The Emotional Processing Scale: Scale refinement and abridgement (EPS-25)☆

Introduction 

In a previous paper [1], we described the development of a 38-item self-report Emotional Processing Scale (EPS) that incorporates Rachman's conceptualization of emotional processing [2] along with other psychological mechanisms that may impede emotional processing [3].

Maximum likelihood (ML) factor analysis yielded an eight-factor solution relating to styles of emotional experiencing (Discordant, Externalized, Lack of Attunement), mechanisms controlling the experience and expression of emotions (Suppression, Dissociation, Avoidance, Uncontrolled), and, finally, signs of inadequate processing (Intrusion) reflecting Rachman's conceptualization [2]. Internal reliability was moderate to high for six of eight factors, and convergent validity was satisfactory.

The EPS was designed to identify emotional processing styles and deficits and measure change in emotional processing dimensions during therapy. To date, it has been used to explore differences between diagnostic groups [4], [5] and there is ongoing research in the area of predicting postnatal depression [6]. It has been translated into nine languages and validated in Italian and Japanese samples [7], [8].

While the EPS-38 appeared promising as a multifaceted measure, scale development is an ongoing process. Two of the factors (Avoidance and Externalized) possessed relatively lower internal reliability than the other six subscales. In addition, feedback from those using the EPS-38 suggested that an abridged version would confer benefits from both a research and a clinical perspective. Thus, the aims of the present research were to (i) refine the scale by trying out items from a new item pool; and (ii) shorten the scale to enhance its clinical and research utility.

Study 1: Item selection and scale redevelopment 

Methods 

Participants 

Participants (N=690) were recruited from a number of settings to respond to a 53-item version of the questionnaire (Table 1).

•Healthy control group (n=310): recruited from workplaces and the community.

•General medical practice group (n=86): individuals attending an appointment at their local medical practice.

•Mental health group (n=180): individuals referred by general medical practitioners to a clinical psychologist or counselor for a range of mental health problems.

•Pain group (n=114): individuals attending a hospital outpatients appointment (fibromyalgia, n=52; rheumatoid arthritis, n=34; chronic lower back pain, n=28).

Table 1. Demographics of sample

		Group
		Control (n=310)	Medical practitioner (n=86)	Mental health (n=180)	Pain (n=114)
	Gender, n (%)
	Males	86 (30%)	14 (19%)	64 (36%)	8 (7%)
	Females	201 (70%)	59 (81%)	114 (64%)	105 (93%)
	Total	287 (100%)	73 (100%)	178 (100%)	113 (100%)
	Missing data	23	13	2	1
	Age
	≤25	92 (33%)	9 (13%)	42 (26%)	31 (27%)
	26–35	75 (27%)	8 (11%)	43 (26%)	59 (52%)
	36–45	47 (17%)	14 (19%)	32 (20%)	6 (5%)
	46–55	39 (14%)	15 (21%)	30 (18%)	4 (4%)
	56–65	16 (6%)	12 (17%)	8 (5%)	8 (7%)
	66+	10 (4%)	14 (19%)	8 (5%)	5 (4%)
	Total	279 (100%)	72 (100%)	163 (100%)	113 (100%)
	Missing data	31	14	17	1
	Education (highest formal qualification obtained)
	None	5 (2%)	10 (13%)	8 (9%)	34 (31%)
	1 or more GCSE (or equivalent)a	78 (28%)	36 (45%)	47 (51%)	29 (26%)
	1 or more A levela	54 (19%)	13 (16%)	23 (25%)	5 (5%)
	First/higher degree	134 (48%)	21 (26%)	14 (15%)	16 (14%)
	Other	9 (3%)	0 (0%)	1 (1%)	27 (24%)
	Total	280 (100%)	80 (100%)	93 (100%)	111 (100%)
	Missing data	30	6	87	3

a The General Certificate of Secondary Education (GCSE) is the name of a set of qualifications, generally taken by students at age 14–16 in England, Wales, and Northern Ireland. The A level (Advanced level) is a qualification in England, Wales, and Northern Ireland, usually taken at age 16–18.

Item pool 

An additional 14 items were added to the existing 38-item scale in order to:

•Seek to improve the Externalized and Avoidance factors and include items relating to an anti-emotionality/rationality construct [9] (e.g., “I was extremely rational and kept emotions out of the picture”).

•Try out additional items relating to emotional styles, regulation of emotion, and signs of inadequate processing.

Item analysis 

Negatively keyed items were reverse scored prior to data analysis. Three items (15, 33, 52) were removed because they had corrected item-total correlations <.20 [10] and failed to show statistically significant between-group (mental health vs. pain vs. medical practitioner vs. control) differences.

After excluding participants who did not provide complete responses to all 49 remaining items, we were left with a sample of n=603, giving a participants-to-items ratio >10:1 [11].

Exploratory factor analysis (EFA) was used because our primary aim was to explore and refine the underlying structure of the items [12]. As our sample was heterogeneous, it should be noted we have assumed the same factor structure across the groups. Examination of individual item skew and kurtosis (mean skew=0.27, S.D.=0.41, range=−0.55 to 1.43; mean kurtosis=−0.83, S.D.=0.47, range=−1.35 to 1.22) confirmed the suitability of the ML factor extraction procedure [12]. An oblique rotation (Promax) was chosen to allow for correlation between factors. Given the suitability of the data, we conducted a ML factor analysis [13] with Promax rotation (κ=4) selecting a nine-factor solution to explore the underlying structure of the 49 items. Items with loadings <.40 were discarded (n=12) as were any factors comprising three or fewer items (n=5). The factor analysis was recomputed on the remaining 32 items and a five-factor solution was selected. The same criteria were applied as previously, leading to the removal of five additional items. The factor analysis was recomputed (again selecting a five-factor solution) producing an eight-item first factor, a six-item second factor, a five-item third factor, and fourth and fifth factors, each with four items.

The rationale underlying the next step was to shorten the scale. We strived for a balanced scale with equal numbers of items per factor. To achieve this, (i) the three lowest loading items were removed from Factor 1; (ii) the only remaining reverse scored item (8) was removed; (iii) an item (20) was added to Factor 4 by returning to the initial eight-factor solution and selecting the item that possessed the next highest loading in relation to the other four items; (iv) an item (17) was added to Factor 5, using the same criteria as for Factor 4. The factor analysis was recomputed on these 25 items. One further factor analysis confirmed that a stable, five-factor, five items-per-factor structure was obtained accounting for 59.4% of the total variance (Table 2).

Table 2. Maximum likelihood analysis with Promax rotation (κ=4)

	Item no. and description		Factors
			I	II	III	IV	V
	I	Suppression (α=.84)
	41	Kept quiet about feelings	91
	44	Bottled up emotions	83
	50	Tried not to show feelings	70
	36	Could not express feelings	51
	20	Smothered feelings	39
	II	Unregulated emotion (α=.76)
	13	When upset difficult to control what I said		76
	37	Felt urge to smash something		71
	21	Reacted too much to what people said or did		68
	29	Wanted to get own back on someone		63
	38	Hard to wind down		46
	III	Impoverished emotional experience (α=.82)
	24	Seemed to be a big blank in feelings			93
	3	Emotions felt blunt/dull			76
	23	Hard to work out if I felt ill or emotional			58
	16	Strong feelings but not sure if emotions			54
	7	Feelings did not seem to belong to me			51
	IV	Signs of unprocessed emotions (α=.85)
	10	Emotional reactions lasted more than a day				79
	48	Thinking about same emotion again and again				79
	6	Unwanted feelings kept intruding				60
	42	Repeatedly experienced the same emotion				60
	46	Overwhelmed by emotions		31		42
	V	Avoidance (α=.74)
	25	Tried to talk only about pleasant things					78
	47	Tried to avoid things that might make me upset					67
	26	Could not tolerate unpleasant feelings					60
	9	Avoided looking at unpleasant things					51
	17	Talking about negative feelings made them worse					30
	Eigenvalue		8.7	2.3	1.4	1.3	1.1
	% of Variance		34.9	9.1	5.8	5.3	4.3

Factor loadings in italics indicate the assignment of items to factors. Only loadings ≥.30 (after rounding) are shown. Loadings ×100 (decimal points have been omitted).

Study 2: Between-group differences 

It was expected that individuals with physical or mental health problems would tend to score more highly on the EPS than healthy controls. A series of paired t tests were conducted using the same sample as in Study 1.

Results 

Table 3 shows that the mental health sample scored significantly more highly than the control group on all EPS subscales. The chronic pain group scored more highly on two of the EPS subscales (Impoverished emotional experience, Avoidance). The medical practitioner group scored significantly more highly than healthy controls on three of the subscales (Impoverished emotional experience, Signs of unprocessed emotion, Avoidance). The total EPS scores of all three groups were significantly higher than those of the healthy control group. Where significant differences were found, we controlled for age and gender and all differences remained significant.

Table 3. Comparison of EPS subscale and total scores across subsamples

		x̄ (S.D.)	Vs. healthy controls
			x̄ difference (95% CI)	t (df)	P value
	Suppression
	Control	3.5 (2.0)	–	–	–
	Mental health	5.1 (1.8)	1.5 (1.2 to 1.9)	8.3 (485)	<.001⁎
	Pain	4.1 (2.1)	0.6 (−0.1 to 1.0)	2.5 (422)	.02
	Medical practitioner	4.1 (1.9)	0.6 (0.1 to 1.1)	2.4 (393)	.02
	Unregulated emotion
	Control	3.2 (1.9)	–	–	–
	Mental health	4.4 (2.0)	1.1 (0.8 to 1.5)	6.2 (486)	<.001⁎
	Pain	3.5 (2.1)	0.3 (−0.1 to 0.7)	1.3 (422)	.19
	Medical practitioner	3.5 (2.1)	0.3 (−0.2 to 0.8)	1.2 (121.8)	.23
	Impoverished emotional experience
	Control	2.5 (1.8)	–	–	–
	Mental health	4.1 (2.0)	1.6 (1.2 to 1.9)	8.7 (322.5)a	<.001⁎
	Pain	3.4 (2.1)	0.9 (0.5 to 1.3)	4.2 (173.7)a	<.001⁎
	Medical practitioner	3.2 (2.0)	0.7 (0.2 to 1.1)	3.1 (392)	.002⁎
	Signs of unprocessed emotions
	Control	4.0 (2.2)	–	–	–
	Mental health	6.0 (1.9)	2.0 (1.6 to 2.4)	10.7 (416.5)a	<.001⁎
	Pain	4.0 (2.2)	−0.0 (−0.5 to 0.5)	−0.1 (422)	0.95
	Medical practitioner	4.7 (2.3)	0.7 (0.2 to 1.2)	2.6 (393)	.01⁎
	Avoidance
	Control	3.2 (1.8)	–	–	–
	Mental health	4.8 (1.6)	1.6 (1.3 to 1.9)	10.1 (387.1)a	<.001⁎
	Pain	4.2 (2.0)	0.9 (0.5 to 1.4)	4.3 (178.7)a	<.001⁎
	Medical practitioner	3.9 (1.9)	0.7 (0.2 to 1.1)	3.1 (394)	.002⁎
	Total EPS score
	Control	3.3 (1.5)	–	–	–
	Mental health	4.9 (1.3)	1.6 (1.3 to 1.8)	12.3 (408.7)a	<.001⁎
	Pain	3.9 (1.6)	0.5 (0.2 to 0.9)	3.2 (422)	.001⁎
	Medical practitioner	3.9 (1.5)	0.6 (0.3 to 1.0)	3.4 (394)	.001⁎

The sample size differs from the total sample of participants due to missing data across subscales.

a Levene's test suggested that variances were not equal; modified t test was used. ⁎ Significant at a Bonferroni-adjusted critical P value of .05/3 to allow for three pairwise comparisons.

Study 3: Test–retest reliability 

Discussion 

This short communication describes the refinement of the Emotional Processing Scale from a 38-item, eight-factor structure to a 25-item, five-factor structure (see Baker et al. [1] for a detailed description of the emotional processing model). Although new items were devised in an attempt to enhance the Externalized factor, the factor analysis did not support its retention. Three of the five factors relate to emotional control or dysregulation (Avoidance, Suppression, Unregulated emotion). A new factor (Impoverished emotional experience) emerged that incorporates items from two previous factors (Discordant and Dissociation) from the preliminary eight-factor structure. This factor captures some aspects of the alexithymia construct [14]. The Signs of unprocessed emotion factor relates to Rachman's conceptualization of the concomitants of inadequate emotional processing (persistent, intrusive emotional phenomena) [2] and is similar to the original Intrusion factor. Overall, internal reliability for the scale was high (α=.92) and ranged from moderate to high for individual subscales.

EPS-25 scores significantly distinguished a healthy control group from mental health, pain, and medical practitioner groups. It should be acknowledged that a potential limitation of the item selection procedure is that we brought back two items that we had previously discarded (due to possessing loadings <.4). However, we felt that the advantages of having equally weighted factors justified this decision.

To conclude, we hope that a refined and shortened version of the EPS will confer benefits clinically, and from a research perspective, in terms of ease of administration, scoring, and completion. Further psychometric evaluation on new samples using confirmatory factor analysis is ongoing.