Spontaneous trait inferences (STIs) occur when an individual is attending to another person’s behavior and this produces trait inferences in the absence of explicit intention to infer traits or form an impression of that person (Uleman, Saribay, & Gonzalez, 2008). For instance, if John gets an A for a test, others tend to make a spontaneous trait inference that John is smart. That is to say, STIs usually give rise to properties of a trait being activated unconsciously, which influences an individual’s impression formation of others (Uleman, Rim, Saribay, & Kressel, 2012)

As it is difficult for STIs to exist independently, other inferences are bound to participate in individuals’ understanding of others’ behavior. When confronted with others’ behavioral information, individuals will also make spontaneous situational inferences (SSIs; Todd, Molden, Ham, & Vonk, 2011). For instance, if John gets an A for a test, others will not only make a spontaneous trait inference that John is smart, but will also make a spontaneous situational inference that the test was easy. Lupfer, Clark, Church, DePaola, and McDonald (1995) employed the probe recognition paradigm to demonstrate the existence of SSIs. After reading situation-implying sentences, their participants took longer to correctly indicate that situational (vs. control) words were not presented in previous sentences.

Researchers using the probe recognition paradigm have found that if a trait is inferred spontaneously at the stage of encoding behavioral sentences, it would be extremely hard for participants to respond that this trait was not presented in the previous sentence. In Uleman, Hon, Roman, and Moskowitz’s (1996) study, participants who read trait-implying sentences made more errors and their reaction time was longer compared with when they read control (non-trait-implying) sentences. Most previous researchers of STIs have focused solely on behavior, and ignored the actors’ features (Todorov & Uleman, 2002). However, in interpersonal communication, as individuals are sensitive to actors’ features (Freeman, Nakayama, & Ambady, 2013), some stereotypes associated with actors can be automatically activated. This may influence interpretation of actors’ behavior (Johnson, Freeman, & Pauker, 2012).

STIs and SSIs can occur quickly (Todd et al., 2011) without intention or awareness. Both are highly efficient, but not uncontrollable (Crawford, Skowronski, Stiff, & Scherer, 2007). For example, stereotype activation influences STIs (Ramos, Garcia-Marques, Hamilton, Ferreira, & Van Acker, 2012). Wigboldus, Sherman, Franzese, and van Knippenberg (2004) examined how a supraliminally or subliminally presented stereotype influences STIs. They associated the actor in behavioral sentences with social categories, which led to either stereotype-consistent or stereotype-inconsistent behavior. They found that when the behavior was stereotype-inconsistent, participants took less time to make a correct response, which meant that STIs were inhibited. As this inhibition occurred even when the stereotype labels were subliminally presented, Wigboldus and colleagues (2004) thus explored how cognitive load moderated the effect of stereotypes on STIs. Participants memorized a random five-digit number in a high cognitive load condition, and a random one-digit number in a low cognitive load condition. The experimental task was to recognize probes according to behavioral sentences. Wigboldus et al. (2004) found that whereas stereotypes did not influence STIs in the low cognitive load condition, they inhibited STIs in the high cognitive load condition. Ramos et al. (2012) added situational information into the behavioral sentences to explore the simultaneous influence of stereotypes on STIs and SSIs. Their results showed that when the actor’s behavior was stereotype-consistent, STIs were facilitated, and when the behavior was stereotype-inconsistent, SSIs were facilitated.

Ramos and colleagues (2012) thus contended that inferences were facilitated if they increased the coherence of the available information, and that inferences played a central role in this processing. However, this view neglects the effect of stereotypes. The nature of the process may be better revealed in the encoding flexibility model (Allen, Sherman, Conrey, & Stroessner, 2009), according to which, stereotype-consistent information is encoded in a conceptual or gist-based manner, whereas stereotype-inconsistent information benefits from extensive data-driven encoding. Thus, we concluded that in the impression formation process, gist-based encoding corresponds to trait inferences, whereas data-driven encoding corresponds to situational inferences (Jerónimo, Garcia-Marques, Ferreira, & Macrae, 2015). Therefore, we proposed the following hypothesis:
Hypothesis 1: When a stereotype is consistent with the actor’s behavior, spontaneous trait inferences will be facilitated, whereas when a stereotype is inconsistent with the actor’s behavior, spontaneous situational inferences will be facilitated.

Previous results have shown that subliminally activated stereotypes can also influence information processing (Moskowitz, Stone, & Childs, 2012). Further, supraliminally or subliminally activated stereotypes can influence STIs (Wigboldus, Dijksterhuis, & van Knippenberg, 2003). However, Ramos et al. (2012) examined only the influence of supraliminally activated stereotypes on STIs and SSIs. We set out to establish if the utility of stereotypes is still powerful when the influence of stereotypes on STIs and SSIs is examined simultaneously. Therefore, we proposed the following hypothesis:
Hypothesis 2: Whether stereotypes are activated supraliminally or subliminally, this will result in an identical pattern: When the actor’s behavior is stereotype-consistent, spontaneous trait inferences will be facilitated, and when the actor’s behavior is stereotype-inconsistent, spontaneous situational inferences will be facilitated.

Method

Participants

We recruited 48 undergraduate students to participate in the experiment. They comprised 20 women and 28 men ranging from 19.68 to 25.1 years (M_age = 22.39 years ±2.71), and they each received 30 RMB (US$5) for their participation.

Materials

To provide the materials for our experiment, five psychology postgraduates wrote 60 behavioral sentences, each of which contained trait and situation properties, as well as corresponding trait and situational words. In the pretest, these sentences were assessed by 60 participants, half of whom were asked to indicate on a 7-point scale (1 = not applicable, 7 = very applicable) to what extent each trait word applied to each behavioral sentence. The other half were asked to indicate on the same 7-point scale to what extent each situational word applied to each behavioral sentence. The rating scores of trait words (M = 5.47, SD = 0.97) and situational words (M = 5.51, SD = 0.94) were both significantly higher than the middle point of 4 (trait words: t(29) = 20.46, p < .001; situational words: t(29) = 21.43, p < .001), and the difference between the rating scores of trait and situational words was not significant. This result indicated that the behavioral descriptions can reflect corresponding trait and situational words and allow for both STIs and SSIs.

The six trait-implying and six situation-implying sentences that were developed from the pretest each described one specific behavior of an actor. For each sentence, two category labels were selected, one for which the behavior was stereotype-consistent, and one for which the same behavior was stereotype-inconsistent. In a pilot study, 50 participants indicated on a 7-point scale to what extent the behavior in each of the resulting 12 sentences was stereotype-inconsistent (1) or stereotype-consistent (7). Results showed that the stereotype-consistent-behavior sentences (M = 5.60, SD = 1.35) were assessed as being significantly more stereotype-consistent than the stereotype-inconsistent-behavior sentences, (M = 1.98, SD = 1.01), t(49) = 37.15, p < .001.

Each sentence was linked with three stereotype labels for the actor (consistent vs. inconsistent vs. neutral) and two probe words (trait vs. situation), namely, the hypotheses were tested in 36 experimental trials. To prevent participants from conjecturing about the aim of the experiment, and to avoid any kind of association between the type of sentence and a correct response, 36 filler trials consisting of other sentences and probe words were also presented.

Procedure

We divided participants randomly into two groups, and administered the experiment using E-Prime software. They were seated in a laboratory with each of them facing a computer screen, and completed the experiment individually. Participants were then presented with one of the two versions of the experiment. All materials in each version were identical, but the stereotype labels in the behavioral sentences were presented in different ways. We then described the experiment to participants as a study on reading speed. We told them that they would see a series of sentences with the words printed in black, and that, following each sentence, a word would appear on the screen in blue. Participants’ task was to indicate, as rapidly and accurately as possible, whether the word was included in the sentence that they had just seen. Participants were instructed to press the “F” key if they thought that the word was in the preceding sentence, and if not, press the “J” key. Participants were told to keep their index fingers on the keyboard throughout the experiment.

In the supraliminal version of the experiment, each sentence was presented in the center of the screen for 1,500ms, after which a blank screen was shown for 500ms. The probe word then appeared in the center of the screen until participants responded, after which a blank screen appeared for 1,000ms to discriminate between trials. There were eight practice trials before the experiment began. Sentences in the practice trials had a structure similar to the experimental sentences, but with neutral content. The stereotype labels in these sentences were not used in the experimental sentences. Participants could practice repeatedly until they learned how to respond correctly. After the practice trials, participants were informed in an introduction that the experiment would begin. The first three trials were also filler trials to prevent startup problems. All requirements in the practice and the real experiment were the same, and the experimental stimulus was presented in a random order. Error rates and reaction time were recorded. The flow chart of the experiment process is set out in Figure 1.

In the subliminal experiment, we divided the behavioral sentences into two parts: actors’ stereotype labels and behavior. After a mask XXXXX of 500ms was presented, the stereotype labels were presented subliminally for 17ms (Wang, Huo, & Wang, 2012), followed by a blank screen for 500ms, after which each sentence was presented for 1,500ms. The actor in each sentence was represented in the text by an X (e.g., X helped an old neighbor to carry luggage to the railway station.). The probe word then appeared in the center of the screen until participants responded. A blank screen appeared for 1,000ms to discriminate between trials. The flow chart of the experiment process is set out in Figure 2.

Results

We used the statistical method previously used by Ramos et al. (2012) to perform separate analyses for STIs and SSIs. As we were investigating how subliminal and supraliminal stereotype labels influenced spontaneous inferences, we have presented our statistical analysis results in two conditions. We examined separately the influence of subliminally and supraliminally activated stereotypes on STIs and SSIs.

Error Rates

As error rates were highly skewed, a square root transformation was performed to reduce skew (Cohen, Cohen, West, & Aiken, 2003). We removed three participants’ results because the error rates were so high (over 30%).

In the supraliminal experiment, we analyzed the error proportions for both trait and situational probes in a one-way repeated measures analysis of variance (ANOVA) with three levels (stereotype label: consistent vs. inconsistent vs. neutral). The subliminal experiment results were similar to those in the supraliminal experiment. The main effect of stereotype consistency was not significant for either trait or situational probes. That is, error rates for the trait-detection and situation-detection words in the consistent, inconsistent, and neutral stereotype conditions were not statistically significant, whether the stereotype was activated supraliminally or subliminally.

Response Time

Outliers were defined as responses that were less than 200ms or greater than 3,000ms. They represented 12% of the experiments and were eliminated from the analysis. We also excluded time for wrong responses, which represented 7.75% of the experiments.

In the supraliminal experiment, we performed a one-way repeated measures ANOVA with three levels (stereotype label: consistent vs. inconsistent vs. neutral) for correct response times. For trait probes the effect was not significant. However, for situational probes, the effect was significant, F(2,21) = 6.11, p < .01, η² = .22. Results of planned comparisons revealed that participants took longer to correctly reject situational probes in the stereotype-inconsistent condition (M = 939.86, SD = 257.68) than in both stereotype-consistent conditions, M = 845.94, SD = 202.59, t(22) = 2.83, p = .01, d = 0.59, and the neutral condition, M = 851.69, SD = 163.19, t(22) = 2.64, p < .05, d = 0.55. The correct response times in the stereotype-consistent and neutral conditions did not differ significantly. Comparing average response times for trait and situational probes, we found that the response times for situational probes (M = 879.16, SD = 207.82) were significantly longer than the response times for trait probes, M = 802.95, SD = 164.26, F(1, 22) = 12.51, p < .01, η² = .36.

The subliminal experiment results were identical to the supraliminal experiment results. We performed a one-way repeated measures ANOVA with three levels (stereotype label: consistent vs. inconsistent vs. neutral) for the correct response times. The ANOVA results were not significant for trait probes. However, for situational probes, the within-subjects ANOVA (stereotype label: consistent vs. inconsistent vs. neutral) was significant, F(2, 21) = 4.58, p < .05, η² = .17. Results of planned comparisons revealed that participants took longer to correctly reject situational probes in the stereotype-inconsistent condition M = 1034.38, SD = 369.20, than in either the stereotype-consistent condition, M = 920.80, SD = 238.23, t(22) = 2.53, p < .05, d = 0.53, or neutral condition, M = 963.74, SD = 302.04, t(22) = 2.22, p < .05, d = 0.46. The correct response times in the stereotype-consistent and stereotype-neutral conditions did not differ significantly. When we compared the average response times for trait and situational probes, we found that the response times for situational probes, M = 972.98, SD = 303.16 were significantly longer than the response times for trait probes, M = 900.98, SD = 246.94, F(1, 22) = 6.52, p < .05, η² = .23.

Discussion

Our results showed that no matter how stereotype labels were presented, there were no effects of stereotype consistency on the error rates of trait and situational words among participants. We did not find this surprising because of the simplicity of the task and the low overall error rates. Participants took longer to indicate that the relevant situational word was not included in the preceding sentence when the behavior in the sentence was stereotype-inconsistent than when the behavior was either stereotype-consistent or neutral. This showed that when a stereotype was inconsistent with the behavior, SSIs were facilitated. However, no effects of stereotype consistency were found for participants’ reaction time to trait words. That is, when the stereotype was consistent with the behavior, STIs were not facilitated. In addition, when comparing the average response times to trait and situational probes, we found that the response times for situational probes were significantly longer than were the error rates for trait probes. This result may indicate that when participants were confronted by the others’ behavioral information, SSIs (vs. STIs) may be induced more easily.

We found that when the stereotype was consistent with the behavior, STIs were not facilitated. This result did not correspond to our assumption, or to Ramos et al.’s (2012) findings, for which there are several possible reasons. The materials that we used were different from those used by Ramos et al., whose behavioral descriptions were composed of complete sentences and situational continuations (e.g., The economist gave money to charity, knowing that donations were tax deductible.). As the whole sentence structure was complex, more cognitive resources were required for the meaning of the sentence to be understood. However, although our sentences implied trait and situation simultaneously, their structure was relatively simple. In addition, culture influences the tendency to make STIs and SSIs. For example, Zárate, Uleman, and Voils (2001) found with Anglo versus Latino samples that only the former group showed evidence of making STIs. In their cultural study, Na and Kitayama (2011) found that individuals with an independent model of the self may develop a spontaneous tendency to infer a personality trait from another’s behavior, but those with an interdependent model of the self may not show such a tendency, and that although European Americans showed clear evidence of STIs, Asian Americans did not. Saribay, Rim, and Uleman (2012) also found that participants primed with an independent (but not interdependent) self-construal inferred traits more strongly than they did situational properties. Thus, there is a strong possibility that Chinese individuals with interdependent self-construal may not make STIs and SSIs equally (Zhu, Zhang, Fan, & Han, 2007. When we compared participants’ average response times for trait and situational probes, we found that the response times for situational probes were significantly longer than were the error rates for trait probes. This may indicate that, for Chinese individuals, SSIs may be more easily induced than STIs.

In line with our assumption, and the encoding flexibility model (Allen et al., 2009), when the stereotype was inconsistent with the behavior, SSIs were facilitated. As stereotypes are inflexible, individuals usually do not change their stereotype when confronted with stereotype-inconsistent information. Thus, they need other information to keep the coherence of the information as a whole, and SSIs will be facilitated.

In addition, in this study, whether stereotypes were activated supraliminally or subliminally, the pattern of the stereotype did not differ: When the stereotype was inconsistent with the behavior, SSIs were facilitated, but when the stereotype was consistent with the behavior, STIs were not facilitated. This result showed that the effect of stereotypes on spontaneous inferences is strong enough to influence an interpretation of others’ behaviors unconsciously. It is possible that the influence of stereotypes on spontaneous inferences is uncontrollable, namely, when the actor’s stereotype is activated, it interacts with behavioral information, and influences the impression formation result. Our results extend previous results on the utility of stereotypes.

There are several limitations in this study. As the stereotype labels that we used are not specific, to further explore the effect of stereotypes on spontaneous inferences, future researchers should use more classified stereotype labels. As we examined the influence of stereotypes on spontaneous inferences with Chinese participants for the first time, future researchers should explore if the effect of stereotypes on spontaneous inferences is different for participants from different cultures or with different self-construal types.

Allen, T. J., Sherman, J. W., Conrey, F. R., & Stroessner, S. J. (2009). Stereotype strength and attentional bias: Preference for confirming versus disconfirming information depends on processing capacity. Journal of Experimental Social Psychology, 45, 1081–1087. https://doi.org/10.1016/j.jesp.2009.06.002

Cohen, J., Cohen, P., West, S. G., & Aiken, L. S. (2003). Applied multiple regression/correlation analysis for the behavioral sciences (3rd ed.). Mahwah, NJ: Erlbaum.

Crawford, M. T., Skowronski, J. J., Stiff, C., & Scherer, C. R. (2007). Interfering with inferential, but not associative, processes underlying spontaneous trait inference. Personality and Social Psychology Bulletin, 33, 677–690. https://doi.org/10.1177/0146167206298567

Freeman, J. B., Nakayama, K., & Ambady, N. (2013). Finger in flight reveals parallel categorization across multiple social dimensions. Social Cognition, 31, 792–805. https://doi.org/10.1521/soco.2013.31.6.792

Jerónimo, R., Garcia-Marques, L., Ferreira, M. B., & Macrae, C. N. (2015). When expectancies harm comprehension: Encoding flexibility in impression formation. Journal of Experimental Social Psychology, 61, 110–119. https://doi.org/10.1016/j.jesp.2015.07.007

Johnson, K. L., Freeman, J. B., & Pauker, K. (2012). Race is gendered: How covarying phenotypes and stereotypes bias sex categorization. Journal of Personality and Social Psychology, 102, 116–131. https://doi.org/10.1037/a0025335

Lupfer, M. B., Clark, L. F., Church, M. B., DePaola, S. J., & McDonald, C. D. (1995). Do people make situational as well as trait inferences spontaneously? Unpublished manuscript, University of Memphis.

Moskowitz, G. B., Stone, J., & Childs, A. (2012). Implicit stereotyping and medical decisions: Unconscious stereotype activation in practitioners' thoughts about African Americans. American Journal of Public Health, 102, 996–1001. https://doi.org/10.2105/AJPH.2011.300591

Na, J., & Kitayama, S. (2011). Spontaneous trait inference is culture-specific: Behavioral and neural evidence. Psychological Science, 22, 1025–1032. https://doi.org/10.1177/0956797611414727

Ramos, T., Garcia-Marques, L., Hamilton, D. L., Ferreira, M., & Van Acker, K. (2012). What I infer depends on who you are: The influence of stereotypes on trait and situational spontaneous inferences. Journal of Experimental Social Psychology, 48, 1247–1256. https://doi.org/10.1016/j.jesp.2012.05.009

Saribay, S. A., Rim, S., & Uleman, J. S. (2012). Primed self-construal, culture, and stages of impression formation. Social Psychology, 43, 196–204. https://doi.org/10.1027/1864–9335/a000120

Todd, A. R., Molden, D. C., Ham, J., & Vonk, R. (2011). The automatic and co-occurring activation of multiple social inferences. Journal of Experimental Social Psychology, 47, 37–49. https://doi.org/10.1016/j.jesp.2010.08.006

Todorov, A., & Uleman, J. S. (2002). Spontaneous trait inferences are bound to actors' faces: Evidence from a false recognition paradigm. Journal of Personality and Social Psychology, 83, 1051–1065. https://doi.org/10.1037/0022-3514.83.5.1051 

Uleman, J. S., Hon, A., Roman, R. J., & Moskowitz, G. B. (1996). On-line evidence for spontaneous trait inferences at encoding. Personality and Social Psychology Bulletin, 22, 377–394. https://doi.org/10.1177/0146167296224005 

Uleman, J. S., Rim, S., Saribay, S. A., & Kressel, L. M. (2012). Controversies, questions, and prospects for spontaneous social inferences. Social and Personality Psychology Compass, 6, 657–673. https://doi.org/10.1111/j.1751-9004.2012.00452.x 

Uleman, J. S., Saribay, S. A., & Gonzalez, C. M. (2008). Spontaneous inferences, implicit impressions, and implicit theories. Annual Review of Psychology, 59, 329–360. https://doi.org/10.1146/annurev.psych.59.103006.093707

Wang, P., Huo, P.-F., & Wang, L.-H. (2012). Processing levels and prerequisites of subliminal perception: Evidence-based visual masked priming [In Chinese]. Acta Psychologica Sinica, 44, 1138–1148. https://doi.org/10.3724/SP.J.1041.2012.01138

Wigboldus, D. H. J., Dijksterhuis, A., & van Knippenberg, A. (2003). When stereotypes get in the way: Stereotypes obstruct stereotype-inconsistent trait inferences. Journal of Personality and Social Psychology, 84, 470–484. https://doi.org/10.1037//0022-3514.84.3.470

Wigboldus, D. H. J., Sherman, J. W., Franzese, H. L., & van Knippenberg, A. (2004). Capacity and comprehension: Spontaneous stereotyping under cognitive load. Social Cognition, 22, 292–309. https://doi.org/10.1521/soco.22.3.292.35967

Zárate, M. A., Uleman, J. S., & Voils, C. I. (2001). Effects of culture and processing goals on the activation and binding of trait concepts. Social Cognition, 19, 295–323. https://doi.org/10.1521/soco.19.3.295.21469 

Zhu, Y., Zhang, L., Fan, J., & Han, S. (2007). Neural basis of cultural influence on self-representation. Neuroimage, 34, 1310–1316. https://doi.org/10.1016/j.neuroimage.2006.08.047

This study was sponsored by the Peak Discipline Construction Project of Education at East China Normal University.

Pei Wang, Faculty of Education, East China Normal University, 3663 Zhongshan North Road, Putuo, Shanghai 200062, People’s Republic of China. Email: [email protected]