Dissociative Self-Preference Model

To address the above two pivotal questions, we propose the Dissociative Self-Preference Model. Its evolutionary backbone generates divergent hypotheses for explicit and implicit preferences. Explicitly, preferences for favorite other should prevail over preferences for self. Explicit other-preferences are evolutionarily adaptive, because they convey caring for others [12], which fosters interpersonal bonds [13]. Interpersonal bonds, in turn, boost evolutionary fitness [14]. Implicitly, however, preferences for self should prevail over preferences for favorite other. Implicit self-preferences are also evolutionarily adaptive, because they facilitate self-favoring automatic behavior in live-or-die situations [15] as well as subtle and automatic self-favoritism in everyday situations [16]. Such action maximizes self-protection [17], which, in turn, boosts evolutionary fitness [14].

We report a study in which we examine the two hypotheses derived from the Dissociative Self-Preference Model. Preferences for favorite other will predominate over preferences for self at the explicit level. However, preferences for self will predominate over preferences for favorite other at the implicit level. We test these hypotheses on a large and divergent sample, using five established implicit measures for generalizability purposes.

Ethics Statement

The study was run as part of a research grant awarded by the German Research Foundation (DFG; www.dfg.de) to the second author (grant identifier: GO 1107/4-1). The DFG's board of ethics passed the research proposal that underlies the present study. DFG-funded projects do not require additional approval by other ethics committees (e.g., from universities). Irrespective, we informed participants at the outset that the study concerns the psychological representation of their favorite person and that they therefore will be asked to complete corresponding self-report measures and sorting tasks. Participants were also informed that participation is anonymous, consented to participate in this study by clicking on a button, and were shown a debriefing screen at the conclusion of the study session.

Sample and Procedure

We used a diverse internet sample in terms of age, occupation, and family status comprising 1,519 German volunteers (69% women; M_age  = 37.8 years, SD_age  = 13.3; this description does not include participants filtered by standard algorithms for the computation of implicit scores, such as the IAT's d-score, see Table 1). At the beginning of the study, participants provided the full name of “the person for whom you have the most positive feelings–the person who is most valuable and likable to you” (favorite other). Following that, we wondered whether participants had chosen themselves as their favorite other. 42 participants (3%) had done so, and we excluded their data from our analyses. Next, participants reported their own name. Subsequently, participants completed one of five implicit self-other preference measures and one explicit self-other preference measure. For each participant, we determined randomly which implicit measure (s)he had to complete. Also, for each participant, we determined randomly whether (s)he received the implicit or the explicit measure first.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Key characteristics of the five implicit measures.

https://doi.org/10.1371/journal.pone.0041789.t001

Explicit Measure

Participants responded to the following eight items on rating scales ranging from 1 =  “[own full name]” to 11 =  “[favorite other's full name]”: “For whom do you have…” (1) “…more positive feelings?”, (2) “…more negative feelings?” (reverse-scored), and “Who is more…” (3) “… worthy to you?”, (4) “…pleasant to you?”, (5) “…likable to you?”, (6) “…worthless to you?” (reverse-scored), (7) “…unpleasant to you?” (reverse-scored), and (8) “…unlikable to you” (reverse-scored) (α  = .89).

Implicit Measures

Different implicit measures may assess somewhat distinct facets of self-preference [18]. To overcome this potential limitation, we administered five established implicit measures: Implicit Association Test (IAT; [19]), Single Block IAT (SBIAT; [20]), Go-Nogo Association Task (GNAT; [21]), Affective Priming Paradigm (APP; [22]), and Name-Letter-Task (NLT; [23]).

Attitude-categories of the category-based measures (IAT, SBIAT, GNAT) were “[own full name]” and “[favorite person’s full name].” Attitude-stimuli of the stimuli-based measures (IAT, SBIAT, GNAT, APP) were idiosyncratically generated by each participant. Specifically, participants were instructed to provide “three words that clearly and spontaneously signify [favorite person’s full name], and three different words that clearly and spontaneously signify yourself.” The instructions set two additional constraints. First, the chosen words for self and favorite other had to be equivalent (e.g., initials for self and for favorite other). Second, each word had to signify distinctively one attitude-category and not the other (e.g., only applicable to self, not to favorite other). Instructions also included examples suggesting the use of first names, last names, nicknames, or initials.

Valence-categories of the category-based measures (IAT, SBIAT, GNAT) were “Positive” and “Negative.” Valence-stimuli of the stimuli-based measures (IAT, SBIAT, GNAT, APP) were “worthy,” “pleasant,” and “likable” for the “Positive” category and “worthless,” “unpleasant,” and “unlikable” for the “Negative” category (the APP additionally included three neutral distracter-stimuli). These valence-stimuli were chosen to match the items of the explicit measure.

Table 1 provides a detailed description of the key characteristics of the implicit measures (e.g., number of trials, sequence of stimuli). In the IAT, participants are instructed to sort–as fast as possible–representative items to four categories. Two of these categories are the rivalry attitude-objects (here: self vs. favorite other) and the other two are opposing valence categories (here: positive vs. negative). In half of the critical trials, participants had to press the same key for sorting correctly the “self” items as well as the “positive” items, and another key for sorting correctly the “favorite other” items as well as the “negative” items. In the other half of the critical trials participants had to press the same key for sorting correctly the “self” items as well as the “negative” items, and another key for sorting correctly the “favorite other” items as well as the “positive” items. Comparison of the reaction times between the two halves of the critical trials provides an implicit index of self-other preference. Specifically, the degree to which a person is faster at trials where “self” and “positive” (and “favorite other” and “negative”) are equated, compared to trials where “favorite other” and “positive” (and “self” and “negative”) are equated, is an indicator of self-preference over other-preference [24].

In the IAT, the two halves of critical trials are administered in separate blocks, one after the other. The SBIAT is a variant of the IAT that administers all its critical trials randomly within a single block. This can address several alternative explanations of IAT effects [20]. The GNAT also builds on the IAT, but each trial has a very narrow time window in which participants can complete the sorting task. Correspondingly, the GNAT does not capitalize on response latencies to derive preferences, but on the number of errors participants make at critical trials. Specifically, the degree to which a person makes less errors at trials where “self” and “positive” (or “favorite other” and “negative”) are equated, compared to trials where “favorite other” and “positive” (and “self” and “negative”) are equated, is an indicator of self-preference over other-preference [25].

In the APP, participants are instructed to indicate–as fast as possible–whether valence items are positive or negative. Shortly before a valence item is presented, however, an attitude item (here: from the categories “self” or “favorite other”) is presented. Presentations of valence items and attitude items are randomized, so that four types of trials emerge: self-positive trials, self-negative trials, favorite other-positive trials, and favorite other-negative trials. Comparison of the reaction times between self-positive and favorite other-negative trials with favorite other-positive and self-negative trials constitutes an implicit index of self-other preference. Specifically, the degree to which a person is faster at self-positive and favorite other-negative trials, compared to favorite other-positive and self-negative trials, is an indicator of self-preference over other-preference [22]. APP is the only implicit measure in our study that does not use attitude-categories, while nonetheless using (idiosyncratically generated) attitude-items. As such, APP is sensitive to the degree to which participants have chosen attitude-items that reflect valence per se, independently of the attitude-items' link to self/favorite other. For example, the attitude-items “anxious” or “intelligent” carry valence per se (in contrast to attitude-items such as [first name] and [initials], which were provided as examples in the instructions). Thus, within the standard algorithm to compute APP scores, we embedded the exclusion of participants who–contrary to instructions–chose attitude-items that reflected valenced traits of self or favorite other (e.g., “intelligent,” “anxious”). Note that such exclusion was unnecessary in the case of those implicit measures that utilize attitude-categories (IAT, SBIAT, GNAT), because these measures are much more robust against such confounds [26].

In the NLT, participants indicate, for all letters of the alphabet, how positive or negative they perceive them to be. The degree to which participants rate their initials more positive than the initials of their favorite other is an indicator of self-preference over other-preference [27]. In order to calculate preference indices for the implicit measures, we strictly followed past research (Table 1).

Sex and Age Differences

We tested for sex differences in explicit and implicit preferences, using a MANOVA with sex as the IV and explicit and implicit preferences as simultaneous DVs. Despite sufficiently large samples to reliably detect even small differences, we found no sex differences for explicit preferences, F(1, 1,458)  = 1.28, p  = .26, as well as for implicit preferences, F(1, 1,458)  = 1.08, p  = .30.

We also tested whether explicit and implicit preferences varied as a function of participant age. Despite sufficiently large samples to reliably detect even small effects, age was uncorrelated with explicit preferences, r(1,456)  =  -.03, p  = .23, as well as with implicit preferences, r(1,456)  = .01, p  = .66.

On the Nature of Other-Preferences

Do explicit other-preferences reflect genuine beliefs or conscious self-presentational efforts to deceive others? Additional analyses suggest the former. First, internet studies minimize self-presentation [30]. Second, many of our current participants (N₁ = 520, N₂ = 793) completed two prior studies, each including a different, well-validated measure of self-presentation (datasets were linked via anonymous key-strings). Comparison of correlation coefficients using Fisher z transformation revealed that the influence of self-presentation on implicit and explicit preferences did not differ significantly with regard to the first index, z  = .74, p  = .46 (Social Desirability Scale–17 [31]) as well as the second index, z  = .85, p  = .40 (Impression Management Scale [32]). Third, the influence of self-presentation on implicit and explicit preferences was generally low, -.08< rs < -.02. Finally, the influence of self-presentation on implicit and explicit preferences was significantly lower than the influence of self-presentation on the golden standard measure of explicit self-esteem (Rosenberg Self-Esteem Scale [33]), which we had additionally administered at the end of this study, zs  = 5.75, ps <.001 (comparison of correlation coefficients again utilized Fisher z transformation). Together, conscious self-presentational efforts to deceive others are unlikely to account substantially for other-preferences (Figure 1).