,**p< .01, ***p<.001.
Louise Sundararajan, Chulmin Kim, Martina Reynolds et al.
82
Over Distance from Experience
Focal Self (I, myself, my own): For the Expressive Writing group, the more one made use
of this variable, as evidenced by weighted mean across writing days, the less likely was one to
perform well on working memory (r=-.38, p<.05) at follow up. This memory impairment
could be due to the high self focus characteristic of rumination (Watkins and Teasdale, 2001).
No significant correlation of this variable with outcome measures was found for the Control
group, a writing condition which probably did not invite ruminative reflections as much as the
expressive writing condition that focused on stressful autobiographical memories.
Affect Non-Focal (cry, understanding): For the Expressive Writing group, weighted mean
of the use of this variable was positively correlated with increased heart rate (r=.40, p<.05),
an indication of reactivity, at the final interview. For the Control group, progressive increase,
from day 1 to day 2 of writing, in this type of expressions was strongly and positively
correlated with an increase in self reported frequency of being sick for the past two months
post writing (r=.47, p<.01).
Note. diff=difference score between the writing days (Day2–Day1). sum=weighted mean across the
writing days (Day1+Day2). maxHRd=difference in maximum level of heart rate from baseline.
SUDSpk=subjective units of distress, peak score. sick2m=frequency of being sick for the last 2
months. Dr12m=frequency of doctor‘s visit post writing, two months to a year. Dr2m=frequency
of doctor‘s visit within 2 months post writing. WM=working memory (reversed scale such that
improvement is shown as downward bar; impairment as upward bar). *p<.05, **p<.01, ***p<.001.
Figure 4d. Study 1, partial correlations, between outcome measures and Over Distance categories of
SSWC (Sundararajan-Schubert Word Count), for (top) Control and (bottom) Expressive Writing
groups.
0.466**
sick2m
SUDSpk
-0.360*
SUDSpk
-0.505***
0.532***
Dr12m
0.334*
Dr12m
0.354*
Dr2m
-0.6
-0.3
0.0
0.3
0.6
Focal Self(diff)
Affect Non-
Focal(diff)
Affect Non-
Focal(sum)
Denial(diff)
Denial(sum)
Low Activation(diff)
Low Activation(sum)
MEAN
-0.6
-0.3
0.0
0.3
0.6
Focal
Self(sum)
Control
MEAN
0.397*
maxHRd
0.381*
maxHRd
-0.6
-0.3
0.0
0.3
0.6
Focal Self(diff)
Affect
Non-Focal(diff)
Affect Non-
Focal(sum)
Denial(diff)
Denial(sum)
Low Activation(diff)
Low Activation(sum)
MEAN
-0.6
-0.3
0.0
0.3
0.6
Focal
Self(sum)
Expressive
Writing
-0.380*
WM
MEAN
Over distance from experience
Figure 4d: Study 1, partial correlations, between outcome measures and Over Distance categories of SSWC
(Sundararajan-Schubert Word Count), for (top) C ontrol and (bottom) Expressive Writing groups.
Note. diff=difference score between the writing days (Day2–Day1). sum=weighted mean across the writing days
(Day1+Day2). maxHRd= difference in maximum level of heart rate from baseline. SUDSpk=subjective units of distress,
peak score. sick2m= frequency of being sick for the last 2 months. Dr12m=frequency of doctor’s visit post writing, two
months to a year. Dr2m=frequency of doctor’s visit within 2 months post writing. WM=working memory (reversed
scale such that improvement is shown as downward bar; impairment as upward bar). *p<.05, **p<.01, ***p<.001.
Language, Emotion, and Health
83
Denial (―doesn‘t bother me‖): For the Expressive Writing group, weighted mean of the
use of denial was positively correlated with increased heart rate (r=.38, p<.05), an indication
of reactivity, at the final interview. For the Control group, the more one used expressions of
denial, as evidenced by weighted mean, the less likely was one to report stress, as evidenced
by strong negative correlation with SUDS(r=-.51, p<.01). But the health cost of Denial comes
through in the change score, which showed that those who progressively used more denial,
from day 1 to day 2 of writing, tended to report less stress (r=-.36, p<.05) at the final
interview, but had more health center visits (r=.33, p<.05), two months to a year post writing.
This is consistent with the protocol of repression (Weinberger, 1990), which is characterized
by a combination of temporary relief of subjective stress, on the one hand, and long term
health cost, on the other.
Low Activation (bored, drowsy): No significant correlations of this variable were found
for the Expressive Writing group. For the Control group, those who used progressively more
expressions of low activation, from day 1 to day 2 of writing, tended to have more health
center visits, both within two months post writing (r=.35, p<.05) as well as two months to a
year post writing (r=.53, p<.001).
DISCUSSION
Overall, there was no major disacrepancy between the two groups in terms of the shifting
balance of cool and hot systems associated with various categories of language use. Both
groups showed higher activation with the use of less than optimal representations—in the
case of under distance categories (figure 4c), both groups were able to reap a mixture of cost
and benefit, with the latter outweighing the former as characteristic of regulated activation;
whereas in the case of over distance categories (Figure 4d), both groups evinced dysregulated
activation at follow up.
There was however a group difference in nuance attributable to instruction set. In light of
the fact that the controls did not have any guidance as to how to express their emotions,
whereas the Expressive Writing group was instructed specifically to do so, the differences
between the two writing conditions may thus fall along the divide between automatic versus
controlled processing (Philippot, Baeyens, and Douilliez, 2006)—the latter, but not the
former, can be expected to extend or reinforce the cool system. This is our tentative answer to
the question: What good does expressive writing do? Cool system effect may explain some
subtle differences in outcome between the two groups.
Consider first the sum total of E (expressions of self and emotions). For controls, higher
proportion of E was related to reduced reactivity at follow up (Figure 4a). In contrast, being
told to explicitly write about emotions might have increased sensitivity to arousal for the
Expressive Writing group, which therefore reaped a mixture of health cost and benefit post
writing, with benefit outweighing the cost—temporary increase in heart rate at the final
interview, but long term reduction in health center visits post writing (Figure 4a). This is an
example of regulated activation, a boon that can be expected from the cool system. But the
most common effect of the cool system is neutralization of the hot system. For instance,
increased frequency of Core affect from day 1 to day 2 was associated with health cost at
Louise Sundararajan, Chulmin Kim, Martina Reynolds et al.
84
follow up for the Control group, but not for the Expressive Writing group (see Figure 4a),
possibly due to the cool system effect of the instruction set, which neutralized the hot system.
Another case in point is optimal emotion representation (Figure 4b), which can be
expected to be associated with cool system effect. The cool system effect of this type of
language use was manifest, for the Control group, in terms of reduced health center visits and
decreased subjective experience of stress, whereas for the Expressive Writing group, it was
manifest in the efficiency of sign use, which consists of letting information proliferate
maximally- as evidenced by significantly more output on attention to affect and facilitative
distance variables than the controls (see Figure 2)—while keeping energy cost at the
minimum, such that no health cost or benefit showed up on the ledger (see Figure 4b).
Study 2
This is a reanalysis of a published study of children‘s expressive writing (Fivush, et al.,
2007), in which 9 to 13-year old children engaged in three consecutive days of writing, for 15
to 20 minutes each day, under emotional and non-emotional instructions (n=56 each)--the
former were asked to write about their deepest thoughts and feelings; the latter about how
they spend a typical day.
Outcome Measures
The following outcome measures were completed by the children one day before writing,
and again two months after writing:
The Birleson Depression Inventory: a self-report on childhood depression (DEP).
The Spence Children‘s Anxiety Scale: a self-report on childhood anxiety (ANX).
The Children‘s Somatisation Inventory: a self report on psychophysiological symptoms
such as headaches, dizziness, and so on (PHY).
The Strengths and Difficulties Questionnaire: The sum of the items of this questionnaire
measure the child‘s overall difficulties. The questionnaire was filled out by the child and the
teacher, resulting in two versions, the child‘s (CSD) and the teacher‘s (TSD), respectively.
Similar to study 1, both groups benefited from writing, showing lower anxiety,
depression, difficulties and somatic symptoms from baseline to follow-up, in comparison to
the non-writing group. With that finding already established in a prior study (Reynolds,
Brewin, and Saxton, 2000), the authors proceeded to analyze individual differences in
writing.
Surprisingly children who discussed emotions and explanations more in their narratives
subsequently showed higher levels of depression and anxiety. The authors attributed this to
children‘s lack of language skills to benefit from expressive writing. To further investigate
this phenomenon, we used SSWC to re-analyze the texts. In Study 2, we included children
who did not complete all three diary entries, resulting in a slightly larger sample size (n=115;
emotional instruction, n=58; non-emotional instruction, n=57) than the original study
(n=112).
Language, Emotion, and Health
85
RESULTS
Over View
Language use: Similar to Study 1, the two groups of children wrote very differently, as
measured by the SSWC variables (Figure 5).
Over all, children wrote in similar ways as adults (Study 1, Figure 2), especially with
regard to over-distance type of language use. It is in the under-distance type of language use
that some differences emerged: Children, in both writing groups, had higher output than their
counterpart in the adult sample (Figure 2) on High Activation and Violent Words. While
children showed no group difference in the frequency of use of Somatic category, both groups
had higher output on this category than the adult sample (Figure 2), in which controls
significantly outperformed the Expressive Writing group. The group differences in children
can be summed up in Figure 6.
As shown in Figure 6, the two writing groups of children did not differ in word count, but
differed significantly in representations of self and emotions as well as core affect. The over
all patterns of group difference in the child sample (Figure 6) are quite similar to the adult
counterpart (Figure 3), in both weighted mean as well as difference scores, if we interpret the
day 1 to day 2 difference in the adult sample (Figure 3) as equivalent to that between day 2 to
day 3 in the child sample (Figure 6)—in both the Control group decreased output while the
Expressive Writing group increased output, on the last day, on core affect and the expressions
of self and emotions.
Note. *p<.05, **p<.01, ***p<.001. Ns=Not significant at .05. In=Invalid comparison due to low
baseline.
Figure 5. Study 2 (N=115), group comparison, based on weighted mean across writing days, on
variables of SSWC (Sundararajan-Schubert Word Count), by writing task.
SSWC Variables
SSWC Variables (with tokens)
Expressive Writing group
Mean (SD)
Control group
Mean (SD)
Significance
Significance
Attention to Affect
Affect Focal (happy/sad) 1.56 (0.86) 0.43 (0.65) ***
Valence Focus (miserable) 0.37 (0.44) 0.04 (0.09) ***
Facilitative Distance from Experience
Reflexive Self (ourselves) 0.22 (0.27) 0.43 (0.42) ***
Detached Self (someone) 1.63 (1.20) 0.64 (0.50) ***
External Attribution (sexy) 1.11 (0.66) 0.69 (0.64) ***
Under Distance from Experience
Affected Self (me) 3.17 (1.35) 1.33 (0.92) ***
Violent Words (swear) 0.36 (0.47) 0.10 (0.26) ***
Somatic (headache) 0.17 (0.27) 0.18 (0.29) Ns
Suffering (traumatized) 0.02 (0.08) 0.01 (0.05) In
Emotion as Action (love/hate, used as verb) 0.40 (0.44) 0.15 (0.29) ***
High Activation (excited) 0.22 (0.33) 0.10 (0.20) ***
Over Distance from Experience
Focal Self (I, myself, my own) 6.29 (1.03) 7.18 (1.30) ***
Affect Non-Focal (cry, understand) 0.72 (0.33) 0.36 (0.22) ***
Denial (“doesn’t bother me”) 0.05 (0.06) 0.01 (0.03) ***
Low Activation (bored) 0.03 (0.05) 0.05 (0.06) Ns
Figure 5: Study 2 (N=115), group comparison, based on weighted mean across writing days, on
variables of SSWC (Sundararajan-Schubert Word Count), by writing task.
Note. *p<.05, **p<.01, ***p<.001. Ns=Not significant at .05. In=Invalid comparison due to low baseline.
Louise Sundararajan, Chulmin Kim, Martina Reynolds et al.
86
Note. diff1=difference score between the writing days (Day2–Day1); diff2=difference score between
the writing days (Day3–Day2). sum=weighted mean across the 3 writing days. *p<.05, **p<.01,
***p<.001.
Figure 6. Study 2, group differences in global variables of SSWC (Sundararajan-Schubert Word
Count), by writing task.
However, unlike Study 1, children‘s word count had health consequences, as shown in
Figure 7a.
Word Count: For the Expressive Writing group, weighted mean of word count across
three writing days was positively correlated (r=.39, p<.01) with the child‘s anxiety at follow
up two months post writing. For the Control group, increase in word count on the second day
and the third day of writing were positively correlated with anxiety (r=.32, p<.05) and
depression (r=.33, p<.05), respectively, at follow up.
Expressions of Self and Emotions (E): For the Expressive Writing group, no significant
correlation was found between outcome measures and the percentage of E. For the Control
group, percentage of E was strongly and positively correlated (r=.46, p<.01) with depression
at two months follow up.
Core Affect: For the Expressive Writing group, no significant correlation was found with
outcome measures. For the Control group, an increase in expression of Core Affect from day
1 to day 2 of writing was positively correlated (r=.28, p<.05) with anxiety post writing.
-32
-16
0
16
32
Word Count
(diff1)
Word Count
(diff2)
MEAN
-1.4
-0.7
0.0
0.7
1.4
Representations
of Self and
Emotions (diff1)
Representations
of Self and
Emotions (diff2)
***
MEAN
-0.2
-0.1
0.0
0.1
0.2
Core Affect
(diff1)
Core Affect
(diff2)
Control
Expressive
Writing
MEAN
0
150
300
450
600
Word Count (sum)
MEAN
0
5
10
15
20
Representations of Self and Emotions (sum)
***
MEAN
0.0
0.2
0.4
0.6
0.8
Core Affect (sum)
Control
Expressive
Writing
***
MEAN
Figure 6: Study 2, group differences in global variables of SSWC (Sundararajan-Schubert Word Count), by writing task.
Note. diff1=difference score between the writing days (Day2–Day1); diff2=difference score between the writing days
(Day3–Day2). sum=weighted mean across the 3 writing days. *p<.05, **p<.01, ***p<.001.
Language, Emotion, and Health
87
Note. diff1=difference score between the writing days (Day2–Day1). diff2=difference score between
the writing days (Day3–Day2). sum=weighted mean across the 3 writing days. ANX=outcome
measure of anxiety. DEP=outcome measure of depression. *p<.05, **p<.01, ***p<.001.
Figure 7a. Study 2, partial correlations, between outcome measures and global variables of SSWC
(Sundararajan-Schubert Word Count), for (top) Control and (bottom) Expressive Writing groups.
Results for components of optimal emotion representation—attention to affect and
facilitative distance from experience--are shown in Figure 7b.
Attention to Affect
Affect Focal (happy/sad): For the Expressive Writing group, weighted mean of bona fide
emotion terms was negatively correlated (r=-.31, p<.05) with somatic symptoms post writing.
For the Control group, weighted mean of this type of language use was positively
correlated (r=.33, p<.05) with depression post writing, but difference scores tell a more
nuanced story. An increase in this type of expression from day 1 to day 2 was negatively
correlated (r=-.29, p<.05) with depression post writing, whereas an increase of the same from
day 2 to day 3 was strongly and positively correlated (r=.48, p<.01) with depression at follow
up. Valence focus (miserable/pleasant): For the Expressive Writing group, weighted mean
across writing days of valence focus was negatively correlated (r=-.30, p<.05) with
difficulties rated by the child post writing. For the Control group, an increase in use of this
category from day 2 to day 3 was positively correlated (r=.33, p<.05) with difficulties rated
by the child at follow up post writing.
Facilitative Distance from Experience
External Attribution (sexy, wonderful): For the Expressive Writing group, no significant
correlation was found with outcome measures. For the Control group, an increase in
0.317*
ANX
0.326 *
DEP
0.458**
DEP
0.284*
ANX
-0.6
-0.3
0.0
0.3
0.6
Word Count (diff1)
Word Count (diff2)
Word Count (sum)
Representations of
Self and Emotions
(diff1)
Representations of
Self and Emotions
(diff2)
Representations of
Self and Emotions
(sum)
Core Affect (diff1)
Core Affect (diff2)
Core Affect (sum)
Control
MEAN
0.391**
ANX
-0.6
-0.3
0.0
0.3
0.6
Word Count (diff1)
Word Count (diff2)
Word Count (sum)
Representations of
Self and Emotions
(diff1)
Representations of
Self and Emotions
(diff2)
Representations of
Self and Emotions
(sum)
Core Affect (diff1)
Core Affect (diff2)
Core Affect (sum)
Expressive Writing
MEAN
Figure 7a: Study 2, partial correlations, between outcome measures and global variables of SSWC
(Sundararajan-Schubert Word Count), for (top) Control and (bottom) Expressive Writing groups.
Note. diff1=difference score between the writing days (Day2–Day1); diff2=difference score between the writing days
(Day3–Day2). sum=weighted mean across the 3 writing days. ANX=outcome measure of anxiety. DEP=outcome measure
of depression. *p<.05, **p<.01, ***p<.001.
Louise Sundararajan, Chulmin Kim, Martina Reynolds et al.
88
expression of emotion with external attribution from day 2 to day 3 of writing was negatively
correlated, strongly with anxiety (r=-.36, p<.01) and moderately with somatic symptoms (r=-
.29, p<.05), post writing.
Note. diff1=difference score between the writing days (Day2–Day1). diff2=difference score between
the writing days (Day3–Day2). sum=weighted mean across the 3 writing days. ANX=outcome
measure of anxiety. DEP=outcome measure of depression. PHY=outcome measure of
somatisation. CSD=overall difficulties reported by the child. *p<.05, **p<.01, ***p<.001.
Figure 7b. Study 2, partial correlations, between outcome measures and Components of Optimal
Emotion Representation of SSWC (Sundararajan-Schubert Word Count), for (top) Control and (bottom)
Expressive Writing groups.
Under Distance from Experience
Affected Self (me, ―making me . . .‖): For the Expressive Writing group, weighted mean
of a passive stance across three writing days was strongly and positively correlated ( r=.45,
p<.01) with anxiety post writing. For the control group, weighted mean of this type of
language use was correlated strongly and positively (r=.37, p<.01) with depression post
writing, but negatively (r=-.40, p<.01) with difficulties rated by the child.
Emotion as action (love/hate, used as verb): For the Expressive Writing group, no
significant correlation was found. For the Control group, weighted mean of this type of un-
reflective expression of emotion across three writing days was strongly and positively
correlated (r=.43, p<.01) with depression post writing; similarly, an increase in this type of
expression from day 2 to day 3 of writing was strongly and positively correlated with
depression (r=.39, p<.01) at follow up.
Suffering (devastated, traumatized): For the Expressive Writing group, an increase from
day 2 to day 3 of expression of trauma was positively correlated (r=.33, p<.05) with
depression post writing. For the Control group, weighted mean of this type of expression
DEP
-0.291*
0.482**
DEP
0.331*
DEP
0.331*
CSD
-0.6
-0.3
0.0
0.3
0.6
MEAN
ANX
-0.363**
PHY
-0.292*
-0.6
-0.3
0.0
0.3
0.6
Control
MEAN
PHY
-0.310*
CSD
-0.298*
-0.6
-0.3
0.0
0.3
0.6
MEAN
-0.6
-0.3
0.0
0.3
0.6
Expressive
Writing
MEAN
Attention to affect Facilitative distance from experience
Figure 7b: Study 2, partial correlations, between outcome measures and Components of Optimal Emotion Representation of
SSWC (Sundararajan-Schubert Word Count), for (top) C ontrol and (bottom) Expressive Writing groups.
Note. diff1=difference score between the writing days (Day2–Day1); diff2=difference score between the writing days (Day3–Day2).
sum=weighted mean across the 3 writing days. ANX=outcome measure of anxiety. DEP=outcome measure of depression.
PHY=outcome measure of somatisation. CSD=overall difficulties reported by the child. *p<.05, **p<.01, ***p<.001.
Language, Emotion, and Health
89
across three writing days was positively correlated (r=.29, p<.05) with the child‘s difficulties
rated by teacher; and the same pattern of correlation (r=.31, p<.05) at follow up was found for
an increase in expression of trauma from day 2 to day 3 of writing.
Violent Words (kill, rape, swear words): For the Expressive Writing group, an increase in
the use of violent words from day 1 to day 2 of writing was strongly and positively correlated
(r=.41, p<.01) with child‘s difficulties rated by teacher. No significant correlation with
outcome measures was found for the Control group.
Note. diff1=difference score between the writing days (Day2–Day1). diff2=difference score between
the writing days (Day3–Day2). sum=weighted mean across the 3 writing days. ANX=outcome
measure of anxiety. DEP=outcome measure of depression. TSD=overall difficulties reported by
the teacher. CSD=overall difficulties reported by the child. *p<.05, **p<.01, ***p<.001.
Figure 7c. Study 2, partial correlations, between outcome measures and Under distance categories of
SSWC (Sundararajan-Schubert Word Count), for (top) Control and (bottom) Expressive Writing
groups.
High Activation (excited, nervous): For the Expressive Writing group, no significant
correlation was found. For the Control group, an increase in expressions of high activation
from day 1 to day 2 of writing was positively correlated with anxiety (r=.29, p<.05), and with
difficulties rated by the child (r=.33, p<.05); similarly, increase in expression of high
activation from day 2 to day 3 was positively correlated with depression (r=.35, p<.05) at
follow up.
0.366**
DEP
0.311*
TSD
0.288*
TSD
0.393**
DEP
0.427**
DEP
0.291*
ANX
0.352*
DEP
CSD
-0.399**
0.334*
CSD
-0.6
-0.3
0.0
0.3
0.6
Control
MEAN
0.446**
ANX
0.412**
TSD
0.332 *
DEP
-0.6
-0.3
0.0
0.3
0.6
Expressive
Writing
MEAN
Under distance from experience
Figure 7c: Study 2, partial correlations, between outcome measures and Under distance categories of SSWC
(Sundararajan-Schubert Word Count), for (top) C ontrol and (bottom) Expressive Writing groups.
Note. diff1=difference score between the writing days (Day2–Day1); diff2=difference score between the writing days (Day3–
Day2). sum=weighted mean across the 3 writing days. ANX=outcome measure of anxiety. DEP=outcome measure of
depression. TSD=overall difficulties reported by the teacher. CSD=overall difficulties reported by the child. *p<.05, **p<.01,
***p<.001.
Louise Sundararajan, Chulmin Kim, Martina Reynolds et al.
90
Over Distance from Experience
Focal Self (I, myself, my own): For the Expressive Writing group, no significant
correlation with outcome measures was found. For the Control group, weighted mean of self
focus across three writing days was positively correlated (r=.29, p<.05) with depression at
follow up. This is consistent with findings of the connection between high self focus and
depression (Watkins and Teasdale, 2004). However, difference score indicating an increase in
self focus from day 1 to day 2 of writing was negatively correlated (r=-.29, p<.05), with
depression at follow up.
Note. diff1=difference score between the writing days (Day2–Day1). diff2=difference score between
the writing days (Day3–Day2). sum=weighted mean across the 3 writing days. ANX=outcome
measure of anxiety. DEP=outcome measure of depression. TSD=overall difficulties reported by
the teacher. CSD=overall difficulties reported by the child. *p<.05, **p<.01, ***p<.001.
Figure 7d. Study 2, partial correlations, between outcome measures and Over distance categories of
SSWC (Sundararajan-Schubert Word Count), for (top) Control and (bottom) Expressive Writing
groups.
Affect Non-Focal (cry, understanding): For the Expressive Writing group, weighted mean
of expressing emotion in cognitive and behavioral terms across three writing days was
negatively correlated (r=-.33, p<.05) with child‘s difficulties rated by teacher. The difference
score tells a more nuanced story: An increase in cognitive and behavioral expressions of
emotion from day 1 to day 2 was strongly and positively correlated (r=.38, p<.01) with
difficulties rated by the child, whereas an increase of the same from day 2 to day 3 was
negatively correlated (r=-.35, p<.05) with anxiety at follow up. For the Control group,
increase in cognitive and behavioral expressions of emotion from day 1 to day 2 was
positively correlated (r=.30, p<.05) with anxiety at follow up.
DEP
-0.292*
0.290*
DEP
0.302*
ANX
-0.6
-0.3
0.0
0.3
0.6
Focal self(diff1)
Focal self(diff2)
Focal self(sum)
Affect Non-
Focal(diff1)
Affect Non-
Focal(diff2)
Affect Non-
Focal(sum)
Denial(diff1)
Denial(diff2)
Denial(sum)
Low
Activation(diff1)
Low
Activation(diff2)
Low
Activation(sum)
Control
MEAN
0.379**
CSD
ANX
-0.352*
TSD
-0.325*
0.312*
TSD
-0.6
-0.3
0.0
0.3
0.6
Focal self(diff1)
Focal self(diff2)
Focal self(sum)
Affect Non-
Focal(diff1)
Affect Non-
Focal(diff2)
Affect
Non-Focal(sum)
Denial(diff1)
Denial(diff2)
Denial(sum)
Low
Activation(diff1)
Low
Activation(diff2)
Low
Activation(sum)
Expressive
Writing
MEAN
Over distance from experience
Figure 7d: Study 2, partial correlations, between outcome measures and Over distance categories of SSWC
(Sundararajan-Schubert Word Count), for (top) Control and (bottom) Expressive Writing groups.
Note. diff1=difference score between the writing days (Day2–Day1); diff2=difference score between the wri ting days (Day3–
Day2). sum=weighted mean across the 3 writing days. ANX=outcome measure of anxiety. DEP=outcome measure of
depression. TSD=overall difficulties reported by the teacher. CSD=overall difficulties reported by the child. *p<.05, **p<.01,
***p<.001.
Language, Emotion, and Health
91
Low Activation (bored, drowsy): For the Expressive Writing group, weighted mean of
expressions of low activation was positively correlated (r=.31, p<.05) with child‘s difficulties
rated by teacher at follow up. For the Control group, no significant correlation with outcome
measures was found.
DISCUSSION
For both groups, there was a positive correlation between length of the written text and
anxiety and depression at follow up. The two writing conditions also did not differ in the
health consequences of less than optimal representations of self and emotions (Figures. 7c and
7d)--both groups seemed to be susceptible to activation of the hot system. However,
children‘s susceptibility was different from the adult sample (Study 1). In the adult sample,
dysregulated activation was associated with over-distance type of language use (Figure 4d),
while regulated activation with under-distance type of language use (Figure 4c); the reverse
was the case with children—Under distance expressions were associated with extensive
symptoms characteristic of dysregulated activation (Figure 7c), while the use of over-distance
type of expressions was associated with a mixture of cost and benefit (Figure 7d), which to
some extent approximates regulated activation, although the capacity for children to regulate
the hot system was not as evident as the adult sample. This contrasting pattern may stem from
the developmental needs of children, who seemed to be in particular need for mental distance
such that under-distance type of language use would cause more activation of the hot system
than over-distance.
Children‘s need for mental distance to regulate the hot system may explain why both
writing conditions benefitted especially from language use that entails facilitative distance
from experience (Figure 7b, right panel). This also explains why language use such as
―happy‖ or ―sad‖--that has attention to affect as its primary referential focus, and that was
associated with the cool system in the adult sample (Figure 4b, left panel)-- activated the hot
system in the child sample, as evidenced by associated symptoms at follow up (Figure 7b, left
panel). That attention to affect activated the hot system for children is particularly true of the
Control group, which showed a positive correlation between increase in depression at follow
up and a higher percentage of expressions of self and emotions in child‘s writing; the same
correlation holds for anxiety and core affect (see Figure 7a).
This suggests that while children were equally vulnerable as adults to the health cost of
less than optimal representations, they were less able to reap the health benefit from optimal
representations of self and emotions—unless they got help. Results of the Expressive Writing
group showed that the instruction set could help by reinforcing and extending the cool
system. Thus when children wrote with the explicit instruction to pay attention to their
thoughts and feelings, their cool system prevailed where the hot system would have been
dominant otherwise: With attention to affect type of language use, the Expressive Writing
group showed negative correlation with symptomatology at follow up, in sharp contrast to the
controls who showed positive correlation with the same (Figure 7b, left panel). Similarly,
whereas children in the Control group reported depression and anxiety at follow up if they
devoted a large proportion of their writing to expressions of self and emotions and core affect,
the Expressive Writing group had more output of these expressions (see Figure 5) without
Louise Sundararajan, Chulmin Kim, Martina Reynolds et al.
92
reaping any health consequences (see Figure 7a). It seems that with the help of the expressive
writing instructions, language in emotional writing could be decoupled from the hot system,
and became simply information. The cool system effect strengthened by the writing
instruction of expressive writing may explain why the protocol produced by the Expressive
Writing group of children was more similar to that of the adult sample in Study 1 than the
control group of children.
How do the results presented here square with the findings of Fivush, et al. (2007) that
the more children wrote about emotions, the more symptoms of depression and anxiety they
showed at follow up? The two studies do not necessarily contradict each other, since the
original study focused on the content of children‘s writing, whereas our re-analysis shifted
from the what to the how question. Take for instance the hypothetical topic of love, which can
be expressed by multiple categories of language use as indicated by italics below:
a. ―I love you‖ which would fall under the category of Emotion as Action;
b. ―My love is a red, red, rose‖ which belongs to the category of Affect Focal;
c. ―You are beautiful‖ which falls under External Attribution;
d. ―I am so excited to see you‖ which would fall under High Activation.
The results of Study 2 predicts the following protocol: If the writing came from the
Expressive Writing group, use of (a), (c), and (d) would not be associated with
symptomatology, whereas high frequency of (b) would be associated with a reduction in
somatic complaints at follow up. By contrast, results of the Control group confirmed the
finding of Fivush, et al. (2007) that the more the child wrote about emotions, the more anxiety
and depression the child showed at follow up: With the exception of (c) which was associated
with a reduction of anxiety and somatic complaints (Figure 7b, right panel), all the other
forms of expressing emotions, (a, b, and d), were associated with an increase in
symptomatology (Figure 7b, left panel; Figure 7c) at follow up.
Together, results of Study 1 and the Expressive Writing condition of Study 2 confirmed
our prediction--optimal representations of emotion are under the sway of the cool system,
while less than optimal representations are under varying degrees of dominance by the hot
system. Although the Control group in Study 2 confirmed the contention of Fivush, et al.
(2007) that some children might not have the necessary language skills to benefit from
emotional writing, the results of Study 2 in general suggested a conclusion that was opposed
to that of the original study, namely that explicit instructions of expressive writing may help
these children. While we need to be cautious with the interpretation of certain low baseline
categories, such as Suffering, Denial, and Low Activation, the overall picture that emerged
from these results tells a nuanced story about how the connection between language use and
health is robust but varies along developmental parameters.
SUMMARY AND CONCLUSION
The basic idea behind the semiotics of Charles Peirce is integration (Sundararajan, 2008),
a notion that is consistent with Pennebaker‘s cognitive reorganization thesis (1985) as well as
his inhibition theory (1993), both converging on the claim that integration of feeling and
Language, Emotion, and Health
93
thinking, thought and experience is essential to health. The unique contribution of Peircen
semiotics lies in its capacity to model the dynamics of integration in terms of complexity. The
basic insight is that it is process (known as sign action), not content, that determines the
efficiency of the sign as a representation of experience. This shift of focus from what to how,
from the content of the information per se, to the modes of processing and representation of
emotion information is consistent with the claim of Philippot and colleagues (e.g., Neumann,
and Philippot, 2007) that regulation of emotion can be achieved by a change in processing
mode without ostensible modification of the emotional information content. Thus instead of
the conventional content analysis that focuses on what is said in the text, it is now feasible to
approach language as modes of information processing with varying degrees of complexity.
Based on the Peircean model of complexity, we proposed a taxonomy of 15 categories of
language use in emotion expressions. These are further grouped into four types of language
use, or processing modes, each with its unique balance between the cool and hot systems of
emotion.
Based on the insight of Peirce that the mind is in signs, and not the other way around
(Colapietro, 1989), our approach entails a figure and ground reversal that puts language on
center stage as the main actor, and group differences as the contextual factors--along with
other contextual factors such as individual differences and developmental needs--that affect
the shifting balance between hot and cool systems associated with different types of language
use. An analogy is gene (also a code) expression which is affected by environmental contexts.
Our theory-based predictions were implemented by a Language analysis program, SSWC
(Sundararajan-Schubert Word Count), which re-analyzed texts from two studies of expressive
writing, one by adults and one by children. Results of both studies showed that the categories
of SSWC map out the semantic space of affect in a logically consistent and intelligible
manner. The taxonomy of SSWC functions not as a dictionary so much as a prism, which
renders visible the various modes of representing emotions. Its proposed classification of
language use functions not as a dream book with fixed interpretations, so much as a theory-
based system of coding, that tags the various modes of representation so as to see whether and
how they vary systematically along various parameters such as instruction sets, individual
differences, and developmental needs.
The findings presented here are tentative, but, if confirmed by future replications, may
have far reaching implications for both theory and research on the language and health
equation. The current research in the field tends to approach expressive writing as a unitary
phenomenon, on the presence (the experimental condition) or absence (the control condition)
of which hangs the balance for health outcomes. This approach renders expressive writing a
black box, thereby making it difficult to investigate the qualitative differences between
different types of emotion expression, and their ramifications for health. The semiotic
approach to language makes it possible to shift our focus from the whether to the how
question. Instead of asking whether expressive writing in general has health benefits
(Frattaroli, 2006), we examined how the language and health equation may vary
systematically under different contexts and conditions.
Louise Sundararajan, Chulmin Kim, Martina Reynolds et al.
94
ACKNOWLEDGMENTS
We thank Anna Graybeal and Robyn Fivush for providing the data for our analysis.
REFERENCES
Averill, J. R. (19375). A semantic atlas of emotion concepts. JSAS Catalogue of Selected
Documents in Psychology, 5, 330 (Ms. No. 421).
Barrett, L. F., and Russell, J. A. (1998). Independence and bipolarity in the structure of
current affect. Journal of Personality and Social Psychology, 74, 967-984.
Berlyne, D. E. (1960). Conflict, arousal, and curiosity. New York: McGraw-Hill.
Bohm, D. (1994). Soma-significance: A new notion of the relationship between the physical
and the mental. Psychoscience, 1 (1), 6-27.
Bootzin, R. R. (1997). Examining the theory and clinical utility of writing about emotional
experiences. Psychological Science, 8, 167-169.
Bucci, W. (1995). The power of the narrative: A multiple code account. In J. W.
Pennebaker (Ed.), Emotion, disclosure, and health (93-124). Washington, DC: American
Psychological Association.
Burbridge, J. A., Larsen, R. J., and Barch, D. M. (2005). Affective reactivity in language: The
role of psychophysiological arousal. Emotion, 5, 145-153.
Colapietro, V. M. (1989). Peirce‟s approach to the self/A semiotic perspective on human
subjectivity. Albany, NY: SUNY Press.
Campbell, J. (1982). Grammatical man. New York: Simmon and Schuster.
Campbell, R. S., and Pennebaker, J. W. (2003). The secret life of pronouns: Flexibility in
writing style and physical health. Psychological Science, 14, 60-65.
Clore, G. L., Ortony, A., and Foss, M. A. (1987). The psychological foundations of the
affective lexicon. Journal of Personality and Social Psychology, 53, pp. 751-766.
Conway, MA., and Pleydell-Pearce, C. W. (2000). The construction of autobiographical
memories in the self-memory system. Psychological Review, 107, 261-288.
Damasio, A. (1999). The feeling of what happens. New York: Harcourt Brace.
Deacon, T. W. (1997). The symbolic species. New York: W. W. Norton.
Estes, W. K. (1972). Reinforcement in human behavior. American Psychologist, 60, 723-729.
Fisch, M. H. (Ed.). (1982). Writings of Charles S. Peirce: a chronological edition (5 Vols).
Bloomington: Indiana University.
Fivush, R., Marin, K., Crawford, M., Reynolds, M., and Brewin, C. R. (2007). Children‘s
narratives and well-being. Cognition and Emotion, 21, 1414-1434.
Forgas, J. P. (2001). Affect and information processing strategies: An interactive relationship.
In J. P. Forgas (Ed.), Feeling and thinking/The role of affect in social cognition (pp. 253-
282). Cambridge: Cambridge University.
Frattaroli, J. (2006). Experimental disclosure and its moderators: A meta-analysis.
Psychological Bulletin,132, 823-865.
Frijda, N.H. (2005). Emotion experience. Cognition and Emotion, 19, 473-498.
Frijda, N.H. (2007). The laws of emotion. Mahwah, NJ: Erlbaum.
Language, Emotion, and Health
95
Graybeal, A. (2004). Expressive writing as a therapeutic intervention for adult children of
divorce. Unpublished doctoral dissertation, University of Texas at Austin.
Graybeal, A., Sexton, J. D., and Pennebaker, J. W. (2002). The role of story-making in
disclosure writing: The psychometrics of narrative. Psychology and Health, 17, 571-581.
Hoopes, J. (Ed.). (1991). Peirce on signs. Chapel Hill, NC: The University of North Carolina.
Johnson, M. (2007). The meaning of the body. Chicagoand London: University of Chicago
Press.
Kihlstrom, J. F., Mulvaney, S., Tobias, B. A., and Tobis, I. P. (2000). The emotional
unconscious. In E. Eich, J. F. Kihlstrom, G. H. Bower, J. P. Forgas, and P. M.
Niedenthal, Cognition and emotion (pp. 30-86). New York: Oxford University.
King, L. A. (2002). Gain without pain? Expressive writing and self-regulation. In S. J. Lepore
and J. M. Smyth (Eds.), The writing cure: How expressive writing promotes health and
emotional well-being (pp. 119-134). Washington, DC: American Psychological
Association.
Krystal, H. (1988). Integration and self-healing: Affect, trauma, alexithymia. Hillsdale, NJ:
Analytic Press.
Labouvie-Vief, G. (2003). Dynamic integration: Affect, cognition, and the self in adulthood.
Current Directions in Psychological Science, 12, 201-206.
Labouvie-Vief, G., and Marquez, M. G. (2004). Dynamic intergration: Affect optimization
and differentiation in development. In D. Y. Dai and R. J. Sternberg (Eds.), Motivation,
emotion, and cognition: Integrative perspectives on intellectual functioning and
development (pp. 237-272). Mahwah, N.J.: Lawrence Erlbaum.
Lambie, J. A., and Marcel, A. J. (2002). Consciousness and the varieties of emotion
experience: A theoretical framework. Psychological Review, 109, 219-259.
Lane, R. D. (1991). LEAS scoring manual and glossary. Department of Psychiatry,
University of Arizona Health Sciences Center.
Lee, B. (1997). Talking heads/Language, metalanguage, and the semiotics of subjectivity.
Durham: Duke University Press.
Legrand, D., and Ruby, P. (2009). What is self-specific? Theoretical investigation and critical
review of neuroimaging results. Psychological Review, 116, 252-282.
Low, C. A., Stanton, A. L., and Bower, J. E. (2008). Effects of acceptance-oriented versus
evaluative emotional processing on heart rate recovery and habituation. Emotion. 8, 419-
424.
Metcalfe, J., and Mischel, W. (1999). A hot/cool-system analysis of delay of gratification:
dynamics of willpower. Psychological Review, 106, 3-19.
Neumann, A., and Philippot, P. (2007). Specifying what makes a personal memory unique
enhances emotion regulation. Emotion, 7, pp. 566-578.
Ott, H. (1972). Hermeneutic and Personal Structure of Language. In J. J. Kockelmans (Ed.
and Tr.), On Heidegger and Language (pp. 169⌐194). Evanston: Northwestern
University Press.
Parmentier, R. J. (1994). Signs in society/Studies in semiotic anthropology. Bloomington:
Indiana University Press.
Peirce, C. S. (1931-58). Collected papers of Charles Sanders Peirce (8 Vols, edited by C.
Hartshorne and P. Weiss). Cambridge, Mass: Harvard University Press.
Peirce, C. S. (1961). The collected papers of Charles Sanders Peirce (7 Vols, edited by C.
Hartshorne and P. Weiss). Cambridge, Mass: Belknap.
Louise Sundararajan, Chulmin Kim, Martina Reynolds et al.
96
Pennebaker, J. W. (1985). Traumatic experience and psychosomatic disease: Exploring the
roles of behavioral inhibition, obsession, and confiding. Cannadian Psychology, 26, 82-
95.
Pennebaker, J. W. (1989). Stream of consciousness and stress: Levels of thinking. In J. S.
Uleman, and J. A. Bargh (Eds.), Unintended thought (327-350). New York: Guilford.
Pennebaker, J. (1993). Putting stress into words: Health, linguistic, and therapeutic
implications. Behavioral Research Therapy, 31, 539-548.
Pennebaker, J. W., and Beall, S. K. (1986). Confronting a traumatic event: Toward an
understanding of inhibition and disease. Journal of Abnormal Psychology, 95, 274-281.
Pennebaker, J. W., Mayne, T. J., and Francis, M. E. (1997). Linguistic predictors of adaptive
bereavement. Journal of Personality and Social Psychology, 72, 863-871.
Pennebaker, J. W., and Seagal, J. D. (1999). Forming a story: The health benefits of narrative.
Journal of Clinical Psychology, 55, 1243-1254.
Philippot, P., Baeyens, C., and Douilliez, C. (2006). Specifying emotional information:
Regulation of emotional intensity via executive processes. Emotion, 6, 560-571.
Reynolds, M., Brewin, C. R., and Saxton, M. (2000). Emotional disclosure in school children.
Journal of Child Psychology and Psychiatry, 41, 151-159.
Russell, J. A. (2003). Core affect and the psychological construction of emotion.
Psychological Review, 110, 145-172.
Sabelli, H. (2005). Bios: A study of creation. Singapore: World Scientific.
Scheff, T. J. (1979). Catharsis in healing, ritual, and drama. Berkeley: University of
California Press.
Sloan, D. M., and Marx, B. P. (2004). Taking pen to hand: Evaluating theories underlying he
written disclosure paradigm. Clinical Psychology: Science and Practice, 11, 121-137.
Smyth, J. M., True, N., and Souto, J. (2001). Effects of writing about traumatic xperiences:
The necessity for narrative structuring. Journal of Social and Clinical Psychology, 20,
161-172.
Sundararajan, L. (2008). It‘s turtles all the way down: A semiotic perspective on the basic
emotions debate. Journal of Theoretiocal and Philosophical Psychology, 28, 430-443.
Sundararajan, L., and Schubert, L. K. (2005). Verbal expressions of self and emotions: A
taxonomy with implications for alexithymia and related disorders. In R. D. Ellis and N.
Newton (Eds.), Consciousness and Emotion: Agency, conscious choice, and selective
perception (pp. 243-284). Amsterdam: John Benjamins.
Teasdale, J. D. and Barnard, P. J. (1993). Affect, cognition, and change: Remodeling
depressive thought. Hillsdale, NJ: Lawrence Erlbaum.
Tucker, D. M. (2007). Mind from Body: Experience from Neural Structure. Oxford: Oxford
University Press.
Watkins, E., and Teasdale, J. D. (2001). Rumination and overgeneral memory in depression:
effects of self-focus and analytic thinking. Journal of Abnormal Psychology, 110, 353-
357.
Watkins, E., and Teasdale, J. D. (2004). Adaptive and maladaptive self-focus in depression.
Journal of Affective Disorders. 82, 1-8.
Weinberger, D. A. (1990). The construct validity of the repressive coping style. In J. L.
Singer (Ed.), Repression and dissociation (pp. 337-386). Chicago: University of Chicago
Press.
Wiley, N. (1994). The semiotic self. Chicago: University of Chicago.
Language, Emotion, and Health
97
Williams, J. M. G., Barnhofer, T., Crane, C., Hermans, D., Raes, F., Watkins, E., and
Dalgleish, T. (2007). Autobiographical memory specificity and emotional disorder.
Psychological Bulletin, 133, 122-148.
Zinken, J., Sundararajan, L., Butler, L., and Skinner, T. C. (2006, August). Why is language
indicative of mental health state? An exploration of lexicon and syntax. In L.
Sundararajan (chair), Self and emotion as semiotic signs: Rediscovering Charles Sanders
Peirce. Symposium conducted at the 114th Annual Convention of the American
Psychological Association, New Orleans.