G. Peeters / Causal and Dispositional Inferences CAUSAL OR DISPOSITIONAL INFERENCES FROM VERBS/RELATIONS: Overview of Some Published and Unpublished Material Relevant for the Issue of Implicit Causality in Verbs Guido Peeters Catholic University of Leuven Center for Social and Cultural Psychology (CSCP) Technical Report April 2010 Abstract. Some published and unpublished research from the CSCP (former LESP) that is relevant for the understanding of implicit causality in verbs, is reviewed and discussed. The reviewed research deals with the formation of impressions of personality on the basis of information about interpersonal relations between the target persons of whom impressions are formed. The obtained impressions often do not fit current theory and research on causality in verbs. However, the outcomes from both research strands may be reconciled if it is assumed that (a) both impressions of personality as well as causal inferences from verbs are consistent with the covariation principle of attribution, and (b) perceived covariation is reversed when verbs or interpersonal relations from which inferences are drawn, are embedded in a cognitively inconsistent or "imbalanced" configuration of information. Author' s address: C.S.C.P. Tiensestraat, 102, B-3000 Leuven, Belgium. Email: guido.peeters@psy.kuleuven.be 1 G. Peeters / Causal and Dispositional Inferences CAUSAL OR DISPOSITIONAL INFERENCES FROM VERBS/RELATIONS: Overview of Some Published and Unpublished Material Relevant for the Issue of Implicit Causality in Verbs The aim of this paper is to compile some ideas and outcomes from my and my students' research that may be relevant for an issue that commonly has been referred to as the issue of causal inference from verbs. However, before going into the issue, some conceptual ambiguity should be resolved. Ambiguity resides in the observation that the term "causal inference" refers to a cognitive category, while the term "verb" refers to a linguistic category. Hence, the expression "causal inferences from verbs" should be understood as "causal inferences from cognitive contents designated by verbs." It follows that the relevant research literature may not be limited to literature focussing on verbs. Any theory or research dealing with causal inferences from perceptual or cognitive contents may be relevant, at least as far as the perceptual or cognitive contents in question can be conceived as possible referents of verbs. In this way some of my research on inferences from interpersonal relations may be relevant. Indeed, as it has been observed by Chomsky (quoted by Bever, 1970), the noun-verb distinction may reflect a universal distinction between entities and relations between the entities as basic units of cognitive organization. Hence, in the following, the terms verb and relation are used quasi interchangeably. Also the terms "subject" (agens or agent) and "object" (patiens or patient) refer to a linguistic context. Considering that relations (verbs) can be formalized as vectors with an origin and a terminal, the linguistic "subject" and "object" correspond to cognitive entities associated with respectively the origin and the terminal of a relation. Hence the terms "origin entity" and "terminal entity" in my papers on relational information processing can be considered equivalent to "subject" and "object" respectively. 2 G. Peeters / Causal and Dispositional Inferences §1. Some Serendipidity from Research on Personality Inferences from Relations: §1.1. Are trait inferences from verbs or relations "causal" by default? Being intrigued by the impressive literature on the formation of full-fledged impressions of a target person's personality on the basis minimal information about some personality traits attributed to the target, I wondered if impressions of personality could not be formed as well on the basis of information about interpersonal relations between the target persons. Thus, I asked participants to form impressions of A and B who were presented in one condition as two people who like each other, and in another condition as two people who dislike each other (Peeters, 1976). As mentioned, the main aim of the study (not relevant for present purposes) was to investigate the formation of impressions of A and B's personalities on the basis of information about liking/disliking relations between A and B. However, for the sake of generalizability several additional manipulations (between participants) were involved; among which there was an explicit causality manipulation involving two conditions: - No explicit causality condition: "A and B like/dislike each other" - Explicit causality condition: "A and B like/dislike each other "because of their characters" The results showed a huge effect of the manipulation of the liking/disliking relations, but no apparent effect of the explicit causality manipulation. This absence of any effect of explicit causality suggests that trait inferences from like-dislike relations are implicitly handled as causal inferences. Apparently trait inferences from verbs/relations are causal inferences by default §1.2 Content and direction (to subject or object) of inferences. The next study on personality inferences from relations involved an extended and modified replication of the previous study (Peeters, 1983). Again trait inferences were drawn from affective relations between A and B, but no manipulation of explicit causality was involved. 3 G. Peeters / Causal and Dispositional Inferences A complex design was used of which only four conditions are relevant to the present purposes. These relevant conditions formed a 2 by 2 factorial design involving the following factors: 2 (AB Relation: 'A likes B', 'A dislikes B') X 2 (BA Relation: 'B likes A', 'B dislikes A') Notice that, in addition to the symmetric relationships used in the previous study, the present design involved also asymmetric relationships (e.g., A liking B while B disliking A). The use of asymmetric relationships enabled to distinguish between inferences to the subject (agent) and inferences to the object (patient). The results showed again strong relation effects, though not all of them being relevant to the present purposes. In order to explain the relevant findings, we should first elaborate briefly on the nature of the traits that were inferred. In the literature on social perception and cognition two trait dimensions have been advanced in different research contexts and carrying different labels (e.g., see Abele, Cuddy, Judd, & Yzerbyt, 2008). In my research I have labelled them OP or "Other Profitability": +OP (likeability, warmth, communion, social evaluation, etc.) versus -OP (dislikeability etc.) on the one hand, and SP or "Self Profitability": +SP (power, competence, agency, intellectual evaluation, etc.) versus -SP (powerlessness etc.) on the other hand. Notice that there is some semantic correspondence between OP value of traits and interpersonal like/dislike relations. For instance, the verb "Liking" is clearly related to the adjective "likeable" that belongs to the OP dimension. Hence, inferring OP saturated traits from liking/disliking relations would be a case of what attribution theorists have called "correspondent inferences" (Jones & Davis, 1965). A most obvious expectation was that participants would draw correspondent inferences. For instance, one could expect that A and B who like each other would be perceived as friendly, warm and likeable (+OP), and that A and B who dislike each other would be perceived as unfriendly, cold, and dislikeable (-OP). When A likes B while B dislikes A, one could expect that the likeability associated with liking and the dislikeability associated with disliking would either cancel each other or show a negativity effect resulting in -OP impressions of A and B. Alternatively, considering that "liking" is a "state verb" (Semin & Fiedler, 1991)--even an "experiencer-stimulus" state verb (Brown & Fish, 1983)--the literature on causal inferences from verbs would allow to expect 4 G. Peeters / Causal and Dispositional Inferences correspondent inferences to the object (patient). It would make that if A likes B while B dislikes A, B should be perceived as +OP (likeable) and A as -OP (dislikeable). However the results showed a quite different and non-obvious pattern. In spite of the fact that liking is a state verb, stronger inferences to the subject than to the object (patient) were obtained. Moreover, inference contents differed between subject and object: - Correspondent OP (likeability, warmth, communion, social evaluation) was attributed to the subject --as if the liker of an "other" would be likeable, and the disliker of an "other" would be dislikeable. - Non-correspondent SP (power, competence, agency, intellectual evaluation) was attributed to the object--as if the liked target is powerless and the disliked target is powerful Conclusion: OP seems associated with the origin (source, subject), SP with the terminal (target, object) of affective relations (liking/disliking, solidarity relations, communal relations, etc.). These associations are non-obvious. One could wonder why. A reasonable explanation may be that in real life the difference between origin (subject) and terminal (object) inferences is masked because affective relations are expected to be symmetric (reciprocal) --cf. Brown (1965). Only when participants were forced to draw inferences from asymmetric affective relationships, the given pattern of differential associations of OP and SP with subject and object was revealed. §1.3. Parenthesis on a replication study highlighting roles of grammatical and logical subject A largely extended and modified replication of the study discussed in §1.2. (Peeters 1983) was run by Peeters & Wyer (1985). The aim was to investigate whether a particular cognitive bias (obtained by Peeters, 1983) could be overcome by priming. This aim was not relevant to the present purposes. Let it just be mentioned that the expected priming effect was not obtained because participants managed to interpret even the primes in the given biased way. 5 G. Peeters / Causal and Dispositional Inferences A relevant outcome was that, as far as a comparison was possible, the new data were, in general, compatible with the former data from Peeters (1983). There was, however, a most interesting exception that may shed light on the relevance of the distinction between grammatical and cognitive roles associated with the sentence subject. In the sentence "A helps B", A functions as a grammatical subject and also as a cognitive "origin entity" associated with the origin of the "helping" relation from A to B. In the sentence "B is helped by A", it is no longer A, but B, that functions as the grammatical subject. However, on the cognitive level, A still functions as the origin entity associated with the origin of the "helping" relation from A to B. In Peeters & Wyer (1985) the latter cognitive function of A was termed "logical subject." Thus in the active sentence "A helps B" both the grammatical and the logical subject coincide in A, but in the passive sentence "B is helped by A", grammatical and logical subject are separated, B being the grammatical and A the logical subject. Evidence was obtained that participants tended to distribute inferences across the grammatical and the logical subject. Thus the sentence "A helps B" triggered attributions to A, and the sentence "B is helped by A" triggered attributions to both A (logical subject) and B (grammatical subject). Unfortunately the study did not focus on the nature or content of the attributions. Participants were not asked to produce personality descriptions of A and B. So we have only a suggestive outcome that may deserve attention, but it is not further dealt with in the following paragraphs. Let it just be mentioned that also in the literature on causality in verbs, the possible impact of active versus passive form has been considered (e.g., the salience hypothesis of Kasof & Lee discussed by Rudolph & Försterling, 1997), but, according to Rudolph & Försterling (1997) that impact would be very limited. §1.4 Conclusions from Serendipity Trait inferences from verbs/relations may be causal inferences by default. Correspondent "OP" trait inferences from liking versus disliking relations seem predominantly subject inferences rather than object inferences. The object inferences seem less prominent and belonging to a non-correspondent "SP" dimension. 6 G. Peeters / Causal and Dispositional Inferences §2. Validation and Extension of the Serendipitous Findings In the following a series of experiments are summarized from a thesis by Demol (1999). The aim was to check the validity of the 1983 findings by having more verbs than just liking and disliking as stimuli. Analogously to the OP and SP classes of traits, two classes of verbs/relations were used: (1) Affective relations (e.g., liking-disliking) that are correspondent to OP. They correspond to the class of "solidarity relations" (Brown, 1965) that tend to be perceived as symmetric (or reciprocal: if A likes B, perceivers assume that B likes A) (2) Status relations (e.g., influencing, hierarchical relations, power relations, dominance, etc.) that are correspondent to SP. Brown (1965) has reviewed evidence that status relations tend to be perceived as asymmetric (if A influences B, perceivers assume that B doesn't influence A). As extensive descriptions of the experiments are available in Demol (1999), only schematic outlines of hypotheses and outcomes are reported, but a substantial general discussion is added.. In the presentation, I use concepts and terminology from the revised ASD (Action State Differentiation) model, which is an extension of Brown & Fish's (1983) model by Rudolph & Försterling (1997). The model involves the following distinction between verbs (based on semantic roles). - ES (Experiencer Stimulus) verbs that correspond to the State Verbs (SV) of the LCM (Linguistic Category Model) of Semin & Fiedler (1991; see also Semin & Marsman, 1994); e.g.: A (experiencer) likes B (stimulus). - SE (Stimulus Experiencer) verbs that correspond to the State Action Verbs (SAV) of the LCM. E.g.: A (stimulus) impresses object B (experiencer) - AP (Agent Patient) verbs, e.g.: A (agent) kicks B (patient), A (agent) helps B (patient). - AE (Agent Evocator) verbs, e.g.: A (agent) punishes B (evocator of the punishment). 7 G. Peeters / Causal and Dispositional Inferences Notice that AP verbs and AE verbs are Action Verbs. Action Verbs (AV) are also dealt with in the LCM model that distinguishes between Interpretative Action Verbs (IAV, e.g.: A helps B) and Descriptive Action Verbs (DAV, e.g.: A kicks B). §2.1. Experiment 1: Extended and modified replication of Peeters (1983) Hypotheses (H): 1 - H(ypothesis) 1: Peeters (1983) will be replicated for affective relations => (H1a) correspondent OP attributed to origin (logical subject) => (H1b) correspondent SP attributed to terminal (logical object) 2 - H(ypothesis) 2: Reverse outcomes are expected for status relations => (H2a) correspondent SP attributed to origin (logical subject), => (H2b) correspondent OP attributed to terminal (logicalobject) Rationale: A influencing B implies sufficient power (SP) in A to influence B (=H2a), and/or leniency in B making B willing to comply with another's requests (=H2b) Results: - Hypotheses confirmed - Additional findings (= referred to as new future hypotheses H3 and H4) => (H3) Correspondent OP also attributed to object of affective/solidarity verb (If A (dis)likes B, correspondent likeableness not only attributed to A but also to B) => (H4) Reverse SP also attributed to object of status verb (If A influences B, then low power attributed to B--may be due to the fact that in the instructions to the participants, the influence relation was formulated using passive form: "B does what A wants") §2.2. Experiments 2-5: Partial and modified replications of experiment 1 8 G. Peeters / Causal and Dispositional Inferences Experiments 2-5 are in point of fact secondary analyses of partial data from another study (Peeters, 2005) on effects of symmetry and asymmetry of relations--not at stake in the present presentation. §2.2.1 Experiment 2. Verbs (relations) used: - likes/dislikes (vindt sympathiek/heeft hekel aan) Nature of the verbs (relations) according to the literature: - Solidarity relation (Brown, 1965) -- tends towards symmetry (implies that correspondent subject attributions and correspondent object attributions are similar) - State verb of the ES (experiencer-stimulus) type with implicit causality in the object (according to Rudolph & Försterling, 1997). This causality is in line with H1b and/or H3, but not with H1a. And so neither with Peeters (1983) nor with part of the above Experiment 1. Results => (H1a) CONFIRMED correspondent OP attributed to origin (subject) => (H1b) CONFIRMED BUT NOT SIGN.. correspondent SP attributed to terminal (object) => (H3) CONFIRMED Correspondent OP also attributed to object of affective/solidarity verb (If A (dis)likes B, correspondent likeableness not only attributed to A but also to B) The confirmation of H3 fits Rudolph & Försterling. However the confirmation of H1a and the lack of significant evidence for H1b do not so; they suggest a subject inference bias.. §2.2.2 Experiment 3 Verbs (relations) used: - encourage/discourage (bemoedigen/ontmoedigen) Nature of the verb (relation) according to the literature: 9 G. Peeters / Causal and Dispositional Inferences - Solidarity relation (Brown, 1965) -- tends towards symmetry (implies that correspondent subject attributions and correspondent object attributions are similar) - State verb of the SE (stimulus-experiencer) type with implicit causality in the subject (according to Rudolph & Försterling, 1997). This is in line with H1a but not with H1b and H3.. Results => (H1a) CONFIRMED correspondent OP associated with origin (subject) => (H1b) NOT CONFIRMED SP not significantly attributed to terminal (object) => (H3) NOT CONFIRMED No evidence of correspondent OP attributed to object of the given affective/solidarity verb (If A (dis/en)courages B, no significant effect on likeableness attributed to B) The confirmation of H1a and disconfirmation of H1b and H3 fits Rudolph & Försterling. §2.2.3 Experiment 4 Verbs (relations) used: - logical object has/hasn't him- or herself influenced by the logical subject (Object laat zich/ laat zich niet beinvloeden door subject-- Note that the logical object is the grammatical subject and presented as the experiencer, the logical subject or grammatical object being the stimulus that causes the experience ) Nature of the verb (relation) according to the literature: - Status relation (Brown, 1965) -- tends towards asymmetry (implies that correspondent subject attributions and correspondent object attributions are mutually opposed, e.g. if the one is powerful and dominant, the other may be weak and submissive. In order to remind that, we use the term "correspondent reverse attributions") - It is not immediately clear how Rudolph and Försterling (1997) would categorize this verb. At a first glance one might consider that the verb is a state verb of the ES (experiencerstimulus) type with implicit causality in the (grammatical) object (Rudolph & Försterling, 1997). However, the grammatical object is the logical subject, and the grammatical subject is the logical object, which would mean that the verb would be of the SE type. Assuming that 10 G. Peeters / Causal and Dispositional Inferences Rudolph & Försterling's expectation deals with subject and object as semantic roles, and that semantic roles reflect logical rather than grammatical functions, we assume that Rudolph & Försterling would expect confirmation of H2a but not of H2b and H4. Results => (H2a) CONFIRMED correspondent SP attributed to origin (logical subject though grammatical object) => (H2b CONFIRMED correspondent OP attributed to terminal (logical object though grammatical subject) => (H4) CONFIRMED correspondent reverse SP attributed to terminal (logical object though grammatical subject) The confirmation of H2a (grammatical object inference) fits Rudolph & Försterling, but the confirmation of H2b and H4 does not so. §2.2.4 Experiment 5 Verbs (relations): - subject does/doesn't exert influence on object (heeft invloed op/heeft geen invloed op-assumed to be status verbs with implicit causality in the subject) Nature of the verb (relation) according to the literature: - Status relation (Brown, 1965) -- tends towards asymmetry (implies that correspondent subject attributions and correspondent object attributions are mutually opposed. In order to remind that, we use the term "correspondent reverse attributions") - State verb of the SE (stimulus-experiencer) type with implicit causality in the (grammatical and logical) subject (Rudolph & Försterling, 1997). This is in line with H2a but not with H2b and H4. Results => (H2a) CONFIRMED correspondent SP attributed to origin (subject) => (H2b CONFIRMED correspondent OP attributed to terminal (object) 11 G. Peeters / Causal and Dispositional Inferences => (H4) CONFIRMED correspondent reverse SP attributed to terminal (object) The confirmation of H2a (grammatical object inference) fits Rudolph & Försterling, but the confirmation of H2b and H4 does not so. §3. General Discussion Semin & Marsman (1994) proposed two independent types of causality implicit in interpersonal verbs: event instigation and dispositional inference, dispositional inferences being specified as inferences about personality. The research reviewed in this paper concerns trait inferences from verbs, which means: dispositional inferences rather than event instigation. In agreement with Semin & Marsman's studies, we found huge evidence of subject inferences, and less compelling (more variable) evidence of object inferences, which argues for a subject inference bias of dispositional inferences. Indeed, the reviewed evidence is in line with a strong subject inference bias for trait attributions, though also object inferences of traits were observed, but they were more variable and less consistent across studies. §3.1. The subject inference bias of dispositional inferences Rudolph & Försterling (1997) ignored the subject inference bias of dispositional or personality inferences. A bias is a constant, and Rudolph & Försterling considered causal and dispositional inferences only as far as they vary as a function of variables such as verb types, acitve/passive voice, sentence context, subject/object reference of adjectives derived from the verbs, etc. They concluded that variation of implicit causality, including dispositional inferences as far as they varied, could be accounted for by Kelley's (1967) covariation model. Perceived or assumed covariation between verb and subject involves subject inferences, and percieved or assumed covariation between verb and object involves object inferences. However, they ignored the constant subject inference bias of dispositional (personality) inferences superimposed to the condition-bound variations of subject- and object inferences. 12 G. Peeters / Causal and Dispositional Inferences The reasons why Rudolph & Försterling (1997) ignored the subject inference bias of dispositional inferences may be twofold. First, they were primarily interested in effects of conditions and may have considered a systematic bias as irrelevant. Second, only one study (Semin & Marsman, 1995, Study 2) out of 36 reviewed studies was specified as a study involving "dispositional inference" as a dependent variable. Hence the bias might be disposed of as a chance effect. However, the subject inference bias of dispositional inferences cannot be disposed of as a chance effect because it has shown up in studies on the formation of impressions of personalities of targets as a function of information about interpersonal relations between the targets. For instance, as reported above, Peeters (1983) obtained personality impressions formed of targets A and B on the basis of interpersonal liking and disliking relations between A and B. Liking and disliking are state verbs with the subject in the role of experiencer and the object in the role of stimulus (ES verbs of Brown & Fish, 1983 and Rudolph & Försterling, 1997). Hence, if asked why "A likes B" respondents may answer "B because B is likeable." However when participants were asked to attribute traits to A and B, they tended to attribute likeable traits to A rather than to B, and, to a lesser extent, power-related traits to B-as if participants considered that "power rather than leniency is attractive." In order to explain the discrepancy of dispositional inference data with current theory and research on causality in verbs, one might consider that dispositional trait inferences should not be dealt with as causal inferences. However this explanation seems contradicted by the observation that no difference was obtained between a condition where perceivers where informed that given interpersonal liking and disliking relations were caused by the traits they had to infer and a control condition without any suggestion about the causal role of the traits (Peeters, 1976--see §1.1 above)). A weaker version of this explanation may build on Semin & Marsman (1995) who dealt with dispositional or "personality" inferences as a separate class of causal inferences, to be distinguished from other causal inferences called "event instigation." A main difference between both classes of inferences seems to be that the dispositional inferences involve a huge subject inference bias. However, this bias contradicts other causal inference rules as well as common sense--provided that A says that B is likeable, it seems not evident to infer that A rather than B is likeable. Hence it requires an explanation. 13 G. Peeters / Causal and Dispositional Inferences Peeters (1983) explained the surprising character of the obtained subject inference bias in that subject inferences were misinterpreted as object inferences because perceivers go beyond the information given in a way as to confound subject and object roles. For instance, liking/disliking relations tend to be perceived as reciprocal. If A likes B, perceivers assume that B likes A in turn. Thus, if likeable traits are attributed to B, it may be because B is the subject in assumed "B likes A", and not at all because B is the object of A's liking. Consistently Peeters (1983) found that if "A likes B", likeable traits are attributed to B only if B likes A in turn. The present explanation may account fairly well for the surprising character of particular manifestations of the apparent subject inference bias of dispositional inferences. However it does not yet account for the apparent subject inference bias itself. §3.2 Towards an explanation: The role of cognitive imbalance and the covariation principle In an impressive review and reanalysis of the extant literature, Rudolph & Försterling (1997) demonstrated quite convincingly that causality implicit in verbs varied consistent with the basic attribution mechanism of perceived covariation, particularly covariation with low vs. high consensus and covariation with high vs. low distinctiveness (Kelley, 1967; see Rudolph & Försterling's article for a suited explanation). As mentioned yet, Rudolph & Försterling ignored the subject inference bias of dispositional inferences. However, in the following, I will demonstrate that the covariation principle may account for dispositional inferences included the ones attributed to a subject inference bias. Therefore we proceed from more recent work of Rudolph on implicit verb causality, particularly on the role of imbalance repair (Rudolph & von Hecker, 2006). Cognitive balance theory (Heider, 1958) was a pioneering elaboration of the cognitive consistency concept that dominated social psychology during the sixties of the 20th century. Hence in Rudolph & Hecker's article the term "balance" refers to the basic concept of cognitive consistency rather than to the specific elaboration of cognitive consistency that has become associated with the label "balance theory." Humans tend towards cognitive balance, which implies that they tend also to "repair" imbalance. Proceeding from Brown & Van Kleeck (1989), who advanced "imbalance repair" as an explanatory principle, Rudolph & von 14 G. Peeters / Causal and Dispositional Inferences Hecker (2006) investigated the role of imbalance repair in causality implicit in verbs. Specifically, they investigated whether imbalance repair of unbalanced cause-effect relationships would affect perceived covariation and hence perceived causality. For instance, "A likes B" implies perceived causality in B. Consistently it has been found that the ES verb "likes" is characterized by high perceived consensus (many people are expected to like B) and high distinctiveness (B is one of few people liked by A). The balance concept implies that the cause is evaluatively correspondent with the verb. E.g.: A may be expected to like B because B is nice. "A liking B because B is arrogant" would be an imbalanced statement because the valence of the cause is inconsistent with that of the verb. In this case, balance repair is achieved by a reversal of perceived covariation: few people like arrogant guys (low consensus) and A seems someone who likes arrogant people in general (low distinctiveness). Consistently causality is situated in A (A has unusual tastes and preferences that make him feeling attracted to arrogant people). The effect of imbalance repair Rudolph & von Hecker observed for cause-effect unbalance, may extend to other forms of imbalance as well. The designs used in the reviewed experiments that yielded the apparent subject inference bias of dispositional inferences all involved both balanced and imbalinced conditions. However the imbalance did not concern unbalanced cause-effect relationships but the violation of expectations about the going together of particular interpersonal relationships. In the following we shall have a look at the relations (verbs) used in the various experiment reviewed in this paper. As it was explained above, the verbs used represented two classes of interpersonal relations according to Brown (1965): solidarity relations (tending to be perceived as symmetric) and status relations (tending to be perceived as asymmetric). (1) Solidarity relations Solidarity relations tend to be perceived as symmetric (if A likes B, perceivers expect that B likes A); hence asymmetric solidarity relations (A likes B but B dislikes A) are cognitively inconsistent of "unbalanced." The designs used in the reviewed experiments involved both symmetric (balanced) and asymmetric (unbalanced) relationships And as it was mentioned yet, unexpected outcomes such as the apparent subject inference bias were owed to the unbalanced conditions. Hence we shall have a closer look at possible perceived covariations 15 G. Peeters / Causal and Dispositional Inferences of unbalanced liking relations as a case in point of unbalanced solidarity relations in general and consider the related causal attributions. Given that A likes B while B dislikes A, one may expect low consensus and high distinctiveness: (a) Low consensus. Perceivers would not expect that people would like someone by whom they are disliked. So there must be something special to A. For instance, A may be an extremely nice "+OP" person, which fits H1a. In addition, People do not expect that people would dislike someone who likes them. So there must be something special to B too. For instance B may be an extremely nasty "-OP" person, which also fits hypothesis H1a. (b) High distinctiveness. Perceivers would not expect that A is a masochist who likes everyone by whom he is disliked. Thus B may be an exception that has properties that make him attractive to A in spite of his rejection of A. We may exclude OP or likeability from those properties, because the low consensus suggests that B is rather -OP or dislikeable, and A may not like B because B is dislikeable. So B must have other attractive properties, and, considering that power attracts (power is sexy), an obvious candidate is +SP, which fits hypothesis H1b. By an analogous rationale we may attribute -SP to A: B is not expected to dislike others by whom he's liked, but A is an exception because A is a weakling, and also that rationale fits hypothesis H1b. Nevertheless, it may happen sometimes that perceivers ignore the above "low consensus" implications, which leaves room for correspondent OP attributions to the object (hypothesis H3) that sporadically were observed. (2) Status Relations Status relations tend to be perceived as asymmetric (if A influences B, perceivers expect that B does not influence A); hence symmetric status relations are cognitively inconsistent or "unbalanced." In the following we have a closer look at possible covariations (and related causality) of unbalanced influencing relations as a case in point of status relations in general Given that A influences B while also B influences A in turn, one may expect low consensus and high distinctiveness: 16 G. Peeters / Causal and Dispositional Inferences (a) Low consensus. Perceivers may expect that only few people are able to exert influence on someone by whom they are influenced themselves. Hence A must be extremely influential, and so must be B, because both are able to influence an influential other. Being influential means having power, which means: +SP, and this fits hypothesis H2a. (b) High distinctiveness. Perceivers may not expect that highly influential people give in to the influence of another person. Hence A and B's willingness to accept another's influence implies a special attribute. One possibility might be that A and B are weaklings, which actually happened to the extent that hypothesis H4 was confirmed. However, that possibility can be ruled out by the above implications of the low consensus rule. Hence an alternative explanation may be that A and B are open-minded and benevolent "+OP" persons who are prepared to meet other people's wishes, and this fits hypothesis H2b. We may conclude that imbalance repair and covariation principles may provide an avenue towards the explanation of some intriguing findings, particularly the apparent subject inference bias of dispositional inferences. However, this conclusion raises a problem that may not bother the present scientific community, but that bothers me. Namely, the conclusion seems incompatible with some of my earlier work on cognitive programs. In the study that serendipitously revealed the subject inference bias of dispositional inferences (Peeters, 1983), I demonstrated that inferences can be drawn from relations consistent with two independent cognitive programs called the Self-Other (SO) Program and the Third Person (3P) Program (for a review about cognitive SO and 3P programs, see Peeters, 2004). Personality inferences were found to be drawn exclusively consistent with the SO program. This means that in order to draw inferences from relations between A and B, perceivers conceive of the relations as relations between self and other. For instance, given the relation "A helps B out of a problem", perceivers may infer that A is generous only if they take into account that A is helping an "other" person. No generosity would be inferred if A would be helping out himself. Inferences drawn consistent with the 3P program do not rely on the self-other distinction. For instance, imagine that the problem of which A helps B out is a mathematical problem. Perceivers may attribute mathematical competence to A irrespective of whether A is helping out another person or himself. The connection of personality inferences with the SO program has consequences for the Correspondent Inference theory of causal attribution (Jones & Davis, 1965). Specifically, 17 G. Peeters / Causal and Dispositional Inferences Jones and Davis seem to have conceived their theory consistent with the SO program, whereby they may have sticked to the "other" part of the program, which means: to relations with the "other." For instance, associating helping with generosity, classic correspondent inference theory assumes implicitly that "helping" is conceived as "helping others" rather than as "helping oneself out." In this way, correspondent inference theory may be improved by fully taking into account the implications of the SO program. Considering that correspondent inference theory is an attribution theory shaped consistent with the SO program, it has been advanced that the covariation principle would be its 3P shaped counterpart. Particularly Kelley's ANOVA model of causal attribution would be shaped consistent with the 3P program (Peeters & Hendricks, 2002). However, this division is challenged by the present role assigned to the covariation principle in the explanation of dispositional "personality" inferences. Apparently covariation is relevant to inferences consistent with the SO program as well. However, it is to be taken into account that within the scope of the SO program variation is limited to only two alternative entities: self and other. For instances, "helping out" varies only between "helping out the self" and "helping out the other." OP dispositions (generosity etc.) are bound to "helping out the other" and SP dispositions (self-confidence etc.) to helping out the self. Any other possible variation, such as about the nature of the problem (mathematical, technical, social, etc.) of which the self or the other is helped out, involves the 3P-program. REFERENCES Abele, A. E., Cuddy, A. J. C., Judd, C.M., & Yzerbyt, V. Y., Eds (2008). Fundamental dimensions of social judgment. Special Issue. European Journal of Social Psychology, 38-7, 1063-1224. Bever, T.G. (1970). The cognitive basis of linguistic structure. In J.R. Hayes (Ed.), Cognition and the development of language. New York: Wiley. Brown, R.'1965). Social Psychology. NewYork: The Free Press. 18 G. Peeters / Causal and Dispositional Inferences Brown, R., & Fish, D. (1983). The psychological causality implicit in language. Cognition, 14, 237-273. Brown, R., & Van Kleeck, M.H. (1989).Enough said: Three principles of explanation. Journal of Personality and Social Psychology, 57, 590-604. Demol, V. (1999). Persoonlijkheidsinferenties uit Interpersoonlijke Relaties: Is er een verschil tussen Affectieve en Statusrelaties? - Licentiaatsverhandeling, K.U.Leuven: LESP. Heider, F. (1958). The psychology of interpersonal relations. New York: Wiley Jones, E. E., & Davis, K. E. (1965). From acts to dispositions: The attribution process in person perception. In L. Berkowitz (Ed.), Advances in experimental social psychology (Vol. 2, pp. 219-266). New York: Academic Press. Kelley, H. H. (1967). Attribution theory in social psychology. In D. Levine (Ed.), Nebraska symposium on motivation (Vol. 15, pp.192-238). Lincoln, NE: University of Nebraska Press. Kasof, J., & Lee, J. Y. (1993). Implicit causality as implicit salience. Journal of Personality and Social Psychology, 65, 877-891. Peeters, G. (1976). In search for schemata underlying the inference of traits from interpersonal relations. European Journal of Social Psychology, 6, 191-205. https://lirias.kuleuven.be/bitstream/123456789/128992/1/1976EJSP_Traits_From_Relations.pdf Peeters, G. (1983). Relational and informational patterns in social cognition. In W. Doise & S. Moscovici (Eds.), Current issues in European Social Psychology (pp. 201-237). Cambridge: Cambridge University Press. https://lirias.kuleuven.be/handle/123456789/166030 Peeters, G. (2004). Thinking in the third person: A mark of expertness?. Psychologica Belgica, 44, 249-267. https://lirias.kuleuven.be/bitstream/123456789/113693/1/2004PsBELexpertness.pdf 19 G. Peeters / Causal and Dispositional Inferences Peeters, G. (2005). Communion (solidarity) and power conveyed by social relations: A matter of content or structure? Extended draft of a paper presented at the 7th Meeting of the European Social Cognition Network. Vitznau, September 1-4, 2005. 34 pp. https://lirias.kuleuven.be/retrieve/28310/2005_ESCON_Relations_full+text.doc Peeters, G., & Hendrickx, A. (2002). The similarity between the perception and interpretation of information in a profane and a religious context. The International Journal for the Psychology of Religion, 12, 41-52. https://lirias.kuleuven.be/bitstream/123456789/115252/1/2002PsRELIGION.pdf Peeters, G., & Wyer, R. (1985). The selective activation of cognitive programs in the processing of relational information: The role of subjective orientation, priming, and grammatical form. (Internal Report No 1). Leuven, Belgium: K.U.Leuven, Laboratorium voor Experimentele Sociale Psychologie. 53 pp. https://lirias.kuleuven.be/retrieve/17577/1985REPORT_RPM_WYER.pdf Rudolph, U., & Försterling, F. (1997). The psychological causality implicit in verbs; A review. Psychological Bulletin, 121,192-218. Rudolph, U., & von Hecker (2006). Three principles of explanation: Verb schemas, balance, and imbalance repair. Journal of Language and Social Psychology, 25, 377-405. Semin, G. R., & Fiedler, K. (1991). The linguistic category model, its bases, application and range. In W. Stroebe & M. Hewstone (Eds.),European Review of Social Psychology, 19, 491508. Semin, G.R., & Marsman, J.G. (1994). "Multiple inference-inviting properties" of interpersonal verbs: Event instigation, dispositional inference, and implicit causality. Journal of Personality and Social Psychology, 67, 836-849. 20